Report No. 44190-KE Kenya Poverty and Inequality Assessment Executive Summary and Synthesis Report April 2009 Poverty Reduction and Economic Management Unit Africa Region Document of the World Bank TABLEOFCONTENTS ACKNOWLEDGEMENTS .......................................................................................................... EXECUTIVESUMMARY ...............................................................................................................Vm M Scale of the PovertyAnd Inequality Challenges ......................................................................... x Access to IncomeEarning Opportunities .................................................................................. xv Trends inHuman CapitalAnd Access to Services ................................................................... xx The Role of Institutions: Focuson The Budget ...................................................................... xxii .. DisadvantagedGroups.......................................................................................................... xxvii 1.Conclusions............................................................................................................................. xxx CONTEXT ................................................................................................................................... 1 Introduction and Objectives........................................................................................................ 1 Land, Peopleand Space: An Overview....................................................................................... 1 Economic Trends ........................................................................................................................ 4 6 2.Selected Aspects ofPolitical EconomyAnd Governance .......................................................... THEEXTENTAND NATUREOFPOVERTYAND INEQUALITY .............................. 11 Introduction............................................................................................................................... 11 A National Snapshot: The 2005-2006KIHBS .......................................................................... 12 Almost Halfthe PopulationLives inPoverty and mostly in RuralAreas ............................ Variation inPoverty incidenceamongProvinces and Districts............................................ 12 Variation inPoverty Incidencewithin Districts.................................................................... 18 19 Non-Monetary Aspects of Well-being...................................................................................... 21 ................................................................. Access to Improved Water.................................................................................................... Inequalitiesin Access to Servicesand Outcomes 24 29 Profile of Monetary Poverty ...................................................................................................... 34 Provinceby Province............................................................................................................ 49 Focusonthe Slums ............................................................................................................... 56 74 3.ImpactsoftheOF Political Crisis on Poverty andInequality......................................................... DUVERS OBSERVED TRENDS ................................................................................... 80 Qualitative Findings.................................................................................................................. 81 Distribution of Welfare IMPROVEMENTS, 1997-2005/6...................................................... 87 Labour Market Trends .............................................................................................................. 89 Changes inthe Distribution of Assets: HumanCapital, Land and Access to Credit ................97 Trends inthe Urban Slums -Insights from Viwandani andKorogocho ................................ 116 Vulnerability ........................................................................................................................... 120 Corruption as a Driver of Poverty........................................................................................... 126 4. POLICIESAND INSTITUTIONSTO ADDRESSPOVERTYAND INEQUALITY 130 .. Growth Agenda-expansion of income earningopportunities............................................... 130 Improving Rural Outcomes ................................................................................................. 135 Progress on the LandAgenda ............................................................................................. 137 Mobility inthe Labour Market ...............;........................................................................... 139 11 .. Improvements inthe InvestmentClimate........................................................................... 139 Access to Financial Services ............................................................................................... 141 Measures Aimed at Disadvantaged Groups ............................................................................ 143 Youth .................................................................................................................................. Women................................................................................................................................ 143 144 SlumResidents ................................................................................................................... 145 Social Protection ................................................................................................................. 147 The Budget as an Instrumentfor PromotingPoverty Reduction and Addressing Inequality. 152 ImprovingExpenditureComposition .................................................................................. 154 DecentralizedSpending...................................................................................................... Scenarios and Options: M A M S .......................................................................................... 157 158 Core Poverty Programs ....................................................................................................... 160 Public ExpenditureManagement.. ...................................................................................... Benefit Incidenceo f Public Educationand Health Spending............................................. 162 166 Monitoringand Evaluation..................................................................................................... 177 REFERENCES ................................................................................................................................. 185 LISTOFBOXES E-Box 1: Impacts o fthe Crisis and Inflation. 2008 E-Box 2: Philip Lagat-Escaped Poverty through Diversification inFarming ......................... ........................................................................ xi xvii Box 1-2: Youth and Politics ......................................................................................................... Box 1-1: Provincial Bias inthe Historical Patterno f Public Good Provision................................ 9 Box 2-1: Data Sets Used inthe KPIA .......................................................................................... 10 12 Box 2-2: DefiningPoverty in Kenya ............................................................................................ 15 Box 2-3: LivelihoodGroups Distinguishedinthe PPA-N.......................................................... ............................................................................................................ 19 Box 2-5 : Youth Unemployment - SelectedInternational Experience......................................... Box 2-4: Location Matters 55 70 Box 2-6: The Impact of Conflict on Poverty inKenya ................................................................. 75 Box 3-1: Joseph MbauniNderitu-Remained Casual Employment and Ill-health ................................................................................................ Poor due to land Fragmentation, Reliance on 84 Box 3-2: Philip Lagat-Escaped Poverty through Diversification inFarming ............................. 86 Box 3-3: Informal Financial Services ......................................................................................... 107 for SustainedGrowth................................................................................................................... Box 4-1: Growth Essential for Poverty Reduction-Equity and Equality o f Opportunity Essential 131 Box 4-2: ImprovingAgricultural Productivity: Factors Affecting Adoption Practices of Maize Farmers Box 4-3: Expanding FinancialAccess via MobilePhones: M-PESA ....................................... ........................................................................................................................................ 136 142 Box 4-4: A Tale of Two Slums: Nairobi and Dakar .................................................................. 147 Box 4-5: Key Elements ofNutrition Interventions ..................................................................... 150 Box 4-6: Mechanisms to Support Local Spending ..................................................................... 159 Box 4-8: Benefit IncidenceMethodology .................................................................................. Box 4-7 Uganda's PAF: Achievements and Shortcomings ......................................................... 162 167 Box 4-9: Delays in Conducting Labour Market Surveys ........................................................... 180 iii LISTOFFIGURES E-Figure 1: Povertyby Province. 2005/6 .................................................................................... xiii ... E-Figure 2: Provincial Distribution of Service Gains, 2003-2005/6 ............................................ xv E-Figure 3: Individual EarningsRelativeto the Poverty Line, 2005/6 ....................................... xvi E-Figure 4: ReducedDistancesto Various Infrastructure, 1997-2007 ...................................... E-Figure 5: Index ofRealTrends in Sectoral Spending, 1999/2000-2006/7 (1999=100) ......xviii ... xxiv E-Figure 6: BenefitIncidenceof GovernmentEducation Spendingby Level andQuintile, 2005/06 (percent)......................................................................................................................... xxv E-Figure 7: Benefit Incidenceof GovernmentHealth Spendingby Level and Quintile, 2005/06 (percent) ...................................................................................................................................... xxvi E-Figure 8: UnemploymentRate, Youth Versus Adult, by Province(percent of active incohort) ................................................................................................................................................... xxVll E-Figure 9: Level ofEducationCompleted(percentage of cohort), 2005/6 ... E-Figure 10: PovertyTransitions intheNairobi Slums, 2003 -2006 ...................................... ........................... xxviii xxix Figure 1-1: Kenya and Selected Comparator Countries: Age DependencyRatio ......................... 2 Figure 1-2: Age Structure of Population, SelectedCohorts, 1988, 2005 and 2030 Figure 1-3: EthnoLinguistic FractionalizationIndex ..................................................................... ........................ 3 4 Figure 1-4: Annual Real GDP per Capita Growth, 1981-2006 ..................................................... 5 Figure 2-1: Lorenze Curve inKenyaand Some Comparisons(2005-2006) ................................ 16 Figure 2-2: Ratio of ConsumptionPer Capitato the Poverty Lineby Decile .............................. 17 Figure 2-3: Populationand Poverty by Province, 2005-2006 ...................................................... 18 Figure 2-4: Material of the Walls of the MainDwelling inRuralAreas by Decile 21 Figure 2-5: Proportion of Enrolled Students by Gender, Ages 6-14: 2005/6 Figure 2-6: Provincial Distribution of Service Gains, 1998-2003................................................ .............................. ..................... 23 Figure 2-7: Provincial Distribution of Service Gains, 2003-2005/6............................................. 25 Figure 2-8: GradeAttainment Profile, Rural-Urban,Ages 15-19: Kenya 1993-2005/6..............26 27 Figure 2-9: GradeAttainment Profile by Quintile, Ages 15-19; 1993-2005/6 Figure 2-10: Distribution ofNet Enrollment Rates by Quintile - SelectedCountries.................28 ............................. Figure 2-11: Water ServiceAccess, Rural and Urban, Top andBottom Quintiles...................... 28 Figure 2-12: Determinants of HouseholdConsumption, RuralandUrban 2005/6...................... 30 35 Figure 2-13: Tegemeo OLS Regressionon Poverty Figure 2-14: Average HouseholdSize andNumber of Children per Household, by Decile........36 ..................................................................... 38 Figure 2-15: EducationofHouseholdHeadby Decile ................................................................. Figure 2-16: PovertyIncidenceand Gender ofthe HouseholdHead, District Level...................40 39 .......................................... Figure 2-18: PovertyRates by Migration Status .......................................................................... Figure 2-17: Where Migrants Come from and Go to, Nairobi Slums 43 44 Figure 2-20: Credit by Province:Tegemeo .................................................................................. Figure 2-19: Fractionof Adults Involved inOff FarmActivities: Tegemeo ............................... 45 Figure 2-21: PatternsinRoads Access by Province, 2004........................................................... 46 ..........49 Figure 2-23: Proportion Children Sick Last Month...................................................................... Figure 2-22: Age-Sex CompositionofNairobi, Kenya, Informal Settlement Populations 57 Figure 2-24: PopulationBreakdownand Poverty Level by HouseholdType, 2006.................... 59 Figure 2-25: Years of Educationof HouseholdHead, by HouseholdType, 2006 ....................... 60 60 Figure 2-26: Years HouseholdHeadhas Stayed in Informal Settlements, 2006 61 Figure 2-27: Enrollment by OVCs (By One or BothParents) ...................................................... .......................... Figure 2-29: Skills Profiles, by Age Cohort (Youth vs.Adults) Males and Females, 2005.........65 Figure 2-28: Age-Specific FertilityRates, 2003........................................................................... 63 Figure 2-30: KeyLabor Market Indicatorsfor DifferentAge Groups, by Gender(percent) .......65 66 iv Figure 2-31: UnemploymentRate. Youth versus Adult. by Province (percent of active incohort) ....................................................................................................................................................... Figure 2-32: UrbanYouth UnemploymentRates in SelectedAfrican Countries (Year Varies) 70 Figure 2-33: Youth (15-29 years) Activities inNairobi Slums. 2004 .......................................... ..67 Figure 2-34: AverageNumber of Months HouseholdRelied on PurchasedStaples, 2007 ..........72 78 Figure 2-35: Changes Cumulative Density Function of ConsumptionPer Capita 79 Figure 3-1: Comparisons of SubjectiveWell-Being, 2000-07: Tegemeo .................................... ....................... 81 Poverty Trends, 1997-2005/6........................................................................................................ Figure 3-2: Contributions of Changes inAverage Consumption, and Changes in Inequality to Figure 3-3: Total Population, Labor Forceand Wage Employment ('000 People). 1981-2006...90 89 Figure 3-4: Broad Sectoral Shares of Employment, 1997-2005/6 .............................................. Figure 3-5: ........91 and Good Quality Jobs in200506 ................................................................................................. Figure 3-6: Trends inEmployment andUnemployment. 1998-2005/06 by Gender. Urban/ Rural, Extent of Goodand BadQuality Jobs, EarningsRelative to the Poverty Line 94 Figure 3-7: Distribution of Employment inthe Informal Sector................................................... 94 97 Figure 3-8: Land Inequality. Regional Averages, 1986-1990, and Kenya 1997 ........................ Figure 3-9: First Responseto Major Shocks (Among HouseholdsReporting Shocks) .............104107 Figure 3-10: Rural Poverty Transitions, 1997-2007 ................................................................... 110 Figure 3-11: Rural Transitions and Persistence, Top and Bottom Quintiles, (percent) 2000-2007 ..................................................................................................................................................... 110 Figure 3-12: Transitions across IncomeQuintiles ....................................................................... 111 Figure 3-13: Summaryof effects of changes inkey variables over time, on poverty (fixed effects) Figure 3-14: Conditional on SeekingExtension, Who Was Asked for Advice, 2007................113 ..................................................................................................................................................... Figure 3-15: ReducedDistances to Various Infrastructure, 1997 - 2007 ................................... 114 Figure 3-16: Effect of Marginal Changes in SelectedSources ofIncome on Inequality ............115 Figure 3-17: Proportion of Slum Residents Observedin2003 but not in2006 ......................... 116 Figure 3-18: Transitions inNairobi Slums, 2003-2006.............................................................. 118 118 Figure 3-19: Percentageof Householdsthat Reportedat Least One Shock. by Decile, 2005/6.120 Figure 3-20: Number of DifferentShocks inthe PrecedingFive Years, 2005/6 ....................... 122 Figure 3-21:Average Number of Shocks over the PastFive Fears. by Decile and Poverty Status, ..................................................................................................................................................... 123 Figure 3-22: First Responseto a Major Shock, Among HouseholdsReporting a Shock (percent) Figure 3-23: ReportedSeverity of Shocks, as Percentageof Monthly Poverty Line, 2005/6.... 124 ..................................................................................................................................................... 124 Figure 3-24: Extent of ConsumptionSmoothing, Within Villages andAcross Villages within .Districts, 1997-2004 .................................................................................................................... 125 Figure 4-1: Annual Growth in GDP at FactorCost UnderAlternative Scenarios, 2007-2012 Figure 4-2: Compositionof Sectoral Spending, 2006/07 ............................................................ (percent)....................................................................................................................................... 134 154 Figure 4-3: PossibleTrade-offs betweenSpendingon Growth on Infrastructureand Social Services. on MDGtargets for Human Development andPoverty, 2012..................................... 158 Figure 4-4: BenefitIncidenceof GovernmentEducationSpendingby Level and Quintile, 2005/06 (percent)......................................................................................................................... 168 Figure 4-5: BenefitIncidenceof GovernmentEducationSpendingby Level and RichestPoorest Quintile, 1992/93 and 2005/06 (percent) ..................................................................................... 169 Facility. 2005/06.......................................................................................................................... Figure 4-6: Quintile Distribution of Benefit Incidenceof GovernmentHealth Care Spendingby 173 LISTOFMAPS E-Map 1: District EstimatesofPoverty. 2005/6 .......................................................................... xiv Map 2-1: District Estimatesof Poverty. 2005/6 ........................................................................... 20 LISTOFTABLES E-Table 2: Access to Selected Services by Quintile. 1993 ..2005/6.......................................... E-Table 1: Inequality inDistribution of ConsumptionGains. 1997. ............................ 2005/06 xii xxi Table 1-1: Decompositionof Sources ofEconomic Growth. SelectedSub.periods. 1997.2006. Annual PercentageRates of Change ................................................................................................ 5 Table 1-2: SectoralShares inValue-Added GDP. and Contribution to Growth, 1997-2006. Percent ............................................................................................................................................. Table 1-3:Provincial Location of DifferentGroups (Percentage Shares)....................................... 6 7 Table 2-1: Poverty Estimates(2005-2006) Table 2-2: Average Rural andUrban Expenditure by Decile, 2005-6 ........................................ ................................................................................... 13 Table 2-3: Inequality Measures, Kenya and Selected Comparators............................................. 16 18 Table 2-4: Trends inSelectedMDG Indicators,Relativeto 2015 andMedium Term PlanTargets ....................................................................................................................................................... ........................... 23 Table 2-6: Water Service Levels by WSB -RuralAreas............................................................. Table 2-5: Access to Selected Services by Quintile between 1993 and 2005/6 26 29 Table 2-7: Water Collection Time by Water ServiceBoard-Rural ............................................ 31 Table 2-8: Vaccination CoverageRates of One-year-oldsby Wealth Quintile and Gender. 1993- 2003 ............................................................................................................................................... Table 2-9: HIV PrevalenceRates across Regions and Wealth..................................................... 32 . 33 Table 2-10: Poverty andHouseholdSize. andNumberof Children ............................................ 37 Table 2-11: Poverty and FemaleHeadship- Selectedhighlights 41 Table 2-12: IDPs inKenya, as of April 2008 ............................................................................... ................................................ 44 Table 2-13: An Overview of Provincial Differences, percent, 2005/6 Table 2-14: Employment Status by Age, 2005/6 .......................................................................... ......................................... 48 Table 2-16: Youth Time use inUrbanAreas, Selected African Countries (percent of Youth)....68 Table 2-15: Ratio ofYouth to Adult UnemploymentRate by Province and Gender, 2005/06 ....66 68 Table 2-17: Relative Poverty Incidence inHouseholdswith Unemployed/ Inactive Youth, 2005/6 (percent)......................................................................................................................................... .. Table 2-18: Level of EducationCompleted(percentage of cohort), 2005/6................................ 69 69 Table 2-19: Youth Earnings as Percentageof Average Earnings, Non Agricultural, by Sex Table 2-20: Top Ten Jobs for MaleYouth, 2004 ......................................................................... ......71 ................................................................. 73 Table 2-22: One Monthand Twelve Months Change inPrice Indices ........................................ Table 2-21: Top Ten Jobs for FemaleYouth in Slums 73 77 Table 2-23: Impactsof Price Shocks on PovertyMeasures ......................................................... 78 Table 3-1:Poverty Transitions inPPA-IV Householdsby Livelihood Group Table 3-2: Major Reasons for Falling into Povertyby Livelihood Group (% Households).........83 .............................. 85 Table 3-3: Major Reasons for EscapingPovertyby Livelihood Group (percent households).....86 Table 3-4: Inequality inDistribution of ConsumptionGains, 1997-2005/06 ............................. 88 (percent), 1997 and 2005/6............................................................................................................ Table 3-5: Trends inthe Compositionof Employmentby Main Sectors, Urban, Rural and Total 92 Table 3-6: Distribution ofthe Employedby Economic Sector, Shares and Changes, 1997- 2005/6 ............................................................................................................................................ 92 vi Table 3-7: Decomposition of Changes in Per Capita Value Added GDP* (Percentage Points per Year). 1997. 2005/6 ..................................................................................................................... 93 Table 3-8: Individual Monthly Earnings: Medianand Gini- Coefficient, 2005/6 Table 3-9: Composition of Employment by Education Completed (percent), 2005/6 ..................95 ........................ Table 3-10: Urban Real Wages by Education Levels 2005/6 as Percentageo fAverage 98 Table 3-11: Trends in SelectedHealth Indicators ......................................................................... ..............95 99 Gini Coefficient, 1996-2005/6.................................................................................................. Table 3-12: Changes in Land Inequality over Time, all Households, by Province and Nationally, 103 Table 3-13: Changes inLand Inequality Over Time, Landholders only, by Province and Nationally, 1997-2005/6 ...I ........................................................................................................ 103 Table 3-14: Land Gini coefficients, Selected Latin American and African Countries (year varies) ..................................................................................................................................................... Table 3-15: Credit Access and Average Loan-Sum, Previous 12 Month.................................... 104 Table 3-16: Share of Households Accessing Credit from Various Sources (percent)* ...............106 106 Table 3-17: Financial Service Access Strands(percent) ............................................................. 108 Table 3-18: Types of Shock inthe Past five Years, Percent o f Households Reporting Any Shock ..................................................................................................................................................... 121 Table 3-19: Percentageof Overall Households (Unweighted) Severely Negatively affectedby Specific Shocks, 2000-2005 ........................................................................................................ 122 Table 3-20: Relative Financial Costs of Grand Corruption Episodes inKenya 127 Table 3-21: Annual BriberyEncounters, Share of Group, 2006 (percentage) ........................... .......................... 128 Table 3-22: ReportedAnnual Expenditureson Bribery, By ReportedIncome 129 Table 4-1: Description of MAMS Simulations .......................................................................... .......................... 132 Table 4-2: Evolution of SelectedMacro Indicators-Base Case, and Trend (Inthe Absence of Temporary Slowdown), 2007-2012 Table 4-3: MDG, Poverty and InequalityIndicators, 1990, 2007, 2015 Target, and 2012 ........132 ............................................................................................. 133 Table 4-4: Barriers to Business inKenya- Manufacturing FirmsReportingFactor as Major or very Severe, 2007, Relativeto 2003 (%) Table 4-5 Central Government Fiscal Outcomes, Percent of GDP, 1999/2000-2006/07 ...........140 ..................................................................................... 153 Table 4-6: RealTrends in Sectoral Expenditure 1999/2000=100 .............................................. 155 Table 4-7: Functional Composition of Central Government Spending, Percent of GDP, 1999/00- Table 4-8: Sectoral Expenditures as Percentageof Total Budget, 1999/2000 -2006/07 .............156 2006/7 .......................................................................................................................................... Table 4-9: Core Poverty Programs, 2002/03 - 2006/07, percent................................................. 156 160 Table 4-10: Semi-autonomous Government Agencies and Extra-BudgetaryFunds, By Sector, 2007 ............................................................................................................................................. 163 Table 4-11: Budget under Spending: Recurrent and Development, percent, 1999/2000 -2006/07 ..................................................................................................................................................... Table 4-12: Budget Execution Rates by Sector 1999/2000-20006/07 ........................................ 164 Table 4-13: Net Primary Enrolment Ratesby Wealth Quintile, SelectedCountries..................165 167 Table 4-14: Benefit Incidence o f Education SpendingRelative to Mean Household Total Consumption, 2005-06 ................................................................................................................ ..................... 169 Table 4-16: Distribution of Public Health Facilities by Province, 2004..................................... Table 4-15: UnitDistrict Spendingon Primary Education by Province, 2005/06 170 172 Consumption, by Quintile, 2005/06............................................................................................. Table 4-17: Kenya: Per capita Health Subsidy Relative to Per Capita Total Household 174 Table 4-18: BenefitIncidence of Government Health Spendingby Poorest and Richest Quintile Table 4-19: Private and Public Spendingon Health, 2005/6...................................................... and Level of Care, 1992-93 and 2006-06 (percent share) ........................................................... 175 Table 4-20: IP-ERS: APR Indicators Including Performance Revised Targets ........................ 175 178 vii ACKNOWLEDGEMENTS This report has been underpinnedby a phasedprocess. In 2005, an initial concept note was discussed in2005, inanticipation of the Kenya Integrated Household Budget Survey and Participatory Poverty Assessment. Over the period 2005 through 2008, technical ' support was provided to the Kenyan National Bureau of Statistics, from DFID, USAID, GTZ as well as the World Bank and others. This included training on poverty measurement and different types of analysis. The actual analysis underpinning the KPIA beganinMarch2008, following the release of KIHBS data to the World Bank. Several areas of work were carried out jointly with local partners and experts including insights from urban slums, with African Population and Health Research Institute (APHRC) and on understanding dynamics, with Tegemeo Institute, for whose cooperation we are grateful. A full list of background papers and authors follows on the next page. Generous support from GTZ andDFID helpedto facilitate the work. This report benefits from discussions and substantive comments received during discussions with the Government of Kenya in early July, including from HE the Prime Minister and his office, Ministry of Planning, Ministry of Finance, members of the National Economic and Social Commission, and several line agencies. This assessment was preparedbyJeni Klugman(TTL). The team of contributors included Tavneet Suri (MIT); Isis Gaddis (KIHBS regression analysis and provincial profiles); Markus Goldstein (gender); Leonard0 Garrido (labour analysis); John Randa and Tracey Lane (fiscal analysis); Lionel Demery (benefit incidence analysis); Johan Mistiaen (poverty levels and decomposition); Nadeem Karmali (non-monetary dimensions); Joost de Laat, University of Quebec at Montreal (UQAM) (slumdynamics); Hans Lofgren and Martin Cicowiez (MAMS simulations); David Ndii (corruption); Raymond Muhula (history and governance); Marieke Kleemans (land and vulnerability); Anne Christensen (water and sanitation); Wendy Ayres (monitoring and evaluation). Sectoral advice and contributions were provided by several Bank colleagues, including Michael Mills, Shobhana Sosale and Ernma Sorensen (human development); Sumila Gulyani (urban slums); Sahr Kpundehand Steve Ndegwa (governance); Frank Fulgence K. Byamugisha, Andrew Mwihia Karanja, Christine E. Cornelius (rural development), and MichaelFuchs (FSD). Our peer reviewer, Jesko Hentschel provided valuable feedback and advice, as did Kathleen Beegle and Peter Lanjouw (DEC) and David Ferrand (FSD), and Praveen Kumar and Cevdet Denizer (PREM). Editorial and administrative support was ably provided by Dora Harris, Marjorie Kingston and Rosemary Otieno. The Government of Kenya has provided comments on the report inearly Marchof 2009 and these were taken into account by Cevdet Denizer who finalized the report inApril 2009. ... Vlll EXECUTIVESUMMARY 1. This assessment of poverty and inequality comes at an important juncture for Kenya. The December 2007 elections and subsequent pronouncements o f the newly formed GrandCoalition have underlinedthe salience o f these issues to ordinary Kenyans, and for policy makers. The violence in early 2008 highlighted the importance o f addressing poverty and inequality as major goals in their own right, but also for instrumental reasons -as major goals intheir own right, the persistent inequalities spark conflict, which i s welfare reducing, and this conflict in turn will harm prospects for growth. The onset o f the global credit crunch has also shown how poverty and public service delivery related vulnerabilities could be exacerbated by external shocks. Cumulatively, these factors underline the value o f appropriate diagnostics about the patterns o f poverty and inequality in informing public debates, strategies and actions to overcome exclusion from the benefits o f growth and development in Kenya as well as designingpolicies to minimize the impact o f the current global crisis. 2. The Medium Term Plan, which sets out the government's policy and program framework within the context of Vision 2030, is a major opportunity to take this agenda forward. The stated focus is on poverty and inequality-and the related themes o f provincial variation, youth and gender-as laid out in the President's foreword and the Minister's statement. This i s welcome, but will need to be carried through to the actual program. The draft plan is presently under revision. It i s expected that this assessment should help provide guidance as stakeholders and policymakers proceed to review the planmore fully, andprioritize the actions therein. The draft plani s also expected to include policy measures to mitigate the effects o f economic slowdown due to the global downturn and considerably higher inflation than in 2007/08 and could benefit from the findings o fthis report. 3. The KPIA attemptsto address several broad policy-motivatedquestionswith a focus on new findings which deepen and extend our knowledge base. What has happened to poverty and inequality over time and what are the major dimensions o f poverty and inequality? Have the opportunities generated by economic growth broadly or narrowly shared, with major population groups largely excluded, and who manages to escape from poverty, and why? And what i s the pattern o f public investments and services, and how do these interact with poverty? The report closes with some policy directions that warrant considerationinlight o fthe diagnostic. 4. The KPIA builds on publishedresults, inparticular the Basic Report on the Kenya Household Budget Survey, and a rich range o f sources, including those associated with leading local think tanks and qualitative data. A series o f World Bank reports-most notably on growth, the investment climate and rural development-have just been finalized, to which the KPIA should be seen as a companion report. ix 5. Selected findings are highlighted inthe Executive Summary: first, the scale of the challenges; second, drivers of poverty and inequality; third, the budget as an instrument for addressing poverty and inequality objectives; and fourth, specific disadvantaged groups. It does not attempt to capture all the report's analyzes, which are extensive and should be of interest to a wide audience. Rather the purpose here i s to briefly present the most salient messages for policy-makers. The focus i s on the mediumterm, but framed by immediate post-crisis needs and priorities. The interested stakeholder is strongly encouragedto readmuch more of the report. SCALE OF THE POVERTY ANDINEQUALITY CHALLENGES 6. There are major underlying structural inequalities inKenya, whichhavetheir roots in colonial times and have persisted in different ways since independence. Kenya needs to address these disparities in order to enable inclusive growth and development, otherwise the vicious circles of elite capture, skewed distribution, social exclusion and poverty will persist. 7. The scale ofthe challenges canbe summarized as follows: In2005/6, almost 47 percent-or 17 million-Kenyans were unable to meet the cost of buying the amount of calories sufficient to meet their recommendeddaily nutritionalrequirementsand minimal non-food needs. Many Kenyans are very poor-indeed almost one out of every five could not meet the cost of this minimal food bundle even if their entire budget is spent on food. The vast majority of the poor-14 million-live inruralareas. Over the long term, little inroads have been made in reducing poverty-the officially estimated poverty rate was 48 percent in 1981. This disappointing record is not surprising inlight o f the weak growth performance over the long term, and high levels of inequality. Alternative sources are used to triangulate theresults, andconfirmthe direction of change. The acceleration of economic growth after 2003 likely reduced poverty; although national data is not available to confirm this, Tegemeo shows reductions inruralpoverty continuing through 2007 (Nyoro et a12008). Inequality in household consumption i s high. In 2005/6, the ratio of consumption between the top and bottom 10 percent stood at 20:l and 12:l in urban and rural areas, respectively. This compares to 5:l in Tanzania and 3.3:l in Ethiopia, for example. There are also significant horizontal differences, across groups, inparticularprovinces. Shocks are a fact of life for many households, and are especially frequent for the poorest. The most common shocks over the period 2000-2005 were, in order of importance, food price inflation, droughts and floods, illness, and death inthe family. 8. The available data pre-datestwo major events in2008: the post election crisis and the acceleration in inflation. E-Box 1 outlines how these may have affected the 1The official monthly threshold was KSh 1,562 and 2,913 per adult equivalent in rural and urban areas respectively in2005/6. X headline figures. The crisis also underlined the fact that the economic recovery which beganin2003 remains vulnerable to political and other risks. The KPIA draws attention to some potentially damaging longer term repercussions of the crisis. Inparticular, prior to the crisis migration was an important strategy used by households to improve their well-being. The feasibility o f this strategy has beenaffected by the risk of conflict. One policy implication is thus the need to help ensure against barriers to mobility, including social tensions, which requires further analysis of the sources of such tension, especially in densely populated rural areas. It is critical that the recent violence and the extensive displacement that occurred not be a harbinger for ongoing curtailment of voluntary labour movement around the country. A shutting down of mobility inresponse to job and other opportunities not only negatively affects individual households directly, by restricting their options, but would likely also have more general adverse effects on growth and economic activity. E-Box 1: Impactsof the Crisisand Inflation, 2008 The costs associated with the post election violence were large, including about one thousand deaths and about 300,000 displaced. In the short term, public services in some areas were disrupted; markets have responded through, among other things, layoffs and increased prices; and civil and social capital and networks were damaged. In the absence o f data post-dating the conflict, the 2005/06 KIHBS is used for heroic first attempts at quantification. These estimates abstract from second round and equilibrium effects and adjustments at the household level, and ignore any changes in nominal wages and incomes that could have happened. With these caveats, initial effects relate to: *Increasedunemployment -ifunemployment rose, say from 12.5 to 15 percent, given the poverty risk o f households with an unemployed member, this would increase the poverty rate significantly, o f the order o f 9 percent. *Inflation-as reported by the KNBS, prices have been rising rapidly since January. Food and non- alcoholic drinks make up over half the index. The data also suggest significant increases in the relative prices o f food and transport over the first half o f 2008. Our analysis o f the KIHBS shows that the food share o f the poor is 70 percent and o f the poorest quintile it is 77 percent. Because many Kenyans have consumption levels just above the poverty line (about 10 percent o f the population are within 10 percent o f the line, and 20 percent within 22 percent), and spend a significant share o f their budget on food, cost o f living increases without corresponding increases in wages and incomes mean that significant numbers o f Kenyans would fall below the threshold. The impacts of food price increasesdependon the distribution o f net buyers and sellers o f grain. Tegemeo analysis shows that the bottom deciles typically purchase grain for more than three months every year, compared to a single month for the top decile. Only 20 percent o f the bottom decile sell grain most years, compared to 60 percent o fthe top decile (Nyoro et a12008). This suggests that impacts o fmaize price rises would tendto be negative for the poor. While overly simplistic, these parameters suggest that the first round impacts on poverty were significant, both for those who were already poor, as well as large numberso f "new poor." This would be offset to the extent that nominal incomes also rose, and to the extent that consumption patterns can adjust inthe face o f relative mice changes. PatternsandDriversof PovertyandInequality over Time 9. The KPIA looks at well-being and inequality in multiple dimensions, in outcomes, like consumption and human development, and in access to income earning opportunities. Trends in the level and distribution of access to productive assets like landand capital (credit) are investigated, as i s access to basic services, using a x i range of data sources. A cross cutting theme is provincial and district variation, for which an initial snapshot i s provided. In addition to spatial disparities, the analysis also identifies several significantly poor groups, including, for example, urban households headed by widows. Youth and slum residents, represent particularly large (and vocal) groups of the poor, with more immediate as well as longer term structural policy implications, are addressed inthe final section ofthis summary. Trendsin Poverty, 1997-200M 10. The resumption of economic growth after 2003 likely brought benefits in terms of poverty reduction. However because our national baseline predates the recovery by several years, we cannot assess how the payoffs were distributed across households. For poverty, we have two data points, 1997 and 2005/6, but this does not provide a linear trend over the intervening period, which included several years of stagnationand recession, and only two years of economic growth. 11. Levelsof inequalityin Kenyawere already highwhen the economic recovery began in 2003. Inequality matters because different types of inequality tend to be inter- connectedand reinforcing, and meanthat some groups have inferior opportunities (World Bank 2007). International evidence also suggests that high levels of inequality damage prospects for sustained growth, and serious horizontal inequalities across groups are associatedwith conflict (Stewart 2002,2008). 12. Average welfare gains over the period 1997-2005/6 were very much concentrated-in particular amongst the wealthiest quintiles, urban residents, and, in terms of provinces (E-Table 1). For many other groups, the aggregate improvement over the period was very limited and practically stagnant. Most striking are the findings that the poorest quintile lost out in absolute terms while gains for the second bottom quintile were only about one percent annually. Even for the middle quintile, growth in consumption was below average. E-Table 1: Inequality in Distribution of ConsumptionGains, 1997 -2005106 1997 2005106 YOchange Area Urban 4,436.59 5,493.96 23.8 Rural 1,962.94 1,993.21 1.5 Quintile Poorest 708.07 702.99 -0.7 2 1,159.83 1,268.59 9.4 3 1,680.82 1,846.75 9.9 4 2,439.84 2,778.18 13.9 Wealthiest 5,758.93 6,895.72 19.7 National 2,349.36 2,698.12 14.8 Source: Staff estimates based on KIHBS (2005/06), W M S (1997). Notes: Mean adult equivalent per capita expenditures. 1997 is adjusted to 2005/06 values usingthe ratios o fthe urban and ruralpoverty lines. xii Spatial Disparities are Significant 13. There is striking provincial variation in incomes, poverty and human development, as well as disparities in access to services. E-Figure 1 shows the provincial differences in poverty incidence, which i s highest in North-Eastern, while Coast, Western and Eastern are all above average. Central and Nairobi have rates o f poverty below the national average. Shares in total poverty differ, given variable population size, with the largest absolute contributions coming from Rift Valley, followed by Eastern. The map illustrates the considerable district level variation in poverty incidence in 2005/6. There i s also considerable variation in poverty within districts. E-Figure 1: Poverty by Province, 2005f6 (Size ofthe Bubble shows Contribution to Total Poverty) I 3 60 - G2 IUnitfI.' stern 17.7 50 - v* 0 E 40- Ccntral 8.1 Nyanza 14.2 Rift Valley 25.6 Nairobi 3.7 10 - 0 1 I , I I , 0 5 10 15 20 25 3c Distributionof population(percent) 14. These spatial disparities in poverty arise for a range of reasons. Naturally, there are climatic and agro-ecological differences, but there are also different levels of access to services. Moreover, estimated rates o f return to household characteristics like education or employment differ significantly by province. Average household characteristics vary between provinces, which can either reinforce or offset these differences inreturns. For example, average household size is higher in Western than in Central province; and the likelihood o f owning land also differs. Our regression models do not tell us why the returns are different. However, the results do show that location i s a powerful predictor o f living conditions andpoverty. ... Xlll E-Map 1: District Estimatesof Poverty, 2005/6 15. The KPIA looks at two major dimensions of disparity-by province, and by quintile-in access to various services over the decade to 2003, and 2003-05/6. For the earlier period, there was stagnation and deterioration of access for poor groups. For example, access to safe water stagnated nationally in the decade to 2003. For education, province-level differences were not as stark as for water and sanitation but Coast and NorthEasternprovinces still lagged far behindthe national average. xiv 16. Service gains were distributed more broadly during the more recent- 2003-2005/6-period for which data is available (E-Figure 2). Rift Valley, Nyanza and Western all saw large increasesinhousehold electrification, relative to the low levels reported in 2003. Water and sanitation access increased. Gross primary education has increased in all provinces, although given the size of Rift Valley province, the gains therein are most substantial in absolute terms while the relative increase was largest in North-Eastern, which had been far behind. E-Figure 2: ProvincialDistributionof Service Gains, 2003-2005/6 Central Coast Eastern H Nyanza 0 Rift Valley 100% 80% 60% 40% 20% 0% -20% --.- - - - _ _ I - - ~ - ~ --*-"...----4 Electricity Water Sanitation Gross Primary Enrollment ACCESS INCOMEEARNING TO OPPORTUNITIES 17. The labour market is the primary transmission mechanism of economic growth to householdincomes, through the quantity, quality and distribution of income earning opportunities. Since labour i s the most abundant asset of the poor, a major determinant of the poverty reduction impact of economic growth is its effect on work opportunities. When growth enhances earning opportunities for the poor, its benefits are widely shared and poverty tends to decline. The KPIA builds on the recent analysis by UNDP and others to examine drivers in the Kenyan labour market, and examines the evolution of rural income earning opportunities more deeply, utilizing panelanalysis. 18. The aggregate jobs performance in Kenya over the past decade suggests some good news as well as a series of major challenges. Comparing the national cross- sectional survey results of 1998 and 2005/6, the broad storyline is that: There was netjob creation. Rural employment grew from 8.4 to 9.4 million- about 1.5 percent annually-although this was largely small scale subsistence and informal work-paid employment hardly grew inrural areas. 0 Unemployment fell from almost 15 to 12.5 percent. However, the urban rate rose from 18.5 to 20.6 percent and the share of youth (15-29 years of age) among the unemployedrose markedly, from 60 to 72 percent. xv 0 Inactivity increased rapidly, from 22.4 to 30 percent o f the working age population, due mostly to full-time study, but also ajump inthose reported as discouraged (Le., not working nor actively seeking ajob). 0 Female unemployment is significantly higher thanmale rates, inboth absolute numbers and relative terms (14 versus 11 percent), and three quarters of unemployed women are youth. There i s also a large gender wage gap, especially at lower levels of education. 19. Having a job inKenya is no guarantee against poverty, of course. The bulk of workers-about two-thirds of the total-are either on their own account and unpaid family workers. The minimal share of private modern employment, a mere 6 percent of the labour force, is striking. About 41 percent were located inurbanareas, andtwo thirds were male. 20. Amongst those Kenyans who reported earnings in the KIHBS, only about 41 percent (3.3 million) had jobs in 2005/06 with earnings over the poverty line.2 We can call these good qualityjobs, while recognizing that having such ajob does not necessarily enablean individualwith dependentsto live abovethe poverty line (E-Figure 3). Almost 6 out of 10workers who reported earnings hadjobs with earnings below the poverty line, or working poor, some of whom were also working less than40 hours a week. E-Figure 3: Individual EarningsRelative to the Poverty Line, 2005/6 Public & Modern Private -.. - - .. .... .... Informaland Small Scale Ag. E - - - - Public & Modern Private,Paid 0 10 20 30 40 50 60 70 80 90 100 percent of workers aged 15-64 with data on earnings 21. Earnings inequality is high in Kenya, with a gini index of 0.693. Earnings in agriculture are systematically lower for all types of employment, especially for those working on their own account or inthe family. 22. What is driving observed trends in income earning opportunities, and how do these link to poverty and inequality? The KPIA uses Tegemeo's panel data-which tracked the same rural households at four points in time over the decade to 2007-to investigate drivers out of and into poverty (Nyoro et a1 2008). While this picture obviously excludes urban developments, the predominance o f rural poverty in Kenya means that understanding these trends i s central to understanding the majority of poverty in Kenya. There are several significant factors associated with escaping rural poverty, * Earnings are missing for 31 percent of workers in the KIHBS sample, mainly own account and unpaid family workers. Gini coefficient varies betweenzero (perfect equality) andone (perfect inequality). xvi some o f which are related to overall market development, others to public investment and other programs: 0 The most notable correlate o f moving out o f poverty is income diversification, both on and o f f farm, which in practice are mainly additional crops being grown and/or informal business opportunities. 0 Changes ineducation (beyond primary school) are important. 0 Distances to markets and infrastructure matter, in particular distance to the closest fertilizer seller andthe distance to piped water. Distance to the nearest fertilizer seller can be seen as a proxy for market access (Suri 2007), though it matters more for farming households. It also captures, at least in part, the costs o f using improved methods (fertilizer and seed). That fertilizer access is significantly correlated with off farm diversification may be due to its correlation with access to technology and markets, as well as possibly enabling adults to produce crops more efficiently and spend more time working off-farm. 23. The Participatory Poverty Assessment (PPA) captured some interesting stories which reflect some important factors which were mentioned by communities inrecalling escapes from poverty. E-Box 2 presents one such example. E-Box 2: Philip Lagat-Escaped Povertythrough Diversification in Farming Philip is a 46 years old farmer in Chepkongony village, Nandi district. He never went far with his education because he came fiom a poor background. In 1990 he was working as a casual laborer inthe tea farms ofwell to do villagers. He usedto earn apaltryKsh200 per monthfrom casual work. He saved and bought a local cow for about Ksh 800 in 1992; he then stopped working as a casual labourerto take care of his family farm and the cow. In 1995, he got ajob ina prominentteafarmer's farm as acasualwith a slightlybetterpayofKsh300. Inthe same year, he set up a tea seedlingnursery inhis farm. Fromthis nursery, he transplantedup to 3000 tea seedlings into his farm. In 1997, he started harvesting his own tea. His cow was also providing some milk. His total earnings per month from tea and milk increasedto Ksh 4,000. He stopped working as a casual labourer and decided to concentrate in farming. By the year 2000, he hadbought a Friesiandairy cow. He also diversifiedinto growing coffee. He joined SouthNandiCoffeeCooperative Society which helpedinmarketinghis berries. By the time we visitedhim (2006), Philipwas earningup to Ksh40,000 per monthfiom sales of his tea, coffee, andmilk combined."My family is not doing very well but we are comfortable." Source: Mango et a1(2007). 24. For sustainedimprovementsinwell-being, the PPA also underlinesthat poor households need to avoid or at least minimize the impacts of shocks. This can be addressed, at least to some extent, through appropriate economic management (to combat inflation), broader policies to encourage economic growth (and reduce the incidence o f job loss) and improved governance and economic conditions more generally (to alleviate crime and security problems). The most common household responses to shocks are, in order o f importance, spend cash savings, sell assets (animals), work longer hours, reduce food consumption and receive help from family and friends. Government assistance was very rare. This points to the absence of, but also the potential for, publicly supported safety nets, especially when shocks affect entire districts or provinces. This would need to ensure appropriate targeting and accountability mechanisms, and should be considered in the context o f reallocations o f spending away from the sizeable items in the budget which do not support growth andpoverty reduction (see below). xvii 25. High underlying levels of inequality are a drag on poverty reduction, although there is some positive momentum in narrowing the gaps in.accessto some key services and opportunities. The rural panel data show that the differences between the rich and poor that existed in 1997-including in market access through improved infrastructure, and extent o f diversification o f farm and off-farm activities-had tended to narrow by 2005/6. Among the good news i s that the gaps between the better off andpoor rural households inaccessing off-farm employment have converged over the past decade. A similar pattern of both overall improvements in access and equity is seen with respect to different modes o f infrastructure, including distance from a motorable road and distance to piped water. 26. Some unevenness in the pattern of development is inevitable,as not everyone is able to take advantage o f opportunities to the same extent. Indeed Tegemeo analysis suggests that rising o f f farm income, in particular salaries and remittances, tend to be associated with worsening inequality, although other sources o f income tend to offset inequality. Infrastructure 27. Improved infrastructure access has been associated with movements out of poverty. One reason why infrastructure variables are important is that households that are closer to a fertilizer seller, a motorable road and/or piped water grow more crops (in absolute number) and are more likely to diversify. E-Figure 4 below presents the good newsthat access to all these key services improved over the decade, andmore rapidly for the poor. The initially observed gaps in access had significantly narrowed by the end o f the period. This likely reflects the significant increase in public investment in infrastructure, described below. E-Figure 4: Reduced Distancesto Various Infrastructure, 1997-2007 1 t 0.2 0 1 0 1997 2000 2004 2007 -Poorest ---Poorest Fertilizer ----Richest - - Fertilizer E xtens ion Extension Poorest Motorable R oad +Richest R ichest Motorable Road Source: Nyoro et a12008 xviii Diversification 28. There has been a very strong upward trend in diversification of economic activities among rural households over the decade. Given the importance of this strategy as a positive driver out o f poverty, it is critical to identify which factors are associated with rural households pursuing this route. Tegemeo's diversification regressions show that the key determinants o f the participation o f households in off-farm activities are driven by availability o f credit, education and access to roads. There are also large differences across provinces, in that households in Eastern appeared to diversifymuchmore frequently than those inNyanza and Western. Age-sex composition o f the household also matters, inthat households with young adults are more likely to be involved inoff-farm activities. 29. On-farm diversificationis also important. The number o f crops plantedinthe main season has increased dramatically over time-the extent o f increase ranged from doubling over the decade in Eastern and Central, and close to tripling in Nyanza and Western. The number o f crops per field inthe main season evidenced dramatic increases, doubling in Western, and increasing by about half elsewhere. In terms o f determinants, on the negative side, the number o f crops tends to be fewer among households with more uneducated adults, andthose with less land. Access to Credit 30. One robust finding of the rural panel analysis is the importance of access to credit as a correlate of escaping from poverty. However, most (70 percent of) Kenyans do not have access to any form o f credit, with even lower rates o f access in arid lands. For those who do access loans, there is significant disparity inaverage amounts- with non-poor households borrowing 42 percent above the average amount, and poor households 77 percent below; the advantage for urbanhouseholds is even larger, whereas female headed are way below average. 31. At the same time, the financial services sector in Kenya is increasingly dynamic. There is evidence o f significant innovationandexpansionover the past several years, including in group-based mechanisms and the use o f mobile phones. Thus for example, while youth, like most Kenyans have very limited access to financial services, microfinance institutions tend to be more constrained by the lack o f business support services for start-up businesses than by liquidity, suggesting the need for greater attention on that front. 32. While access to credit is important, the traditional Development Finance Institutions in Kenya have not promotedaccess in ways that are either pro-poor or sustainable. The more promising routes appear to be grounded in the continued expansion o f group-based lending approaches. Inequality inLand Ownership 33. Only 20 percentof the land is classified as mediumto highpotential,but this i s where about three out of four Kenyanslive. Given the rising population size and the largely rural nature o f income earning activities, pressure on land i s growing. Land is xix also a highprofile issue, inpart because o f the well-documented illegal appropriations o f public landinthe late 1980sandthroughout the 1990s. 34. The KPIA focuses on levels and trends in land inequality, but does not attempt to fully deal with land questions. Land inequality as measured by the gini coefficient was 0.61, based on reported size o f ownership in 1997. Nationally, there was a 36 percent increase inthe gini over the period 1997-2005/6, to about 0.83 for the entire population. The worsening was especially marked inCoast and Nyanza provinces, while levels of inequality are remarkably high not only in Nairobi, but also the Rift Valley. Even among landholders only, it is notable that the reported levels o f inequality are much higher than the estimated consumption ginis, and highalso by international standards. 35. Rates of poverty among landless agricultural laborers are very high, although the evidence also shows that landlessness and poverty should not be equated in Kenya. Approximately equal shares o fthe poor have, and have no, land, and many o f the non-poor do not have land-and hence landlessness per se is not a good proxy for poverty. 36. The increase in land inequality in most parts o f the country over the past decade has been at a rate that suggests that this issue is rapidly becoming more serious in objective terms as well as in terms o f political salience. There i s a well-established policy agenda and hopefully now improved momentum to address this challenge. It i s recommended that the accountability o f the relevant institutions as well as the issues o f taxation o f unused arable land be addressed in the upcoming discussion o f the land agenda. There is a need to update policy proposals to reflect recent events, especially displacement and prospective returns. 37. Insum, the analysis ofdrivers ofpoverty and inequality underline the importance o f economic growth, as well as the need for attention to the distribution o f opportunities. The rural agenda highlights importance o f ongoing improvements to infrastructure, especially motorable roads, to improve access to markets, and the potential for extension services to enable diversification-both factors are associated with off-farm activities, andescapesout ofpoverty. TRENDS HUMAN IN CAPITAL ANDACCESS SERVICES TO 38. On the role of human capital in poverty trends in Kenya, the KPIA finds contrasting patterns. On the positive side, since 2003, we see the expansion o f basic education and the steady decline in the share o f firms reporting inadequate skills as a major or severe constraint to growth. On the other hand, maintaining quality inschooling is a major challenge, as is the expansion o f the system at the secondary and upper levels as primary education alone is not necessarily a route out o f poverty. And there are major health challenges, most prominently malaria and HIV, but also other causes o f morbidity and mortality which adversely affect, among other things, labour productivity, and thus individual and household earnings prospects. 39. In 2003, both the level and trends in most Millennium Development Goals were dire in Kenya (E-Table 2). Over the preceding decade, there had been general regress, and in particular, access o f the poorest quintile worsened on several fronts, including water, sanitation, telephones and schooling. Rates o f access for the bottom xx quintile, which were far below average, deteriorated further. Significant provincial level variations were evident. For example, the risk of infant death in Nyanza and North EasternProvinces are over six times that incentral province. E-Table 2: Access to Selected Services by Quintile, 1993 200516 -- All Bottom Second Third Fourth TOP 93 98 03 05 93 98 03 05 93 98 03 05 93 98 03 05 93 98 03 05 93 98 03 05 Electricity 1 1 1 5 1 6 1 6 0 0 0 0 0 0 0 2 0 0 0 8 1 2 4 2 0 4 7 6 0 5 7 5 3 NetPrimary Enrollment 75 85 79 81 71 76 61 73 74 85 80 80 72 88 84 84 76 90 88 86 81 90 86 89 Net Secondary Enrollment 7 10 13 18 4 4 4 6 6 5 7 12 5 9 12 21 6 12 16 30 17 22 29 52 GrossSecondary Enrollment 19 23 23 40 14 9 9 16 13 18 19 34 15 21 22 50 15 23 29 59 37 44 41 84 Source: DHS, 1993, 1998,2003, andKIHBS 2005/6. *DHS analysisusesasset indices,KIHBSusesa consumptionper-capitaindex. 40. However, important gains began to emerge in the 2003-2005/6-period. As i s well known, the introduction of free primary education in 2003 was associated with significant increases in enrollment. This has happened across all provinces, and, for example, the surveys show a 50 percent increase in gross primary enrollment in traditionally lagging North Eastern. Tegemeo's rural analysis shows that over the decade to 2007, the fraction of adults inrural householdswith no education had fallen, and more rapidly for the poor. Although only 70 percent of children in the poorest quintile were enrolled in primary school in 2006/6, the prognosis for broad based improvements in labour quality in Kenya appears to be good, and trends in the distribution of education appear to be an equalizing force. 41. The data clearly suggests that more schooling is associated with better jobs and thus a greater likelihood of not being poor. Kenyans with higher education tend to find jobs in the formal sector. In the KIHBS, about three-quarters of undergraduates and 81 percent of post graduates has formal sectorjobs, as opposedto 14percent of those with secondaryeducation and 5 percent of those with primary education. Also, the crude risk of unemployment decreases as education attainment increased. Likewise, for urban wages, there appears to be a significant premium for higher education, especially for men. On the demand side of the labor market, there has been a marked decline in the share of firms reporting inadequate skills as amajor or severe constraint to growth. 42. It remains the case that lack of education is associated with poverty, and to this extent, the Government and families' strategy of getting children into school makes economic sense, as well as having other justifications. At the same time, an important finding from the diagnostic of drivers of poverty over time is that more than primary education i s needed for individuals and households to move out of poverty, especially in urbanareas. This underlinesthe pressures onpost. Indeedthose with completed primary education earn only about half the average. Hence it is not surprising that the increased xxi number of primary graduates is putting great pressure on the limited secondary school places available. Inrecent years, the transition rate for children from primary to public secondary schools has risen to 51 percent; although the enrolment rate for children inthe bottom two quintiles i s much lower than average (see below). Moreover, maintaining quality in schooling is also a major challenge, which the government has recognized and i s seekingto addressthrough new hiring andtraining o f teachers. 43. Looking ahead continued efforts are needed to raise enrolments among the poorest, for whom the net primary enrolment rate is still only 70 percent. It i s particularly low in Coast, Rift Valley and North Eastern provinces, where the policy could be focused. Impact evaluation studies have shown that targeted support to poor households, for example through the provision of school uniforms, and actions to improve health (for example de-worming) does raise school attendance. 44. Access to other services has also expanded. Lookingby province at household electrification, for example, Rift Valley, Nyanza and Western all saw large increases relative to the low levels reported in 2003. Water and sanitation access increasedacross all provinces inthe 2003-2005/6 period. Still, about half of Kenyans do not have access to water and sanitation services meeting minimal standards o f quality. Access is even lower in rural areas, and significantly lower among the poor, with significant regional variation. 45. Major healthchallengesremain. Most prominent are malaria andHIV, but also other causes o f morbidity and mortality which adversely affect, among other things, labour productivity, and thus individual and household earnings prospects. Health outcomes in Kenya are weak, especially for the poor. This i s associated with poverty, lack of access to improved water, and low access to, and usage of, preventive services. Major health shocks (including death) are common and are causally related to falling into poverty. Inthe next section we highlight the extent to which public health services are reaching the poor. THEROLEOFINSTITUTIONS:FocusONTHEBUDGET 46. Institutions-including those relatedto the public sector-play a key role in mediating the effects of inequality onto outcomes. The design and especially the implementation of government policies and programs will affect the distribution of benefits. 47. The KPIA looks at various aspects of governance,includingcorruption, since this is an obstacle to poverty reduction in Kenya. Most public discussion in Kenya about corruption makesunequivocal connections to poverty and inequality. Corruption is especially associated with inequality-in particular, upward social mobility and accumulation of wealth. While robust empirical evidence has been lacking on the poverty links, the KPIA brings together some relevant evidence, focusing on petty corruption. The culprits are typically frontline service providers, that is, those with whom people are most likely to interact, e.g., police officers. 48. Household surveys suggest that corruption at lower levels of government is pervasive in Kenya, and the urban poor are especially vulnerable to bribery encounters. According to Transparency International's annual survey, in2006 Kenyans xxii encountered bribery in one out of every two official transactions, and paid, on average, 2.5 bribes annually. This i s common across all socio-economic groups, although the incidence i s significantly higher among urban residents, because they deal with officialdom more often. The good news is that the 2007 TI Bribery Index shows that reforming agencies were reducing corruption ~ignificantly.~ 49. Progress in combating corruption is important for the poor. The recent National Corruption Perception Survey found an inverse relation between the burden of bribery and level of household income (Kenya Anti-Corruption Commission, 2006). Indeedthe burdenof bribesis comparableto the share ofhousehold spendingon water, for example. A companion report (ICA 2008) shows that corruption is also a heavy burden'for informaloperators, which is a major source o f employment for the poor. The Budget 50. The budget is a major potential instrument to improve the pro-poor orientation of public policy. The poverty diagnostics suggest that market access, through improved infrastructure, and education can be important drivers out o f poverty, although we also saw large provincial differences in access to basic services, and outcomes. Inthis context, we examine trends inthe sectoral allocation of the budget over the period 1999/00 to 2006/7 (E-Figure 5). Analysis i s constrained by the lack of functional budget classifications and the absence of provincially disaggregated fiscal data, butthe highlights are as follows: 0 There was a significant increase in real spending levels, in a context of increased revenue efforts and progress in improving the debt financing mix. Total sectoral spending rose from less than 14 to about 20 percent o f GDP. Given high economic growth, this meantthat the budget more than doubled in real terms. 0 Overall spending on priority sectors4efined as agriculture and rural development, infrastructure, and health and education-retained a steady share at about 58 percent of the sectoral budget. Allocations for infrastructure rose relatively fast, which is welcome, and startedto redress the depreciation of the 1990s. Less welcome i s the very rapid rise inpublic administration, quadrupling over theperiod. 5 1. The Kenyanbudget has become increasingly decentralized, with a series of extra- budgetary funds-including the Constituency Development Fund (CDF) and Local Authority Transfer Fund (LATF)-and allocations to the district level via line ministries. The spatial allocation of these budgets is of prime interest given the spatial disparities in poverty outcomes and access to services. Unfortunately, however the available fiscal data do not allow one to gain a comprehensive picture of provincial budget allocations. Preliminary analysis suggests that the existing allocations, analyzed on a per capita basis, have not tended to favor the poorer provinces. The key recommendations which emerge See hm://~~~.tiken~a,org/viewdocument,as~?ID=269. xxiii relate to the need to improve the transparency and comprehensiveness o f information about decentralized spending, and to review existing formula in light o f explicit objectives for growthandpoverty reduction. E-Figure 5: Index of Real Trends in Sectoral Spending, 1999/2000-2006/7 (1999=100) +Agriculture and Rural de1 400 - n 350 - -ePhysicalInfrastructure --tHealth -a- Education 200 -0-PublicSafety Lawand 150 Order 100 4- Public Administration -h-SectoralTotal 52. Core Poverty Programs were introducedin 2002 as a novel way to improve the pro-poorfocus of spendingbut havenot beenvery successful. The criteria are not clearly specified and the rationale for minimum spending targets is not obvious. In practice this has not mattered, since the objective o f ensuring full disbursement o f CPP spending has not been met. However, the approach warrants reconsideration, in context o f more comprehensive efforts to improve transparency and the coherence o f pro-poor spending efforts. 53. More generally, the analysis shows significant scope for more pro-poor and pro-growth budget allocations, across both spatial and functional dimensions. Increased allocations for agriculture-related public goods (such as extension), in the context o f ongoing efforts to improve the relevance and accountability o f services appears, fiom the diagnostics, to have important payoffs at the farm level. 54. Major improvements are needed in fiscal management-in particular in the comprehensiveness and transparency of the budget information, to enable more informed discussions about the budget, and improved monitoring. At present it is virtually impossible to gain a consolidated account o f spending by province, for example. This stands in sharp contrast to Ethiopia, which is much larger, and much more decentralized, but which also manages to generate quarterly out-turns, by woreda and region, with minimal lags. 55. The large and persistent deviation of actual spending from plans approved by parliament, especially for development budgets, is a major issue. This significantly reduces the credibility o f the budget as a policy instrument and indicator o f government intentions. This underlines the need for further progress in improving the budget planning process, but also to the importance o f increasing the capacity o f key xxiv actors to improve scrutiny o f the budget, in particular the relevant parliamentary committees. 56. The financing and provision of education is a major element of the Government'sdevelopmentand povertyreductionstrategy. Education absorbs about 30 percent o f sectoral spending. Public spending on primary education in Kenya i s generally progressive, that is, the benefits poorer households receive relative to their income are far greater than those received by the better-off (E-Figure 6). From a gender perspective, the record o f girls attending school has been fairly good over time. E-Figure6: Benefit Incidenceof Government Education Spendingby Level and Quintile,2005/06 (percent) I 80 70 60 1 50 40 PI 30 20 10 I 0 Primary Secondary Tertiary All education 57. At present, the bottom quintile benefits significantly from primary education,but relativelyless from secondary school spendingand hardly at all from tertiary spending. To frame appropriate interventions, research is needed on the main constraints facing poor families in enrolling children in secondary school. Candidate factors include poor educational attainment at primary level, costs o f schooling, distance and perceptions about returns to secondary schooling, as well as health-related factors. The relatively sparse distribution o f secondary schools makes access difficult and costly to poorer rural Kenyans. The government recently announced free secondary education, and bursaries have potential to make access affordable. However recent evaluations suggest that implementation should be made more transparent and actions taken to ensure that poorer households are included, especially since that i s now the explicit aim o f the scheme. 58. Analysis of health spending from the perspective of poverty and inequality suggests that poor people tend to use lower levels of services. Yet the share o f the budget allocated to higher levels o f care has remained large, with front line rural services relatively neglected. Overall health spending i s regressive, in that the bottom quintiles get less thanproportional benefits from total public healthspending (E-Figure 7). Halfo f the spendingon referral hospitals goes to the top quintile, and less than 10 percent to the bottom quintile. The bottom quintile's benefits from spending on primary facilities are xxv proportional (20 percent). Alongside this pattern, analysis o f private spending on health show very significant disparities. Public subsidies do little to offset this. E-Figure 7: BenefitIncidenceof Government HealthSpendingby Level and Quintile,2005/06 (percent) 60 I 50 49 poorest second 0third 40 Percent I 30 27 20 10 0 Referral hospital District/ provincial Primaryfacilities Allpublic facilities hospital ;ource: staffestimatesbasedon budgetdata and KIHBS2005/06. 59. The key health spending challenges relate to the inadequate quality o f care at the primary level, and enabling access to district and provincial hospitals. The latter in turn would call for two types o f intervention, to improve access to district and provincial hospitals-better all-year roads, and to improve the implementation o f fee exemptions, which do not function well. Weak budget management capacity o f the line ministry, evident inthe low disbursement rates, clearly needs to be addressed. Monitoring; and Evaluation 60. It is crucial to underline the importance of progress in public sector performance and in particular capacity for monitoring and evaluation. The government has introduced several reforms aimed at strengthening monitoring, including the new statistics legislation and the creation o f an autonomous KNBS. However, performance in general has been weak, and it is recommended that the government, through its medium term plan and related initiatives, move more actively to strengthen underlying capacity and to help promote more evidence-based approaches. It is also noted with concern that the draft Medium Term Plan, 2008-2012 neglected the topic o f monitoring and evaluation altogether. 61. Among the policy options outlined in the KPIA are short term, crisis-related initiatives, as well as long overdue measures to improve access and dissemination, and measures to enhance the role o f non-state actors. Among others, we flag the needs to finalize the preparation o f data dissemination and access policy for official data, and release to the public micro-data files o f all household surveys (including the 2005/06 KIHBS), so that policy makers can benefit from broader analysis. We also recommend xxvi increased involvement o f non-state actors in designing and administering surveys. This i s important because the KNBS has often been overwhelmed in dealing with the demand for surveys from development partners, researchers, and others, which can lead to delays indata collection andprocessing, andquality problems. DISADVANTAGED GROUPS 62. A better understanding of the profile of disadvantage can inform ongoing public and policy debates. The Executive Summary presents brief highlights and policy implications related to two groups-namely unemployed youth and the slums. These groups are obviously not the sum total o f those who tend to have been excluded from the benefits o f development in Kenya, which importantly includes pastoralists (on whom much analysis i s available) and some groups o f women, urban widows, for example. At the same time, unemployed youth and slum residents represent particularly poor and large groups, and given the high profile nature o f their situation a selective focus i s justified at this point. We highlight needs for strategic approaches and targeted programs, informed by the diagnostic. In general, the directions suggested here are merely indicative, and more work i s neededto develop concrete options. Youth Unemployment and Joblessness 63. The overall unemployment rates for youth are very high, double the adult average, at about 21 percent(E-Figure 8). Youth fare the worst incomparisonto adults inWestern andNyanza, where the youth unemployment rate is about triple the adult rate respectively; North Eastern province has the smallest ratio (1.5). As noted above, three quarters o f unemployed women are youth. Thus youth unemployment is clearly a generalized issue inKenya, and not restrictedto young males inlarger towns. E-Figure 8: UnemploymentRate, Youth Versus Adult, by Province(percent of active in cohort) 60 50 40 30 20 10 0 North Nairobi Coastal Rift Western Central Eastern Nyanza Eastern Vallev EX3Youth 15-29 tEZZl Adult 30-64 -Youth NationalAveracae J(tAdult NationalAverage 64. While the incidence o f poverty is higher among households with unemployed youth, youth who are unemployed tend to be better educated than employed youth, suggesting that there may be an element o f search and/or frictional unemployment. Moreover, the relatively good levels o f education may limit take-up o f public works schemes (E-Figure 9). xxvii E-Figure 9: Level of EducationCompleted(percentage of cohort), 200516 Employed ~ I I I 1 I I I Jnemployed I I I I i 0 20 40 60 80 65. Looking at joblessness-i.e., individuals who do not have a job, excluding students-suggests that the magnitude o f the problem i s even larger for youth. If we aggregate the rates o f reported unemployment and inactivity, the magnitude is such that 38 percent o f youth-almost two out o f five young Kenyans-are neither in school nor work. Indeed it i s notable that the jobless number exceeds that in paid work. The provincial picture worsens for Central if we look at inactivity as well as unemployment, while Western andNyanza remain the worst. 66. The recent crisis has prompted the government to renew its commitment to addressing youth issues, considering the potential impact that would accrue from engaging them in public works schemes. International experience suggests that the success o f these programs depends on several factors, which need to be carefully considered inthe Kenyan context. For example, the wage should be set at lower than the prevailing market rate for unskilled workers to allow for self-selection. A low wage would ensure that only the unemployed would participate, thus achieving better targeting by the program. However, especially for youth, this level would be very low in Kenya. Lessons should also be drawn from existing public works programs, such as "Roads 2000," "Trees for Jobs" and possibly others to informthe designo f a scale up. 67. Public works are, at best, a temporary measure which can help alleviate the crisis, and make some inroads into the large numbers o fjobless youth. However the empirical findings o fthe KPIA underline that there are no simple solutions to youth unemployment, and suggest the need for further analysis, drawing on regional and international experience. In order to effectively consider longer term structural issues, existing skills development and microfinance initiatives aimed at unemployed youth need to be reviewed. Firms in Kenya have rated workers' skills as a much lower constraint to their growth compared to tax and regulation, infrastructure services, corruption and cost o f finance (see I C A 2008). While new market entrants are increasingly well educated, the incidence o f unemployment among youth is high. Further data and analysis- quantitative and qualitative-is needed to inform appropriate policy and programmatic responses. xxviii The Slums 68. Analyzes of informal settlements in Nairobi underlines that poverty incidence is h i g h 4 3 percent of Nairobi's slum residents fall below the poverty line in2006. At the same time, working with APHRC,we report significant heterogeneity within and across slums - within Korogocho, for example, the difference in the poverty rate betweenthe poorest andwealthiest village i s 31percentagepoints. 69. Slum residents score poorly on non-income measures of poverty as well- unemployment levels are high (26 percent), only a minority have secondary level education or more (24 percent), and living conditions are appalling with only 3 percent of households living in housing with a permanent wall, access to piped water, and an electricity connection. There is very significant mobility into and out of the slums, with about halfthe sample moving out every year. 70. Among slum residents who did not exit, we find considerable number of transitions, both into and out of poverty. In particular, despite the overall decline in the headcount, among the minority who hadbeen non-poor in2003, only about halfwere still above the poverty line in 2006. Conversely, almost a third (31 percent) of those poor in 2003 were no longer below the poverty line in2006 (E-Figure 10). E-Figure 10: Poverty Transitions in the Nairobi Slums, 2003 -2006 Above poverty line 2003 Belowpoverty line 2003 71. What are appropriate policy options in this context? Clearly poverty rates are very high, and enabling access to better income earning opportunities is a priority, including consideration of public works types programs, targeted at specific groups, in particular youth. At the same time, amenities in slums warrant attention. Garbage disposal seems to be an obvious candidate. But why are living conditions so bad? Recent researchpoints to the lack of home ownership and a highturnover rate-about 85 percent of Nairobi's slum residents are tenants and the average resident stays in a home for about five years. By comparison, inDakar, only a quarter i s tenants and the turnover rate i s low, with residents having stayed in the same home for an average of 19 years. Strikingly, living conditions inNairobi's slums are far worse than those in Dakar, even though they are poorer andworse educatedthan their Kenyan counterparts. 72. The diagnostic suggests that innovative approaches are needed in Nairobi. For instance, the widely-used approach of upgrading infrastructure in the slums, without dealing with the tenure issue, is arguably unlikely to work-if improvements in xxix infrastructure are likely to raise rents and benefit the "shack owners" but displace many of the current tenants. Inthis sense, the large scale Kenya Slum Upgrading Program (KENSUP) launched in2004 may warrant strategic review to assess the scope for greater innovation. CONCLUSIONS 73. The KPIA examines the most recently available evidence to inform current debates about the nature of poverty and inequality in Kenya, and appropriate policy directions. We seek to add value to existing knowledge by focusing on and drivers of poverty, by exploiting panel data collected over time. We confirm that there has been some improvement in poverty overall since 1997, but poverty i s still very pervasive, especially in rural areas. Inequality is large and appears to have risen over time. The main correlates o f poverty resonatewith those found in earlier studies-including family size, lack of education and frequency of shocks. The analysis does not reveal significant poverty reducing momentum in the labour market, at least through 2005/6. The picture that i s emerging presents major policy challenges. There was little growth inemployment inmodern sectors, whichremain arelatively small share ofthe economy. Earnings were low, especially inagriculture, andthe level of inequality inearnings was large. 74. On the policy front, there is some good news-for example the expansion of infrastructure appears to be bringing benefits to the poor, by increasing their access to markets and services. The continuing emphasis on education is also bringing dividends. There appears to be some momentum interms of diversification, both on and off farm, at least innon-pastoral areas. 75. Yet there are major challenges for policymakers. As the authorities proceed to finalize the Medium Term Plan, one clear implication from the analysis i s that the poverty and inequality agenda i s cross-cutting, and there is scope for innovation of traditional approaches. Among the topics warranting further consideration are appropriate and coherent responses to the very large regional variation, and the present exclusion of large groups, including many youth, as well as slum residents, from the benefits of growth and development. 76. Among the major challenges warranting further consideration are those of land inequality. Corruption is by now also a well-established issue, but continued progress is critical. There are also important messages related to budget allocations-the scope for greater pro-poor allocations, but also the need for much better information and transparency to facilitate analysis, in particular to get a better handle on spatial dimensions. The more general need for improved access to relevant data access is underlined, to inform policy making and to enable strengthened monitoring and evaluation. XXX 1. CONTEXT INTRODUCTION AND OBJECTIVES 1.1 This assessment o f poverty and inequality comes at an important juncture for Kenya. The recent elections and pronouncements o f the newly formed Grand Coalition have underlined the salience o f these issues to ordinary Kenyans, and for policy makers. The violence in January-March, 2008 highlighted the importance of addressing poverty and inequality as major goals intheir own right, but also for instrumental reasons-since growth i s unlikely to be sustained in the presence o f severe inequalities which increase the risk o f conflict. The global slowdown that began to affect Kenya after August 2008 has also shown how external shocks can exacerbate poverty and public service delivery related vulnerabilities. Cumulatively, these factors underline the value o f appropriate diagnostics about the patterns o f poverty and inequality in informing public debates, strategies and actions to overcome exclusion and to design policies to minimize the impact o f the current crisis. Considerably slower economic growth, projected at around 2.3 percent for 2008, and higher inflation at 25 percent, have implications for poverty requiringnewways to tackle the problem. 1.2 In this context, the KPIA attempts to address four broad policy-motivated questions o f major relevance to Kenya. First, what has happened to poverty and inequality over time, with a focus on new findings which deepen and extend our knowledge base; and second, what are the major dimensions o f poverty and inequality. These first two questions are addressed inPart 11. The third broad issue revolves around such questions as whether the opportunities generated by economic growth were broadly or narrowly shared, with major population groups largely excluded, and who manages to escape from poverty, and why? These are addressed in Part 111, on drivers o f observed trends. And fourth, what i s the pattern o f public investments and services, and how do these interact with poverty? This i s tackled in Part IVYalongside selected policy implications o fthe diagnostic. 1.3 This part is designed to set the stage for the analysis, and begins with an overview o f Kenya's demographic, social, economic and spatial characteristics, reviewing longer term and recent trends in the economy and labour market. Those familiar with Kenya may wish to move directly to Part 11. LAND, PEOPLEAND SPACE: ANOVERVIEW 1.4 Kenya has an area o f almost 58 million hectares. Only 20 percent o f the landi s classified, on the basis o f rainfall, as medium to high potential, andthe rest is mainly arid 1 and semi-arid lands(ASALs). About three out of four Kenyans lives withinthe relatively scarce medium to high potential areas. At the same time, while Kenya's urban population has grown steadily in recent decades, reflecting internal migration, by 2005, about 80 percent o f Kenyans still lived in rural areas. So high rural population density, rather than large urban shares, characterizethe current picture, although of course this is evolving over time. 1.5 The spatial distribution of the medium and high potential lands i s uneven. For example, while nearly all of Nyanza and Western provinces, and 85 percent of Central Province, are classified as medium to high potential, less than 25 percent of land in the remaining provinces i s classified as such. North Eastern province i s entirely low potential. 1.6 On the population front, the rate of growth has decelerated, but remains an issue, especially for the poor. The rate averaged 2.6 percent annually in recent years. It i s expected to decelerate to 1.9 percent by 2030, with the fertility rate falling from 5 in 2005 to 3.2, and life expectancy at birthrising from about 54 in2005 to 61 in2030. The increase in the working age (15-64 years) share and reduced fertility will reduce the age dependency ratio which had been among the highest inthe world-from 0.83 in 2005 to 0.63 in2030 (Figure 1-1). Figure 1-1: Kenya and SelectedComparator Countries: Age Dependency Ratio , , , , , , , / , , , , , / , , / , , , , , , , , , , , , ! 5 5 ! $b$b $b $c $o $ $ $8$8$3 02 I I / / I I I I I I I I I I I I I O O ( D Q ) ( u I C ) C O r 8 6 8 (u Source: Staff calculations based on World Bank, WDI. 1.7 The share of youth (15-29 years) has risen slightly inthe last two decades andis expected to remain around 30 percent. The infant (0-4) and child (5-14) ratio to total population fell between1988 and 2005, and i s expectedto fall further (Figure 1-2). 2 Figure 1-2: Age Structure of Population, Selected Cohorts, 1988,2005 and 2030 I m////A ........................ ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' I " " " 1 2030 ....................... ..................... ...................... ...................... ..................... ...................... ..................... l i ...................... . . ....................... . . . .................... ..................... . . . . . . . . . . . . . . . I 1988 ..................... ..................... . . . . . . . . . . . . . . I I 20% 40% 60% 80% 100% I 0-40 5-14 0 15-29 30-64 >64I I Source: Kenya Population Census, 1988;KIHBS'OS and Staff Estimations. 1.8 Together with other Sub-Saharan African countries, such as Uganda, Chad, Liberia, Cameroon, Zaire, the Republic o f Congo and Madagascar, Kenya tops the rankings of ethnic, linguistic and religious fractionalization. More than 40 ethnic groups have been identified in Kenya, with the two largest tribes, Kikuyu and Luo, representing only slightly more thanone third oftotal population. 1.9 A well-established measure of ethnic diversity is the so-called ethno-linguistic fractionalization index, or ELF.' The lower ELF is, the higher the probability that two randomly selected people are from the same grouphribe. So the larger the geographical area, the higher we expect ELF to be. And for Kenya, the national and provincial ELF'S are typically higher than the district ones. Where this is not the case, it points to more heterogeneity-e.g., the ELF for Kiambu in Central Province is higher than the homogeneous other Central Province districts (and province overall) becauseof Kiambu's proximity to Nairobi and the number of migrants who have settled there to work in Nairobi. Figure 1-3 shows the results nationally, and by province and district. Kenya's overall ELFi s 0.885. The estimates presented here for Kenya are staff calculations based on census data (1989). 3 Figure 1-3: Ethno LinguisticFractionalization Index 1 0 0 9 Kenya ELF = 0.881 0 8 0 7 0 6 Nyanza 0 6 0 5 0 5 0 4 0 4 0 3 0 3 0 2 0 2 0 1 0 1 0 0 0 0 Source: Staff calculationsbasedon Kenya Population Census, 1989. ECONOMIC TRENDS 1.10 The Kenyan post independence story began with vigorous economic growth in the 1960s (averaging over 2 percent per capita) and the 1970s (3 percent). However, duringthe 1980s average growth slowed to 0.4 percent, and the economy actually shrank in the 1990s. Over the past 25 years, rates o f growth have been at best uneven, inthe absence of any period of sustained high, or even moderate, growth,6 as depicted inFigure 1-4. Alongside rapid population growth, income per capita by 2000, inpurchasing power parity terms, had fallen to levels registered at the end o f 1970s! 1.11 Overall, during the years in focus in this assessment, 1997-2006, real annual growth per capita was close to zero-only about 0.2 percent. However, given the earlier years o f recession, this average conceals the economic recovery which began in 2003, driven by horticulture, electricity and communications in the primary, secondary and tertiary sectors, respectively. 6See Mwega and Ndung'u (2002); N'Dulu, Benno and Stephen O'Connell (2000); World Bank (2003); Economic Survey (various years); Kumar (2006). 4 Figure 1-4: Annual Real GDP per Capita Growth, 1981 2006 - p 3% g *6F2% 1% 20 -1% 0% $ -2% -3% -4% -Annual realper capita GDP growth -trend (HP filter)1 Source:Economic Survey, CBS. 1.12 A recent report on growth (World Bank 2008a) decomposed the sources of growth, by sub-period, into changes in physical investment, changes in human capital (years of schooling and returns thereto), and a residual, called "total factor productivity" (TFP).7 Over the long term, 1960-2006, physical capital accumulation contributed about one-third of per capita growth, while humancapital accumulation explained about half. 1.13 For the period 1997-2006, there was some positive contribution to growth from humancapital, but physical investmentper capita did not rise and TFP was negative. Per capita growth had recovered more recently-2003 to 2006 - reflecting both increased investmentincapital, and TFP improvements. The latter likely partly reflected increased utilization of existing capacity. This trend i s consistent with the Investment Climate Assessment (World Bank 2008b, ICA 2008 hereafter). Based on firm level surveys in 2003 and 2007, the ICA found significant improvement in labor productivity and use of capital inputs, as well in TFP, and capacity utilization improved significantly. On average, firms tended to increase employment without a commensurate increase in physical capital. Table 1-1: Decompositionof Sourcesof EconomicGrowth, Selected Sub-periods, 1997-2006, Annual PercentageRatesof Change Period GDP Physicalcapital Human capital Total Factor Per capita Per capita Per capita Productivity Growth 1997-2003 -0.7 -0.2 0.6 0.1 2003-2006 2.6 0.8 0.1 1.8 1997-2006 0.1 0.0 0.4 -0.3 Source: Staff calculations. 'SeeWorld Bank (2008a). 5 1.14 The extent of structural change in the Kenyan economy has been surprisingly limited over the long term through the more recent period. The tertiary sector-that is, services-account for about half of total value added in the Kenyan economy, followed by primary activities, at about 30 percent (Table 1-2). The growth performance of the secondary sector has beenremarkably weak. Indeed, while there was evidence of recent TFP improvements at the firm level, there is no clear sign of productivity improvements resulting from reallocation of labor and capital across sectors to uses that are more productive. Table 1-2: SectoralShares in Value-Added GDP, and Contribution to Growth, 1997-2006,Percent Share in Value Added GDP (YO Total Contribution of Activity to Value points) Growth Added GDP Growth Rate (YOpoints) Primary Secondary Tertiary (Yo points) Primary Secondary Tertiary 1996 29.1 19.5 51.4 1997 27.8 19.5 52.7 -1.0 -1.5 -0.2 0.8 1998 28.9 18.7 52.3 3.3 2.0 -0.1 1.4 1999 29.8 17.8 52.4 2.1 1.5 -0.5 1.1 2000 29.5 17.7 52.8 0.4 -0.2 -0.1 0.7 2001 30.9 17.0 52.1 4.2 2.7 0.1 1.4 2002 29.7 17.3 53.0 0.6 -1.1 0.4 1.2 2003 29.5 17.8 52.7 3.0 0.7 1 .o 1.3 2004 28.9 17.8 53.3 3.8 0.5 0.7 2.6 2005 29.4 17.7 52.9 5.0 1.9 0.8 2.3 2006 29.2 17.7 53.1 5.7 1.5 0.9 3.2 ~~ ~~~ Source: Staff CalculationsBasedon Economic Survey. 1.15 Agriculture remains a critical sector in Kenya, for growth as well as poverty reduction. It accounts for a disproportionately large share o f total emdovment-about 60 percent4ouble its share of GDP, which immediately suggests that agricultural productivity and incomes are low. Four out of five Kenyans reside in rural areas, and most rely directly or indirectly on agriculture for their livelihoods. As we will see below, most of the poor still live inrural areas, and most people inrural areas are poor. 1.16 Finally, by way of introduction to broad economic developments, the employment elasticity o f growth is relevant. Per capita growth i s decomposed into several components, viz output per worker, employment rate, participation rates and the age dependency ratio for the period 1997-2005. The resulting estimate employment elasticity of growth i s just 0.54 percent. This begins to provide some clues about underlying challenges that we explore further below. SELECTED ASPECTS POLITICALECONOMY GOVERNANCE OF AND 1.17 Kenya has enjoyed decades of relative peace compared to most of her neighbors. At the same time, the aftermath of the disputed 2007 Presidential elections indicated deep-seatedunderlying tensions and the risks of future conflict. It i s far beyond the scope of the KPIA to fully understandthe deep roots and proximate causes of the crisis in early 2008. However, we attempt to frame the analysis in a context which 6 explicitly recognizes major elements in the history, political economy and governance which have affected patterns o f development inKenya. 1.18 At the outset, it is important to recognize that the Kenyan people do have a strong sense o f nation. According to Afrobarometer Survey Round I11(2005), 44 percent o f Kenyans think that they are as much Kenyan as they are members of their respective ethnic communities. Only 6 percent think ethnic first, whereas 24 percent put nation first. On the other hand, commentators have noted major economic and political cleavages in Kenya and that government policies over time, by favoring specific groups or coalitions thereof, tended to exacerbate these differences. * One example i s the promotion o f local vernacular, rather than the national language, with indigenous languages occupying prominent roles in the school c~rriculum.~Another is the pattern o f public investment andservices (Box 1-1). lo 1.19 Administratively Kenya i s divided into nine provinces which, except Nairobi, also broadly coincide with the patterns o f ethnic residence. Interms o f location, Kenya's larger ethnic communities tend to be associated with particular provinces, as summarized inTable 1-3." Table 1-3: Provincial Location of Different Groups (Percentage Shares) Share of `Major Share of Share of Rift Share of Total Major Location' Nairobi Valley Group Population Location Population Population Population 1969 2003 1969 2003 2003 1969 2003 Kikuyu 20.1 24.2 Central 96 91.8 34.2 15 21.2 Luhya 13.3 15 Western 88 83.8 1.3 7 11.2 Luo 13.9 10.4 Nyanza 63 53.4 2.5 NA 2.3 Kalenjin 10.9 7.9 Rift Valley 51 41.4 2.5 51 41.4 Source: Kanyinga (2007), taken from Nellis (1974), cited in Stewart (2008). 1969 figures from census, 2003 from DHS. 1.20 As shown below, Kenya is characterized by sharp inequalities across some key socioeconomic dimensions. A useful way to think about these is as horizontal inequalities across groups, which can be defined in different ways, including class, livelihood, ethnicity and location (Stewart 2002). Of course groups are constructed inthe sense that these reflect prevailing norms and attitudes, institutions and structures, and the boundaries between groups are fluid, not fixed. At the same time, analysis along "horizontal" lines can provide insights, and for Kenya, a salient and measurable group dimension is provincial location. 1.21 The KPIA looks at various differences across provinces, which are often significant, even controlling for other factors. Horizontal inequalities are important both * Earlycommentators included Rothchild (1969); Nellis (1974). For more recent discussions on land issues see: Kariuki, S. (2004). For athorough discussionon ethnicity and nation building inKenya see Miguel(2004). lo For a valuable review of regional differences in the provision of services see: Society for International Development (SID).Pulling A~art:Facts and Figures on Ineaualitv inKenya. Nairobi: SID, 2004, p.21. ''Langer,A.,A. MustaphaandF.Stewart (2007). 7 for economic reasons-the effects on growth andpoverty-and the effects on perceptions and political stability. There is also increasing international evidence that horizontal inequality i s related to conflict--e.g., Rwanda, Northern Ireland, Sudan and Nepal (Stewart 2008). Colonial Legacy andpost independence bargains underlineInequalities 1.22 British colonial settlement from the late nineteenth century through 1965 had two distinct outcomes inKenya. First, infrastructure development was unbalanced. Areas inhabited by the settlers tended to have better access to education, roads, and so on, as well as potential for commerce. Thus at independence, Central province had better access to education, health services and transport. Second, colonial settlement in these areas led to the emergence o f a class society based largely on land ownership. These two outcomes continue to characterize Kenya's patterns o f development. Subsequent horizontal inequalities were based on these initial advantages as well as relative representation in the public and private sector in the post independence regime. This is true despite the neglect o f Central province during much o f the M o i period. Provincial imbalances were exacerbated by personal rivalries and created a tendency toward ethno- regional rivalry that continues to pervade politics inKenya (Cohen, 1995; Muhula 2008). 1.23 Since independence, political bargaining and coalition formation has largely assumed regional dimensions. Several studies have underlined the existence and potential dangers inherent in these patterns. Rothchild (1969) and Nellis (1974) documented distinct imbalances in the pattern o f Cabinet and Public Service appointments inthe immediate'post independence years. These were updated by Barkan and Chege (1989), SID (2004) and Kanyinga (2006), who also pointed to their long-term implications for economic growth andpolitical stability. 1.24 Perceptions o f political inequality are evident. The 2005 Afrobarometer found that almost four in ten Kenyans felt that their ethnic community had less, or much less, political power than other communities. There is notable regional variation in the magnitude o f this perception. In Rift Valley, 42 percent o f respondents felt that their ethnic community had less political power than other groups in the country, as did 38 percent o f those in Nyanza and 72 percent in Coast, whereas only 7 percent in Central Kenya felt that way. While these data were collected during a period o f heightened political activity (the referendum), they are nevertheless o f interest. An Overview of GovernanceStructuresand Patterns 1.25 The Kenyan constitution provides for a strong executive, relative to the parliament and the judiciary. The system i s also very centralized. Whereas at independence, the majority o f public service delivery was undertaken by local authorities, revenue reforms in the 1970s increased local dependence on the central government. Responsibility for important public services was transferred to central line ministries. 1.26 The Kenyan constitution also confers vast powers to the President. These include allocation o f ministerial positions, and appointments to such key offices as Court 8 o f Appeal and High Court Justices, top military commanders and Ambassadors without reference to parliament. Successive Presidents have tended to seek to entrench their rule through assignation of strategic administrative positions, and directing resources to provinces aligned with the ruling group. There appears to have been a recurring pattern whereby regime insularity, and exclusion of nonaligned regional elites, increases over the tenure o f the regime. Thus between 1963 and 1978, and 2003 to 2007 Central province was disproportionately represented in Cabinet, Military, Diplomacy, and Public Service. A similar pattern was evident between 1979 and 2002, with Rift Valley dominating. These periods also coincided with the deliberate exclusion o f specific provinces whose political elite were considered disloyal to the ruling group. For example, the record shows that the proportion of principal secretary appointments from Nyanza and Eastern provinces fell during 1969-1972.12 A similar pattern can be observed more recently, when differences within the National Alliance Rainbow Coalition emerged. As a result, perceptions o f Central province hegemony escalated, and became manifest duringthe divisive referendumin2005, the 2007 election andthe events thereafter. Box 1-1: ProvincialBiasinthe HistoricalPatternof Public Good Provision The historical data suggest that the distribution of public goods such as education facilities, health, water and physical infrastructure in Kenya have tended to follow patterns of access to political power. Under Kenyatta, Central had better access to public services compared to other provinces. Similarly, under Moi, Rift Valley province received disproportionate share of government investment in public service delivery. In 1969 about two thirds of primary school aged children in Central Province were enrolled inschool, but less than one-third inNyanza and Rift Valley. During the Kenyatta regime, Central Kenya had the highest percentage of professionally qualified teachers, other than Nairobi. However this advantage was eroded during the 1990s when, for example, Central had the highest pupil-teacher ratio (34), compared to 28 for Rift Valley. Similarly, for health care, Central province was favored during the Kenyatta regime, then the Moi era brought similar advantages for Rift Valley province. By 2002 Rift valley province had about 6,000 people per health facility-the least of any province. Nyanza had about 9,000 while Westernhad 11,000 and sparsely populatedNorth-Easternprovince about 14,000 people per health facility. InPart I1and IV below, we examinemorerecent evidenceto see whether these historical patternshave persisted. ISource: Central Bureauof Statistics.Kenva. 1994. Cited inAlwv. Alwiva and Susanne Schech. I 1.27 The performance o f the national parliament has been mixed. Important oversight organs like the Public Investment Committee and the Parliamentary Accounts Committee do exist but have not hlfilled their p0tentia1.l~The fact that parliament does not control its own calendar and can be prorogued and dissolved by the president undermines its independence. Parliament has little scrutiny o f the budget, and members only encounter the budget bill on the day it is presented by the Minister for Finance. Annual reports o f the Controller and Auditor-General routinely reveal politically l2Nellis, J (2004), p. 14-15; Kanyinga (2006), p. 274-275. This coincided with a rift between Central and Nyanza province's elites, leading to a new opposition party, and the alleged iilvolvement of military officers from Eastern in the 1971 mutiny. A similar pattern occurred between 1985 and 2002, with Rift Valley securing more representation, Central share in Cabinet dropping from 20 to 4 percent, while the Nyanza's from 16to 4 percent over the same period. l3Global IntegrityReport (2007) indicatesthat Kenya scored only 25 out of 100 on the questionof whether government acts on the findings of supreme audit agencies such as the Controller and Auditor General. Government tends to ignore recommendations of PAC and PIC for prosecution of major offences. See Global IntegrityReport on the web: http://report,globalintegrity.org/Kenya/2007/scorecard/75. 9 motivated and economically unviable projects, while recommendations o f the Public Accounts Committee are rarely implemented. On the other hand, the fact that Kenyan MPs are the best paid on the continent and their continued agitation for additional benefits tends to erode their moral authority. It i s perhaps not surprising that a recent study on the strength o f legislatures worldwide ranked the Kenyan parliament at 126 out o f 158.14 Box 1-2: Youth and Politics Youth aged 15-29 years comprise over 30 percent o f the Kenyan population. They are a rapidly growing and increasingly vocal group, and an important political constituency intheir own right. Surveys point to a politically active as well as a highly volatile demographic. About 68 percent o f youth indicate that they were very interested in public affairs, while 62 percent indicated that they discuss political matters frequently (Afro-barometer). Several factors account for the political salience o f this group, including its relative mobility, levels o f education and, as we see below, high o f unemployment andjoblessness. Indeed they may well be the most frustrated and disaffected group o f Kenyans, and therefore the most vulnerable to political mobilization. An emerging phenomenoninKenyanpolitics is the rise o f youth militias. Some o f these militias have been known to work for individual politicians. The rise o f the Mungiki in Nairobi, Nakuru and parts o f Central province, as well as the emergence o f the Land Defense Force inWestern province and parts o f RiftValley could be attributed to the socio-economic and political exclusion o f the youth. High unemployment and, poverty, and a manipulative and predatory political environment collectively heighten perceptions o f exclusion. 1.28 Youth engagement in politics has rarely assumed organized forms in terms o f agenda setting. Traditionally, the youth have been used as party agents and as party "foot soldiers", including during the post election violence in 2008. At the same time, and despite the high number o f young people registering as candidates for parliamentary nominations, there was no organized articulation o f a unified youth agenda. 1.29 The post election violence in 2007 demonstrated how violently engaged youth were in the conflict. According to Youth Agenda, a local NGO, while the youth were responsible for the planning o f only about 7 percent o f the violence, they were responsible for the execution o f about 55 percent o f election related ~iolence.'~The same source reported similar trends in the funding for violence. About 60 percent o f funding for violence was sourced from non-youth sponsors, while youth sponsorship o f violent activity accounted for only 7 percent. These figures are circumstantial evidence o f a chain inwhich youth tendto act as conduits or agents (Muhula 2008). 1.30 To the extent that Kenya's ruling elites continue to use political power as a source o f patronage, group based advantages could be expected to be promoted. This in turn would imply that horizontal inequality would remain a persistent feature o f socio- economic and political structures in Kenya, reflected in unequal access to services and economic opportunities. These issues could be addressed and policies to address to do are the focus o fthe final part o fthe report. 14 ''Youth Fishand Koenig study on strengthoflegislatures. Agenda, 2008, p.25. 10 2. THE EXTENTAND NATUREOF POVERTYAND INEQUALITY ' INTRODUCTION 2.1 This report examines the most recently available evidence about the extent and nature of poverty and inequality in Kenya. Inequality has been increasingly recognized not only in mainstream development debates (WDR 2006), but also in the specific context of Kenya. 2.2 The basic results do not reveal major surprises. We see some improvement in poverty overall since 1997, poverty is still very pervasive, especially in rural areas. Inequality i s large and growing. The main correlates of poverty resonate with those found in earlier studies-including family size, lack of education and frequency of shocks. We seek to add value to existing knowledge by focusing on regional variation, youth, the slums and drivers of poverty, by exploitingpaneldata collected over time. 2.3 This assessment adopts a multi-dimensional approach. Alongside the most common methods used to measure poverty, based on consumption and incomes levels, we seek to measure other aspects of well-being and deprivation. We have assembled comparable indicators for education, health and access to basic services, as well as measures of vulnerability. Participatory approaches are exploited to better capture the nature of poverty, andrisks as well as opportunities perceived by the poor. 2.4 A range of data sources, both quantitative and qualitative, are drawn upon. The recently released Kenya Integrated Household Budget Survey (KIHBS 2005-2006), and the basic official report on poverty (KNBS, 2007a and 2007b) is complemented by a range of other household and firm surveys, and census datasets. Inparticular, we draw on a rich panel data set that is representative of rural Kenya (with the exception of the pastoralist areas which are about 5 percent of the population), which is associated with the Tegemeo Institute. And given the heightened importance of understanding disadvantaged groups inurbanareas, we also exploit a series of local censuses that have been carried out for several years in two large Nairobi slums, by the Africa Population Health and Research Center (APHRC). To complement the quantitative analysis, the Kenya National Bureau of Statistics (KNBS's) Fourth Participatory Poverty Assessment (PPA-IV), which gathered more qualitative information on how communities perceived their living standards, is exploited to provide deeper insights. Box 2-1 summarizes the main data sources used. 11 Box 2-1: Data Sets Used in the KPIA Poverty and inequality are complex and multidimensional country-specific problems. Developing a better understandingo f the scope, cause and possible solutions to these problems requires drawing on a multitudeo fdifferent informationsources. The following datasets are used inthis report: Representativeness Korogocho of householdswhose head had ever married. 2.5 This part begins with anational snapshot of consumption poverty and inequality in2005/6, using the KIHBS. These results are complemented by a focus on geographic variation and inequality. While almost half of Kenya's population lives in poverty, this population is heterogeneous facing diverse problems and living conditions. To build a multi-faceted poverty profile, we distill salient characteristics of poor households including demographic composition, gender dimensions, education levels, migration status, health and nutrition, asset ownership, access to key services and infrastructure. We then turn to focus on three major themes: provincial variation and the slums and youth. Finally but not least, we explore expectedimpacts of the recent conflict on the levels and profile ofpoverty and inequality inKenya. ANATIONAL SNAPSHOT:THE2005-2006 KIHBS 2.6 The national snapshot uses the KIHBS and adopts the official KNBS method, where household consumption is compared against the official poverty lines (Annex 1). The overall rural and urban poverty lines are, respectively, 1,562 and 2,913 Kenya Shillings (KSh) per month per person (in adult equivalent terms, which at the time was approximately US$0.75 and US$1.40 a day per person) and include minimumprovisions for both food and non-food expenditures. The lower food poverty lines-KSh 988 in rural and KSh 1,474 in urban areas-correspond to the average expenditure needed in 2005-2006 to attain the minimumrecommendeddaily intake o f 2,250 kilocalories. AlmostHalfthe PopulationLivesinPovertyandmostly inRuralAreas 2.7 The KIHBS suggests that in 2005/6 almost half of the population (47 percent) lived inpoverty, o f which 85 percent were in rural areas (Table 1.1). Poverty incidence 12 was significantly lower inurbanthan rural areas (one third versus one half). This is true regardless of the poverty line or measure. The rural expenditure distribution i s also much steeper around the poverty threshold, which suggests that a marginal increase inthe cost of meeting the basic needs basket o f goods would result in a relatively larger rural share falling into poverty. The sensitivity o f poverty measures to price and other changes is addressedindetail later inthis chapter, inthe context of crisis impacts and food inflation. 2.8 Food poverty was widespread in 2005/6, with over 16 million Kenyans unable to meet the cost o f buying the amount of calories sufficient to meet recommended daily calorie requirements. Indeed, almost one out o f every five Kenyans was in severe poverty, in that they could not meet the cost of a basic food bundle even if they spent their entire budget on food (Table 2-1). The poverty gap captures the average expenditure shortfall of the poor relative to the poverty line, which inthis case was about 17percent. 2.9 The official data suggest a decline inpoverty over time-from an estimated 51 percent in 1997, to 47 percent in 2005/6.'6 While higher poverty headcount figures for 2000 are sometimes reported, which would inturn suggest a more rapid decline, these are not reliable, and do not refer to these inthe presentreport. Poverty inKenya during2005- 2006 was estimated at 46.6 percent from the KIHBS data. The 1997 WMS is the most immediately preceding national household survey designed to estimate poverty levels. Alas, these poverty estimates are not strictly comparable over time because of modifications in instrument design as well as differences in the measurement approach used. Table 2-1: Poverty Estimates (2005-2006) Number of Poor PovertyMeasure Headcount(YO) (million) Poverty Gap (YO) National Overall 46.60 16.60 16.60 Food 45.80 16.30 Severe 19.50 6.90 Urban Overall 34.40 2.50 11.70 Food 40.40 2.90 13.00 Severe 8.30 0.60 2.50 Rural Overall 49.70 14.10 17.80 Food 47.20 13.40 16.20 Severe 22.30 6.30 6.90 Source: KNBS(2007) and staff estimates basedon the KIHBS. Triangulation 2.10 It is important to assess alternative sources of data andmethods, to see whether the overall officially reported trends are robust. We have reviewed against three such l6This is statistically significant: the 95 percent confidence intervals around the 1997 figure are 50.7 and 53.6, and aroundthe 2005 estimate, 45.4 and47.9 percent. 13 sources-Tegemeo panel, APHRC urbanslums, and household asset-basedpredictions- and conclude that the trends are broadly consistent. 2.11 First, the Tegemeo panel, which is representative of non-pastoral rural Kenya, collects data on incomes that allow estimation of poverty headcounts. This sample i s not designed to be nationally representative, and the results should not be taken as such. Over the period 1997-2007, poverty measured in terms of income fell from 51 to 36 percent. The Tegemeo sample also evidences household assets (livestock, bike, radio, other). Unlikethe WMS / KIHBS comparisons, Tegemeo has information for four points intime, viz 1997,2000,2004 and2007, from the same panelofhouseholds, and is ableto identify sub-periodsof growth and poverty reduction, which is investigatedinPart 111. 2.12 Second, APHRC has undertaken extensive demographic, social and economic analysis of trends in two slums inNairobi. Data on livelihood for two points 2003 and 2006 show a marked decline in poverty-albeit from very high rates, to levels that are still almost double above the national urban average, that i s from 73 to 62 percent over the three year period (de Laat et a12008). This is investigated further below. 2.13 Third, household assets-based predictions, following Stifel and Christensen (2007), suggest that nationally, the predicted poverty rate fell from 51 to 47 percent over the period 1997-2005/6.l7 This methoduses an economically intuitiveandinexpensive methodology which has been used in the absence of regular collection of household consumption data (Stifel and Christensen, 2007), and in cases where comparisons of household survey monetary based results could be valuable (Kijima, Y. and Lanjouw, P. (2003) Poverty in India during the 1990s: A Regional Perspective, Policy Research Working Paper No. 3141,DECRG, The World Bank and Christiaensen, L.,Lanjouw, P., Luto, J., and Stifel, D. (2008) Estimating Poverty Across Non-Comparable Surveys: Evidence@om Vietnam,Russia and Kenya, draft, forthcoming). Estimates using the asset data from the 1997 WMS, earlier DHSs and KIHBS suggestoverall improvements. 2.14 It thus seems that the direction of the officially reported trend of decline in poverty is robust, and, given the Tegemeo results, that this trend continued through 2007. However all the estimates are based on data that was generated prior to the recent election crisis and food price rises. We return to both of those major issues in the final section of this part. It is also important to underlinemore general caveats that qualify the trend of improvements, most notably the significant provincial variation and rising inequality. Muchof this assessment tries to better understandthese underlyingconcerns. l7Thanks are due to David Stifel, Assistant Professor, at Lafayette College, for re-estimating earlier results using KIHBS data, and to Luc Christiansenand Peter Lanjouw for support and advice on this point. 14 Box 2-2: Defining Poverty in Kenya Broadly speaking, measures o f poverty can be either objective (based on observed or reported income, expenditure or assets) or subjective. Much o f this report relies on the former, while also drawing extensively on qualitative information. The latter approach recognizes that Kenyans' individual, community and social perceptions about poverty fundamentally matter. There have been several studies in Kenya which attempt to combine quantitative techniques o f poverty analysis with more participatory approaches, the latter bringingthe perspective o f poor people themselves and their communities. Back in 1994 Narayan and Nyamwaya (1996) found that the community perceptions o f poverty led to estimates similar to those obtained from quantitative surveys. Since then a rich literature has unfolded offering deeper insights into the synergies and potential contradictions between these two empirical approaches (Krishna et al, 2004, Ade Freeman et al, 2005, Lay et al, 2007, Place et al, 2007, Mango et al, 2007, Republic o f Kenya, 2007, and World Bank, 2006, 2007). The qualitative work helps to highlight what are the most important indicators to be tracking, and that differences across places are key. The KNBS's Fourth Participatory Poverty Assessment (PPA-IV) gathered qualitative information from communities. By conducting this exercise on a sub-sample o f the KIHBS, the KNBS opened up potential for a mixed analysis o f the poverty situation, combining structured and participatory approaches, with a broad geographical coverage which sets it apart from other recent studies. The PPA-IV covered 71 clusters in 18districts inbothrural andurbanKenya. In general, the PPA-IV teams found a consensus among the community groups on what it means to be poor. For both rural and urban communities, extreme poverty is closely linked to food consumption. This confirms the intuition o f the approach taken indefining a nutrition-anchored poverty line inthe analysis o f the KIHBS data. Non-food consumption emphasized by the communities consisted o f mostly o f clothing and primary schooling costs (like uniforms) o f children. The majority described poor people as those who are never sure o f their next meal, who depend on others for survival, have dilapidated housing, are poorly dressed and cannot even afford second-hand clothes. Poor people do not own property (such as land) and generally work on other household's farms to earn some living. Most poor people were said to be always inbadhealthandunableto affordto educatetheir children. The PPA-IV found a surprising uniformity across locations inthese perceptions, though some dimensions were community-specific. In Kilifi, for example, the communities considered the lack o f legal title deed (for farm land) as a manifestation o f poverty. Pastoralists, such as Wajir, placed more emphasis on livestock ownership. In urban Nairobi and Mombasa communities identified poverty in terms o f low wages and unemployment. These findings o f the PPA-IV are strikingly similar to the conclusions o f that conducted a decade earlier (Narayan andNyamwaya, 1996). Inequality in Kenya is High, especially in RuralAreas 2.15 The fact that the national consumption decile ratio (richest to poorest) rose from 13 to 19 between 1997 and 2005/6 indicates large and growing inequality. Here we take a broader brush view of inequality to complement the foregoing picture of spatial concentration of poverty. We look initially at measures of vertical inequality, that is across individuals and households, as opposedto horizontal inequality across groups (see Part Iand further below, whenwe examine provincialdifferentiation). 2.16 Different measures can be usedto capture inequality, including Lorenz curves/ Gini coefficients and decile-type ratios (Ferreira and Ravallion, 2008). Figure 2-1 provides some international comparisons of the Gini co-efficient which illustrate that inequality inKenyais markedlyhigher comparedto countries with similar internationally comparable dollar a day poverty rates such as Ethiopia and Malawi. And as we show below, rates o f inequality are even higher for reported earnings, and such key assets as 15 land. However the gini misses the large differences at the top and bottom o f the distribution, which are better reflected in decile ratios. The levels o f measured income inequality in the Tegemeo analysis are far higher than in the KIHBS measures, which may derive inpart from the differences between income and consumption. Figure 2-1: LorenzeCurve in Kenya and Some Comparisons(2005-2006) Year Populationbelow Gini $1a day Kenya 05-6 NA 45.2 Ethiopia 99-00 23 30 Malawi 04-05 20.8 39 Rwanda 2000 60.3 46.8 Tanzani 00-01 57.8 34.6 '02002 Uganda NA. 45.7 c"mu,-W*Mm m m Source: Staff estimates based on the KIHBS 2005-2006 data and WDI 2007. 2.17 Comparing the average expenditure o f the top and bottom deciles further underscores the relative inequalities within rural and urban areas. Average expenditure o f the bottom rural decile i s only 8 percent o f the top rural decile; the ratio inurban areas i s much larger--close to 1:20! Table 2-2: Average Rural and Urban Expenditure by Decile, 2005-6 2005-2006 1997 Decile Rural Urban Rural Urban Poorest 466 1,110 444 1,009 2 813 1,888 609 1,525 3 1,038 2,404 755 1,809 4 1,244 2,955 914 2,123 5 1,458 3,578 1,088 2,455 6 1,719 4,288 1,283 2,869 7 2,039 5,009 1,531 3,502 8 2,473 6,058 1,862 4,469 9 3,147 8,202 2,402 6,422 Richest 5,741 22,823 4,589 13,756 Total 2,331 6,673 1,558 4,033 Decile Ratio (1/10) 0.08 0.05 0.10 0.07 Source: Staff estimates based on KIHBS 2005-2006data and W M S 1997data. Note: Deflated per adult equivalent consumption. 2.18 Figure 2-2 shows the dramatic difference in consumption relative to the poverty line across expenditure deciles. The richest households (top decile) spend about four times the poverty line. 16 Figure 2-2: Ratio of ConsumptionPer Capita to the Poverty Line by Decile 4.51 4 3.5 3 0 2 2.5 2 1.5 1 0.5 0 Poorest 2 3 4 5 6 7 0 Q Richest 2.19 I t is interesting to view inequality levels and trends in a broader international and regional perspective. Table 2-3 shows that both the gini and decile ratio in Kenya are not as high as levels observed in Latin America, where inequality is renowned for its entrenchednature, and damaging effects on social cohesion and growth (World Bank, World Development Report, 2006). Levels of inequality inKenya are, not surprisingly, much higher than in formerly socialist and neighboring Ethiopia and Tanzania, and also the levels reported in East Asia, and the U S and Australia, for example. It is also worth highlighting that national statistics show a widening of inequality between 1997 and 2005/6. The pace of this change may be an issue of concern for Kenyans and policymakers. 17 Table 2-3: Inequality Measures,Kenya and Selected Comparators Year Gini Decile Ratio Latin American and Caribbean Brazil 2004 57 49.8 Honduras 2003 53.8 35.2 Bolivia 2002 60.1 157.3 Mexico 2004 46.1 24.6 Sub Saharan Africa SouthAfrica 2000 57.8 31.9 Ghana 1998-1999 40.8 14.3 Tanzania 2000-2001 34.6 9.3 Ethiopia 1999-2000 30 6.5 EastAsia and Pacific Vietnam 2004 34.4 6.9 Indonesia 2002 34.3 7.9 Thailand 2002 42 12.4 OECD Australia 1994 35.2 12.7 United States 2000 40.8 15.7 UnitedKingdom 1999 36 13.6 Kenya 1997 42.5 13.6 Kenya 2005-2006 45.2 14.0 Variation inPovertyincidenceamongProvincesandDistricts 2.20 Kenya's development has been historically characterized by significant spatial variation. Not surprisingly then, in2005/6, poverty incidence inNairobi and Central was way below the national average, and only 12percent o f the total poor reside inthese two provinces. Conversely, poverty incidence is higher in Coast and Western provinces, which together account for a quarter of the poor in Kenya. Poverty incidence i s most pronounced in the sparsely populated North Eastern province. Figure 2-3 provides a provincialbreakdown of the population andthe numbers o fpoor, basedon the KIHBS. 1 Figure 2-3: Populationand Poverty by Province, 2005-2006 80 7 0 ] N.Eestern4.9 -Le 0 6 0 - Hcstrin 13.9 Fastern 17.7 5 0 - v 8 40- (:entral 8.1 Nyanza 14.2 -a 3 0 - 1 RiftValley 25.6 eICI 20- 0 Nairobi 3.7 I O - 0 4 0 5 I O I 5 20 25 3( Distributionof population(percent) Source: Staff estimates basedon KIHBS. Size of bubble representsprovince's share oftotal poverty 18 2.2 1 Spatial disparities are pronounced at the district level-poverty incidence varies from 93 percent in Turkana to 12 percent in Kajiado district. Indeed, a quarter of the poor population are concentrated in just seven districts,18 half o f the poor live in just under a third of all districts, and three quarters of the poor nationwide reside in half of Kenya's districts. Variations inthe depth o f poverty are also significant: the poverty gap inTurkana is 70 percent comparedto 2 percent inKajiado (KNBS,2007b). We examine spatial variation further below. 2.22 Returning to the qualitative data, the PPA-IV defined five livelihood groups, described in Box 2-3 (Republic of Kenya, 2007). Perceived poverty incidence and its dynamics differed systematically across these groups. In both pastoralist and "mixed marginal fanning" with livestock districts, about 60 percent were considered poor. Poverty was perceived to be lowest among the mixed farming communities (Groups 1 and 2), where about two fifths ofhouseholdsare considered poor. The PPA-IV underlines high poverty incidence in districts characterized by certain kinds of livelihood. Pastoralistsinparticular have beenfaring badly. Box 2-3: LivelihoodGroups Distinguishedin the PPA-IV Livelihood group I:mixedfarming, highpotential. The mixedfarminghighpotentialdistricts were mainly characterizedby food and cash crop (tea, coffee andsugarcane) farmingas well as livestock rearing. They also had mixed farming in the marginal areas where food crops and livestock rearing dominated. These districts includedNyeri, Kirinyaga, KisiiCentral, NandiandButeremumias. Livelihood group 2: mixed marginal Q'ow potential) farming withfishing. The predominantactivities were mixedfarming offoodcrops, some cashcrops and livestock. They also hadfishingandsome cashcropping activities inthe low potentialareas. This group ofdistricts includedKisumu, MigoriandBusia. Livelihood group 3: mixed marginalfarming with livestock. Largeportions ofthese districts were marginal areas characterizedby mixedfarming of food crops, cashcrops (palm, coffee andpyrethrum) and livestock farming. Some of them were also occasionedby casual wages derived fiom large ranches. These districts includedMakueni, Kilifi, Tharaka, Marakwetand Laikipia. Livelihood group 4: pastoral. These districts were largely characterizedby pastoralismwith some pockets of agro-pastoralismand mixed farming (food crops) in some areas. They included Tana River, Marsabit andWajir districts. Livelihood group 5: Urban districts. NairobiandMombasa. VariationinPovertyIncidencewithin Districts 2.23 A poverty map can combine population census information with household survey data to estimate poverty and inequality at finer levels of geographic detail than what is feasible with the survey data alone.l9 For Kenya, the 1999population census and the 1997 WMS were used to estimate poverty incidence at the location level, which averaged 12,000 inhabitants. These suggest considerable variation in poverty within l8In 2005/6, the districts with the largest number of poor people were: Machakos, Makueni, Nairobi, Bungoma, Nakuru, Kwale and Turkana. The seven districts with the highest poverty rates are Turkana, Marsabit, Mandera, Wajir, Samburu, TanaRiverandKwale. 19Smallareapoverty mapswere preparedby KNBSbasedon the 1999PopulationandHousingCensusand the 1997WMS. 19 districts-as well as concentration of poverty-in 1997, two thirds of the rural poor were concentrated in one third of Kenya's 2,070 locations. Poverty estimates in over 60 percent of urban locations were significantly different from their corresponding district level averages. This indicates substantial within district variation Le. poor people in - better off districts, and vice versa. The slum areas in Nairobi, for instance, are home to the bulk of the poor in the country's capital. Analysis below based on complementary date sources (APHRC) suggests that poverty levels are high in the slums but also that there is very significant village level variationwithinthe slums (see below). 2.24 Map 2-1 presents the 2006 District level estimates obtained from the KIHBS which illustrates the considerable geographic variation inpoverty incidence at that time. Itis not possibleto compute 2006 locationlevelpoverty estimatesbecausethe KIHBS and 1999 population census data are too far apart intime. 20 Map 2-1: District Estimatesof Poverty, 2005/6 2oThe KNBSreport "Geographic Dimensions ofWell-Being inKenya: Where are the Poor? FromDistricts to Locations, Volume 1, (2003) shows these patterns as well. 20 NON-MONETARY ASPECTS WELL-BEING OF 2.25 Critical dimensions o f individual and family well-being are not captured in monetary measures. Inthis section we examine access and outcome indicators relating to education, water and health services o f Kenyans, drawing largely on Demographic and Health Surveys and the KIHBS. Kenya's aggregate levels o f human development are highrelative to countries at similar levels o f income. However, the progress has neither been linear over time, nor equitably distributed. In particular, for some key indicators including access to primary education, there was major regress in the 1990s. Deterioration insome important indicators, like child nutrition, was most severe for worst o f f groups. Mirroring the inequality in the distribution o f monetary measures, there are also significant disparities inthe distribution o f access and outcomes. National averages typically mask key provincial andor wealth differences for social services, water and sanitation, and utilities such as electricity andtelephones. 2.26 We begin by giving a sense of housing conditions. KIHBS includes data on materials used in each part o f the housing structure (roof, walls and floor). Conditions are very poor for much o f the population. There i s also significant variance in quality across expenditure deciles in rural areas (Figure 2-4), which is starker in urban areas. Poor households and rural households are much more likely to use firewood, and richer (and urban) households tend to use charcoal. Most households use paraffin as a fuel for light. Electricity is far more prominent in urban areas, especially amongst the upper deciles. Similar findings are obtained by Tegemeo: the rural poor have worse housing (roofs, walls, floors), and poorer households have worse sources o f water during both the wet and dry seasons and worse sources o f lighting. Figure 2-4: Material of the Walls of the Main Dwelling inRural Areas by Decile 100% 90% 80% Other 70% Tin GrasslStraw 60% Corrugated Iron Sheets 50% Wood only 40% MudlCement 30% MudWood BricklBlock 20% Stone 10% 0% 2 3 4 5 6 7 0 9 Poorest Richest wrce: Staff calculations based on KIHBS, 200516 21 2.27 Trends in key aspects of non-monetary wellbeing can be summarized using selectedMDGs. These show recent improvements ineducation, a good record on gender equity inschooling, butworsening child mortality rates. During the 1990s through 2003, primary enrolment fluctuated and secondary school enrolment remained low. The late 1990s was a period of regress for education, when the large educational gains for younger ages earlier in the decade had been eroded by half by 2003. In 2003, less than 40 percent of children aged 15 to 19 years had attained at least the eight grades requiredto complete their primary education. As investigated below, there have been large gains since 2003 with the adoption of free primary education (FPE). At the same time, the latest available (2005/6) figures suggest that retention of primary students remains a barrier to meetingthe education MDG. Gender equality inprimary school has recovered to high levels, while youth literacy gender parity is almost at the MDGtarget. The proportion of one-year olds vaccinated against measles has fallen substantially since 1993. On current trends, the likelihood of meeting this MDGby 2015 seems low. Other antigen coveragerates (Polio, DPTl,DPT2, DPT3, and BCG) all show similar declines. Infant and child mortality have concomitantly risen since 1993; so that the target of reducing the 1990-levels by two-thirds i s receding further. Maternal mortality improved across the two available datapoints, though at a rate which impliesthat the likelihoodofreaching the MDGtarget i s remote. 2.28 Available data shows that there has not been a substantial change in the prevalence of underweight, stunting or wasting during the last 15 years. A Composite Index of Anthropometric Failure (CIAF) calculated using the 2003 DHS indicated that 35.5 percent of children were either stunted, underweight or wasted and the CIAF ranged from 22 percent inNairobi to 46 percent inNorth Eastern Province.21 About 55 percent of pregnant women, 48 percent of women of reproductive age (15- 49 years) and 73 percent of children under 5 years are anaemic, and 76 percent of children under five years suffer from Vitamin A deficiency.22 Research indicates that malnutrition i s an underlying factor that contributes to about half of all child and infant deaths inKenya. A substantial number of deaths could be prevented, particularly in Western Kenya and Nyanza, by improving children's nutritionalstatus (UNICEF 2008). 2.29 The decade to 2005/6 involved stagnation and decline in the health status of Kenyans. This is inturn can be traced back to the lack of economic growth described in Part I,and extensive poverty. High population growth rates, poor access to basic health services and high prevalence of HIV, also contributed to poor health status in Kenya. Available data suggest major deterioration between 1990 and 2003: life expectancy dropped by about five years, the infant mortality rate increased from 60 to 77 per 1000 21Nutrition situation analysis, UNICEF, June 2008 22UnitedNations Demographic Projections(2006), 1999 Census, Iodine Status Survey2006, KIHBS05/06, DHS,2003; Anemia, Vitamin A, andZinc status survey, 1999. 22 live births, and the child mortality rate from 100 to 115. There are at least 1.2 million Kenyans now livingwith HIV/AIDS, andthe TB incidence rate i s estimated to be 620 per 100,000 - one o f the highest rates in the world. Kenya was clearly not on track to meeting the health MDGs. Table 2-4: Trends in SelectedMDGIndicators,Relative to 2015 and Medium Term Plan Targets 2015 1993 1998 2003 2005 MDG MDG2: Education Net primary enrolmentrate 75 85 79 81 100 Primarycompletionrate, % 19 20 19 35 100 Youth literacyrate(%ages 15-24) 92 93 87 92 100 MDG3: Gender Equality Ratiogirls to boys inPrimaryEd(%) 99 95 91 98 100 Ratio literate females to males(% ages 15-24) 92 94 95 99 100 MDG4: Child Mortality Under-5mortality rate(per 1,000) 96 112 115 32 Infantmortalityrate(per 1,000 livebirths) 62 74 77 21 One-yr-oldsimmunizedagainstmeasles(%) 84 79 73 *** 100 M D G 5: Maternal Mortality Maternalmortality ratio (/100,000 live births) * 590 414 91t Births attendedby skilledhealthstaff(%) 54 53 69 ** 90tt Source: 1993, 1998,2003 DHS & 2005/6 KIHBS. * Denotes Data not available. HMISAnnual Report 2003 -2004. The target that 90 percent of births should be attendedby skilled staff by 2015 was adoptedat a special sessionofthe UnitedNationsGeneralAssembly in 1999. 2.30 One positive aspect o f Kenya's record i s the general achievement o f gender parity in access to many basic services. In the case of education, for example, consistently over time, there are no substantive gender differences in school enrollment between ages 6 and 14 (Figure 2-5). According to Table 2-4, this dipped and came back-but interestingly, the literacy rate has been increasing over time, suggesting that girls are now getting more out o f school, which could reflect better attendance. Figure 2-5: Proportion of Enrolled Students by Gender, Ages 6-14: 200516 200516 0.5 0 4 - 6 7 8 9 TI 11 0 Q 14 -Male Female Source: KIHBS200516 23 Inequalities in Access to Services and Outcomes 2.31 There are very large differences among Kenyans in terms o f outcomes and in access to basic services. For example, the risk o f infant death inNyanza is over 6 times greater than an infantin Central Province. The relative risks o f dying o f a child aged 1-5 years in Nyanza, Western and North Eastern Provinces are 7-8 times greater than for child o f the same age in Central. Here we look at two major dimensions o f disparity in access to service -by province, andby quintile. 2.32 Figure 2-6 shows how service gains and losses during 1998-2003 were distributed across provinces. Overall improvements were skewed towards Nairobi and Central provinces. Gains were concentrated in Nairobi for electricity, sanitation and telephone ownership. There were similar gains for Central province across the same three services, plus water, while Eastern andNyanza provinces had gains inaccess to safe water. Deteriorations are most marked for Western and Rift Valley for water. The highlightsfor humandevelopment related indicators are as follows: Nationally, access to safe water stagnated in the decade to 2003. Western province has experienced a halving in access. While all provinces have seen improvements in household sanitation, the national gains have been mostly driven by improvements in the capital and Coast province. For education, province-level differences are not as stark as the water or sanitation numbers but Coast and North Eastern provinces still lag the national average. 0 While secondary enrollments have shown solid growth across the decade, the provinces do display wide differences, especially because Nairobi has a net secondary enrollment which i s more than double the national average. 2.33 For other aspects o f well-being and connectedness, we see that: Electrification across the country increased by 50 per cent between 1993 and 2003, though levels remain very low especially in rural areas. Western and Nyanza Provinces have experienced drops in coverage rates across the decade, while the gaps between the capital and the rest o f the provinces have remained stark. 2005/6 data show marginal gains in electrification, which were positive inall provinces. 0 Mobile telephony has touched all provinces, with the largest gains inNyanza, Central and Eastern Province. The KIHBS did not collect data on telephone ownership, though recent trends suggest that telephone ownership, across individualsnot households, may be approachingone-third. 24 Figure 2-6: Provincial Distribution of Service Gains, 1998-2003 100% 80% 60% 40% 20% OYO -20% -40% I Electricity Water Sanitation Telephone Gross Primary Enrollment Source 1998 and 2003 DHS. NorthEasternProvinceomitted, since datanot collected in 1998. Safe water sources include `Piped into dwelling', `Piped into compound', `Public tap', `Covered Wells' and `Bottled water'. Improved sanitation facilities include `Flush toilets' and `Ventilated improvedpit latrines'. Telephone ownership is at the householdleveland includesmobiles in2003. 2.34 Figure2-7 below shows how the service gains were distributedacross the more recent 2003-2005/6 period, compared the most recent DHs to the KIHBS. Rift Valley, Nyanza and Western all saw large increases in household electrification, relative to the low levels reported in 2003. The doubling of access in Western Province to 4 percent represented a large contribution to the overall gain. Water and sanitation numbers all increasedacross the 2003-2005/6 period. 2.35 Gross primary education has increased in all provinces. However, given the size of Rift Valley province, the gains therein are most substantial in absolute terms. Comparison of the surveys suggests that North Eastern province has seen a 50 percent increaseingross primary enrollment. 25 Figure 2-7: Provincial Distribution of Service Gains, 2003-2005/6 -40% 1.......................... ~.............................. ~ ......~ . . ~ . . .................~ ..... ~........ ~ . ~~~~~ Electricity Water Sanitation Gross Primary Enrollment 2.36 Following common practice (Filmer and Pritchett, 1998), household wealth indices provide further insights into trends in disparities inhuman development. Higher quintiles o f course have better access to basic utilities by construction, and comparisons o f the 2005/6 consumption quintiles against the DHS asset-based quintiles suggests that the asset based distribution has some bias when we look at utilities. It i s nonetheless notable that the gains in improved sanitation facilities, electricity access and telephone ownership were restricted entirely to the wealthier quintiles. Here we look at the overall patterns, andthen at specific services. 2.37 A striking overall finding for the decade to 2003 is that access o f the poorest quintile to education worsened. Rates o f access for the bottom quintile, which were already way below average, and far behindthe top quintile, worsened further. By way o f contrast, the period after 2003 was a welcome reversal-in that there were large gains in access to improved water and schooling. Table 2-5: Access to Selected Services by Quintile between 1993 and 200516 Electricity II 15 I 6 1 6 0 0 0 0 0 0 0 ? 0 0 0 8 1 2 4 2 0 4 7 6 0 5 7 5 3 Net Primary Enrollment 75 85 79 81 71 76 61 73 74 85 80 80 72 88 84 84 76 90 88 86 81 90 86 89 GrossPrimary Enrollment 102 112 111 119 100 101 91 118 IO3 113 116 125 100 119 120 119 103 119 12s 118 106 110 106 109 Net Secondary Enrollment 7 10 13 18 4 4 4 6 6 5 7 f ? 5 9 12 21 6 i 2 16 30 17 22 29 52 GrossSecondarv Enrollment 19 23 23 40 14 9 9 16 13 I 8 19 34 15 21 22 50 I$ 23 29 59 37 44 41 84 Source: DHS, 1993, 1998,2003, and KIHBS 2005/6. * Denotes missingdata. DHS analysis uses asset indices, KIHBS uses a consumption per-capita index. 26 Disparitiesin Education:Rural-Urban, Provincialand by Quintile 2.38 Rural-urban disparities are shown inthe number o f years of education attained by youth, inthis case teenagers aged 15-19 years (Figure 2-8). These emerged after the late 1990s, andpersistedthrough 2005/6. Figure 2-8: Grade Attainment Profile, Rural-Urban, Ages 15-19: Kenya 1993-2005/6 - 1993 200516 1 1 0.75 0.75 0.5 0.5 0.25 0.25 0.25 0 0 0 1 2 3 4 5 8 7 8 9 1 2 3 4 5 8 7 8 9 1 2 3 4 5 8 7 8 9 1 2 3 4 5 6 7 8 < 1 -Urban -Rural 1 Source: DHS 1993-2003, KIHBS 2005/6. 2.39 Education gains over the decade to 2003 bypassedthe bottomquintile. In2003, their net primary enrollment rate lagged the average by a massive 18 percentage points, comparedto 6 points inMozambique and Uganda 10points. The alarmingtrend between 1998 and 2003, when primary enrollment rates fell, was most severe for the lowest quintile. By 2005/6 the difference had narrowed to 7 percent, which suggests that FPE was enabling larger gains for poorer households (see Part IV). 2.40 Despite fluctuations inenrollment since 1993, overall educational attainment for 15-19 year olds changed little, either way. However the average hides important changes across groups. Figure 2-9 divides the population into three groups - bottom 40 percent, middle 40, and top 20 percent. Over time, we see large improvements inthe attainment across the board 1993 and 1998, but after 1998 the position o f the poorest 40 percent was eroded and the richest quintile widened their lead. Hence, there are stark differences between grade eight completion rates of the poorest versus the wealthiest, 20 versus 60 percent. Urban versus rural differences are also substantial. But, as with enrolment, gender biases are mutedwith girls showing marginally higher attainment rates. 27 Figure 2-9: Grade Attainment Profile by Quintile, Ages 15-19; 1993-2005/6 1993 I 1998 2003 2005/6 I 1 1 1 1 0.75 I75 0 75 0.75 0.5 0 5 0 5 0.5 0.25 I 2 5 0 25 1.25 0 0- 0 1 2 3 4 5 6 7 8 91 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 6 9 1 2 3 4 5 6 7 8 9 1 I-Poorest 40% Middle 40% -Richest 20% I Source: DHS 1993-2003, KIHBS2005/6. 2.41 The differences in attainment by quintile can likely be traced, at least inpart, to constraints at the household level. Careful evaluations have found, for example, that the costs of school uniforms are a major barrierto poor families enrolling their children, even in the absence of tuition fees (Evans, Kremer etc). Private spending on education averages about 6 percent of meanmonthly consumption inthe KIHBS. We return to the burdenofprivate spendinginPart IV below, on policy implications. 2.42 Finally, these patterns of education access are put in cross-country context (Figure 2-10). Inequality in Kenya appears to be relatively high (second to Tanzania) in 2003, though there has beensome improvement inthe more recent period. Figure 2-10: Distribution of Net Enrollment Ratesby Quintile -Selected Countries -KE-TZ -MZ -UG-BD -CO 1.25 1 1 0.75 ~ Poorest Poorer Middle Richer Richest Source:DHS 2002-2005 28 Access to ImprovedWater 2.43 About halfthe populationinKenya does not have improved access to water and sanitation services meeting minimal standards of quality. In this section, we analyze patterns in access to water, which is much lower in rural areas, and significantly lower among the poor. There is significant regional variation. 2.44 Overall, access to improved water sources in Ken a is 56.5 percent, with 82 percent coverage inurban areas and 50 percent inrural areas!3 Coverage i s characterized by relatively high service levels. Of those with access to improved water sources-28 percent has access to piped water-public and private-at the national level, 17 percent of the rural population has access to pipedwater and 73 percent of the urbanhas access to this service level. 2.45 The provision of water is organized around Water Service Boards (WSBs). There are large geographical variations at this level, especially inthe quality of services. Rural coverage varies from 63 percent in Lake Victoria North to 40 percent inNorthern. Inurbanareas, significantly higher service levels are providedinAthi andTanawhere 70 percent o f households have access to private piped water, compared to about 25 percent inLake Victoria SouthandLakeVictoriaNorth. Table 2-6: Water Service Levels by WSB -Rural Areas Lake Lake Rift Vic. Vic. Athi Tana Coast Northern Valley South North All Private piped 15.0 21.7 4.2 7.0 11.0 7.1 2.8 10.2 Public piped 9.1 8.7 37.8 5.1 3.05 4.3 3.3 7.8 Other improved 35.1 16.7 20.5 28.2 30.4 35.6 56.7 34.2 Not improved 40.8 53.0 37.5 59.8 55.6 53.0 37.3 47.8 All 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Note: Populationweights. 2.46 Inbothrural and urbanareas those inthe poorest quintileshave lower access to private piped water. It is notable that people in the poorest quintiles in urban areas have almost as good access of as those inthe richest quintiles inrural areas. 23Thiscoverage figure does not consider minimum quantity which according JMP should be at least 20 liters per personper day. 29 Figure 2-11: Water Service Access, Rural and Urban, Top and Bottom Quintiles RuralAreas -Poorest Quintile RuralAreas -RichestQuintile Private piped Not Private improv piped 36% 25% Public piped 6% Other improved 33% Urban Areas -Poorest Quintile Urban Areas -Richest Quintile Not Not improved Private Other improwd 24% .22% . Publicimproed/ 5% 7% \ Public piped Primte 37% piped 80% 2.47 To examine whether women are systematically discriminated against on this front, access to different service levels o f water was examined by gender o f the household head and by the majority gender in the household. Although female headed household appear to have somewhat lower access to private piped water in rural and urban areas, the effects are not very strong and are probably caused by lower household wealth. 2.48 Kenyanhouseholds spend significant amounts o f time collecting water. Water collection times per trip (back and forth) are substantial with average collection time around 50 minutes inrural areas and 20 minutes inurban. 30 Table 2-7: Water CollectionTime by Water Service Board -Rural Walking time Queuetime Total time WSB Mean Median Mean Median Mean Median Athi 30.6 20.0 34.8 10.0 65.5 40.0 Tana 39.2 20.0 12.9 2.0 52.1 30.0 Coast 30.4 20.0 15.2 10.0 45.6 30.0 Northern 43.9 20.0 11.3 4.0 55.2 30.0 RiftValley 52.4 30.0 21.3 0.0 73.6 40.0 LakeVic. Sth 28.9 20.0 10.0 1.o 38.9 28.0 L.Vic. North 25.8 20.0 9.7 '1.0 35.5 22.0 All 34.3 20.0 15.4 2.0 49.7 30.0 Note: Householdweights. There are a few very highoutlierson the collectiontime variables. Therefore mediansare includedalongwith means. Source: Staff analysis basedon KIHBS. 2.49 Among households who pay for water, the poor spend a larger share of their budget on water than the better off. The unit cost of water is higher for households using low quality services than for those using services of higher quality. 2.50 Regression analysis suggests that most o f the variation in access to water services is explained by community fixed effects. On the other hand, intra-community variation in economic status only explains a small part o f the intra-community variation inaccess. This suggeststhat targeting efforts could focus ontargeting poor communities, rather than on targeting the poor within a given community: once a service reaches a community, it i s likely to reach most inhabitants, regardless of their economic status. Similar conclusions seem to hold for sanitation services, although the results are not as strong as for water. Health 2.51 We have already outlined concerns around the trends in health indicators; moreover, health disparities remain high. As noted above, in the context of the MDG record, national vaccination coverage rates of one-year-olds fell markedly inKenya inthe decade to 2003.24 If disaggregated by quintile and gender, the deterioration is actually worst for the poorest groups, both relative to their starting point and relative to richer quintiles. Interms of gender, the declines appear to be slightly worse for boys. Malaria continues to be the leading cause of morbidity and mortality in Kenya. Other major causes include acute respiratory infection, malnutrition, diarrhea disease, HIV/AIDS,and tuberculosis (TB). 2.52 Maternal mortality also remains very high at 414 per 100,000 live births. Although the relatively longhistory of population programs inKenyahas ledto a number of successes, the total fertility rate is still high at 5.5 lifetime births per woman. This report does not investigate the reasons behind levels and patterns of outcomes, though it 24Part of this decline can be traced to the fact that NE province, which has low coverage, was excluded fiom the baseline. 31 i s noted that cultural norms in some areas, including North-Eastern and Coast, affect,the demand for medical care. This inturn points to the need for, for example, more female staff inhealth facilities. 2.53 Health outcomes vary significantly by gender, quintile and location. For example, in 2003: 0 HIV prevalence among women 15 to 49 years old is almost twice that of men inthe same ages, andthe HNinfection rate of girls 15 to 19 years old is six times higher than that o f boys inthe same age; infant and under-five mortality rates among the lowest socio-economic quintilewas 50 percent higher than inthe richest quintile; and the incidence of moderate and severe malnutrition was almost four times greater inthe bottom versus the top quintile. 2.54 Table 2-8 shows trends in access by quintile and province for vaccinations, using the series of DHSs. Between 1993 and 1998 the coverage rates for one-year-olds decreased for all antigens; with Polio3 experiencing a 12 percent drop. By 2003 the coverage rates dropped further, averaging 7 percent less across all antigens. These national drops were most severally felt for quintiles one and two; on average coverage rates reduced by a massive one-third across DPT3, Polio3 and Measles for the bottom quintile. MoHadministrative datareport more recent improvements, noted below. Table 2-8: Vaccination Coverage Ratesof One-year-oldsby Wealth Quintile and Gender, 1993-2003 Wealth Quintiles Antigen Year lS' 2"d 3rd 4'h 5th F M Total 1993 94.1 94.9 95.6 97.8 99.1 95.9 96.6 96.2 1998 93.5 92.7 98.3 97.4 99.0 96.7 95.0 95.9 BCG 2003 70.0 88.7 93.2 95.7 96.2 88.2 86.6 87.4 1993 93.O 95.1 95.8 97.0 98.2 95.4 96.0 95.7 1998 92.6 94.7 96.6 97.2 98.1 95.6 95.6 95.6 DPTl 2003 73.6 93.1 93.5 95.8 95.2 89.8 ' 88.5 89.1 1993 82.4 91.1 82.5 86.8 93.0 87.1 87.0 87.0 1998 68.1 78.2 86.3 84.2 84.2 78.1 80.1 79.5 DPT3 2003 57.4 71.1 86.3 82.1 74.1 74.1 71.9 73.0 1993 81.2 88.3 81.7 86.5 93.0 85.7 86.4 86.0 1998 66.2 74.5 82.6 76.9 74.0 73.0 76.0 74.4 Polio3 2003 55.7 69.2 75.7 73.3 67.1 68.5 66.2 67.4 1993 80.6 82.4 82.4 86.1 88.2 84.7 83.1 83.9 1998 64.3 79.6 84.7 83.8 88.7 96.7 95.0 79.2 Measles 2003 54.8 68.1 79.3 80.3 88.0 72.1 73.2 72.7 32 2.55 As noted above, HIV prevalence is a major issue, and is higher among women across all provinces, and for all age groups up to 45 years. Interestingly, reported prevalence rates are higher for the better off, Table 2-9: HIV Prevalence Rates across Regions and Wealth Women Men Total Nairobi 11.9 8 9.9 Central 7.6 2.1 4.8 Coast 6.6 4.5 5.6 Eastern 6.1 1.6 3.9 Nyanza 18.3 12.3 15.4 RiftValley 6.9 3.4 5.2 Western 5.8 3.6 4.8 NorthEastern 0 0 0 I I Poorest 3.9 3.9 3.9 Second 8.5 4 6.3 Third 7.1 2.5 4.9 Fourth 9.7 4.1 6.9 Richest 12.2 7.3 9.7 Total 8.7 4.6 6.7 2.56 According to official data reported in the recent health sector program,25there are signs of improvements in some important health indicators inthe last few years. For example, HIV prevalence among adults had reportedly fallen, although the data about to be released on this front would need to be reviewed in order to be confident about ongoing trends. Based on MoH data, there were several positive trends in access indicators between 2004/5 and 2005/6 including rising share of women o f reproductive age receiving family planning commodities from 32 to 43 percent; rising share of infants fully immunized, from 58 to 80 percent and increased BCG vaccination rate for newborns, from 84 percentto 96 percent. 2.57 There are also reportedly signs of a decrease in malaria morbidity due to increasedcoverage and usage of insecticide treated nets (ITNs) by pregnant mothers and children, and more effective diagnosis and treatment. Over the past five years, 13.5 million I T N s have been distributed, with the percentage of children sleeping under them rising rapidly to 52 percent in 2006, from 5 percent in2003. The proportionof pregnant mothers using ITNs has risen from 4 percent to 44 percent. It has'been estimated that seven children have been saved for every 1,000 mosquito nets used, and scaling up o f malaria control interventions has reducedchild deaths due to malaria by 44 percent inthe areas o f highest malaria risk. ''HealthSector SupportProject, WorldBank ReportNo.: 40610-KE, November 15,2007. 33 PROFILE OF MONETARY POVERTY 2.58 We now turn to look at the profile of poverty in greater depth, exploiting the insights from both the cross-tabulations and regression analysis. Many of the findings echo those which have been found in the past, though there are some surprises, as we show below. The cross-tabulations provide important insights, but potentially miss complex associations between multiple factors by only considering the correlation betweenpoverty and any household characteristic one at atime. To isolate the effect of a specific variable on poverty-holding all other influences constant-multivariate regressiontechniques can help. 2.59 In this section we use various sources to illustrate and explain the observed patterns of monetary poverty-in particular the KIHBS, Tegemeo and APHRC. The KIHBSdata, for example, is usedto seek to explain variations inhousehold consumption (or more specifically, the logarithm of household consumption per adult equivalent).26 Our selection of the explanatory variables is based on the poverty profile, economic theory and factors that have tended to be associated with consumption and poverty in Kenya and other countries. Previous work in Kenya using multivariate methods (Geda et. al, 2001 and Okwi et. al, 2006) has found that poverty i s closely related to three main groups o f factors: household characteristics (especially household size, female headship and educational attainment); the employment profile of the household; and the characteristics of the location of the household (soil quality, length of growing period, distance to tarmac road). 2.60 Cross-sectionalanalysis cannot determine causality as this would requirethat all the explanatory variables are truly exogenous. However, the findings do reveal major correlates, and are indicative for the discussion inthe next part where we seek to identify the drivers of poverty. In the national analysis using KIHBS, we allow for structural differences between urban and rural areas by running separate regressions for these areas. The overall regressionfit is higher inurbanthan inrural areas, indicating that the KIHBS rural regressions still lack important variables, probably relating to land quality, rainfall and livestock holdings.27 2.61 The variables which emerge as statistically significant inour regression analysis using the KIHBS are illustrated graphically in Figure 2-12. The model uses observed variables to explain household consumption, relative to reference categories. The reference categories, which mainly relate to the head, are: Christian, no religion or not specified, married or cohabiting with spouse present, raised in current place of living, headi s not working, Februaryto April (season of survey) and Western (province). 26The regression is a semilogarithmic specification, the estimated coefficient o f a continuous variable, can be interpreted as the percentage effect o f a unit-change in the exogenous variable on consumption. However, Halvorsen and Palmquist (2001) show that the coefficients o f dummies needto be transformed in order to obtain exact percentage effects. We present exact effects here, and the regression table in the annex shows the underlyingcoefficients and standard errors. 27Inthe text, significant refers to 5 percent and marginally significant to 10percent. 34 Figure 2-12: Determinantsof HouseholdConsumption,Ruraland Urban 2005/6 -100% -75% -50% -25% 0% 25% 50% 75% 100% I Novemberto Jaquary 35 2.62 To elaborate upon and update the national picture emerging from the KIHBS, a series of regression were estimated in a step-wise fashion to measure the most significant correlates of poverty status (Nyoro et a1 2008). The results, which are representative of rural non-pastoral Kenya in 2007, broadly confirm the KIHBS findings, and add some important nuances related to rainfall and infrastructure which are not available in the KIHBS. Themost important observedfactors are as follows (Figure 2-13): Age-sex composition and education of household members. Access to credit, landacreage (wealth) and main season rainfall. 0 Group membership. 0 Salaried andbusinesshnformal activities. Province dummies, even after controlling for rainfall. 0 When access to market type variables are added, distance to fertilizer and to motorable road are most important. 0 Finally, diversification measures (not available for 1997 data) are included. These show that the number of crops per field matters as does the fraction of income derived from salaries and remittances. Figure 2-13: Tegemeo OLS Regression on Poverty Dummyfor whether anyone m Fractron o f adults m the HH Fraction ofadults HHmTolred man mformnVbusmess actiriiy 1 Distance to closest feniluerseller : Mamseason ramfall Acres cropped m the m m season& D u m for whether aHHin Fraction of adults HHw t h coqleted fonn four education Fraction ofadults m the HHuithsome secondw education Fraction o f adults m the HHm hcompleted pnrwry educanon Fraction of adults m the HH~nth some pnrrary education Fraction o f Fraction o Fraction ofHH that IS femle, aged 17-39 Fraction ofHH that 1s d e , aged 17-39 Fraction ofHH that IS f e d e , aged 6-16 Fraction ofHHthat IS d e , aged 6-16 Fraction o f H H that is fenule, aged 0-5 Fraction o fHHthat IS d e aged 0-5 -30 -20 -10 0 10 20 30 40 50 60 70 Source: Nyoro et a1(2008) 2.63 We turn now to explore these different aspects in turn, utilizing both the multivariate results as well as cross-tabulations, and reflect on their overall significance inthe population as a whole. These lead to a focus on horizontal inequality, interms of provincialvariation. 36 Demographics:Age and SizeMatters 2.64 There are several striking demographic features of poverty in Kenya, which echo those which have been found earlier in Kenya, and elsewhere. Specifically, as laid out inTable 2-10, and explored further below: While only 18 percent of Kenyans live in households that include three or more young children (under six years), almost two thirds of these live in poverty. Households are often large, and size matters-almost half the population lives inhouseholdswith at least seven members, and about 60 percent of these are poor. Poverty incidence monotonically declines with household size-only about 14 percent of the poor live in households with fewer than 5 members (Figure 2-14). The finding from Tegemeo panel analysis over the decade to 2007, that household size was decreasing over time for the non-poor and increasing for the poor, raises concerns. 0 Females seem to be running much bigger households-with correspondingly higher poverty rates, as one would expect. Poverty incidence increases with the age of the household head. Compared to the national average, households headed by someone aged over 44 are significantly more likely to be poor; 57 percent of the poor live in such households. However the negative association between the head's age and consumption does not hold inmultivariate analysis o f KIHBS. 0 The presenceof youth raisesthe probabilityofthe household beingpoor. Table 2-10: Poverty and Household Size, and Number of Children Poverty Headcount Distribution ofthe Population share (%I Poor (%) (%I Number of children under 6 None 37.7 23.6 29.2 1 43.2 25.3 27.3 2 48.7 26.7 25.6 3 or more children 63.4 24.4 18.0 Household size 1 10.8 0.4 1.9 2 22.6 1.7 3.4 J 25.9 3.8 6.8 4 30.3 7.8 11.9 5 38.1 11.8 14.4 6 46.6 15.0 15.0 7 or more 59.8 59.6 46.5 Total 46.6 100.0 100.0 Source: WB estimates based on KIHBS2005-2006 data. 2.65 Poverty in Kenya has a predominantly young face. The population pyramid is such that over half the population is aged under 20 years, and they account for two thirds ofthe poor inKenya. 37 Figure 2-14: Average Household Size and Number of Children per Household,by Decile 5 - 7 Source: WB estimates basedon KIHBS2005-2006 data. 2.66 The multivariate analysis confirms that household demographics are strongly linked to poverty, although some o f the effects require further investigation. The following appear to be important: Household size and composition: Adding household members has a negative effect on consumption per adult equivalent, confirming that large households are more likely to be poor. However, the effect is non-linear and the marginal effect declines as households size increases. Surprisingly, a higher share o f female adult membersis linked to higher consumption levels in urban areas, holding all other factors constant. This contradicts the view often expressed in the PPA-IV that a highproportiono fwomen inthe householdincrease the likelihood o fbeingpoor. Age of the household's head While poverty overall increases with age, once we control for other factors, such as size and education, age per se has a positive impact on consumption in urban areas. This probably reflects earning premiums based on work experience. Households with More Education and Vocational Skills are Less Likely to be Poor 2.67 The human capital profile o f the household has strong positive effects on consumption. Returns to education increase with the level attained. The pattern differs between urbanand rural areas inexpected, but notable, ways: 0 In rural areas, even some primary education o f household members has positive and significant returns. 38 Inurban areas, significant positive returns to education manifest only from some secondary education onwards. Returns to university education are higher in urban than in rural areas, mirroring higher-skilled employment opportunities incities and towns. 2.68 Figure 2-15 shows the education of household head by decile, and reflects that the vast majority do have primary education. However, the patterns do vary dramatically. Over half of heads inthe poorest decile report no education at all and a further 40 percent have only primary education. Likewise, Tegemeo analysis shows that the more educated i s the household head, the lower the poverty rate. Figure 2-15: Educationof HouseholdHead by Decile 100% 90% Other 80% Universityor College 10% Secondary 60% 0SeniorPrimary 50% Junior Primary 40% - BNone 30% 20% 10% 0% . Poorest I 2 3 4 5 6 1 8 9Richest I I Source: Staff estimates based on KIHBS 2.69 Interestingly also, having a household member with vocational training increases consumption by 14 percent in rural and by 19 percent in urban areas. We examine the importance of education further below, inthe context of youth. Poverty Incidence is higher among Female-headedHouseholds 2.70 One fourth of all households in Kenya are headed by a woman. Female headed households tend to be significantly larger in size, especially among women who are divorced, never married or whose husbands are away. Compared to male headed households, their poverty incidence averages 5 percentage points higher. Inurbanareas, poverty incidence among female headed households is 48 percent, compared to 31 percent among male headed. Female-headedpoverty incidence is worse in districts that have a higher proportion of female headed households (Figure 2-16), though we did not identifythe reasonsfor this pattern. 39 Figure 2-16: Poverty Incidenceand Gender of the HouseholdHead, District Level 0 20% / I I 0% 10% 20% a% 40% 50% 60% 70% 80% 90% im% Fernah Headd (Poverty Incldence) 2.71 Our analysis suggests that the reason for female headship matters. Specifically, as shown inthe results inTable 2-11andby some complementary analysis: The subgroup with the highest poverty incidence (69 percent) i s that o f married female household heads. In one third o f these cases, the husband i s chronically illor disabled, and they are disproportionately poorer when the household size exceeds three. Another third are cases where the husband has been away from the household for only a few months. This may arise because recent migrants are initially unable to send significant remittances home, since poverty incidence among married female headed households whose working (and remittance sending) husbands have been away for longer periods is not above the national average. While households where the spouse has died constitute a substantial number (13 percent of the total population), poverty incidence among this group is only marginally higher thanthe national average. 0 For households headed by widows, poverty incidence i s significantly higher (54 percent) and especially in urban areas where 2 out o f every 3 o f these households are poor, compared to 52 percent inrural areas. About 14 percent o f all Kenya's poor live in households headed by widows, in which the children are known as Orphans andVulnerable Children. Poverty i s much lower (28 percent) among households headed by a woman who has never been married. 40 Table 2-11: Poverty and FemaleHeadship-Selected highlights Share of the Population Population PopulationLiving Livingin Poor Living in in FemaleHeaded FemaleHeaded Female Headed Poverty Households Households Households Incidence(YO) MarriedMonogamous 2,721,081 1,293,992 29 48 HusbandHome 560,355 366,506 6 65 HusbandAway 2,160,726 927,486 23 43 MarriedPolygamous 944,199 552,405 10 59 HusbandHome 231,051 193,781 2 84 HusbandAway 713,147 358,624 8 50 LivingTogether 34,137 12,154 0 36 Separated 486,450 249,482 5 51 Divorced 319,128 156,083 3 49 Widow 4,207,061 2,262,827 45 54 Never Married 597,654 169,119 6 28 Missing 43,445 16,568 0 38 Total 9,353,155 4,7 12,630 100 50 Source: Staff estimates basedonKNBS (2007) and KIHBS2005-2006. 2.72 Regression analysis shows that controlling for a range of other factors, male- headed householdshave on average 19 percent higher consumptionthan female-headed householdsinrural areas, but there is no significant difference inurbanareas. The factors at work here will have to be investigatedfurther. 2.73 The regression results show that widow(er)s have significantly lower consumption levels inurban areas but higher consumptioninrural areas. What is behind the high poverty rates of urban widow headed households? Additional investigations foundthat: The hypothesis is that urban widows are poor rural migrants does not hold. However, while urbanwidows who are householdheads are muchmore likely to be migrants than rural widows overall, there is no large difference by poverty status: 46 percent of poor urban widows come from the district in whichthey nowresidecomparedto 42 percentof non-poorwidows. Poverty among urban widow-headedhouseholds is especially high when the headis young or elderly. Interms of outside financial support, the average transfers (includingofficial sources and family friends) are higher for urban widows than for rural widows. However, the mediantransfer for urban widows is zero, while it is 1000KShfor rural widows, which suggests that at least half o f urbanwidows get no transfers whatsoever. Poor rural widow-headedhouseholdsalso have an asset endowment inthat they are more likely to own landthan non-widow headedhouseholdsand are morelikely to have a title. 41 These results provide some clues, but further analysis is needed to understand the relatively highpoverty of some sub-groups of widow headedhouseholds. 2.74 While female headship i s not uncommon, the majority o f Kenyans do not live in householdsheadedby a woman. Inan effort to get at a broader picture, we examined the effect of a larger share o f female adults in the average household through our consumption regression. The coefficient on this variable, 0.38 i s positive and significant in the urban sample, and particularly strong for Nairobi.28 One attempt to put this in context is to consider the following example: imagine a household with three adult members, two men and one woman. Ifthe household instead had two women and one man, predicted consumption per capita would be around 12.5 percent higher. This result, and its concentration inurbanareas, also bears further investigation. Migrant Households tend to be better off 2.75 There is significant mobility within Kenya: at the time of the KIHBS, almost one third of household heads lived ina district other thanthe one they were raised, and 20 percent lived in a different province. Nairobi and Rift Valley Province are home to the bulk ofmigrants that were borninanother province. 2.76 Migration appears to be a good strategy in terms o f household well-being: poverty incidence among migrant households is only 28 percent, although they also could have migrated becausethey were wealthier in the first place. Compared to non-migrant households, poverty incidence among migrant households is lower in all Provinces, except for Nairobi, where there i s no significant difference. Poverty i s also lower among migrants who were born in Central Province. Coast and North-Eastern Province are characterized by the largest gap in poverty incidence between migrant and non-migrant households, where the latter are more thantwice as likely to be poor. 2.77 These broad patterns hold in our regression analysis, which is more suggestive, though not conclusive, about the direction of causality. Controlling for education, sector of employment and a whole host of factors associated with poverty status, households that have migrated from an urban area within Kenya have higher consumption than non- migrant households. However, rural households which migrated from another rural area do not have significantly higher consumption, controlling for other observable factors. More detailed analysis elsewhere controlling for possible selection bias and other econometric problems, has found a positive impact of internal migration on household welfare and poverty reduction (see Giles, various on China, MacKenzie on Mexico). One issue not investigated here i s whether in migration, especially in land-scarce areas, poses pressures on host communities, and for example, on unskilledwages. 2.78 For the much smaller number of households that have returnedto urban Kenya from abroad are substantially better off (81 percent higher consumption per capita). Inaddition to all ofthe control variables that have already beendiscussed, it is important to keep inmind that these regressions include controls for the gender ofthe head. 42 Receiving transfers from abroad increases consumption by more than one-third, while remittancesfrom within Kenya are only marginally significant inrural areas (4 percent). 2.79 In the regression analysis, ethnic fractionalization at the district level-as approximated from the 1989 census-is inbothurbanandruralareaspositivelyrelatedto consumption. This probably mirrors the attraction of relatively prosperous districts to migrants from other parts of Kenya. It also suggests that migration and ethnic diversity are conducive to economic development. 2.80 Nairobi's slums are characterized by significant mobility of individuals and households, in and out. The APHRC analysis (detailed below) shows that, for the 120,000 individuals inthe two slums, consistently, over the 13 rounds of data collection between 2002 and 2006, about half of the sample left the slum each round. Interms of inflow, over half came from rural Kenya, but notably, about half came from urban areas that were not slums. Among those exiting, about 40 percent went back to rural Kenya, and almost 20 percentmoved out of slums altogether. Figure2-17: Where Migrants Come from and Go to, Nairobi Slums Source: de Laat et al, 2008 2.81 Poverty rates are far above the slum average for in-migrants-which suggests that well-being likely improves over time for those in the slums, although out migrants also have above average rates o f poverty-which may also motivatehecessitate their departure. 43 Figure 2-18: Poverty Ratesby Migration Status 85 80 75 60 Overa II Inmigrants Outmigrants Source: de Laat et al, 2008. 2.82 Since the time of the KIHBS, the recent crisis has created a large number of migrants-many of whom were involuntary-known as internally displaced people, or IDPs. As of April 4,2008, there were 168,094 IDPs in 181camps. This figure had fallen by more than half since mid January. However, the numbers were rising again in April, and are evidently not stable.29 In addition to the IDPs living in camps, many Kenyans displaced by violence have sought shelter with relatives, or through other means (e.g. renting), and are termed `integrated' IDPs. It i s obviously very difficult to precisely estimate the size of this group. However, by comparing the numbers of persons counted inthe initial waves of displacement (typically to local police stations) with those inthe camps, KRC estimates that there are 150,000 such people, which i s likely a lower bound estimate. The profile o f these involuntary migrants is likely very different fiom that which we observe inthe earlier household surveys. Table 2-12: IDPs in Kenya, as of April 2008 Number of IDP Camps 181 Current numberofIDPs incamps 168,094 Maximumno.of IDPs incamps during crisis 3 10,000 Estimated number of `integrated' IDPs: 150,000 Source: KenyaRedCross. TheEconomicActivitiesof theHouseholdSignificantlyAffect Consumption 2.83 Employment projle: Not surprisingly, employment status and activity are correlated with consumption. Rural households whose head is,engaged in agriculture have about 7 percent higher consumption levels than households whose head is not working, though the statistical significance i s marginal. Premiums in the service sector are high in rural areas, amounting to 29 (miscellaneous services) or even 43 (public administration) percent. However, only a small part o f the rural population is employed *'Figures courtesy of the Kenya Red Cross, who manage the camps, and produce a daily count o f registeredresidents. 44 inthe service sector. Counter-intuitively, employment inindustrial activity inurbanareas i s negatively linked to consumption levels. This result merits some further scrutiny. We examineemployment andunemployment patterns further below. 2.84 Diversification of incomes appears to be important, confirming the results of earlier work, including Lay, M'Mukaria and Mahmoud. (2007). Households with at least one member working for a household-owned non-farm enterprise have about 11 percent higher consumption inurbanareas, inrural areas the effect is smaller (7 percent) but still significant. This confirms that non-agricultural self employment i s an important livelihood and can be a pathway out of poverty (see Mango 2008, World Bank 1995), a proposition that we discuss below. 2.85 The picture on income diversification among the poor provides some important clues for the analysis of dynamics in Part 111. Tegemeo's findings are clear, and confirm expected patterns. The highlights for the decade to 2007, which are explored further below, are as follows: 0 Crop income i s very important-for the non-poor, it constitutes 38 to 49 percent of total income, and for the poor even more significant, from 47 to 87 percent. The share of income derived from off farm activities is high andrising for the poor, but falling for non-poor households (whose income share from salaries i s rising). Poorer households have much lower fractions o f adults involved inbusinesses or in salaried activities. The fraction of adults in a household involved in business activities has been rising fast over time, which is welcome. Even more notable is that the rate of increase has been fastest among the poorer households. 0 On average, the fraction of adults ina household involved insalaried activities has beenrising over time, but mostly only for the non-poor households. Figure 2-19: Fraction of Adults Involved in Off Farm Activities: Tegemeo 1 Non-Poor Poor Non-Poor Poor Business and 1 Buslness and Salary Salary Informal Informal [92000 0 2 0 0 4 H2007 Source:Nyoro et a1(2008). 2.86 Agricultural practice and landholding: Tegemeo analysis shows that both land holdings and land cropped actually fell over the decade to 2007, andmore rapidly for the 45 poor. This is consistent with evidence from the PPA about demographic and land pressures. The issue of land, and inparticular land inequality, i s discussed further below. Interms of improved practices, Nyoro et a1 (2008) found that poor households are less likely to practice zero tillage, composting and on farm feed preservation. 2.87 The multivariate analysis of KIHBS confirms that land- and agriculture-related factors are linkedto consumption in rural areas but the effects are smaller in magnitude than expected. An increase inlandholding size per working age adult from one acre to four acre increases consumption by about 9 percent inrural areas. Mean holdings are 2.4 acres per household in rural areas (including landless households), and the standard deviation is 9.3. Having ownership rights for land holdings increases consumption by about 7 percent, which i s a smaller measured effect than having land titles (11 percent). Households practicing irrigation have on average 8 percent higher consumption levels. We investigate these patterns further below in Part 111. Not surprisingly, the regressions show that there are seasonal patterns of consumption, though the effects are only significant for rural areas: average consumption levels appear to be lowest between Februaryand April, just before the onset of long rains. 2.88 Tegemeo analysis shows that there are big differences in poverty rates across those who tried to get credit and those who did not (noting that the correlation between those who applied and those who received credit i s very high). The importance of the credit variables appears to have increased over time, with some divergence across the distribution as access increasedmore rapidly for the highest quintile. Figure 2-20 shows the trends by province. Interestingly, Tegemeo analysis shows that poor households are much less likely to be group members. Overall, the sample shows group membership decreasing over time, more rapidly for the non-poor. Figure 2-20: Credit by Province: Tegemeo 0 8 0.7 0 6 0 1997 0.5 ~ 2 0 0 0 0.4 a2004 0.3 m 2007 0 2 0 1 0 Coast Eastern Nyanza Western Central Rift Va IIey Source: Nyoroet a1(2008) 46 HouseholdLocation Matters 2.89 We have already noted the large differences in poverty rates and inequality betweenrural and urban areas. Moreover, several province fixed effects are significant, which implies that all other factors being equal, households in different parts of the country enjoy very different consumption level's, specifically: 30 0 Households living inNairobi have 31 percent higher consumption levels than urban households in Western province, which serves as reference category. Urban areas in Central and Eastern province also appear to be better off, whereas urban households in North Eastern province have 33 percent lower consumption levels thanthose inWesternprovince. In rural areas, households in Central province enjoy 31 percent higher consumption levels than those in Western province (other observed factors being equal). 2.90 O f course, these provincial fixed effects capture a myriad o f regional factors, ranging from differences in institutions, public services and infrastructure to agro- ecological variations. To understandhow marginal effects vary between different parts of the country, we interact each of the non-geographic independent variables with each of the provincial dummy variables. This is done separately for rural and urban areas. A Wald test is usedto test the joint significance of interaction effects for each province and to test for equality of interaction parameters for each pair of provinces. The results indicate that inter-provincial differences in the way the explanatory variables affect consumption are important-location matters a 2.91 To set the stage for the analysis by province, Table 2-13 summarizes key characteristics interms of natural endowments, human capital and economic structure. 30The regressions control for many factors that impact on welfare and poverty. Hence, province fixed effects do not merely mirror the differences inthe characteristics associated with welfare and poverty levels observed inthe different provinces. 31Thus the nullhypotheses o f no inter-provincial differences in marginal effects can always be rejected at the five percent level. 47 Table 2-13: An Overview of Provincial Differences, percent, 2005/6 Nairobi Central Coast Eastern North Nyanza Rift Western Eastern Valley Urban share 100 10 37 6 17 14 12 8 Share of medium and N a 69 5 3 97 18 90 highpotential land Average consumption(1) 7,355 2,931 2,456 2,109 1,450 2,280 2,361 1,909 Decile ratio 0.05 0.08 0.06 0.08 0.09 0.09 0.05 0.11 Poverty rate 22 31 59 50 74 47 49 53 Unemploymentrate 21 10 17 9 40 7 13 12 Net Secondary school 38 30 11 15 7 19 18 13 enrolmentrate LandGini (value) 0.99 0.74 0.87 0.73 0.82 0.87 0.77 Source: (1) and (2) World Bank staff estimates basedon KIHBS 2005-2006 data. Notes: basedon deflatedmonthly per adult equivalentterms. 2.92 Figure 2-21 compares access to roads for the population, both nationally and in each province. The data, from 2004, pre-date the more recent scale up in roads investment, but do provide a sense of the pattern of access at that point. The figures were computed using GIS data on the roads network andthe population ineach sub-location in Kenya. The analysis was conducted on two sets o f roads data for 2004: (a) nearest distance to any road; and (b) nearest distance to a bad road (defined as roads with earth or non-graded surfaces). 2.93 The results show that overall, with over 90 percent of the population living within 3 kmof any road, access inKenya is high. However, there is substantial variation among provinces with lower rates of access in Coast, Eastern and Rift Valley Provinces. At the same time it is recognized that parts of Rift Valley (Turkana, West Pokot, parts of Baringo, Samburu), Coast (Tana River) and Eastern (Marsabit, Moyale, parts of Isiolo). Provinces include vast areas with low population densities because of the harsh climate and terrain and communities that migrate regularly to search for pasture. Inthis sense it i s not surprising that road density i s relatively low compared with other regions. 2.94 Moreover, analysis by type of road surface shows that "access" is largely to roads with earth or non-graded surfaces, which can be classified as bad roads. Nationally, about half of the access is to bad roads-with relatively high proportions of badroads inWestern, Nyanza andEastern. 48 Figure 2-21: Patterns in RoadsAccess by Province, 2004 National (Any) National (Bad) Western (Any) Western (Bad) Rift Valley (Any) Rlft Valley (Bad) 8 ~ 3 k r n Nyanza (Any) 8 c 4 k m 03< 5km Nyanza (Bad) 03c 10km O c 15km Eastern (Any) 0 c 2 0 k m Eastern (Bad) Coast (Any) Coast (Bad) Central (Any) Central (Bad) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Proportlon of the Populatlonthat Ilve8 wlthln (x) km of a road Province by Province 2.95 To explore provincial variations further, regressions were estimated separately for Kenyan's eight provinces. 32 This section gives a short profile of each province bringing together evidence from various data sources and types of analysis. We utilize KIHBS, and the rich analysis oftrends at the provincial level undertakenby Tegemeo- which are representativeof rural, non-pastoralist areas. The rates of poverty were shown inFigure 2-3 above. 2.96 Before beginning with the province-by-province overview, we draw attention to large province level differences in household access to infrastructure that emerge from the analysis of Tegemeo data: Coast and Rift Valley do poorly on distance to fertilizer seller, while Rift Valley is very heterogeneous. Nyanza starts off worst on motorable roads in 1997, but ends up better than Coast and RiftValley in2007. Central i s the best throughout. 2.97 The area surrounding the modern capital and regional business center of Nairobi hosts more than two-thirds of enterprises accounting for 44 percent of formal employment (see World Bank, 2005). Nairobi is a major centre for East Africa, interms 32Econometrically, there is no difference between running the regressions by province or interacting all non-geographic explanatory variables with provincial dummies. However, the specification using interaction effects is convenient for econometric testing while running the regressions by province eases the interpretation o f coefficients. 49 of transport, multi-lateral organizations and commercial activity. Between 1997 and 2005/06, households in Nairobi enjoyed the most rapid consumption growth, totaling about 45 percent. This was associatedwith a strong decline inmonetary poverty, from 50 to 22 percent comparing the WMS 1997 and KIHBS. 2.98 However, the Kenyan capital i s also characterized by massive inequalities in income and wealth distribution. Expenditures of the richest decile are almost 20 times higher thanthose of the poorest decile, the gini index for Nairobi is 0.46. Unemployment among youth aged 15 to 24 years stands at 42 percent, with many of them living in Africa's largest slums. Living conditions inNairobi's informal settlements are appalling, characterizedby widespread poverty and limitedor nonexistent public services. Housing units in the slums are shabby and crowded, yet rents are high, absorbing a sizeable fraction of household budget (see World Bank, 2006). Recent data from two large informal settlements (Korogocho and Viwandani) show that less than 3 percent of households living inthese areas had access to aprivate latrine, only 10 percent to garbage disposal services. As a result, open spaces and drains are flooded with human excreta and waste, leaving slum dwellers vulnerable to poor hygiene and illnesses. Over and above, lack of safety i s a serious problem and most slum households feel insecure and at riskto crime(World Bank, 2006). 2.99 Alongside such striking inequities, the overall vitality o f the private sector in Nairobi attracts the highest-skilled segment of the labor force. Around 55 percent of adults living inthe Kenyan capital have at least some secondary schooling, and another 7 percent have benefitted from university education. Stifel and Christensen (2006) have shown that in 1997 more than primary education was necessary to make a positive difference to consumption among households inNairobi. The results from our regression analysis confirm this finding, as only tertiary education has significantly positive returns inNairobi. Nairobi also shows highest returns to work experience, as approximated by the age of the household head. 2.100 CentralProvince bordersNairobi is the traditional homeland of the Kikuyu. As noted in Part I,infrastructure development benefitted from colonial rule, and the post independenceregime. The province has poverty estimates far below the national average -with a headcount index of 31 percent. However, there was no significant poverty reduction between 1997 and 2005/06 indicating that even lower levels of poverty will be muchharder to achieve. According to Tegemeo's panelanalysis, over the decade to 2007, average incomes rose inCentral, but inequality also tendedto rise. 2.101 Non-monetary poverty indicators and correlates are favorable in central province. Central province has high primary and secondary enrollment rates, infant and child mortality rates are the lowest inthe country and fertility is markedly lower than in most other partsof Kenya. These are all good for loweringpoverty, since the regressions show that there is a strong negative link between household size and consumption. Despite the fact that the province i s fairly homogenous ethnically, there is a positive relationship betweenethnic fractionalization and consumption, showing that districts with mixedpopulationtendto bemore prosperous. 50 2.102 From a geographic perspective, Central Province benefits from its proximity to Nairobi and the mild climate of the central highlands. A fair share of the land is high- potential, with tea, coffee and horticulture being important cash crops alongside the dairy industry. Financial services are relatively advanced and particularly the semi-formal financial sector, through a prominent role of SACCOS, which reduces financial exclusion rates (relative to Nairobi). About 70 percent o f the adult population are economically active, which is more than in any other provinces. The results of the regression analysis are in line with the notion that Central province i s strategically positioned. Urban areas show significant positive returns to non-farm household enterprise and wage employment, while rural areas reflect positive returns to human capital even at the primary level. There are also highpremiumsto employment inpublic administration. 2.103 Coast Province is home to the Mijkenda and other smaller, predominantly Muslim, groups. Urbanization is high, with 37 percent of the population living incities and towns. Mombasa, Kenya's second largest city and chief port, serves as the main gateway for cargo traffic into the interior of East Africa, particularly the inlands of Kenya, Uganda, Rwanda and Northern Congo. Prior to the 2008 election crisis, Mombasa also welcomed more than one million tourists each year. However, despite significant beach tourism and the hotel industry along the Indian Ocean, the province is characterized by high and rising levels of poverty, with headcount poverty having increased from 55 percent in 1997 to 59 percent in 2005/06. Small area estimates of poverty and welfare show that gains from the tourism sector are largely confined to the areas directly surrounding the main hotel complexes and do not trickle down to the hinterland. Unemployment is high, as 17 percent of adult population and 26 percent of youth population are without ajob. 2.104 The region also lags behind in non-monetary poverty indicators, particularly in those related to education: almost two thirds (63 percent) of adult population has not completed primary education. Primary and secondary enrollment rates are among the lowest inthe country, only the impoverished areas of North Eastern Province fare worse. Inaddition, there are significant gender disparities, as the ratio ofboys to girls inprimary school stands at 10 to 9 (gross enrollment). On the positive side, access to basic public services and amenities, particularly water and electricity, i s comparatively high, mainly because a large share of the population lives inurbanized settings. 2.105 Together with Eastern and North-Eastern Province, Coast has the smallest proportion of arable land. Only 5 percent of the land i s classed as high potential for crop production and much of the inland areas are agro-pastoralist. Coast Province also has a high fraction of Government land that has not been allocated (see SID, 2006). Perhaps due to its relatively poor quality, land is not an important determinantof consumption in this province as none ofthe land-relatedvariables comes out significant. 2.106 According to the Tegemeo rural analysis, average incomes in Coast province actually deteriorated between 1997 and 2004, then recovered through 2007. In this sample, trends ininequality were uneven over time, with an initial improvement followed byworsening, thenimprovement again. 51 2.107 North-Eastern (NE)Province, sparsely populated and bordering Ethiopiato the north and Somalia to the east, is the most impoverished region in Kenya. Almost three quarters of the population are estimated to be poor, and the high values of the poverty depth and severity indices indicate a sizeable consumption shortfall of poor households relative to the poverty line. The province receives very little rainfall, and its arid and semi arid lands are inhabited mainly by nomadic pastoralists from the Somali ethnic group. The fertility rate stands at 7 children per women, about twice as much as in Central Province. 2.108 Due to its relatively isolated location and dispersedpopulation, the region has long been disadvantaged inpublic service and infrastructure provision (see World Bank, 1995). This has manifestedin a very low asset and endowment base. Only 4 percent of the population use electricity, less than one third has access to safe water. A massive 88 percent of adults have not completed primary education. Despite the introduction o f free primary education, only half of the children are enrolled in primary school and only 7 percent receive secondary education. At the same time, there may be little incentive to enroll: the KIHBSregressions also indicate that there are hardly any returns to education in rural NE.33Gender disparities are well pronounced, as the boy-girl-ratio inprimary school is 100to 61, 2.109 Pastoralism is the predominant livelihood in North-Eastern Province, but economic inactivity rates reported in the KIHBS are the highest in the country. The province also suffers from chronic food insecurity. A fair share of the population depends on food aid, especially when livestock is lost after periods o f prolonged drought. Cattle rustling, poaching and banditry are pervasive. Along the Kenya-Somali border, there are several refugee camps to shelter for people displaced by the civil war in Somalia. People living inthese camps are vulnerable to attacks by Somali fighters and local bandits. 2.110 Eastern Province is the second largest administrative area, stretching from Lake Turkana at the northern border to Ethiopia down to the south. It is also the least urbanized province, with almost 94 percent of population living inrural areas. The area is inhabited mainly by the Kamba and Meru, as well as various pastoralist groups. The regressions show that households following traditional beliefs are significantly worse off. Between 1997 and 2005/06 Eastern Province experienced seen a decline in poverty, but still more than halfof the population falls below the poverty line. 2.111 Eastern Province has the second highest life expectancy in the country (after Central), but other nonmonetary poverty indicators are not so favorable. Almost 60 ' percent of adults have not completed primary education, only 15 percent of children are currently enrolled in secondary school. Infant and child mortality rates are below the national average but considerably higher than inCentral Province. 2.112 Most of the land in Eastern Province is low potential, though there are some pockets of arable land around Embu, Makueni and Mbeere (see SID 2006). In the regressions investigating province level effects, and ownership increases consumption by 33Though the small sample size makes it difficult to precisely estimate coefficients. 52 around 16 percent and land titling adds another 13 percent in the province. However, landholdingsize is not significantly correlated with consumption inEastern. 2.113 In terms of trends, Tegemeo's panel analysis found that Eastern's average incomes rose through 2004, but worsened from then onwards. Trends in inequality varied over the decade to 2007, worsening initially and again more recently. 2.114 Rift Valley,the largest and most populous province, covers almost 30 percent of Kenya's total land area. The Great Rift is the traditional homeland of the Kalenjin, but it i s a very diverse province, and there are significant numbers of Kikuyu, Samburu and Maasai living in this area. The province is characterized by an unequal distribution of income and land, with a consumption gini of 0.42 and a land gini o f 0.87. Inter-ethnic clashesand land conflicts are frequent inthis part of Kenya. 2.115 As noted inPart I, Valley was politically favored under the second Kenyan Rift President DanielArap Moi (1978 - 2003), a Kalenjin fiom Baringo district. Yet poverty in Rift Valley Province remained remarkably stable over the past decade, with a headcount index of around 49 percent. Rift Valley has the third highest poverty severity index (after Coast and North Eastern) showing that many poor households are far away from the poverty line. 2.116 The province is characterized by significant geographical variations. The volcanic soils of the Great Rift are amongst the most fertile in the country, Central Rift Valley is the leading producer of maize and wheat, the highland area around Kericho is the center of Kenya's tea industry. However, much of the land in the northern and southern districts is agriculturally unproductive and dominated by livestock rearing. In the regression analysis, Rift Valley Province has the highest returns to land holding size. Land titling has an additional positive correlation with consumption, while land ownership does not come out significant. As inCentral Province, rural areas benefitfrom positive returns to education from the primary level and there is a positive effect of non- farm enterprise employment on consumption. 2.117 In terms of trends, the Tegemeo rural analysis suggests positive directions in Rift Valley, where we see the combination of rising incomes and slowly falling inequality. This stands incontrast to the WMS-KIHBS findings of somewhat rising rural inequality. 2.118 The densely populated area ofNynnza Province surrounds Lake Victoria inthe south-western corner of Kenya. The Luo are the dominant ethnic group, followed by the Kisii. The province experienceda strong decline inmonetary poverty between1997 and 2005/6, with headcount poverty falling from 62 to 47 percent. The decline i s most pronounced for urban Nyanza, where poverty fell by 22 percentage points over this period, from 63 to 41 percent. 2.119 About 90 percent ofNyanza's land i s high potential but high population density and high fertility rates imply that average plot sizes are small and declining (see World Bank, 1995). Agriculture is dominated by maize and sugar production. The informal 53 financial sector plays a crucial role in providing access to financial services. Overall unemployment rates are the lowest in the country although, as shown below, youth unemployment is a major issue. As expected, the regressions show that returns to land holdings are positive, though smaller than inRift Valley. Non-farm self-employment is extensive in rural areas, and in the regression analysis this is related to higher consumption. At the same time, urban areas suffer from low returns to education comparedto most other provinces. 2.120 Tegemeo's panel analysis shows that the reported incomes of rural households inNyanza are the lowest of all the provinces, butrose over time, while the gini increased till 2004 and then moderated. Levels of livestock ownership are lowest inNyanza. It is also notable that average land size was falling rapidly, to become lowest across all the provinces by 2007. Interms of trying to get credit and receiving credit, Nyanza starts off the worst and ends up secondworst, just ahead o f RiftValley in2007. 2.121 The decline in monetary poverty contrasts against daunting health indicators. Nyanza has the highest HIV/AIDSprevalence rate among all Kenyanprovinces and this, coupled with endemic malaria and water-borne diseases, inflicts a devastating toll on the population. Average life expectancy inNyanza Province i s 16 years less and under-five mortality four times higher, than in Central Province. One child out o f five born in Nyanza does not live to see hisher fifth birthday (see UNDP 2006 and SID 2006). 2.122 Western Province lies at the border to Uganda and i s inhabited mainly by the Luhya ethnic group. The area, which includes Kakamega rainforest, receives strong rainfall throughout the year and has a low degree of urbanization. Despite fertile lands, Western Province i s among the poorest inKenya and has experienced only slow progress over the last decade. Between 1997 and 2005/06, the headcount index fell from 57 to 53 percent. The Tegemeo analysis suggests that Western province's growth performance was uneven, with some recovery after 2004, and inequality generally improving. 54 Box 2-4: Location Matters To illustrate the importance of inter-provincial inequality, consider the following example: Our reference household is Patrick (34 years old), his wife Mary (28 years old) and their three children (7 years, 5 years and 4 years). Patrick has completed primary education, while Mary did not complete primary education. The family owns a small plot o f 2.5 acres-but does not have a title-in a rural village and i s mainly involved in subsistence agriculture. However, Mary also works a few hours per week for a small non-farm enterprise (see Annex for a full list o f reference characteristics). What resources would Patrick and hisfamily require to meet the poverty line? The monthly rural poverty line i s 1562.18 KSH per adult equivalent to meet the cost o f purchasing 2250 calories and basic non-food necessities. Patrick and his family would be considered 3.54 adult equivalents, as the children are assumed to cost less. Hence, the above householdwould need 5530 KSHper month to meet the poverty line. What consumption levels would our regression models predictfor Central and Western provinces? Let us assume first that Patrick lives with his family in Central province. Our regression model for Central province predicts that this household would have total consumption o f 9455 K S H per month. Hence, Patrick and his family would be able to spend each month an amount o f around 3925 K S H beyond o f the poverty line. Now assume that the family resides in Western province. Our regression model for Western province predicts that the same household would have total consumption o f only 5501 KSh per month. Since the poverty line would remain the same, Patrick and his family would now be just below the poverty line and hence classified as poor. Inthis example, the difference inoutcomes (despite same characteristics) is drivenby the following factors. Central province has a much more favorable `province effect' (the intercept in our regression model), which captures unobserved factors associated with location, such as climate, rainfall, soil quality, average distance to roads and markets, local institutions, etc. Second, the positive relationship between ethnic fractionalization and consumption is much larger in Central than in Western province (our reference household is considered to live in a district with moderate ethnic fractionalization, e l f = 0.304). Third, Central Province has considerably higher returns to education, even at the primary level. On the other side o f the coin, Central province is strongly disadvantaged with respect to household demographics, which offsets the above mentioned effect to a large extent. Of course, this is an artificial example and there are a few things to bear in mind. First, the effect is not uniform across households but depends on the underlying household characteristics. Second, average household characteristics do also vary between provinces-and these effects can either reinforce or offset the effects illustrated above. For example, average household size is higher in Western than in Central province, hence Patrick would be more likely having to provide for another dependant if living in Western Province; on the other hand, Patrick would be less likely to own land ifhe lives in Central Province. Third, our regression models do not tell us why the returns are different. This could be related, again, to agro- ecological variations, differences in infrastructure or public services, institutional factors, and hence forth. However, the results do show that inter-provincial variations in returns to household characteristics are important and that location i s a powerful predictor o f living conditions andpoverty. 2.123 Agriculture inWesternProvince is dominated by subsistence farming and maize i s the preferred staple crop, often intercropped with beans. Commercial farming in Kakamega and Busia districts centers around sugar cane, which is processedeither inthe large sugar factories inthe area or insmallholder mills.Vihiga district has tea plantations, and fishing is important inBusia at Lake Victoria. However, average productivity is low due to high population growth, land fragmentation and resulting small farm sizes. The 55 regressions show that land holding size is positively related to consumption, but there is no separate effect of land ownership or land titling. Irrigation increases consumption by about 20 percent, which i s about the same magnitude as inNyanza. There are significant premiumson employment inthe service sector. 2.124 Health indicators in Western are worse than the national average, though not as poor as inNyanza. WesternProvince also lacks behind in infrastructure access. Only 4 percent of the population has access to electricity, while around 6 percent use piped water. Focus on the Slums 2.125 Nairobi, growing at an estimated annual rate of about seven percent, is one of the fastest growing cities inAfrica. Most of the growth of the city ofNairobi is a result of rural-urban migration (APHRC, 2002), and many migrants move into one of the city's informal settlement (slum) areas. A 2002 UNconference defined a slum household as a group of individuals living under the same roof lacking one or more of these conditions: (1) access to improved water; (2) access to improved sanitation facilities; (3) sufficient- living area, not overcrowded; (4) structural quality/durability o f dwellings; and (5) security of tenure. Under this definition, the UN estimates that a large majority, more than 70 percent of people living inNairobi reside in slums (UNHabitat, 2004).. Using a more stringent definition, the Central Bureau of Statistics estimated that about 30 percent of the city's population lives in slums (Gulyani and Talukdar, 2008). Inthis section we use the terms "slum" and "informal settlement" interchangeably. 2.126 Nairobi, on average, is much better off than Kenya in general, on virtually all indicators of well being. However, this average is misleading, since Nairobi is highly unequal. Many of the wealthiest Kenyans live in the capital, and yet there are large pockets of relative and absolute deprivation. In this section we focus on the latter and provide an overview o f some of the most salient aspects of poverty and livelihoods inthe slums. The primary data source i s the Demographic Surveillance Survey collected by APHRC since 2002 on the well-being of the 13,000-15,000 households (44,000 individuals) that live at any one point in time in two of Nairobi's main informal settlements Viwandani and K o r o g ~ c h o Inparticular, we exploit the rounds undertaken . ~ ~ in 2003 and 2006 which collected information on livelihoods. This data source is complemented by the Nairobi Cross-sectional Slum Survey (NCSS) carried out in 2000 by APHRC on a representative sample of 4,564 slum households, and the Nairobi Informal Settlement Survey (NISS) carried out in 2004 on a random sub sample of 1,817 households inViwandani and Korogocho whose household headwas ever married.35 2.127 The 2000 NCSS showed that unlike the usual pyramid shape of the population found elsewhere in Kenya, there are more men than women (56 versus 44 percent) in 34This is in part based on Poverty Dynamics and Mobility in Nairobi's Informal Settlements, APHRC Working Paper (2008) by de Laat, J., 0.Faye, T. Suri, andE.Zulu. "ThissurveywascollectedbyJoostdeLaat,incollaborationwithAPHRC. 56 Nairobi's informal settlements, particularly o f child bearing ages (20-34 years). Children, especially those o f primary school ages, are relatively few. Figure 2-22: Age-Sex Composition of Nairobi, Kenya, Informal SettlementPopulations Male-Female I I 80+ 75-79 70-74 65-69 60-64 55-59 50-54 45-49 40-44 35-39 30-34 25-29 20-24 15-19 10-14 - 5-9 0-4 Source: APHRC, 2002 (representative sample o f all informal settlements) 2.128 Amenities in the slums are very poor. As we explore further in Part IVY improving the state o f amenities and public services is arguably one o f the biggest challenges to improving the quality o f life o f people living inthe informal settlements. 0 Water. According to the DSS 2006 data, less than 6 percent o f households had access to piped water in their homes. 2 percent relied primarily on a water tank (rain harvesting), and 92 percent received their water from open taps where they purchase water. Recent focus group discussions conducted with women and girls that the cost and availability o f water emerged as a major concern in both Korogocho and V i ~ a n d a n iSlum-dwellers pay approximately eight times more . ~ ~ for water than their non-slum counterparts inNairobi as the latter pay a standard rate o f 120 shillings for up to 10,000 liters o f water. Discussants in both communities also provided numerous accurate examples o f illnesses that might be provoked by water problems, such as cholera, typhoid, and diarrhea. And despite this, participants voiced their powerlessness to take the necessary actions to avert or alleviate these conditions. 0 Toilets. Less than 3 percent o f households had a private latrine. Instead, the majority (53 percent) relies on pit latrines shared with other households, or shared `other' latrines (e.g. public pit latrines and `flying-toilets'; see below). Maps for both slums show that water points and toilet facilities tend to be clustered in certain areas whilst other areas are completely un-serviced with these facilities. Many people have to pay to access the toilet and water facilities. As noted by Nyamwange (2007) in major Nairobi slums like Kibera, Mathare Korogocho, 36Undie, C., John-Langba, J., Kimani, E. (2006). "The Place o f Cool Waters': Women and Water in the Slums o fNairobi, Kenya", Wagadu:A Journal of Transnational Women'sand Gender Studies, vol. 3 57 Kangemi and Mukuru, `toilet facilities are more expensive than food. A research on poverty eradication in Kibera last year showed that toilet facilities cost KSh5 (8US cents) per visit per family member, regardless of the nature of the call. It costs a slum dweller dearly to cater for a child with a running stomach!' Since some slum dwellers cannot afford to pay for these services to meet their daily needs, they use what are normally known as `flying' toilets where they defecate into plastic bags and throw the plastics on roof tops or in rivers. Poor drainage and the rainy season can cause floods in residential areas, becoming an environmental hazard.37 Lighting. Access to electricity is improving inthe slums. In2006, 32 percent had access to electricity. The remainder households continue to rely on kerosene for lighting. GarbageDisposal. Most areas of the informal settlements continue to suffer from incredible amounts of litter and waste in open areas and proper garbage disposal remains a huge challenge. In fact, less than 10 percent of households reports relying on garbage disposal services, and less than 20 percent disposes of garbage inpits/dumps.The remaining 71percent of households either disposes of garbage inpublic openspaces (30 percent) or inthe river (41 percent). 2.129 The 2000 NCSS also revealed that morbidity risks for all major illnesses (fever, cough, diarrhea) are higher for children in the slums compared to children elsewhere in Kenya. They also have less access to healthcare, including immunization, and subsequently face higher mortality rates than even their rural counterparts. For instance, infant, child, and under 5 mortality rates were about 20, 65, and 35 percent, respectively, which are high compared to rural Kenya. Figure 2-23 based on the NISS 2004 shows that childrenbornto a migrant household head living inthe Nairobi slums are much more likely to be sick if they also live inthe informal settlement (40 percent were sick in the month prior) as opposedto living inthe rural area of origin (less than30 percent).38 37See Mudege, N. and E. Zulu (2007), "Discourses of illegality and exclusion: When water access matters", unpublishedmanuscript APHRC), andNyamwange, John (2007) "Nairobi Slums like Kibera and Mathare, - where toilet facilities, are more expensive than food", hm://www.rio 1O.dk (Accessed 27 November 2007). 38See Konseiga, A. Family migration: a vehicle of child morbidity in the informal settlements of Nairobi city, Kenya? 2005 mimeo 58 Figure 2-23: Proportion Children Sick Last Month (2-year intervals) Rural -----Urban Source: de Laat et a12008 2.130 In 2006, 62 percent o f households in the two slums had incomes below the official poverty line (per adult equivalent) with substantial variation across household types (Figure 2-24). There are various ways to examine the o f poverty profile in the slums, but a relevant typology appears to be family type, since there are significant differences depending on marital status and whether the partner i s alive or present, or not. At the same time, however, we need to bear in mind that household composition may also reflect survival strategies-in particular split migranthouseholds, but we do not have information to observe those who remain inrural or other areas. 2.13 1 Looking at patterns o f poverty by family type, poverty incidence i s highest (74 percent) among households headed by widows, who make up 5 percent o f the population. Families in which there is both a husband and wife, make up 38 percent o f the total population and their poverty incidence i s also high (71 percent). Having both parents present apparently does not compensate the household more than proportionally for the extra expenditures required. Households headed by divorced women also have above average poverty incidence. At the opposite end are members o f households headed by never married men, whose poverty incidence stands at 42 percent, divorced men (45 percent), and `married-split' (47 percent). However, since these latter households, which make up 25 percent o f the population, are married men whose wife and most o f their dependent children live in rural areas, this measured poverty incidence likely underestimates their poverty level, given dependents to support elsewhere. Figure 2-24 also shows a broad correlation betweenhousehold size and the likelihood o f being poor. 59 Figure 2-24: PopulationBreakdown and Poverty Level by HouseholdType, 2006 I' I 4 80 0% 3 5 70 0% - r 3 60 0% 2 5 50 0% .I -1 2 2 400% (Adult Eq) 0 1 5 30 0% +Below p o ~ r l line y 20 0% 0B f l d 0 5 10 0% c 0 0 0% Source: de Laat et a12008. 2.132 Not only are poverty levels high, the poverty gap is also substantial. Although there is less variation across family types in the size o f poverty gap than there is in poverty incidence, female headed households living below the poverty line have expenditures that are particularly low, 43 percent shortfall for widows and 41 percent for divorcees. 2.133 There is surprisingly significant inequality: the wealthiest decile ineither of the two slums had expenditures (per adult equivalent) 14 times higher than the bottom. The gini coefficient in2006 was 0.39. 2.134 The pattern of poverty rates is correlated with variation in the education levels of household heads. Widows, for example, have on average only 3.9 years of education compared with 6.0 years for widowers. Interestingly, education hardly differs between marriedcoupleswho live together as `married-joint' versus separately, or `married-split' . Figure 2-25: Years of Educationof Household Head, by HouseholdType, 2006 10 Married Divorced/W idowed Single Source: de Laat et a12008. 60 2.135 There i s substantial variation in the duration that household heads o f the different types o f households have stayed in the Nairobi slums. At one extreme, are widowed men and women with average durations o f more than 20 years and divorced .men and women with durations o f around 15 years. At the other end, never married men have stayed on average less thannine years inthe slum. Figure 2-26: Years Household Head hasStayed in Informal Settlements, 2006 (by family type) 10 16 .............................................. ............................................. 14 .............................................. ............................................. 12 ........... .................... ............................................. 10 ........... .................... ............................................. g ........... .................... .................... .......... 6 ........... .................... .................... .......... 4 ........... .................... .................... .......... 2 ........... .................... .................... .......... 0 - Mamed Divorced/Widowed Single Source: de Laat et a12008. 2.136 Lastly, expenditure data collected in 2003 suggests that food expenses absorb the largest share o f household expenses, both for those under (56 percent) and above (48 percent) the poverty line. Energy follows after food (15 percent and 11 percent), followed by rent (13 percent and 9 percent). There is also variation across household types. For example, households headed by a widowed man spend as much as 11percent on alcohol. Determinantsof PovertyLevelsinthe Slums 2.137 To explore more closely the correlates o f poverty in the informal settlements, probity estimations are usedto identify predictors o f whether a household lives below the poverty line. These suggest that: 0 There are surprisingly few significant differences in poverty levels between households o f different ethnicities. Without controlling for individualhousehold characteristics such as age, education, duration and so on, only the Somali community is significantly poorer. However, once controlling for these, Somali households are significantly less likely to be poor. The differences across ethnic groups appear to arise mainly because o f the correlation with family size. 0 There is significant variation inpoverty levels across villages within the same informal settlement. The difference in the poverty rate between the poorest and wealthiest villages within Korogocho is more than 31 percentage points. 61 Controlling for individualhousehold characteristics, this gap grows to 36 percentage points. The village differences are smaller in Viwandani-1 1 percentage points between the wealthiest and poorest, which increases to 14 percentage points once controlling for individualhousehold characteristics. 2.13 8 When looking at individualhousehold characteristics associated with poverty, we see that: 0 Poverty i s inversely correlated with age and duration o f residence in the settlement, 0 Married joint households-where husband and wife jointly live in the informal settlement-are significantly worse off than households whose head is never married or split families (where the wife lives inthe rural area). 0 In all cases, the presence o f children o f all ages, but particularly teenagers, makes households more likely to be poor. More surprisingly, each additional adult increase poverty risk by more than 22 percentage points. 0 Higher levels of education reduce poverty risk by more than 23 percentage points for those with more than secondary, relative to those heads who never attended school. However, the effect o f some secondary is much smaller, only a reduction o f 7.5 percentage points. And there is no significant difference between those with some primary and those who have never attended school. 0 Lastly, the experience o f household shocks such as theft or robbery i s negatively correlated with poverty, which has the obvious explanation that wealthier households are more attractive targets. ChildrenandTeenageYouth inthe Slums 2.139 Primary enrollment levels in the slums are high with little variation across householdtypes. O n average, 93 percent o f children 5-14 years were enrolled inschool in 2006. Unlike most other places in Kenya, however, a substantial proportion o f these are not in public schools. According to the NISS 2004, about one-third o f children enrolled were attending a private or non-formal school. Although slightly more expensive interms o f fees, these schools offer benefits, such as not needing to buy a school uniform. Teachers frequently lack the official training o f public school teachers, most are locals working on contract basis. In informal interviews, parents comment on how private school teachers are frequently more motivated. 2.140 The share o f youth aged 15-19 inthe slums still enrolled i s highrelative to other areas o f Kenya, 73 percent. There i s substantial variation across family types. For example, only 56 percent o f these living with a never-married man are at school, compared to 78 percent o f youth ina `married-joint' family. It i s notable that 83 percent o f 15-19 year olds living with a widower and 74 percent living with a divorced woman are attending school, despite their high poverty levels. It appears that these households invest relatively more in the education o f their children. Finally, enrollment o f OVC (children without one or both parents) i s lower than non-OVCs, especially among 15-19 year olds, 66 percent versus 75 percent. 62 Figure 2-27: Enrollment by OVCs (By One or Both Parents) 100% .---- - 80% I 4 60% NotOVC 40% ~-ovc - 20% I 0% (6-14yrs) (15-19 yrs) Source: de Laat et a12008. Focus on Youth 2.141 Kenya's youth are living in a rapidly changing society, and in some respects they enjoy greater opportunities thantheir parents did. Indeedfor some, the recent period has been a boom -- unprecedented levels o f access to education, newjob opportunities in services, cell phones and so on. But for many, for varied reasons, this has not been the case. Youth also face serious challenges, including high rates o f unemployment and underemployment, HIV/AIDS and other health risks, gender discrimination and, for some, the risks o f violent conflict (see World Bank Youth Assessment 2006). 2.142 In the current political, social and economic climate of Kenya, concern for youth i s important. Without this and policies and programs to help address the problems that youth face, there i s a risk the social fabric will fray and undermine the country's development prospects. Thus the well-being and prospects for youth have emerged as a central challenge for Kenya. While there have been some encouraging trends-in terms o f skills, for example-the youth unemployment rate i s about double that o f adults and joblessness i s very high. 2.143 There are more youth (inabsolute and relative terms) than ever before inKenya, signaling a major demographic shift in the labor force, which particularly affects the prospects o f young people as well as having broader implications. Almost one third Kenyans (30 percent) i s between 15 and 29 years, and the total number reaches almost 11 million people (compared to 8.5 million in 1999). The current generation o f Kenyan youth has growing levels o f human capital and exposure to the world, and thus different career expectations and potential than their predecessors; yet job creation has not kept pace in quantitative or qualitative terms. The Country Social Assessment found that youth unemployment, especially among males, is a major contributor to fmstration and tension in urban areas. Youth represent a politically as well as demographically significant group that will have a key role in shaping the country's development path and democratic transition. A young workforce has the potential to accelerate productivity growth and spur better governance, but without adequate opportunity to contribute 63 productively, youth can turn to risky behavior with possible implications for the incidence of crime, drug use, and HIV/AIDS, for example (World Bank 2005d). 2.144 Becoming economically independent is one of the hallmarks of the transition to adulthood. Indeed, finding a decent job offers the surest hope for millions of young Kenyans to lift themselves out of poverty, and to break the cycle of intergenerational transmission of poverty. Hence in this section we focus on youth in the labour market. We begin with a recap of levels and trends on the human development side, before reviewing the picture with respect to activity, jobs and joblessness in 2005/6. It seems valuable to categorize the youth-by age, location, level of educationhocio-economic background-to be able to unbundlethe issues and work out where to focus. Getting a handle on trends and prospects facing youth, and differentiation among youth, has important policy andprogram implications. The latter are addressedinPart IV. 2.145 The analysis draws heavily on quantitative data that is available for 2005/6. Several indicators are used to cast light on different aspects o f the school to work transition (WDR 2006). We look at education participation and the incidence of child labor, as well as health related indicators. Labor force participation and unemployment rates are examples of indicators that capture the ease/difficulty facing young people inthe transition. The quality of employment (formal vs. informal, family work vs. paid work, etc.), earnings and returns to education are examples of indicators to measure the success of the transition. O f course, these different aspects are closely related. But analyzing these different indicators helps clarify the obstacles in the transition and could provide guidanceto identify the issues and develop appropriate policy responses. HumanDevelopment 2.146 The 2003 DHS shows that youth, on average, enjoy the best health of any age group, as might be expected. However, high fertility rates and reproductive health challenges put young women at risk, and young people are disproportionately likely to contract HIV/AIDS. Fertility is high among 15-24 year olds, with about 25 percent of 15-19 year old women already having a child or being pregnant (the likelihood of this is the same in rural and urban areas). Given that international evidence shows that 15-19 year old girls have double the risk of maternal mortality that 20-24 year olds face, this poses a major threat to young women (ICRW 2005). At the same time, only 40 percent of demand for family planning methods is satisfied among 15-19 year olds is satisfied, and 49 percent among 20-24 year olds, relative to 60 to 75 percent for other age groups (national figures, for which age breakdowns are not available show that in urban areas, overall met demandi s 74 percent, versus 60 percent inrural areas). 2.147 Fertility is considerably higher in rural than urban areas -for every age group, but most evident for 20-24 year olds. Indeed, their fertility rate is almost double the corresponding urban rate, which is striking, especially given the similar levels for 15-19 year olds. 64 Figure 2-28: Age-Specific Fertility Rates, 2003 +Urban -+Rural -+Total 300 I 250 200 150 100 50 I 0 IS19 20-24 25-29 30-34 35-39 4044 45-49 Source: DHS. 2.148 Many Kenyan youth-two out of five-are infull time education. Almost 60 percent of teenagers 15 and over (54 percent for females), and about 18 percent of all 20- 24 year olds are full-time students. 2.149 The skills profile of the labor force does not suggest significant advantages for youth, as yet. Certainly, as shown in Figure 2-29, the literacy rate among male and female youth and primary education attainment are higher than those of adults; but the latter cohort maintains higher equal attainment in secondary and university education among males and quite similar attainments among females. However, the highest current (2005/6) ratio of full-time students to total people in cohort among both, the youth and children, compared to that in 1998 indicates an expected increased in attainment among the youth, closing the current gap with adults aged over 30. The share of youth who are full time students rose from 24 to 31 percent between 1998 and 2005/6. Similarly, the share of children (5-14) in full time education 74.5 percent in 2005/6 compared with 70 percent in 1998. Figure 2-29: Skills Profiles, by Age Cohort (Youth vs. Adults) Males and Females, 2005 70 i I 60 60 p E 50 a 2 50 E .a a .I B 40 4e 40 i c 30 ! P 30 LI * 2o 20 10 P10 0 I Illiterate Prifiary Secondary University 0 I ~ j Illiterate prirrary Secondary University Source: Staffcalculations based on KIHBS '05. 65 2.150 At the same time, very few youth are entering the labour market now without any education. Kenya's forthcoming ICA finds that the importance o f skills constraints to firms has fallen significantly: only about 8 percent of manufacturing firms report the shortage o f skills as amajor or severe impedimentto growth and between2003 and2007, the share has fallen significantly. Youth haveDistinctLabor Market Participationand EmploymentPatterns 2.151 Labour market rates vary by age and gender (Figure 2-30), although there have not beenvery significant changes since 1998. Activity rates decreased somewhat across the board, more so for young women.39 Unemployment rates remained roughly constant for males but fell for all female cohorts, young and adult, mainly due to reduced participation. Figure 2-30: Key Labor Market Indicators for Different Age Groups, by Gender (percent) 100 I 100 1 I 20" +Activityage 15-29 90 9 0 - 80 I 1 80 - -A- Activityage 30-64 70 0"5c.-e 70 - 60 I11 -+-Activityage30-64 '- 3 Employmentage 50 1 .--Employmentage 1629 +---- . _ _ _ % -e- Employmentage 40 +- Employmentage 2e 3064 Y. 2 IC 40 - 30 +Unemployment age 3 0 - -M- Unemploymentage 2 0 - p" 20 i 1 -e+ Unemploymentage 10 I G . Unemploymentage 0' 10 7 3064 --i 1998 2005 L Source: Staff calculationsbasedon ILFS 1998 & KIHJ3S2005/6. 2.152 While the overall employment rates are hugely different between adults and youth (88 versus 55 percent), the type of employment does not differs markedly by age. About the same percentage of youth and adults are in paid employment. Not surprisingly, few youth (in fact a negligible number) are business owners, and a larger share than adults are apprentices. Both adult and young women are more likely to be working on their own account, and muchless likely to be inpaid employment. Table 2-14: EmploymentStatus by Age, 2005/6 Age 15-29 Age 30-64 Male Female Total Male Female Total Employed(share ofcohort) 46.2 41.4 43.8 87.1 79.0 83.0 As Share o ftotal employed: Paidemployment 40.7 26.4 34.3 42.0 19.2 31.4 Business Owner 1.o 1.1 1.1 3.2 1.5 2.4 OwnAccount Worker / UnpaidFamilyWorker 57.2 71.6 63.7 53.9 78.5 65.4 Apprentice 0.2 0.3 0.3 0.1 0.0 0.0 Other 0.8 0.6 0.7 0.8 0.8 0.8 missing 3.9 20.6 11.4 4.9 15.6 9.9 Source: Staff calculationsbasedon KIHBS2005. 39However this may be partly due to survey differences: some of those classed as employed "for family gain" in 1998 may havebeen(more correctly) classed as inactive in2005/6. 66 2.153 At the same time, examining the labour market using such crude indicators as participation and unemployment has limitations, especially in low income countries. This is even more so for youth because their labor market attachment is weaker and schooling options are larger. Hence in addition to unemployment, we estimate ajobless rate (the share of young people not employed and not in school), as a complementary measure that could be of interest in debates about Kenya's poverty and inequality challenges. Youth Unemploymentis high, and Joblessness is even Higher, Especiallyfor Males 2.154 Consistent withKenya's economic growth recovery since 2003, as noted above, the aggregate unemployment rate had fallen in recent years-as did the rate for youth. However the situation still presents very major challenges. Most obviously, the overall unemployment rates for youth remain high, far above average, at about 21 percent. O f additional concern is the sizeable share of youth-about 25 percent-whom are neither in school nor work. Indeedit i s notable that the number inactive and not at school exceeds that inpaid work. The proportion is lower for males (9 percent), possibly because some part of the 18 percent of females recorded as "inactive, non full time students" are engagedinchild-rearing. 2.155 Figure 2-31 illustrates the divergence between youth and adult rates of unemployment, by province. The worst rates overall are inNorth-Eastern, and the lowest adult rates inNyanza. We look at the differences interm of ratios to adults, and include a measure of joblessness (neither in school nor work). Overall, Kenya's youth unemployment rate is about double that for those aged 30-64 years, though the differential varies by province Table 2-15. Youth fare the worst in comparison to adults inWesternandNyanza, where the youth unemployment rate is about triple the adult rate; North Easternprovince has the smallest ratio (50 percent higher). Interestingly however, the picture worsens for Central if we look at inactivity as well as unemployment, while Western andNyanza remain the worst. Figure 2-31: UnemploymentRate, Youth versus Adult, by Province(percent of active in cohort) 6o 1 Nonh Nairobi Coastal kRValley Western Central Eartern Nyanra Eastern I-Youth 15-29 enAduh30-64 -Youth NationalAverage +Aduh NationalA v e w 1 Source: Staff calculations based on KIHBS 05/06. 67 2.156 Thus, looking at joblessness-Le. the percentage o f young people who do not have a job, excluding students-suggests that the magnitude of the labour market problems i s large. Ifwe aggregate the rates o f reported unemployment and inactivity, 38 percent o f youth-almost two out of five young Kenyans, are neither in school nor work. The rates o f youth joblessness relative to adult rates are presented in the right hand column of the table below. Table 2-15: Ratio of Youth to Adult Unemployment Rate by Province and Gender,2005/06 Provinces Unemployment Unemploymentplus Inactivity Male Female Total Male Female Total Nairobi 3-47 2.11 2.70 3.48 1.74 2.38 Central 2.06 1.98 2.01 3-90 4.17 4.05 Coastal 1.72 1.76 1.73 3.60 1.92 2.47 Eastern 1.75 2.03 1.88 4.00 3.23 3.53 NorthEastern 1.38 1.65 1.49 2.32 1.06 1.41 Nyanza 2.35 3.80 3.15 6.32 4.94 5.46 Rift Valley 2.01 1.72 1.83 4.44 2.80 3-42 Western 2.13 4.22 3.01 5.01 4.29 4.60 All Provinces 1.98 2.18 2.09 4.16 2.80 3.30 Source: Staff calculations basedon KIHBS 05/06. 2.157 Young women have a higher likelihood of unemployment than their male counterparts. This presents somewhat of a puzzle since inmost African countries, male youth unemployment rates tend to be higher (Denu, Tekeste, and van der Deijl2005). 2.158 Table 2-16 compares youth time use andjoblessness in Kenya to other African countries; in general, Kenyan youth seem to be in the middle o f the spectrum on the various indicators. However, the share of youth neither in school nor employed in 2005/06 was much higher than in Uganda and Ethiopia, but lower than in several other countries, including for example Mozambique andZambia. Table 2-16: Youth Time use in Urban Areas, Selected African Countries (percent of Youth) Only in Only Jobless School Working Unemployed Inactive (unemployed + inactive) BurkinaFaso 34.8 36.6 14.5 13.2 27.8 Burundi 58.6 14.8 11.6 15.0 26.6 Cameroon 48.3 20.1 18.5 11.2 29.7 C6te d'Ivoire 34.8 28.3 14.4 22.6 37.0 Ethiopia 46.1 26.4 13.5 14.1 27.6 The Gambia 37.0 16.9 2.3 43.3 45.5 Kenya 40.6 34.8 23.8 13.6 36.5 Madagascar 39.6 37.1 16.1 5.4 21.4 Malawi 55.1 14.8 2.8 26.4 29.1 Mozambique 29.8 20.2 36.1 12.8 48.9 Uganda 40.6 31.7 2.3 21.9 24.2 Zambia 35.8 16.6 19.8 27.4 47.3 Source: World Bank (2006b), and KIHBS2005/6. 68 2.159 The extent of unemployment and inactivity among youth prompts questions about the appropriate diagnosis of the problem. Ifunemployment is concentrated among the better-off who are supported by their families and can afford to allow time searching for an attractive job, it may not be a pressing issue from a poverty reduction perspective. On the other hand, if unemployment i s high among the very poor, and if inactivity actually reflects disguised unemployment (Le. discouraged workers who have given up searching for a job or cannot afford to do so), it has different policy implications. Another possibility i s that the jobless who say that they are available to work but are not actively searching for a job could be engaged in some productive activities, such as helping with a family business. 2.160 We looked at the poverty status of households with unemployed youth. A simple cross-tabulation shows the poverty headcount in all households in comparison to those which have youth and unemployed youth in the household. Households with at least one youth member are poorer than the national average. There is an additional effect seen for households with at least one unemployed youth-with the poverty incidence risingto 55 percent. Table 2-17: RelativePoverty Incidencein Householdswith Unemployed/Inactive Youth, 2005/6 (percent) All households 46.7 All households with at least one youth 49.5 Households with at least one unemployed youth 54.9 Households with at least one unemployed or `other inactive' youth 51.6 Urbanhouseholds with at least one unemployed or `other inactive' youth 36.5 Source; Staff calculations based on KIHBS, 2005/6. 2.161 At the same time, the crude correlations betweeneducation andunemployment, and employment, suggests that the unemployed youth indeed do have higher average levels of schooling: about 40 percent had completed secondary school, compared to 26 percent of the employed. The difference i s most marked for unemployed female youth. This could imply that secondary educated youth are waiting for jobs they perceive as commensurate with their skills and aspirations. Table 2-18: Level of EducationCompleted(percentage of cohort), 2005/6 Unemployed Employed Male Female Total Male Female Total None 0.0 0.0 0.0 0.1 0.2 0.1 Primary 63.6 57.5 60.1 72.7 74.0 73.3 Secondary 36.0 42.1 39.5 26.9 25.3 26.1 University 0.4 0.4 0.4 0.3 0.6 0.4 Post graduate 0.0 0.0 0.0 0.0 0.0 0.0 Source: Staff Calculations based on KIHBS `05. 69 Box 2-5 : Youth Unemployment-Selected International Experience More than 18 million young people in Sub-Saharan Africa-all youth (15-24)-were unemployed in2003 (Garcia and Fares, 2008). This rate o f youth unemployment exceeds that o f all regions inthe World, except the MiddleEast andNorthAfrica. The unemployment rate for youth exceeds that of adults in all countries except Burundiand The Gambia. Young workers are particularly disadvantaged in Cameroon and SBo Tom6 and Principe, where the youth unemployment rate is more than five times that o f adults. The situation on the continent suggests that even with fast improvements in average educational attainment supply-side measures alone are insufficient to facilitate adequatejob creation, particularly inthe face o f major demographic pressures. International experience suggests that youth unemployment i s a pervasive problem, and even those youth who do acquire formal education frequently experience great difficulty when entering the labor market. The youth unemployment rate indeveloping countries tends to be substantially higher than that o f adults. Based on data from 60 developing countries around the world, Fares, Montenegro, and Orazem (2006) found that the typical youth spends about 1.4 years inunemployment or temporary jobs before settling into stable work. And employment alone does not necessarily signal a positive outcome, since many youth are unpaid (for example a family business) or low-paying jobs with limited prospects, and frequently in unhealthy or unsafe conditions (Denu, Tekeste, and van der Deijl 2005). In Sub-Saharan Africa the total estimated transition duration for school leavers ranges from one to seven years. Cross-country analysis shows that the youth unemployment rate tends to increase as the share o f youth in the labor force increases (Fares, Montenegro, and Orazem 2006). Downturns in overall economic demand also affect youth prospects. A deterioration in adult employment prospects tends to be associated with disproportionately worse outcomes for young people-and again, this particularly affects those with the least education (World Bank 2006a). 2.162 The situations in Ethiopia and Tanzania provide contrasts. About 40 percent of young adults in Dar es Salaam are jobless (Kondylis and Manacorda 2006), outstripping the rate in Addis Ababa (34 percent)-but the adult male unemployment rate in Dar es Salaam i s negligible, yet high inAddis. This suggests that while youth unemployment in some countries, including Tanzania, may be a transitory phenomenon, in Ethiopia it i s a structural issue that disproportionately impacts youth. Figure 2-32: Urban Youth UnemploymentRatesin Selected African Countries(Year Varies) I 9 0 1 I 80 70 60 $ 30 20 10 0 Source: Leibbrandtand Mlatsheni (2004), except Ethiopia, UBEUS 2004, own calculations; "youth" defined as ages 15-24. 70 YouthEarn much less thanAdults 2.163 Youth everywhere generally earn lower wages than adults, which may be explained by their limited work experience: in Kenya, monthly youth wages, for teenagers, averaged about 20 percent of that o f adults. However the ratio increases significantly in favor of youth for 20-24 years olds, to about 50 percentm4' This also helpsexplain why most youth continue to live with their families. 2.164 There are some interesting distinctions around this average, inparticular: 0 The gap is muchlarger inrural areas-where youth earn 15 and 41 percent of adults for teenagers and 20-24 year old respectively-compared to urban areas, where the corresponding figures are 32 and 56 percent. Well educated youth tend to fare much better, especially for women, and university educatedyouth. For teenagers, the gaps are muchlarger for males, relative to females, for most groups. On the other hand, some 20-24 year old men do appear to be doing relatively well, includingthose working on their own account and apprentices, relative to adult males inthe same types o fjobs. Table 2-19: Youth Earnings as Percentageof Average Earnings,Non Agricultural, by Sex 15-19 years 20 -24 years Male Female Total Male Female Total Youth 20.1 24.4 20.8 44.3 59.8 48.6 -InPaidEmployment 19.9 21.8 20.0 31.6 43.0 35.2 -Ownaccountworker I business owner 22.2 40.5 28.2 92.0 111.1 93.8 -Apprentice/ UnpaidFamily 14.2 18.3 14.5 99.5 42.9 61.1 -Other 17.2 26.0 19.2 32.9 30.9 26.2 Urban 32.7 37.0 31.9 47.0 77.3 56.3 Rural 14.7 15.5 14.6 41.7 41.2 40.9 NoEducation -- _ _ _ _ 18.5 9.9 12.1 Primary 21.9 22.2 21.0 41.5 47.9 42.8 Secondary 26.5 36.0 28.7 46.7 77.1 56.1 University _ _ _ _ -- 31.9 259.1 143.6 Source: StaffCalculations,basedon KIHBS '05. Youth Employmentinthe Slums4' 2.165 Most residents in Nairobi's slums are engaged in economic activities. Among adults aged 30 and above, 95 percent of men and 73 percent of women were reported to have engaged in some income generating activity in the month prior to the survey. The figures for youth are lower, especially among females, 66 and 39 percent respectively for 40Because these wages are not adjusted for time worked, they do not allow for an exact comparison o f hourlywages. 41The above mentionedNISS 2004 survey collecteddetailed informationon employment. Recallthat this is a randomsample inNairobi's slums that excludeshouseholdswhose headshave nevermarried. 71 males and females. About 9 percent of male youth are neither still inschool nor engaged inanincome activity comparedto 43 percentof female youth. Figure 2-33: Youth (15-29 years) Activities in Nairobi Slums, 2004 (NISS 2004) Jobless Income activity e3 Inschool Male Female 2.166 Besides formal schooling, a common source of training inthe slums i s informal vocational training, or apprenticeships. Among youth, 37 percent of men and 27 percent of women have received at least 3 months of this kind of informal vocational training. Higher formal education i s associated with increased likelihood of informal vocational training. 2.167 More education i s associated with better employment prospects. O f those with less than a primary education, 35 percent were engaged in an employment activity the previous month compared with 50 percent and 58 percent, respectively, for those with completed primary and more than primary education. Interestingly, informal vocational training i s associated with even greater increases in employment. Among women of all formal education levels, only 31 percent of those without informal vocational training were employed, compared with 54 percent of those with informal vocational training. The gain among menis similar, 54 percentversus 79 percent, respectively. 2.168 Among men, more than a fifth work as casual/factory workers, which i s generally unskilled without benefits or security. More lucrative is skilled vocational work by mechanics, masons, welders, carpenters, and electrician, and one in six is employed in this category. The highest earnings go to drivers more than double that of casual workers. Security guards also earn relatively high salaries, but this must compensate for long work shifts (generally 12 hours), only one day off per week off, frequent night shifts, and increased risk. More than 1 in 11 men, aged over 30 living in the slums, works as a security guard, reflecting the generally insecure situation in Nairobi. 72 Table 2-20: Top Ten Jobs for Male Youth, 2004 Monthly Earnings Avg Years Male Youth: 15-29 years YOof total (Ksh) Education 1 casuaVfactory worker 22.5 4,869.1 9.6 2 mechanic/mason/welder/carpenter/electrician 15.3 5,655.5 9.1 3 businessperson 10.5 4,656.1 9.2 4 selling (hawker, kiosk etc) 8.4 4,827.8 8.9 5 machineoperator 5.3 6,797.6 10.1 6 jua kali 4.7 3,454.2 8.1 7 tailor 3.4 5,373.4 9.7 8 security guard 3.3 6,185.6 9.6 9 driver 3.1 10,411.8 10.0 10 constructionworker 2.1 3,055.6 8.2 Total employment share 78.64 Source: de Laat et a1(2008). 2.169 Female youth earn much less than male youth-although the difference is smallest in casual/factory work-and are generally employed in very different jobs apart from retail and casual work. Note that the two highest paying jobs are being a teacher (and women working in this profession have the highest years o f education, 12 years on average) and brewinghelling alcohol (which i s done by women with the lowest average years o f education). Table 2-21: Top Ten Jobs for FemaleYouth in Slums Monthly Earnings Avg Years Female Youth: 15-29 years O hof total (Ksh) Education 1 business person 20.3 3,443.7 9.1 2 selling (hawker, kiosk etc) 19.8 2,870.9 8.7 3 casualworker 14.5 4,303.9 9.5 4 hair dresser 8.2 3,922.9 9.0 5 tailor 7.7 2,709.6 9.2 6 helps family (business) 4.0 1,437.5 7.0 7 house maid 3.4 1,261.5 6.2 8 hoteVrestaurantworker 3.2 3,204.5 7.7 9 teacher 2.6 5,300.0 11.8 10 brewinghellingalcohol 1.9 6,183.3 5.8 Total employment share 85.47 Source; de Laat et a1(2008). 2.170 The government i s increasingly focusing on the needs o f the growing youth cohort, evidenced for example by the new "Marshall Plan" for youth to ensure that issues affecting young people are approached in a coordinated, multi-sectoral manner. Alongside broad-based economic growth, which offers the most important route to sustainable job creation, other interventions can improve employment opportunities. 73 These include measures to enhance access to education, credit, and information, as well as second-chance programs like basic skills and literacy training. 2.171 Empowering young women will also be central-although the gender gap i s shrinking, young women are still less likely than young men to be employed (and also less likely to be in school). Job quality also tends to be very low; about 35 percent of employed young women are engaged as domestic workers. Another 20 percent are own account workers. The two groups combine for a total o f over 1.3 million women. Increasing opportunities for women by raising their human capital and by ensuring a favorable policy environment will have major payoffs for growth andpoverty reduction. 2.172 From a policy perspective, among the highlights that have emerged from this analysis of youth are the following: Itis important to separate at least by five year cohort, and gender. Unemployment is especially high for teenage women. While far fewer youth are in employment, compared to adult employment rates, those working have a broadly similar pattern in terms of types of employment. 0 For the younger group ofthose working innon-agricultural activities, earnings are way below average-and very low also relative to poverty line-even for those with secondary education. This is consistent with returns to experience, but could be associatedwith frustration anddrop out. The vast majority of youth appear to be living at home-very few are household heads and few reported having kids. 2.173 In sum, youth are a large group with some distinct challenges as well as advantages. We returnto this theme inPart IV below. IMPACTS OF THE POLITICAL CRISIS ON POVERTY AND INEQUALITY 2.174 All of the analysis so far has been based on data which preceded the crisis in early 2008. This section attempts to map out what might have been the effects of the crisis, although there are no post-crisis household survey data available to measure these effects ex post. 2.175 It is useful to think about the poverty impacts of the conflict in a systematic way. International experience suggests that conflict reallocates public resources away from basic services, and leads to the loss of income opportunities and livelihoods, all of which tend to increase poverty. Cross-country econometric analyzes suggest that the income and poverty effects are large-that, relative to the absence of war, incomes are 15 percent lower and absolute poverty 30 percent higher (Collier, 2006). In the case of Kenya, the duration o f the conflict is obviously far less than the typical civil war (which averages seven years), but the short term disruptions were large, and there is some risk of mediumandlonger term effects associatedwith hysterics andimpacts onconfidence. 74 2.176 A typology o f effects, building on Stewart (1997), can help construct a more integrated approach. This includes direct, short term as well as mediumterm dimensions, across several aspects o f well-being, viz. access to basic services, income earning opportunities and empowerment. The crisis has had direct impacts on several dimensions that are relevant to poverty and well being. In the short term, public services in some areas have been disrupted; markets have responded through, among other things, layoffs and increased prices; and civil and social capital and networks have been adversely affected. These have affected the poor, and also caused many families who were not previously poor to fall below subsistence levels o f consumption. There i s also the impact o f rapid price increases that has been observed since January and these have been large and they have had immediate effects on the welfare o f Kenyan's as further discussed below. Box 2-6 provides a summary. 2.177 There are several points and caveats to underline at this point: 0 For many o f the direct, as well as indirect impacts, there will be disparities betweenwar-affected and other regions, and among ethnic groups. 0 The magnitude o f the medium and longer term effects is largely unknown at this point, and it seems appropriate to develop two scenarios--one based on an early cessation to unrest, and the other, where some uncertainty persists. 0 Given the lack o f capacity o f the poor and near-poor to smooth consumption over time, there i s a real risk that even short term effects could have longer term negative repercussions--e.g, ifthe household's only asset is sold. 0 It is well-established in the literature that short term shocks can also have irreversible effects, including for human capital formation (e.g. if kids are pulled out o f school, babies not immunized, HIV-positive people go off their ARVs). Finally, some effects may not be linear. Some may dissipate, whereas the ill- effects o f others compound over time-e.g. the effect o f rent-seeking and patronage on economic efficiency. Box 2-6: The Impact of Conflict on Poverty in Kenya Public Services Markets Civil / Social New Forms of Inequality Direct Impacts Reducedprovision ' Destructionof Destructionof Directappropriation of and access to private capital social of assets, land, basic services (shops and capital sources of factories) and (institutions, livelihood(business Lack o f accessto localmarkets values, premises) ARVs Redundancies networks) due to reduced through violence Breakdowno f economic and population Mass displacement - public activity; displacement up to 350,000 IDPs order +increased estimated crime and insecurity 400,000 (KAM) Displacementof 350,000 affected; 150,000 of whom are 75 Public Services Markets Civil / Social New Forms of working age Sharpprice increases for essentialgoods Medium/ Macro Reducedrevenue Stagnantor Exclusion and Rent-seekingby longterm due to disrupted falling economic heightened those with access to impacts economic activity activity competition at patronage andrisks and reducedstate Agricultural local levels legitimacy output affected by disruption to family and Crowding out of planting season community links development severed, social spendingdue to Depreciation capital damaged reducedrevenue -loosensthe and increased Capital flight constraints on security spending opportunistic and Unfavorable criminal business behavior, climate, drop in weakens long term support networks investments for small businesses Household Diminished Inflation Local and access to public Unemployment. communities community services: weakened or level health, education, Reducedfarm destroyed policing etc output Existing Increasedgender safety nets and Violence coping mechanisms insufficient or breakdown. 2.178 Inthe absence of data post-dating the conflict, we can use the 2005/06 KIHBS to present some first attempts at quantification for those short term aspects amenable to monetization. At this point, the approach i s very narrow: viz the loss o f jobs due to redundancy and displacement, and the price shocks which have increased the cost o f living. Moreover, we only estimate aggregate impacts, abstracting from regional and group differences in effects, which are likely to be large. These estimates are analogous to micro-simulations, and totally abstract from second round and equilibrium effects and adjustments at the household level. We also totally exclude any nominal adjustments on the wage or income side, which will offset the price inflation to at least some extent. 0 First, looking at the increase in unemployment, even if temporary, the KAM estimates of layoffs and the extent o f internal displacement, suggest that unemployment at the individual level rose by the equivalent o f 25 percent. This means individual unemployment after the crisis o f 14.6 percent and the fraction o f households with at least 1 unemployed member will then be 26 percent. The initial (pre crisis) incidence o f poverty in these households with at least 1 unemployed member i s 54.2 percent and in the other households', poverty incidence i s 44 percent. Given the poverty incidence in households with at least 76 one unemployed member and the increase inthe share o f households with at least 1unemployedmember, there would be a rise inpoverty o f 9 percent. Second, we can estimate the price impacts on poverty. As is reported by the KNBS prices have been rising rapidly since January and by May, the CPI on a year on year basis has increased by some 30 percent. While there are some technical problems with the CPI and the methodology suffers from upward drift, especially when there i s volatility in the series, it is nevertheless the case that inflation has accelerated since January. Food and non-alcoholic drinks make up little over half o f the index and it is clear that price inflation o f these items, as well as others, will have a direct and immediate impact on poverty. It also appears that the relative prices o f both food andtransport have risen, which would tend to adversely affect lower income groups. Table 2-22: One Month and Twelve Months Change in Price Indices Broad Item Group YO YOChangeon Previous YOChangeon Previous month (May O8/Apr 08 ) year (May O8/May 07) Weights Food& Non-alcoholic drink 50.50 4.6 44.2 Alcohol & Tobacco 2.97 -0.2 13.0 Clothing & Footwear 9.00 1.1 5.2 Housing Costs 11-74 0.6 6.5 Fuel& Power 4.18 3.1 17.9 Household Goods & Services 5.82 1.4 8.4 Medical Goods and Services 1.59 1.o 8.3 Transport & Communication 5.75 0.7 20.3 Recreation& Education 6.02 1.1 6.8 Personal Goods 2.45 1.2 6.9 Average all Groups 100.00 3.5 31.5 Source KNBS. 2.179 According to the KIHBS, food represents about 70 percent o f households spending o f the poor, and 77 percent for the bottom quintile. It is o f course relevant whether or not the poor are net sellers or consumers o f food, in understanding the distributional impacts o f food price increases. We can look at the case o f grain for several years, using Tegemeo data (Nyoro 2008). This shows that the poor and the bottom decile in particular are much more likely to be net purchasers o f food-only 25 percent o f the poor said that they sell maize most years, compared to 50 percent o f the non-poor. Likewise, the lower income groups relied on purchased staples for a larger parto fthe year in2007 (Figure 2-34). 77 Figure 2-34: Average Number of Months HouseholdRelied on PurchasedStaples, 2007 3.5 3 2.5 2 1.5 1 0.5 0 Non-Poor Poor Top Quintile Bottom Quintile Source: Nyoro et a1(2008). 2.180 Given this situation and the fact that there are technical problems related to the CPI, meaning that the reported CPI overstates inflation, we run some simulations assuming certain rates o f inflation. If,for example, the cost o f living has increased by say 20 percent, all those families who were living within 20 percent o f the poverty line would fall into poverty. We estimate that the poverty headcount would have risen by 22 percent. The depth o f poverty also worsens for the already poor-by about 35 percent. Poverty severity gives greater weight to those farthest below the line, and shows an even more significant increase. The impacts corresponding to a smaller price shock are also estimated-with a 10 percent increase in prices, the effects on the various poverty measures are about half as large. Of course, this totally abstracts from any adjustments in nominal incomes or second round effects which may offset or worsen direct impacts. These simple extrapolations do not include any behavioral changes or dynamic effects. Table 2-23: ImpactsofPrice Shockson Poverty Measures Initial status Price shock of Percentage Price shock of Percentage 10 Yo increaseswith 20% increases with 10% shock 20% shock Incidence 46.7 51.5 10.3 56.9 21.8 Depth 0.166 0.196 18.1 0.224 34.9 Severity 0.085 0.097 20.7 0.117 37.6 2.18 1 Another way o f illustrating the situation is to view the cumulative density for the basic per capita expenditure inKIHBS. It also shows two distribution neutral shifts o f this per capita expenditure: one where every household has 20 percent lower consumption and one where every household has 20 percent higher consumption. Under the increased consumption, the poverty rate would be 35.4 percent and under the decreased consumption the poverty rate would be 59.6 percent. 78 Figure 2-35: Changes Cumulative DensityFunctionof ConsumptionPer Capita i0,41 - -PCE 0.6 -PCE increasedby 20% PCE decreased by 20% 0.2 0 I I , , 2.182 The foregoing, while overly simplistic, suggests that the short term impacts on poverty could be very large, both for those who were already poor, as well as large numbersof "new poor." Ifwe simply addthe effects on poverty attributable to increased unemployment and increased prices, the rise is of the order of 30 percent. This would effectively erode the gains of growth over the preceding period, and obviously put many households into a crisis situation due to their loss of livelihoods and opportunities, and difficulty insustaining themselves at a minimumthreshold of subsistence. 2.183 Putting these figures into absolute numbers, the total increases simply from these two effects, and abstracting entirely from the loss of life, assets and livelihoods, and worsening hardship among the already poor, could be of the order of five million Kenyans falling into poverty. Of coursethese effects will have beendistributedunevenly across differentdistricts and province, as well as across groups. 2.184 We now turnto explaining drivers of change over time. 79 3. DRIVERSOFOBSERVEDTRENDS 3.1 So far, this assessment described the levels and profile o f disadvantage in Kenya. The analysis emphasized the importance o f the location where people live in explaining observed outcomes, as well as the systematic disadvantage that faces major groups. Certain groups, including, for example, urbanwidows and pastoralists, seem to have been largely excluded from sharing in the fruits o f development in Kenya. It appears that many o f these factors have persisted over time, and some may even have worsened. 3.2 Like in many African economies, another important issue in Kenya is the frequency and magnitude o f shocks, due to weather (e.g. drought), health (e.g. catastrophic illness or death in the family) and other major events. We examine the profile o f vulnerability to shocks, household coping mechanisms and, using panel data, interactions with households' prospects. 3.3 This part attempts to explain trends inpoverty and inequality inKenya over the past decade. We are especially interested in income earning opportunities, as the labour market i s the primary mechanism by which growth inthe economy is shared. Inequality in both opportunities and outcomes (see WDR 2006) plays a significant role. How did initial conditions affect trends in poverty and inequality? And how did trends in the distribution o f private assets, and access to public services, like education, health and water, affect observed outcomes? 3.4 Inthis part, first, some insights are drawn from qualitative surveys. Secondly, we present results o f analysis which decomposes the contributions o f changes in the average and distribution o f consumption to aggregate trends in poverty over the period 1997-2005/6. Third, we review trends in the level and structure o f employment and earnings. Fourth, we will look at the distribution o fprivate assets, including the access to land, capital (credit) and skills (education). Fifthly,we will look more closely at trends in rural sector, drawing inparticular on the results from the Tegemeo panel survey. Sixthly, we will focus on shocks and how this affects households' potential to move out o f poverty. The seventh section looks in detail at the distribution o f public expenditure including schools, hospitals and roads. The final section reviews evidence on how corruption affected poverty and inequality in Kenya. This leads to the final part o f the report which discusses the implications for policy. 3.5 The best source o f data for understanding trends over time comes from panel data, that is, surveys which track the same people and households over time. Through partnership with the Tegemeo Institute and the APHRC this type o f data was made available. These cast significant light on drivers in rural areas in the case o f Tegemeo's panel, and large urbanslums inthe case o f APHRC. We also use national householddata 80 from the late 1990s and 2005/6, which provide nationally representative cross-sections. Now, some qualitative finding will be presented. QUALITATIVE FINDINGS 3.6 Ina country as complex and diverse as Kenya, it is very important to consider "the numbers" within a fuller sense of the context. We are able to access some useful recent surveys that indicate the extent o f dynamics and associated drivers as perceived by various groups o f Kenyans which are reviewed here to set the stage for a broader review o f trends and underlying factors. We begin with Tegemeo's rural analysis o f levels and trends in subjective measures, results from the survey commission in seven districts for the World Bank's Country Social Assessment (2007) and the PPA-IV collected by KNBS. 3.7 The Tegemeo sample collects qualitative information as well as comprehensive data on incomes and economic activities. Recall that this is a panel, so we can look at both measures for the same households over time. Several interesting points emerge from the analysis (Nyoro et a12008). 3.8 The subjective ratings tend to correlate closely with the monetary measures. Individual families were asked to rate their position relative to other households in the area-whether they are better off, worse off, about the same. The results show that the poor regard themselves as worse off. As poverty decreases the fraction that say they are poor decreases. And the non-poor are far more likely to say they are better off. Thus if we cross the family's rating with income, the people who say they are worse off have about a third o f the income o f people who say they are better off. This holds across the years 2000,2004 and 2007 (Figure 3-1). Figure 3-1: Comparisons of SubjectiveWell-Being, 2000-07: Tegemeo 80 70 60 50 40 0 better of 30 63worseoff , 20 Haboutthe 10 O N Non-Poor Poor Non-Poor Poor Non-Poor Poor 2000 2000 2004 2004 2007 2007 Source: Nyoro et a12008. 81 3.9 The poor were much less likely to report improvements over time, although interestingly, the gap between poor and non-poor closes over time. In 2000, the poor were more likely to say it has worsened, whereas in 2007 they switched to saying it has stayedthe same. Ifwe cross general well being inthe area with income, those who report being the same or worse off have very similar incomes. Those who report that conditions have improved have incomes about 20 percent above the average. This is consistent with income growth rates: those who report a worsening experienced the lowest growth rates (andintwo ofthe three periods, negative growthrates). 3.10 Tegemeo's findings thus suggest consistency between subjectiveh-elative and absolute measuresof well-being, andpoint to the value of further combined analysis. 3.11 An analysis of six rural districts in a World Bank Country Social Assessment (2007a) sought to understandthe major drivers o f rural welfare, coping responses of the communities, and associated social impacts. The major poverty drivers were identified as: declining natural resources (land fragmentation and degradation, and loss of assets such as livestock); health shocks, particularly the onset ofthe HIV AIDS epidemic; and institutional weaknesses-widespread corruption (including throughout the judicial system), deteriorating government services, inefficient marketing institutions, and weak local government capacity. 3.12 Households responded to these adverse developments by diversifying their economic activities off the farm and increasing reliance on community organizations. The CSA also found sharp increases in domestic violence associated with gender and inter-generational conflicts, and, in the wider society, increasing crime and ethnic tension. These major concerns are not fully explored inthis report. 3.13 The PPA-IVYdescribedabove, uses a "stages of progress'' approachto ascertain benchmarks from the poorest through to relative wealth-these begin with purchasing food, and range up through aspects of housing and education. There was general consensus in each village on what constituted the successive stages o f households' progress. Correlates of progress perceived by rural households include asset accumulation, livestock purchase, repairs to housing and water storage. Urban communities had slightly different perceptions of the stages of progress, tending to emphasize housing conditions, small-scale enterprises, and schooling, as well as asset accumulation. 3.14 The PPA-IV sought to identify the extent of, and reasons for transitions across the progress benchmarks. The data, though not representative of the population as a whole, offers insights, which can be reviewed interms of the principal livelihoodor main economic activity, inthe district (following Mango et a1 2007). Five broad groups were identified: (1) mixed farming (cash and food crop, livestock) inhigh potential areas; (2) mixed farming and fishing inlow potential areas; (3) mixed farming and livestock inlow potential areas; (4) pastoral districts; and (5) urbandistricts ofNairobi and Mombasa. 82 3.15 The sampled communities perceived a worsening inpoverty over the past eight years, and this i s most evident among pastoral communities. While these results should be treated with caution, we can learn a lot about the dynamics of poverty.42 Out of the group of householdsconsideredto be poor in2005/06, 70 percent were also poor in 1997, which shows significant persistence of poverty. Taking 1990 as the reference, 30 percent ofhouseholdswere consideredchronically poor (Mango et al, 2007). 3.16 The extent of chronic poverty varies by livelihood group, and is worst amongst marginal farmers, pastoralists and urban dwellers. For these groups, a significant minority o f households is trapped inpersistent poverty. The pastoralists inparticular are in a dismal situation: very few (only 3 percent) reported having escaped from poverty over the past eight years, while more than 20 percent fell into poverty (Table 3-1). Table3-1:PovertyTransitionsinPPA-IV Householdsby LivelihoodGroup 1997 200516 - (Percent of households) Percent Percent Percentage Livelihood Remained Escaped Became Remained poor at the poor at pointchange group poor poverty poor non-poor beginning the end inpoverty Mix farm, high 31 10 11 48 41 42 1 F d f i s h , low 24 7 14 55 31 38 7 Fadlivestock, low 43' 11 13 33 54 56 2 Pastoral 38 3 22 36 42 61 19 Urban 42 14 15 29 56 57 1 Average 35 9 14 42 44 50 5 Source: Mango, et a1(2007). 3.17 There i s some `churning' inthe PPA sample with householdsmoving inand out of poverty. Yet very few experiencedtwo poverty transitions over the past 15 years: only 2 percent of all householdsbecamepoor and then escaped poverty, and 2 percent escaped poverty andthenbecame poor (Mango et al, 2007). 3.18 PPA-IV investigated the causes of poverty transitions, interviewing 2,161 rural and 204 urban households, were prompted to recall the sequence ofevents associated with falling into or movingout ofpoverty, or staying poor. Starting with householdswho were persistently poor throughout the past 15 years (632 households), the broad factors that were typically identifiedwere: Limited employment opportunities: This cause of persistence o f poverty was reported most frequently, especially among the urban poor (73 percent of whom identified this as a key factor), and to a lesser extent for other livelihoodgroups. Only pastoralistsdid not see this as an important reasonfor persistence of poverty. 42 This is a qualitative survey conducted at one point in time, and for detailed poverty comparisons we needto usepanelsto uncover the deeper dynamics ofpoverty at the householdlevel, as we will below. 83 0 Health factors: Poor health conditions in general and death o f the main income earner in particular (including burial costs) were seen as important factors leading to persistent poverty, especially inhigh potential areas. Demographic and landpressures: All groups noted high dependency ratios in the household as a factor. Crop farming communities felt that land subdivision and exhaustion trapped households in poverty. This echoes the views of Narayan andNyamwaya (1996: p 3), who consider poverty to be an intergenerational phenomenon associated with land subdivision and soil erosion. Other unanticipated shocks: For pastoralist communities, drought was a major factor behind persistent poverty. Theft and loss of property was important in the recent histories o f households in the farminghshing livelihoodgroup (Box 3-1). Box 3-1: Joseph Mbauni Nderitu-Remained Poor due to land Fragmentation,Relianceon Casual Employmentand Ill-health Joseph was born inKarigu-ini village in 1959, one of six brothers. His parents could not afford to continue with his educationbeyond primary school. They had a small piece of land which they subdivided amongst the sons. Because hehas avery small pieceof land, he spends most ofhistime selling his labour. `Poverty is part andparcel of my life' he said. As a casual labourer, his status has always been uncertain. His wife is also a casual worker. In 1996, he fell from a building and broke his leg. He never received proper medication due to lack of money and since then his leg has been weak. He therefore cannot do hard physical work. In 1998 his wife gave birthto a third child who requires drugs and monthly clinic visits. H e has two children attending school and he benefits from free primary education. However, the children are a big burden. Since their land is very small, whatever little proceeds they get from harvest i s consumed within three months. They then resort to buying food from the market using earnings from local work. At the time of interview, the family's only livestock were three chickens. 3.19 Those householdswho hadbecomepoor emphasizedthe following (Table 3-2): 0 Health factors: As with the chronic poor, those who fell into poverty mentioned poor health conditions and death of the main income earner as causative events. Crop farming households inparticular pointed to this cause, which was less commonly cited by pastoralists. High dependency: This was emphasized by most groups, except pastoralists. Life cycle events, such as family formation and ageing, when associated with growing dependency, were seen to cause economic hardship andpoverty. Group speczjk factors: Pastoralistsrecalled drought, theft and other livestock problems as causal events. One third of urbanhouseholds emphasizedloss o f employment. 84 Table 3-2: Major Reasonsfor Falling into Poverty by Livelihood Group (YOHouseholds) Group Mix Fardfish Farmilivestock, Pastoral Urban Overall farm, low low high Many dependants 47 44 49 21 51 41 I11health 45 65 42 20 31 40 Death 44 73 25 23 9 37 Drought 0 0 21 67 0 24 Unexpected loss 0 11 24 57 0 24 Land 37 11 29 0 0 20 Livestock related 0 0 19 40 0 17 Loss of employment 12 8 12 0 34 16 Crop related losses 12 0 0 21 0 11 Alcoholisddrug 19 0 14 0 0 11 Marriage expenses 10 0 13 11 0 9 Lack of inheritance 10 0 0 0 0 6 N 168 80 136 148 35 567 Source: Mango, et a1(2007). * Numbers do not add up to 100percent because more than one reason could be cited. 3.20 In escaping poverty, the following factors were mentioned most frequently (Table 3-3): Diversijkation of income sources-off farm: Developing a small local business was the most important single factor cited, including petty trading, kiosk retailing, operating taxis and running local hotels and bars, reaffirming earlier findings (see e.g., Lay et al, 2007). Diversijkation of farm incomes: Many farmers had diversified their crops and believed this contributed to their transition out of poverty. Similarly, about a third of the pastoralists also had become better off attributed their welfare improvement to diversifying their livestock. Improvement infarmingpractices; About a quarter of crop farmers cited crop commercialization as a favorable change. Expanding land under cultivation was also important. Social support: About a quarter of households inmost groups cited help from relatives and friends as important drivers out of poverty, although we do not know whether this was to enable investments (e.g., pay for schooling costs) or to cope with shocks (Box 3-2). 85 Box 3-2: Philip Lagat-Escaped Poverty through Diversification in Farming Philip is a 46 years old farmer in Chepkongony village, Nandi district. He never went far with his education because he came from a poor background. In 1990 he was working as a casual labourer in the tea farms of well to do villagers. He used to earn a paltry Ksh 200 per month from casual work. H e saved and bought a local cow for about Ksh 800 in 1992; hethen stopped working as a casual labourer to take care of his family farm and the cow. In 1995, he got ajob inaprominent tea farmer's farm as acasualwith aslightly betterpay ofKsh300. Inthe same year, he set up a tea seedlingnursery inhis farm. Fromthis nursery, he transplantedup to 3,000 tea seedlings into his farm. In 1997, he started harvesting his own tea. His cow was also providing some milk. His total earnings per month from tea and milk increasedto Ksh 4,000. H e stopped working as a casual labourer and decided to concentrate in farming. By the year 2000, he had bought a Friesian dairy cow. He also diversified into growing coffee. Hejoined SouthNandi Coffee Cooperative Society which helped inmarketing his berries. By the time we visited him (2006), Philip was earningup to Ksh40,000 per month from sales of his tea, coffee, ISource: and milk combined."My family is not doing very well but we are comfortable" he said. Mango et a1(2007). Table 3-3: Major Reasonsfor EscapingPoverty by LivelihoodGroup (percent households) Group Mix Farm/fish FarmAivestock, Pastoral Urban Overall farm, low low high Business: ruralkommunity 44 72 47 50 62 51 basedenterprises Business:small city-based 6 0 8 14 14 8 enterprises Regular employment in 18 26 22 47 62 28 private/public sector Crop diversification 30 50 20 12 19 26 Crop intensification 11 0 0 0 0 6 Crop commercialization 26 24 27 17 0 23 Increasedlandunder 29 0 36 0 0 23 cultivation Livestock diversification 18 13 12 29 0 15 Livestock commercialization 0 9 20 21 21 12 Increasedherd size 0 24 0 0 0 10 Few dependants 22 13 20 0 14 18 Help from friends andrelatives 20 30 25 31 24 25 incountry Inherited property 20 0 33 10 8 20 142 46 121 42 37 388 * Numbers do not addupto 100percentbecausemore than one reason could be cited. Source: Mango, et a1(2007). 3.21 Inthe rich context provided bythese various Kenyancommunities, we now turn to examine more quantitative evidence, to get a better understanding o f the significance o f different drivers. We look both at transitions out o f poverty, through, for example, job opportunities and income diversification, as well as on negative shocks, like poor health. 86 DISTRIBUTION WELFARE OF IMPROVEMENTS,1997-2005/6 3.22 Itis always usefulto look at the distributionof improvements inwell-being over time. However inthe case of Kenya, as emphasizedabove, the nationally representative snapshots that are available cover several years of economic stagnation and recession, anda recovery only inthe two years priorto 2005. This means that average changesover the period cannot be interpreted as ongoing trends, certainly not with respect to growth, nor necessarily with respect to distribution. In addition, as discussed in Part I1of this report, differences in survey methodology, particularly with respect to the measurement of household expenditure and earnings, raise comparability issue across the two national household survey rounds. We are simply unable to tell from the data that is available. Nonetheless, looking at the changes over time provides an illustrative starting point, whichis complementedbythe results ofurbanslum andrural panels. 3.23 With these caveats firmly in mind, the distribution of changes in consumption between the two official surveys is examined.43 Table 3-4 compares monthly adult equivalent expenditures per capita for 1997 and 2005-2006 for various sub-groups of the population. 3.24 Average growth was 15 percent but this did not benefit the whole society: The poorest quintile lost out in absolute terms and gains for the second bottom quintile were only about one percent annually. Also for the middle quintile, growth inexpenditure was below average. 3.25 Average welfare gains were very much concentrated, in particular amongst the wealthiest quintiles, urban residents, and, interms of provinces, Nairobi (especially), and Nyanza and Eastern. For many other groups, the aggregate improvement over the period was very limited and practically stagnant. Since the period covered is almost nine years, an aggregate change of less than about five percent arguably amounts to stagnation-and this applies on average to rural residents, andthose inCentral, Coast and the Rift Valley (Table 3-4). 43The 1997 expenditures are adjusted using the ratio of poverty lines. The earlier cautions about the comparability ofthe consumptionaggregatesover time shouldalso be borne inmind. 87 Table 3-4: Inequality in Distribution of ConsumptionGains, 1997 -2005/06 1997 2005/06 YOchange Area Urban 4,436.59 5,493.96 23.8 Rural 1,962.94 1,993.21 1.5 Province Nairobi 5,065.96 7,355.14 45.1 Central 2,841.20 2,930.62 3.2 Coast 2,509.92 2,45 5.78 -2.2 Eastern 1,83 1.41 2,108.69 15.1 Nyanza 1,734.40 2,280.12 31.5 RiftValley 2,269.36 2,360.56 4.0 Western 1,727.91 1,909.32 10.5 Quintile Poorest 708.07 702.99 -0.7 2 1,159.83 1,268.59 9.4 3 1,680.82 1,846.75 9.9 4 2,439.84 2,778.18 13.9 Wealthiest 5,758.93 6,895.72 19.7 National 2,349.36 2,698.12 14.8 Source: WB estimates basedon KIHBS(2005/06), W M S (1997). Notes: Mean adult equivalent per capita expendituresare used and 1997 is adjustedto 2005/06 values usingthe ratios ofthe urban andrural poverty lines. 3.26 Further analysis of the impact of inequality on the record of poverty reduction over the period, 1997-2005/6 uses a method associated with Datt and Ravallion (1997). The observed poverty reduction is decomposed into changes in the average and changes in its distribution. These results are based on keeping the poverty line constant in real terms (again using the ratio of poverty lines as the deflator). The "growth" effect i s the simulated impact of the increase in meanper capita expenditures on the headcount (with distribution constant) and the distributioneffect is the simulated impact on the headcount of a change inthe distribution (keeping meanexpenditures constant). 3.27 These results suggest that, between 1997 and 2005/6, average consumption increased, but so did inequality. If inequality had not worsened, poverty could have fallen by almost 20 percent nationally, instead of 5.5 percent that resulted in practice. The drag imposed by rising inequality was even larger inrural areas: ifinequality had not risen, the headcountwould have fallen by 15 instead of 3 percent. This i s consistent with the foregoing results which showed that the per capita expenditures of the richest 40 percent of the population grew much faster than average, whereas the lower quintiles did not experience much, if any, improvement. The same decomposition was estimated by province and suggested that the extent to which inequality was a drag on poverty reduction varied by province. Itwas most pronouncedwithinWestern andRift Valley. 88 Figure 3-2: Contributionsof Changes in Average Consumption,and Changes in Inequality to Poverty Trends, 1997-2005/6 150 , Tolrl Urban Rural 131 5 5 1 -5O00 -150 j -149 -19.9 .21 6 .250 J Source: Staff estimates based on W M S and KIHBS. Note: Mean per capita equivalent adult expenditure. 3.28 This decomposition is suggestive, andunderlinesthat inequality is a major issue not only inits own right, but interms o f its impacts on poverty reduction. However, it i s important to recall that it cannot be concluded from the national data that the growth recovery increased or decreased inequality, since we cannot distinguish the sub-period after 2003. LABOUR MARKET TRENDS 3.29 The labour market is the primary transmission mechanism of economic growth to household incomes. Since labour is the most abundant asset of the poor, a major determinant of the poverty reduction impact of economic growth is its effect on the quantity, quality and distribution of employment opportunities. When growth enhances earning opportunities for the poor, its benefits are widely shared and poverty tends to decline. InKenya, as inmany other developing and industrial countries, jobs are at the heart of lively economic and political debates. Major policy documents since independence have stressed this theme, most recently in the `Vision 2030' and the Medium Term Plan (2008-2012). Recent studies on the topic include the UNDP sponsored investigation of the 2005/6 KIHBS, published as `An Employment Targeted Economic Program for Kenya'. This section introduces some basic labour market trends to help set the stage for analysis which follows. There is a general caveat that conventional labour market concepts are difficult to apply in Kenya, as in other developing countries.44 44 The most obvious constraint is that almost half o f the working-aged population is self-employed in agriculture, where it i s difficult to measure net earnings. Recorded unemployment excludes those Kenyans who cannot afford to be unemployed. Measuring labor force participation as a discrete condition is fi-aught with difficulties, especially for women, and also because this is often not measured consistently across 89 3.30 Figure 3-3 presents the labour challenge invery simple and stark terms, for both the stock and flow showing population and job force against the number of jobs as reported in the official Economic Surveys and for two points in time, Labour Force Survey andKIHBS. The flow rates diverge, with the population increasing by nearly one million every year, and the labour force by close to half that number, compared to a growth in wage employment annually which has hovered at levels well below 100,000. Thus, inabsolute terms, the implied gaps are enormous, and growing. The urbanlabour force grew relatively faster, reflecting rural-urban migration and entry of women into the labour market, which dominated the effects o f individuals delaying entry due to continuing schooling. Figure 3-3: Total Population, Labor Force and Wage Employment(`000 People). 1981-2006 40,000 35,000 -$1 cp 30,000 8 a 25,000 20,000 0 s 15,000 10,000 ccp -s: U 5,000 0 CI 8 0 3 0 n cp years -WagEmployment +Population +Labor force + A Population(ILFS'98 &KIHBSOS) Labor force (ILFS'98 & KMBS'05) +Paid Employment(ILFS'98 KIHBSO5) Source: Staff Calculations Basedon Economic Survey, ILFS'98 andKIHBS '05/06. 3.31 The overall record ofjobs performance inKenya over the past decade suggests some good news as well as a series o f major challenges. Interms of identifying trends, and the labour intensity of growth, we are handicapped by the fact that between the survey points there were several years of stagnant or declining economic activity as well as recovery. For the analysis however we only have the endpoints. Comparing the cross- sectional survey results of 1998 to 2005/6, the broad storyline is that: There was net job creation in both rural and urban areas. Rural employment grew from 8.4 to 9.4 million between 1998 and 2005/06 (1.5 percent annually). Urban employment rose from 2.1 to 2.4 million during the period (2.1 percent annually). 0 Unemployment fell from almost 15 to 12.5 percent. The total number of unemployed fell almost by 150,000 people, from 1.83 to 1.68 million, despite surveys. For example, recent analysis o f Tanzanian data suggests that an observed decline in "activity" can be traced at least inpart to differences inhow the relevant question was asked. 90 a rise of 150,000 in urban areas. Improvements were concentrated in rural areas where the rate fell from almost 14 to 10 percent, whereas the urbanrate rose from 18.5 to 20.6 percent. The share of youth (15-29 years of age) among the unemployedrosemarkedly, from 60 to 72 percent. 0 Inurban areas, the observation that both unemployment rate and the number of jobs rose reflects that the economically active population increased at a higher rate (2.4 percent annually) thanthe number ofjobs created (2 percent). 0 The share of the working age population recorded as inactive (those at work, nor actively looking for work) increased rapidly, from 22.4 to 30 percent of working age population. This is mostly due to the cohort share classified as full-time students, which rose from 11 to 18 percent between 1998 and 2005/06. There was also a jump in those reported as discouraged (Le., not working nor actively seeking a job), from 0.5 percent of the working age population to almost 3 percent in2005/06. 3.32 Female unemployment i s higher in both absolute numbers and relative terms than male unemployment (13.8 versus 11.2 percent). The female unemployment rate did nonetheless fall significantly between 1998 and 2005/6, since the level had stood at almost 20 percent. The decline was partly explained by the slower annual increase of female labour force participation (0.9 percent, compared to 1.6 percent per year for males) as relatively more women decided to continue their full time studies. The data also register a higher increase in the number of discouraged females. It is notable that three out of four unemployed women are youth aged 15-29. In addition, about one quarter of employed women were underemployed, in the sense that they worked, on average, less than28 hours per week. 3.33 Trends inemployment shares between 1997 and 2006 suggest lack of structural change inthe economy. Almost seven million Kenyans worked in agriculture in 2005/6, roughly one million more than in 1997. Employment in secondary activities was around 750,000, of which about 440,000 were in manufacturing, rising slightly as a share, from 5.5 to 6.3 percent. Figure 3-4: Broad SectoralShares of Employment, 1997 -2005/6 2005/06 1997 4 I 0% 20% 40% 60% 80% 100% Primary m Secondary Tertiary Source: Staff calculations based on W M S and KIHBS. 91 3.34 Looking at the sectoral composition of employment over 1997-2005/6, we see that the share of primary activities-viz agriculture, fishing and so on-rose almost 1.3 percentage points. Secondary employment-mainly manufacturing, and other activities, like construction-remained roughly constant at about 6 percent. Tertiary employment (Le. in services) shrank. The phenomenon can be traced inpart to the reduction of public employment which was not fully compensated by private sector expansion (Table 3-5). Table 3-5: Trends in the Composition of Employmentby Main Sectors, Urban, Rural and Total (percent), 1997 and 2005/6 1997 200516 Rural Urban Total Rural Urban Total Primary 71.9 12.2 58.4 73.1 7.8 59.7 Secondary 4 13.1 6.1 4.2 13.2 6.1 Tertiary 24 74.6 35.5 22.6 79 34.2 3.35 Looking behind these broad categories, there has been some sectoral variation (Table 3-6). Agriculture maintained its share of employment. Employment in manufacturing increased. Mining, and transportation, storage and communication also rose, but from a much lower base. The shares of several important sectors decreased. This reduction i s most pronounced in construction and community, social and personal services. As noted above, the last is associated withtrends inpublic employment. Table 3-6: Distribution of the Employedby Economic Sector, Shares and Changes, 1997 200516 - 1997 2005106 YOChange Agriculture, Hunting, Forestry& Fishing 59.5 59.4 0.0 Mining& Quarrying 0.2 0.6 175.2 Manufacturing,Electricity, Gas and Water 2.6 3.8 47.5 Construction 3 2.5 -18.4 Trade, Restaurants,Hotels, Financial,Other Services 32.4 30.3 -6.5 TransDortation. Storage.Communications 2.3 3.3 46.9 Source: Staff calculations based on WMS'97 & KIHBS'05/06. 3.36 A Shapley decomposition can be used to break down per capita GDP growth into three components: output per worker, level o f employment and inter-sectoral shifts o f labor (see PREMPR 2007). The method i s counterfactual, isolating changes under ceteris paribus assumptions. For instance, the contribution o f changes in output per worker on per capita GDP i s estimated in the absence o f changes in the level of employment or sectoral shifts o f employment. 3.37 The decomposition results suggest some good news, as well as possible causes for concern, which we investigate further below: Most welcome are the significant annual increases in labour productivity in the private sector; However, the inter-sectoral shifts in employment almost totally offset sectoral productivity improvements-in that secondary private sector employment 92 shrank, with the net effect that per capita improvements in output were negligible; Public sector employment shrank (Manda (2004); Bigsten and Kimuyu (2002); Soderbom et a1(2004); 0 Changes inthe capital-labour ratio were negative; and Labour force participation declined, due to increased schooling. Table 3-7: Decompositionof Changes in Per Capita Value Added GDP* (PercentagePointsper Year), 1997 -2005/6 Contributionsfrom changes in: Output per Sectoral compositionof worker Employment employment Total Changein Per Capita GDP: 0.24 SectoralContribution 1.8 0.4 -0.5 1.6 Public sector 1.2 -0.6 -0.6 0.0 Private Sector 0.6 1.o 0.0 1.6 Primary 0.4 0.5 -0.2 0.8 Secondary -0.1 0.1 0.1 0.0 Tertiary 0.3 0.4 0.1 0.8 Participation rate -1.5 Working Age PopulationIPopulation 0.1 Decompositionof changein output per worker, due to: 1.8 Changes inTotal Factor Productivity 2.5 Changes inCapital Labor Ratio -0.8 (*)Value Added GDP is the sumofGDP by activity, without deducting imputed financial services andnet subsidiesusedinoverall GDP calculations. 3.38 This picture is broadly consistent with developments on the policy side. A series o f economic reforms inthe 1990s opened up the economy and curtailed the size o f the state sector and its engagement in direct production (see World Bank 2008a for details). The structural reforms intended to improve the efficiency o f the Kenyan economy were associated with significant retrenchment o f workers from both public and private enterprises. There was significant loss o fjobs inthe formal private sector, mainly inmanufacturing firms that were unable to compete without subsidies, especially textiles and agricultural processing firms (Manda 2004). So the net employment effect reflects a smaller but more productive secondary sector. 3.39 Amongst those Kenyans who reported earnings, about 41 percent (3.3 million) hadjobs in2005/06 with earnings over the poverty line.45 We can call these good quality jobs, while recognizing that having such a job does not necessarily enable an individual with dependents to live above the poverty line (Figure 3-5). Almost 6 out o f 10 workers who reported earnings hadjobs with earnings below the poverty line, whom we can call working poor. Underemployment, those working less than 28 hours a week, affects about 45Earnings data is missing for 31 percent of workers in the sample, mainly own account and unpaid family workers. 93 21 percent of workers, o f whom about 40 percent are youth. The share of working poor andthose underemployed likely overlap to some extent (Figure 3-6). Figure 3-5: Extent of Good and Bad Quality Jobs, Earnings Relativeto the PovertyLine Public& Modern Private Public& Modern Private, Paid 0 10 20 30 40 50 60 70 80 90 100 percent of workers aged 15.64 with data on earnlngs Source: Staff calculations, basedon KIHBS `05. 3.40 Figure 3-6 show that there is a higher share of working poor in rural areas-indeed only about 20 percent had earnings above the rural poverty line in 2005/6. In urban areas, a higher share of women is low paid. Figure 3-6: Trends in Employmentand Unemployment. 1998-2005h6 by Gender, Urban / Rural, and Good Quality Jobs in 200506 1800 Urban 1WO 2WYS y 1200 g 1998 800 f W O 4w 200 0 Male Female -Good Quality 0Employment &Unemployment Source: Staff calculations, based on KIHBS `05. 3.41 Looking by province, Western and Coast registered the fastest annual average increases in employment, at 5.4 and 3 percent. Rift Valley accounts for almost one quarter of total employment. Central and Eastern provinces registered highest rates of increase inthe number of unemployed (7.8 and 4 percent, respectively). Nairobi provided by far the highest fraction of good quality jobs (60.5 percent of total employment in the province). This presents some contrasts and puzzles vis a vis the provincial distribution of consumption gains that was presented above. Coast had employment growth but stagnant consumption per capita, which may reflect rapid population growth and/or a predominance of `bad quality' jobs. Eastern was a relative gainer in average consumption, but increases in unemployment were marked also, while Central also experienced rising unemployment. More work is needed to bring together the regional poverty and labour market stories. 94 3.42 About 50 percent of workers in paid employment had "good quality jobs" compared to almost 62 percent for business owners for which earning data was available. Among both own account workers and unpaid family workers, the share with good quality jobs as definedhere i s muchlower, possibly about 20 percent.46 3.43 Earnings in agriculture are systematically lower for all types of employment, and lowest for those working on their own account or inthe family. Levels of inequality in earnings are generally high in Kenya, with an overall measured gini index of 0.693. Within categories, inequality is higher among agricultural workers (gini of 0.71) than amongst non agricultural workers (gini of 0.62). Compared to neighbors and other regions, the extent of inequality inKenya is very high. Interestingly inKenya, inequality i s sharpest in rural areas and, overall, among those with lower levels of education. It i s also highamong those working intrade and services (Table 3-8). Table 3-8: Individual Monthly Earnings: Median and Gini- Coefficient, 2005/6 (In nominal KSh./month) Non -Agriculture . Agriculture Paid Employee Median 4,114.9 1,371.6 Gini 60.1 51.3 Working Employer Median 12,344.6 1,714.5 Gini 59.0 66.1 Own Account/UnpaidFamily Worker Median 4,509.2 685.8 Gini 63.6 77.1 TOTAL Median 5,066.7 1,200.0 Gini 62.4 71.0 Source: Staff calculations, basedon KIHBS'OS usingADePT Labour. 3.44 Not surprisingly, relatively educated workers are more likely to work in the public and modern private sectors. At the same time, more than one quarter of those working inagriculture finished secondary school (Table 3-9). Table 3-9: Composition of Employmentby Education Completed(percent), 20096 Public Modern Private InformalSector Agric & Other None 0 0 0.2 0.1 Primary 17.2 38.7 62.8 72.6 Secondary 74.6 54.9 36.1 26.7 University 6.6 4.9 0.7 0.5 Post Graduate 1.7 1.5 0.2 0.1 Total 100.1 100 100 100 Source: Staffcalculations, basedon KIHBS2005/6. 46 This may partly reflect that, not surprisingly, many workers in the "unpaid family" category were not paidincash, andthe inkindnature of earningsalso of small scale farmers. 95 Structureof Employment 3.45 In2005/06, most workers (58 percent) were small scale farmers, in pastoralist activities and other primary employment. Recently publishedanalysis of the KIHBS by UNDP and others has extensively investigated the profile of employment in 2005/6. Here we look at the structure o f employment in terms of both "status"-that i s whether the individual is a paid employee, businessowner, working on own account, or unpaid in the famil and sector ofthe economy. Beginning with trends inemployment status we see that:47- Paid employment in rural areas stagnated between 1998 and 2005/06 at around 2 million workers whereas urban paid employment rose in absolute terms, by over 170,000 or an annualrate of 1.8 percent. As defined by the surveys, business owners represented fewer than 2 percent of total workers in 2005/06. At the same time, 'there appears to have been significant growth, albeit from a very small base, from 46,000 in 1998 to over 200,000, which i s bothnotable and welcome. Own account and unpaid family workers, together, fell slightly as a share, but at seven million in 2005/06 still represents about two-thirds of employed Kenyans. 3.46 About 927,000 people worked in the public sector (central government, public sector owned firms, parastatals, publicly employed teachers and other public activities). There are indications that public employment shrank between 1998 and 2005/06. Economic survey data shows a decrease in public employment from over 711,000 in 1998to about 650,000 in2005/06.48 About 73 percent o fthe public employees are male. 3.47 The minimal share of private modern employment in 2005/06 data, a mere 6 percent of the labour force (671,647 people), is striking. About 41 percent of these modern formal workers were located inurbanareas. Two thirds were male. 3.48 The term informal sector includes a diversity o f activities from the very small scale, with low returns and prospects, to much more lucrative pursuits. Survey data indicates the existence of a two-tier informal sector inKenya, with significant differences inboththe capital usedto start activity andthe averageearningsperworker. 3.49 Informal sector absorbed 3.3 million of workers in 2005/06, of which 1.4 million (44 percent) were located in urban areas. It was evenly divided by gender. The sectoral distribution of informal activities, as shown in Figure 3-7 below, reveals that 47Changes in employment by status represent a minimum estimate of variation, since for 2005/6 we are unableto allocate employment status of almost 10 percent of workers (comparedwith a negligible number of missingdata o f employmentby status in 1998). 48The Economic Survey is more accurate in estimatingpublic employment than in estimating private and informal employmentandwages. 96 services accounts for the highest share, followed by manufacturing. Informal sector jobs tend to be associated with lower educational attainment. Average earnings o f informal workers average only about one third o f earnings inthe modem private sector. Figure 3-7: Distribution of Employmentin the Informal Sector I Transport. & Other Other Services Construction Trade, Restaurants& 44% Source: Staff calculations basedon KIHBS 200516. 3.50 To sumup, the data do not reveal significant poverty reducing momentum inthe labour market, at least through 2005/6. The picture emerging presents major policy challenges. There was little growth in employment in modern sectors, which remain a relatively small share o f the economy. Earnings were low, especially in agriculture, and the level of inequality inearnings was large. CHANGESINTHEDISTRIBUTION ASSETS: OF HUMAN CAPITAL,LAND ACCESS AND TO CREDIT Human Capital:focus on Education and HIV 3.51 Since 2003, there have been important gains in primary and secondary enrolment. The prognosis for broad based improvements in labour quality in Kenya appears to be good, and trends inthe distribution o f education is an equalizing force. 3.52 Inrecent years, the transitionrate for children from primaryto public secondary schools has risen to 51 percent. The gross enrolment rate at the secondary level i s about 30 percent, and 79 percent o f those who enter secondary school complete all four years. The increased number o f primary graduates is putting great pressure on the limited secondary school places available. 3.53 Tegemeo's rural analysis shows that over the decade to 2007, the fraction o f adults in rural households with no education had fallen, and more rapidly for the poor. However large differences in levels of education between upper and lower income 97 quintileswere still observable in2007. Furtheranalysis is presentedbelow about the role of education as a driver out of poverty. 3.54 The data clearly suggests that more schooling is associated with betterjobs and thus a greater likelihood of not being poor, Kenyans with higher education tend to find jobs in the formal sector. Inthe KIHBS, about three-quarters of undergraduates and 81 percent of post graduates has formal jobs, as opposed to 14 percent of those with secondary education and 5 percent of those with primary education. Also, the crude risk o f unemployment decreases as education attainment increased. Likewise, for urban wages, there a significant premiumfor higher education, especially for men. 3.55 At the same time, we needto raisethe caution that as education has expanded in Kenya, completion of primary schooling is enough to escape from poverty. Indeed,those with completed primary education earn only about half the average. Also notable is the crude gender wage gap, which is marked especially at lower levels o f education (Table 3-10). Table 3-10: Urban RealWages by Education Levels2005/6 as Percentageof Average (percentunlessotherwise specified) Male Female Total No Education 18.2 7.9 11 Primary 59.4 42.6 52.8 Secondary 124.5 98.9 115.1 University 343.1 328.6 338.9 Post graduate 453.8 435 450.1 Average urban earnings(KSh.Month) 18,475 13,525 16,552 Source: Staff calculationsbased on KIHBS 2005/6. 3.56 Likewise, worker-level data for formal sector firms suggests that the Kenyan force is relatively well-educated (ICA 2008). The typical worker in the modal firm in Kenya's ICA sample has between7 and 12 years of schooling. While this is lower than South Africa, it is higher than the other African comparators. In Kenya, 85 percent of firms report typical education levels of more than seven years, compared to 65 percent in Tanzania and 63 percent inUganda (World Bank, 2008~). 3.57 Shortage of skilled workers i s the least important constraint in Kenya's investment climate (World Bank 2008b). Over the past three years, the proportion o f firms reporting skills shortages as a major or severe impediment to growth has dropped from 28 to 8 percentY4' and i s now only half the share of neighboring Tanzania. However, there are some differences across firms: those in manufacturing are twice as likely as retail firms and more than six times as likely as service firms to report skills shortages as a constraint. Larger, foreign-owned and exporting firms in the manufacturing sector are more likely to report that lack of skills i s a major or severe impedimentthanother firms. 49Similar results are obtainedifthe panelportion ofthe sample is used. 98 3.58 The I C A notes that three hypotheses could explain why firms do not have immediate concerns with the formal education o f the workforce, either that firms have made the necessary input-mix adjustments that are compatible with a low skills workforce; that firm-based training can substitute for poor formal education, and/or that other constraints in the business environment dominate skills issues. It i s difficult to empirically validate any of these possibilities, though it i s noted that about 41 percent o f firms inthe sample provided training to their workers. Inthose firms, nearly two-thirds o f skilled workers and about halfo f unskilledworkers received training. Health 3.59 As outlined in Part 11, many o f Kenya's health indicators are poor and health disparities remain high. Malaria continues to be the leading cause o f morbidity and mortality in Kenya. Other major causes include acute respiratory infection, malnutrition, diarrhea disease, HIV/AIDS and among the highest estimated TB incidence rates in the world. Health inequalities are also an extremely serious concern, as health outcomes vary significantly indifferent gender, socio-economic and geographical groups. 3.60 There have been signs o f improvement. For example, HIV prevalence among adults fell from over 10 percent in the late 1990s to 5.1 percent in 2006, and there are signs o f reduced malaria morbidity due to increased coverage and usage o f insecticide treated nets (ITNs) by pregnant mothers and children, and more effective diagnosis and treatment. While still low, TB control has also improved with the case detection and treatment success rates improving (Table 3-11). Table 3-11: Trends in Selected Health Indicators 1993 1998 2003 Latest MDGTarget pregnantwomen HIV prevalenceamong 15-24year old 7.4 11.9 10.6 5.1 <4.5 Malaria prevalenceage >5 (%childrenwith fever) 41.8 42.3 41.6 TB treatment completionrate 80 85 90 Source: DHS and MoH. 3.61 There are now at least 1.2 million Kenyans living with HV/AIDS. The negative impacts o f untreated HIV/AIDS on individuals and households in Sub-Saharan Africa have been documented in a range o f settings and countries (see e.g., Beegle, 2005). Research has documented the decline in work performance associated with untreated HIV/AIDS inthe case o f tea pluckers in Kenya (Fox, Rosen et al. 2004), and declining labour supply in rural areas around Eldoret (Thirumurthy, Graff Zivin and Goldstein, 2005). Given the medical effectiveness o f antiretroviral therapy (ART) in restoring immune function and extending survival, and with over 160,000 Kenyans receiving ART in 2007, attention has turned to evaluating various socio-economic impacts of ART on individuals, households, andprivate sector companies. 3.62 One important issue i s the degree to which ART restores physical functioning so that an individual can continue with or return to her normal daily activities after 99 initiating therapy (Larson et al, 2008). Returning to normal activities includes reducing both absenteeismand "impaired presenteeism," the term usedfor the loss of productivity when an illor disabled individual shows up for work but accomplishes less. Recent careful investigations in Kenya using a detailed panel data set (Larson et a1 2008) compared tea plantation workers receiving ART through their company hospital and a matched comparison population of employees in the general workforce. They examine the pattern of impaired presenteeismover time pre andpost-ART, the monthoftreatment initiation, and 12months post-ART. They find substantial impaired presenteeism, mainly through women shifting to less strenuous assignments during the year prior to ART. Men returnedto a nearly similar work pattern as the generalmale workforce after three months on ART through the end of the study. Women however continued to spend fewer days plucking tea but more days on other tasks than the general female workforce throughout the first year on ART. 3.63 A similar rural study near Eldoret in Western province finds a sharp labour supply response to ARV treatment (Thirumurthy, Graff Zivin, and Goldstein, 2005). Using a sample o f patients at a rural health centre and a random sample of households in the catchment area as controls, labour supply of patients was found to decline sharply in the months before treatment, but returning to close to normal levels of labour supply within six months of the initiation of treatment. This, and a related study (Graff Zivin, Thirumurthy and Goldstein, 2006) also show strong intergenerational effects of treatment. First, as adult patients return to work, their children, especially boys, reduce the amount of time they spendworking inthe labour market. Second, inthe same time period as increased parental labour supply, children of patients increase their school attendance by over 20 percent. Finally, the nutritional status of children of patients also increases sharply, particularly for children who were the most malnourished. 3.64 These results suggest that there are important productivity gains and intergenerational human capital effects associated with ARTs. It also suggests that the estimated one million Kenyans who are HIV positive and not on ARTs are likely less productive thannormal, with potentially serious impacts on the next generation as well. 3.65 To sum up the role of human capital in Kenya, we see contrasting trends. On the positive side, we see the expansion of basic education and the steady decline in the share of firms reporting inadequate skills as a major or severe constraint to growth. On the other hand, maintaining quality inschooling is a major challenge, as is the expansion of the system at the secondary and upper levels as primary education alone is not necessarily a route out of poverty. And there are major health challenges, most prominently malaria and HIV, but also other causes of morbidity and mortality which adversely affect, among other things, labour productivity, and thus individual and household earnings prospects. We examine ill-health impacts further below in the context of shocks. 100 Land as a Driver of Povertyand Inequality 3.66 Landis central to our assessment of poverty and inequality inKenya-both for its role in determining the pattern o f development, and its significance to social and political stability. To quote the recent Ndung'u report: Land retains afocal point in Kenya's history. It was the basis upon which the struggle for independence was waged. It has traditionally dictated the pulse of our nationhood. It continues to command apivotal position in the country's social, economic,political and legal relations. 3.67 The landquestion inKenyaneeds to be considered inthe context of history and governance. Some salient historical factors were highlighted in Part I, here we recall some basic facts on land: that during colonial times, land was appropriated by settlers across the country; the earlier communal land system was blocked into trust, crown and private lands, and a cash economy was introduced. At least since the 1980s, land administration and management, inparticular injustices and favors in land allocation and corruption inlandrelated institutions, have becomemajor issues inpublic debates. 3.68 As noted at the outset, whereas only 20 percent of the land is classified as mediumto high potential, about three out of four Kenyans are concentrated therein, with the rest living in vast areas of the arid and semi arid lands (ASALs). Given the rising population size and the continuing largely rural nature o f income earning activities, access to land is critical to popular perceptions about well-being. The illegal appropriation of public land has been an especially visible crime which has excited huge passion, not least because, as the Ndung'u Commission found, the practice of illegal allocations of land increased dramatically during the late 1980s and throughout the 1 9 9 0 ~ . ~ ~ 3.69 Average landholding size for smallholders is defined by the KNBS (2001) as landholdings between 0.2 - 12 ha, is 1.2 ha while for large farms, it is 700 ha (CEM 2003). The contribution of smallholders to marketedproduction has risen over time, to an estimated 70 percent in 2003. Very little (only about 1.2 percent) of the cropped land is irrigated. 3.70 The KPIA does not attempt to fully deal with the land question. Our scope is limitedto examining levels andtrends inlandinequality inKenya. This is useful because disparitiesinaccess to landloom large incurrent debates, anda robust empiricalbase has been absent. However, the long list of land issues, with which we do not deal here, includes controversies around internally displaced peoples (IDPs), past and present; resettlement schemes (as in the Rift Valley); squatters and land fragmentation; slum settlements, insecure tenure/overlapping rights by individuals/communities; disinheritance by women and other vulnerable groups; environmental degradation; destruction of forests and desertification; and allocation of public land to individuals 50 Southall, Roger (2005), The Ndung'u Report: Land & Graft in Kenya, Review of African Political Economy, 103, March2005, pp.142-51. 101 (land grabbing). Much of this agenda is being pursued in related work by the Land Reform Transformation Unit of the Ministry of Lands and Housing which is managing the completion ofthe National LandPolicy andpreparation for its implementation, and is actively engaged in by national NGOs, like the Kenya Land Alliance and various development partners. 3.71 There i s a rich international literature on land inequality, much of which has been motivated by the high levels of land inequality in Latin America, and the associations with access to credit (Binswanger 1986) and education (Mariscal and Sololoff 2000). Previous work (e.g. Deinenger and Squire 1998) has focused on inequality among landholders. However this misses important inequalities vis a vis the landless. Erickson and Vollarht (2004) give a stark example to show the pitfalls: suppose that only two people in a country of a hundred people own all the land, and share it equally. The gini among landholders would be zero-perfect equality-yet 2 percent of the population own all the land! Inthe KPIA, we present measures of inequality both for the whole sample- including those without land and those who did not engage in farming over the preceding year-and restricted to the land holding group. We would ideally adjust for security of tenure, but the data at our disposal does not allow. 3.72 We provide the summary results for the whole sample, by province and over time, and for the sub-Sam le of landowners, drawing on the WMS and KIHBS for 1997 and 2005/6 re~pectively.~'Inprinciple, especially given the relative scarcity of medium andhighpotential landinKenya, we should want to adjust for quality. This canbe done at various degrees of sophistication. Ravallionand van de Walle (2008) for example look at Vietnam's evolving land distribution and calculate an "irrigated-equivalent" measure of land. No adjustment has been done in the present case, though we would note that the analysis excludes North-Eastern province, which i s largely arid, and only one percent of the landis irrigated. 3.73 The overall increase in inequality based on reported size of ownership is striking: nationally, there was a 36 percent increase over the period, to about -83when we look at landholding inthe entire population (Table 3-12). The worsening was especially striking in the Coast and Nyanza provinces. Levels of inequality in the latter year are remarkably high not only in Nairobi, but also the Rift Valley and Coast. The Nairobi figure is very high-perhaps partly reflecting the very high rates of tenancy (see Gulyani 2007). 3.74 We do not investigate the causes of increasing landequality, although it is noted that an underlying factor i s population pressure and diminishing land size. Smallholder farms are being subdivided while medium and large farms tend to remain intact. ''We do not separately report North Eastern, because o f the very limited number of observations (122 compared to more than 1000 for each o f other provinces). 102 Table 3-12: Changes in Land Inequality over Time, all Households, by Province and Nationally, Gini Coefficient, 1996 -2005/6 1997 200516 Percentage change National 0.612 0.832 35.9% Nairobi 0.757 0.993 31.1% Central 0.546 0.744 36.4% Coast 0.500 0.865 73.1% Eastern 0.601 0.73 1 21.6% Nyanza 0.475 . 0.815 71.8% Rift Valley 0.642 0.870 35.4% Western 0.579 0.769 32.7% 3.75 We also estimated inequality only among those who own land, inpart because of the importance of off-land activities in the economy, and also to enable international comparability. This measure excludes both the landless agricultural workers and those whose primary occupation i s not farming. 3.76 Measures of inequality are lower, though still high, when restricted to landowners only (Table 3-13), and the extent of increase in land inequality over the period i s large, but less than when we looked at the whole sample. Interestingly, this sample shows a decline in inequality in Nairobi, possibly because we are excluding renters. However, similar to what we observed for the overall sample, there was a significant worsening inNyanza and Coast, and very high levels of inequality inthe Rift Province, with an estimated land gini of close to .77, which i s very high by international, as well as Kenyan, standards. Table 3-13: Changes in Land Inequality Over Time, Landholdersonly, by Provinceand Nationally, 1997 200516 - 1997 2004 Percentage change National 0.612 0.711 16.1% Nairobi 0.757 0.642 -15.2% Central 0.546 0.610 11.9% Coast 0.500 0.686 37.3% Eastern 0.601 0.606 0.8% Nyanza 0.475 0.737 55.2% Rift Valley 0.642 0.769 19.7% Western 0.579 0.695 19.9% 3.77 Evenamong landholders only, it is notable that the reported levels of inequality are much higher than the estimated consumption ginis. Our estimated ginis for earnings are of similar orders of magnitude. 3.78 Putting the landholding results in regional and international perspective, Kenya's rates of land inequality were significantly higher than South and East Asia for example, but close to Sub-Saharan and North African averages. The more recent estimates for Kenya bring the country closer to levels similar to those observed in the region which i s most renowned for land inequality, LatinAmerica. 103 Figure 3-8: Land Inequality, RegionalAverages, 1986-1990, and Kenya 1997 I I Source: WDI, and staff estimates for Kenya. 3.79 Table 3-14 presents some individual country comparators, and suggests that levels of land inequality in Kenya are approaching those in Latin America, e.g., Bolivia, and are much higher than neighboring Uganda, for example, though lower than that reported for Tanzania. Table 3-14: Land Gini coefficients,Selected Latin American and African Countries (year varies) Argentina ,856 Bolivia ,768 Colombia .829 Madagascar .804 Tanzania .790 Uganda ,549 Venezuela ,917 Kenva -711 Source:WDI, and staff estimates. 3.80 One important question is whether the general presumption that landless rural households are particularly poor holds true in Kenya. In India, the landless would typically comprise both the ultra poor casual agricultural wage laborers, but also those who might be involved innon-agricultural livelihoods, and who might not necessarily be poor. In fact among the poor, about half are landless, but half are not. Interestingly, a higher share (close to two-thirds) o f the non-poor is landless. This shows that landlessness per se i s not a good proxy for poverty in Kenya. It would be useful to distinguish gender of the household head-given the issue of female disinheritance, although this would not pick up widows who are disinheritedand move to the towns. 3.81 In 2005/6, average net earnings of those employed in small scale agricultural activities (including those raising livestock) was only about one-quarter of employees working in other activities (public, private sector and informal non agricultural workers). Disparities were observed also within farmers-in particular landless versus those with land. The average net earnings of landless farmers was almost 30 percent below of that of land holders and more than 40 percent less than that of land holders who claimed to 104 have land titles. A regression of rural net earnings of agricultural workers on standard variables (gender, experience, education, etc.) shows a highly significant, positive impact of holding land, andthe size o f the land. Everything else constant, holding landincreases net earnings 340 percent over that of the landless. Initial investigations found land texture and land slope coefficients to have expected signs, but their impact is not significant. 3.82 The analysis of landinequalitythus highlights some key messages. First, levels of land inequality are high, indeed higher than observed inequality in consumption and earnings for example, and high also by international standards. Second, that there has been a rapid worsening in land inequality in most parts of the country over the past decade-at a rate of change that suggests that land inequality is rapidly becoming more serious in objective terms as well as in terms of political salience. The size of smallholdings is becoming a productivity issue. Third, however, the evidence also shows that landlessness and poverty should not be equated-approximately equal shares of the poor have, and have no, land, and many of the non-poor do not have land-and hence landlessnessper se is not a good proxy for poverty, although there are very high rates of poverty among landlessagricultural laborers. Access to Financial Services 3.83 As we shall see below, using longitudinal data, access to credit is key lever for rural households to escape from poverty. In this section we provide a sense o f the dimensions of access to financial credit, using the KIHBS, and financial services more generally, using the 2006 FinAccess survey.52 3.84 The key patternsobservedfor access to credit in2005/6 using the KIHBSare: 0 Most Kenyansdo not have access to any form of credit, with muchlower rates of access inarid landsinparticular. Levels of access across poodnonpoor and female/male headedhouseholdsare similar-close to 30 percent for all groups. The differences which exist are as expected--e.g., non-poor access is about 12percent higher thannonpoor. There is significant disparity in average loan amounts-with non-poor households 42 percent above the average, and poor 77 percent below; the advantage for urban households is even larger, whereas female headed are way below average (Table 3-15). The differences betweenthe top and bottom decile are more remarkable, viz more than 2.5 times the average, versus less than half the average amount. This is a simple cross-tabulation without any controls, but the patterns are clear. 52The FinAccessdataset includes4070 adult respondents, ofwhich450 are from urbandistricts, 2651from rural, 577 semi-arid, and 392 from ariddistricts. 105 Table 3-15: Credit Access and Average Loan-Sum, Previous 12 Month Percent of Households Average Loan-Sum (all AccessingCredit LoansCombined) * Relative to (KSh 000) Mean All households 3 1% 36.8 1.00 Non-poor 32% 52.1 1.42 Poor 28% 8.6 0.23 Rural 31% 22.1 0.60 Urban 29% 84.2 2.29 Female-headed 30% 16.3 0.44 Poorestquintile 27% 4.0 0.11 Richest quintile 33% 94.9 2.58 Source: KIHBS Note: Includesall sources of credit (formal and informal). * 3.85 The patterns inaccess and amounts may inturnbe traced back to the sources o f credit. We see that formal sources (like banks) are relatively unimportant across all groups. Micro-finance lenders and SACCOs are accessedby about 20 percent on average (but only 9 percent for the poor). We see that poor households rely disproportionately on personal relations and their horizontal links are likely to be to others who are not well off either. A similar pattern appears for female headed households. We separate out friends and money lenders-since there are likely to be very different rates o f interest-and see that the poorest are more likely to turn to moneylenders. For total amounts by source (not shown), personal relations remain the largest single source, and especially for those at the bottom o f the distribution. Since most households borrow from only one source, this picture i s not so different from access by source (Table 3-16). Table 3-16: Share ofHouseholdsAccessingCredit from Various Sources (percent)* All 4.8 4.5 0.3 0.1 13.9 1.6 48.3 13.1 2.2 3.2 1.7 0.2 13.3 3.7 households Non-poor 6.7 5.6 0.4 0.1 18.4 1.9 44.5 9.0 2.1 4.0 1.7 0.2 13.9 3.5 Poor 1.4 2.7 0.1 0.1 5.6 0.9 55.5 20.7 2.6 1.8 1.6 0.1 12.3 4.0 Rural 3.9 4.1 0.3 0.0 11.5 1.5 50.8 14.6 2.0 1.6 1.8 0.2 14.4 3.8 Urban 7.8 6.1 0.4 0.3 21.9 2.0 40.3 8.1 3.0 8.4 1.2 0.3 9.8 3.0 Female- 3.2 3.1 0.0 0.1 8.3 1.9 54.2 13.9 1.4 1.1 2.1 0.2 16.6 4.4 headed Poorest 0.3 2.2 0.0 0.0 1.9 0.5 58.7 24.9 4.0 1.0 2.3 0.0 7.3 5.0 quintile Top quintile 10.2 6.6 0.3 0.1 26.9 2.1 36.4 5.0 2.6 6.4 1.0 0.3 12.2 3.2 Source: Staff estimates basedon KIHBS. Note: Figures do not add to 100 becauseof multiple responses. 3.86 Not surprisingly in this context, very few households are able to borrow in the face o f shocks, particularly shocks that affect their friends and neighbors as well. This i s 106 explored further below, but suffice here to note that borrowing can be an appropriate strategy inthe face of temporary shocks, but not to permanentcuts inincome. 3.87 Figure 3-9 shows the most common first response is to run down savings and sell assets, including livestock, and many cut consumption as a first response. Only 3 percent borrowed, which is lower thanthe share that resortedto prayer. At the same time, there i s a complementary need to address the underlying risks (e.g. price and weather) in order to improve access and cost of credit. Figure 3-9: First Responseto Major Shocks(Among HouseholdsReporting Shocks) Other P Consume Less Other assistance Family assists Borrow Start new business Work more Sell animals Sell assets or land Spend cash saGngs I 0% 5% 10% 15% 20% 25% 30% Source: KIHBS, 2005/6. 3.88 Further analysis has been undertaken using the 2006 Financial Access Survey (FinAccess), which collects more detailed information about financial services, not only credit. In fact, a higher share o f the sample saves than borrow. It shows that informal financial services-ROSCAs and ASCAs, plus friends and family, and local shops-are the most common mechanismsthrough which Kenyans access credit (Box 3-3). Box 3-3: Informal Financial Services Many Kenyans belong to informal groups, which can help serve their savings and credit needs. These include: -- welfare -- rotatingor clan groups that assist each other, for example, intimes of sickness or to pay for funerals; credit associations and individual savings and lending groups, and a similar structure inwhich an outside manager is retained; and, -- investmentclubs that invest money for a group inshares or infixed assets suchas land or houses. The use of informal groups is highest amongst business owners in rural and semi-arid districts, which suggests that these are a substitute for the financial services that urban entrepreneurs tend to obtain from formal Droviders. 107 3.89 FinAccess confirms several important points about access to financial services: (a) A large proportion of the Kenyan population has no access to financial services, whether formal or informal; (b) The share of the population that uses no financial services is much higher in arid districts; (c) Within areas, access is further differentiated by source o f income andpoverty proxies (see below); and, (d) Although the population share that is served is similar in urban, rural, and semi-arid districts, the mix of services is different. In urban areas, respondents rely more heavily on services from banks and semi-formal sources (SACCOs and MFIs) while in rural and semi-arid districts, there is greater reliance on services provided via informal groups (Table 3-17). Table 3-17: FinancialService Access Strands (percent) Service Provider Total Urban Rural Other Semi-Arid Arid Formal:Bank, post office 18.5 43.8 18.1 15.8 8.7 Semiformal: SACCOMFI 8.1 22.7 17.2 10.2 4.1 Informal: ROSCAs/ASCAs 35 39.1 54.5 58.1 31.1 Excluded 38.3 36.9 34.4 34.1 64.3 3.90 Most Kenyans-regardless of where they live-have no access to any type of financial services. However, when access to credit i s considered, the picture is even bleaker. Only wage earners have even limited access to formal sources of credit and although there i s evidence o f declining access to credit as one goes from urban to arid areas, it i s the low levels of credit for every district type that i s most striking. Even for those who have access to financial services in some form, these services are limited to savings and payments products. 3.91 More detailed analysis of financial exclusion in Kenya (Johnson and Nino- Zarazua, 2007) usedmulti-variate analysis, and found: Type o f employment i s important-in particular formal, and especially public, sector employees. Those who are older, better educated and male are more likely to use formal services. Women are more likely to use informal services-viz ROSCAs, MFIs and group basedmechanisms. Some wealth variables, like owning a radio or a bicycle, are correlated with a lower likelihood of exclusion from both formal and informal financial services. There are provincial patterns in usage, and SACCOs and group based mechanisms in some provinces (Western, Rift, Coast and NE) reduce the risk of overall financial exclusion. 108 3.92 While the extent of exclusion from financial services, and inparticular credit, is quite striking, it is important to bear inmindhowever is that the financial services sector in Kenya is very dynamic. There is evidence of significant innovation and expansion over the past several years, including group-based mechanisms and mobile phones. In Part IV we explore some policy implications of the foregoing. RURAL DYNAMICS:INSIGHTSFROMTHETEGEMEO PANEL 3.93 Rural dynamics are critical to understanding trends inpoverty and inequality in Kenya. About three in four Kenyans live in rural areas and agriculture and primary activities account for about 60 percent of employment, though only about 30 percent of GDP. Part I1has shown that in rural areas, rates of poverty are much higher, levels of inequality are high, and access to various basic services and human development outcomes tend to be lower. Earlier work has drawn attention to factors tending to depress rural productivity and growth inKenya, ina combination of exogenous and policy related reasons (World Bank 2003; Ndulu and O'Connell, 2000; Tegemeo X) including the weather, land pressures, poor access to credit, weak infrastructure, and reduced effectiveness of extension services. InPart IV we look at the current status of key policy related factors. 3.94 The incidence of rural income poverty (albeit importantly excluding pastoralists) fell over the past decade-the Tegemeo data indicate a fall inthe headcount from 54 percent in 1997to 38 percent in2007. This section aims to better understandthe dynamics associatedwith this picture, andthe underlying drivers. As outlined above, the Tegemeo panelprovides a unique opportunity to investigate the importance of both initial conditions and changes over time. Inthis section we reviewoverall trends inincome and poverty, andthenhone inon poverty dynamics and, inparticular, drivers. 3.95 Even in a context of overall growth and poverty reduction, Tegemeo's panel suggests that most (55 percent of) rural households experienced poverty at least once between2000 and 2007 (75 percent over the decade to 2007) (Figure 3-10). Likewise, PPA found that some 60 percent had experiencedpoverty between 1997 and 2005. Inthe Tegemeo sample, about 25 percent of households escaped poverty over the period 2000- 2007. About one fourth stayed poor and about one in ten became poor, indicating both poverty persistenceand additional vulnerability to falling into poverty during a period of overall improvements. 109 Figure 3-10: Rural Poverty Transitions, 1997-2007 Stay Poor S tay Non-Poor P oor to Non-Poor Non-Poor to Poor Source:Nyoro et a12008. 3.96 Were rural households able to significantly increase their incomes as opposed to moving around the poverty threshold? The Tegemeo study examined whether a household i s able to transit from the bottom to the top quintile over the decade to 2007. Perhaps not surprisingly, Figure 3-11 shows that only a negligible share of the poorest households are able to make this transition, and likewise, in the opposite way, from the top to the bottom quintile. Yet this also shows that there is significant movement, inthat only a small minority o f those who started in the bottom and top quintiles (8 and 9 percent respectively) stay intheir baseline quintile. Figure 3-11: Rural Transitionsand Persistence, Top and BottomQuintiles, (percent) 2000-2007 Poorestto Poorest Richestto Richest Richestto Poorest Poorestto Richest Source:Nyoro et a12008. 3.97 A final depiction o f transitions is used to give a broader sense of the extent of movement across quintiles over the decade. Figure 3-12 shows transitions across income quintilesand ineffect summarizes the transition matrix across quintiles. It shows that the 110 poorest quintile was most likely to stay inthe same quintile, and ifthere was movement, most likely to move the second quintile, then most likely to move to the third, and so on. Thispatternemerges quite clearly across quintiles. Figure3-12: Transitions across IncomeQuintiles 10 9 8 7 Poorest 2007 6 5 r 2 4 3 3 4 2 Richest 2007 1 0 Po0rest 2 3 4 Richest 1997 1997 3.98 A series o f regressions were estimated to test which factors are most significant in explaining both transitions into and out o f poverty, and the persistence o f poverty (Nyoro et a1 2008). Regressions which control for various observed, household and community level characteristics are used. Fixed effects control for unobserved factors at the household level and show the effects o f changes in the variables on changes in poverty.53 3.99 A step-wise procedure was adopted so as to examine the role o f specific factors. For all the regressions, versions were estimated with province dummies, but for models including the infrastructure variables, province dummies were excluded because there was not always enough variation in the infrastructure variables within province. Interestingly, province dummies do not affect the results much-which means that the province effects explain the dependent variable (in this case poverty transitions), but do not correlate as much with the observed variables used on the right hand side. This suggests underlying factors are at work at the provincial level. 53A step-wise procedurewas adopted so as to examine the role of specific factors. For all the regressions, versions were estimated with province dummies, but for models including the infrastructure variables, province dummies were excluded because there was not always enough variation in the infiastructure variables within province. Interestingly, province dummies do not affect the results much-which means that the province effects explain the dependent variable (in this case poverty transitions), but do not correlate as muchwith the observedvariables used on the right handside. This suggestsunderlyingfactors are at work at the provinciallevel. 111 3.100 The results are similar to those for the profile of poverty-in particular education and credit matter (though the importance o f the latter varies across specifications), as do changes off farm income. In terms o f off farm income, variables related to the fractions o f adults involved in businesshnformal activities and in salaried activities are important, as i s the number of crops per field. For some o f the variables (like distances), however, there is not enough variation for a given household over time for this to show statistical significance-only the dummy for whether a household is close to a motorable road is significant. 3.101 As we saw above, there is significant churning, that is movements inand out of poverty. Which factors determine whether a households transit? Tegemeo regression analysis ontransitions usedinitial conditions in 1997.54 3.102 A full set of changes were estimated, as well as province dummies (Figure 3-13 summarizes). The results confirm the story in Part 11, but help to isolate the most significant and robust drivers over the decade: 0 From a policy point of view, the factors associatedwith households emerging out o f poverty are of special interest. The regression analysis shows that education, wealth (land) and access to extension matter, as well as income diversification off farm-both business and salaried activities. Among those households who suffered poverty throughout the period, initial conditions were key, in particular lack of access to credit, lack of access to infrastructure and less land. Often bad weather was additionally correlated with the persistenceofpoverty over time. On the other hand, for those who managed to stay out of poverty throughout the decade, the key factors were education, diversification, land, credit, and infrastructure. Inparticular, the distance to an extension agent is a significant correlate. In terms of the initial conditions in 2000, all the infrastructure variables (fertilizer, piped water, extension, motorable road) and business activities are important. There are also some significant province differences. 54 Resultsusingjust the post 2000 data and initialconditions for the year 2000 are similar, except that the measureofdiversificationappears to havebecomemoreimportantovertime. 112 Figure 3-13: Summary of effects of changes in key variables over time, on poverty (fixed effects) Dummyfor whether a Fractionof adults in Dummyfor whether a HH is closeto a motorabl Fractionof adults inthe HH with completedprimary education Fractionof adultsinthe HH with some primary education Fractionof HHthat is female, aged 17-39 Fractionof HH that is male, aged 17-39 Fractionof HH that is female, aged 6-16 Fractionof HHthat is male, aged 6-16 Fractionof HH that is female, aged 0-5 -30 -20 -10 0 10 20 30 40 50 3.103 Another way of looking at transitions is to ask the question what, conditional on being poor in 2000, affects the probability that a rural Kenyan household would have escaped poverty in 2007? For this sub-sample (538 households), few changes are significant, because there is not enough change over time that varies much across households. For this reason, many of the variables that are significant in other specifications lose explanatory power. For those who escapedpoverty, we do see: 0 Initially higher levels of diversification and good weather are important (or the converse for being poor in2007). 0 Changes in education and distances to markets and infrastructure matter, in particular distance to the closest fertilizer seller and the distance to piped water. Distance to the nearest fertilizer seller can be seen as a proxy for market access (Suri 2007), though it matters more for agricultural households. It also captures, at least inpart, the costs ofusing improvedmethods (fertilizer and seed). 3.104 How does distance to extension services affect usage, and what i s source of advice? Tegemeo analysis shows a positive correlation between proximity and seeking advice: about three-quarters of the quintile of households that are closest in distance to extension agents sought advice in 2007, compared to about half of the quintile that are furthest away. Figure 3-14 shows that conditional on seeking extension, it was most fiequently public agents whom were asked for advice-about half of cases of those who sought advice went to public agents, and about 10 percent went to private agents or neighbors. 113 Figure 3-14: Conditionalon Seeking Extension, Who Was Asked for Advice, 2007 60 4- cn 40 30 - 20 First Advisor Sought 10 tSecondAdvisor Sought 0 d U Source: Nyoro et a12008. 3.105 We see a very strong upward trend in diversification over the decade. Which factors are associated with rural households successfully being able to diversify their sources o f income? The key determinants o f households' participation in off-farm activities are availability o f credit, education and access to roads (Nyoro et a1 2008). Age-sex composition of the household also matters, in that households with more adults in the age range 17-39 are more likely to be involved in off-farm activities. There are also large differences across provinces, in that households in Eastern appear to diversify muchmore frequently thanthose inNyanza and Western. 3.106 Infrastructure variables are important - in that households that are closer to a fertilizer seller, closer to a motorable road, closer to piped water grow more crops (in absolute number) are more likely to diversify. That fertilizer access is driving off farm diversification may be associated with its correlation with access to technology and markets, as well as possibly enabling adults to produce crops more efficiently and spend more time working off-farm. The good news is that access to all these key services improved over the decade, and more rapidly for the poor. The initially observed gaps in access hadclosed by the end o fthe period Figure 3-15. 114 Figure 3-15: ReducedDistancesto Various Infrastructure, 1997 2007 - 9 T 1,6 8 t L 4 7 1,2 6 1 5 4 3 2 0 - L O -Poorest 1997 2000 2004 2007 +P -- Fertilizer - Fertilizer -Poorest Extension +Richestt --- --Richest Extension oores t Motorable Road R ic hes Motorable Road Source: Nyoroet a12008. 3.107 Looking more specifically at trends in crop diversification, we see that the number of crops planted inthe main season by province has increaseddramatically over time-the extent of increase ranged from doubling over the decade in Eastern and Central, and close to tripling inNyanza and Western. The number of crops per field inthe main season dramatically increased, doubling in Western, and increasing by about half elsewhere. Interms of determinants, on the negative side, the number of crops tends to be fewer among householdswith more uneducatedadults, andthose with less land. 3.108 Finally, but not least, the Tegemeo panel is utilized to examine sources of change in inequality. By looking at the marginal effects of various sources of income on the income gini, we can also identify sources of increased inequality over time. Estimates suggest that a 1 percent increase in off-farm income increases the gini by 0.0325, whereas livestock and crop income have effects which are smaller in magnitude butwhich tendto reduce inequality (-0.01 and -0.02, respectively). 3.109 Interestingly, the disequalising effect of off-farm income tended to increaseover the decade to 2007. Specifically, the marginal effect of off-farm income on the gini rises from 0.039 to 0.058, for livestock income changes from -0.013 to -0.001 and for crop income changes from -0.0255 to -0.057. This suggests that rising off farm income would tend to be associatedwith worsening inequality, whereas other sources of income tend to offset inequality. Livestock income is not very important, but crop income is an important driver. 3.110 For off-farm income, it is useful to distinguish between salaries (and remittances), versus informal business activities (note we can only do this from 2000 onwards). Salary and remittances have a large disequalising marginal effect (0.0487 in 115 2007), informal income a much lower positive effect (0.0088) and livestock and crop income similar to earlier (-0.0006 and -0.0570, respectively) (Figure 3-16). Figure 3-16: Effect of Marginal Changes in Selected Sources of Income on Inequality n 0.08 0.06 .. - 0.04 0.02 . 0 ---- Off Farm Livestock -0.02 -0.04 -0.06 -0.08 1997 172007 Source: Nyoro et a12008 3.111 These marginal effects are interesting, although caution is neededin interpreting causality. It may not be the off-farm effect per se, but rather the underlying (unobserved) factors that facilitate the householdsparticipation inoff-farm activities that are associated with inequality. It should also be underlined that some variation in the distribution of opportunities will tend to characterize any development process-the issue i s more about how to broadenthe distribution of opportunities, rather thancurtail inorder to equalize. 3.112 To conclude, the analysis of rural dynamics suggests some good news and some evidence of positive effects of government programs. But it is important to underlinethat there remain a large group of chronically poor, for whom initial conditions are still important-viz, ifpoor at the outset, muchmore likelyto bepoor adecade later. Those who stay poor are characterized by low education and lack of access to basic infrastructure. TRENDS INTHE URBANSLUMS-INSIGHTSFROMVIWANDANI AND KOROGOCHO 3.113 Part I1provided an overview of poverty levels, inequality, and living conditions inthe Nairobi slums. It painted a pictureof a poor working population that is relatively educated, whose children are schooling at relatively high levels, but that lacks access to basic sanitation services such as garbage disposal and toilets. Not surprisingly, child under 5 morbidity and mortality outcomes are among the worst in the country. Still, 116 there have been some improvements between2003 and 2006, which are explored in this section, along with the drivers of changes inpoverty and exit out of the informal slums.55 3.114 The incidence of poverty fell from 73 to 62 percent over the three year period, and this improvement appears to have benefiteda broad range of household types. This was also associatedwith an increase invarious household assets. For example, ownership of radios increased from 70 to 82 percent, of televisions from 11 to 24 percent, and o f (mobile) phones from 9 to 47 percent. Adolescents aged 15-19 years in the slums also witnessed a very large increase in enrolment from 44 percent in 2003 to 77 percent in 2006. However, the poverty gap remained large-at 36 percent. 3.115 While the level of amenities continued to be very poor through 2006, there were some improvements after 2003. For example, households having access to garbage disposal services increased from 2 to 10 percent, access to electricity from 11 to 32 percent, and the 20 percent of households having access to improved (shared) toilet facilities in 2006 constituted an increase from 8 percent in 2003. Lastly, the level of crime also seems to have dropped sharply. In 2006, less than 3 percent o f household reported having experienced a mugging inthe year prior, compared with 9 percent inthe past 4 months in2003.56Theft similarly dropped from 9 to 7 percent. 3.116 Exitrates from the slums are very high. On average across household types, 62 percent of members observed in 2003 were not observed 3 years later (Le., exited the slum). The same proportion holds when looking just at the household head. Not surprisingly, mobility i s the greatest among single male and female adults (76 and 71 percent, respectively). Mobility i s the lowest among members of households headed by widowed and divorced women (49 and 51 percent, respectively), which i s likely a reflection of their lower economic opportunities inNairobi as well as limited access to land amongthis group inthe rural areas. 55This is based on Poverty Dynamics and Mobility in NairobiS Informal Settlements, APHRC Working Paper(2008) by de Laat, J., 0.Faye, T. Suri, andE.Zulu. 56The change of time fi-amemay have led to some confusion among respondents.However, note that the 2006 timeframe (with lower reportedcrime) hada shorter recall window. 117 Figure 3-17: Proportion of Slum ResidentsObservedin 2003 but not in 2006 (By householdheadtype) Source: de Laat et a1(2008). 3.117 The poverty incidence in2003 among those who were observed inbothrounds was higher than those who exited; 77 percent o f stayers versus 71 percent o f those who exited. Similarly, poverty incidence in 2006 among those who stayed was also higher than those who were observed in2006 but not in2003 (entry): 64 versus 58 percent. This i s consistent with the observation that there i s a poor group of households, such as those headed by widowed or divorced women, who are less mobile Figure 3-17 3.118 Even among slum residents who did not exit, we find considerable number o f transitions, both into and out o f poverty. Inparticular, despite the overall decline in the headcount, among the minority who hadbeen non-poor in2003, only about halfwere still above the poverty line in 2006. Conversely, almost a third (31 percent) of those poor in 2003 was no longer below the poverty line in2006 (Figure 3-18). Figure 3-18: Transitions in Nairobi Slums, 2003-2006 Above poverty line 2003 Below poverty line 2003 Source: de Laat et a1(2008). 118 3.119 To identify the correlates of poverty dynamics in these informal settlements, controls include household composition and head characteristics, based on 2003 values. The results show: 0 Households that were poor in 2003 were only 8 percentage points more likely to still be below the poverty line in2006, confirmingthat there i s substantial income mobility. Among the households that were poor in 2003, several initial conditions are significant predictors of whether the household i s below the poverty line in2006, namely: o age and duration in slum of the head reduce the likelihood that the household will be poor three years later-which suggests that longer duration inthe slum i s not an obstacleto moving out of poverty; o households with infants (0-4 years) present in 2003 or children (up through age 17) are more likely to be inpoverty three years later; and o interestingly, relative to householdswhose headwas marriedjoint in2003, divorcedseparated household heads in2003 are 10 percentage points less likely to be inpoverty in2006. 3.120 Somewhat surprisingly, human capital variables in 2003 are not significant predictors o f poverty in2006 (again conditional on beingbelow the poverty line or not in 2003). Nor does the death of a household member over the period increase the likelihood of being inpoverty. 3.121 Finally, econometric results showing determinants of exit are reviewed. It is recalled that about 80 percent of residents in these slums are individual tenants. The results suggest: surprisingly, poverty status in 2003 is not a significant predictor of whether a household was likely to have exitedby 2006; 0 the ethnicity fixed effects show hardly variation across ethnic groups interms of who exits; there is, however, substantial variation across villages within the slums; household heads who were older in2003 andwho had inthe slum longer were less likely to exit; whereas, relative to joint households, household heads who were never married, or married but not living with their spouse, were more likely to have exited by 2006; importantly, the presence of children of all ages makes households less likely to exit whereas the presence of adults does not. Coupled with the fact that the presence of children makes households more prone to poverty, this suggests that householdswith childrenmay not have the resources to leave the informal settlement andmove to locations with better amenities; household shocks such as mugging and robbery that happened the year prior to 2003 are not correlatedwith exit inthe 2003 to 2006 period; household heads are much more likely (27 percentage points) more likely to exit ifahouseholdmemberaged 5 or older diedinthis period; and 119 none o f the humancapital variables are significant predictors o f exit. 3.122 Exit from the slum seems very sensitive to access to amenities at the village level. For example, while households with a private toilet in 2003 were 11 percent less likely to have exited, they would be 76 percent less likely if everyone inthe village had a private toilet. Similarly, while household access to garbage disposal services or electricity in 2003 i s not associated with changes in exit by the household, an increase in the share o f households within the same village having access to electricity by 10 percentage points makes the household head 12 percent less likely to exit. With every 10 point increase in the proportion o f households within the village who do not need to dispose their garbage in a public space or the river, the household head is 10 percent less likely to exit. dispose their garbage in public between 2003 and 2006 reduced exit rates by 7 percent. VULNERABILITY 3.123 The foregoing analysis has highlighted the significant churning in and out o f poverty in Kenya. Over the decade to 2007, overall rural income poverty fell, but more thanhalf ofhouseholds experienced poverty at some point. Shocks may be an important driver of households falling into poverty, or deeper into existing poverty. This emerged as an important theme inthe PPA IV. Such shocks can arise from many sources, which are examined below, and can affect households' monetary wellbeing through loss o f productivity and/or increased costs o f living. Households hit by shocks may respond in ways that affect future wellbeing (e.g., if assets are sold or children taken out o f school). Inthis section we examine the prevalence andprofile ofshocks, usingthe KIHBS, which is complemented by reporting on findings of more detailed analysis o f the extent o f insurance to shocks. However, we should underline that the data presented was collected prior to the crisis inearly 2008, which itself created a major shock for many families. In Part IVYwe review social protection interventions inthe Kenyancontext. 3.124 Inthe KIHBS, 85 percent of households had reported at least one major shock, o f any type, over the preceding five years, o f whom about half had experienced three shocks. The overall frequency o f shocks i s evidently higher for the lower deciles. More than nine out of ten households in the bottom decile experienced at least one shock (Figure 3-19). Figure 3-19: PercentageofHouseholds that Reportedat LeastOne Shock, by Decile, 2005/6 I I 9s 120 3.125 The most common shocks over the period 2000-2005 were, in order of importance, food price inflation, droughts/floods, illness, and death in the family, followed by the death of livestock (Table 3-18). Examinedby quintile, weather related shocks are by far the most frequent, followed by food price effects, then health. HIV shocks are reportedly few, and more often inthe higher expenditure quintiles, although it i s possible that these are included under deaths, which affect lower quintiles more. Table 3-18: Types of Shock in the Past five Years, Percentof HouseholdsReportingAny Shock Expenditure Quintile Type of shock Poorest 2 3 4 Richest All Drought or Floods 41.2 29.9 23.7 18.4 11.4 24.9 Crop diseaseor pests 6.3 5.0 6.2 5.5 3.9 5.4 Livestock died / stolen 26.4 23.6 18.3 17.1 9.8 19.0 Business failure 5.5 7.2 7.5 8.3 10.0 7.7 Loss o fjob or salary 4.5 4.4 5.7 7.7 10.5 6.6 Endofassistance or remittances 3.5 1.7 2.4 1.7 2.2 2.3 Large fall incrop selling price 5.7 8.6 8.5 7.6 4.5 7.0 Large rise inprice o f food 32.8 30.7 33.5 29.2 28.8 31.0 Agricultural inputprices 8.3 11.6 12.2 11.5 7.4 10.2 Chronic/severe illness or accident 25.2 25.3 28.8 26.3 20.6 25.2 Birth 3.5 3.3 2.9 2.2 1.5 2.7 Death o f householdhead 4.7 3.8 4.5 4.3 2.4 3.9 Death o f working adult 2.5 1.6 2.3 2.0 1.5 1.9 Death o f other family member 20.6 20.7 23.8 23.4 24.8 22.7 Break-upofthe household 2.1 2.2 2.0 2.2 2.1 2.1 Carjacking/Robbery/burglary/assault 2.6 3.9 4.1 3.8 8.7 4.6 Jailed 2.5 1.8 3.1 1.8 2.0 2.2 Fire 1.6 2.7 1.4 2.5 0.9 1.8 Dwelling damaged, destroyed 2.3 2.5 2.4 1.6 0.6 1.9 HIV/AIDS 0.4 0.7 0.6 1.1 1.7 0.9 Severe water shortage 13.2 10.0 8.4 7.7 8.1 9.5 Other shocks 4.3 4.3 5.0 5.4 6.4 5.1 Source: Staff calculations based on KIHBS. 3.126 Are there provincial patterns in shocks? Crime i s more of an issue in Nairobi and Western. Events that would be expected to negatively affect agriculture- droughts/floods and water shortages-especially affect households in Coast, Eastern, North-Eastern and Rif Valley, and in Coast and Western, crop disease also shows up. Nyanza was especially affected by high rates of chronic illness, whereas households in North-Eastern reported to be adversely affected by the withdrawal of regular assistance (including relief). Food price inflation was an important issue everywhere, with only Nairobi reporting ratesbelow 23 percent (Table 3-19). 121 Table 3-19: Percentage of Overall Households (Unweighted) Severely Negatively affected by Specific Shocks, 2000-2005 Nairobi Central Coast Eastern North Eastern Nyanza RiftValley Western HIVI AIDS 1% 1% 0% 1Yo 0% 1Yo 1Yo 1Yo Dwelling damaged 0% 1% 2% Fire Endofregular assistance Crop diseaselpests Crime victim Other Job loss Crop price fall A g input price rise Business failure Water shortage Livestock diedlstolen Illness I accident 15% 21% Source: Staff calculations based on KIHBS 3.127 To complement the foregoing, we show the frequency o f severe shocks for the sample as well as rural-urban, poor-non poor and top versus bottom decile. This reveals some interesting patterns: first, that almost half o f households experienced three shocks inthe preceding five years; second that the frequency o f shocks is muchhigher inrural areas; and third, the richest also experience a large number o f shocks. About one fifth o f the poor experienced three shocks-and less than 10 percent did not report any (Figure 3-20). Figure 3-20: Number of Different Shocks in the Preceding Five Years, 2005/6 I 60 50 , I I 40 30 0Top Decile I 0 Bottom Decil 20 Non-Poor 0 Poor lo 1 2 3 4 5 I Source: Staff calculations based on KIHBS. 3.128 To simplify the analysis, we adopt a rough typology o f shocks that are common inKenya, namely: Economic-these are mainly related tojob loss, or increasedprices. 122 Randomevents, related to adverse weather, can cause crop failure and death of livestock among the direct effects. Eastern andNorth Easternprovinces are heavily affectedby these shocks. Health-related risks, such as chronic disease, injuries and death o f family member (due to e.g. TB, malaria or AIDS. Health shocks impact households through both decreased labour-supply and sometimes significant out of pocket medical expenses. The long-term effects o f such shocks, as a result for example o f reduced schooling for children inhouseholds where a caregiver is chronically sick, can be significant. Graff Zivin, et al. 2006 found decreased school attendance associated with aids morbidity, though mortality, as they receive ARVs. Risks related to crime, violence and human security was o f growing concern to Kenyan communities and households even before the December elections. Statistics from 2006 showed that Kenya ranks among the most crime-prone countries in Africa, with this increasingly being a rural as well as an urban phenomenon.Crime and violence has not only a direct economic impact on poor households, but can also create psychologicaltrauma. 3.129 The distinction between idiosyncratic and covariate shocks is important to note. An example o f the former would be illness in the family, and examples o f covariate shocks would be the post-election violence, an economic downturn and food inflation, as well as, typically, droughts and floods. 3.130 Looking at the share o f households reporting shocks in the past five years, economic shocks are by far most common, followed by random or weather related events. This is true for the whole population, as well as for the poor, and the bottom decile (Figure 3-21). Figure 3-21: Average Number of Shocks over the Past Five Fears, by Decile and Poverty Status, Kenya Poor Non-Poor Top Decile Bottom Decile Source: Staff calculations based on KIHBS. 3.131 Finally, the first response to a shock is examined for the five most frequent shocks inthe sample, namely food inflation, illness, droughts/floods, death inthe family 123 and livestock loss (Figure 3-22). The most common responses are similar and can be listed inorder of importance as: * spend cash saving; *sell assets (animals); * work longer hours; * reduce food consumption; and * receive help from family and friends. 3.132 The relative importance of each mechanism varies by the type of shock-for example, family and friends help out most at the death o f a family member and cash saving is usedmost ifa householdmember has a chronic illness or an accident. Very few households borrow in the faces of shocks-about 3 percent reported accessing credit in the face of a shock. Governmentassistancewas also very rare. - 8 Other UReduced non food expenditures Received help from farnilylfriends 1Drought or Livestock Floods lwl;t;ioe/ Large ris in pd) 4ilh;Il;r~ot~r;{ hronidsev e Death of IM Spent cash saving ~~ Source: Staff calculations based on KIHBS. Figure 3-23: Reported Severity of Shocks, as Percentage of Monthly Poverty Line, 200516 120 100 80 Kenya 60 * Poor Non-Poor 40 Bottom DeciIe Top DeciIe 20 0 Total Loss Demographic Random Crimeshocks Economic Health Other Shocks Shocks Shocks Shocks Shocks Source: Staff estimates based on KIHBS. 124 3.133 The foregoing results provide a sense of magnitudes, but are merely illustrative in that we are simply presenting cross-tabulations for a single point in time, using retrospective reporting. More detailed analysis has been undertaken by Suri (2008) for Kenyan rural households to investigate whether or not households are able to smooth their food consumption over time inthe face of shocks. The government generally does not assist families affected by shocks. Does this matter or can households smooth consumption without public assistance? 3.134 There are many studies which have sought to measure the extent of risk-sharing ina developing economy, indeedthe literature on this topic is vast (see Morduch, 1995) for an overview) -perfect risk sharing implies that all individual shocks can be smoothed, away, but aggregate shocks cannot. Suri (2008) finds, basedon analysis ofthe Tegemeo panel and data on rainfall shocks for the period 1997 -2004 that for maize consumption and crop consumption, there is close to perfect risk-sharing for idiosyncratic shocks. It also seems that households manage to smooth risk well within villages. Villages, on the other hand, are rather far from perfectly insured, even ifthere is evidence o f some risk-sharing betweenvillages. Figure 3-24 shows the extent to which individual households were able to smooth maize and food consumption over time in the face of shocks. The regressions used household consumption in logs and control for individual fixed effects, to focus on co-movement of changes in household consumption with changes in group (village or district) consumption. The figure shows the standard error bands, which indicate that the confidence intervals are muchwider at the district level. Figure 3-24: Extent of ConsumptionSmoothing,Within Villages and Across Villages within Districts, 1997-2004 (Co-movementof 1Means Perfect Risk Sharing) 1 2 1 0.8 0 6 0 4 0 2 0 Withm Villages Wthm Dlstncts Source: Suri, 2008. 3.135 Hence we finish up this section by underlining that shocks are a fact of life in Kenya, for many households. It is of course a concern that the poorest are affected so frequently by shocks. Several of the shocks that were highlighted can be addressed, at least to some extent, through economic management (re inflation), broader policies to encourage growth (and reduce the incidence o f job loss) and improved governance and economic conditions more generally (to reduce crime and security problems). For 125 weather related shocks, which are frequent, the evidence suggests that rural, non-pastoral households in Kenya are able to smooth food consumption extremely well; however, once these shocks become generalized-that is, beyond the village, to the district or province, smoothing does not take place through inter- household transfers or self insurance. This points to the role of publicly supported safety nets when shocks are large enough that they are aggregate to districts or provinces. We returnto this theme in the next Part, under social protection. CORRUPTION AS A DRIVER POVERTY5' OF 3.136 There is a vocal consensus that combating corruption is one of Kenya's most critical governance and development challenges. International assessments continue to rank Kenya as one of the most corrupt countries in the world. Kenyans themselves put corruption an issue of foremost concern. 3.137 Much public discussion in Kenya about corruption makes unequivocal connections to poverty. Corruption is especially associated with inequality-in that upward social mobility and accumulation of wealth. While robust empirical evidence of these seemingly obvious links has been lacking, we do have results which are at least indicative, as describedbelow. 3.138 The use of public office for private gain is the most widely accepted definition of corruption. This definition admits a broad range of abuse of power, ranging from briberyto embezzlementand fraud, to subversion of policy, legislation and institutions to serve private interests. Scholars and analysts have proposed different taxonomies of corruption. A popular taxonomy, which we use here, characterizes corruption by scale, as petty, bureaucratic, or grand. We focus here largely on petty corruption, which refers to relatively small but frequent bribes paid to obtain routine administrative and social services. The culprits are typically frontline officers, e.g. health workers, police officers on the one hand, andthe general citizenry onthe other. 3.139 Corruption undermines growth in two ways. First, it is inefficient. It absorbs resources, for instance, management time spent negotiating bribes. It slows down business, for instance transportation and cargo clearance. This puts firms in corrupt environments at a disadvantage, particularly with regard to competing in the global market. Second, it distorts resource allocation, by diverting scarce resources, including capital and human resources, from productive activities to "rent seeking." These tend to reduce productivity and investment. 3.140 Corruption can be regressive in several ways. Wasteful spending and cost escalations can crowd out pro-poor spending; and tax evasion reduces the resources available to finance public services that directly benefit the poor, and undermines progressive taxation. Bribing for services is one of the most ubiquitous forms of corruption. This premium hurts the poor more than the better off. A bribe of one hundred shillings to obtain a "police abstract" for instance, may be inconsequential to a ''Thissectiondraws on analysispreparedby DavidNdii. 126 middle-income person, but meanforegoing a meal for the millions o f Kenyans below the poverty line. 3.141 Cross country econometric studies consistently find statistically significant differences in investment and growth, inequality and poverty across countries (source). A 2-point difference on a corruption scale o f 1 (high) to 10 (low) accounts for 4 percentage point difference in investment rates, which in turn translates to a half percentagepoint difference ingrowth of per capita incomes. The impact has been found to be even more significant on the incomes of the poor. An increase of one point, on a scale of 1to 10 is associatedwith a 10percent difference inthe growthrate of the income of the bottom20 percent of the population (source). The findings also suggest significant impact on or at least correlation with, inequality. 3.142 These differences are quite significant. Indeed the growth impacts imply that the difference between a score of 6, versus 2 where Kenya is ranked on Transparency International's CPI, effectively costs Kenya one percentagepoint o f economic growth per year. A 3 percent per capita growth rate will take 24 years to double a per capita income of US$500, a growth rate of 4 percent per year cuts this to 18years. 3.143 Kenya has suffered grand corruption on a large scale, the most infamous cases being the Goldenberg and Anglo-Leasing scandals. Apart from undermining public trust, these episodes have been costly. Most obvious is the opportunity cost in terms of public expenditure. These are compared to the Government outlays on education and health in the respective periods (Table 3-20). The cumulative losses identified in CAG audits i s three times education outlays, ten times health outlays andjust over twice the combined outlays over the period. The Goldenberg fraud alone cost the equivalent of 60 percent of the education outlays, and just over twice health outlays and about half the combined outlays. The potential loss from Anglo-Leasing amounts to 15 percent of the education outlays, just under half of the health outlays and about 12 percent of both, while the actual losses were equivalent to 5 percent, 12 percent, and 3.6 percent of education, health and the combined expenditures respectively, aggregated for the whole period. More generally, perceptions of political risk increased significantly as a result o f these scandals, especially since the audit findings show that while Goldenberg was the largest, itwas byno meansthe only major fraud going on atthat time. Table 3-20: Relative Financial Costs of Grand Corruption Episodes in Kenya KSh. Cost as % ofpublic spending billions Education Health Both Goldenberg(1991-93) 27 58 217 46 CAG (1991-97) 475 295 968 226 Anglo Leasing(1997-2004) 14.5 3.9 12.2 2.9 Total(1991-2004) 519.5 94 346 74 Note: The health and educationexpendituresare cumulative for the correspondingperiodinbrackets. 3.144 Household surveys suggest that corruption at lower levels of government is pervasive in Kenya. According to TI-Kenya's annual survey, in 2006 Kenyans encountered bribery inone out of every two official transactions. They paid, on average, 127 2.5 bribes per personper year. Just over a third (36 percent) o f the bribes were relatedto law enforcement issues, primarily to police andjudicial officers, 29 percent were paid to obtain services, 24 percent in matters related to regulatory compliance (e.g. licenses). Business and employment related matters accounted for 7 percent and 3 percent o f the bribes reported respectively. Another survey conducted by the Kenya Anti Corruption Commission, also in 2006, reports broadly similar findings. One third o f the survey's respondents who sought public services paid a bribe. 3.145 The experience o f bribery i s fairly frequent across all socio-economic groups. Table 3-21 presents the percentage o f respondents in the category reporting bribery experience (Le., were asked for or offered a bribe) e.g., 53.7 percent o f urban youths aged 18-24 who were interviewed encountered bribery during the year. A couple o f stylized facts do emerge. First, the incidence is significantly higher among urban residents, because urban residents deal with officialdom more e.g. most rural people do not have electricity or water connections, hence they do not deal with utilities. Secondly, the incidence among low income groups is lower than average in rural areas, but significant higher than average in urban areas. This suggests that the urban poor are especially adversely vulnerable to bribery encounters. Table 3-21: Annual Bribery Encounters, Share of Group, 2006 (percentage) Rural Urban National Youth 18-24 41.4 53.7 47.4 25-29 51.4 62.9 57.5 Type of Employment Unemployed 45.8 59.0 51.4 Self-employed 50.0 56.8 53.3 Family enterprise (farm & nonfarm) 53.2 67.9 60.6 Private sector (business) 59.2 54.1 55.8 Non-profithommunity 62.5 52.6 57.1 Public Sector 55.6 53.7 54.5 Monthly Income (KSh) Less than 4,999 42.5 61.5 48.6 5,000-9,999 47.1 61.9 54.4 10,000-24,999 62.4 50.5 55.8 25,000-49,999 48.1 53.4 52.0 50,000 + 33.3 48.3 47.2 Total 48.4 58.0 53.0 Source: National Corruption Perception Survey 2006 by the Kenya Anti-Corruption Commission. 3.146 The 2006 National Corruption Perception Survey found that the poor spend a higher share o f their income inbribes. Indeedthere is an inverse linear relation between the burden o f bribery, and level o f income. Indeedthe burden o f bribes i s comparable to the share o f household spending on water, for example. Onthe other side, the burden is still notable, but not large, for upper income groups. 128 Table 3-22: ReportedAnnual Expenditureson Bribery, By Reported Income Monthly Income Expenditureon bribery p.a. % of income(median) 1000andbelow 1549 1001- 5000 1673 4.6 5001-10000 2731 3.0 10001-25000 3693 1.8 25001-50000 4015 0.9 50001-75000 5955 0.8 over 75000 8745 All 3185 Note: Missingbecausethere i s no median(Le. denominator) inthe open ended income categories i.e. below 1,000 and above 75,000. Source: National Corruption PerceptionSurvey 2006, Kenya Anti-Corruption Commission. 3.147 There is good news, however. Recent evidence shows that reformingagencies have made progress inreducing reported levels of corruption, and there are signs of reducedtolerance for corruption. 3.148 We now turnmore directly to addressthe implications ofthe poverty and inequality diagnostic for policies and institutions inKenya. 129 4. POLICIESAND INSTITUTIONS TOADDRESS POVERTYAND INEQUALITY 4.1 It is impossible for this report to cover the full breadth and depth o f the policy and program agenda implied by the poverty and inequality challenges facing Kenya. Instead we address only selected aspects-viz. fiscal policy, with an emphasis on allocations and distributive dimensions, governance, access to finance and land, youth, as well as monitoring and evaluation. We begin by highlighting key points on the rural development and related growth agenda, which are covered more fully in companion reports beingpublishedconcurrently by the World Bank. 4.2 The Medium Term Plan (MTP) for the period 2008-2012 provides the government's medium term policy and program framework, within the context o f Vision 2030. The stated priority i s poverty and inequality-and the related themes o f provincial variation, youth and gender-as laid out in the President's foreword and the Minister's statement. This is clearly very welcome. However the centrality o f this goal i s not carried through to the actual program, which treated poverty as a more marginal concern, limited to some aspects o f the social pillar. It is expected that this assessment should help provide guidance as stakeholders proceed to review the plan more fully, and prioritize the actions therein. GROWTH AGENDA-EXPANSIONOFINCOMEEARNINGOPPORTUNITIES 4.3 Prolonged periods of slow growth and stagnation, compounded by periodic contractions inthe economy, have limitedthe pace o f poverty reduction inKenya. Indeed, over the quarter century since the early 1980s, the decline in the headcount has been minimal, and the absolute number o f poor inKenya has risen. 4.4 The economic recovery since 2003 brought benefits in terms o f poverty reduction. However because our baseline predates the recovery by several years, we do not yet know how the payoffs o f macro level achievements have been distributed across households. We do know, however, that levels o f inequality in Kenya were already high in 2003, when the economic recovery began, and there has been no abatement over the period 1997-2005/6 overall. 4.5 UsingMAMS, a model developed at the World Bank for the analysis of MDG and development strategy options, we simulate Kenya's economy for a set of scenarios for the period 2007-2030 with a focus on an initial period, 2007-2012.58The purpose i s to 58For more details see, Bourguignon, Diaz-Bonilla and Lofgren (2008). Additional simulation results are providedinAnnex Tables. 130 explore impacts of alternative policies and exogenous conditions on growth, selected MDGs, and inequality. MAMS models the evolution of these key outcomes, based on exogenous changes and linkages based on labor and capital endowments. The numerical results should be viewed as approximate given uncertainty about parameters and, more broadly, the important roles of local context, and policy specifics that are not included. Nor do these simulations take into account the evolution o f inequality over time, inaccess to assets and services, which can also be influenced by institutional choices and functioning (Box 4-1). Box 4-1: Growth Essentialfor Poverty Reduction-Equity and Equality of Opportunity Essentialfor Sustained Growth Slobal experience suggests the strong link between overall economic growth and the speed of poverty reduction. A recentWorld Bank study of 14 countries confirmedthat the pace of overalleconomic growth is the main factor that determines how quickly poverty declines. The cases included Indonesia and Vietnam, Bangladeshand India, Tunisia and Brazil, as well as Uganda and Senegal. In these countries, successful sustained and rapid economic growth strategies proved to be successful poverty reduction strategies as well. Yet equity, Le., equality of outcomes such as income, health, and security, and equality of opportunity are essential ingredients of sustained growth strategies. The benefits of brisk growth generally are spread widely but not evenly. The rural poor do gain. But the experience of sustained growth in the modern era clearly suggests city-dwellers gain more-and to some extent this is inevitable. As people move across the rural-urbandivide, measured inequality will tend to increase. Peoplecare about bothkinds of equality, but they will tolerate unequal outcomes provided governments take steps to contain them. If one group is persistently and flagrantly excluded ftom the fruits of growth, country experiences show that they eventually find a way to derail it. Conversely, evidence from many countries suggests people will make great sacrifices for the sake of economic progress if they believe their children and grandchildren will enjoy a fair share ofthe rewards. Recent investigationshave underlinedthat various forms of inequality can matter. Inparticular, disparities across groups, alsoknown as horizontalinequality, can become entrenchedand self-perpetuatingover time, especially in the absence of corrective public actions (Stewart 2002, 2008). These in turn have wider adverse repercussionsnot only for the inclusivenessof growth, but also for its sustainability. International experience, including the cases of Ireland, Sudan and Rwanda, suggest that high group inequality can increases the riskof conflict(Stewart 2008). Hence, global experience suggests that governments must safeguard equality of opportunity and contain inequality of outcomes. Equality of opportunity has been shown to be best served by providinguniversal access to public services like health and education, and by meritocratic systems in government and the private sector. The latter goal may be served by redistribution through the tax system, and appropriate spending programs, includingthe funding of service provisionand public sector investment. At the same time, the full impacts of measuresto redistributeneedto be considered, includingpotential adverse effects on incentives, investmentandrisk taking. Source; Pro-PoorGrowth inthe 1990s. WorldBank. 2005. The GrowthReport: Strategies for Sustained GrowthandInclusiveDeveloDment.CommissiononGrowthandDevelopment.World Bank.2008 4.6 Inthe MAMS analysis, for the base scenario, which serves as a benchmark for comparisons, we assume growth rates inreal GDP at factor cost of 4.5 and 5.5 percent for 2008 and 2009, respectively, and a higher rate o f 6 percent starting from 2010 extrapolating from the growth recovery which took place between2003 and 2007. Except 131 for the two years 2008 and 2009 under the trend scenario, GDP growth i s endogenous for all non-base scenarios. The content of the scenarios is summarized inTable 4-1. Table 4-1: Descriptionof MAMS Simulations Base Base (business-as-usual) Same as base exceptfor: Trend Trendgrowth o f6% also in2008 and2009 pet-lo Decline inpetroleum import prices 2009-2012 aid-inf Infrastructure expansion financed by additional grant aid tax-inf Same infrastructure expansion as aid-inf buttax-financed Combi Combining: petro world price decline (same aspet-lo) and grant aid increase (same as aid- inj), with government efficiency gains, allocating all savings to infrastructure 4.7 Under the base scenario, most macro aggregates grow at 5-6 percent per year (Table 4-2) a rapid pace by historical standards, maintaining relatively constant GDP shares. In per capita terms, private consumption grows at a rate of 2.2 percent annually. The exchangerate appreciates slightly over time. Table 4-2: Evolution of SelectedMacro Indicators Base Case, and Trend (In the Absence of - Temporary Slowdown), 2007-2012 Base Trend-- no slowdown in 200819 Absorption 4.9 5.4 Consumption private 4.9 5.4 Consumption government -- 5.8 6.0 Investment private 4.8 5.4 Investment government -- 2.1 3.0 Exports 5.5 6.0 Imports 4.3 4.8 GDP at factor cost 5.6 6.1 Real exchange rate (index) -0.6 -0.6 4.8 Reflecting the links modeled between incomes and access to services and outcomes, rapid growth inGDP means that MDGs 2 (completion of primary education), 4 (under-fivemortality rate), 7a (water access) and 7b (sanitation access) as well as poverty and (Gini) inequality all improve (Table 4-3).59 However under the base scenario, Kenya i s not on track to achieve any of these MDGs by the 2015 deadline. 4.9 Looking more specifically at the government's own target under the Medium Term Plan - to reduce poverty to 28 percent by 2012 - we can see the following using simple modeling techniques from the MAMS. First, given the existing distribution (gini), the rate of economic growth would have to be much higher than presentprojected, about 12 percent annually through 2012. Second, if inequality continues to worsen, as had been 59 InTable 4.3, we show results for two poverty lines, overall poverty and food poverty (see Part 11).Inthe discussion, we focus on the overall poverty line. The two indicators tendto move inthe same direction inthe differentscenarios. 132 the case over the 1997-2005/6 period overall, the rate o f growth would have to be even higher. On the other hand, iflevels of inequality were somewhat lower, say about .35, then the rate of growth needed to reach the poverty target in 2012 would also be somewhat lower. These estimates are very simplistic, but also illustrative of the rates of growth needed, and the role that inequality plays in the short term in determining poverty reduction outcomes. 4.10 The trend simulation assesses how Kenya's economy would have evolved inthe absence of the recent crisis - in 2008 and 2009 the growth rates are raised from projected values of 4.5 and 5.5 percent, respectively, to the long-run rate of 6 percent. Annual GDP growth up to 2012 is 0.5 percent higher (for the period 2012-2030, the annual growth gain i s minor, at 0.1 percentage points), and MDG indicators would also have been higher (Table 4-3). Table 4-3: MDG,Poverty and Inequality Indicators, 1990,2007,2015 Target, and 2012 2015 Trend 1990 Target 2007 Base no slowdown Combi Primary Completion 25.0 100.0 69.7 75.4 75.6 78.2 Child Mortality 97.0 32.3 97.7 84.8 82.9 68.7 Improved Water 45.0 72.5 63.4 64.7 64.9 66.5 Sanitation 40.0 70.0 48.9 52.1 52.7 58.9 Poverty 44.8 22.4 46.5 37.7 36.6 35.2 Food poverty 19.4 14.9 14.3 13.8 Gini 0.452 0.43 1 0.430 0.432 4.11 What difference do recent oil price rises make to growth projections? In 2007, oil represented 7 percent of Kenya's total import bill; for all simulations, we impose a 50 percent increase inthe real foreign-currency price of imported oil. Inone simulation, pet- lo, we gradually reduce this price by 25 percent. This terms-of-trade gain adds a quarter of a percent to the GDP growth rate up to 2012. The impact is on domestic final demand (absorption is stronger)-it grows by an additional 0.75 percentagepoints as Kenya, for a given level of GDP, is able to import more and/or export less. The real exchange rate appreciates slightly. 4.12 How government spending i s financed can affect incentives for individual economic actors as well as macro-economic variables. Additional spending could be financed by additional external assistance and/or increased domestic tax effort. MAMS allows investigation of these scenarios. Inthe simulation aid-inJ; the government receives an increase in grant aid (0.9 percent of GDP in 2009 to 3.6 percent in 2012), using the resulting fiscal space for spending on infrastructure. Under aid-inJ; GDP growth gains 0.5 percentagepoints and is accompaniedby expansion, not only in government demand, but also in private consumption and private investment. The combined impact of these changes i s a decline in the poverty rate by 0.2 percentage points by 2012 compared to base. Under tax-inJ; the government finances the same increase in infrastructure spending with higher domestic taxes (direct and indirect). The GDP growth gain is weaker (0.2 percentage points) and accompanied by contraction of private consumption and private 133 investment as the government raises additional revenues from its population. Accordingly, the poverty rate in 2012 is simulated to be 1.3 percentage points higher than for base. In sum, the financing simulations suggestthat over reliance on domestic sources can weaken economic growth, due to associated contractions in private consumption and private investment. 4.13 The final simulation, combi, brings together a series of changes to see the impacts on growth. It combines a decline inpetroleum world prices (same as for pet-lo), additional grant aid (same expansion in foreign currency as for aid-inj), with government efficiency gains (growth rate for "other government," part of which i s assumed to be unproductive, i s cut by 50 percent), using the resulting fiscal space to expand public spending on infrastructure. The results include an annual GDP growth gain of close to one percentage point accompanied by a two percentage point gain in domestic final demand growth, permitting private consumption and investment to grow. The estimated poverty rate in 2012 would be 1.1 percent lower relative to the base simulation. This simulation illustrates what Kenyacan achieve inthe mediumruninthe absence of major shocks, with strong domestic efforts to prioritize government spending, and greater support from donors. Figure 4.1 compares annual growth in GDP at factor cost 2007-2012 for the different scenarios. It indicates that the losses from the recent turmoil are significant in the medium run but that this can be offset by a combination o f domestic actions, additional aid and favorable world conditions. Figure 4-1: Annual Growth in GDP at Factor Cost Under Alternative Scenarios, 2007-2012 (percent) 6.4 6.2 6.0 5.8 5.6 5.4 5.2 5.0 base trend pet-lo eff-inf eff-hd aid-inf tax-inf combi Note: See Table4-1 for definitions. 4.14 In the fiscal analysis below, we use the MAMS to present some broad results about alternative allocations of government spending. 134 ImprovingRuralOutcomes 4.15 Our analysis has underlined that agricultural performance i s central to understandingtrends inpoverty and inequality inKenya while highlightingthat for many, diversification both on- and off-farm is key. On the positive side, export crops, domestic horticulture and livestock products have performed well over the past few years. And, as we saw, there i s good news from Tegemeo's analysis o f a representative panel o f rural households over the decade to 2007. However, overall agricultural performance has been dragged down by slow productivity growth inmaize, which accounts for the largest share o f cropped area inthe small farm sector. 4.16 Land constraints and decreasing farm size imply the need to develop and disseminate low-cost yield- and value-increasing technologies. Achieving productivity growth for staple crops like maize appears to be a straightforward proposition: enabling smallholders to access improved technologies and relevant advice. Most maize farmers are already using improved seeds and chemical fertilizers, because it i s profitable for most to do so andthese inputs are relatively easily available. 4.17 An upcoming Agricultural Development Review (World Bank 2008c) provides a more comprehensive agenda. Here we draw attention to the main obstacles to moving out o f rural income poverty, namely the lack o f useful advisory services and weak infrastructure. The cost structure i s high in part due to high transport costs and port inefficiencies. Poor roads, particularly rural roads, raise input prices and marketing costs and thus create disincentives for farmers to behave more commercially and diversify. Useful advisory services are needed, to introduce new technologies and enterprises to help raise overall farm productivity through increased commercialization. 4.18 However, policymakers need to be thinking off farm as well. One finding which comes out strongly is the importance o f on and off farm diversification as an escape route from poverty. Successful diversification also appears to be associated with infrastructure variables-in that households that are closer to a fertilizer seller, closer to a motorable road, closer to piped water are more likely to diversify-and access to credit. 4.19 There was significant agricultural liberalization inthe early 1 9 9 0 though direct ~ ~ intervention by the Government has been maintained in some areas. It appears that where government policy has facilitated the market, as inthe tea and dairy sectors, performance has been good. Among problem areas are the coffee sector, which continues to suffer from politicized small-holder marketing arrangements, and direct intervention in the pyrethrum. 135 Box 4-2: ImprovingAgricultural Productivity: FactorsAffecting Adoption Practicesof Maize Farmers Any improvementinpractices or technology has a distributionof returns, inthat not every farmer benefits the same way. The reasonsfor such distributions are various, and can be anythingthat creates a differential return to hybrid over non-hybrid(like weather, soil fertility).Understandingthese differences is important to inform policy and programs designedto encourage farmers to use these technologies. A recent study estimated the benefits to hybrid, relative to non-hybrid maize across the sample of Tegemeo farmers between 1997 and 2004 (Suri 2008). The objective was to understand why some people are not using hybrid seeds-whether this can be traced to constraints (like distances or costs), or irrational behavior on the part ofthe farmer. The studyrevealedthat farmer-specific comparative advantages inthe useofhybrid maizeplay an importantrole inexplainingyields andadoption decisions. Lookingat the entire distributionof returns to the hybrid technology for these farmers, it was possible to sort farmers with low returns fromthose with highreturns. The periodunder review is sevenyears, andthe analysis showedthat: There is a group of farmers where returns from hybridmaize would be very high, yet they do not adoptthis improvedmethod. For these farmers, it seems that the importantconstraints that prevent adoption are related to market access and infrastructure constraints, measured by the distance to the closest seedfertilizerdistributors. This group comprisesabout20 percentofthe sample. For a larger fraction of farmers-about 60 percent-the returns to hybridmaize are much smaller, but sufficiently large to provide incentives to these farmers to choose to adopt. This group can afford to use hybrid maize, even when the mean returns to hybrid are low, since it helps insure them against bad outcomes. For these farmers, since they seem to have the cash and location neededto access hybrids, they would benefit greatly from new and improvedhybrid strains, even ifthey were costlierthancurrentlyavailablevarieties. The other maize farmers, the remaining20 percent, are termed marginal. They tend to switch in and out of use of hybridmaize fiom periodto period, but obtain essentially the same yields under hybrids andnon-hybrids. Inthis important sense their failure to adopthybrids is rational. This research suggests that policies and programs to promote agricultural productivity need to take into account the different types of farm level and individualconstraints that face farmers. There is a group of maize farmers-an estimated 20 percent of the sample-who would significantlybenefit from improved seeds, but needto overcomethe distance from seed and fertilizer distributors which significantly increase the costs of seed fertilizer. Alleviatingthe constraints facingthe first, potentially productivegroup, would greatly increase yields for these farmers -but policy makersneedto bear inmindthe relatively small share of farmers. In some cases, generalmeasures may be effective, but for marginalgroups, more tailoredand targetedprogramsmaybe needed inorder to enableproductivityimprovements. Indeedthe results suggest that extensionof existing hybrids to the marginal farmers would not be effective-because of the limited returns thereto. For this group, research and development is needed to identify bettedappropriate technologies that would actually work for them. 4.20 At the same time, some trends inthe 1990s were damaging to rural development. Public spending on infrastructure, including rural roads, fell to very low levels. The reversal inthis pattern after 2003, which i s explored indetail below, i s thus very welcome. Cutbacks in government services like livestock disease prevention and control led to resurgence o f problems that had previously been contained, and spending allocated to extension services went largely to central administration rather than to meet operational needs inthe field. 4.21 The Agricultural Development Review argues for renewed focus on agricultural productivity, strengthening input and output markets, completing the shift in the role o f public sector from direct intervention to market facilitation, and increasing public 136 investments in agricultural development. It argues that Kenya's small-farm sector will otherwise face a very uncertain and untenable future. Among the cross-cutting policy recommendationsto promote rural development are the following: Improvement in rural infrastructure and market access. Productivity diversification and market performance are significantly influenced by transport costs. Over time, the performance o f major markets for agricultural products and inputs has improved significantly, but the transport costs from farm gate to primary and secondary markets remain high. Addressing inefficiencies and problems with Mombasa Port related services are also important (see also the ICA and growthreports). Strengthening the performance of traditional marketing systems. About 90 percent of food produced i s consumed domestically, largely via internal and traditional marketing channels. Public investment priorities on this front include: o upgrading the urban wholesale and retail market facilities to reduce the costs and risks to farmers andtraders and to leverageprivate investments; o ruralroads, marketinfrastructure, andstorage facilities; and o extension services, including to build the capacity for producer organizations. 4.22 While public expenditure on agricultural research, extension, irrigation, animal health and veterinary services, market information, grades and standards, capacity strengthening of extension staff, and so on, can support agricultural growth, a critical determinant of the effectiveness of these services is whether these are accountableto the farmers and dynamic and responsiveto market conditions. Progress on the LandAgenda 4.23 InPart I11we concluded that there has beenarapidworsening inland inequality in most parts of the country over the past decade-at a rate of change that suggests that land inequality is rapidly becoming more serious inobjective terms as well as interms of political salience. The predominant forms of landholding inKenya are private, communal and public lands, andparks and reserves. Official figures indicate that about 19percent is private registered land, 65 percent is communal (so-called Trust) land, which is held by county councils on behalf o f the communities. The rest is unalienated public land and parksandreserves. 4.24 Inrecent years there have beenlively public debates around land. A number of civil society organizations-such as the Kenya Land Alliance-and individuals have been engaged in the policy and legislative discourse. There has been no shortage of major policy initiatives, including: Presidential Commission into the Land Law System of Kenya ("Njonjo" Report 2002). 137 Constitution of Kenya Review Commission (Chapter on Land) Commitment for aNational LandPolicy Process. Presidential Commission into the Illegal/Irregular Allocation of Public Land ("Ndung'u'' Report, 2004). Kenya Economic Recovery Strategy for Wealth and Employment Creation, 2003-2007, which commits to implementing the Njonjo Report. National Land Policy formulation process, which commenced in 2004, and provides for a multi-sectoral approachinpolicy implementation. LandReform Support Program, which integrates landpolicy development with institutional reforms andpreparation for implementation. 4.25 The conversion of land from communal to individual ownership and registration before and after national independence involved some o f the land that i s claimed by pastoralists. This has resultedin long standing conflicts, especially in Rift Valley, and i s one of the areas being addressed by the ongoing National Land Policy process. One important avenue that has beendiscussed, but not adopted, i s taxation o f the large tracts of unusedlands. 4.26 A long standing problem has been the access of women to land especially through inheritance. While the constitution provides for equal access, there is a clause which exempts devolution of property at death. As culture and traditions support male inheritance and as family laws are not sensitive to gender, access o f women to land ownership has been adversely affected. The draft National Land Policy attempts to address this problem by taking measures to counteract discriminatory cultural and traditional practices. This would follow the footsteps of countries such as South Africa, Mozambique and Nigeria whose constitutions contain clauses that provide for gender equality with a clarification that the provisions supersede any existing legal provisions including incustomary law (Deininger 2003, pp 59-60). 4.27 Several organizations-including Ministry of Lands and Housing as well as a myriad of other central and local government authorities involved inthe administration of private lands-have been regarded as inefficient, corrupt and ineffective (Ministry of Lands and Housing, 2005, Issues and Recommendations Report of the National Land Policy). Authority for land allocation which is currently anchored in the presidency and delegated to the Commissioner of Lands is to be given to an independent National Land Commission and to district and community land boards. These organizations would be made responsible also for land administration and made more accountable to the people they are meant to serve. How this accountability would come about is a key issue requiring further attention. The Ministry of Lands, currently responsible for land administration, would be restructured and streamlined to focus on policy and regulatory issues. 4.28 Kenya now faces a major challenge of enabling a socio-economic environment nationally and locally, so that trust and mutual coexistence is the norm, and peace prevails. This is a minimum prerequisite to future growth and poverty reduction. Thus, the land sector assumes great importance and requires concrete progress in addressing real and perceived injustices. Clearly there has been much debate and policy effort around land 138 issues in Kenya. Substantial analytical work and much o f the land policy formulation process has been done. However, there i s a need to update policy proposals to reflect recent events, especially displacement and prospective returns. The foregoing evidence about levels and trends inlandinequality, and its relationto poverty, should helpto inform upcoming policy development. Mobilityinthe LabourMarket 4.29 There are no major obstacles inthe labour marketto an acceleration of sustained growth and poverty reduction. Regulations are not a major obstacle to hiring and firing workers, and skills are not a binding constraint. 4.30 The mainpolicy message to underline on the labour market side is that domestic migrationand mobility are integral to growth and poverty reduction. There is an extensive tradition of labour mobility in Kenya, and adoption of this strategy on the part of householdsappears, from the evidence, to be an important correlate of householdswho are not poor. Likewise, those districts which are more ethnically diverse tend to have lower rates ofpoverty. 4.31 It is thus critical that the recent violence and the extensive displacement that occurred not be a harbinger for ongoing curtailment of voluntary labour movements around the country. A shutting down of mobility in response to job and other opportunities not only adversely affects individual households directly, by restricting their options, but would likely also have more general adverse effects on growth and economic activity. Improvementsinthe InvestmentClimate 4.32 The poverty diagnostic has shown the importance of off-farm incomes diversification as an escape route from poverty. Broadly speaking, this occurs either through individuals engaging directly in often informal activities, or through being employed as wage labour. The prospects o f being able to pursue either route dependvery much on the environment and incentives for individuals and firms to enter and expand their activities. 4.33 It is far beyond the scope of the KPIA to review Kenya's investment climate in detail-which has been done in the 2008 ICA-although . we do underline that factors which dissuade firms from making investments that generate new jobs also tend to push down wages, because investors need a higher rate of return to be willing to invest. The investment climate includes a broad range of factors, from poor infrastructure through having to pay bribes, and there i s no magic bullet. In Kenya, the key factors have been documented through firm surveys which ask about perceptions as well as actual decisions in terms of investment, employment and so on (World Bank 2008b). High taxes, transaction costs-such as registering a business, and getting goods in and out o f customs-and high investment risks, in unreliable infrastructure (roads, energy) and crime top the list of factors which not only affect profitability but also tend to encourage informality. Licensing obligations pose a burdeninterms oftime and money. 139 4.34 Looking at Kenya's scores on key indicators o f doing business, and trends between 2003 and 2007 reveals both good news as well as areas of concern.60 The top ranked constraints are tax rates and infrastructure (electricity andtransport), bothof which had become more binding inrecent years. Incontrast, telecommunications has eased as a constraint andtoday is among the least problematic. Crime and corruption improved their ratings, although still remaining among the most pressing problems. We showed above that corruption is especially damaging for the poor. The cost of finance was the second most important problem for manufacturing firms in2003 but since then has improved. As noted above, lack of skilled labour entered low on the list at number 14, and labour regulations at number 16 (Table 4-4). Table 4-4: Barriers to Businessin Kenya- Manufacturing Firms Reporting Factor as Major or very Severe, 2007, Relative to 2003 (YO) Indicator Rating 2007 Dynamics Tax Rates 1 57 Transportation 2 53 Electricity 3 53 Corruption 4 51 tttJ Competitorsfrom the InformalSector 5 4Q Crime,Theft and Disorder 6 47 J Tax Administration 7 43 Access to Finance(availabilityand cost) 8 36 J Customs and Trade Regulations Q 34 Business Licensingand Permits 10 28 MacroeconomicInstability 11 26 ttJ Telecommunications 12 24 .1 Access to Land 13 17 Labor Regulations 14 16 PoliticalInstability 15 16 tt.1 inadequatelyEducatedWorkforce 18 Q Source : ICA Survey. Note: meansworsening. 4.35 Problems which are especially acute for the informal sector include crime and insecurity, poor access to credit, services and infrastructure. Various legal, regulatory barriers specifically affect the operations of the informal sector, and even if not generally enforced, may be abused by officials. The rapid growth in informal enterprises in Kenya underlines their importance in production and employment, especially but not only in urban centers. As noted above, the extent of informalization is in part a reflection of choices at the firm level to avoid the financial burdenof registration and taxation plus the minimum capital requirements, which according to the ICA are the main reasons why firms do not choose formality. However, informality also has costs-in particular lower rates of access to finance. An important question for policy makers in Kenya, and elsewhere, i s how to maximize the potential contributions of informal enterprises to broad based growth, and income earning opportunities. Alongside continued progress to improve the business climate, there is a role for proactive programs to reach out to disadvantagedgroups, includingyouth and women. 6oThe 2003 sample oversampledlargerf m s , so trends shouldbe interpretedwith care. 140 Access to Financial Services 4.36 Access to financial services, or at least to credit, has emerged as a robust variable at the household level linked to escape from poverty. At the same time, among the poor, access was largely limited to the near poor-that is, those close to the poverty line-and that significant numbers of Kenyanssuffer financial exclusion. 4.37 The ERS and the MTP (2008-2012) recognize the critical importance of a sound financial system, and laid out three broad objectives for financial sector policy: greater stability, improved efficiency, and better access for more Kenyan households and businesses. In recent years, the Government has moved away from direct control and interventions inthe financial sector infavor of a more "market-based'' approach. This has resultedalready ina very rapid growth of total productive sector credit and has stimulated a range of new innovative approachesto the delivery o f financial services (citation). Here we briefly reviewthe policy context andthe scope for public action to improve access.61 4.38 Given the findings from the foregoing poverty analysis, it is tempting for policy makers to intervene in credit markets. However, the documented failures of DFIs in Kenya, alongside the manifest dynamism of private and local level activity in the sector, should warn against such a conclusion. The key challenge is to promote rural access to the financial sector in ways which are sustainable, in a context of significant growth in competition inthe lower-income partof the financial market. 4.39 Among promising ongoing developmentsare: 0 Growth of group-based financial services, supported by training efforts extendedbyNGOs and others. 0 Mobile telephone-basedtechnology, that is extending rapidly, and which at last count had about 1.6 million users,62 These provide a means of electronic payment which utilizes existing infrastructure which already reaches most Kenyans, includingthose inrural areas (Box 4-3). 0 Doubling of the number of ATMs in the past two years to more than 700. Although the costs of opening and maintaining a bank account-a prerequisite for accessing ATM networks-are prohibitive for most people, the price o f ATMtransactionshas also recently beenfalling. The Government is bidding out work to supply social protection payments to people in a number of ASAL districts.63 If the mechanisms established to reach program recipients are also accessible to others, this will contribute to marketdevelopment. "Thefullagendaoffinancial sectorreformiscoveredintheWorldBank2008aandtheFinancialSector DeepeningTrust StrategyPaper (2008-2010) (FSD Kenya, December2007). 62Mobilephone-basedservices andbank accountbasedservicesare distinguished,basedon whether a bank account is required to use the service. The former include the unbankedto whom it provides a means of access. 63 For example, transfers to orphans andvulnerable children usingmobiletelephone or other technology are currentlybeingexploredby the DFID/CGAPchallengefund. 141 Box 4-3: ExpandingFinancialAccess via MobilePhones: M-PESA In 2007 Safaricom Kenya launched the M-PESA mobile money transfer service which instant and at relatively low cost, using an account which accepts payments via a network of agents. The service is designedto respondto the basic financial services needs of Kenyans making remittances, for example and the widespread availability of mobile phones. Many people inside and outside Kenya are watching the progressmadeby serviceslike M-PESA with great interest. Giventhe already highusageof similar servicessuch as airtime transfer, the three millionKenyanswho are unbankedcurrent cell phone users could inprinciplebecome mobile phone banking customers inthe short term. This new distributionchannel alone would increase the level of formal fmancial inclusion in Kenya from 19 percentto almost 33 percent of adult Kenyans, and further still, as the percentageof peopleusing cell phones continuesto rise, this potentialnumberwill rise higher still. However the commercialrealities will slow the impact of mobile banking on financial access. While M- PESA is f m l y positionedas a paymentservice, it has limits as a savingsmechanism. There is expectedto be a tie inwith the bankingindustryby the end of 2008, but the chargingmechanismis set to be similar to the current moneytransfer tariff, whichis relativelyexpensive. There couldbe a bigger breakthroughfrom the expansionof agency basedbanking driven by the banks through, for example, low income people in beingable to access accountsthroughPOSterminals inretail outlets. There are severalnew POS networks presentlyunder development andexperts are expectingsome impactsinthe nextyear or so. 4.40 Inworking out how to respondto the challenge of expanding access to credit, it i s important to underline that the performance of the five DFIs-namely the Agricultural Finance Corporation (AFC), IDB Capital (formerly the Industrial Development Bank), the Industrial and Commercial Development Corporation (ICDC); Kenya Industrial Estates (KIE)andthe KenyaTourist Development Corporation (KTDC)-that were establishedin the 1950s-70s has been very di~appointing.~~recent assessment of the DFI's situation A by the Treasury analyzed both the main causes of poor performance and also the consequences of the deteriorated financial condition. This noted, among other things, the problems created by weak management and poor political and management decision- makingfor extendedperiods especially inthe 1990s. 4.41 Inthe light of experience, it seems important, rather thanrushing aheadto set up new funds in Kenya, to assess the principles and standards that govern the work of successful DFIs inAfrica. These inturn could guide the re-specification o f the regulatory arrangements for Kenyan DFIs. A guiding principle would be that these should not exist merely to channel scarce budget funds. They would also be designedto have the financial credibility and management competences to attract additional non-budget funds to enhance their delivery against clearly defined mandates, all of which would need to have strong public policy justifications. The Government has beenconsidering the adoption of a comprehensive strategy that i s designedto buildstrong and credible institutions that will quickly be able to stand on their own feet and to attract significant private and donor funding in support of the development mandates that the various DFIs have been or will be assigned. 64 See Financial and legal sector technical assistanceproject, component I - Financial and Legal Sector Strategy Development, Development Finance Institutions (DFIS). A strategy paper for the reform of Kenya's DFZS, by an advisory team from the Kenya Treasury, April, 2007; and Kenya's Development FinanceInstitutions,a DiagnosticReview andRecommendationsfor Restructuring,June, 2004. 142 MEASURES ATDISADVANTAGED AIMED GROUPS Women 4.42 High measured female unemployment is a striking feature o f Kenya's labor market, especially young women. Women represent half o f Kenya's workforce, but have tended to predominate both among the low paid and the unemployed. Because women experience relatively greater difficulties insecuring productive employment, appropriately targeted active programs could be considered. Women have lower levels of educational attainment and lower access to capital and assets than men, one major concrete obstacle which i s more intractable, their domestic work burden. Policy and programmatic interventions should be designedwith these concerns inmind. 4.43 We saw above that women have less access to credit, and access smaller amounts. Targeting women, as some micro-finance institutions (MFIs) have done internationally, can combat gender. The Kenya Women's Finance Trust provides microfinance specifically to women. Evidence on female microfinance borrowers in Bangladesh, where MFIs are well established, indicates significant poverty reduction impacts, as well as better human development outcomes for their children (Littlefield, Morduch, and Hashemi, 2003). Specific efforts to eradicate gender discrimination in lending could be useful since. For example, for banks often "require the permission o f husbandsinorder to grant credit to wives" (SCGA p. 42). 4.44 A Ministry o f Gender and Children Affairs and the Gender Commission have been established, but the impacts are not yet evident. Increasing efforts to reachwomen in areas where they appear to be under-served could help reduce gender inequality. There have been some significant strides thanks to civil society. There has been a burgeoning o f women's associations at the community level in recent years, and there is now a vibrant segment o f civil society focusing on gender issues. Although the local organization and voice has not yet filtered upward to the regional or national level, this is a strength on which to build. Organizations like the Federation o f Women Lawyers, for example, conducts legal advocacy work for women. 4.45 Reproductive health i s an important issue inits own right. It i s also critical vis a vis reducing demographic pressures on families which, as we have seen, contribute significantly to the risk o f being poor. The family planningprogram, established within the Ministry o f Health in 1967, contributed considerably to the decline inthe highfertility rates from an estimated 8.1 in 1977/78 to 4.7 in 1998, although it i s not clear that there has been improvement since then. Contraceptive Prevalence Rate (CPR) for all methods and modern methods were estimated in2003 at 39 percent and 32 percent, respectively, and i s not on a clear upwardtrend. 4.46 The 2004 KSPA survey data indicate that only 73 percent o f all health facilities are offering temporary methods o f family planning services. Analysis o f the trend data on unmet need for family planning among currently married women in Kenya has remained high, at about 24 percent since 1998. Among the key issues to be addressed at the policy and program levels are addressing regional and socio-economic disparities in 143 Contraceptive prevalence, including through private sector and community provision and supporting demandand broadening the method choice. The low level of integration of FP with HIV/AIDSservices could also be addressed. Youth 4.47 The recent crisis has prompted the government to renew its commitment to addressingyouth issues, considering the potential impact that would accrue from engaging theminimmediate employment schemes. MOYS hasprepareda "Marshal Planfor Youth Employment and Development, March 2008," aimed at the creation o f immediate and medium-term employment opportunities for youth. The Plan proposes six immediate measures, out of which four aimto engage youth inlabor-intensive public works programs with the objective of creating close to 370,000 jobs in 2008 alone. Other immediate measures aim at peacebuilding and reconciliation. 4.48 One of the main government interventions directly targeting youth in Kenya has been the Youth Enterprise Development Fund (YEDF), which aims at increasing young entrepreneurs' access to loans, and facilitating supply chains linkages. The YEDF itself does not directly target the most vulnerable youth. A Youth Empowerment Program (YEP) i s also under preparation. The government, together with partners, i s currently redesigning the program in light of the recent post-election crisis. The program will focus on small-scaled public works schemes, with the dual objective of creating immediate employment for disadvantagedunemployed youth and rebuilding essential physical assets intargetedurbanandperi-urbanslums, and include a skills development component. The program will be piloted in selectedareas, to generate evidence that can guide future scale- upofthe program. 4.49 International experience suggests that the success o f these programs depends on several factors: 0 Relying on self targeting, through setting the wage rate low-international experience suggests that the rate should be set below the prevailingmarket rate for unskilled workers. A low wage would ensure that only the unemployed would participate, thus achieving better targeting by the program. However, especially for youth, this rate would be very low; 0 Eligibility to participate in the program may be restricted to a specific age group, and betailored to encouragewomen to participate inthe program; 0 The labor intensity (share of wage bill in total cost) should be maximized, although this is determinedby the type of works produced and the labor-based technology used; and 0 Institutional capacity i s crucial for the design, and efficient implementation of these programs. 4.50 The above factors needto be analyzed and applied to Kenya, and lessons should be drawn from existing public works programs, such as "Roads 2000" and "Trees for Jobs." Public works are a temporary measurethat could help alleviate the crisis. Inorder to effectively consider longer term structural issues, existing skills development and 144 microfinance initiatives aimed at unemployed youth should be reviewed. As noted above, firms in Kenya have rated workers' skills as a much lower constraint to their growth compared to corruption, cost o f finance, tax and regulation. While new market entrants are increasingly well educated, the incidence o f unemployment among educated youth i s high. Further data and analysis-quantitative and qualitative-is needed to inform appropriate policy and programmatic responses. 4.51 Easing the school to work transition is an important policy objective. Strengthening linkages between education and work can increase the relevance o f training opportunities and improve employability. For example, helping master craftspeople to improve their own skills and increasing their access to new technologies could result in better opportunities for apprentices. The existing voucher program in Kenya illustrates such an approach; however, one review concluded that the private sector would be better suited to implementing this kindo fprogram than the government (Godfrey 2003). 4.52 Youth have very limited access to financial services. However, liquidity does not seem to be an issue for Microfinance Institutions (MFIs) who are rather constrained by the lack o f business support services targeted at start-up businesses. Therefore, more analysis i s needed to determine the most constraining factors to facilitate the entry o f unemployed youth into the labor market, whether formal or informal, including self employment through a combination o f relevant entrepreneurial training, business sensitization, mentoring and credit. These options offer significant medium to long-term solutions to youth unemployment provided the policy, regulatory and institutional constraints to their proper functioning are well identified and addressed and the appropriate remedies carried out. SlumResidents 4.53 Analysis o f informal settlements in Nairobi found that poverty incidence in the slums was high in late 2006, with around 63 percent o f slum residents falling below the poverty line. At the same time, there is heterogeneity within and across slums-in Korogocho, for example, the difference in the poverty rate between the poorest and wealthiest village i s 31percentage points. 4.54 Slum residents score poorly on non-income measures o f poverty as well- unemployment levels are high (26 percent), only a minority have secondary level education or more (24 percent), and living conditions are appalling with only 3 percent o f households living in housing with a permanent wall, access to piped water, and an electricity connection. 4.55 Strikingly, living conditions in Nairobi's slums are far worse than those in the slums o f many other African cities. Slum residents in Dakar, for instance, enjoy better living standards compared to those in Nairobi, even though they are poorer and worse educated than their Kenyan counterparts (Box 4-4). A closer look at different aspects o f living conditions--quality ofhousingunits, infrastructure, tenure, and the neighborhood- reveals that Dakar outperforms Nairobi in each dimension. Specifically, housing units in Nairobi are smaller, more crowded, and constructed with worse building materials. 145 Further, as we saw in Part 11, access to various kinds of infrastructure is very low in Nairobi-by contrast, 84 percent of Dakar's slum residents have apipedwater connection, 82 percent have electricity, and 73 percent havetheir garbagepicked up. 4.56 What are appropriate policy options in this context? Clearly enabling access to better income earning opportunities is a priority, including consideration of public works types programs, targeted at specific groups, like youth (see above). At the same time, it appears that amenities in the slums warrant attention. But why are living conditions so bad? Recent research points to the lack of home ownership and a high turnover rate- almost 92 percent of Nairobi's slum residents are tenants and the average resident stays in a home for about five years. By comparison, in Dakar, only a quarter are tenants and the turnover rate is low, with residents having stayed in the same home for an average of 19 years. The high tenancy rate inNairobi, within a complex political economy, has created a situation in which none of the three stakeholders-the tenants, the absentee "shack owners," and the government as landowner-have invested to improve quality of living conditions. 4.57 The diagnostic suggests that innovative approaches are needed in Nairobi. For instance, the widely-used approach of upgrading infrastructure in the slums, without dealing with the tenure issue, i s arguably unlikely to work. Improvements in infrastructure may raise rents and benefit the "shack owners," but displace many of the current tenants. 4.58 In 2004, the government launched a program to improve slums-the Kenya Slum Upgrading Program (KENSUP). The objective is to provide physical, social, economic, organizational and environmental improvements as a means o f enhancing amenities in the slums. The intention is to construct 150,000 housing units yearly, of which about 60,000 will target slum dwellers, with the rest redressing inadequate urban and rural housing. However, KENSUP's plan to build new housing units for slum residents i s already facing difficulties. The multi-storey walk-up apartment buildings with three-room units that are being constructed are relatively expensive and will reach very few of the current slumresidents. 4.59 The slum situation calls for an approachthat combines infrastructure investment with efforts to alter the tenure mix, for instance, by converting a significant proportion of slums tenants into owner-occupiers that have a greater incentive to invest intheir housing and their neighborhood. There are various ways inwhich this result can be achieved but there will often require a tri-partite negotiation between the government (the official landowner inmost cases), the "shack owners" andthe tenants as a starting point. 146 Box 4-4: A Tale of Two Slums: Nairobiand Dakar A study comparing the slums of Nairobi and Dakar suggests that incomes, education and jobs are not sufficient to ensure good living conditions. Slum residents in Dakar enjoy better amenities than those in Nairobi, even though they are poorer and worse educated. Conceptualizing quality o f living conditions as a composite o f four interacting factors-the unit, infrastructure, tenure, andthe neighborhood's condition and location-and graphing it as a Living Conditions Diamond, allows Gulyani and Talukdar (2008) shows that on each o fthe four dimensions, living conditions inNairobi's slums are worse than those inDakar. As described in Part I1 and World Bank (2007), infrastructure access in Nairobi is appalling and dramatically lower than inDakar. For instance, inNairobi only 22 percent have access to electricity and 19 percent have access to piped water in the form o f an in-house connection or a yard tap. By contrast, 84 percent o f Dakar's slumresidents have a pipedwater connection and 82 percent have electricity. Further, housing units in Nairobi are smaller, more crowded, and constructed with worse building materials. The explanation lies inunderstanding tenure. Although landtenure is largely "informal" inboth cities, the tenure mix-proportion o f renters versus owners-and turnover rate differ significantly in the two cities. Source: Gulyani andTalukdar (2008); Gulyani, Talukdar and Jack (2008). Social Protection 4.60 The Ministry o f Gender and Children Affairs (MGCA) i s responsible for development o f a National Social Protection Strategy (NSPS). The latest draft (November, 2007) defines social protection as a response to drivers o f poverty and vulnerability. In Part 111, we investigated the frequency and severity o f shocks, such as rise in food prices, drought, floods, illnesses, injuries, employment losses, crop failures and thefts. In the long term, the strategy aims to facilitate the development o f a comprehensive social protection system, with those identified as poor and vulnerable to poverty being the priority. In the short and medium term, the objective o f the strategy is to meet the immediate needs o f the poorest and most vulnerable. Furthermore, given limited financial and human resources, the initial social protection interventions should begin where the needis greatest. 4.61 While there are major immediate needs associated with the recent post-election violence, these have been documented well elsewhere (Damage Assessment etc.), we focus here on the long term elements o f the strategy, and focuses in particular on the largest programs-viz those aimed at supporting pastoralists and Orphans and Vulnerable Children, and combating malnutrition, as well pensions for civil servants which, while not targeted at the poor, represent the largest social protection program in practice, and consume a considerable part o f the budget. ExistingPrograms 4.62 Kenya's social protection policies and programs are generally underdeveloped- in that fewer than one in ten individuals in the workforce is covered by retirement schemes, for example. More generally on the labour market side, there i s an extensive network o f retraining programs, but these are typically low standard and poorly coordinated. Public works are very limited in scope. The KIHBS shows that the most common responses in the face o f major shocks can be listed in order o f importance as: 147 spend cash saving, sell assets (animals), work longer hours, reduce food consumption and receive help from family and friends. Government assistanceis rarely mentioned. 4.63 Inrecent years, with external support, the Government has embarked on various programs that aim to help offset the risks faced by vulnerable groups in the population, and the impacts of shocks, and as noted above, is developing a comprehensive strategy. These fall into two broad categories: Inresponse to the risk of drought; development programs in the north and in pastoral areas, the national drought contingency fund and a pilot cash transfer program for the chronically food insecure in the arid and semi-arid regions (Hunger Safety Net Program). About 2 million Kenyans (6 percent of the population) permanently rely on food relief, with higher numbers in drought years (e.g., 3.5 millionin2006). 0 New policy instruments to directly tackle poverty and vulnerability, in particular a cashtransfer program for Orphans and Vulnerable Children(OVC). 4.64 Here we examine the main initiatives: pastoralists, OVCs, programs to combat malnutrition, andtransfers to retiredcivil servants. Pastoralists 4.65 Part I1 confirmed the well-known point that pastoralists are the most disadvantaged group in Kenya in terms of levels of consumption and access to basic services. The Arid Lands Resource Management Project (ALRMP), which is supported by various donors, includes a Drought Management component. An early warning system covers 28 districts and 3.5 million people in the ASAL areas and monthly household surveys inform the government where there is stress, such as food shortages and animal diseases. Another component is Community Driven Development, covering 11 districts, which supports income generating activities for vulnerable and marginal groups identified by the community, such as widows and orphans. There is a 30 percent matching element from the communities. 4.66 The ALRMP has been operating for over a decade and has reportedly reduced child malnutrition, lowered food aid requirements and increased access to water, school and health facilities. The program has put in place governance and accountability structures at community level, which have empowered communities to manage their own development process, and to access other resources such as the CDF and food for work programs. Because of limited opportunities for investment in these dry remote areas, promoting income generating activities has been a challenge. Increased investment in infrastructure can encouragediversification and income generatingactivities. 4.67 While the ALRMP covers 520 communities through 2,000 micro projects, reaching over one million beneficiaries, the extent and level o f deprivation in ASALs is still huge, as evidenced in the foregoing analysis on North Eastern province. It should also be noted that there have been critiques of the relief food program, pointing to lack of flexibility and problems with community based targeting (errors of both inclusion and 148 exclusion), and weak accountability and mechanisms for redress for those having complaints (Kenya HumanRights Commission, 2006, Occasional Paper, 2006). Orphansand VulnerableChildren (OVCs) 4.68 Families missing one or more parents are more likely to be poor. Among the national efforts to fight HIV/AIDSis a food and nutritionprogram, supported by the WFP, which includes efforts to increase enrollment and school attendance of OVCs. Sub- districts are selected based on HIV prevalence and poverty levels, and beneficiaries are selected based on the nutritional status of the patients and socio-economic and demographic criteria. In 2009 the program will target 66,000 beneficiaries. An evaluation in 2005 suggested overall improvements in nutrition among program participants and in the degree of recovery for those on ART. The program effectively targeted the most food insecure and HIV/AIDS infected households, although Improvements in beneficiary targeting were suggested. The study also showed evidence that the program funds were thinly spread. Exit from food support was difficult, especial1 for women headed or elderly headed households, with a high number o f , orphans. & 4.69 A cash transfer program for OVCs, implementedby the Ministry of Gender and Children Affairs (MGCA), started in2004 in3 districts. The secondphase of the program was launched in 2006, and i s currently reaching over 20,000 households in 37 districts. By 2012, the program is envisaged to reach 100,000 households or 12.5 percent of the estimated 2.4 million OVC countrywide. An ongoing operational and impact evaluation of this program suggests some success in targeting poor households, but room for improvement inidentifyingthe poorest household^.^^ CombatingMalnutrition 4.70 Itwas underlinedinPartI1that Kenyahas shown little, ifany, improvement over the last 20 years in terms of meeting MDGs with respect to combating malnutrition. While a National Food and Nutrition Policy was prepared, it has not been approved in parliament, nor is there any agreement on the implementation mechanisms. Budgets for the related programs have not been identified. Box 4-5 describes some lessons from UNICEF experience. 4.71 The recent increases in food prices have exemplified the social, economic and political salience of this issue. Based on analysis of the KIHBS, the poor in 2005/6 devoted about 70 percent of their expenditure to food; for the bottom quintile, the share was even higher 77 percent. Estimates o f food poverty-Kenyans with insufficient 65WorldFoodProgram,Draft Country Program -Kenya, May, 2008. 66World Food Program, Reducing the Burden on HIV/AIDS: Experience of a Food Aid Intervention in WesternKenya,March, 2005. Oxford Policy Management, Kenya OVC-CT program - Operational and Impact Evaluation-Baseline Survey Report,Alex HurrellandPatrickWard, draft report, February2008. 149 consumption to cover basic food costs-were already around one fifth of the populationin 2005/06. Box 4-5: Key Elements of Nutrition Interventions UNICEF has noted that the following nutrition interventions have proven to be cost effective, feasible and to preventmalnutrition before it happens: 1. Promote child survival and development through the support o f exclusive breastfeeding for the first six months o f life, adequate and timely complementary feeding o f young children and increased feeding during and after illnesses. 2. Prevention o f micronutrient deficiencies through vitamin A supplements for children and lactating women using a range o f contact points to maximize coverage; iron-folate supplements for pregnant women; universal consumption o f adequately iodated salt; and production and consumption o f micronutrient rich foods including, where appropriate, fortified foods. 3. Increased dietary diversity through consideration o f nutrition in food and agriculture programs; and incorporation o f nutrition inschool curricula. These actions need to be supported by a functional nutrition information system, a strong nutrition communication strategy and broader strategies to increase food and livelihood security. Clearly strong commitment o f decision makers at national, provincial, district and community levels is needed. An effective coordination and monitoring body for nutrition also appears to be a promising route. Source: UNICEF, Nairobi, Kenya 2008. 4.72 There are a series of programs already in existence that are designed, and succeed, to varyingextents, to meet nutritionneeds. To summarise briefly: 0 School Feeding Program, in food-insecure areas i s a GoK program supported by the World Food Program (WFP). Selected schools are in unplannedurban settlements o f Nairobi and the most food-insecure sub-districts in ASALs with the lowest enrolment and completion rates and high gender disparities. In2008 this program supported 1.2 million children, but due to increasedinternational costs, has been cut and in 2009, the program will cover 750,000 beneficiaries, and is projected to shrink further. An evaluation in 2007 suggests that school feeding contributed to increased enrolment, attendance and completion rates from 2004 to 2007. Targeting i s a challenge as is capacity inlogistics, and data collection, analysis andreportingq6* 0 Emergency Food-Aid programs: reaches around 1.8 million people in ASAL areas and 250,000 refugees annually. A recent evaluation suggests that this support allowed households to stay in their communities and retain assets, rather than sell assets, such as cattle, to meet their basic consumption needs. 0 Hunger Safety Netprogram (HSNP): i s a pilot program being implementedby the GoK with support from DflD which aims to provide a regular, predictable and guaranteed amount of cash to chronically food insecure households in the ASALs. Four districts have been selected for the pilot phase-Turkana, World Food Program, Draft Country Program -Kenya, May, 2008. 150 Mandera, Marsabit and Wajir-with cash transfers to 300,000 people, to learn how to develop the most effective targeting and payment mechanisms. It will test a range o f community-based and categorical targetingapproaches. Pensions 4.73 The Civil Servants Pension Scheme (CSPS) is a defined-benefit scheme covering approximately 393,000 members including civil servants, military personnel, police and prison staff, members o f parliament, employees o f the judicial system and teachers. This formula results in pensions which average about 75 percent o f basic wages at retirement. The National Social Security Fund (NSSF) i s a mandatory provident fund for employees o f private sector companies, parastatals and other quasi-government authorities. The combination o f low contribution levels (maximum contribution o f KSh 400 per month, which i s less than 2 percent o f the average formal sector wage earnings), high operating expenses, and poor returns on its investments has resulted in low lump-sum benefits provided at retirement. 4.74 Earlier analysis found that the NSSF suffers from weak governance, the pensions provided are too small to provide long-term security and are effectively insolvent due to decades o f poor investment performance. 69 Despite the large number-lY342--of schemes, the actual coverage o f the Kenyan population is very low, at only 8.2 percent o f the labour force, and limitedto those with formal employment. Reform options should be considered inthe context o f assessing pensions policy as a whole and the need to develop a long term strategy. As the system is presently designed, state-sponsored pensions are both expensive (as a share o f the budget), regressive (in terms o f the distribution o f benefits) and yet fails to provide meaningful income security for its members. Muchwork has already been done to address these, but major decisions have not yet been taken. In Kenya, one o f the most notable expenditure trends since 2000 has been that pension payments to public servants has almost doubled as a share o f GDP, to 1.1 percent in 2006/07. Far more domestic resources are being spent on retired civil servants than on transfers to vulnerable groups. The largest o f the other programs described above i s emergency food aid which i s largely externally financed, and which is, by its very nature, limited to offsetting the worst repercussions o f food shortfalls. In this context, further development o f the social protection strategy, informed by the findings about poverty and vulnerability inthis assessment, would be welcome. 4.75 To sum up, it seems that the evolution o f Kenya's social protection system has a significant distance to go, in order to reach its stated objectives. Very few o f those hit by shocks received government assistance, and the total coverage i s extremely limited at present-the only groups which are directly reached are those with HIV and OVCs and people living in the arid lands. There are pilots and plans with respect to youth (see above). None o f these programs presently reach more than a fraction o f those in need. Priorities for moving ahead, given budget and institutional constraints, and in light o f empirical poverty analysis, warrant further attention. 69World Bank 2006 Kenya: An Assessment of Pension Reform Directions. 151 THEBUDGETASAN INSTRUMENT REDUCTION FORPROMOTINGPOVERTY AND ADDRESSINGINEQUALITY 4.76 The government budget is a major policy tool to reduce poverty and improve equity. This section examines how this instrument has been used in Kenya since 1999/2000 in light of pervasive inequality and spatial dimensions to poverty. We review aggregate performance and trends in public spending and then provide a more detailed analysis of the distribution of spending. Weaknesses in fiscal management are a considerable problem, as we show below. The latter part of the chapter examines the spatial aspects and spending allocations by province, inkey sectors (education, health and water), and high profile programs, in particular the CDF. The analysis faces major constraints, including the fragmentation of the budget and lack of reporting at a consolidated general government level, lack of reporting in an internationally recognized method (such as the IMF's Government Finance Statistics Manual), and lack of reporting using a standardfunctional classification. 4.77 Kenya's fiscal management inthe 1990s was characterized by the dominance of recurrent expenditures, widespread tax exemptions favoring political elites, divergence o f actual spending from the budget approved by parliament, and reliance on domestic financing. Short term domestic debt stock grew rapidly and short term interest rates were high, resulting in increasing interest payments. As economic growth slowed the macro- fiscal position became increasingly unsustainable and threatened the ability of the government to meet its recurrent obligations. 4.78 In2002, the government embarkeduponanEconomic Recovery Strategy (ERS), which was also the first national poverty reduction strategy. The major economic policy objectives included improved accountability in the use of public resources, and restructuring and refocusing public spending toward priority activities including agriculture and rural development, infrastructure, and health and education. Deepening of the Medium Term Expenditure Framework (MTEF) process, Public Expenditure Management (PEM) reforms, and utilizing an annual Public Expenditure Review (PER) were expected to inform and improve the resource allocation process. There was an increased awareness of the need for increased development expenditure to address the country's increasingly run-down capital assets. 4.79 There has been significant success on the macro-economic front overall - but much less progress in meeting the government's stated allocative objectives. Table 4-5 shows the evolution of public finances over the period 1999/2000 - 2006/7, including reductions inthe stock of domestic debt to GDP and inthe ratio of short term treasury bills to longer term treasury bonds and the increased grant element of external financing. While the primary budget surplus fluctuated over time, over the entire period it fell from 3.3 to only 0.7 percent of GDP in2006/07. There was a strong revenue effort, and inreal terms, total revenue rose at an annual average of 6 percent due to economic growth and higher corporate profitability as well as measures such as the tax amnesty scheme and improved tax administration. 152 Table 4-5 CentralGovernmentFiscal Outcomes,Percentof GDP, 1999/2000-2006/07 19991 20001 20011 20021 20031 20041 2005 20061 2000 2001 2002 2003 2004 2005 20106 2007 Revenues 19.9 21.9 19.9 20.8 22.3 22.3 21.5 21.4 Expenditures 19.5 23.5 22.3 24.5 22.7 22.2 24.8 23.1 Recurrent 16.8 20.1 19.9 20.5 19.8 18.9 20.4 18.7 Development 2.6 3.4 2.4 4.0 2.9 3.3 4.4 4.4 Overall balance 0.1 -1.0 -2.6 -3.0 -0.3 0.1 -2.4 -1.6 Primary balance 3.3 2.2 0.3 0.3 2.2 2.4 0.3 0.7 Financing 0.0 0.5 2.6 3.O 0.0 -0.5 2.3 2.0 Domestic 1.3 0.1 3.9 4.3 0.7 -0.5 1.8 3.0 External -1.5 0.8 -1.1 -1.0 -0.7 0.0 0.0 -0.2 Stock of domestic debt, net 17.6 16.6 19.7 23.0 21.3 18.4 18.1 18.4 Source:IMF. 4.80 Public spending grew more rapidly than the economy. Central government expenditure and net lending as a share of GDP increased from 19.5 percent of GDP in 1999/2000 to almost 25 percent in 2005/06. Inreal terms, sectoral expenditure increased by an average of 8 percent annually. 4.81 Development spending grew more slowly, but rose at a rapid rate from 1999/2000, albeit from a very low base. The domestic effort in financing public investment has increased significantly, inboth absolute terms, and relative to donors. In 1999/2000, 80 percent of development spending was externally financed. By 2006/7, domestic sources financed two-thirds, so that the government spent 3 percent of GDP on development, comparedto external development financing of about 1.4 percent of GDP. 4.82 The main highlights for recurrent expenditure trends, which also include staffing costs for social services, between 1999 and2007 are as follows: Wages and salaries constituted the largest share of total expenditure, but declined from 36 percent of total spending in 1999/2000 to about 30 percent in 2006/07, as a result of a hiring freeze and salary restraint. At the same time, in real terms, the personnel bill increased by an annual average of 5 percent, and has maintained its share of GDP (around 7 percent). Expenditure onpensions for public servants almost doubled as a share of GDP, to 1.1 percent in2006/07. Interestpayments fell as a share of total expenditure, as a result of both debt restructuring and debt stock reduction. Expenditure on defense including national security intelligence increased by 9 percent inreal terms. The share of operations and maintenance expenditure in GDP rose by half, to 4.5 percent, which begins to redress the pattern of under-spending inthe 1980s and 1990s. 153 ImprovingExpenditureComposition 4.83 The fiscal accounts suggest that in practice, progress towards a more growth- oriented and pro-poor budget has been limited. First, we provide a snapshot for the most recent year available (2006/7), then look at trends. In 2006/07, the largest share goes to education, followed by public administration and infrastructure (Figure 4-2). Figure 4-2: Composition of SectoralSpending, 2006/07 Total Spending by Sector 2006107 A f i 16.6 Education Public Administration QPhysicalInfrastructure 0PublicSafety Law and Order .National Security UAgricultureand Rural Development .Health oGeneral Economic Services I Source: Ministry ofFinance, Quarterly Budget and Economic Review, various issues 4.84 Table 4-6 summarizes real trends insectoral expenditure, normalized to 1999/00. The patterns are striking-real spending on infrastructure more than tripled, and on agriculture doubled-yet more than quadrupled on public administration. Environment, health, education andpublic safety experiencedincreases far below the average doubling. 154 Table 4-6: Real Trends in SectoralExpenditure 199912000=100 19991 20001 20011 20021 20041 20051 20061 2000 2001 2002 2003 2005 2006 2007 Agriculture and Rural Dev 100.0 158.2 150.7 124.7 101.6 151.5 180.5 Agriculture 100.0 175.0 147.1 137.5 100.4 156.7 209.8 Environment 100.0 136.6 156.0 77.8 78.2 81.7 103.9 Lands and Settlement 100.0 117.0 156.4 202.0 140.1 265.9 159.6 PhysicalInfrastructure 100.0 156.6 133.2 104.5 211.9 251.3 316.9 LocalGovernment 100.0 153.1 139.2 129.8 111.8 211.9 235.7 Roads, Public Works andHousing 100.0 102.1 106.3 68.0 146.4 169.0 220.4 Transport andCommunications 100.0 173.8 176.5 153.1 239.7 246.3 370.4 Energy 100.0 742.6 345.8 129.0 646.0 731.7 708.6 Health 100.0 118.6 137.7 147.3 156.7 178.7 158.6 Education 100.0 101.1 109.8 125.6 139.9 150.2 162.5 Public Safety Law and Order 100.0 139.1 137.9 134.3 168.9 193.1 177.4 Public Administration 100.0 201.9 156.2 204.1 303.4 288.8 400.9 Grand Total 100.0 131.1 131.3 137.8 170.0 188.0 211.4 4.85 Overall spending on ERS priority sectors-defined as agriculture and rural development, infrastructure, and health and education-increased at the same rate as total public expenditure, and thus since 2002/03 has remained a steady share at 57.5 percent to 2006/07. Over time, we see the following trends (Table 4-8): 0 Infrastructure increased from 11 to 17 percent of spending, indeed doubling from the 2002/03. Actual spending could have been higher had the absorptive capacity constraints been better resolved. For example, in 2006/07 only 62 percent of the recurrent budget and 73 percent of development budget were actually spent. o Public spending on roads, which had declined to only 3.6 percent of spending by 2002/03 recoveredto 7.5 percent by the end ofthe period. o Spending on water resources management and development rose from less than0.1percentbefore 2002/03 to 2.7 percentin2006/07. o The share of energy production rose from 0.6 to 2.2 percent of total public spending between 2002/03 and 2006/07. The share of public administration spending almost doubled from 9.1 percent to 17.2 percent of spending. This line includes funds for special programs and funds for AIDS programs, which are under Office of the President, although these amounts are not significant as such. Even after netting out CDF funds, spending tripled in real terms. Part of this spending arises because costs, like those associated with privatization and restructuring of public bodies, and tax refunds, show under the MOF vote. The rest of public administration netting out Ministryof Finance spending is 26 percent o f PA expenditure and spending underthis categoryhas tripledsince 2003 155 0 The share of spending on education fell from 37.1 to 28.5 percent. Also falling has been the share of spending in the health and agriculture sectors. One possible explanation for priority sectors not maintaining their shares could be lack of absorptive capacity (see below on budget executionrates). It also could partly reflect the fact that the sector is labour intensive and that public sector wages have been constrained. For health, development partners support is largely spentoff-budget andnot through government systems. Table 4-7: FunctionalCompositionof Central Government Spending,Percentof GDP, 1999/00- 2006/7 19991 20001 20011 2002 20031 20041 20051 20061 2000 2001 2002 12003 2004 2005 2006 2007 ARD 0.9 1.4 1.3 1.1 0.8 1.1 1.2 Infrastructure 1.5 2.3 1.9 1.4 2.7 3.0 3.3 Health 1.o 1.2 1.4 1.4 1.4 1.5 1.2 Education 5.0 4.9 5.2 5.8 5.9 6.0 5.7 Public Safety Law and Order 2.1 2.9 2.8 2.7 3.1 3.3 2.7 Public Administration 1.2 2.4 1.8 2.3 3.1 2.8 3.5 Economic Services 0.3 0.3 0.5 0.5 0.5 0.5 0.9 National Security 1.4 1.7 1.9 1.9 1.8 2.0 1.6 Sectoral Total 13.4 17.2 16.8 17.2 .................. 19.4 20.2 20.0 Note: Table shows central government spending only. Table 4-8: SectoralExpendituresas Percentageof TotalBudget, 1999/2000 -2006/07 19991 20001 20011 20021 20041 20051 20061 2000 2001 2002 2003 2005 2006 2007 Education, Science & Technology 37.1 28.6 31.1 33.8 30.6 29.7 28.5 Public Administration 9.1 14.0 10.8 13.5 16.2 14.0 17.2 Physical Infrastructure 11.1 13.2 11.2 8.4 13.8 14.8 16.6 Local Government 2.3 2.6 2.4 2.1 1.5 2.6 2.5 Roads, Housing and Public Works 7.2 5.6 5.8 3.6 6.2 6.5 7.5 Transport and Communications 0.9 1.3 1.3 1.o 1.3 1.2 1.7 Water Resources& Management 0.0 0.0 0.0 1.0 2.3 2.0 2.7 Energy 0.7 3.7 1.7 0.6 2.5 2.6 2.2 Public Safety Law and Order 15.9 16.9 16.7 15.5 15.8 16.4 13.4 Health 7.8 7.0 8.1 8.3 7.1 7.4 5.8 Agriculture and Rural dev 6.9 8.3 7.9 6.2 4.1 5.5 5.9 Gender, Sports, & Social Services 0.0 0.0 0.0 0.4 0.6 0.6 0.7 Other sectors 11.5 11.9 14.2 13.8 11.7 12.1 11.9 Sectoral Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Memorandum items ERS Priority Sector 62.8 57.2 58.3 57.2 56.3 57.9 57.5 "Total" as % o f GDP 13.4 17.2 16.8 17.2 19.4 20.2 20.0 CDF Allocation 2.1 2.3 2.7 Local Authority Transfer Fund 4.0 5.0 5.2 Source: Ministry of Finance: Quarterly Budget and Economic Reviews, various issue. Notes: Agriculture includes Livestock and Fisheries, and Cooperatives. Other sectors include Ministries o f Labor and Human resources, Tourism and Information, Trade and Industry, Defense and National Security Intelligence services. Actual expenditures by local authorities, except 2006107, which is estimated. Total excludes bank restructuring costs, net lending and interest payments and extra-budgetary funds. 156 Scenariosand Options: MAMS 4.86 The allocation of domestic government spending across different functions is under the control of the government. Inparticular, the government can choose to spend more or less of a givenbudget on productive purposes which are likely to promote growth and reduce poverty, versus unproductive purposes, like some administrative costs. Indeed this has been an explicit policy objective of the Kenyan government since at least 2003, withthe adoptionofthe Economic Recovery Strategy. 4.87 Using MAMS, we can examine the potential payoffs of alternative budget allocations. A series of plausible scenarios are examined-which do not involve absolute cuts in spending, but affect the rates of growth over time. Intwo simulations, which we label eff-inf and eff-hd, the annual growth rates for "other government" (which includes defense and general administration) i s halved, from around 8 percent under base to 4 percent. This saving can be used for alternate purposes: under eff-inf, the resulting fiscal space i s usedto expand spending on infrastructure, whereas under.eff-hd, the savings are allocated to finance basic services (all levels of education, health, and water and sanitation). 0 Under eff-inz annual growth in GDP, private consumption and investment all accelerateby 0.3 percentage points. MDGs 2, 4, 7a, and 7b all improve. By 2012, the poverty rate i s 1percentagepoint lower than under the base case. 0 For eff-hd, the acceleration of growth inGDP, private consumption and investment are weaker (by 0.1-0.2 percentage points). Government consumption grows more rapidly while investment growth slows down. The improvements are stronger for MDGs2 and 4 andweaker for MDGs 7a, and 7b. 4.88 Figure 4-3 shows an evolution of indices for poverty reduction and other human development outcomes through 2012 for a set of simulations that reallocate spending betweeninfrastructure and HDwith eff-inf and eff-hd as the two extreme cases. The index for poverty reduction shows how much of the MDG 2015 target of halving poverty is achieved, while the index for HD is based on the achievements with respect to a weighted average of the 2015 targets for the education, health and environmental MDGs.~'Inboth figures, the right-most point has the highest spending on infrastructure and the lowest on HD.As one moves toward the left of the graph, spending is switched from infrastructure toward social services. 'OThe weights used are one-third for MDGs 2 and 4 (the rates of net, on-time primary completion and under-five mortality) and one sixth each for MDGs 7a and 7b (rates ofwater and sanitationaccess). 157 Figure 4-3: PossibleTrade-offs between Spending on Growth on Infrastructure and Social Services, on M D G targets for Human Developmentand Poverty, 2012 Share of HD target 0.470 0.465 0.460 0.455 0.450 0.22 0.24 0.26 0.28 0.30 0.32 0.34 0.36 Share of poverty target achievement 4.89 In2012, under either ofthese modestbudget reallocation scenarios, the economy remains far from achieving the 2015 MDG targets for either poverty or human development. Ifthe timeframe is extended through 2030, the MDG targets are more than fully achieved, and to a larger extent for poverty reduction. The general picture of the trade-offs are similar. Reduced spending on infrastructure and more spending on human development tends to slow progress inreducing consumption poverty. This arises via the growth effect, which is modeled so that the impacts of infrastructure on growth are larger than the impacts of human development on growth. Inthe model, the marginal returns from additional spending on human development decline over time due to diminishing marginal returns from, and increasing marginal costs of, government services combined with increasing marginal lossesfrom less spending on infrastructure. 4.90 These results are illustrative but not, of course, by any means conclusive. These do suggest that the complex processes generating MDG outcomes involve roles for public spending both on social services and on infrastructure. The latter both facilitates service access and by raising incomes, enhancing the ability of households to get access to and contribute to the financing of services. 4.91 We now turn to look at specific mechanisms that have been developed over the past few years to try to improve the pro-poor focus o fthe budget, namelythe Constituency Development Fund and Local Authority Transfer Fund, and the concept of Core Poverty Programs. DecentralizedSpending 4.92 Parts I1 and I11investigated the significant regional variation in economic and poverty performance. A key question which arises i s whether public investment worsens or alleviates such disparities. Historical evidence suggests very unequal distributions up through 2003 at least (Kiringai 2007). This section attempts to update the picture. The distribution and impacts of these transfers are of central interest to the KPIA. However it i s extremely difficult to obtain comprehensive data from budgetary and extra budgetary 158 sources to arrive at the total allocation to local administrative units, constituencies, districts and local authorities, aggregated to the provincial level. Nor do we have systematic information on leakages of funds, which tracking exercises suggest have been significant. 4.93 Within Kenya's unitary system of government, an array of funds are used to support decentralized spending include the Local Authority Transfer Fund (LATF), the constituency Development Fund (CDF), the Free Primary Education Fund (FPE), the Secondary School Education Bursary Fund (SEBF), the Rural Electrification Program Levy Fund (REPLF), and the Road Maintenance Levy Fund (RMLF) (readers unfamiliar with the system should refer to Box 4-6). It is not easy to determine how pro-poor are the allocations of the different funds nor the collective outcome of all the funds. The data available i s partial, across programs and years. Box 4-6: Mechanisms to Support Local Spending The Constituency DeveloDment Fund (CDF) Act as amended earmarks 3.5 percent o f ordinary revenue to constituencies for purposes o f development and in particular the fight against poverty at the constituency level. The CDF was initially shared equally among all constituencies, then the formula was revised in 200516 to increase the share o f poorer constituencies: 75 percent o f the fund allocated equally and the balance i s allocated based on the constituency's contribution to poverty. The allocation to constituency Y is ( " ) defined as: CDF, = 0*75210 * +(0.25 * CDF)*W C ~ where WCP is the constituency's contribution to poverty. The Local Authoritv Transfer Fund (LATF) was established in 1999, earmarking 5 percent of national income taxes, to improve service delivery o f the 175 local authorities. This source accounts for about one- quarter o f total local authority revenue. The formula shares 3 percent of the LATF equally, 60 percent is based on population and 37 is percent based on urban population density. The LATF allocation for local authority X can be expressed as: LATF, = 0.03 LATF ( * )+(5* 0.6 LATF) (% 0.37 LATF) * + * * 175 N UN where N i s the national population, n, and un, represent total and urban population o f local authority X respectively, and UNis the total urbanpopulation. The Road Maintenance L e w Fund (RMLF) was established in 1993 and earmarks fuel levy and transit toll receipts. In terms o f allocations: 60 percent goes to international and national trunk and primary roads; 24 percent to secondary and 16 percent to rural roads. The most significant local expenditure is under the district allocation budget. (DAB). For ministries with district offices, all the district allocations are aggregated in one budget head. Once the budget is tabled and approvedby Parliament, ministries publish a DAB, splittingthe allocation to across districts, by line items of expenditure. The DAB guides the exchequer releases to the districts. The bulk is disbursed through the District Authority to Incur Expenditure, A.I.E, which is based on the DAB for each ministry.To get the total district allocation, we aggregate the district AIEs from all the ministries. One basic problem is that it is difficult to derive the DABs by district/province, and also that there are no clear guidelines for allocation or enforcement. Furthermore, when the budget is revised through Supplementary Estimates, DABs are not revised and it is not clear how, for instance, expenditure cuts are implementedat the DAB level. However, the picture is even less transparent in that there are significant expenditures incurred at the regional levels that are not disbursed through the A.I.E. The most significant are: teacher wages and secondary school bursary funds, and wages and drugs for health facilities. 159 4.94 Further analysis is needed to examine the pattern of decentralized spending against Kenya's growth and distributional objectives. In light of these results, consideration could be given to possible revisions of the allocation formulas andthe scope for increasing the share of development budget in local spending. O f major underlying importance i s the needto improve the comprehensivenessand transparencyof the relevant fiscal data, so as to enable analysis andinformeddebate. CorePovertyPrograms 4.95 In 2002, Kenya introduced the concept of `core poverty programs' (CPPs) to protect activities which disproportionately benefit the poor. These were initially selected based on a set of criteria agreed upon by both sector ministries and the Ministry of Finance, and development partners, which were revised in 2003/04 to become more comprehensive and take into account new programs identified in the ERS, and a commitment to allocate not less than 5 percent of the budget to CPPs. The new criteria aimed to cover programs that would increaseincomes of the poor, improve their quality of life, security, and equalit . CPP's development spending in 2006/07 amounted to 7 percent of total budgetY7' of which about one-third of recurrent spending was on education, largely free primary. The road sector was the second largest CPP, followed by health. 4.96 Part of the rationale for the CPPs was that these programs should get priority in budget allocations, be protected against any mid-year cuts, and receive full and timely disbursements. However, these objectives have not beenmet. The average disbursement rate for CPP was only 67 percent, compared to 86.5 percent for the budget overall (see below). The monitoring format for the CPP was improved in2007, but implementation i s still not fully effective, and getting timely and complete reports from line ministries i s a problem. Table 4-9: Core Poverty Programs,2002103 2006107, percent - 2002103 2003104 2004105 2005106 2006107 Utilization Rates: Recurrent 89.2 86.7 77.7 88.7 83.0 Development 54.3 47.8 46.9 52.2 61.7 Total 69.5 64.5 61.7 70.5 70.7 CPP Share of Overall Budget RecurrentCPP: 4.7 6.6 7.7 8.6 9.7 Development CPP: 32.1 34.1 28.6 24.3 25.1 As Share ofTotal Budget: 7.5 10.0 10.9 11.3 14.0 Source: Ministry of Finance. 4.97 Among the other problems evident inthe CPP is the failure of the budget outlook paper as well as many Ministerial PERs to specify the criteria used for selection. As a result, the concept of CPP may have been subject to abuse. For example, the entire Kenya Agricultural Research Institutebudget (KSh. 2 billion) is ring-fenced as a CPP. KARIhas 71 CPP is defined so that the recurrent part finances only non-wage expenditures while development vote might include wages and salaries. 160 considerable administrative staff and over 200 PhDs staff conducting various research programs. It might have beenprudent to protect only those programs o f KARIrelating to poverty reduction but arguably not its entire budget. Other examples include the entire budget of the judiciary, state law office. Even within the education sector a number of "pet projects'' have been hidden under CPP including Kenya Institute for Education, Primary Teachers Training Colleges Provincial and District Administration. 4.98 We briefly compare Kenya's CPP approach to other countries in the region. While the definition of poverty-reducing spending should be country-specific it i s also apparent that the Kenyan definition i s far broader than has been adopted elsewhere. A review of the definitions for 29 low-income countries that have defined "pro-poor" spending and have been tracking pro-poor spending under the HIPC initiative shows that they all included education and health (some limited to basic levels of service), and a number included social protection and rural development. Many include rural roads, or transport in general and road maintenance. About four countries included justice and the judiciary.72 4.99 Regional and international experience suggests that identification of programs benefiting the poor and protecting these through administrative measures i s potentially useful. It should, however, be seen as a transitional measure, in that the link between poverty reduction and public spending is complex, and excessive focus on identified programs and projects has the potential to skew the spending away from the most efficient mix of interventions. It could also detract from policy measuresthat may not cost much in terms of public spending but would significantly improve poverty outcomes. These kinds of issues have tended to shift discussions towards a focus on results and identification of spendingpriorities withina mediumterm planframework. 4.100 For example, Uganda's PAF gives special treatment only to specific programs that are inthe PEAP (PRSP) that are directly poverty reducing, delivering a service to the poor, and have a well developed plan. The five major areas are primary education, primary healthcare, water and sanitation, rural roads and agriculture extension services (Box 4-7). An alternative approach in Tanzania, involves a system o f identifying priority poverty expenditures, to assure budget releases to such items. There is no explicit commitment to increasethe proportion of budgetary funding to these programs, which has meant that the government has been able to reallocate to emerging priorities such as transport, communications and tourism. While the government recognized the inherent limitations of any definition of "poverty reducing expenditures," the 2005/6 budget requested spending entities to prepare budget proposals explicitly linked to PRSP outcomes. ''IMFandWorld Bank(2006) HIPC InitiativeandMDFU-Status ofImplementation. 161 Box 4-7 Uganda's PAF: Achievements and Shortcomings Ugandais well-known for its Virtual Poverty Fund(VPF), which is a tagging mechanism for tracking pro- poor expenditures in the budget. The PAF share grew from 17.5 to 37 percent of a rapidly expanding governmentbudget between 1997/8 and 200213, as budget support increasedfrom $20m in 1998/9to over $130m in2001/2. By providing donors with a level of comfort in terms of allocation, implementation and transparency, the PAF enabledthe donors' shift fromprojectto budget and SWAP support. Recentanalysis has drawn attentionto the achievementsand pitfalls associatedwith Uganda's approach. Here we highlight some aspectsofrelevanceto Kenya. There are specific requirements for local governments and other government departments to report PAF disbursements. Reports are made public and discussedin open quarterly meetings, where civil society, the press and donors are present. And 5 percent of PAF funds are earmarked for enhanced monitoring and accountability. Thus the PAF provided a platform for establishment of an open and transparent process of budgetreportingandreview, andimprovedthe focus on the results of government's programs. A number of the achievements can be directly attributed to the PAF, although it must be emphasizedthat broader reform initiativessuch as the MTEF, SWAPS,and the PRSP combinedwith a supportive political, institutional and policy environment have played a major part. Moreover, some aspects of the PAF are problematic andcouldpotentiallyunderminethe achievementofUganda'spoverty reductiongoals: Among the key lessons drawn by Williamson and Sudharshan (2003) are that VPFs can make a positive contribution to public expenditure management systems by protecting budget disbursements and tracking pro poor expenditures, especially where public expenditure management systems are weak. Such elements of VPFs are relativelystraightfonvardto establish and implement. Fromthe outset a VPF shouldbe simple, and limitedto the identificationof PRSP priority expenditures inthe budget classification system; tracking of performance of these expenditures within transparent budget-wide reporting systems; and the protection o f disbursement, which shouldbe linkedto a systemof limiting overspendinginother partsof the budget. A VPF should supportrather thanreplaces the implementationof such comprehensiveimprovementsinbudget preparationandimplementation. Source: Williamsonand Sudharshan(2003). PublicExpenditureManagement 4.101 There are major weaknesses in Kenya's system and practice o f public expenditure management that undermine policy efforts to address poverty and inequality. These relate to the comprehensiveness o f the budget, the preparation process and execution. Many o f these problems are well-known and much discussed, and the recent IMF Fiscal ROSC (March 2008), and the PEFA (2006) provides an excellent overview. Others appear to have been relatively neglected. Here we provide a brief overview. Comprehensivenessof the Budget 4.102 There is significant spending in Kenya which takes place outside the main budgetary management rules and procedures. Resources may be partly or wholly reflected in the budget as earmarked revenues or transfers and spent by semi-autonomous government agencies (SAGAS) and extra-budgetary funds (EBFs) and Fund Accounts (FAs). Spending by these entities is not included in the annual accounts of central government. These entities have their own supervisory boards and financial management rules and procedures including audit. 162 4.103 There are approximately 101 SAGAS and EBFs operating in a wide range o f activities. The Road Fund is one example that was discussed above. Hospitals and Universities are classic autonomous entities, whose spending is mainly funded from budgetary transfers topped up with user fees. Among the FAs i s the CDF. Table 4-10: Semi-autonomousGovernmentAgencies and Extra-Budgetary Funds,By Sector, 2007 Internal Security 1 Trade & industry 6 Law 1 Personnel Management 2 Justice & Constitution 4 Education 15 Planning 3 Gender, Sports, Culture & Social 5 Information & 4 Services communication Regional development 6 Livestock & Fisheries 2 AIDS 1 Agriculture 4 Water & Irrigation 14 National Heritage 3 Health 5 Environment 4 Science & Technology 8 Road 1 Cooperative development 1 Tourism & wildlife 6 Transport 3 State corporation 1 Home affairs 1 Source: IMF ROSC 2007 BudgetPreparation 4.104 The budget preparation process has become more open and transparent since reforms in 2005/06 (see ROSC and PEFA). The changes included integrating the MTEF and the arinual budget process by bringingthem together inthe Treasury; line ministries producing pre-budget papers (the budget outlook paper and the budget strategy paper); and seeking more political engagement at various stages inthe process. 4.105 Despite this significant progress, several challenges need to be addressed inorder to enable resources to be allocatedtransparently across ministries and in line with national priorities, including: Improving the effectiveness o f sector working groups in meditating the competing demands o f ministries within sectors. Either they need to be made more effective, or a new model for allocating resources across ministries is needed. 0 Strengthening the capacity o f line ministries to prepare clear, feasible, and well-costed ministerial public expenditure reviews. Making the links between ministerial priorities, the budget process and the national priorities as presented in the MTP and its annual progress report explicit. This i s typically not done. One priority i s to strengthen incentives for ministries to allocate resources inline with MTP objectives, andto monitor and report on progress in achieving the-objectives. This could be done through the preparation o f ministerial strategic plans and annual work programs that specify the objectives for the coming year, name the specific activities that will be pursued to achieve the objectives, and provide milestones and indicators through which success can be measured. 0 Core poverty programs need to be better specified. As noted, many o f the MPERs and the budget outlook paper lack adequate descriptions o f the core poverty programs and fail to specify criteria for their selection. The 163 effectiveness o f increased spending on the identified programs in reducing poverty therefore cannot be easily monitored. 0 Procurement planning needs to be part o f budget preparation process (see below). 0 Involvement o f ministries' district offices informulating the budget needs to be deepened. District offices have not in the past participated actively in formulating the budget. As noted above, neither the MPERs nor the budget outlook paper, nor out turns, provide information on district-level spending. Strengthening the linkage between the recurrent and development budgets- these are formulated by different offices o f the government, each without reference to the other. Two unrelated documents are prepared andpublished. 4.106 A basic shortcoming which continues to characterize Kenyan public policy is the disconnect between budget intent and actual spending. The deviation inimplementing the budget as approved by parliament is a major concern, from both an accountability and development point o f view. The trend shows at best only limited improvements over time particularly in the development budget. The overall deviation was 16.7 percent in 2006107, which is worse than the preceding year's gap o f 13.2 percent. Development budget execution has been historically weak: the deviation averaged 43 percent in 1999/2000, but has improved somewhat to 28 percent. Table 4-11 provides the overall summary o f budget implementation and Table 4-12 summaries budget implementationby eight MTEF sectors. 4.107 Over 90 percent o f what i s allocated for recurrent purposes i s normally spent, largely on wages and salaries. However, certain public institutions such as the Ministry of Roads and Public Works, Energy and the State Law Office seem to be persistent under- spenders. Table 4-11:Budget under Spending: Recurrent and Development,percent, 199912000 -2006/07 199912000 2000/01 2001102 2002103 2004105 2005/06 2006107 Total 12.5 10.5 13.3 18.7 11.4 13.2 16.7 Recurrent 7.4 4.5 1.1 3.8 11.4 6.3 11.1 Development 54.1 34.5 59.6 57.2 35.4 33.3 28.2 4.108 There are serious implementation challenges in the development budget. While there has been some improvement, an average o f 60 percent i s still unacceptably low. This i s observed across all ministries and sectors, although health and transport are especially weak. Less than 30 percent o f development budget allocated to the Ministry of Health is spent despite the poor health infrastructure and health indicators. The most worrisome under-spending was on preventive and promotive health care, as cash flow bottlenecks to spending units and slow procurementpreventthe utilization o f funds. 164 Table 4-12: BudgetExecutionRatesby Sector 19991200&20006107 19991 20001 20011 20021 20041 20051 20061 2000 2001 2002 2003 2005 2006 2007 Recurrent Budget Agriculture and RuralDevelopment 103.5 97.9 89.1 72.5 81.2 92.8 94.2 Public Safety Law and Order 92.4 108.9 108.7 93.8 96.6 93.7 86.0 Health 99.2 110.4 113.9 99.8 104.6 96.3 83.5 Education 96.9 99.2 108.0 98.4 98.2 99.0 99.8 PhysicalInfrastructure 83.6 78.7 76.7 68.6 97.4 81.3 62.0 GeneralEconomic Services 131.2 90.0 125.0 86.5 87.6 85.4 96.3 Public Administration 66.2 76.2 73.9 88.4 88.8 88.6 88.4 National Security 102.2 102.6 110.5 99.0 99.4 95.5 85.5 Total Recurrent 92.6 95.5 98.9 91.7 96.2 93.7 88.9 Development Budget Agriculture andRural Development 21.4 59.9 78.0 100.3 25.7 60.9 69.0 Public Safety Law and Order 47.8 87.2 48.6 48.9 60.5 52.7 55.7 Health 36.9 23.9 57.8 22.3 13.3 32.7 22.4 Education 49.4 40.2 52.3 51.4 60.4 66.2 59.3 Physical Infrastructure 50.6 68.2 28.7 24.6 70.0 72.0 73.3 GeneralEconomic Services 59.3 28.9 62.7 70.6 65.3 45.2 80.5 Public Administration 80.2 87.4 11.4 42.1 82.9 85.9 91.3 National Security 33.6 Total Development 45.9 65.5 40.4 42.8 64.6 66.7 71.8 Total 87.5 89.5 86.7 89.5 86.7 81.3 88.6 Source: Ministry of Finance. 4.109 Underutilization of the development budget can be largely traced to ministries' lack of procurement planning which delays the issuance of contracts until well into the new financial year. Preparing procurement plans as the basis for budget preparation and implementation could lead to both more accurate budget submissions, facilitate the procurement process and timely budgetary releases. 4.110 Poor performance has been attributed to the poor inflows from the donor commitments. However, the problem of funds flow within government systems needs to be rectified and the capacity of government officials to implementprojects and programs strengthened. At the same time, greater realism is needed in development budget allocations, especially to those ministries experiencing capacity issues in implementing developmental programs. There is a need to systematically analyze the reasons for ministries' underutilization of resources, especially inthe ministries of health, education, and roads and public works. Clearly identifying the absorptive constraints would enable the government to identifysolutions to overcome them. Emerging Challenges 4.111 Much more needs to be done to ensure that resources are allocated transparently across ministries and in line with national priorities. In addition to strengthening the 165 process, the government needs to focus on reorienting expenditure to better reflect national and sectoral spending priorities. Several broad options have already been highlighted in earlier work-including the World Bank's Public Expenditure Review of 2004, the Budget Strategy paper 2005/06 and various studies by KIPPRA on expenditure efficiencies, including: Identification in the budget estimates of potential savings from improvements in technical and allocative efficiency. This would include scaling back or closing down programs and interventions that are not effective or do not achieve their desiredobjectives. Reorienting the budget across economic and functional categories of expenditure to support growth and poverty reduction. 4.112 Clearly however, further more specific work i s neededto make progress on this important agenda, and to get beyond generalities. It appears that key steps include better costing of ministerial programs and interventions and more systematically evaluating programs in terms of distributive and efficiency as well as institutional impacts, and sustainability. 4.113 We now turnto present some analysis which aimed to understandthe distributive impacts of spendingintwo key sectors, education and health, by way of illustrationfor the betterunderstandingofequity dimensions that is needed. BenefitIncidenceof PublicEducationandHealthSpending Education 4.114 Education is fundamental to the Government of Kenya's overall development strategy, as embodied inthe Vision 2030 and related strategic documents. There is a 2-8- 4-4 structure, 2 years of pre-primary or Early Childhood Development Education (ECDE), 8 years of primary schooling, 4 years of secondaryhechnical education, and 4 years of tertiary education. One key objective i s to improve equity in the education system, this section assessesthe extent to which this i s attainedvia the budget (Box 4-8). 4.115 Education outcomes in any country reflect two main influences: household behavior in deciding to enroll their children in school, and government spending decisions. Both influences have changed over the recent past in Kenya, influencing how education spending is distributedacross the population. 4.116 We saw in Part I1and I11that most provinces have seen significant increases in primary enrolments since the early 1990s. But, in some provinces (Central, Nyanza, Rift Valley) the improvements have been less striking, and that access rates are lower for the poor. The net primary enrolment rate was almost 90 percent among the richest groups in 2005/06, but only just over 70 percent for the poorest, even if the increase in primary enrolments since 2003 has come mainly from poorer groups. The net enrolment rate among the poorest quintile increased sharply in the early 2000s (from just 61 percent in 166 2003 to 73 percent in 2005/06). While overall primary enrolments compares favorably withother countries, enrolments amongthe poorest are lagging (Table 4-13). Table 4-13: NetPrimary Enrolment Ratesby Wealth Quintile, SelectedCountries Kenya Tanzania Mozambique Uganda Bangladesh Colombia Quintile (2003) (2004) (2003) (2006) (2004) (2005) Poorest 61 58 65 72 78 68 2nd 80 66 68 80 88 76 31d 84 73 69 82 89 81 4& 88 82 72 86 92 83 Richest 86 88 79 89 92 85 All quintiles 79 73 71 82 87 78 Source: DHS. Box 4-8: BenefitIncidenceMethodology Benefit incidence brings together data on the spending (subsidies) by the public sector on selectedservices, and informationon the use of such services by households-the former from the budget accounts of the sectors concerned, andthe latter throughhouseholdsurveys. Estimatesofbenefit incidenceare obtainedby J First, estimatingthe (unit) subsidyinvolvedfor every public schoolenrolment (primary, secondary and tertiary) J Second, identifying (firom the household surveys) the households who use the publicly subsidized service-from the poorestto the richest. J Third, imputingto those households the unit subsidy for every enrolled child-the subsidy depending on the levelofeducationat whichthe childis enrolled. 4.117 The evolution of the government education budget, its allocation across the different levels of education, and the enrolment decisions of households affect who benefits from public education spending. Using government recurrent spending by level of education and survey-based estimates of school enrolment, unit expenditure is obtained for primary, secondary and tertiary, and show that tertiary education attracts much larger public spending per pupil than other levels: we estimate that the tertiary unit spending was some 20 times the primary spending in 2005-06, which is similar to estimates from the Ministry of Education, while noting that the high relative weight to tertiary has declined since the early 1990s, when the tertiary unit subsidy was over 30 times the primary subsidy. Secondary spending per pupil has remained about triple primary amount. 4.118 Therehavebeensignificant changes inthe distribution of the education budget in the early 2000s. It is not so much that budget allocations among the sub-sectors have changed (in percentage terms at least), but household behavior has changed, thus modifying their claims on the education budget, as poor households have claimed a growing share of public primary spending. The Free Primary Education policy encouraged a sharp increase in primary enrolments, especially among poorer Kenyans. The share of the poorest quintile of the primary budget increased from 22 percent in 1992/93 to 25 percent in 2005. At the other end of the wealth spectrum, the richest 167 quintileenrolled far fewer children inpublicprimary schools, andits share ofthe spending fell, from 17to 10percent over the same period. 4.119 The poor have also gained more from the secondary spending. The share going to the poorest quintile rose (from 7 to 10 percent between 1992/93 and 2005). The share to the richest quintilefell-from 30 percent to 24 percent. 4.120 The government introduced a Secondary Education Subsidy Scheme (SESS) more than a decade ago. The introduction of Free Primary Education (FPE) in 2003 has resulted in increasing numbers of students completing primary education. Not all, and especially children from poor households and from ASAL regions, have access to secondary education. A major innovation this year has been the announcement of free secondary day schooling in academic year 2008, although reports suggest that still contribute significant amounts to fundtheir children to attend secondary schools. 4.121 Poor people get little from tertiary spending. As is typical inAfrica, the poorest groups do not benefit from spending on tertiary education, with no change over time (at about 2 percent of the tertiary budget). The growth in university enrolments has clearly come mainly from better off households. The richest quintile has seen its share of the tertiary budget increase sharply since the early 1990s. Figure 4-4: BenefitIncidenceof Government EducationSpending by Level and Quintile, 2005/06 (percent) 80 70 60 a so 40 p1 30 20 10 0 Primary Secondary Tertiary AU education 168 Figure 4-5: Benefit Incidenceof Government Education Spending by Level and RichestL'oorest Quintile, 1992/93and 2005/06 (percent) 80 7 0 60 50 40 30 20 10 0 4.122 Relative to their low levels of consumption, the poorest quintile gains from education spending, at least on primary school. The current &kind benefits gained from schooling are equivalent to 35 percent of their consumption. Although larger in absolute terms, the gains of households inthe richest quintile from government education spending amount to only 4 percent o ftheir consumption. Mean Subsidy Mean HouseholdConsumption KShs Subsidy as a share of Consumption(percent) Poorest 1,954 324 34 2,312 6,546 29.9 4.9 0.5 35.3 Second 2,004 573 36 2,612 11,950 16.8 4.8 0.3 21.9 Third 1,708 739 126 2,574 17,402 9.8 4.2 0.7 14.8 Fourth 1,455 912 342 2,710 26,665 5.5 3.4 1.3 10.2 Richest 795 784 1,256 2,835 69,740 1.1 1.1 1.8 4.1 Kenya 1,583 666 359 2,609 26,457 6.0 2.5 1.4 9.9 4.123 The basic benefit incidence estimates presented above assume that government spending per enrolled pupil does not vary geographically-it remains the same wherever the child is enrolled. Education spending and enrolment data by district are available for primary schools. These indicate that disparities inspendingper pupil across the provinces are modest. Eastern Province appears to benefit the most, while Nairobi and North Eastern received the lowest subsidy per pupil. These data, following closely on a 169 significant increase inprimary enrolments, could reflect differences inthe surges inschool enrolments. Table 4-15: Unit District Spending on Primary Educationby Province, 200906 Unit Subsidy Ratio to the mean Ksh Highest Lowest Highest Lowest Coast 6,943 7,455 6,059 1.07 0.87 Central 6,668 7,195 5,664 1.08 0.85 Eastern 7,220 8,359 6,264 1.16 0.87 Nairobi 6,579 Rift Valley 6,692 7,576 2,859 1.13 0.43 Western 6,882 7,350 5,122 1.07 0.74 Nyanza 7,057 9,9 12 3,159 1.40 0.45 North Eastern 6,437 6,596 5,545 1.02 0.86 All Kenya 6,888 9,912 2,859 1.44 0.42 Source: Ministry o fEducation 4.124 There is much greater variation across districts in unit spending especially in Nyanza and Rift Valley. While the majority of districts (52) received disbursements for primary education greater than Ksh6,500, a sizeable minority (17 districts) received less, some significantly less. Inthree districts inparticular, the disbursement was very low-in the Rift Valley Bomet (Ksh4,735) andNandiNorth (Ksh2,859), and inNyanza, Kisumu (Ksh3,159). 4.125 An econometric investigation of whether variations in unit subsidy across districts have affected household enrolment decisions found a significant relationship between district primary school enrolment rates and the district primary unit spending. On average districts receiving higher primary public subsidies also have higher levels of enrolment. This suggests that lower unit spending i s not typically a result of surges in school enrolment. Rather the reverse-when governments spend more, households respondby sendingtheir children to school. 4.126 The following policy messages emerge from this analysis and related studies about education access, especially from impact evaluations: 0 Continued efforts are needed to raise enrolments among the poorest girls and boys. It is particularly low in Coast, Rift Valley and North Eastern provinces, where the policy efforts should be focused. Impact evaluation studies have shown that support to poor households can raise school enrolment. For example, one careful study concluded that providing free school uniforms raised school attendance by 6.5 percent. 0 It is important to consider innovations that can be usedto enable attendance- for example, deworming in some circumstances has been shown to make a markeddifference. 0 Ensure that the gains in primary school enrolments are not lost: Poor people responded to the FPE policy, but poor quality and the failure of children to 170 improve education attainment might eventually discourage attendance. Schooling standards at the primary level should be enhanced. Raise secondary enrolments among the poor. The poorest quintile gains little from secondary school spending. To frame appropriate interventions, research i s needed on the main constraints facing poor families in enrolling children in secondary school-candidates include poor educational attainment at primary level, costs of schooling, access to secondary school facilities and perceptions about relevances and benefits. There are close to 21,000 public and private primary schools in the country, but only about 4,800 at the secondary level. This sparse distribution likely makes access difficult and costly to poorer Kenyansliving inremoterural areas. The government recently announced free secondary education. However secondary school bursarieshave beenunder implementation for over a decade, and a recent evaluation suggests that the scheme is non-transparentand weakly targeted to poorer households. There is now a scheme explicitly aimed to increase access for poor students to secondary education, ensure their retention and enhance their completion rates. Health Spending 4.127 The foregoing analysis has drawn attention to the links between health and poverty in Kenya, which appear to go both ways. We also noted the worrying decline in health outcomesover time, andthe persistenceof inequalities inhealth outcomes. 4.128 According to a number of recent studies (including the National Health Sector Strategic Plan I1(NHSSP 11) and the 2003 Household Health Expenditure and Utilization Survey), access to health services has been limited, particularly in rural areas, and the quality of services has beeninconsistent and generally poor. The Kenya Service Provision Assessment Survey of 2004 found that only 57 percent of facilities could provide a basic package of child, maternal, reproductive health and HIV/AIDS services, and only 10 percent of clinics were able to provide 24-hour delivery services. Distance to facility, and facility access in general appears to be a major constraint for rural households. The distribution of facilities geographically is uneven. While most facilities serve large numbers (on average each facility in Kenya serves over 15,000 people), this varies by province, with Central, Rift Valley, Coast and Eastern being better placed than the others.73 4.129 The delivery of public health services has been constrained by systemic problems as well as inadequate funding. In recent years the focus has been on development expenditures in health, although as we saw above, the record on disbursements is poor. The Public Expenditure Tracking Study (2007) concluded that funding (especially development funding) is not reaching the operations units because of inefficient financial management. However, resource constraints are deeper than just financing. Ensuring that adequate numbers of qualified health staff with appropriate skills 73Note, these data refer only to public facilities. Clearly, development planning inhealthtakes into account the private sector activity acrossthe regions. 171 are available inthe right places remains a considerablechallenge. The distributionof staff across provinces is uneven. A study in 2005 found overstaffing inprovincial and district hospitals, and acute understaffing in many rural health facilities. Around 47 percent of dispensaries were staffed by only one Enrolled Community Nurse, and 3 percent had only support staff not qualified to administer drugs. Unattractive working and living conditions, poor incentives (including allowances), and weak deployment procedures all contribute to the staffing imbalances. 4.130 The KIHBS, alongside fiscal data, is usedto investigate such inequality further. The survey not only obtained information on the living standards of the Kenyan population, it identified which households used services subsidized by the government, including health-care services. By combining this information on the use of health services with data on how the government subsidized these services, we are able to estimate the benefit incidence of health spending in Kenya-at least for the year of the survey, 2005-06. Table 4-16: Distribution of PublicHealth Facilities by Province,2004 FACILITY TYPE CENTRAL COAST EASTERN NAIROBI NORTH NYANZA RIFT WESTERN TOTAL Dispensaries 205 144 325 18 43 180 540 81 1,527 HealthCentres 57 33 58 8 6 80 136 62 440 District Hospitals 12 11 26 1 10 24 21 13 118 Provincial Hospitals 1 1 2 1 1 1 1 8 National and Specialized Hospitals 1 2 3 Rural HealthTraining & DemonstrationCentres 1 15 7 5 6 12 7 53 Total Facilities 277 204 418 29 65 291 710 164 2,158 Facilities % 12.8 9.5 19.4 1.3 3 13.5 32.9 7.6 IO0 Population 3,918,538 2,860,649 5,180,139 2,656,997 1,235,592 4,868,010 8,077,517 3,954,081 32,751,523 Populationper facility 14,095 14,022 12,393 91,620 19,009 16,728 11,376 23,964 15,176 Facilitiesper 100,000 pop'n 7 7 8 1 5 6 9 4 7 Source: HumanResource Mapping Report, 2004. 4.131 The KIHBSfound quite similar incidence of illness across q ~ i n t i l e s .But while ~ ~ about a quarter of the population in all quintiles reported being ill or injured, 75 differences begin to emerge in the response to the illness. Poorer Kenyans are far less likely to seek treatment when sick. The better off are more likely to seek care, but typically chose private care. The poor when they seek treatment typically go to the public sector. Poor Kenyans also turn to private providers, many of whom (62 percent) are modern rather than traditional. When poorer Kenyans get sick and go to a government facility, in most cases (78 percent) they visit a primary facility-ither a dispensary or a health center. Better-off Kenyans are more likely to use hospital facilities-just under a half of those inthe top quintile seeking care ingovernment facilities go to a hospital. 74 Throughout this analysis households are ranked according to their mean per-adult equivalent total consumption. 75 Some have argued that illness is typically under-reported by poorer people because they view many illnesses (such as diarrhea) as a regular feature o f normal life. 172 4.132 Use of government services represents an in-kindtransfer to the household-the transfer being what the household would need if it was required to pay the hllcost of providingthe service. For each visit to a public health facility the household gains the unit cost attached to that visit. Because such costs vary by facility, households who use high- cost facilities gain a greater in-kindtransfer. Because these are seen as current transfers, only the recurrent budgetof the government i s relevant for this purpose. 4.133 To obtain estimates of unit costs in government health facilities we rely on the Shadow Health Budget, prepared for 2007. Each visit to a referral hospital costs on average KSh 1,463, 13 times higher than a visit to a primary facility. A visit to other hospitals was about 8 times more costly to the government. This is similar to that estimated for 1992/93 (Castro-Leal et al, 1999). Households using referral hospitals therefore gain a much greater in-kindtransfer than those usingprimaryfacilities. 4.134 Bringing together the unit costs and reported use of facilities from the KIHBS, estimates are obtained of the in-kind transfer to households across the consumption distribution(Figure 4-6). Figure 4-6: Quintile Distribution of BenefitIncidenceof GovernmentHealth Care Spending by Facility, 200906 60 50 40 I 30 & 20 10 0 Referralhospital District/provincial Primary facilities All publicfacilities hospital Source: Staff estimates based on budget data and KIHBS2005/06. 4.135 The results follow the pattern of use: 0 Poor people gain little directly from the significant spending on the highest level of health care delivery, since they hardly use referral hospitals. Half the consultations at referral hospitals are from the richest quintiles. Poor residents o f Nairobi may not consider use of the referral hospitals an option. The PPA- 173 IV found that 'most of the Nairobi communities felt that Kenyatta National Hospital was not really a public facility because of the high payments for medication and the harsh and negative attitude o f the medical personnel.' (MinistryofPlanning andNationalDevelopment, 2008, p. 79). The use of other hospitals is more widespread. But even here, the better-off use the facilities far more than the poor. The top three quintiles gains about a quarter of the subsidy, eachof the bottomtwo quintiles gainingjust 14percent. Poor people rely on primary care. They get a proportionate share of spending on primary facilities (the poorest 40 percent gain 40 percent of the primary subsidy). 4.136 The net result is that in absolute terms health spending is regressive in Kenya: poor people benefit less that the better off, Mean health spending going to the poorest quintile is only about half that going to the richest. The top three quintiles are far better placed. At the same time, the spending can also bejudged relative to consumption: inthis sense, health spending i s progressive. The in-kindsubsidy the poorest quintile gains fiom using government health facilities represents 6 percent o f their income/consumption. For the richest, the ratio is only 1percent. Table 4-17: Kenya: Per capitaHealth Subsidy Relative to Per Capita Total Household Consumption, by Quintile, 2005106 Healthsubsidy Householdconsumption Healthspendingbenefitas percenl householdconsumption ( K W ("/.I Poorestquintile 397 6,546 6.1 Quintile2 402 11,950 3.4 Quintile3 594 17,402 3.4 Quintile4 637 26,665 2.4 Richestquintile 754 69,740 1.1 All Kenya 557 26,457 2.1 Source: Staffestimatesbasedonbudgetdata andKIHBS2005/06. 4.137 The incidence of health spending has changed little over time. We see no marked changes in the distribution of the government health budget since 1992-93. Indeedsuchchangesthat didoccur are not encouraging since the richest quintilegained an increasing share o f all levels of service. The poorest quintile lost its share of even primary services. 174 Table 4-18: BenefitIncidenceof GovernmentHealth Spending by Poorest and Richest Quintile and Level of Care, 1992-93 and 2006-06 (percent share) Poorestquintile Richestquintile Primary Hospitalbased All health Primary facilities care facilities basedcare All health 1992-93 22 13 14 14 26 24 2005-06 20 13 14 16 30 27 Source: Staff calculations; Castro-Leal, et a1(1999). 4.138 We also examine gender dimensions o f health spending in Kenya. Females receive a larger share o f health spending almost 60 percent. However, poor females do - not do well. Relative to better off women, poorer women the poorest quintile are disadvantaged-particularly with respect to referral hospital care. And poor males get more benefit from referral hospitals than poor females-the opposite pattern from the better off. Better off females use all health facilities (including hospitals) more thanbetter off males. Not surprisingly poorer females get most out o f primary spending. 4.139 Finally, we examine how much households themselves spend on health care, which potentially increase inequality in health care access. We see immediately that households in fact assume the greater burden o f health care spending. Households account for almost three quarters o ftotal health care spending. 4.140 The distribution o f private spending reveals that disparities in private spending far exceed those inpublic spending. The meanpublic subsidy for health care received by the top quintile i s about twice that for the bottom quintile, whereas for private spending the ratio is about seven to one. The burden o f health spending i s greater for the poor. Although poor people spend less in absolute terms on health care, for households in the poorest quintile this amounts to 8 percent o f consumption, compared to the richest quintile, which spend 6 percent on health. Table 4-19: Private and Public Spending on Health, 2005/6 Householdand Total government spendingas MeanAnnual HouseholdSpendingonHealth mean percentofhousehold Care annual consumption health Frequent Medical Mean spending health diagnosis All annual on items and household government publicly treatment Other health health provided drugs spending subsidy health Household Government (Ksh) (Percent) Poorestquintile 55 438 46 539 397 936 8.2 6.1 Quintile2 89 307 126 523 402 924 4.4 3.4 Quintile3 150 410 217 778 594 1,371 4.5 3.4 Quintile4 214 610 553 1,377 637 2,014 5.2 2.4 Richestquintile 544 2,035 1,686 4,264 754 5,019 6.1 1.1 All Kenya 210 760 525 1,496 557 2,053 5.7 2.1 Source: Staff estimates based on budget data and KIHBS 2005/06. 175 4.141 It is evident that the current waiver system is not working. The poorest households spend proportionately more o f their income than the richest, even though they use mostly primary facilities. The recent Public Expenditure Tracking Survey (2007) concluded that `the application o f waivers and exemption guidelines at facility level i s not standard and i s subject to abuse by the facility staff. In particular and against the stated procedures, the well to do groups in the society are being exempted from payments an issue which works against the principles o f equity and access to health services by the poor and the vulnerable groups. As we develop the strategy issues to do with poverty definitions and measurement should be addressed to ensure that the. poor groups access services.76 This i s a major concern to be addressed. 4.142 In sum, our analysis of health spending from the perspective of poverty and inequality suggests that: The pattern o f spending across the levels o f health-care delivery has remained largely unchanged over the past decade, despite policy pronouncements in favor o f primary care. The share o f the budget allocated to the higher levels of care has remainedhigh, with front line rural services suffering as a result. 0 Poor people tend to use lower levels o f services which receive lower budget allocations. Given their limitedaccess to hospital facilities, poorer people gain relatively little compared with the better-off from government subsidized hospital care. 0 While women overall benefit more than men from health spending, but poorer women appear to be disadvantaged relative to men. 0 Inequality inhousehold spending on health i s significant, and public subsidies do little to correct for this. 4.143 From a poverty and inequality perspective, the key health spending policy challenges appear to be two-fold. First, there is a need to improve the quality o f care at the primary level. These are the facilities used mostly by poor people. According to the PPA-IV, `the communities described lower level health institutions (dispensaries and health centers) as offering inadequate services, mainly because some services were not available, while district, provincial and national hospitals have more qualified doctors, are normally better stocked with drugs, and offer more specialized services.' (Ministry o f Planning, and National Development, 2008, p. 77). Second, it seems appropriate to encourage greater use by poor people o f district and provincial hospitals. This in turn would call for two types o f intervention, to improve access to district and provincial hospitals-better all-year roads-to improve the implementation o f fee exemptions which are likely to discourage the poor more than the non-poor. Many communities inthe PPA- IV reported that unofficial payments often were needed to obtain care-in addition to the scale o f charges levied inpublic facilities. These and other costs (transport, waiting time etc.) are likely to discourage use o f these services by poor people. 76Quoted from Muchiri(2008: p. 15). 176 MONITORINGANDEVALUATION 4.144 A reliable and effective monitoringand evaluation system is critical to assessing the impact of policies and programs aimed at reducing poverty. Having accurate information which helps to measure and understandthe causes of poverty and monitor changes over time can help policy-makers design and implement effective interventions. This is important not only in the context of annual planning and budgets, but also the Medium Term Plan and the Vision 2030. Well-structured evaluations allow decision- makers to assess the performance of various interventions, and therefore to make better informed decisions of how to allocate resources and manage programs to achieve stated objectives. 4.145 Kenya's capacity for generation and analysis of statistics was once among the strongest in Africa, but its capacity and performance has eroded over time. During the 1990s, budget cuts led to a reduction in staff, an inability to keep pace with technological innovation, and a fall instaff motivation. The frequency and quality of products produced by the Central Bureau of Statistics declined markedly. Similarly, capacity of line ministriesto collect, process, and disseminate informationdeteriorated, andmany stopped producing data altogether. 4.146 Inrecentyears, demandhas grownfor greater transparency and more systematic information about key dimensions of development. The IP-ERS sets out 31 summary indicators that are intendedto provide a complete picture of progress inpoverty reduction over time (Table 4-20). On the institutional side, the government with support of the development partners, has begun to take steps to strengthen its monitoring and evaluation systems. For example, the government developed a strategic planfor the statistical system in March 2003, and several development partners are supporting its implementation. In the IP-ERS, the government committed itself to the development of a national integrated monitoring and evaluation system (known as NIMES). This is aimed at providing reliable measurement of the efficiency and effectiveness of government policies and programs, linkedto the budget. 177 Table 4-20: IP-ERS: APR IndicatorsIncluding PerformanceRevisedTargets IP-ERS Objective Outcome Indicator Accelerate economic growth RealannualGDP growth rate (%) Achieve and maintain price stability Annual underlyingrate of inflation(%) Control and reducethe burden of domestic debt Stock of domestic debt/GDP ( O h ) Control public sector wage bill Public Sector Wage BiIl/GDP ( O h ) Expand revenue sources Revenue/GDP (%) (including AINGDP ratio) Reform public expenditure management Benchmark score on PEM-AAP Rehabilitate the road network Proportionof road network inbadlpoor condition ( O h ) Increasepower coverage inrural areas Percentageo f ruralhouseholds served (%) Increaserural water coverage Percentageo f ruralhouseholds with safe and reliable access to water (%) Increaseurban water coverage Proportion of urbanhouseholds with safe and reliable access to water (%) Enhance safer road system Number of fatalities on the road annually Enhance Annual growth rateoftourists (%) Strengthen trade and industry Growthof volume o f exports ("A) Reduce infant mortality Fully immunized children as %of less than 1yrs pop. Reduce HIV/Aids prevalence Proportion of pregnant women aged between 15 -24 years attendingANC who are HIV positive (%) Reduce maternal morality Percentageof pregnant women attending ANC (at least 4 visits) (R) Reduce the burdenof disease Inpatient malaria morbidity as a percentageo f total inpatient morbidity Improveprimary net enrolment (NER) Primary net enrolment ("A) Increasenorth easternprovince net enrolment rate Primary net enrolment rate FORNEP (%) Reducethe rate of primary school drop outs Primary school completionrate ("A) Reducethe incidence of primary repetition Primary repetition rate Increasetransition rate o f pupils to secondary school Primary to secondary school transition rate (%) Strengthen employment creation and productivity Measured labor productivity in formal sector Raise incomes of small holders Agricultural sector growth rate ("A) Improve environmental management Forest areaprotected by gazettement(Ha) Introduce universal environmental screening projects Proportionof public sector projects subjected to environmental impact assessments Reduce absolute poverty Proportion of people below absolute poverty line ("4.) Reduce prevalence of underweight children inunder 5 yrs Prevalencerate of underweight children's in o f age underweight 5 yr olds (%) Enhance more expeditious justice Ratio o f concluded casesto reported cases (%) % change inDomestic Crime Index by 2009 % changeinoverall ConuptionPerception Index %of citizens who report that they have accesstojustice systems to resolve disputes Proportion of total public sector spending controlled and Public sectortransparency and devolution of power managedat local level (based on LA'S,LATF, DRCS and CbF) Public sector wage bill/GDP ("A) Fullv functionine national svstemo f M&E ooerating at Creation of national monitoring and evaluation system all Iivelsand &vidina -feeiback to the bud& by t& - . end of 2005. 4.147 Recent initiatives have resulted in various institutional reforms, although these have not yet demonstrated results. A new statistics act became effective inFebruary 2007. This transformed the bureau o f statistics into a semi-autonomous corporate body with a governing board-the KNBS-and sets out the legal framework for the national statistics system. Having autonomy allows the bureau to pay higher salaries than the civil service and to assess and reward performance. This transformation seems to have prompted a large number o f its staff to undertake training in advanced survey and statistical techniques. However, one major reform which has been discussed but not yet adopted is a data access policy, which i s needed to improve access to data by policy-makers, researchers, development practitioners, andthe general public. 4.148 Analyzes o f key aspects o f poverty, inequality and development in Kenya have been delayed or prevented by the absence of recent reliable nationwide data. Specific surveys-including those carried out and analyzed by the Tegemeo Institute and the APHRC-have been carried out over time, and provide invaluable knowledge on 178 important aspects of development. But nationally representative surveys have been either infrequent andor unavailable for analysis. For example, despite evident government interest andthe existence of a national steering committee, the KNBS has not carried out a survey of employment since 1998. Release o f the KIHBS data to researchers has been delayed for over 18 months. However, the KNBS i s now taking steps to address data access issues andthe situation i s expectedto improve in2009. 4.149 The recent crisis has heightenedthe need for reliable andtimely data. While the U N ' s World Food Program and Food and Agriculture Organization collect data through their own monitoring networks, specific follow-up surveys are needed to gauge the impacts o f the crisis, to design interventions to mitigate the impacts, and to assess the outcomes o f the interventions. This is where links to non-governmental and think-tank efforts become evenmore critical. 4.150 In this section we review the status of initiatives to improve monitoring and evaluation in Kenya, and the key challenges to reform. We begin with an overview of available survey and administrative data and how these are being used in practice, including the attempted shift to results based management. The general finding is that while the overall stated directions are welcome, implementation has been very slow. Much needs to be done to generatethe informationthat policy makers needto improve the effectiveness of efforts to eliminate poverty. A BriefOverview 4.151 Inmany ways, Kenya is a data-rich country. Its development challenges have attracted scores of world-class local and international researchers, many of whom have commissioned their own purposive surveys. There are a variety of national-level data, some of which are collected through household and individual surveys, and more general administrative informationcollected through routine official data collection and reporting. Some major statistical products and initiatives relevant to poverty and inequality are highlighted below. 4.152 Surveys aimed at better understanding poverty. To date, the government has assessed aspects of poverty and wellbeing through regular household surveys. Comprehensive household budget surveys were administered in 1992, 1994, 1997 and 2005/6. Demographic and health surveys have been conducted every five years or so starting in 1989. A population and housing census, carried out every decade (the most recent being 1999), includes several dimensions of poverty, including school attendance and attainment, labor force participation, housing, water, and so on. The government has produced several pieces of analytical work, including the basic report on the KIHBS (August 2007), an atlas of ecosystems and humanwell-being (March 2007) that examines the links betweenavailability of environmental resources(land, water, forest products, and 179 wildlife) and household con~umption.)'~ and two volumes of poverty maps, publishedin 2004 and in2005.78 4.153 The timeliness of data release from national surveys is a major issue. Analysis for this poverty assessment was delayed almost two years inthe absence of access to the KIHBSdata-which are still not available to researchersgenerally. Box 4-9: Delaysin ConductingLabour Market Surveys A jobs study planned for 2007 was severely impeded by the absence o f recent data on labor market issues, whether on f m s or workers. Indeed the report concluded that the availability o f data i s probably now inferior to that which was collected pre-1960. I t recommended that the government collect essential information relevant to the labor market on a regular and timely basis. The KNBS does not have a fm schedule for repeating the relevant labour surveys. In practice there are significant lapses o f time between the surveys which, among other things, makes preservation and access to data problematic. Some sector ministries periodically compile jobs related information for their own mandated activities. For example, various ministries collect information on minimum wages, data on skills and school, college, and university graduation, and employment conditions inthe export processing zones. However, these data may not be compatible with the approaches and defmitions used by the KNBS, nor is such information housed inthe KNBS. 4.154 Qualitative Data. Several qualitative surveys have been undertaken in recent years using participatory approaches to deepen understanding of how communities perceive their living standards and the poverty in their midst. One such initiative was the fourth participatory poverty assessment (PPA-IV). The PPA-IV covered a sub-sample of the KIHBS, allowing for a mixed analysis of the poverty situation in Kenya, combining structured and participatory approaches. One of the significant benefits o f the PPA-IV survey is that it permitteda more in-depthenquiry into household histories, exploring how and why the poverty status of households has changed over time (see Part I11above). Another example i s the World Bank's 2006 study o f poverty in Nairobi slums,79which combined quantitative data from households living in Nairobi slums with qualitative informationgatheredthrough community questionnaires and focus group discussions. 4.155 Think-tanks and Non-Governmental Efforts. Several world-renowned survey efforts have been established in Kenya outside o f government, which regularly generate valuable data. For example: 0 Tegemeo Institute has been collecting panel data from rural households since 1997. This data have provided improved insights on the impact of policies implementedsince the mid 1990sonrural livelihoods. The African Population and Health Research Center (APHRC) i s carrying out panelsurveys to investigate the linkages betweenurbanization, poverty, health, 77 The atlas was produced collaboratively with the World Resources Institute, International Livestock Research Institute, Kenya Ministry o fEnvironment andNaturalResources, and the KNBS. 78Central Bureau o f Statistics, "Geographic Dimensions o f Well Being inKenya." 79World Bank, 2006, "Kenya: Inside Informality: Poverty, Jobs, Housing and Services inNairobi's Slums." Report 36347-KE, Water and UrbanUnit 1, Afiica Region, World Bank, May 31. 180 and migrationinto and out of slums throughout the course of life. The surveys are also aimed at investigating the impact of interventions designedto improve the well-being of the urbanpoor in Kenya (see X above). The findings of the surveys are being used to inform policy makers o f effective approaches to reduce infant and child mortality rates, cut incidence of malaria, and slow the spreadof HIV/AIDS inNairobi slums. 0 Transparency International Kenya has been undertaking a bribery survey each year since 2001. The findings of the survey are used to rank public organizations with respect to daily experiences o f ordinary Kenyans with bribery. Since 2006, FinAccess i s conducting a household survey every two years to establish levels of access to and demand for financial services by Kenyans across the country, beginning in2006. Collection and Dissemination of Ministerial-level Data 4.156 Kenya has yet to establish a systematic approachto monitoring and evaluation at the ministerial level. Severalministriesdo collect reasonably timely and reliable data on outputs and outcomes of their programs, which they provide to the KNBS for dissemination through its annual Statistical Abstract and Economic Survey. For example, the Ministry of Health collects and reports information on the proportionof children fully immunized and outpatient morbidity, disaggregated by province.8o The Ministry of Agriculture, the Ministry of Livestock and Fisheries Development and various marketing boards monitors output and sales of major agricultural commodities. 4.157 However, much needs to be done to improve the quality and dissemination of data from even those ministries with relatively strong monitoring systems. For example, the Education Management Information System does not allow one to access data on primary enrolment andcompletion ratesandthe secondaryenrolment rate, beyond a single year. It seems that the ministries may need to rationalize their management information systems to eliminate duplication, to introduce electronic formats and to coordinate data collection and entry. More systematic sharing of information between facilities, district ministry offices, and the central ministry is required. More timely dissemination of informationwithinministriesand to the broaderpublic is needed. 4.158 District-level Monitoring. To supplement data gathered through surveys and routine processes, the government piloted an initiative in 2005 to collect district-level monitoring and evaluation data. The Ministry o f Planning sent questionnaires to district officers seeking information on services (education, health, security, judiciary, agriculture, and environment); infrastructure (water and sanitation, roads and electricity); and poverty alleviation and governance (public safety, law and order, justice, devolution of power, gender, and corruption). This exercise revealed significant weaknesses in capacity for district level monitoring, but also pointed to the way forward. Nearly half the districts did not respond and those that did were not able to present all the information requested, The Ministry of Health was split into the Ministry of Public Health and the Ministryo f Medical Services inApril 2008. 181 because the local offices of central ministries did not respond. Understandably, the data that were most readily available were those that are collected routinely, such as school enrolment. Little data were provided on issues related to governance or the effectiveness of programs, as this an area where no systematic data collection has beendone inthe past. Key recommendations of the pilot included (a) establish well-trained district teams which are conversant with the national monitoring, indicators; (b) improve the concepts and methodologies for obtaining data at the district level; (c) integrate the district M&E system into the planning and budgeting framework; and (d) harmonizing the reporting system for all the funds at the district and constituency level. However, whether or not the government will adequately fund a district-level monitoring program in the future is not yet clear. FiscalData 4.159 Kenyahas made significant progress since 2003 inimproving fiscal transparency andaccountability (see IMF ROSC 2008). Recent legislation-including the Government Financial Management Act (2004), the Public Procurement and Disposal Act (2005), the Public Audit Act (2003), and the Statistics Act-have established some key building blocks for the construction of a sound fiscal managementsystem. 4.160 However, much more is needed for Kenya to establish a sound public financial management system that is in line with international best practice. This has hamperedthe analysis in this assessment, and also constrains our fiscal understanding more generally. Inparticular, as highlighted above: 0 Although the government upgraded the economic classifications in 2005/06 to Government Finance Statistics (GFS) 2001principles, the sectoral breakdown i s not compatible with the GFS functional classification. Program classification has not yet beendeveloped. 0 The coverage and reliability of in-year accounting and reporting remain inadequate. The introduction of the computerized budget execution and accounting (integrated financial management information system, IFMIS) across government is underway but has still not been completed. Making the accountant general's department responsible for the collection, consolidation, and provision of in-year information on budget execution would help ensure that all users have access to the data inatimely manner. Fiscal reporting does not cover all o f general government and there are gaps in the coverage of central government in in-year budget outturn reports. Audited annual accounts of central government operations are still not issuedwithin the prescribedperiod and audit reports for local governmentsremain inarrears. As noted above, the fragmentation of the budget and the lack of reporting on local spendingis a major issue. The public financial management reform program, currently under implementation, is expectedto address these deficiencies inthe coming years. 182 4.161 Results-based Management. Efforts began in 2005 to introduce a results based management (RBM) system aimed at creating a more effective, efficient, and ethical public administration. The Public Sector Reform and Development Secretariat was established in the Office of the President to oversee its implementation. Performance contracts for key public officers that aim to provide strong incentives for all senior officers to ensure that specific targets derived from the ministerial strategic plans are met. Independent observers are to assess performance at the end of the year. General information on the performance contracts and the findings of the assessments is available to the public, but not the content of specific contracts and assessments. The introduction of RBMappears to have improved service delivery in some ministries and agencies. The Grand Coalition government, with 40 ministries will make both the administration of the systemand an assessment of results more challenging. 4.162 Kenya launched its national integrated monitoring and evaluation system (NIMES) inSeptember 2007. The goal is to strengthenthe linkage betweengovernment's planning and budgeting processes, and monitoring and evaluation. Ministries are encouraged to prepare their strategic plans, annual work programs, and budget submissions with monitorable outcomes in mind. The ministry's targets are these specified in the national development strategy. The milestones and outcome indicators will then be specified in performance contracts of public officials. In order to function, this obviously requires better information on results. The NIMES also includes a set of standardized formats for project-level M&E, which project managers will be required to use in quarterly and annual reporting on inputs, outputs, and outcomes o f project expenditures. 4.163 The M&E department of the Ministry of Planning and National Development i s overseeing the implementation of the NIMES. It intends to produce annual progress reports on implementation of the development strategy in time for the annual budget preparation exercise and for the annual staff performance evaluation process. Building Capacityfor Evaluation 4.164 There are several examples of successful evaluations, sometimes undertaken in concert with development partners. For example, the World Bank supported Western Kenya Community Driven Development Project includes funding for an independent evaluation of the project's impact on poverty and well-being, using a randomized evaluation design during roll-out o f the project. There have been a whole series of evaluations of HIV/ART programs (see Part 111) and school-basedinterventions. 4.165 Independent researchers have conducted some highly influential impact evaluations. The poverty lab at MIT in collaboration with civil society organizations has sponsored a number of impact evaluations looking at among other things the effect on school attendance of free distribution at school o f deworming medication, o f meals, and of girl's uniforms. The results of these evaluations have influenced government policy in important ways. For example, the discovery that the free distribution of deworming medication at school led to improved attendancerates encouraged the government to fund a deworming intervention. 183 Challenges,ConclusionsandRecommendations 4.166 The government has candidly recognized the major challenges to improving the comprehensiveness, reliability, and timeliness of economic, social, and poverty data. Some of these were highlighted inthe most recent Annual Progress Report for the IP-ERS, dated May 2006, including inadequate budgetary allocations resulting in limitedcapacity, lack of awareness of the role and importance of M&E and insufficient linkages across initiatives. 4.167 Moving ahead requires better coordination and collaboration within government, across sometimes competing initiatives. Surveys will play an important role, but so will proxies which rely on more frequently available information. As we have seen from the foregoing analysis, there are large disparities inKenya, which are an important part of the development story. This underlinesthe importance of continued efforts to develop district levelmonitoringandcapacity. 4.168 Itis recommendedthat the governmentthrough its mediumterm planand related initiatives, move more actively to strengthen underlying capacity and to help promote more evidence-based approaches, muchmore is needed. Inthis sense, it is a concern that the MediumTerm Plan (MTP), 2008-2012 neglectsthe topic of monitoring and evaluation altogether, and that its implementation matrix contains very few baseline or target values. Most of the MTP targets that are includedare presentedinmore generalterms. 184 REFERENCES Agesa, Richard U., Jacqueline Agesa, and Andrew Dabalen (2006). "The Urban Real Wage Decline and Increased Wage Inequalities in Kenya: 1986-1999." Division of Finance and Economics at Marshall University and The World Bank. January, 2006. Agesa, Richard U., and Sunwoong Kim (2001). "Rural to Urban Migration as a Household Decision: Evidence from Kenya." Review of Development Economics, Vol. 5(1): 60-75. Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg (2003). "Fractionalization.)' Journal of Economic Growth 8 (2): 155-94. Andalon, Mabel, and Carmen Pages (2008). "Minimum wages in Kenya." IZA DP No. 3390, Institutefor the Study of Labor (IZA), Bonn. Appleton, Simon, Arne Bigsten, and Damiano K. Manda (1999). "Educational Expansion and Economic Decline: Returns to Education in Kenya, 1978-95." CSAE Working Paper 90, Centre for the Study of African Economies, Oxford. Ariga, Joshua: "Factors Driving the Growth in Fertilizer Consumption in Kenya, 1990- 2005". Tegemeo Working paper 24/2006. Tegemeo Institute o f Agricultural Policy andDevelopment. 2006 Awuor, Thomas; Njuguna, Joseph; Kute, Allan; Kariuki, John; Ityengi, Didacus 0.; Ferrand, Cyril: "KENYA: Rapid Food Security Assessment." FAO, on behalf of the KenyaSteeringFood Security Group. January, 2008 Barber, Gerald (1987). "The 1986 Urban Labor Force Survey: Description and Structure of Database" Technical Paper 87-08. Long Range Planning Division. Ministry of Planning andNationalDevelopment, Republic of Kenya. October, 1987. Barro, Robert J., and Jong-Wha Lee (2000). "International Data on Educational Attainment: Updates and Implications." CID Working Paper No. 42, Center for International Development, HarvardUniversity, Cambridge MA. Bates, Robert H. (2005). "Beyond the Miracle o f the Market: The Political Economy of Agrarian Development inKenya," Cambridge University Press, Cambridge, New York. 185 Beck, Thorsten, Asli Demirguc-Kunt, and Maria Soledad Martinez Peria (2007). "Reaching Out: Access to and Use of banking Services Across Countries." Journal of Financial Economics, Vol. 85( l),234-266. Beck, Thorsten, and Ross Levine (2004). "Stock Markets, Banks, and Growth: Panel Evidence." Journal of Banking and Finance, Vol. 28, Issue 3: 423-442. Berg, Andy, Jonathan D. Ostry, and Jeromin Zettelmeyer (2006). "What Makes Growth Sustained." IMF Working Paper 08/59, International Monetary Fund, Washington DC. Bigsten, Arne, and Peter Kimuyu (2002). "Structure and Performance of Manufacturing inKenya." PalgraveMacmillan, Hampshire. Bigsten, Arne, Peter Kimuyu, and Karl Lundvall (2004). "What to Do with the Informal Sector." Development Policy Review, 2004,22 (6): 701-715. Bigsten, Arne, and Mans Soderbom (2006). "What Have We Learned from a Decade of Manufacturing Enterprise Surveys in Africa?" World Bank Research Observer 21(2): 241-65. Bosworth, Barry P., and Susan M.Collins (2003). "The Empirics of Growth: An Update." Brookings Papers on Economic Activiv, No. 2: 113-206. Bressie, Kent, Michael Kende, and Howard Williams (2005). "Telecommunications Trade Liberalisation and the WTO." Info-The Journal of Policy, Regulation and Strategyfor Telecommunications 7 (2): 3-24. Broadman, Harry (2007) "Afiica's Silk Road: China and India's New Economic Frontier" The World Bank.2007. Bruno R. and Simon Commander with J. Macharia and J. Ogai (2006). "Institutions of the Kenyan Labour Market and their Consequences." Background paper for Jobs inKenya, GOK-BankReport. Cadot, 0 and J. Nasir (2001). "Incentives and Obstacles to Growth: Lessons from Manufacturing Case Studied in Madagascar." Regional Program in Enterprise Development No. 117, World Bank, Washington DC. Castro-Leal, Florencia, Julia Dayton and Lionel Demery (1999). "Public Social Spending in Africa: Do the Poor Benefit?" World Bank Research Observer, 14(1): 49-72. Cheng, Kevin (2004) "Growth Performance in Kenya during 1980-2004," by K Cheng, in Kenya: Selected Issues and Statistical Appendix, IMF Country Report No. 04/418, International Monetary Fund, Washington, DC. 186 Christiaensen, Luc and Kalanidhi Subbarao (2004). "Towards an Understanding of Household Vulnerability inRural Kenya." World Bank Policy ResearchWorking Paper 3326, World Bank: Washington DC. Christiaensen, Luc and Kalanidhi Subbarao (2005). "Towards an Understanding of HouseholdVulnerability inRuralKenya." Journal of African Economies, Vol. 14, Number4, pp. 520-558. Christiaensen, L., Lanjouw, P., Luto, J., and Stifel, D. (2008) "Estimating Poverty Across Non-Comparable Surveys: Evidence from Vietnam, Russia and Kenya." Draft, forthcoming. Cohen, John. M.,(1995): "Ethnicity, ForeignAid, and Economic Growth inSub-Saharan Africa: The Case of Kenya." HIID Development Discussion Paper No. 520, Harvard Institutefor International Development, Cambridge, MA. Collier, Paul, and Stephen A. O'Connell (2006). "Opportunities and Choices" Draft chapter for Explaining African Growth Project, African Economic Research Consortium, Nairobi. Corden, W. M., and R. Findlay (1975). "Urban unemployment, Intersectoral Capital Mobility and Development Policy." Economica, 42, 59-78. D'Adda, Giovanna, Markus Goldstein, Joshua Zivin, Mabel Nangami, and Harsha Thirumurthy (2008). "ARV Treatment and Time Allocation to Household Tasks: Evidence from Kenya." African Development Review, Forthcoming. de Laat, Joost, 0.Faye, T. Suri and E. Zulu (2008). "Poverty Dynamics and Mobility in Nairobi's Informal Settlements." APHRC Working Paper. Duflo, Esther, Pascaline Dupas, and Michael Kremer (2007). "Peer Effects, Pupil- Teacher Ratio, and Teacher Incentives: Evidence from a Randomized Evaluation inKenya." Available at http://econ-www.mit.edu/files/2099. Easterly, William, and Ross Levine (1997). "Africa's Growth Tragedy. Policies and Ethnic Divisions." World Bank, Washington DC. Erickson, Lennart and Dietrich Vollrath (2004). "Dimensions of Land Inequality and Economic Development." IMF Working Paper 04/158, International Monetary Fund:Washington DC. Evans, David; Kremer, Michael; and, Ngatia, Miithoni (2008) "The Impact of Distributing School Uniforms on Children's Education in Kenya".RAND Corporation, World Bank, Harvard University Economics Department, and, Yale University Economics Department. March2008 Fearon, James, D. (2003). "Ethnic and Cultural Diversity by Country." Journal of Economic Growth 8 (2): 195-222. 187 Forster, Dayo. "Financial Access in Kenya Results of the 2006 National Survey FSD Kenya." FinAccess. The Steadman Group. October 2007. Fox, Louise and Melissa Sekkel (2006). "Job Creation and the Quality of Growth in Africa" (Work in Progress). PREM - Africa Region. World Bank, Washington DC. Freeman, H.Ade, Frank Ellis, and EddieAllison (2004). "Livelihoods and Rural Poverty Reduction inKenya." Development and Policy Review, 2004,22(2): 147-171. Gallup, John Luke, Jeffrey Sachs, and Andrew Mellinger (1999). "Geography and Economic Growth." In Annual World Bank Conference on Development Economics, eds. Boris Pleskovic andJoseph E. Stiglitz, World Bank: Washington DC. Geda, et a1 (2001). "Determinants of Poverty in Kenya: A Household Level Analysis." Mimeo. Gerdin, Anders (2002) "Productivity and Economic Growth in Kenyan Agriculture, 1964-1996.". Agricultural Economics, Vol. 27, No. 1: 7 - 13. Glewwe, Paul, Nauman Ilias, and Michael Kremer (2004). "Teacher Incentives." Processed. Glewwe, Paul, Michael Kremer, Sylvie Moulin, and R. Nyamuru (1999). "Textbooks andTest Scores, Evidencefrom aRandomizedEvaluation inKenya." Processed. GoK (Government of Kenya). 1986. "Sessional Paper Number 1 of 1986 on Economic Management for RenewedGrowth." Ministry of Finance. -------. (1988a) "Urban Labor Force Survey 1986;' Long Range Planning Division and Central Bureau of Statistics, Ministry o f Planning and National Development, Nairobi. July, 1988 -------. (1988b). "Rural Labor Force Survey 1988. Survey Instrument and Enumerator's Reference Manual." Central Bureau of Statistics, Ministry of Planning and National Development, Nairobi. February, 1988. -------. (1991). "Rural Labor Force Survey 1988/89." Long Range Planning Division and Central Bureau of Statistics, Ministry o f Planning and National Development, Nairobi. April, 1991. --------. (1997) "Second Poverty Report in Kenya." Vol. Iand 11, 1997. Ministry o f Finance and Planning, Nairobi. --------, (2000). "Popular Version of the Second Poverty Report in Kenya." Ministry of Finance and Planning, HRSSD, CBS, Nairobi. 188 -------. (2002). "Analytical Report on population projections - Volume VII." Kenya 1999.Populationand Housing Census.August, 2002. -------. (2003a). "Report of 1998/99Force Survey." Central Bureau of Statistics, Ministry of Planning andNational Development. March, 2003. -------. (2003b). "Economic Recovery Strategy for Wealth and Employment Creation (2003-2007)." Ministry of Planning and National Development, Nairobi. June, 2003. --------. (2003~)."Kenya Road Inventory and Condition Survey Study." Ministry of Roads andPublic Works, Nairobi. --------. (2005a). "National Export Strategy Implementation Action Plan 2005-008." MinistryofTrade andIndustry,Nairobi. ---------. (2005b). "National Health Services Strategic Plan 11, 2005-10." Ministry of Health, Nairobi. ---------. (2005~)."Kenya Integrated Household Budget Survey 2005/06: Questionnaire andInterviewer's Manual." Central Bureau of Statistics, Ministry of Planning and NationalDevelopment, Nairobi. ---------. (2006). "Participatory Poverty Assessment." Various Site and District Reports, the Human Resources and Social Services Department and Central Bureau of Statistics, Ministryof Planning andNationalDevelopment, Nairobi. ---------. (2006a). "Sessional Paper: The Development and Management o f the Roads Sector for Sustainable Economic Growth." Ministry of Roads and Public Works, Nairobi. ---------. (2006b). "Public Financial Management Performance Report and Performance Indicators." (PEFA Evaluation) FinalReport. July 2006, Nairobi. ---------. (2006~)."Investment Program for the Economic Recovery Strategy for Wealth and Employment Creation." Annual Progress Report: 2004/2005. May 2006, Nairobi. ---------. (2007a). "Basic Report on Well-Being in Kenya. Based on Kenya Integrated Household Survey - 2005/06." Ministry of Planning and National Development, Nairobi. ---------.(2007b). "Kenya Integrated Household Budget Survey (KIHBS) 2005/06, Vol. I Basic Report." KenyaNational Bureauof Statistics,Nairobi. ---------. (2007~)."Vision 2030." National Economic and Social Council, Office of the President,Nairobi. _________.(2007d). "National IndustrialPolicy," Draft InterimReport.Ministry of Trade and Industry,Nairobi. 189 ---------. (2008a)."Basic Report of the Fourth Participatory Poverty Assessment 2005/06." Planning Department in the Ministry o f Planning and National Development in collaboration with International Livestock ResearchInstitute(ILRI),Nairobi. ---------. (2008b). "Vision 2030, Medium Term Plan, Strategy for National Transformation 2008-2012. Accelerating Equitable Economic and Social Development for a Prosperous Kenya." Ministry of State for Planning, National Development and Vision 2030, Nairobi. ---------. (various years). "Economic Survey." Central Bureau of StatisticsKenya National Bureau of Statistics. Ministry of Planning and National Development, Nairobi. Halverson and Palmquist (1980). "Interpretation of DummyVariables inSemilogarithmic Equations." American Economic Review, Vol. 70 No. 3: 474 -475. Hart, Keith (1972) "Employment, Income and Inequality: A strategy for Increasing Productive Employment InKenya." International Labor Organization, Geneva. Hausmann, Ricardo, and Bailey Klinger (2006). "Structural Transformation and Patterns of Comparative Advantage inthe Product Space." Working Paper 128, Center for International Development, HarvardUniversity, Cambridge, MA. Hausmann, Ricardo, Jason Hwang, and DaniRodrik (2005). "What You Export Matters." NBER Working Paper 11905, National Bureau of Economic Research, Cambridge, MA. Hausmann, Ricardo, Lant Pritchett, and Dani Rodrik (2004). "Growth Accelerations." Working Paper 10566, John F. Kennedy School o f Government, Harvard University, Cambridge, MA. Hausmann, Ricardo, Dani Rodrik, and AndrCs Velasco (2005). "Growth Diagnostics." John F. Kennedy School of Government, Harvard University, Cambridge, MA. Helu, Samuel (2007). "Maritime Services inKenya." Paper presented at the Ministryof Trade and Industry, GoK, and International Lawyers and Economists against Poverty Trade in Services Workshop, Nairobi, March26-27. Herrera, Santiago, and Gaobo Pang (2005). "Efficiency o f Public Spending in Developing Countries: An Efficiency Frontier Approach." Policy Research Working PaperNo. 3645, World Bank, Washington, DC. IEA - Kenya, Instituteof Economic Affairs and KNCHR, Kenya National Commission of Human Rights (2006). "Kenyans' Verdict: A Citizens Report Card on the Constituencies Development Fund(CDF)." 190 IEA - Kenya (Institute of Economic Affairs) (2006). "Kenyans' Verdict: A Citizens Report Card on the Constituencies Development Fund(CDF)." September 2006. URL IP3 (Institute for Public Private Partnerships). 2007. "Policy, Legal, and Institutional Framework for PPPs in Kenya," Consultant report prepared for the World Bank, Washington DC. IMF(International Monetary Fund) (2007). "Kenya-Staff Report for the 2006 Article IV Consultation, SecondReview under the PRGF Arrangement." Appendix 2 (April). Washington, DC. --------. (2005) "Second Review under the Three-Year PRGF Arrangement." November 11,2005. --------. (2007) "Kenya: Draft Report on the "Observance o f Standards and Codes (ROSC). Fiscal Transparency Module." (July 2007). Jayne, T; Ariga, J; Gautam, M; Kopicki, R; Brown, M (2007). "KENYA Agriculture Policy Review: Current trends and Future Options for Pro-Poor Agriculture Growth." FA0 InvestmentCentre. Draft # 3, Johnson, Simon, Jonathan D. Ostry, and Arvind Subramanian(2007). "The Prospectsfor Sustained Growth in Africa: Benchmarking the Constraints." NBER Working Paper 13120, National Bureauof Economic Research, Cambridge, MA. Johnson, Susan; Nino-Zarazua, Max; Ariti, Con0 (2007). "Financial ExclusioninKenya: An Analysis of Financial Services Use." Nairobi, Kenya, Financial Sector Deepening Trust, Kenya. Kabubo-Mariara, Jane, Domisiano M.Kirii, Godfrey K.Ndenge, John Kirimi and Rachel K.Gesami(2006): "Regional andInstitutionalDeterminantsofPoverty: The Case of Kenya." Final Report Presented for Phase I1of the Collaborative Project on Poverty, Income Distribution and Labour Market Issues in Sub-Saharan Africa, January 2006. Kabubo-Mariara, Jane (2002): "Herders Response to Acute Land Pressure and Determinants of Poverty under Changing Property Rights: Some Insights from Kenya." Ecological and Environmental Economics Working Paper Series, No.4. Kamau, Paul (2004). "An economic analysis of gender diversity and urban poverty in Kenya: A case study of Eldoret Town." Electronic Publications from the University of Nairobi, Kenya. Kelley, Allen C., and Charles E. Nobbe (1990). "Kenya at the Demographic Turning Point? Hypotheses and a Proposed Research Agenda." World Bank Discussion Paper 107, World Bank, Washington DC. 191 Kenya Ports Authority. 2005. "The Determinants of Cargo Dwell Time and the Resulting Impact on the KenyanEconomy." December, 2005. Kenya Human Rights Commission, (2006) "Bi-Annual Human Rights Report Kenya." January-June2006. Volume 8, Number 1, Nairobi. Kijima, Yoko and Peter Lanjouw (2003). "Poverty inIndiaduring the 1990s:A Regional Perspective" Policy ResearchWorking Paper No. 3141, World Bank: Washington DC . Ketkar, Suhas L., and Dilip Ratha (2007). "Development Finance via Diaspora Bonds, Track Record and Potential." Policy Research Working Paper 4311, World Bank, Washington, DC. Kiai, Maina; Lumumba, Odenda (2006). "Unjust Enrichment - The Making of Land Grabbing Millionaires." Living Large Series Vol. 2. No. 1, KNCHR and Kenya LandAlliance, Nairobi. Kijima, Y. and P. Lanjouw (2003). "Poverty in India during the 1990s: A Regional Perspective." Policy ResearchWorking PaperNo. 3141, DECRG, World Bank. Kingdon, G., J. Sandefur and F. Teal (2004). "Patterns of Labour Demand. African Region Employment Issues - Regional Stocktaking Review." Centre for the Study of African Economies, University of Oxford. Kiringai, Jane, James Thurlow, and Bernadette Wanjala (2006). "Beyond Recovery: An Equitable Growth Strategy for Kenya." Processed. Klinger, Bailey, and Daniel Lederman (2004). "Discovery and Development: An Empirical Exploration of New Products." Policy Research Working Paper 3450, World Bank, Washington, DC. Kremer, Michael. (2002) "Randomized Evaluations of Educational Programs in Developing Countries: Some Lessons". Department of Economics, Harvard University. Kumar, P. and Leonard0 Garrido. (October, 2006). "Labour Market Outcomes in the Nineties." World Bank PREM -AFTP2. Note prepared for "Jobs inKenya" GOK -BankESWtask. Lay, Jann, Michuki M`Mukaria and Toman 0. Mahmoud (2007). "Boda-bodas Rule: Non-agricultural Activities and Their Inequality Implications in Western Kenya." GIGA Working Paper No. 48, German Institute for Global and Area Studies, Hamburg. 192 Littlefield, E., J. Morduch, and S. Hashemi, (2003). "1s Microfinance an Effective Strategy to Reach the MillenniumDevelopment Goals?" CGAP Focus Note 24, Consultative Group to Assist the Poor; Washington DC. Lokshin, Michael M., Elena Glinskaya, and Marito Garcia, (2000). "The Effect of Early Childhood Development Programs on Women's Labor Force Participation and Older Children's Schooling inKenya. Policy ResearchWorking Paper No. 2376, World Bank,Washington DC. Lofgren, Hans, and Praveen Kumar (2007). "The Challenges o f Achieving Kenya's Vision 2030: A Macro Perspective." In Perspectives in Poverty Reduction and Human Development: Analysis of Alternative Strategies with an Economy-wide Framework, eds. Carolina Diu-Bonilla and Hans Lofgren. Forthcoming. Macharia, J. (2006). "Determinants of Employment Before and After Structural Adjustment Programmes." Masters Degree Dissertation, University ofNairobi. Maloney, W. and J. Nunez (2004). "Measuring the Impact of MinimumWages: Evidence from Latin America." The World Bank, Poverty Section Unit, Latin America and CaribbeanRegion, World Bank: Washington DC. Manda, Damiano (2004). "Globalisation and the Market in Kenya." Kenya Institute for Public Policy Researchand Analysis. KIPPRA Discussion Paper # 31,Nairobi. Manda, Damiano, German0 Mwabu, and Mwangi Kimenyi (2004). "Human Capital Externalities and Private Returns to Education inKenya." Working Paper 2004- 08, Department of Economics, Universityof Connecticut. Mango, Nelson (2007). Poverty Diagnostics and Dynamics. Draft. Marinda, P. (2006): "Child-mother Nutrition and Health Status in Rural Kenya: The Role of Intra-household Resource Allocation and Education." International Journal of Consumer Studies, Vol. 30. McPeak, John (2005). "Confronting the risk of asset loss: What Role do Livestock Transfers inNorthern Kenya Play?" Journal of Development Economics, Vol. 81, pp. 415-437. Miguel, Edward and Kremer, Michael. (2001) "Worms: Education and Health Externalities in Kenya." National Bureau of Economic Research (Cambridge, MA) Working PaperNo. 8481,September 2001. Morduch, Jonathan. (1995). "Income Smoothing and Consumption Smoothing". Jounal of Economic Perspectives9, 103-114. 1995 Muchiri, Stephen, Thomas Maina, Nzoya Munguti, Adam Lagerstedt and Burkard Koemm (2008). "Development ofNationalHealthFinancing Strategy for Kenya." Health Sector MidTerm Review (Mimeo, June 2008). 193 Mukui, John T. (1994). "KENYA: Poverty Profiles, 1982-1992. "Consultant Report Prepared for the Office of the Vice President and Ministry of Planning and National Development, Nairobi. Mukui, John T. (2005). "Poverty Analysis inKenya: Ten Years On." Study conducted for the Central Bureau of Statistics (CBS), Society for International Development (SID), and SwedishInternational Development Agency (SIDA). Mukula, Raymond (2008). "Horizontal Inequality and Ethno-Regional Politics inKenya." Draft, June 4,2008. Muyanga, Milu, Miltone Ayieko and Mary Bundi (2006). "Transient and Chronic Rural Household Poverty. Evidence from Kenya." Mimeo. Mwega, Francis M. and Njunuga S. Ndung'u (2002). "Explaining African Economic Growth Performance: The Case of Kenya." AERC Working Paper #3, African Economic ResearchConsortium, Nairobi. Narayan, Deepa, and David Nyamwaya (1996). "Learning from the Poor: A Participatory Poverty Assessment inKenya" Environment Department Paper No. 034, World Bank: Washington DC. Ndulu, Benno J. and StephenO'Connell (2000) "Africa's Growth Experience: A Focus on Sources of Growth." AfricanEconomic ResearchConsortium, Mimeo. Ndulu, Benno, J., Lopamudra Chakraborti, Lebohang Lijane, Vijaya Ramachandran, and Jerome Wolgin. (2007). "Challenges of African Growth, Opportunities, Constraints, and Strategic Directions." World Bank, Washington DC. Nehru, Vikram and Ashok Dhareshwar (1995). "A New Database on Physical Capital Stock: Sources, Methodology and Results." World Bank, Washington DC. Nyoro, James, Tavneet Suri and Betty Kibaara (2008). "Income, Poverty and Income Dynamics inKenya." Working paper. Odhiambo, Walter; Nyangito, Hezron (2002) "Land Laws and Land use in Kenya: Implications for Agriculture Development." KIPPRA Discussion Paper No. 15. Kenya Institute for Public Policy Researchand Analysis, Nairobi. Oiro, Mwabu, Manda: "Poverty and Employment inKenya." Mimeo. Okwi, Paul O., et a1 (2006): "Geographic Determinants of Poverty in Rural Kenya: A National and Provincial Analysis." A Report prepared for the Rockefeller Foundation, Nairobi, Kenya. Place, Frank, Michelle Adato and Paul Hebinck (2007) "Understanding Rural Poverty and Investment in Agriculture: An Assessment of Integrated Quantitative and 194 Qualitative ResearchinWestern Kenya." WorldDevelopment, Vol. 35, No. 2, pp. 312-325. Place, Frank, Paul Hebinckand Michelle Adato (2004): "Identifying the Rural Poor and Their Ability to Invest in Agriculture." Presented at Q-Squared in Practice: A Conference on Experiences of Combining Qualitative and Quantitative Methods inPoverty Appraisal, Centrefor International Studies, Universityof Toronto. Pollin, Robert; Mwangi wa Githinji, James Heintz (2007): "An Employment-Targeted Economic Program for Kenya." International Poverty Centre in Brasilia (IPC), UnitedNations Development Program (UNDP) and Political Economy Research Institute(PERI), UniversityofMassachusetts, Amherst, MA. Ratha, Dilip, Sanket Mohapatra, and Sonia Plaza (2008). "Beyond Aid: New Sources and Innovative Mechanisms for Financing Development in Sub-Saharan Africa." Policy ResearchPaper4609, World Bank, Washington DC. Renkow, M., D. G. Hallstrom and D. D. Karanja (2004). "Rural Infrastructure, Transactions Costs and MarketParticipation inKenya." Journal of Development Economics: Vol. 73: 349-367. Rodrik, Dani (2000). "Saving Transitions." World Bank Economic Review 14 ( 3): 481- 517. --------. (2007). "The Real ExchangeRate and Economic Growth: Theory and Evidence." John F. Kennedy School of Government, Harvard University, Cambridge, MA. --------. (2008). "Second Best Institutions." John F. Kennedy School of Government, Harvard University, Cambridge, MA. January, 2008. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi (2002). "Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development." Journal of Economic Growth, Vol. 9(2): 131 -165. Sachs, Jeffrey, and Andrew M. Warner (1997). "Sources of Slow Growth in African Economies." Journal of AJFicanEconomies 6(3): 335-76. Sarah Bayne, Sarah; Michael Muragu, DavidNewton, andBob Thomas (2007) "Strategic Conflict Assessment: KENYA." DFID, Kenya. SID, Society for International Development (2006). "Readings on Inequality in Kenya. SectoralDynamics and Perspectives." Regal Press KenyaLtd: Nairobi. SIDA, Swedish International Development Cooperation Agency (2005). "Anintegrated economic analysis for pro-poor growth in Kenya." SIDA Country Economic Report 2005:8, Department for Policy andMethodology, Stockholm. 195 Soderbom, Mans and Francis Teal (2004). "Size Efficiency in African Manufacturing Firms: Evidence from Firm-level Panel `Data." Journal of Development Economics, Vol. 73(l), 369-394. Sonderbom, Mans; Teal Francis; Anthony Wanbugu (2005). `Unobserved Heterogeneity and the relation between earnings and firm size: evidence from two developing countries." Economic Letters 87, 153-159. 2005 Stewart, Frances (2008). "Kenya, Horizontal Inequalities and the Political Disturbances of 2008." Note for Discussion. Stifel, David and Luc Christiaensen (2006) "Tracking Poverty over Time in the Absence of Comparable Consumption Data" Policy Research Working Paper 3810, World Bank, Washington, DC. Stifel, David and Luc Christiaensen(2007). "Tracking Poverty Over Time inthe Absence of Comparable Consumption Data." World Bank Economic Review, Vol. 21, No. 2,317-341. Swallow, Brent (2005): "Potential for Poverty Reduction Strategies to Address Community Priorities: Case Study of Kenya." World Development, Vol. 33(2): 301-321. Tahari, Amor, Dhaneshwar Ghura, Bernardin Akitoby, and Emmanuel B. Aka (2004). "Sources of Growth in Sub-Saharan Africa," IMF Working Paper 041176, International Monetary Fund, Washington, DC. Thurlow, James; Jane Kiringai and Madhur Gautam. "Rural Investmentsto Accelerate Growth and Poverty Reduction in Kenya." IFPRI Discussion Paper 00723. International FoodPolicy ResearchInstitute.October 2007. TMG, Telecommunications Management Group (2007). "Trade in Information and Communication Services: Opportunities for East and Southern Africa, Kenya Case Study." Consultant report for the World Bank. UNDP, Population Statistics and forecasts, at: UNPopulation InformationNetwork. UN Population Division. Department of Economic and Social Affairs. http://www.un.orglpopid UNDP (2007). KenyaNational HumanDevelopment Report. UNDP Kenya, Nairobi. Wacziarg, Romain T., and Karen H. Welch (2003). "Trade Liberalization and Growth: New Evidence." NBER Working Paper 10152, National Bureau of Economic Research, Cambridge, MA. Weinreb, Alexander A.: "First Politics, Then Culture: Accounting for Ethnic Differences inDemographic Behavior inKenya." Population and Development Review 27(3), 437-467. 196 World Bank (1995). Kenya Poverty Assessment. Report No. 13152-KE, World Bank: Washington DC. --------. (2003a) "Kenya: A Policy Agenda to Restore Growth." Country Economic Memorandum. Poverty Reduction and Economic Management 2, Africa Region. August 18,2003. --------.(2003b), "The Kenyan Strategic Country Gender Assessment." Poverty Reduction and Economic Management and ESSD,African Region. October, 2003. --------. (2004a) "Kenya: Accelerating Reforms to Improve the Commercial Legal Frame and Remove Administrative and Regulatory Barriers to Investment." Foreign InvestmentAdvisory Service (FIAS), World Bank Group, March2,2004 --_--_--. (2004b) "Investment Climate Assessment. Enhancing the Competitiveness of Kenya's Manufacturing Sector: The Role o f the Investment Climate." Africa Private Sector Group. November 2004. --------. (2005a). Kenya. Growth and Competitiveness. Report No. 31387-KE, World Bank: Washington DC. --------. (2006a). Kenya. Inside Informality: Poverty, Jobs, Housing and Services in Nairobi's Slums. World Bank: Washington DC. --------.(2006b). "World Development Indicators," Washington DC. --------. (2006~)."Kenya: An Assessment of Pension Reform Directions" - Discussion Draft-Report. --------. (2006d). "Kenya Education Sector Support Project Appraisal Document: Kenya Education Sector Support Project Development Objectives Indicators." --------.(2006e). "Doing Business," Washington DC. 2006. -------_. (2006g). "Youth in Africa's Labor Market" 2 vols. Draft, June 2006. Washington DC. --------. (2006h). "Kenya Education Sector Support Project Appraisal Document: Kenya Education Sector Support Project Development Objectives Indicators." --------. (2007a). "Kenya: DoingBusiness 2007: How to Reform." --------. (2007b). "An Emerging Picture of Jobs in Kenya in The 1990s: Lessons For Today." World Bank.February, 2007. --------. (2007~)."Kenya: AnAssessment ofPensionsReformDirections." (April draft). --------. (2007d). "Doing Business 2008." (September). 2008. 197 --------. (2007e). "Transport Costs in Africa: A Review of the Main International Corridors inthe Four Sub-regions." Ongoing (Septemberdraft). ----_---. (20070. "Connecting to Compete, Trade Logistics in the Global Economy." (November 2007). --------. (2007g). "Africa Regional Communications Infrastructure Program." www.worldbank.ora/rcip --------. (2007h). "Uganda. Moving Beyond Recovery: Investment and Behaviour Change for Growth." Country Economic Memorandum, Report No. 39221-UG. --------. (2007i). Kenya. Unleashing the Potential for Trade and Growth. Report No. 37688-KE, World Bank. Washington DC. Mimeo. --------. (2008a). "Needs Assessment: Key Issues Emerging from the Review o f Damage andNeeds Assessments." April 21,2008. --------. (2008b). "Kenya Agriculture Policy Review." Ongoing. --------. (2008~)."Kenya Tourism Study." Ongoing. -------- (2008d). "Kenya: Investment Climate Assessment 2007." Regional Program for Enterprise Development (RPED) Africa Finance and Private Sector (AFTFP) May, 2008. World Bankand Government of Kenya (1997). "Kenya Public ExpenditureReview." World Bank and IMF (2007). "Kenya: Joint Staff Advisory Note on the Annual Progress Reports of the Investment Programme for the Economic Recovery Strategy for Wealth and Employment Creation." International Development Association and International Monetary Fund, Washington DC. World Economic Forum (2007). "Global Competitiveness Report 2007-2008." Palgrave Macmillan, Hampshire. 198