Report No. 29743-BUR Burkina Faso Reducing Poverty Through Sustained Equitable Growth Poverty Assessment June 7, 2005 PREM 4 Africa Region Document of the World Bank MFB Ministry o f Finance and Budget MESSRS Ministryof SecondaryandHigherEducation and Scientific Research M O H MinistryofHealth MTEF Medium-Term ExpenditureFramework NGO Non-govemment Organization PAP Priority Action Plan PER Public ExpenditureReview PDDEB Ten-year Basic EducationDevelopment Plan PNDS National HealthCare Development Plan PRGB Budget Management Reform Plan PRGF Poverty Reduction and GrowthFacility PRSC Poverty Reduction Support Credit PRSP Poverty Reduction Strategy Paper PRSP-PR Poverty Reduction Strategy PaperProgressReport ROSC Report onthe Observanceof Standards and Codes SP-PPF Permanent Secretariat for the Supervision of FinancialPolicies and Programs STC-PDES Technical Secretariat for the Coordination of Social and Economic Development Programs TOFE Government Financial Operation Table UNDP UnitedNationsDevelopmentProgram UNICEF UnitedNations InternationalChildren Emergency Fund WAEMU West Ahcan Economic and Monetary Union WHO World Health Organization Country Director: A. DavidCraig Sector Director: Paula Donovan Sector Manager: Robert R. Blake ... 111 TABLE OF CONTENTS I INTRODUCTION . .............................................................................................................................. 1 A. Country Backgroundand Linkwith the PRSP................................................................................. 1 I1.B MEASURINGPOVERTY . Available Data andExistingAnalysis.............................................................................................. 3 OVER TIME ......................................................................................... 4 A. B. Priority Survey Designand Comparability o f Poverty Estimates................................................... 5 ~ C. Constructinga Comparable Welfare Indicator............................................................................... 11 D. Measuring Changes inPoverty ....................................................................................................... 14 Conclusions and Recommendations............................................................................................... 18 I11 GROWTHAND POVERTY 1998-2003 . ......................................................................................... 19 A. GrowthPatterns.............................................................................................................................. B. PovertyDeterminants..................................................................................................................... 19 26 C. D. Poverty Dynamics .......................................................................................................................... 29 E. Was Growth Pro-Poor? .................................................................................................................. Growth, Poverty, and Inequality .................................................................................................... 32 35 F. Conclusions and Recommendations............................................................................................... 38 I V THE QUESTFOR SUSTAINEDEQUITABLE GROWTH . ....................................................... 40 A. ProjectingPovertyDevelopments with PAMS.............................................................................. 40 B. The Baseline Growth Scenario....................................................................................................... 40 D. C. Alternative Growth Paths............................................................................................................... 45 Vulnerability to Macroeconomic Shocks....................................................................................... 49 V .E. Conclusions andRecommendations............................................................................................... 51 A.PROVIDING SOCIAL SERVICESTO THE POOR ................................................................... 53 Education........................................................................................................................................ 53 B. Health............................................................................................................................................. 66 C . Social Protection ............................................................................................................................ 81 D. Conclusions and Recommendations............................................................................................... 90 VI EQUITYOFFISCALOPERATIONS . ........................................................................................... 94 A. Fiscal Developments and Choices.................................................................................................. 94 C. B. Theoretical Considerations............................................................................................................. 96 Empirical Results ........................................................................................................................... 96 D. Conclusions andRecommendations............................................................................................... 98 LISTOFTABLES: PARAMETERS FOR THE PRIORITY SURVEYS .......................................................................... 6 TABLE2.NUMBERPRODUCTSUSEDTO CONSTRUCT THECONSUMPTION AGGREGATE TABLE1.SURVEY DESIGN OF ............................................ 8 TABLE3.COMPOSITIONOF THEFOOD .................................................................... 10 TABLE4.IMPACTOFCHANGES INTHE SURVEY DESIGNPOVERTY LINEIN 1998AND 2003 AND METHODOLOGY ON POVERTY STATISTICS .................... 11 TABLE5.ITEMS RECORDEDDIFFERENTLY1998AND 2003 ..................................................................................... IN 13 TABLE6.COVERAGEOFTHE COMPARABLE CONSUMPTIONAGGREGATE TABLE7 OUAGADOUGOU . .................................................................... 13 ..................................................... 14 TABLE8.POVERTY AND INEQUALITY POVERTYAND INEQUALITY DYNAMICS........................................................... CONSUMERPRICE INFLATORS . BY COMMODITYGROUP . 15 TABLE9.REAL GDP: SHARESAND SECTORALGROWTH .......................................................... 20 TABLE10.SUBSECTORSHARES INSECONDARY SECTORAND INFORMALSECTORVALUE ADDED,1998.................... PATTERN, 1998-2003 23 TABLE11.SUBSECTORSHARES INTERTIARYSECTORAND INFORMALSECTOR VALUE ADDED,1998........................ 24 iv TABLE12.GDP.PRiVATECONSUMPTION.AND HOUSEHOLD 25 TABLE13.COMPARATIVEANALYSISOFPOVERTYDETERMINANTS............................................................................ EXPENDITURE.1998-2003............................................ 27 TABLE14.SOCIO-ECONOMIC GROUPS POVERTY, 1998AND 2003 ....................................................................... AND 31 TABLE15.GROWTH AND INEQUALITY ELASTICITY FORTHE POORAND ULTRA-POOR 33 TABLE16 POVERTY DECOMPOSITION POPULATIONSHIFTEFFECT 34 TABLE17 PRO-POORGROWTH .. WITH .................................................................... ................................................ INDEXAND POVERTY EQUIVALENT GROWTH RATE, 1998-2003 ............................... 38 TABLE18.BASELINE MACROECONOMIC FRAMEWORK POVERTYRESPONSE AND ....................................................... 41 TABLE19.INEQUALITYAND HOUSEHOLD EXPENDITURETRENDS 42 TABLE20 POVERTYELASTICITYDECOMPOSITION, TABLE21 PRO-POORGROWTH .. ............................................................................... 2004-2015 .................................................................................. 43 INDEX D~AMICS ....................................................................................................... 44 TABLE22.PRIMARYEDUCATIONSTATISTICS, 1998-2002........................................................................................... 54 TABLE23 SECONDARY EDUCATION . STATISTICS, 55 TABLE24.EVOLUTION LITERACY, OF 55 TABLE25.LITERACY 1994-2003........................................................................................................ 1998-2001...................................................................................... LEVEL EXPENDITURE BY QUINTILES, 2003 ................................................................................ 56 57 TABLE27.PRIMARYSCHOOL GROSS TABLE26.HEADS HOUSEHOLDSHIGHEST OF LEVEL EDUCATION, ENROLLMENT 57 TABLE28.PRIMARY SCHOOLGROSS RATES,1994-2003...................................................................... OF 2003 ................................................................ ENROLLMENTRATES BY EXPENDITURE QUINTILES, 58 TABLE29.SECONDARYSCHOOL GROSSENROLLMENT 2003................................. ................................................................. 58 TABLE30.SECONDARYSCHOOLGROSSENROLLMENT RATES, 1994-2003 ............................ TABLE31.ACCESSTO SCHOOLBY LOCATION REGION........................................................................................ RATESBY EXPENDITURE QUINTILES, 2003 59 AND 61 TABLE32.REASONS FORNOT ATTENDINGSCHOOL TABLE33 REASONSFORSCHOOLDISSATISFACTION . .................................................................................................... 61 ................................................................................................... 62 TABLE34.LABOR 63 TABLE35.INCIDENCEOFCHILDLABORFORCEA R T ~ c ~ P A TRATENBY LEVEL EDUCATION................................................................. P ~ ~ OF TABLE36 DISTRIBUTIONOF SUBSIDIES FOREDUCATION ............................................................................................ 64 TABLE37 PER-CAPITAANNUALPRIVATESPENDINGONEDUCATION.......................................................................... TABLE38 HEALTH TABLE39 PROXIMITYOF A HEALTH 68 TABLE40 REGIONALCOMPARISONOFHEALTH ..... ...................................................................................................................... 63 66 INFRASTRUCTURE ......................................................................................................................... 67 CENTER, 2003 .................................................................................................... STATUSINDICATORS ........................................................................ 69 TABLE41.DYNAMICS HEALTH OF STATUSINDICATORS .............................................................................................. 70 TABLE42.CHILDMORTALITY ................................ 70 TABLE43.MALNUTRITION RATES FORFIVE-YEARPERIODSPRECEDINGTHE SURVEY, 2003 STATUS OF CHILDREN0-5 YEARS OLD AND PROGRAMPARTICIPATION, 2003 ..................72 TABLE44.PARTICIPATION OFNON-MALNOURISHED ..................................................................................................................................................................................... 73 CHILDREN 0-5 Y E A M OLD INMALNUTRITION PROGRAMS, 2003 TABLE45.KNOWLEDGE OF MODERN CONTRACEPTIVES INSELECTED AFRICAN C O ~ T R I E S ...................................... 73 TABLE 46 INCIDENCEOF ILLNESSDNJURYAND HEALTH . .................................... 75 TABLE47.CHOICEOFPROVIDERS............................................................................................................................... CAREUTILIZATION, 1998AND 2003 76 TABLE48.DELIVERIES TYPEOFASSISTANCE BY AND LOCATION TABLE49 PUBLIC EXPENDITURESFORHEALTH . ............................................................................... 77 .......................................................................................................... 78 TABLE50.PUBLIC EXPENDITURES FORHEALTH .......................................................................................................... 78 TABLE51.DISTRIBUTIONHEALTH OF ..................................................... 79 TABLE52.AVERAGEHOUSEHOLD CAREUSERS MPUBLIC SECTOR, BY LEVEL 80 TABLE53.EFFECTSHEALTH SPENDINGONPOVERTY................................................................................... OUT-OF-POCKETPAYMENTSONHEALTH, 2003..................................................... 81 TABLE54.PROBABILITYOF SEEKINGCAREWHEN SICK, 2003 ................................................................................... OF CARE 82 TABLE55.DISTRIBUTION SHOCKSACCORDING TO FREQUENCY ............................................................................. OF 84 TABLE56.COPINGMECHANISMS ......................................................................... 85 TABLE57 SOCIAL SECURITY CONTRIBUTIONRATES .................................................................................................. . USEDBY SURVEYEDHOUSEHOLDS 87 TABLE58.SOCIAL SECURITY .................................................................................................. TABLE59.CARFOBENEFITS,2000-2003 ................................................................................................................... CONTRIBUTIONUTES 88 88 TABLE60 FISCALREVENUES...................................................................................................................................... . 94 TABLE61.FISCALEXPENDITURESFUNCTIONAL ......................................................................... TABLE62.FISCALEXPENDITURES ECONOMICCLASSIFICATION............................................................................. 96 BY CLASSIFICATION 95 BY TABLE63.WELFAREELASTICITYAND PRICEREFORM ..................................................................................... TABLE64 WELFAREELASTICITY AND WELFARE REFORMINDEX............................................................................... . INDEX 97 98 V L I S T OF FIGURES FIGURE1.COMPARINGPOVERTYWITHOUTPOVERTYLINES.1998-2003.................................................................... 16 FIGURE2.COMPARINGPOVERTYWITHOUT POVERTY LINES, 17 FIGURE3 PRODUCTIONOFMAINCEREALSAND COTTON, 1997-2003........................................................................ FIGURE4 CEREALAND COTTONOUTPUT, AND POPULATIONGROWTH FIGURE5 VALUEOFCEREALAND COTTONOUTPUT, ANDPOPULATIONGROWTH ... 1994-2003.................................................................... 21 ....................................................................... 22 ...................................................... 22 FIGURE6.GDPGROWTH POVERTYRATES, 1998-2003....................................................................................... AND 30 FIGURE7 GROWTH-INCIDENCE . ........................................................................................... 36 FIGURE8.ECONOMIC CURVES, 1998AND 2003 POLICIESAND POVERTYREDUCTION ...................................... 4 j FIGURE9.EXPORT r................................................. QUANTITY INCREASE . ..................................................................................................................... 46 ................................................................. 47 FIGURE11.AGRICULTURAL PRODUCTIONINCREASE................................................................................................... FIGURE10 INCREASE INAGRICULTURALPRODUCTION,POVERTY IMPACT 47 FIGURE12 PUBLIC SECTORWAGEINCREASE FIGURE13.CHANGE INTHECOMPOSITIONOFGDP..................................................................................................... . ............................................................................................................. 48 49 FIGURE 14 ADVERSE . WEATHER CONDITIONS, POVERTYIMPACT ............................................................................... 50 FIGURE15 ADVERSEWEATHERCONDITIONS............................................................................................................. 50 FIGURE16 EXPORT FIGURE17 LITERACYLEVEL GENDER ... PRICE SHOCK ............................................................................................................................... 51 BY ...................................................................................... 56 FIGURE18.EDUCATION COMPLETIONRATESBYLOCATION AND AGEGROU~S ....................................................................................... 59 FIGURE19.EDUCATIONCOMPLETION RATES INURBAN AREAS .................................................................................. 59 FIGURE20 EDUCATIONCOMPLETIONRATESINRURALAREAS . .................................................................................. 60 FIGURE21 EQUITY . .......................................................................................... 65 FIGURE22.CNSS: CONTRIBUTIONSANDBENEFITS, ........................................................................................... OF PUBLIC SPENDINGONEDUCATION 2002 87 FIGURE23 CNSSBUDGET . EXECUTION,2002............................................................................................................. 88 L I S T OF BOXES BOX:1:IMPROVEMENTS INTHEDESIGN PRIORITYSURVEYS OF .................................................................................... 4 BOX 2: REVISINGPOVERTY AGGREGATES: LESSONS OTHERCOUNTRIES............................................................ FROM j BOX 3: IMPACTOF DIFFERENT RECALL PERIODSONTHE LEVEL REPORTED OF CONSUMPTION ..................................... 8 BOX 4: HOW IS THE OFFICIAL POVERTYLINECOMPUTED? ......................................................................................... 10 Box5: STATISTICALSIGNIFICANCEOFOBSERVEDPOVERTYCHANGES...................................................................... 15 BOX 6:.THE GROWTH INCIDENCECURVE OR GIC ........................................................................................................ 35 BOX 7: PRO-POORGROWTH INDEXAND POVERTY EQUIVALENT GROWTH RATE ........................................................ 37 BOX 8: GOVERNMENT POVERTY OBJECTIVES AND THE MDG ..................................................................................... 44 BOX 9: MANAGEMENT OF THE EDUCATIONSYSTEM INBURKINA 54 BOX 10:EDUCATIONINBURKINA FASO-2025 ............................................................................................................. FASO...................................................................... 62 BOX 11: PER-UNIT COSTOF PUBLIC SPENDINGON EDUCATION ................................................................................... 65 BOX 12: CROP DIVERSIFICATIONAS A RISKREDUCING STRATEGY FORFARMERS ...................................................... 85 SOURCE: SIAENS AND WODON (2004).BOX 13: THEEFFECT THE COTE D'IVOIRE OF CRISISON BURKINABE HOUSEHOLDS ............................................................................................................................................................... 86 Box 14:BURKINAFASOSOCIAL PROTECTION STRATEGY........................................................................................... 89 This document has a restricteddistribution andmay be usedbyrecipients only inthe performance oftheir official duties.Its contents may not otherwise be disclosedwithout World Bankauthorization. Vi ACKNOWLEDGMENTS This report was prepared by a Bank team led by Jan Walliser (AFTP4). The core Bank team included Rodica Cnobloch (consultant), Boniface Essama-Nssah(PRMPR), Issouf SamakC (consultant), Corinne Siaens (consultant), Moukim Temourov (AFTH2), EmilTesliuc (HDNSP), and Quentin Wodon (AFTPM). Christophe Rockmore (consultant) provided advice and support on data issues. Team support was providedbyJudite Fernandes(AFTP4). The report has benefited from the overall guidance of Robert Blake, Sector Manager (AFTP4), A. David Craig, Country Director (AFC15), and Jean Mazurelle, former Country Manager (AFMBF), as well as from feedback provided by peer reviewers Aline Coudouel (PRMPR) andArvil Van Adams (AFTHD) andthe extendedcountry team. The team benefited enormously from close cooperation with staff at the Ministry of Economy and Development (MEDEV) and the National Institute for Statistics and Demography (INSD), as well as the collaboration with the German Gesellschaft fiir Technische Zusammenarbeit (GTZ). We wish to thank inparticular His Excellency Seydou Bouda, Minister o f Economy and Development, who took a close interest in linking the macroeconomic projections with poverty simulations through the poverty analysis macroeconomic simulator (PAMS); Daniel Bambara, Director General for the Economy and Planning (DGEP), who supported and facilitated the work on PAMS in his department; Antoine-Marie Sic TioyC, Director of the Planning and Macroeconomic Analysis Directorate (DPAM), who contributed actively with his collaborators to linking PAMS with the forecasting tool IAP; Bamory Ouattara, Director General o f INSD, who sharedthe 2003 household data andclosely collaborated with the team; and Michel KonC, Director of Statistics at INSD and expert on priority surveys, whose collaboration in interpreting and analyzing the household data was invaluable for advancing this study. We also thank Rolf Meier and Bakary KindC o f the GTZ project "Advisor to the Ministries in charge o f Economy and Finance'' for their close collaboration, active contribution to improving PAMS, including financial support, and advice. We would like to acknowledge the supplementary funding from the World Bank's central department for poverty reduction and economic management(PREM) to advancepoverty and social impact analysis. vii EXECUTIVESUMMARY Linking growth and poverty is a crucial element for evaluating the effectiveness o f government policies under the PRSP process. Burkina Faso has benefited from more than 3 percent growth in per-capita incomes since the devaluation in 1994. The steady increase in incomes, albeit from a very low level, should over time have lifted some Burkinabk above the poverty line and led to a reduction in poverty rates. To gauge how the rising incomes have affected the population, the government has conducted household surveys from time to time (1994, 1998, and 2003) to collect information o f household consumption and incomes. This study uses the household data from 1998 and 2003 data to (i) consider the measurement o f poverty over time; (ii)study the links between growth and poverty in 1998-2003 and under possible future growth paths; (iii) examine the relationship between poverty and social services; and (iv) illustrate equity considerations inthe execution o f fiscal policy choices. Indoing so, it builds on the existing analysis by the government, donors, andindependent researchers. Measuring Poverty Measuring changes in poverty is not an exact science and subject to data quality issues. Over time, the National Institute o f Statistics and Demography (INSD) has improved the data collection process, but has also made changes to survey designs that have reduced the comparability o f different surveys. Care needs to be taken to construct a comparable measure for householdwelfare to best assess changes inpoverty. A comparable welfare indicator can be constructed for 1998 and2003 since the effects of changes in survey design can be clearly identified and largely eliminated. The resulting consumption measure for households contains equally measured items from the 1998 and 2003 surveys by excluding a limited number o f items measured differently in both surveys. A consistent set o f monthly price data is applied to household consumption items to eliminate the effect of prices changes over time and space. However, a similar methodology cannot be applied to 1994 survey data because o f changes in the recall period in the survey design, the effect o f which cannot be eliminated. Using a comparable poverty measure, we find that the poverty headcount declined by about 8 percentage points between 1998 and 2003. Adopting the previously announced 46.4 percent headcount for 2003 and the related poverty line as baseline, the comparable poverty headcount in 1998 was 54.6 percent. The poverty decline was stronger in rural than in urban areas, and inequality remained largely unchanged on the national level between 1998 and 2003. The conclusion that poverty declined between 1998 and 2003 is robust to changes in the poverty line. Using a household income measure, rather than consumption also allows drawing the conclusion that poverty declined during 1994-98. Growth, Poverty, and Inequality Growth during 1998-2003 was driven by a large expansion o f the primary sector, following the 1997-98 drought. Increases in volume and value of cereal production exceeded population growth, and by 2001 the volume and value cotton production bypassedlevels attained viii in 1997 after recovering from the crisis caused by the "white fly" infestation in 1998. Growth in the primary sector was the result o f increased surfaces, with little change in productivity. The secondary sector expanded at about the same rate as overall GDP, with the informal sector playing a large role in the expansion o f manufacturing (food products, textiles, minerals, and metal works), indicating that growth estimates may be fairly imprecise. In the tertiary sector growth was sluggish owing to government restraint and sluggish growth in trade. Again, informal activities were the major dynamic force. National account data and household data for 1998-2003 line up reasonably well. The corrected household consumption data implies annual average per capita expenditure growth o f about 2.3 percent per year. From national account data, we arrive at changes in real per capita private consumption of about 3 percent per year on average. Given the variety o f errors possible inthe way these data havebeen established, the matchisreasonably close. As regards correlates of poverty, results are similar for the 2003 survey as those found in previous studies. Larger household size, lower education levels, occupation in agriculture, and remoteness tend to be correlated with lower per-capita consumption levels. By contrast, families o f landowners and those receiving transfers from relatives tend to have higher expenditure per capita. The decline in national poverty rates between 1998 and 2003 is largely a result o f the growth in agricultural output, both in subsistence farming and cotton farming. The respective socio-economic groups producing agricultural non-tradables (60 percent o f the population in 2003) and agricultural tradables (18 percent o f the population in2003) registered a sharp decline inpoverty rates of 7.5 and 6 percentage points, respectively. However, inequality increased in rural areas, reducing some o f the potential growth impact on poverty rates. Between 1998 and 2003, the share of subsistence farmers inthe population has declined, whereas the share o f cotton farmers has been on the rise after the confidence crisis related to the "white fly" infestation had been overcome. As a result o f this shift o f the rural population, there were less subsistence farmers among the poor in2003 than in 1998 but more cotton farmers. The urban sector also contributed to declining poverty by absorbing an additional share o f the population without experiencing rising poverty rates. Overall, urban poverty rates declined little. However, this fact masks that the urban population was on the rise and thus urban areas were able to absorb rural population without a dramatic increase in poverty, in part through a decline in urban inequality. Inparticular, the share o f socio-economic groups inthe private non- tradable sector and the informal sector and the unemployed has been rising considerably. The first two groups benefited sufficiently from the growth and declining inequality in the secondary andtertiary sector to absorb those moving into these groups and still see either a sharp decline in poverty rates (private non-tradable) or stable poverty rates (informal sector). However, it is noteworthy that the ranks o f the unemployed in the population and among the poor increased as well, raising some warning flags as regards the employment impact o f urbanization. As regards the trade-off between growth and inequality, technical analysis shows that growth has a significant impact on poverty and that the growth-inequality trade-off is fairly small. A 1 percent growth results in a 2 percent decline inthe poverty gap, the average distance i x between the poor and the poverty line. A 1 percent increase in inequality, measured by the Gini index, would need to be compensated by about 1.4-1.5 percent more growth inBurkina Faso, a number that is small compared to several other countries that have previously been studied. In rural areas 0.8-1.1 percent more growth would suffice to compensate for a 1percent increase in inequality, but inurbanareas it would require 3.3-3.4 percent more growth. Poorer and richer households saw their per-capita expenditure rise during 1998-2003, and thus growth was pro-poor. Using the growth-incidence curve approach, it can be shown that in rural areas inequality increased because expenditure o f richer households grew at a higher pace than those o fpoorer households. The opposite i s true inurban areas, explaining why nation-wide inequality remained unchanged. The faster rise in expenditure among richer rural households likely i s related to rising incomes from cotton farming, which went to relative better o f f households first. Using a variety o f additional technical tools-taking into consideration the responses of poverty measures to growth and inequality changes-confirms the conclusion that growth i s pro-poor, although only weakly so in urban areas since growth there can be quickly overwhelmedby changes ininequality. TranslatingSustainedGrowthinto PovertyReduction The government has set itself the objective to reduce the poverty headcount to 30 percent by 2015. Under a baseline growth scenario building on long-term real GDP growth of about 5 percent, driven predominantly by secondary and tertiary sectors, simulations with the poverty analysis macroeconomic simulator (PAMS) show that the target could be achieved and poverty would decline to about 29 percent by 2015.' However, achieving sustained growth o f five percent under the baseline scenario would be predicated by increases in cultivated land use or agricultural productivity, and significantly higher private investment than in the past to achieve the growth in the secondary and tertiary sectors. Moreover, the baseline scenario assumes smooth growth rates without shocks from climatic conditions or terms o f trade changes. Under the baseline scenario, the inequality would rise over time and the growth- inequality trade-off would become more important. Given that under the baseline scenario growth insecondary and tertiary sectors is more important while urbanization i s limited, more o f the additional income flows to the smaller share of urbanhouseholds, thereby raising inequality. In 2015, 2.5 percent growth would be needed to compensate for a 1 percent increase in inequality. The quality o f growth would not change, however, and remain at least moderately pro-poor. Variations to the baseline scenario show the importance o f cotton and agricultural production for accelerating poverty reduction and limit the changes in inequality. A 20 percent increase inthe level o f cotton production in2004 would give rise to a 4 percentage point decline in the poverty headcount accompanied by rising inequality since the cotton producer are a relatively small group o f the population. A 20 percent increase in the level o f non-tradable agricultural production would reduce the poverty headcount by about 9 percentage points by 2015 and reduce inequality as a larger population group shares into the additional income ' Basedona consistentpovertyheadcountmeasure for 1994, the government's target would ldcely also be consistent with achieving the income-poverty MDG. X created. Increasingpublic sector wages by contrast, would generate a barely measurable poverty impact since the group of civil servants is small and their poverty headcount low. Finally, a growth scenario with the same overall growth rates but a different composition-a larger contribution from the primary sector-reaffirms the previous findings, namely that participation of the rural population in growth is crucial for accelerating poverty reduction and reducing inequality. External shocks can have important repercussions for poverty targets. A reduction in agricultural output levels by 10 percent (with a subsequent continuation on the previous growth path, that i s no recovery o f output), would over time lead to an increase in the poverty rate by 3 percentage points. Similarly, shocks to cotton prices would have a direct impact on poverty-a 20 percent increase would lower poverty by about 1.5 percentage points. Poverty and Social Services Education. The level o f schooling, educational attainment, and literacy in Burkina Faso remains low. According to the latest household survey, just 30 percent o f male adults and 13 percent o f female adults know how to read and write, with significantly lower literacy rates in rural areas. Overall literacy rates rose from 18 to 21 percent. Despite some important progress made over the past 5 years, enrollment and completion rates are still low by international standards. Primary school enrollment have progressed significantly from 35 to 44 percent according to the survey, and secondary school enrollment rates stand at only 16 percent up from 13 percent in 1998. Schooling i s closely tied to parents' wellbeing, with kids from the richest quintile being twice as likely to go to primary school and 5 times more likely to go to secondary school. In terms o f quality o f education, low completion rates-only 30 percent for primary school-remain an important concern and a significant factor inhighunitcosts o feducation. Access to schooling and attendance remains a problem. On average, 45 percent o f children attending primary school and 75 percent attending secondary school must walk for more than 30 minutes to reach the nearest school. 22 percent o f initially registered students were not attending school, most importantly because o f expulsion (48 percent) and high cost (27 percent). Almost one-quarter o f household heads reports dissatisfaction with school conditions, mainly because o f lack o f equipment (73 percent), teacher shortage (15 percent) and quality o f teaching (16 percent). Burkina Faso has a high incidence o f child labor, 44 percent o f children age 5-14 are involved in some type o f economic activity, and children from poor households are twice more likely to work than children from better o f fhouseholds. A benefit-incidence analysis shows that overall the education system is tilted toward higher-income households. Higher income households tend to benefit more from govemment spending on education because they are more likely to have access to schooling and more likely to send their children to school. The discrepancies rise with the level o f education since the gap in attendance becomes more important and the unit costs rise sharply for higher levels o f xi education. As a result of the concentration of private schools inurban areas, private spending on education services is almost 10 times higher inurbanthan inrural areas and heavily concentrated among higher-income households. Health. Health infrastructure has increased during 1998-2003, but accessibility o f health centers may still be an issue for rural population. 35 percent o f the population still have to walk more than an hour by foot to reach the nearest health center. Access is mostly an issue in rural areas as in urban areas almost all households are within 1 hour walking distance o f the nearest health facility. Progress has been made in some key health indicators, reversing the trend o f the mid- 1990s. Most importantly infantmortality declined from 104 to 83 per 1000 live births and is now below the Sub-Saharan average. Under-5 mortality also declined, mostly as a result of lower infant mortality. There are important differences by location and economic status in mortality rates. Among the likely reasons for continued high child mortality is high morbidity as regards malaria among children (64 percent o f those hospitalized for malaria are children between 0 and 5), and prevalence o f diarrhea and repository infections. Also, antenatal care and assisted birth remain much less frequent in rural than in urban areas, and are closely linked with income. However, compared with 1998 the use o f assisted deliveries seems to have risen, with a decline in differences in income groups. Availability of qualified health personnel still appears to be constraint to higher levels o f assisted birth. Child malnutrition, which also is closely linked to mortality, remains widespread. 40 percent o f children aged 0-5 are stunted, almost 17 percent are emaciated, and 33 percent are undenveighted. More dramatically, 44 percent o f stunted children are severely stunted indicating long spells o fpoor nutrition. Prevalence o f malnutrition is higher inrural areas and lower among top expenditure quintiles. 40 percent o f stunted children and 16 percent o f emaciated children have ever participated in a nutrition program. However, participation is twice as high in urban than inrural areas. Burkina Faso places in the upper range of African countries as regards knowledge and use o f modern contraceptives. Knowledge rates reached over 90 percent in 2003, and attained 100 percent in urban areas. Current usage rates among married women, however, are low at about 9 percent, and only 20 percent o f married women have ever used contraceptives. Self reported morbidity, which is a very imperfect indicator, has declined in past years and health care utilization increased. Self-reported morbidity (illness in the past 15 days) in the household survey declined from 7.1 percent to 5.8 percent, with a much larger drop in urban morbidity than in rural morbidity. Among those who were sick, 64 reported using health services, compared with only 42 percent in 1998. Differences in the use o f health care across expenditure quintiles also declined. However, increased use o f care did not translate into higher use o f health centers or secondary hospitals, which were in fact used even less than before. Increases in care usage were generated by use o f traditional healers (rural areas) and private health care (urban areas), as well as some increase in the use o f regional or national hospitals. Among those not seeking care the main reason put forward was the highcost (5 1percent). xii Public health spending i s concentrated on primary care, which accounted for about 60 percent o f total public spending in2002. However, since utilization rates are higher inthe upper expenditure quintiles, the incidence even o f primary spending falls more heavily on higher- income households. Out-of-pocket expenditure represents more than 50 percent o f total health financing, and the average BurkinabC spends CFAF 2,900 per month on health services, mostly on drugs (75 percent). However, the out-of-pocket spending o f the lowest income groups has fallen over time, reflecting the decline in the cost o f care and drugs instituted in 2001-03. Overall, one can show that health expenditures do not have an impoverishing effect on households. Health care demand i s influenced by household size, income, and schooling. Higher- income households are more likely to seek care, as are individuals from larger households, and those having secondary education. Older people, households with a female head, low educational status, anddistance to a healthcenter tend to reduce usage o f care. Social protection. Households in Burkina Faso are exposed to a variety o f risks from economic conditions, regional instability, and climatic conditions. Technical analysis o f household data shows that, overall, the vulnerability o f households to fall into poverty has declined between 1998 and 2003. Households use two main risk management arrangements, informal/traditional arrangements and formal arrangements, to which the government has added risk management programs. Traditional arrangements include coping mechanisms (sale o f assets etc.) and are often not sustainable over the long term. In addition, they reduce the household's capability to face further hardship. Surveys indicate that for many households sale o f livestock and family assistance are the most common coping mechanisms. The crisis in C6te d'Ivoire has put additional strain on some households by removing the cash transfers that were an additional insurance against adverse shocks. In terms o f risk management, there would be significant gains from broader diversification o f crops to reduce risks and the exposure. Formal arrangements are limited to funds covering less than 2 percent o f the population (CARFO, CNSS). They pay pensions, family allowances, and professional risk allowances to civil servants or workers in the formal sector. In addition, there is a wide range o f public program addressing socioeconomic risks, such as health, education, food security, labor market, andsocial assistance. Equityof FiscalOperations Fiscal policy intervenes on both the revenue as well as the expenditure side. A global consideration of government operations can therefore help to review the broad equity considerations involved incertain policy choices. Over the past years, the government has restructured both the revenue as well as the spending side o f the budget. On the revenue side, the introduction o f the common external tariff reduced revenues from duties and shifted a larger burden o f the tax collection to the value-added tax and specific excises. On the expenditure side, there has been a tendency to increase social X l l l ... sector spending and shift more spendingto materials and services. This change has been enabled by the move of donors from foreign-financed investment to budget support, allowing the government to somewhat readjust the balance between recurrent and investment spending. Choosing some aversion parameter for inequality, it is possible to rank reforms of certain household incomes andprices by their equity impact. Onthe price side, it turns out that the most sensitive change would be prices o f subsistence agriculture output, followed by food items. On the income side, equity measures are most sensitive to changes in agricultural income and transfers to rural areas. Based on these findings, the govenunent's policy initiatives in the 1990s could be considered as generally mildly equitable. The move away from import duties to the VAT that exempts unprocessed food likely somewhat reduced costs o f subsistence farming and thus the "price" of this output. The choice o f petroleum excise increases, taxing a non-food item, would also appear to be a relatively more equitable choice than other alternatives. Similarly, the taxation of urbanand formal incomes over agricultural incomes i s inline with the index ranking for changes to householdincomes. Key Recommendations Data Anticipate the impact o f survey design changes on the comparability o f data and enhance the questionnaire to be able to estimate the potential bias. Improve the collection o f price statistics, both within and outside the priority survey, by collecting information on both quantities consumed and the value o f the respective purchases, or by administering price questionnaires at the community level. a Implement the survey quarterly(quarterly panel), to eliminate the influence o f seasonality and short-lived shocks on reportedconsumption. Provide adequate and timely financing for the survey, for an adequate monitoring o f poverty and to ensure the comparability o f the welfare indicators from micro- and macro- data sources. a Improve the timeliness and production o fnational account data to allow following growth trends andthe compositiono fGDPon a timelier basis with fewer revisions. Design surveys to follow informal sector activity to better measure economic growth in urban areas. Fully adopt the international definitions in computing indicators for the health sector, improving the data collection for the DHS, and train health personnel in the use o f classificationand registration. Growth. Build on the strategic vision for broad-based growth o f the PRSP to improve the effectiveness and focus o f government actions that could draw subsistence farmers into market-based and export activities and broaden the poverty-reducing impact o f cotton xiv production through a further improvement in its performance and reach. Pay particular attention to policies that could raise agricultural productivity. e Target growth-supporting policies inthe urbanareas to activities benefiting people inthe informal sector and the unemployed to avoid sharp increases in inequality during the growth process and avoid further increases inunemployment. 0 Undertake a systematic review under the PRSP process o f programs and activities as regards their link with the government's growth andpoverty-reduction objective. e Using PAMS and other tools, systematically incorporate a review o f the poverty and inequality impact o f growth-supporting policies for rural and urban sectors into the PRSP and policy design to recognize how government actions may support economic growth that is equitable. 0 Deepen the study o f exogenous shocks to explicitly identify risks for the poverty reduction strategy and identify possible government policy responses. Social Services. Implement the school construction program without further delay and increase the capacity o f existing schools through provision o f multi-grade and double-shift schools. Further proliferation o f existing public measures, such as free distribution o f textbooks and class materials (especially for girls) combined with careful tracking o f expenditure could lessen the private cost o f education and help increase demand for education. Finalize the development o f a sector-wide MTEF to reflect both resource needs for all levels o f education as well as the equity o f government spending. Enhance the availability o f trained health personnel and proper equipment inneedy areas and introduce a system o f incentives for personnel indifficult areas Further seek to reduce costs o f preventive care and drugs, improve health center management and more systematically study the causes for the important decline inhealth center usage between 1998 and 2003. Improve and develop alternative health financing mechanisms, through community-based prepayment and health insurance schemes. The national programs addressing vaccination, malaria, nutrition, HIV/AIDS should be strengthened and supported by more and stable sources o f financing, including the national budget. Consider strategies to stimulating the demand side o f health care. Prepare a national social protection strategy to address the severest risks faced by the poorest households inthe country. xv I.INTRODUCTION 1. Understanding the channels through which growth affects household income and poverty is o f great importance for developing economies. Since the adoption o f the Poverty Reduction Strategy Papers (PRSP) as a basis for concessional lending to low-income countries, there has been a renewed interest inpoverty and social impact analysis (PSIA) to better target government policies and support growth processes that are pro-poor. By adding to the knowledge on poverty and growth processes in Burkina Faso, this poverty assessment (PA) is a core element o f the Bank's analytical support for PRSP implementation in Burkina Faso and a critical analytical underpinning for the Bank's financial support through poverty reduction support credits (PRSCs). 2. In order to support the monitoring and evaluation of government policies through the PRSP process, the PA seeks to enhance the knowledge on (i) poverty measurement; (ii) poverty trends and their linkages with growth; (iii) poverty-focus o f spending in education, health and social protection; and (iv) the equity o f fiscal activity. Based on this analytical foundation, the P A will explore options to enhance the poverty impact o f growth, respond to macroeconomic shocks, and improve the poverty focus o f spending on health, education, and social protection. A. COUNTRYBACKGROUND LINKWITHTHE PRSP AND 3. Burkina Faso i s a landlocked country o f about 12 million inhabitants with a narrow natural resource base. Low per capita income o f less than US$250 and low educational attainment with a literacy rate barely over 20 percent in 2003 characterize Burkina Faso's state o f development. According to the 2003 household survey, about 46 percent o f the population were below the national poverty line, and most of the poor were inrural areas where 80 percent of the population reside. Agricultural output is subject to frequently changing weather conditions and cotton, the major cash crop, has suffered from wide price swings on international markets. The economic structure and unfavorable initial conditions pose particular challenges for Burkina Faso's authorities to design policies that foster growth, channel additional income to those below the poverty line, and limit exposure o f the lowest income segments to shocks. The largely rural population and low population density also hamper rapid improvements in social services and increases the need for a careful planning and targeting o f social expenditure to reach the most deprivedareas. 4. Overall, Burkina Faso has made considerable progress in macroeconomic stabilization under three successive programs supported under the IMF's Poverty Reduction and Growth Facility (PRGF). The growth rate o f gross domestic product (GDP) has averaged 5.8 percent between 1994 and 2003, against a population growth rate o f about 2.5 percent. Fiscal consolidation after the CFA devaluation in early 1994 resulted in an increasingly stable macroeconomic environment with low inflation in the context o f the peg o f the CFA to the French franc and the euro. The macroeconomic stabilization notwithstanding, the country has remained highly vulnerable to external shocks, including i)rainfall conditions in the Sahel zone; ii)internationalcottonpricedevelopments; iii)politicalinstabilityinneighboringcountries; and iv) lack o fpredictability of donor resources. These shocks affect the poor directly through lower incomes and indirectly through difficulties in executing government investment program and supplying social services. 5. Burkina Faso was among the first countries to present a full PRSP. The PRSP is centered on four pillars: (i)accelerating equitable growth; (ii)ensure that the poor have access to basic social services; (iii)expand opportunities for employment and income-generating activities for the poor; and (iv) promote good govemance. Several long-term sectoral programs, such as the 10-year basic education development plan (PDDEB) and the national health development plan (PNDS) underpin the PRSP objectives. The government is currently in the final stages o f revising the 2000 PRSP. The revised PRSP has the same four pillars as the original PRSP but expands the number o f priority sectors to cover security, environment, HIV/AIDS, capacity building, and small and medium enterprises and industries, including small-scale mining. The PRSP revision based on a participatory process also includes the development o f regional poverty reduction strategies. 6. Progress in implementing the PRSP since 2000 has been satisfactory overall, as laid out in three PRSP progress reports to date. Under the second PRSP pillar, school enrollment rates have increased significantly inrecent years, and the rising trend ingrade repetition rates has been reversed recently. Child vaccination rates and health center visits have improved moderately, and infant mortality rates declined from levels exceeding the average o f sub-Saharan Africa to levels comparable to the rest o f the continent. Modest gains were made in child and matemal mortality rates. However, despite these gains, important challenges remain in putting the strategies on a long-term footing and improve the poverty-focus o f public spending. 7. Incontrast to the medium-term development plansinthe social sectors, the policies under the first and third pillars have received less attention during the monitoring and evaluation process since progress is measured with greater difficulty and over longer time spans. The PRSP aims at lifting people out o f poverty through growth with equity by enhancing competitiveness, creating a better business environment, and by financing selected interventions inthe rural areas to enhance income-generating activities for farmers, women, and vulnerable groups. Although growth rates between 1998-03 reached 5.3 percent on average, the period was also characterized by a drought in 2000 and the onset o f the Ivorian crisis in 2002. Analysis on how growth was linked with poverty reduction i s still incomplete, and few lessons have been drawn from the recent growth experience as to whether the government needs to develop further policy initiatives to improve the poverty impact o f growth. 8. The poverty assessment supplies additional analysis on poverty to support the Burkinabk authorities' revolving monitoring and evaluation process o fthe PRSP inareas where analysis has been incomplete. To this end, it discusses the issues o f poverty measurement; macroeconomic links with poverty reduction in 1998-2003 and beyond; poverty-focus o f social services and social protection; and equity o f fiscal activities. - 2 - B. AVAILABLEDATAAND EXISTING ANALYSIS 9. The poverty assessment relies heavily on data collected for household surveys in 1998 and 2003. The so-called priority surveys (enqugtesprioritaires or EP) were conducted over a period o f 4 months in 1994 (EP I), 1998 (EP 11) and 2003 (EP 111) and collected information on household characteristics, household consumption and components o f household production and revenue. For comparability reasons, outlined inmore detail below, the current study focuses on the 1998 and 2003 data. Inaddition to priority surveys, the poverty assessment relies on national account data producedby INSD and incorporated into the authorities' macroeconomic projection tool instrument automatisk de prevision (LAP). National account data are final through 1997, semi-final in 1998 and 1999, and preliminary in 2000-2003. Other data for the poverty assessment come from the Demographic and Health Surveys (DHS) conducted in 1998199 and 2003, results of surveys on poverty perception conducted in 1998 and 2003, and smaller surveys on the cost o f health and education services undertaken annually by INSDinthe context o f PRSP monitoring. 10. The poverty assessment adds to work that has already been undertaken in different contexts by the authorities, the World Bank, and external researchers. The authorities issued in August 2003 a preliminary descriptive analysis o f the 2003 priority survey, accompanied by a statistical background study.* These two papers carefully describe the poverty profile and poverty determinants. To the extent that this study does not differ-notably as regards the description o f household expenditure and characteristics, access to social services, as well as the analysis o f microeconomic poverty determinants-it does not cover these areas indetail. Inother aspects, most importantly the issue o f comparability o f poverty measures and analysis o f poverty-growth dynamics, this study adds new information and thus presents additional analysis and new conclusions. The poverty assessment furthermore complements the preliminary report on the 2003 DHSa3 11. Other analytical work closely related and complementary to the poverty assessment include the recent public expenditure review (PER), the risk and vulnerability assessment (RVA), and a case study on Burkina Faso as part of broader pro-poor work undertaken by a group o f donors, including the World B a d e 4The recent PER undertaken for Burkina Faso investigates budget allocations in 1998-2002, discusses improvements and remaining weaknesses in budget processes, most importantly program budgeting, and reviews the effectiveness o f spending on health and education. The poverty assessment complements the PER with a discussion o f the links between basic social services, social protection, and poverty as it arises from the 1998 and 2003 surveys, and adds a discussion on the equity o f fiscal policies. In its discussion of social protection issues, the poverty assessment also adds additional information to the RVA. As regards the pro-poor growth case study, it complements this poverty assessment through a review of different methods, their inconsistency, and their impact on measuring poverty; a discussion of the 1994 household survey and the devaluation; and a more detailed coverage o f urbanlabor market issues. INSD(2003), andLachaud(2003). INSD(2004). WorldBank(2004a), WorldBank (2004b), andG r i mandGiinter (2004). - 5 - 11. MEASURINGPOVERTYOVERTIME' 12. The main sources for measuring the level of poverty in Burkina Faso are the three priority surveys conducted in 1994, 1998, and 2003. Each one of these surveys delivers a snapshot on the economic and financial situation o f about 8,500 households, with data being gathered during4 months of the respective survey year. The original objective o f the surveys has been relatively modest, namely to present a picture o f monetary and non-monetary poverty at a moment in time, using a shorter and less expensive questionnaire than, for example, the often applied LivingStandardMeasurementSurveys (LSMS). 13. Over time, INSD devoted substantial and continuous efforts to improve the design and implementation o f priority surveys. Survey improvements encompass all dimensions of survey implementation, from sampling and questionnaire design, to data entry and quality control (see Box 1). As a result of these improvements, INSD enhanced the precision o f the latest poverty and inequality estimates, generating increasingly precise and nuanced information on living standards. However, the changes in design also reduced the comparability o f the poverty estimates. Despite this lack o f comparability, the simple fact that different priority surveys used a number o f common features and were publicly available invitedmany researchers to use them for comparisons of monetary poverty, and to investigate the factors associated with changes in poverty over time. Box: 1: Improvementsin the Designof PrioritySurveys INSD improved the design and implementation of the priority surveys over time. These improvements range from sampling and questionnaire design, to data entry and quality control: By adapting the sampling frame to the new administrative organization o f the country, the regional representation o f the survey went up from 8 strata in 1994, to 10 regions in 1998 and 13 regions in2003. The collection of regional price statistics was improved, to generate more detailed price deflators needed to correct for the large differences inpurchasingpower across regions and areas o f residence inBurkina Faso. 0 The design o f questionnaire was improved by increasing the detail with which consumption and income components were recorded, or by adjusting the recallperiod for the frequency o f certain purchases. The quality o f the data entry process was considerably improved in 2003, with the adoption of the CWIQ technology basedon pre-coded questionnaires readby scanners, substantially reducing data entry errors. 14. In order to lay the foundation for the analysis of the poverty dynamics, this chapter reviews the comparability o f survey features and constructs a comparable aggregate o fhousehold consumption for 1998 and 2003 to measure changes in poverty over time. These consumption aggregates subsequently serve as the basis for the analysis o f growth, poverty, and inequality, as well the discussion on social sectors andpoverty. 15. The revision of the consumption aggregate results in some changes in the poverty incidence compared with preliminary estimates by INSDcirculated inAugust 2003. To focus on poverty dynamics and minimize the differences with original estimates, this study uses a poverty line for 2003 that reproducesthe official INSDpoverty incidence figure o f46.4 percent for 2003, This chapter is drawn from EmilTesliuc andMichel KonC (2004). - 4 - and reconstructs comparable poverty lines and incidence figures for 1998. The revised poverty figures are more consistent with overall developments o f economic output measures from the national accounts and non-monetary poverty indicators, such as literacy rates and health indicators. 16. Burkina Faso is among several countries facing issues o f comparability o f survey data over time and engaging in revisions o f poverty incidence estimates. Most countries have improved their surveys over time, and placed most efforts in generating an improved poverty profile based on the most recent and reliable information. However, in many occasions, the comparability o f those estimates was initially o f lesser concern and became only an issue when poverty rates and national accounts data seemed to move indifferent directions. As a number o f recent studies demonstrate the frequent need to adjust poverty statistics derived from non comparable surveys is on the rise (see Box 2). Box 2: RevisingPoverty Aggregates:Lessonsfrom Other Countries A number of countries revised their poverty statistics with the view o f producing statistics that are comparable over time. Madagascar, Senegal and Cape Verde provide recent examples from Sub-Saharan Africa, while Kyrgyz Republic, India and Russia illustrate that such revisions are commonelsewhere. The case o f Madagascar is similar to the one in Burkina Faso (Patemostro et al., 2001). To ensure comparability o f the poverty estimates produced from three non-comparable surveys from 1993, 1997 and 1999, a joint team o f specialists from INSTAT, World Bank and Come11 University produced a partial but comparable consumption indicator, using only a subset o f commodities recorded similarly in the three surveys. Inthe case o f Senegal, Cape Verde and Kyrgyz Republic, such revision was triggered by the inconsistency between a growing GDP, unchanged inequality and raising poverty, similar to the situation inBurkina Faso. Inthe case of Kyrgyz Republic, the inconsistent poverty statistics were produced from a series of LSMSs whose reported consumption was severely affected by seasonality (surveys implemented in different months), the 1999 Russiancrisis, and the selective attritiono f the well-off households. A joint team o f experts from the Statistical Office and the World Bank usedthe panel element of the Household Budget Survey to estimate comparable, annual poverty rates. The use o f a comparable welfare aggregate has produced poverty statistics consistent with the dynamics o f GDP and inequality. Finally, two hghly publicized statistical experiments in India and Ecuador have illustrated how fragile reported consumption is to changes inrecall period, or inthe level o f detail with which consumption is recorded. In India, shortening the recall period from 30 to 7 days increased reported food consumption by 30 percent, as reported inDeaton (2003). Lanjouw and Lanjouw (2000) report another experiment in Ecuador, where a shorter (24 items) and a longer (97 items) consumption schedule was administered to two non-overlapping random samples. The consumption reported by the household from the poorest decile was 25 percent higher in the sample with the longer consumption schedule. A. PRIORITY SURVEYDESIGN AND COMPARABILITY OF POVERTY ESTIMATES 17. The priority surveys collect information on household consumption and incomes- including consumption out o f own production-for poverty and inequality measurement, and information about non-monetary dimensions o f well-being, such as malnutrition, employment, attainment and access to education, health and basic services, housing conditions and endowment - 5 - with durables, land or livestock. Inaddition, the priority surveys gather detailed information on householddemographics. 18. The primary purpose of the priority surveys was to provide: (i)a description o f the standard o f living of the population; (ii) analysis o f poverty and inequality and the factors an associated with it; and (iii)an examination of the nonmonetary aspects o f well-being. The survey was designed to be representative for the main socio-economic groups, areas o f residence and regions. However, changes in the administrative structure o f the country between each survey wave, from 7 regions and 8 strata in 1994 to 10 in 1998 and 13 in 2003, complicated the comparisons at regional level. 19. Many survey design parameters are similar in the three surveys, from sample size to the implementation of a two-stage stratified sampling, with selection o f areas in stage 1 and random selection o f 20 dwellings from those areas in stage 2 (Table 1). Stratification increased over time as 13 regions became the official administrative subunits o f the country, while clustering was alike across surveys. Table 1. Survey DesignParameters for the Priority Surveys Priority Survey EP I EP I1 EP I11 Year 1994 1998 2003 Period Oct-94 - Jan-95 May-Aug Apr-Jul Sample Size -- Households 8628 8478 8500 Individuals 65014 63509 54034 # Strata 8 10 13 # Clusters 436 425 425 HouseholdsKluster 20 20 20 Source: INSD(1996), INSD(2000)and INSD(2003) 20. Some differences between surveys compromise the comparability o f 1994 data with those collected in 1998 and 2003. These include the implementation o f the surveys in different seasons, and the use of a different recall period for some commodity groups. One can only determine the direction o f the bias induced by seasonality and different recall period on reported consumption in 1994 compared to the 1998 and 2003 surveys but it is impossible to construct comparable databases. 21. Other differences between the surveys can be corrected by applying a uniform methodology. These concern notably (i)changes in the questionnaire design; (ii) changes in the methodology used to construct the welfare aggregate; (iii) changes in the methodology used to construct the poverty line; and (iv) changes in the circumstances under which each survey was fielded. 22. Seasonality. The timing o f the surveys was substantially different between the first and subsequent rounds. EP Iwas fielded inthe post-harvest period, from October 1994 to January 95, while EP I1and EP I11were implemented during the pre-harvest period, from M a y to August - 6 - 1998 and April to July 2003. The local population calls the pre-harvestperiod hardshipperiod or lean period. It is very likely that seasonality influenced the level of reported consumption in Burkina Faso in 1994 compared to 1998 or 2003 in opposite dimensions, thus overstating consumption (and understating poverty) in 1994 compared with 1998 and 2003. As argued in World Bank (2004a), there i s a potentially large seasonaleffect on reported consumption, whose magnitude i s impossible to measure accurately. 23. Other studies from Sub-Saharan Africa reported large differences in household consumption in pre- and post-harvest periods. Dercon and K r i s h n h (2000) follow a panel o f 1450 rural households from Ethiopia during 1994 and 1995, before and after the harvest. They found that food consumption was 10 percent higher during the post-harvest period than in the lean period. Such difference in consumption across seasons within the year is very large compared to average growth o f gross domestic product (GDP). For Burkina Faso, with a per capita output growth o f 2-3 percent per annum in recent years, the difference in reported consumptionindifferent seasonscould be equivalent to at least three years o f average growth. 24. Dercon and Krishnan (2000) found that the higher consumption during the post-harvest period is explained in part by the seasonal response to labor demand (consumption rises when labor demand, hence effort, increases) and to the price level (consumption falls inthe pre-harvest season when prices are high). Shocks are also playing an important role. The study reports a lower incidence o f shocks duringthe post-harvest period, which i s likely due to an improvement inthe householdlivelihoods that allows them to better manage risks. The same factors are likely influencing the consumption behavior o f the household inrural areas o f Burkina Faso, hindering the comparison o f the 1994 consumption aggregate with the ones reported insubsequent years. 25. Recall period. The three priority surveys had the same recall period for all consumption items, except for food. Information on food consumption was collected in 1994 using a 30-day recall period, and in 1998 and 2003 for a 15-day recall period. This apparently innocuous change insurveydesign probably had a largeimpact onreportedconsumption. 26. Longer recall periods likely result in lower reported consumption ("recall bias") owing to consumption in earlier time periods not being remembered well, as explained by Deaton and Grosh (2000). This conjecture has been confirmed for several developing countries. Experiments using households from the Ghanaian LSMS found that for 13 high-frequency purchases, reported expenditures fell at an average o f 2.9 percent for each day added to the recall period. Furthermore, Deaton (2003) describes a nationally representative experiment with different recall in India, where shortening the recall period from 30 to 15 days resulted in 30 percent higher annualized consumption (Box 3). 27. InBurkina Faso, shortening the recallperiod for food items from 30 inthe 1994survey to 15 days inthe 1998 and 2003 surveys likely reduced the amount o f "recall bias" and resulted in more reported food purchases and consumption out o f own production. However, since no controlled experiment took place when the recall period was shortened in 1998, it is impossible to know how much reported consumption in 1994 has been influenced by the recall bias, making it impossible to correct the 1994 consumption aggregate in order to compare it with the ones reported insubsequent surveys. - 7 - Box 3: Impact of DifferentRecall Periods on the Level of ReportedConsumption The Case o fIndianNational Sample Survey (NSS) "Survey questionnaires differ in the length o f the recall period over which respondents are asked to report their consumption. The choice of recall period is often thought to involve a tradeoff between accuracy of memory, which calls for a short period, and the match between consumption and purchases, which i s more accurate when averaged over a long period. But there is little understanding o f the effects o f different recall periods, particularly in poor, agricultural societies. In India between 1989 and 1998, the NSS experimented with different recall periods, replacing the traditional 30-day recall period for all goods with a 7-day recall period for food and tobacco, and with a 365 day period for durable goods and some other infrequentlypurchased items. The sample was randomly divided, and half were given the old questionnaire, and half the new, so that it is possible to make a clean evaluation of the effects o f the change. The shorter reporting period increased reported expenditures on food by around 30 percent, and total consumption by about 17percent, very much inthe right direction to help resolve the discrepancy with the NAS. Because there are many Indians close to the poverty line, the 17 percent increase was enough to reduce the measured headcount ratio by a half, removing almost 200 million people from poverty. What might seem to be an obscure technical issue o f survey design can have a major effect on the measurement o f poverty, not only inIndia, but inthe world. It should be noted, however, that the higher consumption totals associated with the shorter recall period, although closer to the NAS estimates, are not necessarily more accurate. Indeed, the NSS has carried out a series o f controlled experiments with a plausible gold standard in which, for many foods, the 30-day reference periodappears to be more accurate than the 7-day period, see NSSO Expert Group on Sampling Errors (2003)" From: Angus Deaton (2003) "Measuring Poverty in a Growing World" (or "Measurine Growth in a Poor World"), NBERWorking Paper9822. 28. Questionnaire design. From one survey to the next, INSD implemented a number o f changes in the design o f the questionnaire, which may have affected the comparability o f the consumption aggregate. Some components o f household consumption were collected only inthe later surveys. For instance (i)food consumption received as gift was recorded only in the 1998 and2003 surveys, (ii) non-food purchases" were recordedonly in1998 and 2003; and (iii) "other the purchases of "school uniforms" were collected only in 2003. More recent surveys collected the same information as the previous ones, but using a more disaggregated schedule. Such changes were resulted in an increase in items used to construct the consumption aggregate from 55 to 74 between 1994 and 2008, and from 74 to 89, between 1998 and 2003 (see Table 2). To ensure that the resultingpoverty statistics are comparable across the 1998 and 2003 surveys, the consumption aggregate used here includes only those items recorded inthe same way inthe 1998 and 2003 surveys. Table 2. Numberof ProductsUsedto Constructthe ConsumptionAggregate Non-Food and Food Services Health Education Total EP I- 1994 23 22 5 5 55 EPII- 1998 33 31 5 5 74 EP 111- 2003 39 39 5 6 89 - 8 - 29. Construction of consumption aggregate. Over time, the methodology for the construction o f the consumption aggregate changed. In all three survey years the welfare aggregate was constructed using the same components, such as the sum o f the annual consumption o f food, schooling, health, rent and other non-food expenditures (including here household investments and durables). These expenditures were adjusted for regional price differences. However, the way these components have been aggregated, annualized and deflated was different: Durables: In 1994, household investments and purchases o f durables were not annualized but durables were annualized in the 1998 and 2003 surveys (if somebody bought a TV during the last month, this expenditure was multiplied by 12 to arrive at an annual figure). Compared with 1994, this change reduced measured poverty in 1998 and 2003, m:The andincreased inequality. surveys collect information on actual rent paid by tenants, and on estimated "rental values'' of the dwelling from owners. However, a considerable fraction o f the households didnot report the "rental value" o f their house (24 percent in 1994, 14percent in 1998 and 6 percent in 2003). In 1994, households who neither reported rents nor estimated their rental incomes were assigned an imputed rental value derived from a hedonic regression o f rent on dwelling characteristics and location indicators. By contrast, in 1998 and 2003, the households who didneither report nor estimate rents were assigned zero consumption o f housing services. This difference increased measured poverty and inequality in 1998 and 2003 compared with the 1994 methodology. Adiustments for the cost o f living: Only in 1994 the consumption aggregate was expressed in constant purchasing power using the Ouagadougou consumer price index (CPI) to correct for month-to-month inflation during the time o f the survey, and a regional price deflator to correct for regional price differences. Different CPIs were used for food, non-food, education, schooling and rent. In 1998 and 2003, per capita consumption was adjusted only for regional price differences, and no correction was made for month-to-month differences although these were larger than in 1994. Seasonal adiustments: To correct for the fact that the 1998 data have been collected inthe "lean period" per capita consumption o f all households was increased by 12.5 percent. The reason behind such adjustment is to get at an "annual consumption" figure closer to the one derived in the system o f national accounts. However, no similar adjustment was made for 2003, when the survey was implemented during the same timeframe. The introduction of this adjustment has an important impact on the resulting poverty trends. Poverty line derivation. The methodology used to construct the poverty line also underwent changes from one survey to the next, hampering the comparability o f poverty estimates: (i) The structure o f consumption basket used to construct the food component o f the poverty line changed over time and is thus not apt for makingcomparisons over time (Table 3). (ii) The food basket used for the estimation o f the food poverty line represents only a small portion of the total food consumption o f the poor, and includes only four basic staples (com, millet, rice and sorghum). The reason for the limited size o f the food basket underlying the food poverty line is the lack of reliable price information for other staples. - 9 - However, the food poverty line is not representative of the consumption of the poor, and will understate food poverty as it uses only cheaper sources of calories, compared to other food items consumed by the poorest 40 percent of the population. This choice results in a lower poverty line and explains why poverty rates in Burkina Faso are lower than in some neighboring countries with similar per capita GDP (see also Box 4 as regards the constructiono f the poverty line). (iii) The high seasonality o f prices o f the main staples, especially during the lean period, contributes to the fluctuation o f the nominal poverty line, and generates dramatic swings inthe share offood intotal householdconsumptionfiom one surveyto the other (Table3). Table 3. Composition of the Food Poverty Line in 1998 and 2003 Sorghum Millet Com Rice Shares of the staples: -- in 1998 in 61% 34% 4% 1% 2003 35% 37% 19% 10% Average price of the staples, in CFA Francs Ouagadougou -- in 1998 165 208 163 285 in2003 140 145 125 250 Food Poverty Line -- in 1998 in 52295 2003 41153 Ratio non-foodlfood - in 1998 - in 1.009 2003 0.390 Box 4: How is the Official Poverty Line Computed? INSDconstructs apoverty line using the "cost ofbasic needs" approach, inthree steps: First, the food component of the poverty line is estimated by valuing a set o f food items providing the recommended intake o f 2283 calories per capita per day at the prices prevailing during the survey. The food items used by INSD consist o f four staples most consumed by the households in Burkma Faso, namely sorghum, millet, com and rice. A standard conversion table is used to determine the caloric content o f each K g o f product. The composition o f this food basket reflects the share o fthese four staples inthe consumption o f the households for each survey, and thus changes from survey to survey. The basket is evaluated at the prices o f the four staplesprevailing inOuagadougou markets at the time o f the survey. Second, the non-food component is equal to the non-food expenditures observed for the households whose food consumption is close to the food poverty line estimated above. The ratio o f non-food to food consumption is estimated for each household. Then, the share o f non-food consumption inthe total poverty line is chosen to be the share reported by those households whose food consumption is close to the value o f the food poverty line (for instance, 21percent). The ratio obtained thus far is multiplied by the food poverty line to get the non-food component o f the poverty line. Inother words ifthe typical household near the food poverty line consumes food and non-food items of about equal value, the non-food component would be set equivalent to the food poverty line. The official poverty line then is determined as the sumo f the food and non-food components. - 10- 3 1. Finally, the circumstances when surveys were fielded were different. These circumstances may have impacted on the comparability (representativity) o f this "snapshot" consumption aggregate with an ideal annual aggregate, such as those reported in the system of national accounts. The priority surveys give only the "snapshot poverty", not "annual poverty" rates. These surveys differ from the practice o f the LSMSs, which collect information on consumption during a typical month (average consumption over the year). Shocks and other circumstances may lead to substantial differences between monthly consumption and income (captured by priority surveys) and annual consumption and income (reflected in national accounts). Inthe case o f Burkina Faso, the following circumstances are important: (i) in 1994, the survey was implemented during the period when the devaluation o f the CFA franc was still propagating through the economy; (ii)in 1998, the survey was implemented after a severe drought; (iii) in2003, the survey was implementedduringthe peak period of the crisis induced by the civil war inthe neighboringCote d'Ivoire. 32. To summarize, the discussion above outlines a variety o f factors that may have introduced biases in the comparison o f seemingly similarly constructed consumption aggregates and poverty lines used thus far in official statistics. The discussion also distinguishes potential biases introduced between the 1994 and subsequent surveys that could not be eliminated and other changes that appropriate methodological adjustments could correct. Table 4 summarizes how the different factors affect measured poverty. Table 4. Impact of Changes in the Survey Design and Methodology on Poverty Statistics Year Impacton ReportedPoverty Changes in Questionnaire Design Datacollectedduring the pre-harvestperiod 1998/2003 Increasespoverty Shorteningthe recallperiodfor foods to 15days 199812003 Decreasespoverty Higher coverageof food consumption(from gifts) 1998/2003 Decreasespoverty More detailedconsumption schedule all rounds Decreases poverty Expenditureson schooluniformsonly in 2003 2003 Decreases poverty Changesin Methodology: ConsumptionAggregate Durablesannualized in 1998/2003 1998/2003 Small decrease in poverty, increasein inequality Seasonal inflation factors in 1998 1998 Reducespoverty, strongimpact Rentsimputedin 1994, self-reportedafter 199812003 Increasespoverty and inequality Changesin Methodology: Poverty Line Different consumptionbasket all rounds Inconsistentpoverty lines Different inflation for the PL items comparedto rest all rounds Inconsistentpoverty lines B. CONSTRUCTING A COMPARABLE WELFARE INDICATOR 33. This section summarizes the normative choices made in constructing the comparable welfare indicator and presents the rationale for these choices. Given the substantial changes operated in the design o f the priority surveys from 1998 to 2003, the focus i s on measuring the changes o f poverty over time rather than absolute poverty. This understanding should facilitate the comparison o f the figures reported inthis paper with other estimates reported inother studies on poverty in Burkina Faso. As its is impossible to address recall and seasonality biases for the 1994 survey, the poverty assessment also refrains from constructing a welfare indicator for 1994 andfocuses onthe 1998 and2003 surveys instead. - 11 - 34. In absence of a unique welfare indicator for measuring poverty, this study chooses to construct a comparable welfare indicator for the 1998 and 2003 surveys by adapting, with as little changes as possible, the methodology used to derive the original welfare aggregate and poverty lines by INSD in 2003. The resulting welfare indicator, real per capita consumption, covers only the items recorded identically inthe priority surveys 1998 and 2003.6 As outlined in some more detail below and in Annex 2, to achieve comparability (i)the coverage of the indicator has been adjusted to a comparable subset o f goods, durables and other household investments have been excluded from the aggregate, and rental payments have been imputed for those households that do not report or estimate their rental incomes, following the procedure used in 1994 by INSD; (ii) consumption has beennormalized into annual consumption; (iii) both temporal and regional price indices have been applied to correct for differences in the cost o f living, following the procedure used in 1994 by INSD; and (iv) in line with INSD, no adjustments have been in dividing consumption by number o f household members. Finally, it i s noteworthy that no further exogenous seasonal adjustment factor has been applied in 1998 or 2003 since the data were collected at about the same time o f the year. 35. As regards the coverage o fthe consumption aggregate, only few consumption items were excluded from the comparable consumption aggregate for theoretical reasons, notably the purchase o f durables, other household investments, and bulky expenditures associated with ceremonies. Unless their rental value can be imputed, it is standard practice to exclude these items (Deaton and Zaidi, 1999 and Deaton and Grosh, 2000). The exclusion o f these items does not have a large impact on reported poverty, but reduces inequality since they are more likely to be consumed by higher-income households. 36. The components o f the consumption aggregate that were not collected identically in the 1998 and 2003 surveys were also excluded. These are summarized in Table 5. By using only a subset of items recorded similarly in the two rounds there is little loss in precision. The abbreviated consumption indicator still covers 84 percent o f the food consumption and 88 percent o f total consumption reported in 2003, and 92 percent o f food consumption and 93 percent o f total consumption in 1998 (Table 6). The 1998 survey, by incorporating fewer and more aggregated items, measured consumption with less precision, explaining why a smaller share was excluded. However, since all included items were measured identically, the comparability o f consumption measures over time is not compromised although the share o f included items is smaller in2003. 37. The remaining consumption items were annualized and aggregated into 9 commodity groups. Consumption annualization assumes, in keeping with INSD practice, that consumption recorded for the past 15 or 30 days i s repeated 24 times or 12 times, respectively, during the year. Annualized consumption is grouped separately for purchases and subsistence or gifts into: (a) food, beverages and tobacco; (b) clothes and shoes; (c) rent and utilities; (d) house maintenance; (e) health; (f) transport; (g) leisure; (h) health; and (i) other items. These groups match the CPI sub-indices collected by INSD Price Department for the capital city Ouagadougou (see Table 7), usedinthe next steps to correct for the variation inthe cost o f living. An alternative welfare indicator would be household revenue, information on which is also collected by EPs. Informationonrevenue is consideredbelow. - 12- Table 5. ItemsRecordedDifferentlyin 1998 and 2003 (Excluded from the consumptionaggregate) EP I11998code EP 1112003 code Fish andother sea products 10 110,111 Meat (exceptpoultry) 11 112, 113, 114 Vegetables 18 121,122, 123, 124 Pretaporter 15 215,216 Textiles 17 218,219 Schoolunfomes not collected 1.3 Table 6. Coverageof the ComparableConsumptionAggregate (Percent o f fulhon-comparable aggregate) Welfare Full/ Non- Partial/ Aggregate Comparable Comparable Ratio Year (1) (2) - 1) Per capita consumption 1998 109,827 102,568 93% 2003 130,734 114,846 88% Per capita food consumption 1998 59,753 54,992 92% 2003 70,700 59,613 84% Note: The full/non-comparable aggregate is constructed using the same methodology as the partiallcomparable one, without excluding any of the consumption items which were either added or disaggregated in 2003. Both indicatorsdo not includedurables or other householdinvestments. 38. Consumption aggregates are adjusted for differences in the cost o f living both across space and time. First, the regional price index constructed by INSD for each survey is used to transform household consumption from regional to constant Ouagadougou price^.^ Second, each of the 9 components o f household consumption is expressed in June 2003 constant prices using the corresponding price index for each commodity group (see Table 7). This method performs a more accurate adjustment than the use o f the CPI for aggregated consumption, as it implicitly applies household-specific consumption shares instead o f the CPI weights. It thereby takes into account that the poor may consume more food items than the average household when adjusting for price differences over time. The method also assumes that prices are changing proportionally with the changes observed (only) inOuagadougou. 39. Consumption is expressed inper capita terms by dividing household consumption by the number of people residing in the household. In keeping with INSD practice, there is no adjustment for household size, i.e. for economies o f scale in consumption or for differences in the needs of children versus adults. This methodology would tend to increase poverty measures for larger households since a child would need the same consumption as an adult to pass the poverty line. Ignoring economies o f scale inconsumption and smaller needs o f children therefore also tends to overemphasize family size as a determinant for poverty. 7 Regional price indices are difficult to measure owing to the necessity to measure them in larger regional markets and the difficulties incollecting data. This problemmay introduce measurement errors inconsumption aggregates. - 1 3 - Table 7. OuagadougouConsumer PriceInflatorsby CommodityGroup (June 2003=100) ... I. CommodityGroup May-98 Jun-98 Jul-98 Aug-98 Apr-03 May-03 Jun-03 Jul-03 Food, beveragesand tobacco 103.5% 98.2% 103.2% 101.7% 117.4% 106.5% 100.0% 108.6% Clothes and shoes 113.6% 113.8% 113.8% 113.7% 99.8% 100.0% 100.0% 100.9% Rent and utilities 111.3% 112.4% 112.5% 111.4% 101.3% 99.7% 100.0% 101.3% House maintenance 101.6% 101.1% 101.0% 100.4% 99.9% 99.9% 100.0% 99.9% Health 108.2% 108.0% 110.5% 111.3% 100.9% 100.9% 100.0% 100.0% Transport 118.8% 119.7% 119.4% 119.7% 97.2% 98.5% 100.0% 100.4% Leisure 103.2% 103.2% 103.2% 103.2% 100.0% 100.0% 100.0% 100.0% Education 110.7% 110.7% 110.6% 110.6% 100.0% 100.0% 100.0% 100.0% Other goods 113.6% 113.5% 113.5% 113.2% 100.0% 100.0% 100.0% 100.1% Total 109.2% 107.2% 109.3% 108.6% 105.2% 102.0% 100.0% 102.7% 40. The poverty line that divides the population into "poor" and "non-poor" is determined with the objective to generate the same poverty headcount that was estimatedby INSD for 2003 with a non-comparable welfare indicator. Using the 2003 survey and the comparable consumption aggregate, a poverty line of CFAF 72,110 in June 2003 Ouagadougou prices replicates the poverty headcount of 46.4 percent reported by INSD for 2003 (INSD, 2003). By way of comparison, the INSD useda poverty line o f CFAF 82,672 applied to their consumption aggregate, which cannot becomparedover time, to arrive at apoverty headcount of 46.4 percent. c. MEASURING CHANGES INPOVERTY 41. On the basis of the comparable welfare indicator, one can conclude that Burkina Faso experienced substantial reduction in poverty, without significant changes in inequality, between 1998 and 2003 (Table 8). The significance o f these results can be established from the survey design (Box 5) and confidence intervals are presented in Table 8. The poverty headcount went down from an estimated 54.6 percent in 1998 to 46.4 percent in 2003, and both the poverty gap and its severity declined. The reduction in poverty was more pronounced in rural areas, where the headcount fell from 61.1 percent in 1998to 52.4 percent in2003. The reduction o fpovertyin urban areas was substantially smaller, and not statistically significant. Inequality remained unchanged between the two surveys (Gini Index o f 0.444), and was systematically higher in urban (0.484) than inrural areas (0.376). 42. The findings as regards the variation in poverty hold for a broad range o f plausible poverty lines. Figure 1 uses 1'` order stochastic dominance tests to investigate the trends in poverty to illustrate this point. Insetting the poverty line there i s a certain subjective element, the most important being the choice of the caloric requirement that would "anchor" the food component of the line. This raises the question how robust poverty rankings are across time to the choice ofthe poverty line. 43. One way to make robust comparisons without choosing a particular poverty line i s by testing for stochastic dominance, i.e. by comparing poverty incidence curves (or cumulative density curves) that summarize the distribution of per capita consumption. A poverty incidence curve shows the cumulative share of population against real per capita consumption. The points - 14- along the horizontal axis can be considered as the complete set o f possible poverty lines. The proportion o f the poor can be found by reading off the proportion o f the population on the vertical axis whose per capita consumption is less than a given level. If, at any poverty line, one poverty incidence curve i s above the other, the amount o f poverty inthe first population will be always greater than in the other population for any given poverty line. If two poverty incidence curves intersect, then the comparisons will give different results for different ranges o f poverty lines, However, as Figure 1 shows, this was not the case for BurkinaFaso in 1998 and 2003, and thus the results as regards the measured decline in poverty do not depend on the poverty line used. Table 8. PovertyandInequality. PovertyandInequalityDynamics National Rural Urban Estimate Std. Err. Estimate Std. Err. Estimate Std. Err. Poverty Headcount 1998 54.6% l..l% 61.1% 1.1% 22.4% 1.9% 2003 46.4% 1.2% 52.4% 1.3% 19.2% 1.7% Poverty Gap 1998 0.183 0.005 0.208 0.005 0.057 0.006 2003 0.153 0.005 0.176 0.006 0.051 0.005 Square Poverty Gap 1998 0.082 0.003 0.094 0.003 0.022 0.003 2003 0.068 0.003 0.079 0.004 0.019 0.002 GiniIndex 1998 0.444 0.013 0.349 0.011 0.499 0.027 2003 0.444 0.011 0.376 0.012 0.484 0.013 Note: Standard errors take into account the survey design. The poverty gap measures the average percentage distance of the expenditure of the poor from the poverty line. The square poverty gap or poverty severity measures the average squared percentage distance of the poor from the poverty line. Box 5: StatisticalSignificanceof ObservedPovertyChanges The poverty and inequality figures publishedby MSD and quoted in most official documents of the Govemment, the World Bank or other donors are sample estimates. They are not derivedfrom a census, or any other exhaustive research on household welfare. They are estimatedfrom surveys, with some known (im)precision. This precision improves with the stratification of the survey, and worsens with the degree of clustering of the sampled households.Knowing the inverse o f the probability of selection of each household in the sample, the survey strata and clusters, confidence intervals having a given statistical significance(typically 5 percent) can be constructed. A 95-percent confidence interval, for instance, guarantees that the true povertyrate or inequalityindex will be found in this range in 95 out of 100 trials, where by trials we mean a different survey implementedat the same time, under the samedesign, but on adifferent sample ofhouseholds. The sample size of the priority surveys is comparable with intemational practice. Given the population of Burkina Faso, around 10-12 million during 1998/2003, a sample of 8,500 households gives reasonable precision. We use the following survey design parameters to generate the standard errors (and hence, the confidence intervals) of the poverty and inequality statistics producedin this paper: the codes for the primarysamplingunits(425 intotal) indicatethe clusteringofthe survey; and the regions(10 in 1998 and 13 in 2003) indicatethe stratificationvariables. we use the set of expansionweights producedby INSDto extrapolatethe results from the sample to the total population. - 15- Figure 1. ComparingPovertywithout PovertyLines, 1998-2003 Poverty IncidenceCurves - 4 Burkina Faso, National, 1998 and 2003 I ______- -- ------ ,q- 8'"- z C P ps - b aJ n,- 0 - I I I I 0 Per Capita Consumption,'000sCFA FrancsJun-2003 100 200 300 I- 1998 -----2003I Readon y axis the correspondingpoverty headcount for any povertylineon x axis RuralArea Urban Area r .-Q - 3gq - I B, - a 8 ' v - 0 - a I Per 0CapitaCcnsumMion,'000s CFA FrancsJun-200. 100 200 300 Per0CapitaConsumption. '000s CFA FrancsJun-2003 100 200 300 -1998----20031 1998 2003 Readany axis memrrespordingpovertyheadcount fw any poWlineonx axis for any povertylinean x 8us Readon y axisthe mrrespondingpoverty headcount Note: The three poverty incidence curves presented inFigure 1 illustrate the relationship between a given (arbitrary) poverty line and the resulting poverty headcount. The points along the horizontal axis can be considered as the complete set of poverty lines. The proportion o f the poor can be found by reading off the proportion o f the population on the vertical axis whose per capita consumption is less than a given level o f per capita consumption. If, at any poverty line, one poverty incidence curve [I]is above the other [2], the amount of poverty inpopulation [11will be always greater than inpopulation [2] for any given poverty line. Inour case, the poverty rate would be lower in 2003 at any poverty rate, at the national level, as well as inrural and urban areas. The vertical line at CFAF 77,110 representsthe poverty line in2003 prices. - 16- 44. Usingrevenue instead o f consumption aggregates allows comparing 1994 and 1998 data consistently. Up to now, the poverty assessmentdidnot consider 1994 data since its consumption aggregate is affected by seasonality and different recall period for food consumption. However, the 1994 and 1998 surveys also collect detailed information on household incomes, from agricultural sales, wages, profits, rents and other monetary transfers, using a similar schedule and the same recall period (revenues obtained during the last year). This comparison cannot be extended directly to 2003, when the recall period was shortened to 30 days. Figure 2 presents a 1'' order stochastic dominance test for a real per capita revenue aggregate constructed on the basis of the three surveys. The 1998 cumulative distribution curve lies below the 1994 one, suggesting a strong reduction inpoverty betweenthe two years. 45. The same trend as for 1994-98 i s apparent from the comparison o f the 2003 revenue- based curve with 1998 data for moderate poverty lines. However stochastic dominance is no longer valid for extremely low levels o f the poverty line. The latter phenomenon is due to the shortening o f the recall period in 2003 which increased the variance o f the reported revenue, as many farmers or other seasonal workers who where unemployed or sold their harvest duringthe month preceding the survey reported zero cash revenues. The same workers, if asked as in 1994 or 1998 about their cash revenues during the previous year, would have reported positive amounts. The result is that the welfare indicator constructed on the basis o f revenues in 1998 and 2003 i s not comparable. Figure2. ComparingPovertywithout PovertyLines, 1994-2003 Cumulative Distributionof Per Capita Revenues CC _I_....__-.. *IC----- _ _ _ _ _ _ _ _ _ - _ _ - __..__ _ - _-.---- _ - ~ rd ._..a .__..--* _.__..._ .-.--- ..-.--___..-- ./-...-. *..e .*- III I 0 77 100 200 300 PerCapita Revenues, `000sCFA FrancsJun-2003 1994 -----1 gga .._________2003 I Readony axis the correspondingpovertyheadcount for any povertyline onx =is Note: The vertical line at CFAF 77,110 represents the poverty line in 2003 prices. - 17- D. CONCLUSIONSAND RECOMMENDATIONS 46. The construction of comparable welfare measures for evaluatingpoverty dynamics needs to take into account the changes in survey design. Almost all countries have improved their surveys over time, and placed efforts in generating an improved poverty profile based on the most recent and reliable information. However, in the case o f the Burkinabk priority survey, apparently innocuous changes in survey design, such as the change in the recall period, trigger substantial and spurious changes in reported consumption, and hence measured poverty. In addition, the change in the timing of the survey has had a substantial impact on poverty trends given the large differences inconsumption inthe lean (pre-harvest) andthe post-harvest seasons. 47. Adopting real per capita consumption as welfare indicator, in line with the choice o f the Burkinabb authorities, it is possible to correct for changes in survey design and construct a comparable welfare indicator for 1998 and 2003. This indicator i s built on the coverage o f a comparable subset o f goods and the adjustment for price differences over time and between regions. 48. Applying a comparable consumption indicator to measure poverty dynamics between 1998 and 2003, it canbe shown that poverty declined from about 55 percent to 46 percent, inline with growth inoutput. A stochastic dominance curve reveals also that the conclusion regarding a decline inpoverty between 1998 and 2003 is robust to the choice o f the poverty line. 49. A comparable consumption aggregate cannot be constructed from the 1994 survey data because o f biases introduced by different recall periods and seasonality. However, it is possible to compare householdrevenues collected by the 1994 and 1998 surveys. A stochastic dominance curve based on these revenue measures also shows that poverty declined for any choice o f poverty line between 1994 and 1998. Because o f changes in survey methodology in 2003, the same analysis based on comparable revenue measures cannot be carried forward for 2003. 50. To produce more reliable poverty indicators, comparable over time, the following is recommended: anticipate the impact o f survey design changes on the comparability o f data and enhance the questionnaire to be able to estimate the potential bias improve the collection o f price statistics, both outside and within the priority survey, by collecting information on both quantities consumed and the value o f the respective purchases, or by administering price questionnaires at the community level; implementthe survey in quarterly waves (quarterly panel), to eliminate the influence o f seasonality and short-lived shocks on reported consumption; provide adequate and timely financing for the survey, for an adequate monitoring o f poverty and to ensure the comparability o f the welfare indicators from micro- and macro- data sources. - 1 8 - 111. GROWTHAND POVERTY 1998-2003 5 1. Poverty declined significantly during 1998-2003 according to a comparable welfare indicator constructed from two priority surveys. This chapter investigates how the poverty changes during 1998-2003 relate to the growth pattern o f the economy. It first reviews the growth pattern between 1998 and2003, and discusses the sources o f growth, before reviewingin more detail the changes inmeasuredpoverty and their microeconomic determinants. The chapter subsequently links growth and poverty to review how changes inoutput may be transmitted into income changes for different socio-economic groups and thereby influence poverty rates. A variety o f tests are performed to identify the statistical nature o f the link between poverty reduction and growth andto derive key conclusions about the growth process. A. GROWTH PATTERNS 52. Burkina Faso has enjoyed GDP growth averaging 5.5 percent over the six-year period from 1998 to 2003 (Table 9). Overall, output growth over that period was driven by primary sector growth o f almost 8 percent per year, and agricultural growth inparticular, which averaged 10.5 percent over the 6-year period, albeit with considerable volatility related to rainfall patterns. Secondary sector growth was close to average growth rates as manufacturing grew at slightly lower pace than the economy with a fast expansion o f public utilities. The tertiary sector grew less than the economy on average, with other services and transport contributing the bulk o f the growth. Overall, between 1998 and 2003, the structure of GDP has changed little with dominating primary and tertiary sectors each representing 37-40 percent. 53. Growth in the agricultural sector deserves particular attention in analyzing growth patterns and poverty since more than 80 percent o f the population lives in rural areas. Poverty measures pick up to a large extent variations in the well-being o f the rural population, and thus the output o f the rural economy plays an important role in linking household survey results with GDP growth. Inanalyzing agricultural growth rates for Burkina Faso from national account data, one has to keep inmindthat agricultural output of any given calendar year is harvested mostly at the end o f the year and therefore included in that year's GDP figure, but the economic impact o f the harvest on demand, prices, and well-being i s largely felt inthe following year andrecorded in that year's GDP. Consequently, as the household surveys have been collected during May- August (1998) and April-July (2003), they would reflect the size o f agricultural output in the year preceding the survey and its secondary effect in the survey year. For completeness sake, Table 9 therefore also reports GDP figures for 1997. 54. After a drought in 1997-98, the period 1998-2003 was characterized by a substantial growth in cereal production, increasing availability o f basic staples and food security for the rural population (Figure 3). During the 1997-98 crop year, cereal output fell by more than 9 percent during a drought, with production estimated at about 2.2 million tons, a production level that was similar in other drought years in the past decade (1995 and 2000). The lower level o f available cereal output constrained consumption levels of rural households in the year o f the household survey (1998). Cereal production recovered in 1998 and 1999 to 2.6-2.7 million tons, exceeding levels reached prior to 1997. A drought in 2000 again limited cereal output, but output subsequently climbed to record levels in 2001 and 2002, exceeding 3 million tons, and reaching - 1 9 - an exceptional 3.6 million tons in 2003. According to govemment estimations, the production levels in2001-03 exceed levels neededto ensure the self-sufficiency of BurkinaFasoby 500,000 to 1million tons. Table 9. Real GDP: Shares and Sectoral Growth Pattern, 1998-2003 1997 1998 1999 2000 2001 2002 2003 Average 98-03 (Ratio, inpercent of GDP) Gross domestic product 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Primarysector 34.4 37.0 37.5 35.8 38.5 37.7 38.7 37.5 Agriculture 18.2 20.8 20.5 18.5 21.9 21.2 23.C 21.c Livestock 12.0 12.1 13.1 13.2 12.7 12.7 12.1 12.6 Forestryand fishing 4.3 4.1 4.0 4.1 3.9 3.8 3.7 3.9 Secondarysector 17.1 16.3 15.4 16.3 15.4 16.8 17.1 16.2 Manufacturing 12.4 12.4 11.0 11.5 10.9 11.9 12.3 11.7 Electricity, gas, and water 0.5 0.6 0.6 0.8 0.7 0.9 1.o 0.8 Constructionandpublic work 4.3 3.4 3.8 4.0 3.8 3.9 3.9 3.8 Tertiary Sector 41.3 39.8 40.5 42.4 40.6 40.3 39.4 40.5 Trade 10.2 9.0 7.5 7.8 7.5 7.6 7.5 7.8 Transport 5.8 6.1 6.2 6.9 6.7 6.7 6.6 6.5 Non-marketservices 16.5 15.5 17.1 16.8 16.1 15.5 14.7 15.9 Other services 8.7 9.2 9.7 10.9 10.4 10.5 10.6 10.2 Indirect taxes andduties 7.2 6.9 6.5 5.5 5.5 5.2 4.8 5.7 (Percentagechange) Gross domestic product 6.8 8.5 3.7 1.6 6.8 4.6 8.0 5.5 Primarysector -3.7 16.6 5.3 -3.2 15.0 2.4 10.8 7.8 Agriculture -I1.5 24.5 1.9 -8.1 26.5 1.o 17.1 10.5 Livestock 8.1 9.6 12.2 2.5 2.5 4.8 2.5 5.7 Forestry and fishing 3.6 2.8 2.0 3.0 3.0 2.5 3.5 2.8 Secondarysector 20.2 3.6 -2.2 7.7 0.4 14.1 10.4 5.7 Manufacturing 21.2 5.0 -7.3 6.0 0.9 14.7 11.4 5.1 Electricity, gas, andwater 18.5 31.1 4.9 29.3 -1.0 35.2 10.9 18.4 Constructionandpublic work 7.5 -4.8 15.2 8.9 -0.5 8.3 7.4 5.7 Tertiary Sector 10.9 4.5 5.7 6.2 2.4 3.7 5.5 4.7 Trade -2.2 -4.3 -13.1 5.3 2.7 5.2 6.7 0.4 Transport 21.3 13.1 6.4 13.1 2.7 5.2 6.7 7.9 Nonmarket services 14.8 1.6 14.7 -0.6 2.7 0.7 1.9 3.5 Other services 14.8 14.7 8.3 14.3 1.8 6.1 9.2 9.1 Indirecttaxes andduties 12.1 3.6 -1.4 -13.8 5.4 0.0 -1.1 -1.2 I I 55. Cereal production outpaced population growth over the 1998-2003 period (Figure 4). With the exception o f the dip in production during the 2000 drought, cereal output has been risingfaster than population during the time between the two past household surveys. Cereal per capita production since 1998 also matched or exceeded levels recorded after the devaluation during 1994-96. - 20 - Figure3. ProductionofMain CerealsandCotton, 1997-2003 (1997 = 100) 240.0 - 220.0 - 200.0 - 180.0 - 160.0 - 140.0 - & / p-, / ..... . $9 hum 120.0 - 100.0 - 60.0 1 , 1 I I I I I 1997 1998 1999 2000 2001 2002 2003 56. Production o f cotton, the maincash crop, has only recently recovered from the impact o f the "white fly." Cotton output grew sharply after the devaluation, reaching a record level of almost 340,000 tons in 1997-98, up from 150,000 tons in 1993-94. However, owing to the severe impact of the "white fly," which destroyed an important part o f the harvest in 1998-99 and led to a loss o f confidence by farmers, output stayed between 250,000 and 280,000 tons untilthe 2001- 02 season, when cotton output surpassed 375,000 tons. Taken together, the cotton and cereal output growth explain the bulk ofthe growth inagricultural output during 1998-2003. 57. Growth in cereal and cotton output relied mostly on increased land use and very little on increase in productivity. According to agricultural statistics, the productivity o f cereals varied with climatic conditions but remained otherwise largely constant, yielding about 0.8 tons o f cereal per hectare of cultivated area. Similar, cotton yielded around 1 to 1.1 tons per hectare duringthe past 10 years. The increase inagricultural output over the past decade can therefore be largely ascribed to an increase incultivated areas. 58. The total value o f agricultural output rose with the increase in production and also outpaced population growth over the 1998-2003 period (Figure 5). The figure applies farm gate prices for cotton and market prices for cereals inthe following year to total harvests o f the prior year. The value o f production is particularly important for cotton production, since it is entirely sold by farmers and creates cash incomes.* Owingto the price stabilization scheme o f the cotton company, the pass-through o f international cotton price volatility i s limited, although the sharp decline in2001-02 led to a reduction inproduction values at rising output levels. For cereals, the For the links between poverty and cotton production, see Wodon (2004). - 21 - increase in values is an important indicator that those households selling surpluses at the margin were better off in2001-02 than in 1997-2000. Figure4. Cerealand CottonOutput, andPopulationGrowth (1997 = 100) 170.0 160.0 150.0 140.0 130.0 120.0 110.0 100.0 90.0 80.0 70.0 1997 1998 1999 2000 2001 2002 2003 Figure5. Value of Cerealand CottonOutput, and PopulationGrowth (1997 = 100) 200.0 180.0 160.0 140.0 120.0 100.0 80.0 60.0 I I I I I 1997 1998 1999 2000 2001 2002 59. As noted above, secondary sector growth expanded at about the same rate as GDP during 1998-2003. According to the semi-final national accounts for 1998, 37 percent of the value - 22 - added in the secondary sector came from production o f food and tobacco, followed by textile industries and construction and public works with about 20 percent each (Table 10). These three subsectors thus explain 75 percent o f the value added inthe secondary sector in 1998, and in all three informal sector activities play an important role as shown in Table 10. Unfortunately, detailed national accounts in semi-final form are only available through 1999, and the industrial production index, due to its narrow coverage, does not give reliable information o f secondary sector growth, especially as regards new enterprises and informal activities, and therefore does not match national account estimates for 1998-2003. Therefore, national account figures for 2000-03 are therefore largely estimations based on different indicators and assumptions about informal sector activity. ' Table 10. Subsector Shares in SecondarySector andInformalSector Value Added, 1998 (Inpercent) Subsector Share of value added in Share of informal value secondary sector added inthe subsector Extractive industries 2.5 0.0 Foodproducts, drinks, and tobacco 37.5 80.2 Textile industries 20.7 49.7 Paper production 0.7 0.0 Chemical industry 5.O 0.0 Production of non metallic minerals 1.6 73.5 Works inmetal and wood 5.5 60.0 Electricity, gas, and water 6.4 0.0 Construction and public works 20.2 62.2 60. Secondary sector growth performance reflects estimates o f formal and informal activity in the secondary sector. The estimated growth performance in manufacturing (food, textile, paper, chemicals, minerals, works), mostly represents measurable formal and estimated informal sector activity in food andbeverage industries and textiles (cotton ginning for the formal sector). To the extent that informal activity must be estimated and relies only on indicators, national accounts may capture economic activity only imperfectly, and links with household surveys may be more difficult to establish than in the primary sector, which is driven by more easily observable cereal and cotton harvests. 61. In the tertiary sector, growth performance was constrained below average GDP growth rates by relatively slow expansion o f trade and non-market services (government). The restraint in expanding government consumption and the declining share o f trade in GDP in 1998-2000, giventhe subdued export performance, explain the low growth rates. However, in2002-03, trade and transport are estimated to have picked up in lockstep with rising cotton prices and exports and the increased activity following the Ivorian crisis, which diverted merchandise for landlocked Sahelian countries from the rail system and Ivorian roads to BurkinabC roads. As is the case for the secondary sector, large shares o f trade as well as the fastest growing subsector "other services" in the tertiary sector reflect informal activity and their estimations cannot be based on accounting data (Table 11). Again, as will be discussed in more detail below, the imperfect information may hamper the joint analysis o f national account growth estimations and results o f household data. -23 - Table 11. Subsector Shares in Tertiary Sector andInformalSector Value Added, 1998 (Inpercent) Subsector Share o f value added in Share o f informal value tertiary sector added inthe subsector Trade 11 38.4 46.8 Hotels, bars, and restaurants 5.7 23.6 Transport, stocking and communication 10.2 19.6 Financialestablishments 3.3 0.0 Insurances 0.3 0.0 Other market services 17.4 79.2 Non-market services 11 24.7 0.0 62. The uses side o f the national accounts also gives some indications as regards links between national account data and household data. Table 12 combines available population data with national account figures broken down into consumption, investment, and net exports. Customarily, little information is available about private consumption, which is usually treated as the residual after public consumption and investment have been calculated from fiscal accounts and net exports and private investment have been estimated. As a result consumption shares in GDP tend to be relatively volatile. National accounts estimates for Burkina Faso place consumption between 91 and 98 percent during 1998-2003, with a declining trend since 1999, in line with a declining external imbalance (narrowing o f the resource gap). Investment shares have been fairly stable since 1999,with a declining trend o f public investment shares reflecting inpart the fact that more attention is being given to finance recurrent costs o f social services. 63. Despite important limitations o f the data, it i s possible to align national account data reasonably well with expenditure estimates from the household surveys.' Overall, the household survey estimates indicate that real expenditure per capita grew on average by about 2.3 percent per year between 1998 and 2003. This finding can be compared with national account data indicating that real per capita GDP during the same period expanded at an average rate of 2.6 percent per year. It is important to keep inmindthat household surveys took place inthe first half of 1998 and 2003 and therefore reflect only part o f the growth in the respective year, precluding a perfect alignment o f annual national account figures with infra-annual householddata. - 24 - Table 12. GDP, PrivateConsumption,and Household Expenditure, 1998-2003 1998 1999 2000 2001 2002 2003 Average (InpercentofGDP incurrentprices, unlessotherwiseindicated) Consumption 91.3 97.1 97.6 95.2 94.4 93.2 94.8 o/w Public 12.0 11.8 12.7 12.0 12.9 12.2 12.2 Investment 22.8 17.7 18.6 18.5 18.0 18.3 19.0 o/w Public 7.8 9.8 8.5 8.1 7.1 6.3 7.9 Resource gap I/ -14.1 -14.8 -16.2 -13.7 -12.4 -11.5 -13.8 GDP per capita (in CFAF, current prices) 168,890 165,772 163,166 180,245 192,021 198,189 178.047 Privateconsumption per capita(in CFAF, current prices) l33,9 17 141,530 138,558 149,883 156,538 160,413 147,244 InUSDollarsper capita 227 230 195 205 225 266 223 Expenditureper capita(inCFAF, current prices) 2/ 109,827 ... ... ... 130,734 ... I.. InUSDollarsper capita ... ... ... ... ... (Percentage change) GDP (constant 1985 prices) 8.5 3.7 1.6 6.8 4.6 8.0 5.5 GDP per capita(constant 1985 prices) 6.1 I.3 -0.8 4.3 2.2 2.5 2.6 GDP (in current prices) 15.0 0.5 0.8 13.1 9.1 8.7 7.7 GDP per capita(currentprices) 12.3 -1.8 -1.6 10.5 6.5 3.2 4.8 Total consumption (current prices) 15.3 7.0 1.2 10.3 8.2 7.3 8.2 Private consumption 16.1 8.2 0.2 10.7 6.9 7.9 8.4 Publicconsumption 10.3 -1.3 8.5 7.5 16.7 3.2 7.5 Gross investment(current prices) 11.9 -22.1 6.1 12.4 6.3 10.3 4.1 Private investment 15.3 -47.3 30.0 16.4 14.5 18.7 7.9 Public investment 5.9 26.2 -13.1 7.5 -4.3 -2.8 3.3 Resourcegap (currentprices) -12.2 -5.8 -10.1 4.4 1.o -0.2 -3.8 Memo item: Exportsofgoods and services 36.4 -16.3 -5.5 10.1 0.2 15.8 6.8 Consumer price index 5.0 -1.1 -0.3 4.9 2.3 2.0 2.1 Total population31 2.4 2.4 2.4 2.4 2.4 5.3 3.0 Private consumptionper capita (current prices) 13.4 5.8 -2.1 8.4 4.5 2.6 5.5 Privateconsumptionper capita (constant prices, deflatedwith CPI) 8.0 6.9 -1.8 3.1 2.1 0.5 3.1 Expenditureper capita(in constantprices) 21 ... ... ... ... ... ... -. L.5 I/Definedasthedifferencebetweenexportsandimportsofgoodsandservices. 21Fromthe 1998 and2003 priority surveys, adjustedas discissed in the precedingchapter. 3/ Assumes that about 300,000 peopleof Burkinabt origin retumed to BurkinaFaso in early2003. Sources: Burkinabt authorities, IMF, andcalculationfrom 1998 and 2003 priority surveys. 64. More importantly, real household consumption data can be aligned with real private consumption from the national accounts. For that purpose we deflate private consumption per capita in current prices with the changes in the consumer price index (CPI). The calculations - 2 5 - show a real growth rate of per-capita private consumption o f 3 percent during 1998-2003.'0 Giventhe data problems for both national accounts andhousehold surveys, the growth rates can be considered as broadly consistent and confirm that, from an aggregate standpoint, the methodology introduced in the previous chapter at best underestimates growth in private consumption. National account data also point to a possible time pattem o f expenditure growth for households since most o f consumption growth occurred in the period o f recovery after the 1997-98 drought, whereas more recently private per-capita consumption grew more slowly in real terms thanper-capita GDP. 65. Inflation patterns of the past years have been closely in line with cereal production and rainfall given the large weight of food items in the consumer price index. On average, inflation was low at about 2 percent. However, consumer price inflation reached 5 percent in years following a drought (1998, 2001) and was negative when agricultural production recovers. The volatility o f food prices can expose those households to risks that purchase foodstuffs on the market and produce no or insufficient amounts o f food for their own consumption. B y contrast, farming households that can rely on their own food production are less exposedto price changes on markets since they access markets only at the margin. B. POVERTY DETERMINANTS 66. In matching growth experience with expenditure data and discussing policies that can serve to reduce poverty, it i s useful to investigate microeconomic determinants o f poverty. Several authors have already investigated determinants of poverty in 1994, 1998, and 2003, notably Fofack (2002), and Lachaud (2003), and their results are reviewed in the following jointly with additional estimations undertaken for the purpose o f this study. Ininterpreting these results it needs to be kept in mind that these regressions capture correlates o f poverty rather than "determinants" in the strict sense. Nonetheless, we retain below the traditional term poverty determinant. 67. Different studies have identified similar key poverty determinants despite changes in methodology using different household surveys. In terms o f methodology, previous work has usedprobit (Fofack, 2002) and logit (Lauchaud, 2003) models on household data without further correction for inconsistencies in measuring poverty levels. Annex 3 reports results from a regression for the logarithm of expenditure, income, and assets on different variables where household expenditure data has beencorrected as described above. Table 13 summarizes the key findings from these different sources as regards significant poverty determinants. Overall, the different studies found very similar significant poverty determinants, indicating that the qualitative results are fairly robust to poverty measurement issues and have changed little over time. 68. Fofack (2002) and Lachaud (2003) report broadly consistent poverty determinants. Fofack's results suggest that the age-dependency ratio, asset ownership, education and literacy, IoAlternatively, it is possible to review the uses side o f real GDP figures. However, these data turn out to be fraught with problems, notably difficulties indefining the appropriate deflators. Currently available estimates for the uses o f real GDP result in sharp downward trends for consumption shares inreal GDP and a sharp upward trend in public investment shares inreal GDP, both o fwhich are inconsistent with a sustainable real growth path. - 26 - and location (represented by poverty mapping) are the key determinants o f poverty in Burkina Faso. Lachaud similarly finds age dependency, education and literacy, employment pattern, spatial distribution of income or expenditure, as well as migration and transfers are significant in explaining the nature o f poverty. In what follows, we explore the analysis undertaken for this study that adds occupational identifiers and land (see the table in Annex 3) jointly with the finding o f previous studies. Poverty is closely related to household size and particularly the number of children. Households with larger number o f children have lower levels o f per capita consumption and income, and thereby a higher probability o fbeing poor. 69. According to the estimates, an additional child in urban areas is associated with a 14 percent decline inper-capita expenditure. Inrural areas, this effect is larger varying from 18 to 30 percent depending on the age of the child. By contrast the number of young children and children is correlated with higher asset scores indicating the tendency to hold larger savings and assets inlarger families. These conclusions are very similar to those o fthe prior studies. Table 13. ComparativeAnalysis of PovertyDeterminants Fofack(2002) Lachaud(2003) Povertyassessment EP 1994& 1998 EP 1998 & 2003 EP 2003 Regressionvariable Variable I Signifi- Variable Signifi- used I cant used II cant Intercept Yes No Yes Yes Yes Yes Age-dependencyfactor Yes Yes Yes Yes Yes Yes Head's literacy Yes Yes Yes Yes Yes Yes Head's occupation No N.A. No N.A. Yes Yes Spouse's literacy No N.A. No N.A. Yes Yes Spouse's occupation No N.A. No N.A. Yes No Povertymappingvariable Yes Yes No N.A. No N.A. Socio-economicgroup (SEG) No N.A. Yes Yes Yes Yes Level of educationby SEG Yes Yes No N.A. No N.A. SEG mapping Yes Yes No N.A. No N.A. Asset mapping Yes Yes No N.A. No N.A. Geographic location No N.A. Yes Yes Yes Yes Householdhead(age, sex, marriedor not, etc.) No N.A. Yes No Yes Yes Householdtype (monoparentalor not) No N.A. Yes Yes No N.A. Migration 2/ No N.A. Yes Yes Yes Yes Transfers No N.A. Yes Yes No N.A. Access to safe water Yes Yes No N.A. No N.A. Type of fuel for cooking Yes Yes No N.A. No N.A. Landowned for production No N.A. No N.A. Yes Yes urce: Fofack(2002), Lauchaud(2003), World E ik estimates. 1/ Both in urbm'andrural areas where applicable.Significanceis measuredat the IO-percent level. 2/ Migrant from C6te d'Ivoire in2003.Migrant looking for work or landin 1998. 70. The results of the regressions are influenced by the choice of the consumption measure that disallows economies o f scale. As noted in the previous chapter, in keeping with the conventions by INSD per-capita consumption i s established by dividing household consumption by the number of household members. It is therefore implicitly assumed that larger households have the same per-capita consumption needs as smaller ones and that adults and children need to consume equal amounts to be equally well off. The coefficients o f the regressions reported here - 27 - would differ if a different assumption was adopted with larger households having a lesser tendency o f beingpoor and smaller declines inwell-being for additional children. 71. Poverty declines strongly with higher education levels o f the household head, and education levels of the spouse inrural areas. Gains from education for the head are substantial, especially in terms of per capita consumption and assets. A household with a head having completed primary school has an expected level o f consumption that is approximately 13 percent higher in urban and 22 percent in rural areas than that o f a household with an uneducated head. The gains from secondary and tertiary education are even higher. Again, these findings are qualitatively consistent with previous work. 72. Occupation and the sector of activity for householdheads are important correlates for the household consumption and the income per capita as well as the asset score. According to our estimates, urbdrural households having a head with ajob in the services sector have levels o f expected per capita expenditure, income and asset score that are larger by 31, 68 and 38 percent inurban areas and 44, 63 and 28 percent inrural areas than those of households having a head working in agriculture. Household heads inthe industry sector also systematically do better than heads inagriculture but slightly less so than householdheads inservices. 73. As regards gender and poverty, results reported in the annex suggest that households having a female head are associated with significantly lower levels o f per capita consumption and income and asset score. For instance, in rural areas, per capita expenditure is lower by 21 percent and per capita income lower by 34 percent. The effect is even higher in urban areas, Since households headed by females tend to be smaller, the differences in well-being between female headed and male headed households may be even larger ifhousehold economies o f scale for consumption were assumed. 74. Household geographic location explains partly the income and expenditure per capita differences. All variables are measured against the omitted region Hauts Bassin. Controlling for other variables, it appears from the estimation, that the regions Est, Centre Nord, Centre Ouest and Centre have higher per-capita expenditure levels than Hauts Bassin, whereas the regions Mouhoun, Sud Ouest, Plateau central, Nord, Centre Est and Sud have lower levels o f per-capita expenditure. These results are in line with Lachaud (2003), who also uses the 12 regions for his rural regressions. These results regarding the regional differences are little surprising as far as the central and surrounding regions (Centre Nord, Centre Ouest) are concerned since they are enclosing the capital city Ouagadougou. The significantly higher expenditure in the Eastem region (Est) is more surprising. 75. Migration of the head i s associated with higher levels of per capita consumption only if the migration duration is short. A householdhaving a head who migrated for a period o f less than 6 months has levels o f per capita consumption, income and asset score that are significantly higher than households who have not migrated. When the absence is between 6 and 12months, expected per capita consumption, income and asset score are not higher. A household member living in C6te d'Ivoire tends to positively affect expenditure and income levels o f urban households, but not rural households. - 28 - 76. As regards income transfers, notably worker remittances, Lachaud reports that these transfers significantly reduce poverty among poorest households in rural areas in 2003. Using a specific module of the 2003 survey Siaens and Wodon (2004) also find that the Ivorian crisis significantly affected income o f previous beneficiaries but that their expenditure levels were affected to a much lesser extent as households engaged in coping strategies." The poverty level o f those households permanently losing remittances from relatives because they were unable to continue transferring money or returnedfrom Cbte d'Ivoire may therefore rise inthe future once their possibilities to smooth consumptionstreams have been exhausted. Remittances were mostly used for food, health, schooling, andconstruction. 77. Land ownership is positively correlated with expenditure per capita. Landownership (measured in hectares) strongly increases per-capita expenditure and income in both urban and rural areas as households are able to produce food for their own consumption and sale on the market. Moreover, Bardasi and Wodon (2004) show that among farmers, those who did not diversify their crops had lower expected incomes andfaced increased risks. 78. Poverty determinants give indications for formulating policies supporting a participation o f the poor inthe country's economic growth. Overall, given the stability o f the characteristics of the population, little o f the poverty dynamics over a short horizon would be expected to come from changes in the above-mentioned key determinants and most o f it would come from rising incomes. However, poverty determinants inform about possible reasons for the structural impediments to reducing poverty. They thereby give indications how future growth might be channeled to the poorest segments of the population by launching policy initiatives that remove impediments for increasing household income. Increasing the education levels, improving balance in regional growth and income opportunities, increasing dynamism o f modem sectors, assuring fair access to land, and educating families to space births are possible policy interventions inthat direction. C. POVERTYDYNAMICS 79. As outlined inthe second chapter, based on consistent measures o f poverty, poverty rates in Burkina Faso declined by more than 8 percentage points during 1998-2003. The growth experience was broadly in line with GDP growth and growth o f private consumption (Figure 6). However, even though sectoral GDP growth rates are already indicative, it is instructive to review the changes in poverty by socio-economic groups to more closely align growth and poverty dynamics. 80. The decline o f poverty rates during 1998-2003 can be largely attributed to the decline in rural poverty rates (Table 14). In presenting these figures, this study slightly modifies the original data in order to present a consistent set o f data for historical developments and the projections presented in the following chapter. Notably, while maintaining the same national poverty line resulting in a headcount index o f 46.4 percent in 2003, the study uses slightly different poverty lines for urban and rural areas to reflect the different cost o f living and assigns " Out of about 8,100 households, 494 report having a household member inCBte d'Ivoire; 1074 households report receiving remittances; and 375 households received remittances during the 30 days prior to the survey. - 29 - certain small socio-economic groups to urban and rural areas to avoid large fluctuations intheir poverty rates in simulations performed for the next chapter.I2 These adjustments o f the data would tend to result in somewhat higher urban poverty rates and somewhat lower rural poverty rates for both 1998 and 2003 compared with the unadjusted data, and they slightly alter the shares o f urban and rural population shares. Overall, using the adjusted survey figures, rural poverty headcount decreased from 62.2 percent to 52.7 percent, whereas urbanpoverty rates only fell marginally. This finding i s consistent with the GDP growth figures showing the strongest growth in rural output. Moreover, as discussed in more detail below, the population shift from rural to urbanareas accelerated the decline inrural poverty rates and slowed the decline inurban poverty rates. Figure 6. GDP Growth and Poverty Rates, 1998-2003 1 56 r p o v e r t y ratio (ri ht scale) 2 54 135 52 115 50 95 48 75 46 55 44 35 42 1998:Act. 1999: Est. 2000: Est. 2001: Est. 2002: Est. 2003: Act. Source: Burkinabhauthoritiesand IMF (nationalaccounts), andWorld Bank estimates and simulations. 81. Inthe rural sector, it is apparent that the largest groups benefited from a strong decline in poverty headcounts following the expansion in cereal and cotton output during 1998-2003. The break-down o f the population into nine socio-economic groups (SEG) allows analyzing more closely the differential impact o f economic developments on poverty rates. The SEG producing non-tradables in the agricultural sector most closely reflects subsistence farmers, and the 8 percentage point decline in poverty rates (increase in per-capita household expenditure) aligns well with GDP growth figures and cereal production. Moreover, the poverty rates in this group were likely reduced by population shifts either to agricultural tradables or urban SEGs. - l2The national poverty line for the consistent poverty measure is CFAF 72,110. The rural poverty line is CFAF 71,737 and the urban poverty line is CFAF 73,557. We also assume that all family helpers are living in the rural area, and all public sector households are living inthe urban area, - 3 0 - Table 14. Socio-Economic Groupsand Poverty, 1998 and 2003 (Inpercent) ShareofPopulation Share of Poor PovertyHeadcount 1998 2003 1998 2003 1998 2003 Ruralarea. 86.3 79.5 94.1 91.0 62.2 52.7 ULk!L?rea 13. 20.5 5. 4~ - -9.~0 ~ ~21.1 ~------------ - ~ ~ ~ ~ ~20.9-. Public sector (Urban) 4.1 3.6 0.7 0.3 9.1 3.4 Agricultural tradable (Rural) 16.8 18.3 16.4 18.6 53.1 47.1 Other agriculturalnon-tradable(Rural) 65.3 59.6 74.6 71.0 61.8 55.3 Familyhelpersandothers (Rural) 0.6 0.7 0.3 0.6 30.3 42.0 Non labor force (Rural) 3.6 1.o 3.3 0.8 50.4 38.8 Privateformal tradable (Urban) 1.c 0.8 0.1 0.1 8.1 7.7 Privateformal non-tradable(Urban) 1.9 2.6 1.2 0.9 33.3 15.7 Informal (Urban) 5.6 7.4 2.3 3.4 22.8 21.5 Unemployed(Urban) 1.1 6.0 1.o 4.3 47.8 33.1 ource: World Bank estimates. 82. As regards production o f agricultural tradables, which most closely correspond to producers o f cotton, their per-capita household expenditure also expanded strongly, and their poverty rates declined by 6 percentage points, in line with the increase o f the volume and value o f cotton production in 2001 and 2002. The shift o f population into this group, reflecting the "newcomers" among cotton farmers, slowed the decline in poverty rates o f this subgroup and increased the overall contribution o f this subgroup to the poor despite rising incomes. It likely also contributed to the increase in inequality in rural areas (see the Gini index in Table 8) as incomes o f existing cotton farmers rose faster than those for other farmers and newly arriving cotton farmers. Also significant i s the shift o f a large part o f the population in rural areas out o f the "non labor-force group,'' which contributed to the strong decline inrural poverty rates. Rural incomes and expenditure developments also reflect some o f the growth in the tertiary sector, notably trade, transport, and other services categories. 83. Inurban areas, most of the positive impact from growth and declining inequality was overwhelmed by a shift o f the population into the unemployed and informal groups. Declines in poverty rates among these latter two groups were not sufficient to overcome the increase in the number o f people inthese categories, raising their contribution to the poor. By contrast, poverty rates for the non-tradable and unskilled in the formal sector declined substantially. Overall, the picture o f poverty dynamics by SEG described above suggests that the urban tradable sector expanded little and contributed little to growth and income creation during 1998-2003. By contrast, growth in secondary and tertiary sectors, notably manufacturing and other services, reduced poverty rates in the formal non-tradable SEG, and growth accompanied by a decline in inequality managed to absorb an additional 2 percent o f the population in the informal sector without increasing poverty rates o f this SEG. This finding is in line with the strong presence o f informal sector activity among some o f the fastest growing subsectors (food, construction, and - 31 - other services). However, the unemployed urban population grew strongly, making the unemployed the largest contributor to the poor from urban areas in2003. 84. Another possible link between growth and the household survey is explored by G r i m and Giinter (2004), who juxtapose growth in per-capita expenditure of sector o f activity of the household head and growth in these sectors obtained from the household survey's labor m0du1e.l~However, data problems render the results o f this exercise difficult to interpret. Since the information on sector of activity is only available for the household head, it misses important information on other household members and seems to lead to an underestimation of certain employment categories, possibly by missing out systematically on informal activities. For example, although more than 12 percent o f GDP is estimated to be contributed by the manufacturing sector, only 1.7 percent o f household heads are identified as being in this sector. Even assuming that productivity is much higher in manufacturing than in agriculture, the share of households attributed to manufacturing through the labor module seems implausibly l o w since the sector is heavily dominated by food and textile production, much o f which i s informal and involves female labor. D. GROWTH, POVERTY, AND INEQUALITY 85. Inorder to further formalize and analyze some of the findings, it is useful to perform a number o f decompositions o f poverty data. The technical details o f these methods are described inAnnex 4. Inorder to avoid assuming a density function o fhouseholdexpenditure, most o fthe results focus on changes in the poverty gap-the average percentage distance of the expenditure o f the poor from the poverty line-and the square poverty gap (or poverty severity)--the average squared percentage distance of the expenditure o f the poor from the poverty line.14 Poverty gap measures always move inthe same direction as the poverty headcount. The squared poverty gap gives higher weight to the poorest households than the poverty gap. 86. One o f the key questions i s to what extent growth increases the incomes o f those below the poverty line and to what extent it benefits better-off households and increases inequality. Based on 1998 survey data, Table 15 shows the tradeoff between inequality and growth by presentingthe growth and inequality elasticities o f the poverty gap and the square poverty gap, along with the inequality-growth trade-off index o f Kakwani (2000). The growth elasticities represent the percentage change o f the relevant poverty measure for a 1 percent growth without changes in inequality. The inequality elasticities represent the percentage change for the poverty measure in relation to a 1 percent increase in the Gini index. The inequality-growth trade-off index by Kakwani shows how much growth (inpercent) i s needed to keep poverty constant ifthe Gini index rises by 1percent. 87. Growth significantly reduces the average distance between the expenditure of the poor and the poverty line. On the national level and among all poor, the pure growth elasticity for the poverty gap was larger than -2.0, thus a 1 percent GDP growth leads to a reduction in poverty 13 See Grimmand Giinter (2004), Table 6. l4 Defining a density function for household expenditure would be arbitrary and make the results dependent on the density assumptions. Since a decline in the poverty gap is necessary and sufficient for a decline in the poverty headcount, the poverty gap measures are perfect indicators for movements inthe headcount with the added benefit o f not being arbitrary. - 32 - gap by more than 2 percent if there are no changes to inequality. A similar result holds for the severity o f poverty (square poverty gap), which declines by about 2.5 percent for each percent in growth. As regards inequality, the data indicate that a 1 percent increase in the Gini index increases the poverty gap by more than 3 percent. An even stronger elasticity holds for the severity o fpoverty given that this measure gives higher weights to poorer households. 88. On the national level and among all poor, the trade-off between growth and inequality in Burkina Faso implies that a 1percent increase inthe Gini index needs to be compensated by 1.4- 1.7 percent o f additional growth to keep the poor at the same distance from the poverty line. Similarly, 2-2.5 percent of growth are needed to compensate the impact o f a 1percent increase in the Gini index on the square poverty gap. Compared to other countries presented by Kakwani (2000), the growth-inequality trade-off index appears moderate, implying that growth can compensate more easily for changes in inequality than elsewhere. Kakwani (2000) reports, for example, that in 1998 the growth-inequality trade-off was 5.1 for Thailand, 3.1 for the Philippines, 1.7 for Korea, and 1.3 for Laos PDR. Table 15. Growth and Inequality Elasticity for the Poor and Ultra-poor Poverty Poverty Gap Ratio 1 Severity of Poverty Ratio 3eadcount Elasticit Elasticit Tradeoff Elasticit Elasticit Tradeoff Unit Record1998 Poor National 54.( -2.19 3.18 1.56 -2.49 5.22 2.0! Rural 62.: -2.03 2.40 1.18 -2.49 4.07 1.61 Urban 21. -2.27 7.52 3.3 1 -2.66 11.29 4.2~ Ultra-Poor I National 38. -2.58 5.10 1.98 -2.91 7.63 2.6. Rural 42. -2.58 3.96 1.54 -2.90 6.06 2.0' Urban 13.: -2.56 10.75 4.20 -3.06 15.87 5.1, Unit Record2003 Poor National 46.4 -2.03 2.80 1.38 -2.50 4.67 1.8 Rural 52.' -2.02 1.71 0.85 -2.5 1 3.05 1.2 Urban 20.i -2.20 7.36 3.34 -2.44 10.81 4.4 Ultra-Poor National 32.t -2.54 5.89 2.32 -3.00 8.91 2.97 Rural 37.' -2.55 4.67 1.83 -3.04 7.21 2.37 Urban 13. -2.46 10.461 4.251 -2.60 14.571 5.6d Source: World Bank calculations - 33 - 89. Growth-inequality trade-offs play a more important role for the urbanpopulation and the poorest, posing additional challenges for policymakers. In calculating separate trade-off indexes for the ultra-poor, the study follows Kakwani(2000) indefining those with expenditure below 80 percent o f the poverty line as ultra-poor. The results in Table 15 indicate that about one-third of the population has per-capita expenditure below 80 percent o f the poverty line, also indicating the depth o f poverty inBurkina Faso. The growth-inequality trade-off is uniformly higher inthe urbanareas and for the ultra-poor thanfor the average population. Given the size o fthe trade-off inurbanareas, where 4-6 percent ofgrowth areneededto compensate for a 1percent increase in the Gini index, policymakers will need to give particular attention to the inequality aspect o f growth in urban areas since the impact o f growth on poverty can be significantly reduced by rising inequality. Secondly, the higher growth-inequality trade-off among the ultra-poor also calls for policies that support growth and limit changes in inequality at the bottom end o f the income distribution to best reap the benefits o f growth. 90. Another decomposition o f poverty data following Son (2004) shows that the shift in population to urban areas has been pro-poor (Table 16). On average on the national level, both the poverty gap and the squared poverty gap declined by about 3 percent per year during 1998- 2003. These gains were achieved through growth and population shifts. Whereas growth affects were diminished by inequality effects, the shift o f the population from poorer to relatively richer areas helped inreducing the poverty gap and the severity of poverty. Infact, the population shift contributed about half to the net effect inthe decline o f the poverty gap between 1998 and 2003. On net, therefore, the rural-urban migration has had positive effects on poverty thus far as the urban areas have been able to absorb the larger population without raising poverty levels, and overall the gap between the expenditure of the poor and the poverty line narrowed. Table 16. Poverty Decompositionwith Population Shift Effect (Percentage change per year) Square Poverty PovertyGap 1J Gap 1/ ____-___________________________________-----------~-----~--- Growth Effects -3.87 -5.20 InequalityEffects 3.37 PopulationShift Effects -2.39o 1s -1.18 Total effect -2.97 -3.00 Source: World Bank calculations. I/Theindicatorsarecomputedusingthepovertygapandthesquare povertygap of 1998 and2003. - 34 - E. WAS GROWTH PRO-POOR? 91. An important question for policymakers is to what extent growth overall has been pro- poor. Several approaches have been devised to measure the extent to which the poor benefited from growth. This study adopts two different approaches to measuring pro-poor growth: Chen and Ravallion's (2003) growth incidence curves (GIC), and Kakwani and Pernia's (2000) poverty-equivalent growth rate measures. Under the first concept, growth is defined as pro-poor iftheincomeofthepoorriseswithrisingoverallincome,i.e.povertydeclines.Underthesecond concept, growth is classified in different groups according to the accompanying changes in inequality. The second concept is stricter and not all cases defined as pro-poor in the first concept qualify as pro-poor inthe second case. 92. Using the GIC approach (see Box 6), it can be shown that growth has been pro-poor in the sense that all households benefited fiom rising expenditure levels. As shown in Figure 7, all households, grouped by their household expenditure in 1998, had higher expenditures in 2003. However, inrural areas, expenditure growth was above average for most of the households with the highest expenditure levels, whereas expenditure growth rates for 80 percent o f the households at the low end o f the expenditure distribution was generally below average. Inurban areas, the finding i s opposite with fastest growing expenditures among households with the lowest expenditure levels. On the national level, the GIC is dominated by the findings for rural households, which are by far the largest share o f the population. Box 6:.The Growth IncidenceCurveor GIC Ravallionand Chen (2003) propose to call growth pro-poor ifthe poverty measureo f interest falls. The basis o f their measure starts from the changes inincome or expenditure o f the poor using the cumulative distribution function o f income or expenditure. To measure poverty, they apply the Watt's index, thereby avoiding problems with poverty headcount measures. Their pro-poor growth measure can also be derived fiom the "Growth Incidence Curve" which plots the growth rate o f income or expenditure of the pth quintiles and thus describe how the gains from growth i s distributed. 93. The disaggregated GIC findings support the previous finding o f a positive population shift effect on poverty and the summary poverty and inequality measures reported in Table 8. The urban GIC shows signs o f integration o f poorer households into the urban setting without increase inpoverty and a decline in inequality. Inthe rural area, the GIC is consistent with slight increase in inequality and higher expenditure growth among households with higher expenditure in 1998, reflecting the fact that the growing incomes from rising cotton production accrue to farmers that already tend to have larger incomes than average farming households. 94. The pro-poor growth index (PPGI) and poverty equivalent growth rate (PEGR) approaches provide additional insights to the pro-poor growth analysis. The PPGI decomposes the poverty elasticity for the poverty gap and the square poverty gap into growth and distributional components as suggested by Kakwani and Pemia (2000) and measures to what extent growth or inequality effects dominate the growth process. The PEGR indicates the growth rate that would result in identical reductions in poverty gap or square poverty gap as actual growth ifgrowth didnot lead changes ininequality (see also Box 7). - 35 - Figure7. Growth-IncidenceCurves, 1998and 2003 (Percentage change inexpenditure by household, ranked by household expenditure in 1998) 5.0 - National, 1998 & 2003 4.0 - - -? Change in mean expenditure per household 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 -- Urban, 1998 & 2003 6.0- _________--I__ 5.0 - 4.0 - 3.0 - 2.0 - 1.0 - Change in mean expenditure per household 0.0 fl, I , 1 1 I , , 8 ) I 8I ' 3 8I I , I ,rI , ! , I I I I I 1 1 8 8 8I I I' 81 ' , I a ' t' 8 ' ''rr"I , ,' I a" ' 1 1I" "" "" ' ' ' I " ' 1 ''"' 8 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 - 36 - Box 7: Pro-Poor Growth Index and Poverty Equivalent Growth Rate Pro-Poor Growth Index The approach used by Kakwani and Pernia (2000) is to decompose the elasticity (77) o f the poverty gap or the square poverty gap into growth (qg)and income ( V I ) effects between two discrete periods, starting from the basic definition o f an income-based measure o f poverty. 77 is equal to qg+qI .The qgis the pure growth effect, and is the percent change inpoverty when the distribution of income is constant; the qI is the inequality effect, and is the percent change in poverty when inequality change in the absence o f growth. Kakwani and Pemia then defined +-=L ` `I-k `I=1 -as thepro-poor growth index, andthey statedthat (1) when + qI 4 + 1,growth is highly ??g 7, q g pro-poor ; (2) when 0 4 4 1, growth is not strictly pro-poor (i.e., growth may be associated with a substantial 4 positive inequality elasticity) with growth being characterized as pro-poor for a PPGI above 0.66, moderately pro- poor for a PPGIbetween 0.33 and 0.66, and growth weakly pro-poor for a PPGI between 0 and 0.33; when 4 4 0, growth is said not pro-poor. Kakwani and Pernia's formulation does provide easily interpretable results for the episodes when economic recession i s accompanied by poverty declines. Poverty Equivalent Growth Rate While the GIC relies on the first order dominance, the PEGR is basedon the second order dominance. "The PEGR is the growth rate that would result inthe same level o fpoverty reduction as the actual growth rate if the growthprocess was not accompanied by any change ininequality." If y is the growthrateofincomeor expenditure percapita over two periods, recallingthat growth, the PEGR is defined as y' = y4. When y* +y ,then growth is pro-poor; if0+4y* +y is the index o fpro-poor ,thengrowthis trickle-down; and if y + 0 and y* 4 0then growth is immeserizing. 95. Overall, referring to Kakwani and Pemia's criteria, growth between 1998 and 2003 was moderately pro-poor in Burkina Faso (Table 17). On the national level, the growth elasticity o f about 2 for the poverty gap is offset by an inequality elasticity o f 0.7, resulting in an overall elasticity o f -1.34, that is the poverty gap declines by about 1.3 percent for each additional percent o f growth. The resulting PPGI is 0.66 and places Burkina Faso at the border between moderately pro-poor growth and pro-poor growth. This PPGI translates into a PEGR o f 1.5 indicating that instead o f a per-capita income growth o f 2.3 percent observed between 1998 and 2003, a growth rate o f 1.5 percent would have been needed to achieve the same decline in the poverty gap without any change in inequality. Following Son (2004), this performance qualifies as trickle-down growth as the poor benefit from growth but at the same time higher-income households take some o f the gains through rising inequality. Compared with measures derived for the poverty gap in almost all cases the measures for the square poverty gap show somewhat weaker indicators, however without overturning the conclusions. 96. Significant differences persist between rural and urban areas as regards the distribution o f growth. Sensitivity o f poverty measures to rising inequality is significantly larger in urban than inrural areas. As a result the PPGI for urban areas signifies only weakly pro-poor growth, and, - 37 - although growth i s still trickle down, the PEGR is close to zero as inequality changes have the potential to quickly absorb any positive effects of growth on the poverty gap or the square poverty gap. Rural figures, by contrast, are close to national averages given the predominant weight of the ruralpopulation. 97. Inbroad terms, the pro-poor growth analysis confirms that under a variety of definitions growth was pro-poor, at least moderately so. In rural areas some of the bigger gains went to households with higher incomes, reducing some of the benefits o f growth, but overall the impact o f growth on poverty reduction is less sensitive to inequality changes. Inurban areas, the growth impact is weaker and poverty would have increased were it not for the decrease in urban inequality. There i s a significant danger that inthe hture risingurbaninequality could absorb the h i t s o f growth because inequality elasticities are high.Both findings give some basic directions for the design o f policies supporting growth in Burkina Faso-the importance o f rural growth policies oriented toward creating opportunities for subsistence agriculture and urban growth policies limiting inequality of incomes by creating opportunities for those currently in the informal sector or amongthe unemployed. Table 17. Pro-Poor Growth Index and Poverty Equivalent Growth Rate, 1998-2003 F. CONCLUSIONSAND RECOMMENDATIONS 98. Growth in Burkina Faso during 1998-2003 has been driven by a strong expansion o f the primary sector, mostly cereals and cotton, with the secondary and tertiary sector growing at a slower pace. Secondary and tertiary sector growth was strongly linked to informal sector activities. Overall growth in real pre-capita consumption implied by the national accounts of about 3 percent was somewhat higher than the growth inper-capita expenditure on the household level of2.3 percent over the same period. Nonetheless, the data are sufficiently close to have some comfort in establishing a link between the macroeconomic developments and those at the household level. - 38 - 99. As regards the determinants of poverty, the work undertaken in recent years with a variety o f surveys comes to similar conclusions. Poverty is closely related to household sizes, education and literacy, location, employment pattems, migration and transfers, and land ownership. 100. The analysis o f poverty dynamics and pro-poor growth measures reveals that all groups have benefited from growth but that growth was only moderately pro-poor according to the classification by Kakwani and Pemia. Growth benefited producers o f cereals but expanding production o f cotton seems to have boosted the incomes o fthe higher-income households inrural areas more than those for lower-income households, contributing to rising inequality in rural areas. Urban areas successfully managed to absorb an influx o f additional population through higher growth among some lower-income households, which overall prevented urban poverty rates from rising. However, there was also a significant increase in the unemployed urban population. Overall growth inrural areas i s less subject to changes ininequality and poverty gap measures show moderately pro-poor characteristics with growth elasticities that are substantially largerthan inequality elasticities. By contrast, inthe urbanareas growth is only weakly pro-poor as inequality elasticities for the poverty gap and the square poverty gap are relatively large, indicating large benefits from pursuingpolicies that limit urbaninequality. 101. The following recommendations canbe made based on the findings o fthis chapter: 0 Improve the timeliness and production o f national account data to allow following growth trends andthe composition o f GDP on a timelier basis with fewer revisions. 0 Design surveys to follow informal sector activity to better measure economic growth inurbanareas. 0 Based on a more detailed study in chapter 5 below, review medium-term policies for education and social protection in light o f poverty correlates to support policies that improve poverty outcomes by removinglong-runstructural impediments. 0 Recognizing the opportunities and limits o f development o f cotton production, target growth-supporting policies in the rural areas to activities that could draw more subsistence farmers into market-based activities (see chapter 4). 0 Targkt growth-supporting policies inthe urban areas to activities benefiting people in the informal sector and the unemployed to avoid sharp increases in inequality during the growth process. - 39 - IV. THE QUESTFORSUSTAINEDEQUITABLE GROWTH 102. The previous chapters discussed the measurement o f poverty, and how growth and changes in income distribution contributed to poverty reduction in 1998-2003. These findings give indications for areas that government policies could usefully target under the first and third PRSP pillars. This chapter adds additional information on the link between future growth and poverty reduction byperforming simulations with the poverty analysis macroeconomic simulator (PAMS). The simulations allow gaugingthe poverty-impact o f different growth paths and shocks with a view toward attaining the government's objective to reduce the poverty headcount to 30 percent by 2015. A. PROJECTINGPOVERTY DEVELOPMENTS PAMSWITH 103. The PAMS model, developed by Pereira da Silva, Essama-Nssah and SamakC (2002), simulates income changes o f a variety of representative households for any given change in aggregate economic output. To do so, PAMS decomposes aggregate income changes into income changes o f a number o f representative households on the basis o f overall income growth and changes in relative incomes between households representing nine different socioeconomic groups drawn from definitions o f the household survey (see Table 14 in the previous chapter). The version o f the model for Burkina Faso has been developed in close collaboration with the government and the local GTZ advisory project. 104. PAMS projects poverty developments based on any standard consistent macroeconomic model. In the case o f the Burkina Faso model, PAMS can be directly connected to either a standard World Bank RMSM-X model or the macroeconomic projection tool o f the government, called IAP. The macroeconomic model projections are then translated into income changes for a set o f representative households for different labor categories. These representative households reproduce the original poverty headcount and income distribution. To translate GDP growth in income changes, PAMS simulates labor demand and supply for each representative group, determines its wage and ensures that these prices feed through to the production level. It then takes into account the income tax rates and transfers to determine income growth by group and representative agents. In the final step, this information can be used to derive the poverty headcount and determine the level o f inequality. 105. The PAMS model has two shortcomings that should be kept in mind when analyzing its output. First, it i s linked to a macro-consistent model and therefore the quality o f the results will depend on the quality of the model and the macroeconomic variables it produces. Second, the model depends on assumptions regarding the distribution o f income within representative groups and the elasticity o f labor supply and demand used to generate the changes inlabor incomes. B. THEBASELINEGROWTHSCENARIO 106. The baseline growth scenario broadly follows scenarios that have been used for long-term projections in the past by the World Bank and the International Monetary Fund, such as the projections for the Heavily Indebted Poor Country (HIPC) Initiative. It builds on an average - 40 - long-run GDP growth rate o f 5 percent driven by secondary and tertiary sector growth of 6-7 percent and primary sector growth o f 4 percent (Table 18). Inflation i s projected to remain low, at 2 percent per year. Fiscal revenue would increase over time to about 24 percent o f GDP (reducing growth of disposable household income ), assuring a decline in the country's aid dependency over time, and fiscal spending as share o f GDP would increase moderately to allow increased spending, notably on education. Export growth would slightly outpace real GDP growth, implyinga moderate increase inthe export-to-GDP ratio. Table 18. Baseline Macroeconomic Framework and Poverty Response 2003 2004 2005 2006 2001 2008 2009 2010 2011 2012 2013 2014 2015 Act. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Selected macro indicators RealGDP growth 11 8.0 4.8 5.3 5.2 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Primarysector 11 10.8 1.8 4.5 4.5 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Secondarysector 1/ 10.4 6.3 6.7 6.6 6.6 6.6 6.0 6.6 6.6 6.6 6.1 6.7 6.1 Tertiarysector 11 5.5 6.1 5.3 6.8 6.8 6.8 6.5 6.5 6.5 6.5 6.5 6.5 6.5 Fiscalrevenue21 11.3 12.0 12.5 13.0 13.5 11.7 18.6 19.5 20.5 21.6 22.1 23.9 23.9 Public expenditure21 22.0 22.5 22.7 22.9 23.6 24.1 25.0 25.0 25.0 25.0 25.0 25.0 25.0 Exportsof goods 11 10.7 15.8 16.4 8.5 6.3 5.5 5.6 6.0 6.0 6.1 6.1 6.2 6.3 CPI(percentagechange) 2.1 2.2 2.2 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 PovertyIncidence National 46.4 44.1 42.4 40.3 39.2 38.0 36.0 34.8 33.8 32.4 30.7 29.7 28.1 Rural 53.1 51.4 49.6 47.6 46.6 45.3 42.9 41.8 40.7 39.1 31.1 36.1 35.0 Urban 20.5 19.7 11.9 16.0 15.4 13.6 13.3 12.1 12.4 12.1 11.8 11.3 11.0 Demographicstructure Annual growthrate 31 National 5.1 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 Rural 4.6 1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 Urban 7.8 4.3 4.3 4.2 4.2 4.1 4.1 4.1 4.0 4.0 4.0 3.9 3.9 Share of population Rural 19.5 19.1 18.8 18.4 18.0 71.6 71.2 16.9 16.5 16.1 75.1 15.4 15.0 Urban 20.5 20.9 21.2 21.6 22.0 22.4 22.8 23.1 23.5 23.9 24.3 24.6 25.0 Source: World Bank projections with PAMS. 11 Real growth rate in percent. 2/ In percento fGDP. 31The growth rate for 2003 reflects 300,000 total displaced from CBted'ivoire. 107. The baseline scenario builds on a certain number o f assumptions. First, primary sector growth, although assumed to be lower than in the past few years on average, needs to be driven by either a further increase in cultivated land or an increase in agricultural productivity over time. Second, secondary and tertiary sector growth o f 6-7 percent will not be possible without growth o f private investment, improvements inskill levels, and satisfactory infrastructure. Third, the projections exclude any shocks to GDP growth from rainfall patterns or terms of trade, thus ignoring the vulnerability o f Burkina Faso to these factors. Finally, the baseline projections build on modest shift of population from rural to urban areas, with an increase of urban population from 20 to 25 percent by 2015. - 41 - 108. Under the baseline scenario, the poverty incidence i s expected to decrease substantially from 46.4 in 2003 to 28.7 percent by 2015. The poverty headcount thus declines by about 4 percent per year with per-capita income growing at around 2.5 percent, implying a poverty- growth elasticity of 1.5-1.6. By contrast, with about the same rise inper-capita income between 1998 and 2003, poverty declined by about 3 percent per year. The per annum decrease in the rural areas is projected to accelerate slightly from 3.1 percent between 1998 and 2003 to 3.4 percent, while inthe urban areas it is projected to increase sharply from 0.2 to 5 percent over the respective two periods. These projections reflect that under the baseline projections secondary and tertiary sectors are the main engines o f growth and thus urb& labor demand grows more quickly than rural labor demand, representing an important structural shift from past growth experience. 109. The macroeconomic developments affect poverty rates through their effects on the changes inreal incomes per capita across SEGs (Table 19). These aggregate changes are derived from SEG specific income growth, which inturn reflect the links PAMS establishes between the structure o f GDP, labor demand, labor supply, and incomes. As can be seen from Table 19, the baseline structure of GDP growth is projected to benefit the informal sector, the formal tradable sector, and agricultural non-tradable, explaining the steady decline in poverty in rural areas and the decline in urban poverty that contrasts with past experience. Owing to readjustments o f labor supply over time linked to the differential growth rates for rural and urban areas, the income growth fluctuates despite stable GDP growth patterns inthe longrun. Table 19. InequalityandHouseholdExpenditureTrends 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 20I 4 2015 Act I/ Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. Proj. ~~ Inequality and Expenditure Gini Index (inter-group) 44.8 45.7 45.3 47.9 48.2 48.9 49.4 50.3 51.3 52.2 52.9 54.2 56.2 Average per capita income growth 2.3 1.8 3.1 3.4 1.7 2.1 2.9 2.7 2.3 2.3 3.2 2.5 2.1 Per-capita incomegrowth by SEG (in percent) 2/ Public sector, urban (3.6) 1.7 2.1 0.0 2.9 1.4 0.9 1.6 1.7 2.2 2.3 2.4 2.4 2.4 Non-tradable, rural (59.6) 2.7 1.1 3.3 3.2 1.1 1.6 3.2 2.6 2.3 2.6 3.9 2.5 2.0 Tradable, rural (18.3) 1.1 3.1 2.8 2.6 2.5 1.2 1.4 0.9 0.9 0.4 0.1 0.7 0.3 Family helpers, rural (0.7) 0.9 1.0 0.7 1.5 0.8 0.6 1.1 1.0 1.0 1.0 1.3 0.8 0.7 Formal, tradable, urban(0.8) 3.0 2.4 5.3 10.7 3.2 5.0 4.7 4.9 4.9 5.0 5.1 5.1 5.2 Formal, non-tradable, urban(2.6) 4.1 5.2 2.3 4.5 2.5 2.7 0.7 9.8 3.2 0.9 5.4 6.5 2.3 Informal, non-tradable,urban(7.4) 5.3 2.2 5.7 6.8 5.7 9.8 7.6 7.0 7.0 7.0 7.1 7.2 7.2 Unemployed (6.0) 0.1 3.6 0.7 2.5 0.5 0.4 0.6 1.8 1.3 0.1 1.3 1.3 1.3 Non labor force, rural (1.0) 0.4 0.4 0.2 0.6 0.3 0.2 0.3 . 0.3 0.4 0.5 0.5 0.5 0.5 Sources: World Bank projectionswith PAMS. 1 / The percentagechanges in this column indicatethe per annumchanges between 1998 and 2003 21 The numbersin parenthesisindicate the share of each groupto the total populationin 2003. 110. The baseline scenario implies rising growth and inequality elasticities o f the poverty gap and the square poverty gap over time, deepening the growth-inequality trade-off (Table 20). The pure effect of economic growth on the poverty gap will accelerate from -2.0 in2003 to -2.6 by 2015. However, along with the rising growth elasticity, the inequality elasticity rises sharply from 2.8 to 6.6. The substantial contribution of secondary and tertiary sectors and their strong contribution to urban incomes leads to these developments since, as noted above, the urban sector had already much higher inequality elasticities in 2003. Thus, as more income i s channeled to urban areas, inequality rises and the growth-inequality trade-off becomes more important. The rising inequality also reduces the growth elasticity of the square poverty gap, which places a higher weight on unequal income distributions. \ Table20. PovertyElasticityDecomposition,2004-2015 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Poverty Gap Growth elasticity -2.0 -2.1 -2.1 -2.2 -2.2 -2.2 -2.3 -2.3 -2.4 -2.4 -2.5 -2.6 -2.6 Inequality elasticity 2.8 3.1 3.3 3.6 3.8 4.1 4.3 4.6 4.9 5.2 5.6 6.1 6.6 Inequality/Growth Tradeoff 1.4 1.5 1.6 1.6 1.7 1.8 1.9 2.0 2.1 2.1 2.3 2.4 2.5 Square Poverty Gap Growth elasticity -2.5 -2.4 -2.4 -2.4 -2.3 -2.2 -2.2 -2.2 -2.1 -2.0 -2.0 -1.9 -1.8 Inequality elasticity 4.1 5.0 5.3 5.6 5.8 6.0 6.2 6.5 6.7 6.9 1.3 7.6 8.0 InequalityiGrowth Tradeoff 1.9 2.1 2.2 2.3 2.5 2.7 2.8 3.0 3.2 3.4 3.7 4.0 4.3 Source: World Bank projectionswith PAMS. 111. Turning to the pro-poor growth measures presented above, the baseline scenario does not fundamentally change the fact that growth is weakly to moderately pro-poor (Table 21). Subject to some fluctuations on an annual basis given the volatility o f income growth in the PAMS projections, growth on the national level remains moderately or weakly pro-poor. However, with rising inequality over time and with urban incomes rising faster than rural incomes, there is a tendency for a decline in the pro-poor growth index and the poverty-equivalent growth rate. Growth inurban areas continues to be substantially less pro-poor than inrural areas. 112. The baseline economic scenario would allow the government to achieve its objective o f 30 percent for the poverty headcount in2015 1994 (see Ministry o f Economy and Development and UNDP, 2003). Under the baseline scenario, the poverty headcount is projected to decrease fiom 46.4 percent in 2003 to 28.3 percent in 2015. As regards the MDG target, the government previously considered a poverty headcount of about 23 percent in 2015 as being consistent with the MDG. However, the discussion in chapter 2 shows that a consistent MDG target would be between 25 and 30 percent and thus Burkina Faso would be on track for the income-poverty MDGunder the baseline scenario (see also Box 8). -43 - ~~ Box 8: Government Poverty Objectives and the MDG The government has established its own objective for the poverty headcount o f 30 percent by 2015 as well as presented calculations for the income-poverty objective o f 23.2 percent it considers consistent with the respective MDG.The MDGaims at reducing the poverty headcount by 50 percent between 1990and 2015 based on a measure o f US$l per day inpurchasing-power parity (PPP) terms. Since the 1990 poverty headcount is not available, usually the 1994 poverty headcount o f 44.5 percent is used as a starting point since the poverty line in 1994 was also close to a measure o f $1 per day in PPP terms. However, as pointed out in chapter 2, the 1994 poverty headcount is inconsistent with the 2003 headcount measure and a consistently measured poverty rate for 1994 would be likely be closer to 60 percent. An M D G target based on a consistent headcount measure for 1994 would therefore be higher than the MDG target o f 23 percent presented by the government. It i s therefore very likely that under the baseline scenario Burkina Faso would attain the MDGstarting from a consistently measuredheadcount in 1990. Table 21. Pro-poor Growth Index Dynamics 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Pro-Poor Growth Index Poverty Gap Ratio National 0.4 0.9 0.7 0.5 0.5 0.8 0.6 0.6 0.6 0.7 0.4 0.3 Rural 0.9 1.0 0.4 0.9 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Urban 0.9 0.8 0.7 0.5 0.4 0.2 0.6 0.1 0.1 0.4 0.3 0.3 Severity of Poverty Ratio National 0.3 0.7 0.6 0.3 0.3 0.7 0.5 0.5 0.5 0.7 0.4 0.3 Rural 0.5 0.8 0.8 0.5 0.6 0.9 0.9 0.9 0.9 1.1 1.0 0.9 Urban 0.8 0.4 0.5 0.0 0.2 0.8 0.3 1.9 1.0 0.1 0.1 0.0 Poverty Equivalent Growth Rat,e Poverty Gap Ratio National 1.0 5.5 2.1 2.2 1.6 2.5 2.0 2.5 1.7 2.3 1.9 1.6 Rural 2.4 3.5 1.3 1.0 2.1 3.7 3.2 3.0 3.3 4.2 2.9 2.4 Urban 10.4 1.6 4.3 2.4 1.9 0.9 2.9 0.3 0.6 2.0 2.2 2.1 Severity of Poverty Ratio National 0.6 4.3 1.7 1.2 0.9 2.2 1.8 1.9 1.5 2.4 1.9 1.3 Rural 1.3 3.0 2.9 0.5 1.3 3.5 2.9 2.6 3.1 4.7 2.9 2.2 Urban 8.6 0.8 2.8 0.1 0.9 2.6 1.5 5.9 3.9 0.3 0.5 0.3 Source: World Bank projections with PAMS. 113. The baseline scenario is subject to risks but also leaves opportunities for improving the quality o f growth. The-already optimistic-baseline scenario builds on (i) about 5 percent GDP growth per year driven by stronger secondary and tertiary sector growth; (ii) continued high population growth of 2.4 percent per year; and (iii) moderate population shift from rural to urban areas. It would reduce poverty considerably but also lead to a continuous increase in inequality, - 44 - rendering poverty reduction more and more difficult over time. Rural poverty would continue to determine the national trend as GDP growth in rural areas would be below average. In the following we pursue the question whether different scenarios can improve the quality o f growth andthus support a further acceleration o fthe trend observed under the baseline, andwhich types o f shocks could affect the growth andpoverty path. C. ALTERNATIVEGROWTH PATHS 114. The baseline scenario serves as a springboard to study different transmission channels that could improve the quality o f economic growth andpotentially accelerate poverty reduction. Utilizing the PAh4S framework, we conduct some specific simulation exercises to identify possible policy shocks. Ingeneral, government policies could aim at influencing the composition of GDP (for example, through targeted infrastructure investments), the transmission channels a between GDP growth and income (for example, through market regulation or fiscal policies) or enhance the chances o f the poorest and most vulnerable o f participating in the income creation process (for example, through social services and social protection). These channels are summarized in Figure 8. The simulations below are largely focused on variations to the economic weight. Figure 8. Economic Policies and Poverty Reduction 115. Annex 5 presents the summary macroeconomic statistics and poverty headcount for 6 different simulations. The simulations have been selected to illustrate the poverty and inequality impact o f specific variations to the macroeconomic environment and reflect also the behavior of the RMSM-X model used to generate the macroeconomic variables. Although these may not constitute realistic growth paths for Burkina Faso, they indicate how possible interventions by the government or through external events may influence the outcome o f the baseline scenario. - 45 - 116. Export quantity changes. Exports constituted roughly 9 percent of GDP in 2003, and more than 90 percent of exports were derived from agricultural products, namely cotton, livestock and meat, etc. The first simulation investigates how an increase in the level exports by 20 percent in 2004-leading to a permanently higher export level thereafter since subsequently growth o f output is based on this higher level-would influence incomes and poverty. The increase would boost GDP growth in2004 by about 4 percentage points. Figure9. Export Quantity Increase (Difference with baseline scenario) 6.0% 4.0% 2.0% 0 0% -- -2.0% 4 0% -6.0% 2033 2CC4 2- Mo6 2037 2008 2w9 2010 2011 2012 2013 2014 2015 1 . 1 1 1 1 .kp/lnc Gini-Total Kl 117. The results (Figure 9) indicate that the positive export shock leads to an increase in income and inequality, and to a decline inpoverty. The reduction in aggregate poverty from the boost in income to the rural producers o f tradables is important, at about 4 percentage points. Inequality is projected to increase by 9 percentage point over the simulation period. The main reason for this result is that the simulations implicitly assume the additional growth will boost the income o f households already employed in the tradable sector. However, these households make up for less than 20 percent o f the population and the rising income of this group representing less than 20 percent o f the population leads to a one-time decrease in aggregate poverty tends to raise inequality. The simulation underlines that export-led growth will need to be achieved though an expansion o f the tradable sector to those currently producing non- tradables to spread the wealth through evenly. 118. Agricultural production increase. As the previous simulation revealed, boosting the existing sector of tradables reduces poverty but the impact i s relatively narrow. In a second scenario, we increase the production o f the "non-tradable" agricultural sector that employs 60 percent o f the Burkinabb by 20 percent in 2004, again leading to a permanently higher level o f GDP since future growth rates are assumed to be the same as under the baseline. This production - 46 - increase could be alternatively interpreted as a development o f a sector o f tradable products by current subsistence farmers. The results o f the simulation are displayed inFigures 10 and 11. Figure10. IncreaseinAgriculturalProduction,PovertyImpact 50% 45% 40% 35% 30% 25% 20% 15% 10% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 1-- - -R)Act.(%)-BasePo1 Figure 11.AgriculturalProductionIncrease (Difference with baseline) ... - 2UU XIT 2025 x16 XO7 XOB 2039 2010 2011 2012 2M3 2014 ~,, . .Exp/lnc ~I Gini-Total FO ~ 119. Boosting agricultural output o f the large population group currently involved in subsistence agricultural and producing non-tradables can accelerate poverty reduction and reduce inequality compared with the baseline, consistent with Timmer (2003) andDorwad et al. (2004). This result follows from the predominant role of this group inthe population and its higher-than- average poverty rate. Inour scenario, the poverty headcount would be about 9 percentage points - 47 - lower in 2015 if agricultural production by the non-tradable SEG increased 20 percent. Clearly, the production increase would only be possible under certain conditions, such as the increase in agricultural productivity and better integration of the rural areas into the market, including external markets, to permit farmers to sell surplus production and convert additional output into higher incomes. 120. Increase in public wages. To illustrate the aggregatepoverty impact o f public wages, we assume a 4 percent real increase inthe public wages and salaries in2004 and then apply baseline growth rates (see Figure 12). Although the wage increase directly improves per capita incomes in the public sector and somewhat reduces the incidence o f poverty among households headedby civil servants, the impact on aggregate poverty i s not measurable at the level of single digits. This result follows directly from the limitedimportance o fpublic sector households. The public sector provides direct income to about 4 percent o f the population, and the poverty headcount o f this SEG is only 14 percent. Aside from the question how wage increases would be financed, such a policy targets a population group that can only make a very limited contribution to reducing aggregatepoverty. Figure 12. PublicSector Wage Increase (Difference with baseline scenario) 0.1% 0 1% 0 I% 0 1% 0 1% 0 0% t' 0 0% 0 0% 0 0% 0 0% 121, Varying the composition of GDPgrowth. The baseline scenario assumes that the bulk o f growth comes from secondary and tertiary sector growth. Inthe following simulation, we assume that the overall GDP growth rate stays the same but that the growth contribution o f the primary sector is higher, rising by 2 percentage points, and that the contribution of the tertiary sector declines accordingly. - 48 - 122. The results o f this simulation, displayed inFigure 13, show that the change inthe growth composition accelerates the decline inpoverty and reduce inequality. This result is not surprising inlight o f the simulations discussed above but indicates the importance o f sources of growth for achieving poverty reduction at identical GDP growth rates. The change in the growth composition implicitly benefits equally producers o f tradable agricultural goods and non-tradable agricultural goods, with the latter holding the larger share in production. Since the latter also belong to the poorest SEGs compared with some of the urban SEGs expected to benefit from tertiary sector growth, the change ingrowth composition underpins an acceleration of the decline in poverty and reduces income inequality. As noted earlier, the finding is predicated on a relatively modest speed o f urbanization. Figure13. Changein the Compositionof GDP (Difference with baseline scenario) 4 % . - .-.- _ c - -.-- 3.0% - .e c- - e - - 2.0%- -.---- _ * . - D. VULNERABILITY TOMACROECONOMIC SHOCKS 123. BurkinaFaso is exposed to a variety of exogenous shocks, includingrainfall patterns and commodity price shocks. Inthis section, the study considers the impact o f these shocks on long- runpoverty reductionobjectives to give a handle on the magnitude of upside anddownside risks to which the country may be exposed. 124. Adverse weather conditions. To simulate a drought, we reduce agricultural output by 10 percent in 2005, reducing the primary sector growth rate from 4.5 percent under the baseline to 1.3 percent. Growth figures inother years are assumed to remain unchanged, so the level o f GDP i s permanently lower in outer years. Figures 14 and 15 show the impact of this scenario on poverty and inequality. Overall, the shock (andthe subsequent loss owing to the lower level from which the economy grows) adds more than three percentage points to the poverty headcount in 2015. Moreover, since rainfall patterns affect the primary sector, inequality rises above the baseline. - 49 - Figure14. Adverse Weather Conditions,PovertyImpact 25% - 20% - 15% - - - PDAct.(%) -BasePo Figure15. AdverseWeather Conditions (Difference with the baseline scenario) 3 0 % - 2 0 % . 10%- o m - - .. 1 .- - - - - -Bp/lnc Gini-Total P o / 125. Exportprice shock. Cottonis Burkina Faso's most important export good, accounting for about 60 percent of exports. In recent years, cotton prices have fluctuated considerably and prices have been depressed by subsidies indeveloped countries that have increased world supply of cotton. InBurkina Faso, cotton prices for producers have varied less than international prices, but the cotton parastatal has passed on some price variations through lagged changes in the - 5 0 - bonuses.I5 Below, we review the impact o f a 20 percent increase incotton prices (Figure 16) on poverty. Figure 16. Export PriceShock (Difference with baseline scenario) 126. Inabsence of an increase inthe cotton-growing population, an increase inthe income o f extant cotton growers by about 20 percent has an impact on the aggregate poverty headcount o f 1.5 percentage points. Cotton growers contribute only 18 percent o f the poor, but an increase in their incomes by 20 percent would have a measurable impact on the national poverty headcount. The growth incotton incomes would also increase inequality on the national level since it would benefit a relatively narrow group o f the population. If rising cotton prices were sustained over time and draw subsistence farmers into cotton production, it could benefit a larger group and the results as regards the poverty headcount would turn out to be even more favorable. Regarding negative shocks to cotton prices, we find symmetry o f results for short-term and relatively small shocks. Larger and more long-lasting negative shocks lead to an increase in the poverty headcount. E. CONCLUSIONSAND RECOMMENDATIONS 127. Simulations using the poverty analysis macroeconomic simulator show that it would be possible to reach the income-poverty target set by the government. Under a baseline scenario of 5 percent real growth over the long term driven more by the secondary and tertiary sectors than the primary sector, income growth among the poorest would be sufficient to reach the target. However, if, as assumed, urbanization rates increase only modestly, the bulk o f the population in l5 For instance in 2001 and 2002, the cotton price fell respectively by 16.1 percent and 8.9 percent, and SOFITEX eliminated a CFAF 20kg bonus it had paid in 1999l2000. A CFAF 1Oikgbonus was reintroduced with the 2002-03 agricultural campaign. - 51 - rural areas will benefit less from growth under the baseline than urbanpopulation, and inequality will rise. 128. Changes inthe composition of growth can accelerate the trend toward poverty reduction. However, these changes have to benefit an important segment o f the population to be effective. Increasing only cotton output without increasing income growth among those who currently farm mostly for their own consumption, will have its limits as a poverty reduction strategy and instead lead to rising inequality. Instead, growth would need to be broad-based and include the current subsistence-farming sector. 129. Exogenous shocks canbe important obstacles to poverty reduction. A one-time decline in agricultural output with a subsequent return on the previous growth path has a long-term effect on the poverty headcount, and short-term cotton price fluctuations appear to have a aggregate povertyimpact and could affect the wider population. 130. On the basis of the findings of this and the previous chapter, it is recommended that the government undertake the following actions: 0 Build on the strategic vision for broad-based growth of the PRSP to improve the effectiveness and focus of government actions that could draw subsistence farmers into market-based and export activities and broaden the poverty-reducing impact o f cotton productionthrough a hrther improvement inits performance and reach, payingparticular attention to raisingagricultural productivity. 0 Undertake a systematic review under the PRSP process o f programs and activities as regards their link with the govement's growth andpoverty-reduction objective. 0 Using PAMS and other tools, systematically incorporate a review of the poverty and inequalityimpact of growth-supporting policies for rural and urban sectors into the PRSP and policy design to recognize how government actions may support economic growth ' that is equitable. 0 Deepen the study of exogenous shocks to explicitly identify risks for the poverty reduction strategy and identifypossible government policy responses. - 52 - V. PROVIDINGSOCIAL SERVICESTO THE POOR 131. Providing social services to the poor i s one o f the key objectives o f the PRSP. This chapter updates information on social sectors and social protection using the 2003 household survey. As part o f the continuous review of policies under the PRSP process, it provides information to policymakers, complementing the 2004 PER (World Bank, 2004b), for the effective targeting of social spending to the most deprived segments o fthe population. A. EDUCATION 132. Despite significant progress in recent years, educational attainment in Burkina Faso remains low, and the country faces enormous challenges as it strives to achieve universal primary education by 2015. Education in Burkina Faso i s characterized by low enrollment rates inprimaryandsecondary education, highilliteracy rate amongthe adults, andpersistent regional and gender inequalities in educational attainment. Using the 2003 Burkina Faso Priority Survey and official government statistics, this section analyzes benchmark educational outcomes by gender, location, and expenditure quintiles. The analysis includes literacy, educational attainment, school attendance rates by levels o f education, and utilization o f public educational services. The paper also analyzes private and public expenditures on education and the incidence o fbenefits from public spending on education. TheEducationSystem 133. The education system in Burkina Faso consists o f preschool education, six years o f primary education (grades CP1-CM2), seven years o f secondary education (grades 6-Terminale), post-secondary professional and technical education, and four years o f higher education. Preschool education, as well as post-secondary education, is very limited, and they are not included in the current study.I6 The education system in Burkina Faso provides the largest numbero fjobs inthe public sector (see Box 9). 134. Primary education. In 2002, there were about 5,800 primary schools in Burkina Faso, which was 40 percent more than the same number in 1998 (Table 22). While primary education i s mostly provided in public schools (87 percent), the number o f private schools during this period increased by 67 percent, which was almost three times faster than the increase in the number of public schools (24 percent). In 2002, there were about a million o f students and twenty thousand teachers in primary education in Burkina Faso. Public schools are more crowded than private schools, with the student/classroom ratio being about 50 for public and 45 for private schools. The average studentkeacher ratio was 51. l6 In2001, less than two percent of 3-6 year old children inBurkina Faso were attending preschool establishments mainly located in two largest cities o f the country - Ouagadougou and Bobo Dioulasso (Kabore, 2002). As for higher education, university students comprised less than one percent o f the total number of enrolled at all levels of education inthe 1999-2000 academic year (World Bank, 2004a). - 53 - Table22. PrimaryEducationStatistics,1998-2002 1998 1999 2000 2001 2002 Primary schools 4,159 4,860 5,131 5,389 5,804 Public 4,055 4,339 4,s 17 4,697 5,028 .................................................................................................................................................................. Private 464 _" 521 ._.. 614 ... " 776 .... ~ 692__ "" " ~ Classrooms 15,983 17,037 17,456 19,252 20,62 1 Public 14,080 14,917 15,171 16,619 17,658 Private--...........-................ ................. 1,903 2,120 ............ 2,285 2,633 .. 1 "l.." " *" "l_"ll._ ._ .... .........2,963........................... " ...... ~ "" " " ~ Students 816,393 852,160 901,29 1 938,238 1,012,150 Public 727,990 755,090 792,880 819,338 880,211 ................................. Private .............................. ............... ........................................ 88,403 108,411 .............................................................................................................................. 118,900 I 31,939 ~ ~ ~ " ~..... 97,070 ~ Teachers 15,073 16,762 17,294 18,176 19,740 Public 13,519 14,704 15,091 15,779 17,053 . Private .. ................."........... 1,554 ~ ~ " " ............."2,05 8............."._ ........... 2,203 2,397 .. 2,687...................... " StudentMassroomratio 51.1 50.0 51.6 48.7 49.1 Public 51.7 50.6 52.3 49.3 49.8 Private ...... ..................................." 46.5 45.8 47.4 45.2 44.5..................... ~ " " " " ~......."I._.I ......-I".._ "... ._I.I " " Studendteacherratio 54.2 50.8 52.1 51.6 51.3 Public 53.8 51.4 52.5 51.9 51.6 Private 56.9 47.2 49.2 49.6 49.1 Source: MEBA Box 9: Managementof the EducationSystem in BurkinaFaso Currently there are two education ministries in Burkina Faso: Ministry o f Basic Education and Alphabetization (MEBA), including a Minister-Delegate for literacy and non-formal education, and the Ministryof Secondary and Higher Education and Scientific Research(MESSRS), including a minister-Delegate for Technical and Professional Training. Overall, the education sector provides about 26,000 jobs, which constitute almost a half o f the public sector employment in Burkina Faso. MEBA is the largest ministry, and, on-line with the Government priority on expanding primary education, the ministry is expected to enlarge further. In2002, MEBA employed about 21,500 people, 92 percent o f whom were primary school teachers. MEBA has 13 regional (DREBA) and 46 local (DPEBA) education departments, anditruns more than 5,000 primary schools throughout the country. 135. Secondary education. In 1999, the last year for which the data is available, there were about 400 secondary schools in Burkina Faso (Table 23). The share of private schools in secondary education is about 50 percent, which is three times more than the same share in primary education. On average, public schools are larger and have a higher student/classroom ratio than private schools. Public schools are more crowded and the studentkeacher ratio i s much higher inpublic schools (41) than inprivateones (18). EducationalAttainment 136. Adult literacy. The educational attainment of the populationo f Burkina Faso i s low, and there are large disparities across groups (Table 24). The 2003 survey shows that, on average, just 30 percent of adult males and 13 percent o f adult females know how to read and write. The situation is even more severe in rural areas where only one out o f five men and only one out o f - 54 - fourteen women are literate. The data show that despite some progress between 1994 and 2003, fightingadult illiteracy remains abigchallenge for the country. Table 23. SecondaryEducationStatistics, 1998-2001 1998 1999 2000 2001 Secondary schools 381 408 Public 199 212 .................................................Private ............ .. .......................... 182 ................196 . ........................................ ........................................ Classrooms 2,944 3,194 3,211 Public 1,803 1,946 1,952 .............. ............................ Private ....... .................................................................................................................... 1,141 ............ ........................................... 1,248 1,259 .......................................................................... Students 173,205 189,689 199,278 217,176 Public 115,878 124,790 130,711 139,781 .................... ................................................................................................................ .................. Private 57,327 64,899 68,567 77,395 " "-"....... ..................................................................................................................................... Teachers 6,215 6,541 3,771 Public 3,024 3,015 3,001 ................................................................................................................................................................................................................................................................¡ Private 3,191 3,526 Studenticlassroomratio 58.8 59.4 62.1 Public 64.3 64.1 67.0 ................................................................................................................................................................................................................................................................,, Private 50.2 52.0 54.5 Studentiteacher ratio 27.9 29.0 52.8 Public 38.3 41.4 43.6 Private 18.0 18.4 Source: MESSRS Table 24. Evolution of Literacy, 1994-2003 ~~~ Literacy 1/ ~ ~~ 1994 1998 2003 Urban 51.6 50.6 56.3 Male 61.7 59.9 65.7 .. Female " 40.9 42 "............................ 47 " .............." .."............................. ~ ~ Rural 11.8 10.8 12.5 Male 18.8 15.6 18.8 ........ 5.7 6.8 7.2 ^ .....Female " ._ ................................................................................................. ........... " .................................................................................................... Total 18.9 18.4 21.8 Male 27.1 24.8 29.4 Female 11.4 12.9 12.5 Source: World Bank calculations usingthe 2003 priority survey liDefinedasabilitytoreadandwrite;percentof15+year-oldpopulation 137. Literacy is strongly correlated with economic wellbeing (Table 25). The survey data show that individuals from poor expenditure quintiles have less education than their counterparts from rich quintiles. There are also significant income-related disparities in literacy across gender groups and location. In urban areas, for instance, just 31 percent o f adults from the poorest quintile can readandwrite, which is much lower than the same indicator for the richest quintile (70 percent). On average, inthe poorest quintile only one out o f 20 adult females knows how to read andwrite. - 55 - 138. Younger adults, on average, are more literate than their parents or grandparents (Figure 17). The survey data show that young adults of age 30 and less have the highest literacy level across all age and gender groups (close to 40 percent). However, the fact that the 15-19 year old males have a slightly lower literacy level than their 20-29 year old counterparts is an alarming signal andmight indicate the male adult literacy "satiation" point. Table 25. LiteracyLevel by ExpenditureQuintiles,2003 Literacy 1/ Poorest 4 2 Q3 44 Richest Rural 10 11 11 13 19 Male 16 17 18 19 26 ..... Female .. 5 ............................. 6 ............................................................. 6 8 .............. 13 ............... ~ " ..................................................................................... ............................ ................................................... ~ " " " Urban 31 34 44 48 70 Male 18 20 23 29 50 ............ " Female _. _ 6 7 ._ ............... 35 . . ~ 14 ~ ............................................................................................ " ................................ . 11._ "_ " Total 11 13 16 21 42 Male 18 20 23 29 50 Female 6 7 11 14 35 Source: World Bank calculationsusinnthe 2003 onontv survey - . . l/Defined as an ability to readand write; percentof 15+ year-oldpopulation Figure 17. LiteracyLevelby Gender and Age Groups 60+ male female 50-59 3 Y) 40-49 & 30-39 a2 2 20-29 15-19 ?4 1 0 I O 20 30 401 Source: World Bank calculations usingthe 2003 priority survey 139. Typically, the head o f a household i s its main breadwinner, and hisher income- generating ability is determined by the level of education that he/she has achieved (Table 26). The survey data show that 90 percent o f rural and 48 percent o f urban heads of households in Burkina Faso have no education at all. Inrural areas, just two percent o f household heads have completed primary education andonly three percent of them have some secondary education. 140. Enrollment rates. The primary school enrollment rate inBurkina Faso increased from 35 to 44 percent between 1994 and 2003, but it still remains quite low by international standards - 56 - (Table 27)." The survey-based primary school enrollment rate of 44 percent is comparable with the census-based enrollment rate o f 47 percent (2002) and the official enrollment rate o f 47.5 percent. There i s also large divergence inprimaryschool enrollment rates across location, gender groups, and regions. In 1994-2003, for instance, the primary school gross enrollment rate (GER) for urban areas increased from 74 to 102 percent, while the increase in rural areas was very modest (from 28 to 34 percent). Furthermore, while the GER increase helped reduce and practically eliminated the enrollment gap between boys and girls inurban areas, the gender gap inrural areas fbrther widened. Analysis ofenrollment data also shows that remoteregions ofthe country have much lower enrollment rates than the regions closer to the capital. In 2003, for instance, the primary school GER in Central Region was 94 percent, while the rate inthe Sahel was only 22 percent. Table26. Headsof Households HighestLevel of Education,2003 Levelof education Rural Urban Total None 90 48 81 Primary, incomplete 5 11 6 Primary, complete 2 8 3 Secondary, incomplete 3 21 7 Secondary, complete 0 4 1 Total 100 100 100 Source: World Bankcalculationsusingthe 2003 priority survey Table 27. PrimarySchool Gross EnrollmentRates, 1994-2003 Primary educationGER 1994 1998 2003 Urban 14 102 102 Male I 9 106 104 ............................................................................................................................................................... Female 69 99 ................................................................. 100 " ........................ Rural 28 31 34 Male 34 3 1 41 Female - -..... -22 24 ...................................................................... 21 I ~ ~ -......_ ~ ~ Total 35 41 44 Male 41 47 50 Female 29 35 38 Source: World Bankcalculations using the 1994, 1998and 2003 priority surveys 141. There is large divergence inschool enrollment across different income groups indicating that children from better-off households are more likely to attend primary school than those from poor households (Table 28). The survey data show that in urban areas the primary school GER for the richest quintile i s 84, which is 24 percentage points higher than the enrollment rate for the poorest quintile. Inrural areas, the difference in enrollment between these two income groups is 50 percent. The enrollment rate ratio between the richest and poorest quintiles in rural areas is 1.36 for boys and 1.86 for girls, suggesting that income inequality affects girls more than boys. The Burkina Faso Priority Survey defines primary school enrollment for children o f 7-12 year-old and secondary education enrollment for chldren o f 13-19 year-old. - 57 - 142. Enrollment in secondary education in Burkina Faso is very low, with large differences across gender, location and income groups. On average, just 16 percent o f children o f the secondary school age attend schools (Table 29). The most recent official secondary school enrollment rate is 12 percent (2001). The gender gap in enrollment is high inrural areas, where in 2003 just 8 percent of boys and only 4 percent of girls went to secondary schools. As the survey data shows, in rural areas in the last ten years enrollment in secondary education has practicallynot changed. Table 28. Primary School Gross Enrollment Rates by Expenditure Quintiles, 2003 PrimaryeducationGER Poorest 4 2 4 3 44 Richest Rural 21 27 28 28 32 Male 28 31 33 31 38 ...... Female................................... ..... 14 23 22 26 " .......... ..............",.,....".............................. 26................ ~ ~ Urban 60 65 69 79 84 Male 61 62 76 81 89 ....................... "............... Female 59 68 62 76 80 "_ .. ........... __ I... " ....." ......._" ........... ....... ................................................................... ......... ~ " ~ ................................................... ~ Total 23 30 34 37 53 Male 29 34 39 39 57 Female 16 26 28 35 48 Source: World Bank calculations usingthe 2003 priority survey Table 29. Secondary School Gross Enrollment Rates, 1994-2003 Secondary education CER 1994 1998 2003 Urban 37 49 51 Male 45 56 54 ..................................... ..Fe.!??ak ............................................................................................. 29........................................................ ._..................................................................................................................... 41 48 Rural 5 5 6 Male 7 6 8 ........................................................................................................................................................................................ Female 3 3 . 4 ~ Total 11 13 16 Male 14 15 17 Female 9 10 14 Source: World Bank calculations using the 1994, 1998 and 2003 priority surveys 143. As for primary education, secondary education enrollment rates are also highly affected by households' economic wellbeing (Table 30). The data show that children from rich households are more likely to attend secondary school than their counterparts fiom poor households. The income gap in enrollment is especially significant in urban households where only one out of eight kids in the poorest quintile goes to school. Inurban areas secondary school enrollment increaseswith income for both girls and boys, while inrural areasjust enrollment for boysincreaseswith higher incomes. 144. Education completion rates. Figure 18 shows education completion rates for the 15-25 year old cohort. The figure shows that, on average, about 60 percent o f students leave the education system after completing just 5 years o f education. Inrural areas, more than 80 percent o f the initially enrolled students leave school after just 2 years o f education. The primary school - 58 - completion rate is about 30 percent in rural and 60 percent in urban areas. As indicated by Figures 19 and 20, boys drop out of primary school more often than girls, while girls leave secondaryschool at a higherrate thanboys. Table 30. Secondary School Gross EnrollmentRatesby ExpenditureQuintiles,2003 Seconday education GER Poorest 4 2 Q3 4 4 Richest Rural 4 4 6 5 8 Male 5 5 8 6 12 .-.,,.,I Female ..................................................... .................................................... 3 4 4"_ ......................................... 3 4 ............ I " " " " ~ ~ " Urban 13 15 30 34 50 Male 13 19 27 37 55 . .. Female " 11 32 46 " ............................................. "..................... 12...................... " .... .............................. " " ....................................................................... " " _32 ............................................................... ............ ~ " Total 5 5 10 12 27 Male 6 6 11 14 33 Female 3 5 9 10 23 Source: World Bank calculations usingthe 2003 priority survey Figure 18. EducationCompletionRatesby Location \ +rural +urban +total ................ I I , , 0 2 4 years of education 6 8 10 12 14+ Source: World Bank calculations using the 2003 priority survey Figure 19. EducationCompletionRatesin UrbanAreas 0 2 4 6 8 10 12 14+ years o f education Source: World Bank calculationsusingthe 2003 priority survey - 59 - Figure20. EducationCompletionRatesin RuralAreas 100 2 +female +male 80 In 9 2 60 b 2 ~ 0 40 8 s - a 20 0 I 0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 + years of education Source: World Bank calculations usingthe 2003 priority survey Povertyand Access to Education 145. Education andpoverty. Education highly correlates with poverty. As discussed inchapter 3, different regressions on 1994, 1998, and 2003 survey data show that the probability o f a household being poor decreases with the head of household' years o f education. For example, Lauchaud (2003) estimates.that a 10 percent increase in the head of household's years of education reduces the probability of being poor by 0.23 percent. His sensitivity analysis concludes that, all things being equal, a household headed by an illiterate person i s three times more likely to be poor than a household whose head has secondary or higher education. The survey data also reveal a significantly higher return to secondary and higher education as compared to primary education. Noting low returns on education in Burkina Faso (see below), Lachaud (2003) also shows that absence of education does not necessarily translate directly into beingpoor. 146. Constraints to school attendance. Inrural areas, distance to the nearest school is one of the main factors affecting school attendance (Table 31). The survey data show that, on average, in rural areas 45 percent of kids attending primary schools and 75 percent of kids attending secondary schools have to walk more than 30 minutesto reach the nearest school. There are also large differences across regions on access to school. According to the survey, inthe Southwest region, for instance, 72 percent of children attending primary school walk more than 30 minutes to get to school. Inthe case of secondary education, the absolute majority o f students across the country (except inOuagadougou) also spend 30 or more minuteswalking to school. 147. According to the survey, on average, 22 percent o f the initially registered students were not attending schools at the time o f the survey (Table 32). Both for primary and secondary education, expulsion from school (48 percent) was the main reason for missing school, while 27 percent reported the high cost of studying as one o f the main reasons for school absenteeism. Other reported reasons for missingschool were: no needto study (8 percent), difficult access to school (3 percent) and beingsick (3 percent). - 60 - Table 31. Access to School by Location and Region Primary school Secondary school < 30 min >3Omin < 30 min >30min Location Rural 55 45 12 88 Urban ............................... ................-............"91 ........................................9 72 28 ~ ~ Region Sud Ouest 28 72 6 94 Sahel 38 62 6 94 Cascades 43 57 15 85 Centre Nord 46 54 7 93 PlateauCentral 48 52 11 89 Est 48 52 16 84 Centre Est 60 40 23 77 Centre Ouest 65 35 19 81 Centre Sud 68 32 27 74 Hauts Bassins 68 32 37 63 Nord 79 21 28 72 Boucle du Mouhoun 79 21 21 79 Centre 93 7 68 32 Total 62 38 25 75 Source: World Bank calculations using the 2003 priority survey Table 32. Reasons for Not Attending School Reasonsfor not attending, inpercent %not Total would work/ school is attending like to not need toofarho school expelled work too costly to study school sickness other Age 7-13 7 50 4 22 8 7 7 2 100 ...... ..... 14-19 ..................... - 40 _ 47 9 .................. 29 8 2 2 3 100 " " " ..................................................................................................... " " .-..................................................... " Location 100 Rural 24 52 6 23 7 4 4 4 100- Urban 19 38 13 37 8 0 2 2 100 Total 22 48 8 27 8 3 3 3 100 Source: World Bank calculations using the 2003 priority survey 148. Almost a quarter o f households with school age children report dissatisfaction about education (Table 33). Across all categories, major reasons for school dissatisfaction were a lack of textbooks and school equipment (73 percent), teacher shortage (15 percent), and quality o f teaching (16 percent). The survey data also suggest that there are not enough teachers in secondary schools. The quality o f education was a major issue inprivate primary schools, with a third of students reportingbeingnot happy about the quality o fteaching, butwas not an issue for private secondary schools. The results are consistent with survey responses on aspirations by Burkinabk for the education system in2025 (Box 10). - 61 - Table 33. Reasons for School Dissatisfaction Reasons for dissatisfaction Yo of books/ quality of dissatisfied equipment teaching no teachers school building other Location Rural 26 7s 12 17 8 7 urban."..-. ......... ............................................................. 70 22 21" 14...........................................6 ......."..... _ 10 ~ ~ Type of school Primaryschools 24 77 I5 11 8 6 Public 24 78 14 12 8 7 Private 8 89 33 11 0 0 Other 19 65 21 5 12 4 Secondaryschools 26 67 16 24 6 I O Public 29 65 16 27 7 I O Private 22 71 0 29 0 0 Other 20 72 17 12 4 9 Total 24 73 15 16 7 8 Source: World Bankcalculationsusingthe 2003 priority survey Box 10: Education in Burkina Faso-2025 A recent national study "Burkina 2025: Survey on National Aspirations" conducted by Ministry of Economy and Development asked about 1,500 adult respondents to express their vision on how the education system in Burkina Faso would look like in 25-30 years. According to the survey, increased school enrollment rates (34 percent of respondents), better-quality schools (18 percent), and free education (12 percent) were the most frequent answers. When asked about necessary steps to improve educational outcomes in the country, the respondents suggested that reducing tuition fees (35 percent) and increasing the number o f schools (18 percent) would lead to better results. Other most frequent suggestions were improved working conditions for teachers (7 percent) andbetter educational programs (6 percent). Source: "Burkina Faso2025: Survey on NationalAspirations", 2002. 149. Education and labor market outcomes. There is not a significant relationship between education and labor force participation (LFP) for heads o f householdsinBurkina Faso, but labor participation decreases with education for other household members (Table 34). One o f the major reasons for this inverse relationship between LFP and education is the structure of Burkina Faso economy, which, beingmainly basedon subsistence agriculture, does not reward education. The data show that in urban areas the LFP rate o f individuals with primary education is lower thanthe rate of those with no education, but LFP starts rising for individuals with secondary or higher education, confirming the earlier hypothesis that mostly secondary and higher education have a significant effect on labor participation and economic wellbeing (see above). 150. There is high incidence of child labor inBurkinaFaso (Table 35). The survey shows that 44 percent o f children o f age 5 to 14 years old are involved in some sort of economic activity, with work on family land or small business beingthe most common type o f activity (98 percent). There is no significant difference in labor incidence across boys and girls. Children from poor - 62 - households are twice more likely to work (52 percent) than their counterparts in non-poor households (26 percent). The data show that 50 percent o f children inrural areas work on family land, while in urban areas the labor incidence is about 14 percent. As supported by the survey data, better-educated parents are more likely to keep their children at school and away from the labor market than parents with less education. Table 34. Labor Force ParticipationRateby Levelof Education hhhead other men hhwife other women Total Rural 93.6 62.5 84.9 58.8 72.3 No education 93.6 70.6 85.1 64.2 76.9 Primary 93.9 33.8 83.1 25.4 40.6 Secondary or more 94.6 39.2 66.8 26.2 56.4 Urban 85.4 40.8 59.0 27.0 48.5 No education 83.6 49.5 58.9 34.5 55.2 Primary 85.6 35.0 57.1 20.1 37.2 Secondary or more 87.6 40.8 60.7 25.2 52.0 Source: Lachaud(2003). Table35. Incidenceof ChildLabor (Inpercent) Boys Girls E 9 year-old 10-14year-old. 5-9 year-old 10-14year-old Grand total Poor 43.6 58.9 45.1 65.3 51.9 NO"P%!! ...... ................................................... 19.2 "..... 31.7 .......................... 22.0 30.7 25.5 Rural 41.7 58.6 42.6 63.7 50.2 Urban, central 1.4 7.6 5.3 10.6 6.4 urb?!.?..o!!er. .................................10.7 - 16.9"..... " 12.2 ................................... "... _ 14.8............................................................................ 13.5 Education ofthe head ofhousehold None 40.0 55.8 41.0 60.0 48.0 Primary 20.4 29.8 18.5 31.9 34.4 Secondary 8.4 9.1 9.6 7.4 8.7 College 0.0 0.0 1.9 4.7 1.9 Total 37.0 51.2 37.7 54.5 44.0 Source: Lachaud(2003). Costs o f Education 151. Public spending on education. Education is one o f the priority sectors for public spending in Burkina Faso." In 2002, the total spending on education comprised 23 percent o f the state budget, with primary education being the largest budget item (14 percent). Primary education also receives the largest share o f HIPC funds. Since 2001, when HIPC funds became first available, about CFAF 37 billion CFAF (33 percent o f total) have been allocated to primary education. Contributions from international donors are another large source o f funds for '*An analysis of private and public spending on primary education shows that the Government roughly provides about 80-85 percentof totalexpenditures. - 6 3 - education in Burkina Faso. In2002, for instance, from the total o f CFAF 87.6 billion spent on education, CFAF 58.3 billion CFAF (66.6 percent o f total) were from the government budget, CFAF 8.9 billion (10.2 percent) came from HIPC funds, and CFAF 20.3 billion (23.2 percent) were contributed by donors. In addition, Burkina Faso is also eligible for funding through the Education-for-All Fast Track Initiative (EFA-FTI), which in 2004-05 could provide about 20 millionU S dollars towards achieving universal primaryeducation inthe country. 152. Benejt incidence analysis of public spending. The benefit-incidence analysis (BIA) couples household survey data on school attendance with aggregate public spending on education to assess how different income groups compare interms o fthe amount o f subsidythey receive. For the purpose of BIA, the per-student subsidy is calculated by dividing the total public spending by the total number of students." Ingeneral, BIA have to be interpreted cautiously since their estimates o f amounts spent per child are uncertain and the expenditure data not always captures urban-rural andregional gaps accurately. The BIA shows that the total education sector subsidy is unquestionably pro-rich. Inprimary education, for instance, 28 percent o f total subsidies went to the richest quintile, while the poorest quintile receivedjust 14 percent o f total subsidies. The incidence is even stronger insecondary education, where 41 percent of subsidies benefit the richest quintile while the poorest quintile receives only 9 percent o f total subsidies. Highper-unit cost of secondary and higher education (Box 11) and the larger participation of students from better-offhouseholdsinsecondary and higher education contribute to this finding. Table36. Distributionof Subsidies for Education Subsidyper student, Total subsidy, `000 Quintile Students `000 CFAF CFAF % subsidy Primary education Poorest 534 40.4 21,574 14.5 2 634 40.4 25,614 17.3 3 718 40.4 29,007 19.5 4 771 40.4 31,148 21.0 Richest 1,016 40.4 41,046 27.7 Total 3,673 40.4 148,389 100.0 Secondary education Poorest 199 69.0 13,727 9.1 2 268 69.0 18,487 12.2 3 370 69.0 25,523 16.9 4 446 69.0 30,765 20.4 Richest 906 69.0 62,496 41.4 Total 2,189 69.0 150,997 100.0 Sources: World Bank calculationsusingthe 2003 priority surveyand expendituredata World Bank (2004b). 153. Figure 21 shows the results of BIA for boys and girls using concentration curves, which plot the cumulative percentage o f the survey sample of boys and girls, ranked by expenditure quintiles, against the cumulative percentage of subsidy received. The equality line indicates the Pleasenote that the expenditure data is for 2002, while attendance rates are reportedfor 2003. - 64 - case when subsidies are distributed equally across income groups. Given the gender disparity in enrollment, the benefit incidence shows that both in primary and secondary education the pro- rich allocation o f public subsidiesnegatively affect girls more than boys. Box 11: Per-unitcost of PublicSpendingon Education Even if achieving universal primary education i s one o f the main objectives o f the Government's education policy, the expenditure data show that on per-student basis the government spends more on secondary and higher education than on primary education. In 1998-2002, for instance, public spending on primary education increased by 55 percent, while the secondary and higher education spending increased by 68 percent. In 2002, for instance, the budget spending on a college student was 722,000 CFAF, an amount which is 18 times more than what was spent on a pupilinprimary education. I I Perstudent publicspendingin BurkinaFaso, 2002 Higher iI 1 7 2 2 Secondary 69 Primary `000 CFAF I I Source: World Bank (2004b) Figure21. Equityof Public Spendingon Education Primary education Secondary education --cboys--cgirls -equality -m- boys -+- girls -equality I 0 sl 92 q3 s4 q5 0 ql q2 q3 s4 qs expenditure quintiles expenditure quintiles Source: World Bank calculationsusing the 2003 priority survey 154. Private spending on education. On average, spending on education comprises less than four percent o f Burkina Faso households' non-food expenditure (Table 37).20There are large differences on education spending across households in urban and rural areas and across different income groups. On average, an urban household spends annually CFAF 23,500 per 2oUnfortunately, the 2003 Priority Survey data do not allow to separate household education spending by levels o f education. Instead each household reports its total monthly and annual spending on education for all kids and regardless o f their level of education. The per-student household expenditure data used inthis section are calculated by dividing each household's annual spending on education by the number o f children (6-20 years old) currently attending schools. - 65 - child, while a rural household spends 10 times less of that amount. Inrural areas, the ratio o f per capita spending of the richest and poorest quintiles i s 1.5, while inurban areas this ratio is about 6. As confirmed by enrollment figures presented earlier in this report, the ability o f a household to pay for education determines if a child fiom that household goes to school or not. As also noted above, for poor families, the costs o f education can be a critical deterrent to participation in education. Table 37. Per-capitaAnnualPrivateSpending onEducation (InCFA francs) Poorest 4 2 Quintiles 4 3 4 4 Richest Average Rural 1,905 2,299 2,297 2,228 2,950 2,354 Urban 5,089 6,172 10,115 13,356 33,151 23,500 Total 2,096 2,663 3,290 4,387 15,464 6,752 Source: World Bank calculations usingthe 2003 priority survey. B. HEALTH 155. Health status and health indicators in Burkina Faso have progressed little in the past decade. Even though the government and the donor community have made significant efforts towards improving human development indicators, child and infant mortality are still high, nutritional status i s poor, human resources allocated to health sector are unequally distributed and at times scarce. This section uses the result o f the 1998 and 2003 priority surveys and the 1998 Demographic and Health Survey (DHS) to discuss the performance o f the health system with regards to the poor. To this end it reviews health status, utilization o f health care, health expenditure, and determinants o f health care demand. The analysis is complementary to the recently conducted public expenditure review (World Bank, 2004b).21 The HealthSystem 156. The options for treatment inBurkina Faso include both traditional and modern providers. The traditional practice i s neither regulated nor organized, and little information on its use or costs is available. According to the 1998 and 2003 priority surveys (EPs), resorting to traditional providers i s the second common option in rural areas after using a local health center (CSPS), whereas traditional practice is only the last option inurban areas. 157. The modem sector consists of both private and public facilities. The private modem sector i s still poorly developed. In 1999, there were only 300 medical institutions and 105 pharmacies throughout the country, more than 80 percent o f them in urban areas.22The private institutions are runby individuals, non-government organizations (NGOs)and other associations, and the services are delivered by doctors (which are often public sector employees) and pharmacists. The creation o f private clinics and hospitals is to some extent limited by insufficient regulation, or by an unfavorable tax status. *'Throughout the analysis, the consumption aggregate described inchapter 2 is being applied. 22See World Bank (2003). - 66 - 158. The public sector is organized in a three-tier system, The first tier consists o f primary health care centers (called Centres de Santk et de la Promotion Sociale, or CSPS) in the first echelon, and medical centers and medical centers with emergency units (CM and C M A respectively) in the second echelon. The second and third tiers encompass the regional and national level hospitals (CHR and CHN respectively), and the higher level hospitals are supposed to be referral hospitals for the lower level healthinstitutions. CSPSs are the most common option at health district level and are designed to provide preventative and curative care and to be serviced by a nurse assisted by a midwife, andtwo other staff members.23The inpatient services and higher-level care at the health district level are offered by CWCMA, which are also offering child-delivery services. In2002, Burkina Faso had 1051CSPSs, 87 dispensaries, 39 maternities, 28 CMs, and 36 CMAs (see Table 38). Table38. HealthInfrastructure Average radius CHNKHR CMA C M CSPS Dispensaries Maternities Total of servicefor CSPS (km) 1990 11 577 123 16 797 1995 11 17 61 677 116 16 898 10.4 2000 11 31 41 798 145 46 1072 9.4 2001 12 33 36 835 145 46 1107 9.2 2002 12 36 28 1051 87 39 1253 9.0 Source: Ministryof Health (2002) and World Bank (2004b). 159. The CHRs and CHNs offer the most specialized and highest level o f care. They are mostly concentrated inurban areas, and their total number i s almost equal over years: 11during the 1990s, and 12 starting 2001. Interms o f medical equipment, the CHWCHNs are reportedly under-equipped, and the equipment tends to be "poorly maintained and sometimes un~uitable."~~ 160. Interms of accessto facilities, in2001 there were 0.9 CSPS per 10,000population, while only 0.6 C W C M A per 100,000 p ~ p u l a t i o n .In~an effort o f the government to expand the ~ availability o f primary care inall the regions, it has increased the number o f lower echelon health facilities over the past decade. In2002, there were 82 percent more CSPSs and 144 percent more maternity wards than in 1990. However, the number o f C W C M A s decreased from 74 centers altogether in 1994 to only 67 in 2002. Still, the CM-to-CMA ratio did improve from almost 3:l in 1994to approximately 4 5 in2002. 161. Despite certain improvements in the availability o f primary health care facilities, there are big differences across health regions. In 2001, the ratio o f inhabitants per CSPS varied from 8,970 inKoudougou, to 15,025 inFada. Regarding the inhabitants per one CM/CMA, this varies from 84,300 in Gaoua to 251,760 inBobo-Dioulasso. Although the average radius o f service for the CSPSs diminished from 10.7 km in 1993 to 9 km in 2002, there are large geographic differences: in2001, the radius varied from 6.1 kminOuagadougou to 15.2 kminFada. 23Nonetheless, in2002 only 77 percent o f the CSPS were found to meet the personnel norms(World Bank, 2004b). 24See World Bank (2003b). 25The number o f CSPSs includes the dispensaries and maternities. - 67 - 162. Evidence from the 2003 priority survey suggests that health centers are situated within 1 hour by foot for 65 percent o f the population (see Table 39). Most o f the population lives within 2 hours o f a health center regardless o f their location or poverty status. Yet, there are differences in access: 71 percent of the urbanpopulation but only 25 percent rural population lives within half an hour o f a health center, and these figures are 40 and 27 percent for the non-poor and poor populations, respectively. Table 39. Proximity of a Health Center, 2003 (Walking distance inminutes) Rural Urban Non-poor Poor Total Less than 1/2 hour 25.0 70.7 39.0 26.8 33.3 112-1hour 33.4 25.9 29.8 34.6 32.0 1-2 hours 28.0 3.4 20.9 26.5 23.5 2 hours-1/2 day 7.7 * 5.9 6.7 6.3 112-1days 1.2 0.0 0.8 1.2 1.o No access 4.5 * 3.3 4.1 3.7 Source: World Bank calculations using the 2003 priority survey. Note: *=sample size is under 25. 163. In the effort of improving the quality of health care, the government has considerably increased health staffing. In the 1990s, the number o f trained health personnel expanded by one third, mainlydue to an increase inthe number o ftrained nurses and midwives. In2000-02 alone, the number o f doctors and midwives doubled. The health centers fulfilling personnel norms rose from 70 percent in 1995 to 77 percent in 2002. However, human resources in the health sector are managed in a centralized system within the rigid government pay scale, creating disincentives for locating personnel inremote areas.26 164. Concerning drug availability, the distribution o f essential modem drugs improved with the launching o f the CAMEG (or Central DrugProcurement and Distribution Agency) in 1992. Since its creation, the cost o f essential drugs fell by 30-60 percent by 1994. In2003, the private expenditures on .drugs amount for roughly 76 percent o f the household spending on health, a decrease by 7.5 percentage points when compared with 1998. Even though the cost of drugs fell sharply, their increased availability does not imply affordability. Health Status 165. Morbidity and Mortality. Compared with several of its neighboring West-African countries, Burkina Faso has poor health indicators (Table 40). It has among the lowest life expectancies and immunization rates, and among the highest fertility rates, HIV/AIDS prevalence, under-5 and predictedmaternal mortality rates. 26 SeeWorld Bank (2003b) for an in-depth analysis of the human resources. - 68 - Table 40. Regional Comparison of Health Status Indicators Mortality rates Immunizations prevalence Life of children<=I y.0.)(% HIV/AIDS TFR expectancy Matemal Under- DPT 49 year- 15- female, 15- (estimated) Infant 5 Measles 24 year-old old 2002 2002 1995 2001 2001 2001 2001 2001 2001 B.F. 6.3 42.9 1400 104 197 41 46 6.5 9.7 Cote d'Ivoire 4.6 47.8 1200 102 175 57 61 9.7 8.3 Ghana 4.0 54.9 590 57 100 80 81 3.0 3.O Guinea 5.0 46.2 1200 109 169 43 52 1.5 1.4 Mali 6.1 40.9 630 141 231 51 37 1.7 2.1 Niger 7.1 45.7 920 156 265 31 51 1.4 1.5 Source: WDI (2003). 166. Burkina Faso faced improvements in some indicators during the last years while coping with a decline in others (Table 41). To summarize the main trends, total fertility rate and maternal mortality rate, infant and under-five mortality rates decreased. The percentage o f births attended by health staff, and the percentage o f 1year-olds immunized against DPT3 and measles increased. However, life expectancy declined, the percentage o f children without vaccinations by age 1 and the prevalence o f HIV/AIDS for female aged 15-24 increased, as did death rates from transmissible diseases such as measles and meningitis. 167. Burkina Faso i s characterized by high fertility rates. The total fertility rate was almost constant (around 6.8-6.9) during the first half o f the nineties, and it reached a low o f 6.3 in2003. A more detailed picture o fTFR is captured by the 1998 and 2003 DHS:rural rates are 80 percent higher than the urban ones and the rate among female secondary or higher education is only 40 percent o f the rate among females without education. 168. Infant and under-5 mortality rates decreased during the 1998-2003 period. This is a particularly important trend reversal for infant (IMR) and under-five (USMR) mortality rates, which had deteriorated between 1993 and 1998.27 In 2003, IMR reached 83, and U 5 M R 184 deaths per 1,000 live births (Table 42). Nonetheless, despite the (modest) decrease, infant and child mortality rates remain among the highest in the world, and are higher than for the average o f low-income countries. This reversal alone and a continuation o f the current trendwould likely not be sufficient to meet the mortality MDGsby 2015. 27 DHSandofficial statistics differ starkly inabsolute mortality ratesbut bothshow the same trend. - 69 - Table 41. Dynamics of Health Status Indicators 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Population,total (mill) 8.9 9.1 9.3 9.5 9.8 10.0 10.2 10.5 10.7 11.0 11.3 11.6 11.8 Populationgrowth 2.3 2.3 2.4 2.4 2.4 2.3 2.4 2.4 2.4 2.5 2.4 2.4 2.3 Fertility rate, total (births per woman) 7.0 6.9 6.8 6.8 6.5 6.4 6.3 6.2 Life expectancyat birth, total (years) 45.4 45.1 45.3 45.5 44.9 44.2 43.5 42.9 Matemal mortality 350 566 484 Under-5MortalityRate 187 219 184 InfantMortalityRate 115 124 94 93 110 108 105 105 105 104 83 Propnof I-yr-old im"d againstmeasles 41 39 70 43 46 56 61 33 48 43 59 65 64 Propnof 1-yr-old i m ' d against DPT3 32 29 40 48 40 47 37 28 31 42 57 64 67 %Births attendedbyhealthstaff 41 31 36 59 Prevalenceof HIV, female (age 15-24) 5.8 9.7 Prevalenceof HIV 17.2 8.9 11.4 8.6 19.0 16.5 17.6 10.2 Sources: All data are from WDI (2003) unless otherwise mentionedbelow. Mortality rates and percentageof births attendedby health staff for 1993/1998/2003: DHS reports; IMR (other years), immunizationcoverage and prevalenceof HIV (per 100,000 population): Ministry of Health (2002); immunizationrates in2002: World Bank(2003~). Note: MMR is computed per 100,000 live births, USMR and IMR are computed per 1,000 live births, prevalence of HIV is given per 100,000 population. Table 42. ChildMortality Rates for Five-year Periods Preceding the Survey, 2003 years preceding Calendar Neonatal post neonatal Infant mortality Childmortality Under-5 the survey period mortality (NN) mortality (PNN) (1qO) (4q1) mortality (5qO) 0-4 1999-2003 31 51 83 111 184 5-9 1994-1998 46 58 104 112 205 10-14 1989-1993 44 48 92 1I O 192 15-19 1984-1988 46 51 97 120 205 Source: DHS2003. 169. The improvements in mortality rates could mask important differences by location or economic status. For example, there are large ruralhrban disparities for the 10-years preceding the 1998/99 DHS.'' Urban IMR and U5MR are only 60 and respectively 55 percent o f the rural rates, and the gap widens once a child has survived his first year o f life. Moreover, maternal education plays an important role in decreasing mortality as for mothers with primary education the IMR is 15 percent lower, and the U5MR is 26 percent lower. The data o fthe 2003 DHS will show, whether the rising discrepancy o f U5MR between socioeconomic groups observed during 1993-98 has been reversed inrecent year. 170. The priority survey gives some indication, from self-reported diseases, concerning the possible causes for continued high mortality rates of children despite improved vaccination coverage. In 2003, 41.2 percent of sick individuals had malaria. This finding is confirmed by official data: malaria represented the reasons for consultation in43 percent o f the cases, and 63.4 of those hospitalized for malaria are children aged 0-5. Moreover, reportedly malaria causes almost one-third o f the under-5 deaths. Diarrhea and respiratory infections are among the other four main reasons for the morbidity and mortality o f children under 5 years old. During the 28 Data fromthe 2003 DHSwere not available at the time ofcompletion ofthis report. - 70 - 1993-1998 period, the prevalence o f diarrhea increased from 12 to 20 percent, and the prevalence of acute respiratory diseases increased from 11 to 13.5 percent, with a bigger impact in rural areas. 171. As regards HN-AIDS, the official prevalence rates (per 100,000 population) show a decrease in the mid-nineties from 19 percent in 1994 to 10 percent in 1998. In 1997, 370,000 people were affected, while this number climbed to more than 660,000 in 2000. It is estimated that 85 percent o f the infections are through sexual contact, with perinatal (mother-to-child) and blood transmission accounting for most of the remaining. The prevalence rates are 34 percent among TB patients, 16-42 percent among sexually-transmitted disease clinic patients, and 15-60 percent among commercial sex workers. Women 15-24 years old and "adults in their most productive years" are the most affected groups.29 172. Child malnutrition. Child mortality rates and child malnutrition are closely linked. The main indicators for malnutrition are the prevalence o f stunting, emaciation, and wasting. The stunting reflects low height for age (chronic malnutrition), and emaciation low weight for height (presence o f malnutrition, although not necessarily long-term), while emaciation is characterized as low weight for age. The following analysis i s based on the 2003 priority survey, and focuses mainly on stunted andemaciated ~hildren.~' 173. In 2003, almost 40 percent o f the children aged 0-5 are stunted, almost 17 percent are emaciated, and 33 percent are underweight. 44 percent o f the stunted are severely stunted, whereas only 15 percent o f the emaciatedunderweight are severely affected. These numbers reflect an increase in the prevalence o f stunting, and a decrease in the prevalence o f underweighting when compared with the 1998 rates (36.8 and respectively 34.4 percent) from the 1998/99 DHS. 174. The prevalence o f malnutrition i s usually higher in rural areas and lower among the top expenditure quintile of the population (see Table 43). Also, the prevalence o f stunting increases, whereas emaciation decreaseswith age.31These basic findings are confirmed by Lachaud (2003). H e studies the determinants o f the stunting using both linear and non-linear models and notices that income is not necessarily correlated with malnutrition. Moreover, mother's age and educational level, bigger families and urban location positively influence the nutritional status o f the children. A child's age i s positively correlated with malnutrition (and this relation grows stronger as the child grows older). 175. About 40 percent o f stunted children and 16 percent o f emaciated children have participated in a nutrition program. As shown in Table 43, we can ascertain from the priority survey whether a malnourished child ever participated in a nutrition programs (often perceived as food programs) and growth monitoring programs (or GMP, offering mostly but not only 29See World Bank (2003~). 30The data is analyzed from the raw information on the sex, age, weight, and height of the children under 5 years old. Indoing so, we use the ANTHRO software developed for WHO and Center of Disease Control and Prevention (USA) in 1999.Dueto measurement and collecting errors ininputting the data, the programflags improbable values while computing z-scores. These flagged values are omitted from the reported statistics. 3`Bonnet (2001) illustrates that the malnutrition"comes more frombreastfeeding than solid food diet." Therefore, if the birthspacing is poor, malnutrition mightbecome anappendageo folder children. - 71 - weighing and counseling service^).^' The participation rates among urban malnourishedchildren are twice as high as among the rural malnourished, despite the higher prevalence o f malnutrition in rural areas. Although there are no significant differences in the occurrence of stunting and emaciation across expenditure quintiles, the participation in malnutrition programs rises with expenditure levels. Interms of overall participation o f children in nutrition programs (those not malnourished), participation is significantly higher in urban areas and among households with higher expenditure (see Table 44). Table43. MalnutritionStatusof Children0-5 Years OldandProgramParticipation, 2003 (Inpercent) Stunting Emaciation Expenditure quintile Expenditure quintile 1st 1/ 2nd 3rd 4th 5th Total 11st 2nd 3rd 4th 5th Totall Malnourishedchildren,o.w.: 36.7 39.3 38.9 43.1 40.0 39.5 16.2 16.2 17.6 15.2 13.2 15.9 didparticipateinnutritional Rural programs 6.5 9.7 9.4 18.4 18.9 12.0 15.7 13.3 12.5 21.7 17.7 15.7 didparticipateingrowth monitoringprograms 24.6 34.7 32.2 32.3 38.8 32.1 31.7 35.2 35.6 39.3 42.0 35.9 Malnourishedchildren,o.w.: 21.8 21.1 26.4 27.0 24.2 24.7 15.5 15.0 13.6 17.6 13.7 14.8 didparticipatein nutritional Urban programs * 25.6 13.9 23.4 34.7 26.9 * * * 43.4 39.4 36.2 did participateingrowth monitoringprograms * 75.9 75.1 76.1 77.0 76.6 * * * 74.2 93.5 84.4 ~ Malnourishedchildren,o.w.: 36.0 38.0 37.6 40.5 34.5 37.4 16.2 16.1 17.2 15.5 13.4 15.8 did participateinnutritional Total programs 7.1 10.3 9.7 18.9 22.8 13.4 15.2 15.1 13.4 25.3 25.3 18.2 monitoringprograms- did participate in growth I 26.3 36.3 35.5 37.0 48.1 36.2 134.0 38.3 38.7 45.2 59.8 41.8 Source: World Bank calculations usingthe 2003 priority survey. 1/ Inquintile is the poorest, and 5" quintile the richest. Note: *=sample sizes under 25. 176. Reproductive health. Sexual and reproductive health services, including family planning, play an important role in improving child and maternal health through spacing o f births as well as the global effort to contain the HIV/AIDS pandemic. Reproductive health services provide supplies, education and information known to be effective in preventing the spread of HIV infection, at a relatively modest cost. 177. Burkina Faso places in the upper range of selected African countries as regards the knowledge o f modem contraceptives among married women, and past education efforts have increased the knowledge considerably (Table 45). Only four countries show knowledge rates higher than 90 percent, among them Burkina Faso. Moreover, knowledge increased by 27.2 percentage points between 1993 and 2003. However, compared with countries with similar knowledge rates, the usage o f modem contraceptives among married women still lags somewhat behind, despite an important increase since 1998/99. 32 GMPscanpreventthe early andmoderatemalnutrition. - 72 - Table 44. Participationof Non-MalnourishedChildren0-5 Years OldinMalnutrition Programs,2003 (Inpercent) Poorestquintile 2nd 3rd 4th Richestquintile Total A. Particioation in nutrition programs Total 8.9 16.7 16.4 24.6 24.1 17.9 Rural 8.6 16.0 15.4 23.7 21.5 16.3 Urban 13.4 22.5 22.1 28.5 27.7 25.2 B. Particbation in prowth monitorinporograms Total 32.9 45.2 42.9 48.1 60.3 45.6 Rural 30.0 41.4 38.0 41.0 43.0 38.1 Urban 74.1 80.6 74.4 78.6 83.3 80.0 Source: World Bank calculations usingthe 2003 priority survey. Table45. Knowledgeof ModernContraceptivesinSelectedAfrican Countries %ofmarriedwomen that: % ofunmarried, but sexually active women that: Year of survey have knowledge have ever are currently have knowledge have ever are currently of modem used modem using modem of modem used modem using modem contraceptives contraceptives contraceptives contraceptives contraceptives contraceptives Benin 2001 90.7 22.3 7.2 96.0 46.2 17.1 Burkina Faso 1992/93 63.3 10.0 4.2 84.6 42.0 16.0 Burkina Faso 1998/99 77.2 12.2 4.8 92.3 58.3 44.0 Burkina Faso 2003 90.5 19.5 8.6 Cote dIvoire 1998199 86.9 26.7 7.3 98.5 65.7 26.7 Ghana 1998 93.1 37.8 13.3 97.4 43.9 20.3 Guinea 1999 69.1 9.7 4.2 92.6 41.7 31.8 Mali 2001 76.2 20.7 7.0 89.1 41.4 22.0 Mauritania 2000101 68.2 12.9 5.1 72.0 17.6 5.3 Niger 1998 75.1 11.4 4.6 97.6 60.4 44.1 Nigeria 1999 61.9 18.5 8.6 89.1 49.1 35.4 Senegal 1997 82.5 17.3 8.1 96.6 61.8 45.6 Togo 1998 93.4 24.6 7.0 98.9 55.3 22.9 Source:Demographicand Health Surveys, several issues (last years available for selected countries), DHS (2003). Note: The sampleconsistsof women between 15 and 49 years old. 178. According to the 1998/99 DHS, knowledge of contraceptive methods is lower in rural areas but almost universal in urban areas, and it rises with income levels. Current use of contraceptives is higher among unmarried and sexually active women than among married women. For both groups, use of modem birth control methods rises strongly with wealth quintiles. - 73 - Health Care Utilization 179. General cure. Health care utilization provides important insights into the effectiveness o f the health system inoffering treatment to those who are ill,especially the poor. The evidence on utilization patterns comes from the 1998 and 2003 priority surveys. Self-reported morbidity refers in the following to the percentage o f individuals who report being illhjured in the last 15 days, and health care utilization as the fraction o f illhnjured individuals seeking health care during the same recall period.33The data is to be interpreted with caution as the poor tend to underestimate their morbidity and higher-income household tend to overestimate it. Similar observations apply to healthcare utilization rates. 180. The overall self-reported morbidity has decreased between 1998 and 2003 by almost 20 percent, from 7.1 percent to 5.8 percent (Table 46).34 This finding could indicate improved effectiveness of the treatment against diseases but could also be influencedby other factors since self-reported morbidity i s affected by individual characteristics, prices o f and access to health care service^.^' The increasing morbidity across expenditure quintiles in the incidence o f self- reported morbidity in both 1998 and 2003 and across u r b d m r a l locations may therefore also indicate that detection o f diseases and treatment is linked to higher education and incomes. The lower incidence o f self-reported morbidity among the poor would then largely derive from the "luxury good" nature o fhealthcare rather than the fact that the poor are illless often. 181. During 1998-2003, the gap between self-reported morbidity in urban and rural areas decreased, mainly because o f the sharp drop inurban morbidity (by 35 percent) compared to the drop in rural areas (by 11 percent). The urban areas also faced the most drastic reductions in morbidity across all expenditure quintiles. Overall in both rural and urban areas, the morbidity decline was largest among the bottom and the two top quintiles. 182. The health care utilization rate was 64 percent in2003, up from 42 percent in 1998. The health care utilization increased by 18.3 percent between 1998 and 2003. However, the changes in utilization were not uniform, but varied across socioeconomic groups and urbdrural areas. Utilization increases with income in both 1998 and 2003, but its overall distribution i s more equal across quintiles in 2003. Utilization inurban areas is higher than inrural ones, but the gap i s declining. Furthermore, the socioeconomic disparities by location changed, with a marked decline in the difference in utilization rate among the poorest and richest especially in urban areas. 183. Cost o f care seems to continue to be the main reason for the illto refrain from seeking care. Of those self-reported sicWinjured and not seeking care, 51percent mentioned the services 33 In 1998, only the self-reported illhick individuals were asked whether they sought health care. As opposed, in 2003 all the individuals were asked whether seeking health care, no matter o f their health status. It should be mentioned that 88% o f the individuals seeking health care in the last 15 days reported also being illhnjured during the same reference period, and we are referring to this sample inour further analysis. 34 Unfortunately, we cannot assess whether this trend continues or reverses the 1993-1998 one, as the recall period inthe 1993 survey was 30 days. 35 Dow et al. (1997) show that an increase in the price o f health inputs resulted in an improvement in the self- reported health status, despite o f unchanged health status when objectively reported. - 74 - were "too expensive," as only 39 percent considered it "not necessary" to seek care. Among the people thinking the services are too expensive, 51.5 percent were from rural areas. Table46. Incidenceof IllnesslInjuryand HealthCare Utilization,1998 and 2003 1998 2003 p~~~~~ 2nd 3rd 4th Richest Total ,,intile Poorest 2nd Richest Total quintile 3rd 4th quintile Rural 4.1 4.9 6.3 7.7 10.9 6.4 3.3 4.6 5.9 6.3 9.5 5.6 Male 4.5 4.5 6.2 7.5 10.7 6.3 3.5 4.5 5.4 6.3 9.6 5.5 Female 3.7 5.3 6.4 8.0 11.1 6.4 3.1 4.8 6.3 6.3 9.3 5.6 Urban 8.4 5.3 7.3 9.9 12.7 10.7 4.5 4.0 4.7 5.7 8.8 6.8 Male 7.6 5.6 7.3 8.2 11.7 9.9 5.1 4.4 4.3 5.0 8.0 6.3 Female 9.2 5.0 7.4 11.7 13.7 11.5 3.9 3.6 5.1 6.4 9.6 7.4 Total 4.3 4.9 6.4 8.1 11.7 7.1 3.3 4.6 5.7 6.2 9.2 5.8 Male 4.7 4.6 6.3 7.6 11.2 6.9 3.5 4.5 5.3 6.0 8.9 5.7 Female 3.9 5.3 6.5 8.5 12.3 7.3 3.1 4.7 6.1 6.3 9.4 5.9 Rural 27.9 28.1 35.5 44.3 52.4 39.0 43.5 56.7 60.8 67.5 71.1 61.6 Male 27.5 26.7 34.6 41.9 55.2 38.4 44.2 54.6 58.9 64.3 70.4 60.1 Female 28.3 29.3 36.2 46.4 49.9 39.5 42.8 58.5 62.3 70.6 71.8 63.0 Urban 23.8 35.3 43.1 44.9 53.5 49.3 58.8 50.9 49.9 71.0 77.6 71.4 Male 19.6 34.1 43.4 44.7 51.5 47.7 66.4 56.6 42.8 69.9 81.8 73.6 Female 27.5 36.6 42.9 45.0 55.4 50.6 49.7 43.6 55.5 71.9 74.1 69.6 Total 27.6 28.6 36.4 44.4 53.0 41.6 44.6 56.2 59.6 68.2 73.8 63.7 Male 27.0 27.4 35.8 42.4 53.3 40.8 45.9 54.8 57.3 65.2 74.9 62.9 Female 28.2 29.7 37.0 46.1 52.7 42.31 43.2 57.5 61.6 70.9 72.8 64.5 Source: World Bank calculations using the 1998 and 2003 priority surveys. 184. The increased health care utilization rates did not translate into a more intensive use o f . primary health care centers (CSPS). Although among the illseeking care the CSPS has been the most common health care option, its use decreased 15 percent between 1998 and 2003 (Table 47). Inrural areas, the choice o f CSPS decreased from 71.5 to 57.9 percent, and inurban areas it fell from 25.9 to 22.5 percent. This decrease is observed across expenditure quintiles in both rural and urban areas, despite the increase in the number of CSPSs over time. Some decrease in utilization also took place for C M N C M . Instead the sick sought care increasingly at traditional andprivate caregivers, and at regional and national hospitals. 185. Inbothrural and urban areas the illtumed increasingly to private care, whichbecame the primary healthcare choice inurban areas. Inrural areas the relative decline o f the CSPS was also accompanied by an increase o f traditional care, which in 2003 took second place as provider after the CSPS and before private care. The strong increase intraditional care in rural areas was mainly due to the rising demand inthe bottom three expenditure quintiles. Inurban areas, the use o f traditional care declined, and the illsought care increasingly in high-level public institutions and among private caregivers. 186. Overall, the priority survey shows clearly a reduction in the utilization o f the CSPS and CMNCM, which are the main pillars o f the public health system. The alternative choice is more expensive and higher-quality private care in urban areas and traditional healers for the rural population. - 75 - Table 47. Choice of Providers (Inpercent ofthose ilVinjuredand seeking care) 2nd quintile 3rd quintile 4th quintile Richest uintile Total :::::: i!! 1998 2003 1998 2003 1998 2003 1998 2003 1998 2003. ::: CSPS 76.2 42.7 72.1 59.5 73.1 55.4 68.2 64.1 71.6 58.8 71.5 57.9 CMNCM 8.5 4.9 10.6 5.4 9.8 10.6 9.8 7.0 RegionaYnational hospital 1.4 7.1 `100" 3.0 6.0 10.6 3.1 6.8 ~ ~ ~ ~ ~ Private 0.6 9.5 3.3 11.3 3.4 2.5 8.5 5.9 2.3 8.1 TraditionaYother 13.3 35.8 13.6 18.5 12.5 26.5 15.6 16.0 11.6 14.1 13.3 20.1 CSPS CMNCM RegionaVnational Urban hospital Priiate TraditionaYother ** ** ** ** 67.2 53.5 62.4 57.3 47.1 41.6 58.0 49.4 11.9 4.5 15.0 5.8 13.9 11.4 13.2 8.1 3.9 8.3 5.7 9.2 15.5 18.7 9.0 11.8 5.9 7.9 3.4 13.0 13.2 18.0 - -------- 7.8 13.4 11.1 25.7 13.6 14.7 10.3 10.3 12.0 17.3 . . YS. Note: *=the sample size is under 30, and therefore the averagesare not reported. 187. Antenatal care and assisted deliveries. Antenatal care and assisted delivery are a specific utilization o f health care facilities that i s closely linked with child and maternal mortality concems. According to the 2003 DHS, one-fourth o f mothers givingbirth did not have any prior antenatal care, and there are considerable differences across regions. In rural areas 29.6 percent did not receive any antenatal care, compared to only 3.3 percent in urban areas. 45.2 percent o f mothers giving birth in the North did not get any care, compared with 6.4 percent in the Center/South. 188. There are also large differences across income groups and between urban and rural areas inthe percentage o f assisted deliveries (Table 48). Only9 percent o fthe births inurban areas are delivered at home, compared to 64.4 percent o f births in rural areas. The frequency o f assisted deliveries rises in with expenditure (2003) or wealth (1998) quintiles in both urban and rural areas. This suggests that, despite the decrease in the cost o f deliveries, both the physical access and ability to pay represent factors in determining demand for assisted deliveries. However, compared with 1998, the use o f assisted deliveries increased, and the differences between income groups declined. Despite improvements, the availability o f doctors and midwives appears to still be a constraining factor for care in rural areas, as the differences in births assisted by professional staff between urban and rural areas in 2003 were still large, independent o f the income groups. - 76 - Table 48. Deliveries by Type of Assistance and Location (percentage of birthepisodes, five years before the survey) Poorestquintile 2"dquintile 31dquintile qthquintile Richest quintile Total A. Percentof deliveriesat home Total 79.9 76.3 76.9 67.7 23.4 67.4 1998199 Rural 80.0 76.3 77.2 68.8 46.5 73.8 Urban * * * * 7.5 9.3 Total 67.9 59.4 62.6 54.0 38.0 57.1 2003 Rural 70.1 62.4 67.7 62.6 54.9 64.4 Urban 20.3 16.0 14.8 7.3 4.7 9.2 B. Percentof deliveriesassistedbv doctors . Total 0.2 0.6 0.3 0.4 2.9 0.8 1998199 Rural 0.2 0.6 0.3 0.3 0.8 0.4 Urban * * * * 4.4 4.4 Total 0.5 0.8 0.7 0.7 2.1 0.9 2003 . Rural 0.5 0.8 0.5 0.5 0.9 0.6 Urban 0.7 0.5 2.5 1.9 4.4 2.8 C. Percentof deliveriesassistedby nurseslmidwives Total 17.7 20.2 21.6 31.9 72.1 30.2 1998199 Rural 17.6 20.2 21.5 30.7 49.7 24.1 Urban * * * * 87.4 85.8 Total 33.7 41.2 37.9 45.1 61.9 43.2 2003 Rural 31.4 38.0 32.8 36.5 46.7 36.2 Urban 81.7 86.9 86.1 91.5 92.2 89.9 Source: DHS 1998199,priority survey 2003. Notes: *=results are not shown due to small number of cases (or large variance, for the same reason). The quintiles are wealth-basedin 1998, while consumption-based in 2003. Health Expenditure 189. Public and private expenditures on health. In Burkina Faso, the health sector i s financed throughpublic andprivate channels. Public expenditures are financedby the regular government budget, HIPC funds, and donor contributions. The amounts budgeted to health were increasing duringthe 1998-2003 period. However, their share inthe total government expenditures declined somewhat in the past 2 years (Table 49). Total public spending more than doubled over the period owing to the incoming HIPC funds. 190. Private household expenditures, which include drug spending and other contributions that support CSPS operating budgets, continue to be an important part o f the health care financing process, but declined in relative magnitude as public spending rose. According to estimations based on the priority surveys, private health spending amounted to 182 percent o f government spending in 1998, but to only 109percent o fpublic spendingin2003. 191. Public expenditures on health in Burkina Faso are concentrated on primary care, namely on CSPSs and CWCMAs (Table 50). The total budgetary resources to primary level increased from 11,230 to 33,691 million CFAF between 1998 and 2003. This increase did translate into an increase in the share devoted to CSPS/CMAs from 44 percent in 1998 to 60 percent in 2003 (based on expenditure allocations in2003). - 77 - Table49. PublicExpendituresfor Health (millions o f CFAF) 1998 1999 2000 2001 2002 2003* Total public expenditures, ofwhich: 25,290 29,551 32,005 32,231 41,586 56,453 Govemmentbudget 19,414 20,505 22,546 25,263 25,866 31,795 HIPCfunds -- _- -- 1,893 10,008 16,319 Outside financing 5,876 9,046 9,459 5,075 5,712 8,339 Household expenditures (OOP)** 46,035 61,776 Government budget on health as Dercent of: Total govemmentexpenditures 8.7 10.8 11.0 10.7 10.3 Total D U b k expenditures on health as percent of: Total govemment expenditures 9.0 8.7 7.9 9.5 9.9 GDP 2.2 2.4 2.0 2.4 2.7 Sources: World Bank (2004b). Notes: *=Budgeted amount; **= World Bank estimates from 1998and 2003 priority surveys. Table does not include estimateofnon-householdprivate sectorspendingfor health(e.g., private insurance,healthservices financed by private companies, etc.). Table50. PublicExpendituresfor Health (millions o f CFAF) 1998 1999 2000 2001 2002 2003* ODerational Structures Third level 5,579 3,742 5,350 6,294 7,048 7,535 National hospitals (CHN) 3,216 2,380 4,152 4,194 4,607 4,841 Second level regional hospitals - (CHR) 1,608 1,583 2,719 2,813 2,916 3,177 Primary level -CSPS and C M K M A 11,230 16,669 17,858 15,025 23,734 33,691 Outsidefinancing 5,876 9,046 9,459 5,074 5.712 8,339 HIPCfunds _- _- _- 1,893 10,008 16,319 Total public healthexpenditures 25,291 29,551 32,005 32,231 41,586 56,454 Source: World Bank (2004b). Note: *=budgeted amount 192. Incidence of public spending on thepoor. Regardingthe incidence o f public spending on the poor, Table 51 shows that utilization rates o fpublic health institutions among the poorest are much lower than for the richest. Utilization rates in the first quintile are 52 and 60 percent, respectively o f the rates in the fifth quintile in the mentioned years. Second, the distribution o f users by expenditure quintile at either first or higher level o f care is very concentrated with users from the richest 40 percent dominating every level o f care. 193. The tendency for public health services to benefit higher-income groups to a larger extent than lower-income groups is reinforced by the previously mentioned increased use o f traditional services by lower-income groups between 1998 and 2003. In 1998, 81.4 percent o f the poorest sought care in CSPS/CMAs compared with 61 percent o f the richest, whereas in 2003, the preferences o f the poorest tumed towards the traditional sector and only 46.8 percent o f them sought care inCSPS/CMAs compared with 53 percent o f the richest. - 78 - Table 51. Distribution of Health Care Users inPublic Sector, by Level Poorestquintile 2nd quintile 3rd quintile 4th quintile Richestquintile 1st level (CSPS and CMMCMl 1998 9.1 10.7 17.7 26.6 35.9 2003 6.4 16.0 18.1 25.0 33.8 2nd and 3rd level (CHR and CHN) 1998 * 15.6 72.3 2003 ** ** 13.2 17.& 58.4 Source: World Bank calculations based on 1998 and 2003 priority survey. Note: *=sample sizes are under 30. 194. As presented above, the out-of-pocket expenditures still represent more than 50 percent of the total health financing. In order to protect the poor and ensure certain health coverage for vulnerable population against priority diseases, the Ministry o f Health developed several programs. They include programs against malaria, HIV/AIDS, tuberculosis, or programs aimed to improve the immunization rates, nutrition, or reproductive health. Some o f these programs, however, depend on extemal funding, and are therefore subject to uncertainty about how long they canbe offered, or how manypeople canbenefit from them. 195. Effect of private health expenditure on poverty. Since out-of-pocket expenditures are a frequent source of financing care, health spending may expose households to poverty either through foregone income or the displacement o f consumption. The analysis here is based on household level expenditures on health, and it follows Wagstaff and van Doorslaer (2001) in assessingthe effects of the second channel on poverty. 196. In 2003, the private spending on health amounts to 52.3 percent o f the total health financing.36 This situation originates from the fee-for-services approach o f the Bamako initiative "in order to achieve a degree of cost recovery." 37 However, inrecent years the user fees inthe public sector have been lowered particularly for consultations for prenatal and infant care, as well as for certain types o fdeliveries, but they were increasingfor general consultation^.^^ Public health insurance to cover cost o f care i s available for few, and mutual health organizations (MHOS)cover only a limited set o fhealthcare services. 197. The average Burkinabk household spends 2900 CFAF per month on health. Inthe 2003 priority survey, total spending on health comprises (1) payments for consultations, (2) costs o f medical tests, (3) drug expenditures, (4) cost o f hospitalizations, and (5) other health expenditures. Table 52 presents the average nominal household spending on health inthe month preceding the survey. Most o f these expenditures are on drugs (75 percent) and medical tests (10.6 percent). Households intke richest quintile spend on health 22 times more than the poorest households (7217 CFAF compared to only CFAF 335 among the poorest). Accordingly, health spending i s progressive, meaning that the health payments wield an equalizing effect on the income distribution. 36Total health financing include public andprivate financing, as well as HIPC funds and extemal contribution. The donor funds might be underestimated because of poor tracking information. 37See Edmond et al. (2002). 38See World Bank (2004b). - 79 - Table 52. Average HouseholdOut-of-pocketPaymentson Health,2003 (CFAFper month, nominal terms) Poorest .2nd Richest Total Concentration quintile 3rd 4th auintiie index Consultations 9 20 23 81 494 169 0.657 Medical tests 4 13 25 95 967 307 0.728 Drugs 305 800 887 1711 5090 2174 0.477 Hospitalizations 1 6 24 42 412 133 0.708 Other 16 11 81 107 253 115 0.487 Total 335 850 1040 2036 7217 2897 0.521 as % of total expenditures 1.2 2.5 2.6 3.3 5.5 3.4 Source:World Bank calculations usingthe 2003 prioritysurvey. 198. Household expenditure are concentrated on drugs, and drug costs represent almost the entire health spending o f the lowest expenditure quintile. Drug spending represent 91 percent o f health spending in the first quintile compared to 70.5 percent in the richest quintile. While the percentage spent on consultations and hospitalizations might reflect the lower utilization rates faced by the poor, the compositions o f total health expenditures hints to a substitution between drugs andprofessionalcare that could only be analyzed with more detailed data.39 199. To evaluate to what extent health spending contributes to impoverishing households, Table 53 considers welfare aggregate without health expenditures (denoted as "post-payment"), and including health expenditures ("pre-payment"). Health spending for this exercise is considered welfare-increasing spending on goods and services. The poverty lines are computed as 75 percent o f the pre- and post-payment median expendit~re.~' 200. Table 53 illustrates the impoverishing effects o f private spending on health for the 1998- 2003 period. By adding the health expenditures, the poverty incidence increases by 0.4 percent in 1998, and by 0.6 percent in 2003. The percentages o f people enteringthe poverty after the health payments (1.5 percent in 1998 and 1.4 percent in2003) are lower than the percentages o f people leaving poverty (1.9 percent in 1998 and 2.0 percent in 2003). The effects o f health spending on the poverty gap and the severity o fpoverty are small. DemandDeterminantsfor HealthCare 201. Demand determinants for health care are being explored with a probit analysis. The vector o f explanatory variables contains individual and household head characteristics, household characteristics (composition, location, socioeconomic ranking). The analysis is applied to the whole population, as well as separately for under-fives, 6-14 and above 15 year olds. The reason for splitting the sample is to further explore mortality factors for children and the young. 39The poor might choose to postpone therapies and interventions, substituting them with drugs, inorder to continue earning money. 40As mentioned inWagstaff et al. (2001), attention should be paid that the pre-payment poverty line will include a component for health expenditures, while the post-payment line will not. - 80 - Table 53. Effects of HealthCare Spending on Poverty 1998 2003 Povertv lines pre-payment 50285 57420 post-payment 48419 54942 Povertv headcounts pre-payment 29.5 32.5 post-payment 29.1 31.9 Effect on poverty -0.4 -0.6 Povertv eaDs pre-payment 7.6 9.1 post-payment 7.5 8.8 Effect on poverty -0.1 -0.3 NormalizedRoverhi eaos pre-payment 1.51 1.58 post-payment 1.55 1.60 Effect on poverty 0.04 0.02 Severitv of Dover@ pre-payment 3.0 3.6 post-payment 2.9 3.4 Effect on poverty -0.1 -0.2 % enteringthe pool of poor 1.5 1.4 % leavingthe pool ofpoor 1.9 2.0 % staying inthe pool ofpoor 27.6 30.5 Source: World Bank calculationsusingthe 1998 and 2003 priority surveys. 202. Income, household size, and schooling are key determinants o f the probability to seek care when sick. Income is positively correlated with the probability o f seeking care, especially for adults. Household size, gender (for 6-14 year olds), and secondary schooling for the household head also tend to increase the likelihood o f seeking health care. Age, having a female household head (for the adult sample), low educational status o f the household head and distance to a health center have the tendency to decrease the same likelihood. Urban location does not significantly enter the probability to seek care; neither do characteristics of a child's mother. However, there are important regional differences in seeking care that could indicate cultural differences inusingmodernhealth care facilities. C. SOCIAL PROTECTION 203. Social protection plays a key role inreducing vulnerability and protecting the welfare o f the poor, and the right mix o f social protection programs and policies in conjunction with interventions in other sectors and at the macroeconomic level can have a significant impact on poverty reduction (World Bank, 2002). The Bank's social risk management concept recognizes this multi-sector nature o f social protection and proposes that the most appropriate combination o f risk management arrangements (informal, market-based, or publicly provided) and risk management strategies prevention, mitigation, and coping) in any given situation depends on the types o f risks and on the costs and effectiveness o f the available instruments (World Bank, 2001). The main purpose o f this section i s to update and complement information contained in the recent Riskand Vulnerability Study (World Bank, 2004a). - 81 - Table 54. Probability of Seeking Care When Sick, 2003 All sample 0-5 y.0. 6-14 y.0. 0-14 y.0. 15+ y.0. zndquintile 0.039 0.005 0.024 -0.023 0.088* 3`* quintile 0.080** 0.046 0.090 0.060 0.095** 4`h quintile 0.123*** 0.091 0.097 0.071 0.154*** ..- Richest quintile 0.153*** 0.120* 0.180* 0.126** 0.182*** Age -0.002* -0.136*** -0.283*** -0.066*** 0.005 Age sq. 0.000 0.020* 0.014*** 0.003*** -O.OOO** 1:Female 0.021 0.016 0.121* 0.063* -0.001 Owneducation: primary incomplete 0.006 Owneducation: primarycomplete 0.057 Owneducation: secondary incomplete 0.115 Owneducation: second& complete -0.044 1:Hh. head is female -0.070* 0.017 0.063 0.062 -0.098* Age o f hh. head 0.001 -0.003 0.003 -0.002 0.001 Hh.headeducation: primary incomplete -0.071* -0.104 -0.048 -0.092 -0.047 Hh.head education: primarycomplete 0.043 -0.064 0.037 -0.039 0.08 1 Hh.head education: secondary incomplete 0.106 *** 0.184*** 0.100 0.147*** 0.070 Hh.headeducation: secondary complete 0.162** 0.306 0.206** 0.182 Hh.head education: higher 0.059 -0.221 0.224 0.105 0.099 Age o f mother -0.002 Mother's eduqation: primary incomplete 0.003 Mother's education: Drimarv comdete . . -0.018 Mother's education: secondary incomplete 0.061 Lrdhousehold size) 0.069*** 0.141** 0.069 0.136*** 0.074*** Share o fmales 0-5 y.0. 0.144 0.276 0.893 0.687** -0.097 Share o f females 0-5 y.0. 0.154 0.262 1.050** 0.540* 0.069 Share o f males 6-14 y.0. -0.019 0.531 1.355*** 0.878*** -0.176 Share o f females 6-14 y.0. -0.019 0.313 0.766 0.584** -0.010 Share o f males 15-24y.0. -0.056 0.373 1.120** 0.754** -0.171 Share o f females 15-24y.0. 0.000 0.388 0.684 0.592** -0.102 Share o f males 25-34 y.0. 0.125 0.870* 2.454** * 1.405*** -0.050 Share o f females 25-34 y.0. 0.167 0.774* 0.733 0.803** 0.023 Share o f males 35-44 y.0. 0.050 0.997* 1.785** 1.351*** -0.108 Share o f females 35-44 y.0. -0.108 0.05 1 1.193** 0.587* -0.277* Share o f males 45-54 y.0. -0. I40 1.248** 0.349 1.026** -0.334 Share o f females 45-54 Y.O. 0.036 0.421 -0.078 0.218 -0.006 Share of males 55+ y.0. 0.01I 1.535** 1.452 1.727*** -0.222 Distance to healthcenter: 0.5-1 hours -0.028 -0.072 -0.087 -0.065* -0.005 Distance to health center: 1-2hours 0.000 -0.038 -0.077 -0.042 0.034 Distance to health center: 2-12 hours -0.091** -0.094 -0.149 -0.104 -O.lOO* Distance to health center: more than 0.5 days -0.33 1*** -0.539*** -0.429*** -0.486*** -0.208* Distance to health center: no access -0.068 -0.086 0.135 -0.014 -0.113 1:Urban 0.027 -0.215** -0.029 0.060 0.005 -0.133** 0.034 Boucle de Mouhoun -0.122*** -0.053 -0.144** Sahel -0.1 12** 0.025 -0.198 0.017 -0.188*** East -0.007 -0.012 0.154 0.041 -0.072 South-West -0.159*** -0.177* -0.117 -0.176** -0.157** Center-North -0*102** 0.103 -0.238* -0.004 -0.169*** Center-West -0.058 -0.110 0.061 -0.027 -0.084 Plateau Central 0.039 0.033 0.I35 0.101 -0.019 North -0.222*** -0.234** -0.237 -0.203** -0.244*** Center East -0.004 -0.099 0.035 -0.013 0.001 Center -0.043 -0.003 -0.071 -0.006 -0.070 Cascades 0.003 0.039 0.106 0.094 -0.052 Center South -0.212*** -0.153 -0.211 -0.183 -0.227*** Observations 3199 716 444 1280 1919 Robustz-statistics inparentheses ~~~ *significant at 10%;** significant at 5%;*** significant at1% - 82 - Welfare andVulnerability 204. Compared to the poverty level, which i s an ex-post static measure of welfare, the concept o f vulnerability i s forward-looking and it measures the probability o f an individual or a household becoming poor (or poorer) in the nearest future. Vulnerability, as defined above, is a probability, a number between 0 and 1. For instance, if a household has vulnerability towards poverty o f 0.4, this means that it faces a 40 percent chance o fbecoming poor inthe future. 205. Based on the 1998 and 2003 priority survey data Lachaud (2003) analyzes the dynamics of vulnerability and concludes that the level o f vulnerability, i.e. the probability o f becoming poor, declined during the same period. The econometric analysis conducted for the study shows that between 1998 and2003 the level o f vulnerability towards poverty declined from 0.52 to 0.48 for the poor and from 0.35 to 0.33 for the non-poor segments o f the Burkina Faso population. The analysis shows that the share o f population with increasedvulnerability also declined during this period, from 0.70 in 1998 to 0.59 in2003.4' Main Sources of Risk and Vulnerability 206. The Bank's recently completed study on risk and vulnerability identifies the following main types o f risk inBurkina F a ~ o : ~ ~ 0 Economic risks - the economy o f Burkina Faso is heavily dependant on fluctuations o f world prices for cotton and livestock, the country's two main sources o f foreign exchange; 0 Regional instability -Burkina Faso, a landlocked country inWest Africa, totally depends on its neighbors for delivery of goods to and out o f the country, and this geographical dependence makes the country's economy vulnerable to any political instability in the region (therecent military conflict inCote d'Ivoire, for instance); 0 Natural risks and food insecurity - geographic location o f Burkina Faso is a source o f vulnerability by itself: located in Sub-Saharan Africa, the country is generally dry and its soil is infertile. A lack o f rain and water resources has a strong negative effect on agricultural output leading to shortages and food insecurity; 0 Health risks - poverty and poor health go hand by hand in Burkina Faso, where a combination of food insecurity, poor access to clean water and endemic malnutrition results inhighrates o fmorbidity and mortality. 207. A survey conducted in four villages in the center-north o f Burkina Faso shows that all households inthose villages were repeatedly affectedby different natural and social shocks, with droughthain shortage, famine/food shortage and sickness or death o f a household member being the most common (Table 55). The key shocks experienced by households were covariate shocks (drought and famine), while illness, death o f a family member, or death or loss o f animals were idiosyncratic shocks. The data also show that all households experienced a situation o f famine 4'These findings were derivedbasedon the original poverty lines publishedby INSDwhereas the previouschapters of this study are usinga poverty line that is consistent with a revised consumption aggregate. Varying the poverty line would influence the quantitative results, butnot the qualitativeresults of Lachaud(2003). 42World Bank (2004a). - 83 - and repeated drought inthe four years preceding the survey.43Evenifthe survey is not nationally representative, it's results, nevertheless, provide a useful insight on the type o f shocks experienced inmost remote parts o f the country and households' coping mechanisms. Table 55. Distributionof ShocksAccordingto Frequency (Frequency expressedinpercentagesamongthose experiencingthe shock) Source: INSD(2002). RiskManagementArrangements 208. There are two main risk management arrangements in Burkina Faso - informal or traditional andformal arrangements. In addition to this, the country also has numerous public risk management programs that address a variety o f socioeconomic risks. Given the limited scope o f formal arrangements and risk management programs, the majority o f Burkinabk households rely on informal/traditional methods o f coping with risks. 209. Traditional arrangements. Traditional risk management arrangements include actions, such as selling household assets (livestock, land plots, etc.), borrowing from neighbors or relatives, community or family arrangements, migration, sending children to work, and others. These actions are relatively easy to implement, but they are not sustainable because household resources are limited and they are also harmful in a sense that they reduce, often dramatically, households' longer term ability to cope with economic hardship. Furthermore, actions such as sending children to work instead o f school and arranged marriages at a very early age reduce human capital andthe future earning potential o fnew generations. 210. The results o f the four-village survey showed, for instance, that selling livestock was the most common risk coping mechanism, with almost a half o f households (48 percent) reporting selling livestock to survive (Table 56). Other most common mechanisms included (i) seeking assistance from extended families (13 percent), (ii)working more (10 percent) or finding a secondjob (6 percent), and (iii) borrowing from relatives or friends (5 percent). ~~ 43World Bank(2004a). - 84 - Table 56. Coping Mechanisms Used by Surveyed Households Coping MechanismEmployed Percentageo frespondents who used thls coping mechanism Sale of livestock 48 Helpfrom extendedfamily 13 Work more 10 Find a secondjob 6 Take a loan from parentsor friends 5 Other strategies 8 Didnothingto deal with shocks 1 1 Box 12: Crop Diversification as a Risk Reducing Strategy for Farmers Diversification has long been recognized as an essential strategy used by rural households indeveloping countries to deal with risks and shocks. Households diversify by engaging inboth farm and non-farm activities, and within farm activities, b y engaging in both crop and livestock production. Crop production itself can be diversified, as households may produce different speculations, some consumed by the household and others sold on the market. In Sub-Saharan Africa, and especially inWest Africa, the main cash crops tend to be cotton, cocoa, and coffee. Other crops include cereals such as rice, maize, sorghum, and millet,.as well as a range o f vegetables and fruits. The benefits of diversification are well-known. Since different factors affect differentcrops ina different way, the same expected income can inprinciple be obtained at a lower risk by engaging in the production o f multiple crops. For example, in a country such as Burhna Faso whose northern departments are part o f the Sahelian region, droughts are frequent. But a lack o f rain is not likely to have the same effect on all crops. Diversification should therefore help to mitigate the negative impact of weather shocks, while potentially also increasing expected income from crops. Also, the earnings obtained by households for different crops may change over time independently o f weather shocks, especially for cash crops whose prices tend to fluctuate on the international market. Again, diversification may protect farmers from such variations. Using panel data from Burkina Faso on crop production by farmers, and relying on a measurement framework taking into account the potential impact o f risk on household welfare, it was shown that crop diversification in Burkina Faso leads not only to a higher expected mean level o f income, but also to a lower risk. The impacts are relatively large, with a 10 percentage point decrease inconcentration, i.e., more diversification, leading to a gain in risk-adjusted income o f more than 20 percent. These results suggest that support for diversification among farmers would help for reducing poverty in a poor country such as Burkina Faso where households are often subject to shocks, including weather shocks. 3urce: Bardasi and Wodon (2004). 211. The 2003 Priority Survey data also provide useful insights into how households deal with risks. An econometric study that used the survey data to examine how crop diversificationcould serve as a coping strategy against tough weather conditions in Burkina Faso concluded that households can potentially increase their incomes if they diversify the crop range that they produce (Box 12). Another study that usedthe survey data to investigate the effects o f the recent crisis inCote d'Ivoire on Burkinabt householdsconcluded that even ifjust about eight percent o f households were affected by the crisis, these households belonged to the most poor and vulnerable segment o f the population (Box 13). 212. Formal social security arrangements. There are two formal social security institutions in Burkina Faso - CNSS (La Caisse Nationale de Skcuritk Sociale) and CARFO (La Caisse Autonome de Retraite des Fonctionnaires). While the CNSS provides services to both public and private sector employees and their families, the CARFO administers pensions and other social benefits exclusively for public sector employees and their families. CNSS and CARFO together cover about 172,000 people, or less thantow percent o fthe population. ~ ~~~ Source: Siaens and Wodon (2004).Box 13: The Effectofthe C8ted'Ivoire Crisis on Burkinab& Households who was affected? The Burkina Faso Priority Survey (2003) allows to study the consequences o f the crisis in Cote d'Ivoire on Burkinabk households as well as to analyze their coping mechanisms. Firsto f all, the data show that 92.5 percent o f households were not receiving remittances before the crisis, so they were not affected by the crisis. However, amongst the households that did receive remittances before the crisis, 78.5 percent reported being affected by the crises (82.5 percent inthe bottomquintile, versus 70.2 percent inthe top quintile). what were the consequences...? Almost two thirds (62.7 percent) o f the households reported having difficulties to pay for their food as a result o f the decrease inremittances. One household infour (25.1 percent) had to interrupt some type o f economic activity. Households also had to cut back on health care (11.6 percent), schooling (5.4 percent), travel (4.8 percent), and constructiodhousing (4.0 percent). As expected, the impact on the ability o f households to meet their basic food needs was larger for comparatively poorer households. ...and the coping mechanisms? The coping mechanisms deployed or considered by the households who used to receive remittances included: (i)working more (24.8 percent), (ii) obtaining assistance from social programs (16.7 percent), obtaining assistance from other family members (8.4 percent), retuming to C6te d'Ivoire (4.2 percent), migrating to another region within Burkina Faso (2.6 percent), migrating to another country (1.4 percent), or taking other actions (3.6 percent). Some 16.7 percent o f households declared that they would not take any action to deal with the crisis. Probably because o f a lack o f coping mechanisms and resources on which to rely in times o f difficulties, this proportion was higher among the poor (22.3 percent in the bottom quintile) than among better off households (15.3 percent inthe richest quintile). 213. The National Social Security Fund, CNSS (La Caisse Nationale de Skcuriti Sociale), administers three types of social security benefits: pensions, family allowances, and professional risk allowances. Administratively, the CNSS reports to the Ministry o f Employment, Labor and Youth. Everyday activities o f the hnd are managed by a Board o f Directors, which consists o f the representatives o femployers andemployees, retirees, and the Government. 214. The CNSS provides services to: (i) private and semi-public sector employees; (ii) certain categories o f public sector employees that are not covered by CARFO (such as temporaries, contractors, etc.); (iii) job trainees (pensions and professional risk allowances only); (iv) students o f technical professions (professional risk allowances only), and (v) private individuals (given that they had been previously with CNSS for a period of at least 6 months). Independentworkers are not covered by CNSS services. In2002, the CNSS had 23,055 contributing employers, which provided social security benefits to a total o f 138,288 people. 215. The CNSS is funded by contributions from employers and employees according to the rates in Table 57. These contribution rates are applicable to all eamings (salaries, bonuses, in- - 86 - kindpayments, etc.), except family allowances. There is a ceiling of 600,000 CFA for monthly earnings that are subject to social security contributions. In 2002, the total amount o f contributions receivedby the CNSS was 15,206 millionCFAF. Table 57. Social SecurityContributionRates Type of contributions Amount Contributors Pensions 11% of total earnings Equally shared by employers and employees Family Allowances 7% of total earnings Paidentirelyby employers Professional Risk Allowances 3.5% o f total earnings Paid entirely by employers 216. The CNSS provides three types of social security benefits -pensions (old age, disability and survivor), family allowances (prenatal, maternal, and child support) and allowances for professional risks (daily allowances, disability benefits, and limited health insurance). In 2002, the total amount of social security benefits paid by CNSS was 7,033 million CFAF (Figure 22). The largest share of the paidbenefits were pensions (5,100 million CFAF), which was followed by family (1,280 million CFAF) andprofessional risk allowances(653 thousandCFAF). 217. The CNSS is administeredby its mainoffice inOuagadougouand four regional branches, which all together consume almost a half of the fund's budget (Figure 24). In 2002, the CNSS spent more than nine million CFAF for administrative expenses, an amount larger than all benefits paidby the system. Figure22. CNSS: Contributionsand Benefits,2002 I1 7,526 I Family allowances Pensions Professionalrisk allowances 1 DConmbutions mPayments 1 Source: CNSS 218. Establishedby a government decree in 1986, CARFO (La Caisse Autonome de Retraite des Fonctionnaires, or Retirement Fund of Public Sector Employees) administers pensions and other benefits for public sector employees and military personal. Administratively, CARFO reports to three ministries (Public Function, Commerce and Finance), while its everyday activities are managed by an eight-member Board equally composed o f representatives o f the Governmentand employees. - 8 7 - Figure 23. CNSS BudgetExecution,2002 Office lease 2% Benefits Management 46% expenses 48% Sanitary and social action program 4% Source: CNSS 219. Inorder to be eligible for an old age pension, an individual mustbe of 63 years old and have at least 15 years of service in the public sector or military. Contributions to CARFO are made by the state (10 percent of total earnings) and employees themselves (8 percent o f total earnings). Ifa public sector employee i s temporarily detachedto another sector (private sector or NGOs, for instance), the share paid by the employer is set at 14 percent of total earnings. In 2003, the total amount of contributions to CARFO was CFAF 12,482 million (Table 58). Table 58. SocialSecurity Contribution Rates Year Contributions (million CFAF) Total Employee Employer 2000 5,313 5,712 11,025 2001 5,024 6,490 11,514 2002 5,140 6,585 11,725 2003 5,318 7,164 12,482 Year Oldage uensions Disabilitv pensions Survivor pensions Omhans pensions Numberof Total Numberof Total Number o f Total Numberof Total 2000 10,060 7,201 123 43 7,723 1,593 1 1,436 693 2001 10,925 8,088 137 46 8,415 1,800 10,730 673 2002 11,800 8,643 142 53 8,970 1,896 11,140 696 2003 12,445 9,334 161 57 9,605 2,067 11,510 739 - 88 - 221. Public Risk Management Programs.44In addition to traditional and formal social protection arrangements, Burkina Faso also has a wide range of public programs that address different kind of socioeconomic risks. These programs focus mainly on health, education, food security, the labor market, and social assistance program concerning women and children.45 Administered by several ministries, public programs are either universal or they address specific vulnerable groups (women, children, disabled, repatriates, etc.). Public programs are jointly funded from the state budget and by the donor community. Disaggregated hnding o f donor and government funds for each o f the programs i s not available. 222. A major policy challenge in Burkina Faso is the lack o f reliable data to quantify the degree o f risks and vulnerability faced by poor and to access existing interventions. Risk management programs may have a range o f different objectives, such as to increase a calorie intake or school attendance rates, andor to lower unemployment rates. For an analysis o f the effectiveness o f the programs, it is necessary to be clear about the objectives each program services, the intended target group, the number o f beneficiaries covered, and costs disaggregated by administrative costs and transfer costs. None of this information is currently available inthe country for any program. 223. Highdependence on extemal financing sources shows clearly that sustainabilityofpublic programs inBurkina Faso i s almost impossible. There i s no major public program inthe country that is financed entirely from the state budget. Should donor financing shrink due to some economic or political circumstances, Burkina Faso would need to scale down significantly almost all its public programs, including the planned social protection strategy (Box 14). Box 14: Burkina Faso Social Protection Strategy Since early 2002, the World Bank Institute (WBI) and Africa Region Human Development (AFTHD) have brought together government representatives, NGOs, and relevant actors o f several African countries, including Burkina Faso, invarious types of learning events on risk management and social protection. These events included a serious o f risk management courses organized via videoconference and four regional seminars held in Africa and Europe. Since 2002, the Bank has also supported organization o f three national social protection seminars held in Burkina Faso. The main purpose of these activities has been to help the country teams with the preparation of national social protection strategies. The latest draft o f the Burkina Faso social protection strategy presented at a regional social protection workshop organized in June, 2004, in Helsinki, Finland, is a result o f several years o f work by the national team. The document is quite advanced and it can serve as a good base for the final document. The draft outlines well main risks and vulnerable groups inthe county and it contains a good summary o f existing public risk management programs. The draft strategy, however, falls short o f solid analyses o f poverty and vulnerability, and the action plan also needs prioritization on mostly needed social protection programs. The national team is currently revising the strategy, which is expected to be presented by the Government to the NationalAssembly later this year. 224. Lack of critical minimum data base to assess targeting efficiency o f programs is another major issue in Burkina Faso. While estimated numbers o f beneficiaries for some programs are 44This section draws heavily on BurkinaFaso Riskand Vulnerability Assessment. 45See World Bank (2004a) for a detailed analysis and a list o f the programs. - 89 - known (health programs, for instance), there are many other programs that target certain geographical areas, but the numbersof beneficiaries cannot be directly estimated. There are also many cases when targeting does not exist and transfers go to bothpoor and non-poor groups. On the other hand, there are many programs that target small numbers of beneficiaries (micro- finance andjob creation, for instance) having limited impact onpoverty. D. CONCLUSIONSAND RECOMMENDATIONS Conclusions 225. Education. Analyses of the 2003 Burkina Faso Priority Survey show large income, gender andregionalinequalities inaccess to education facilities and education outcomes. Despite significant progress inrecent years, literacy and school enrollment rates in Burkina Faso remain low. Adult illiteracy is widespread, and it is especially common inrural areas where only one out o f five menand one out o f fourteen women canread and write. 226. Achieving universal primary education remains an enormous challenge for BurkinaFaso, with only 44 o f children of primary school age attending schools in 2003, and with large enrollment disparities between boys and girls. The secondary education enrollment rate i s only 16percent, and there are largevariations incompletion rates across urban and rural areas. Access to education facilities also remains an important issue: children in large urban areas have a relatively easy access to educational facilities, however in many regions of the country long distances to school are a deterrent factor in school attendance. As reported in the survey, expulsion from school and high cost of education are the two largest reasons for school absenteeism. 227. Returns to education are low in Burkina Faso, with only secondary or higher education having significant difference in an individual's income-earning potential. The incidence o f child labor, especially inpoor households, is widespread inthe country. 228. In2002, BurkinaFasospent 87.6 billionCFAF ofpublic funds (government, HIPC and donors) on education (about 165 million U S dollars). Primary education is the first priority for public funds. In addition to HIPC money and donor finds, Burkina Faso is also eligible for finding from the EFA Fast Track Initiative (estimated 20 million US dollars in2004-05). While the education system in Burkina Faso could benefit from any resources it can obtain, the country's current capacity to absorb additional funds and to implement the volume o f necessary activities remains a major concern. 229. The benefit incidence analysis o f public spending shows that poor households receive a proportionately smaller share of public funds than richer households because kids from poor households have low enrollment rates. Because o f gender disparity in enrollment, the public spending favors boys more than girls, especially for secondary education. Though education is not one of the most important expenditure items for BurkinabC households, the data suggest that private spending on education makesup about 15 to 25 percent of total spending. - 90 - 230. Health. Burkina Faso has poor health status indicators when compared with its neighbors and exhibits a strong link between health status and poverty. Mortality rates and malnutrition rates are higher inthe poorest quintiles and inrural areas. 231. Poverty is related to access to health care and health care utilization. The poor make less use o f health care and tend to seek care mostly in primary care facilities and from traditional healers. This finding can be partly explained by the out-of-pocket payments as the poor tend to spend most o f their health care budget on drugs and may choose relatively inexpensive care. However, overall, the study didnot find a major impact o fhealth spendingon poverty rates. 232. Use o f reproductive health, antenatal care and assisted birth remains low and unequal among income groups. The percentage o f births delivered at home and the percentage of deliveries by a trained person have increased over time, but remain very low for women inrural areas. Moreover, the usage o fmodern contraceptives i s still extremely low. 233. The outcomes o f public interventions on health are shaped by the interaction between poverty and a health system heavily depending on user fees and outside contributions. The available data point towards a regressive pattern o f expenditures as higher income groups tend to have higher probabilities overall to access public health care. 234. Social Protection. Despite the economic growtho f recent years inBurkinaFaso, poverty is still widespread in the country and households remain vulnerable to a wide range o f socioeconomic risks. The data show that vulnerability towards poverty has slightly decreased between 1998 and 2003. 235. Main risks faced by the population touch all aspects o f everyday life o f ordinary Burkinabk: from high risks of being infected by endemic diseases to droughts, food shortages and malnutrition. In addition to domestic shocks, Burkina Faso is equally vulnerable to shocks from outside the country, such as regional instability and large fluctuations interms o f trade. 236. Traditionally, informal risk management arrangements remain most important for Burkinabe households. The survey data show that selling domestic livestock i s the most common instrument to deal with a shock. Once own resources are exhausted, a household is mostly likely to seek help from neighbors or relatives than from the state. Formal social security systems are limitedto a small fraction o fthe population, and they are very expensive to run. 237. Public risk management programs are numerousbut the lack o f data makes it difficult to assess their effectiveness, and there is an urgent need to evaluate, streamline and consolidate the existing interventions by different Ministries. Recommendations 238. Education. Shortage o f schools and teachers, especially inrural areas, is a major concern. There is a need in Burkina Faso to carehlly plan and execute the school construction program without further delay and increase the capacity o f existing schools through provision o f multi- - 91 - grade and double-shift schools. Accompanying this construction program a sustainable increase inteachers is needed and further efforts to train existifigteacherscouldbemade. 239. There is low demand for education in Burkina Faso. Therefore, further proliferation o f public measures, such as free distribution o f textbooks andclass materials combinedwith careful tracking o f expenditure could lessen the private cost o f education and help increase demand for education. Involving parents' association and village chiefs in school decision making and management would likely also help increase demand for education. Girls' low school enrollment is still an issue and the government needs to broaden its female education awareness programs, andit should buildon the first experience with the programoffering free school kits for girls. 240. It is necessary to reflect further on the equity o f government spending on education. High unitcosts for secondary and higher education combined with the usebypredominantlybetter-off households o f these types o f education divert government spending from basic education. At the same time, given the low returns to primary education, expanding secondary education along with primary education would be necessary to eventually enhance the skills o f the Burkinabb population. The development o f a sector-wide MTEF that is under way should be used to reflect bothresource needs for all levels o f education as well as the equity o f government spending. 241. Health. Addressing inequalities in health status and health care requires improving the efficiency and equity o f health care delivery. Among these measures are (i) enhance the availability o f trained personnel and proper equipment in needy areas; (ii) introduce a system o f incentives for personnel in difficult areas and ensure the quality o f services; (iii) reduce costs o f preventive care and drugs; (iv) strengthen health center management; (v) improve and develop alternative health financing mechanisms, through community-based prepayment and health insurance schemes; (vi) consider strategies to stimulating the demand side o f health. The puzzling decline in health center usage rates over the past 5 years and the increased use o f traditional healers (rural areas) or private services (urban areas) should be a particular focus o f attention to ensure that government infrastructure and personnel investment are well targeted to health demand. 242. The national programs addressing vaccination, malaria, nutrition, HIV/AIDS should be strengthened and supported by more and stable sources o f financing. Besides these programs addressing major health issues, it might also be necessary to enhance programs tackling the provision o f health and nutrition services to the poor. 243. The evidence on reproductive health underscores the nature o f vulnerability among females. It would be useful to (i)disseminate lessons from least-risk maternity home program and further examine financing aspects; (ii)improve availability and use o f services for family planning and maternal care; (iii)examine possible external or alternative financing for such services; (iv) empower community change groups like COGES. 244. Health dimensions o f poverty demand appropriate tools to monitor and evaluate the trend inindicatorsover time. Measurement ofoutcomes is crucial to assess the impact of anypolicy or program in the country. Therefore, steps should be taken to (i) full adoption o f the intemational - 92 - definitions in computing (ii) improving the data collection, as important indicators such as mortality rates are discussed based mainly on DHS data; (iii)training the health personnel inthe use o f classification and registration. 245. Social Protection. Burkina Faso needs a national social protection strategy to address the severest risks faced by the poorest households inthe country. The magnitude o f vulnerability o f the poor in Africa, in general, and in Burkina Faso, in particular, means that it cannot be addressed through any one operation or type o f instrument alone, so the country needs to concentrate on a range o f different mechanisms with which to prepare for and cope with risk. A clear social protection strategy would help identify and prioritize different risks, develop an action plan, and identify key players. A clear vision on social protection would also lead to efficient use o f limitedresources and eliminate unnecessary competition for scarce internal and external funds between different ministries. 46See, for example, problems encountered with the matemal mortality rates. - 93 - VI. EQUITY OF FISCALOPERATIONS 246. Traditionally equity considerations in the area of fiscal operations are being looked at from a sectoral spending perspective through benefit-incidence analysis without much consideration o f the overall impact o f fiscal operations. Although the question of equity is difficult to tackle in a fully comprehensive manner, this chapter gives some indications from welfare theory as regards the possible "aversion" o f the population to certain reform policies and volatility of fiscal operations. It thus adds additional insights as regards the hture trade-offs involved in financing additional expenditure for social services for the poor and investments in infrastructure to create incentives for an acceleration of equitable growth. A. FISCAL DEVELOPMENTSCHOICES AND 247. The fiscal environment of Burkina Faso has been traditionally characterized by low revenue-to-GDP ratios and a strong dependency on foreign aid, Moreover, expenditure have a high share of wages and salaries and thus are relatively inflexible. This general set-up has changed little during the period 1998-2003 studied above. However, the government has made some choices on both the revenue and expenditure side that have resulted in important changes inthe composition ofrevenueandexpenditure. Table 60. Fiscal Revenues Total revenue and grants Total revenue Tax revenue Taxes on income Taxes on goods and services VAT Taxes on trade Import duties Other trade taxes Other taxes Non-tax revenue Capital revenue Grants received Project grants received Source:Burkinabk authorities and IMF. 2/ Changein the percent of total expenditureper annum 248. Onthe revenueside, there hasbeen an important shift towardtaxes on goods and services (Table 60). Following the introductiono f the common extemal tariff (TEC) in2000, the share o f taxes on intemational trade in total revenue and grants has declined markedly, while the value- added tax (VAT) has compensated for some o f the loss in trade taxes. This shift from customs dutiestoward the VAT, institutedunder a commonreformproject ofthe West African Economic - 94 - and Monetary Union (WAEMU), has been inspired by the interest in reducing distortions and included the unification of the VAT rate at 18 percent. Moreover, the government increased specific taxes on petroleum products in 2002, considered to be an efficient way to raise re~enue.~'Despite these tax reforms, the loss from the introduction o f the TEC was not fully compensated and the share o f tax revenue to GDP declined from 12.1 to 11.1 percent during 1998-2003. As regards income taxes, improved enforcement has resulted in a rising share of these taxes in overall revenue. The improved performance o f public enterprises, most importantly the government-owned petroleum import company, has also had resulted in rising dividends, boosting non-tax revenue. Finally, the level o f foreign g h t s in overall revenue has fallen slightly and an increasing share o f grants stems from direct budget support rather than project financing, exposing the treasury to larger risks from the timing o f budget grant disbursements. 249. On the expenditure side, the government has devoted an increasing amount of resources to priority sectors (see World Bank, 2004b and Table 61). These increases reflect the priorities of the PRSP to improve the provision o f social services, with resources being channeled to construction o f additional infrastructure and the hiring o f additional personnel. The economic classification of spending (Table 62) reveals that during this process spending has been redirected toward current expenditure and away from investment spending. Notably, materials and services and the wage bill increased faster than overall spending as the government began to hire more teachers and health workers and sought to improve equipment o f schools. The relative decline in investment spending has been the result o f falling importance of foreign-financed investments, for which the government has made up in part with domestically financed investments. Overall, the development reflect the combination o f a reorientation o f spending and the replacement of project support with budget support, which gave the government larger leeway inallocating resources between recurrent andinvestment spending. * Table 61. FiscalExpendituresby FunctionalClassification 1998 2002 98-02 Actual In % of In% of Actual In%of In % of Annual Level expend. ofGDP Level expend. ofGDP Total expenditure 363.8 100.0 22.1 486.4 100.0 21.8 Gen. public serv. 34.6 9.5 2.1 46.3 9.5 2.1 Defense 23.3 6.4 1.4 33.4 6.9 1.5 Education 47.2 13.0 I2.9 87.6 18.0 3.9 Health 25.3 7.0 1.5 40.8 8.4 1.8 SocialSecurity 1.7 0.5 0.1 3.3 0.7 0.1 l Economic affairs 14.2 3.9 0.9 17.5 3.6 0.8 Other (residual) I 217.5 59.8 13.4 257.5 53.0 4.31 - 3 4 Source: Burkinabk authorities..IMF..and World Bank (2004bl 1/ Change inlevelper annum 2/ Changein thepercentof total expenditureper annum 47The theory onoptimal taxation notes that commodity tax rates should be inversely related to the price sensitivity of consumers ofthese commodities. - 95 - Table 62. FiscalExpendituresby Economic Classification Total expenditure Current expenditure Wages and salaries Materials, services, and other Transfers and subsidies Interest payments Domestic interest payments 3.2 0.9 0.2 5.2 1.o 0.2 10.0 2.3 Foreign interest payments 9.5 2.6 0.6 11.1 2.1 0.5 3.2 -4.1 Capital expenditure 203.9 56.0 12.4 232.0 44.2 9.5 2.6 -4.6 Foreign financed 140.4 38.6 8.5 134.4 25.6 5.5 -0.9 -7.9 Domestic financed 63.5 17.5 3.9 97.6 18.6 4.0 9.0 1.3 Source:Burkinabt authorities and IMF. I 2/ Change in the percent of total expenditureper annum B. THEORETICAL CONSIDERATIONS 250. Welfare economics offers some theoretical foundations for the measurement o f the welfare impact o f fiscal policies. To measure the impact, an assumption has to be made about the so-called social welfare function that specifies how society values the income o f different members based on how adverse it i s to inequality. In what follows we adopt welfare fbnctions that yield scale-independent measuresproposed by Son (2003) and vary the degree o f aversion to inequality. Annex 6 describes details o f the derivation o fwelfare measures for policies that affect household income and expenditure. 251. Below the study considers two measures o f welfare for policy changes. Both measures evaluate income or price changes based on an average income level that generates the same social welfare as the actual income distribution. The elasticity o f this particular income level against changes in income components or prices indicates the sensitivity o f social welfare to changes in income components or prices. Son (2003) has derived two welfare indices from this elasticity. The first measure is an index that measures the social marginal benefit of an increase o f a certain income component by one currency unit (called welfare reform index below). A positive value o f this index indicates that increasing that income component is increasing social welfare. The second index assesses how social welfare changes inresponse to changes inprices o f a certain good or category o f goods (called price reform index below). C. EMPIRICALRESULTS 252. As regards the social welfare impact o f price changes, reforms affecting food prices have a larger impact than those for non-food items. Table 63 shows the values o f welfare elasticity and price reform index for household expenditure components for 1998 and 2003. By definition, because price increases reduce welfare, the welfare elasticity for an increase in prices is always - 96 - negative, while the price reform index ranking used to rank reforms can have positive and negative signs. The fact the price reform index is negative for all types o f expenditure indicates that in Burkina Faso there are no welfare-improving ("income-equalizing or progressive") increases inprices that could be exploited, at least not among the broad categories shown below. Also, in both household surveys and irrespective o f the inequality aversion assumed, it arises clearly that food price increases have the most detrimental impact on social welfare. Inaddition, from the 2003 priority survey it can be seen that the "price" increase o f subsistence food consumption ranks lowest inthe price reform index. By contrast, non-food items were by far the highest-ranking item inthe price reform index. Table 63. Welfare Elasticity and Price ReformIndex Aversionparameter= 1 Aversionparan :ter = 2 Percent Welfare Price Welfare Price Shares Elasticity ReformIndex Elasticity ReformIndex 2003 Total 100.0 -0.275 -0.575 -0.338 -0.478 FoodItems 29.1 -0.316 -0.675 -0.413 -1.189 NonFoodItems 41.4 -0.286 -0.068 -0.298 -0.1 15 FoodAuto-consumed 18.8 -0.274 -1.250 -0.399 -2.279 Non-FoodAuto-consumed -1.574 1998 Total 100.0 -0.66 -0.62 -0.81 -0.98 FoodItems 78.0 -0.82 -1.59 -1.01 -2.17 Health 6.9 -0.04 -0.56 -0.04 -0.46 Education 3.3 -0.02 -0.60 -0.02 -0.57 Rent 11.8I -0.11I -1.31I -0.15 -2 19 253. As regards the changes in the tax structure undertaken in the since the late 1990s, the results broadly confirm the thrust o f these reforms as being in line with fiscal equity considerations. In particular, the broadening o f the VAT at a single rate while exempting non- processed food items and the lowering o f import duties under the TEC is likely to have mildly improved fiscal equity by lowering barriers for some agricultural inputs and foodstuffs without taxing agricultural output. By the same token, the government's choice to tax petroleum products more heavily in 2002 to make up for some revenue losses o f the TEC seems to be justified as they fall under the non-food items with better welfare characteristics. Regarding subsistence agriculture, the findings also give additional impetus to the current discussion at the WAEMU level to lower the levies on imported fertilizer (especially those charged with the highest import duty) and agricultural production material as measures that could lower the implicit price o f food produced and consumed by farmers themselves, considered to be the most beneficial price reform according to the above index. Using the 1998 survey, and inline with some o f the finding o f the previous chapter, household spending on education and health has a relatively modest effect but still tends to disfavor lower-income households. 254. Regarding household income sources, Table 64 provides the values o f welfare elasticity andthe welfare reform index for different income components derived from the 2003 household survey. The welfare elasticities for different income component vary substantially. Agricultural wages have the highest national income share; the welfare elasticity o f agricultural wages are 0.429 and 0.648 for the aversion parameter equal to 1 and 2, respectively; that is, a 1 percent - 97 - increase in agricultural income leads to the increase in welfare by 0.429 percent for the inequality aversionparameter o f 1. Table 64. Welfare Elasticityand Welfare ReformIndex Share IIAversion parameter=l Aversion pars ieter =2 Welfare IWelfare Welfare Welfare Elasticity ReformIndez Elasticity Reform Indei Total 100.0 0.233 -0.767 0.3 13 -0.687 Agri Wages 39.4 0.429 0.091 0.648 0.647 Wages & Salaries 11.2 0.034 -0.695 0.004 -0.964 Private Sector 3.7 0.012 -0.672 0.002 -0.947 Other Private 4.7 0.032 -0.324 0.013 -0.722 NonAgric. Rev. 23.8 0.231 -0.030 0.227 -0.048 Leasing & Divid. 1.4 0.006 -0.572 0.001 -0.896 Transfer Urban BF 3.0 0.029 -0.059 0.028 -0.083 Transfer Rural BF 1.3 0.018 0.306 0.054 3.057 Transfer from CGte d'Ivoire 4.5 0.028 -0.376 0.01 1 -0.759 Transfer FCE 0.4 0.001 -0.8 15 0.000 -0.982 Transfer others 2.2 0.006 -0.731 0.001 -0.960 Transfer Public 3.7 0.020 -0.466 0.007 -0.808 Transfer Local 0.7 0.004 -0.454 0.004 -0.492 I I 255. The findings indicate that higher agricultural wages and (private) transfers to rural areas are equalizing and thus welfare enhancing. This finding is a reflection of the generally lower incomes in rural areas and the equalizing nature o f private transfer payments to rural areas. It also confirms from an income-distribution perspective to some extent the government's tax policy that i s imposing relatively heavy burdens on formal sector wages and salaries, and is seeking increasingly to hlly draw the non-salaried private sector into the surveillance o f the income tax system. 256. Worker remittances from CGte d'Ivoire have a relatively high welfare impact compared with other transfers except those to rural areas. But the fact that the welfare elasticity o f the transfers from CGte d'Ivoire decline from 0.028 to 0.011, respectively for the higher value of the aversion parameter implies that the income from transfers from CGte d'Ivoire are not much concentrated among the poorest class of households. The same applies to public transfers, which seem not to be targeted to the lowest-income households. This finding is consistent with the previously discussed need to develop a social protection strategy that is well targeted and affordable. D. CONCLUSIONSANDRECOMMENDATIONS 257. The broad conclusions that emerge from a consideration o f equity in fiscal operations are that the fiscal reforms undertaken by the government in the late 1990s appear to have been mildly equalizing. This effect arises from the reduction in the tax burden on imports and exemption o f unprocessed food from the domestic taxes (VAT). The change in the regime - 98 - reduced some o f the high barriers for agricultural inputs, although a heavy combined taxation from import duties andVAT remains for some processed goods. 258. On the expenditure side, the findings confirm, from an equity standpoint, the tax policy of the government. The current focus on tax collection among taxpayers in the formal and informal sector in urban areas would appear to be appropriate given the high impact o f agricultural wages on aggregate welfare. By contrast, more could be done to improve the poverty-focus of government transfer programs, andto facilitate remittances from abroad. 259. The recommendations fi-om the above findings cover different areas o f tax and expenditure policy: From an equity viewpoint the ongoing considerations on the WAEMU level o f the import duty regime for agricultural inputs and equipment would appearjustified given that welfare indices show the largest sensibility to the price o f self-consumed food items. Further efforts should be made to target government transfer programs to lower-income households rather thanreservingthem for those inthe civil service or formal employment. A regular review o f the distribution o f tax burdens and expenditure under future household surveys and the PRSP process would seem to be necessary. The annual PRSP review should include a more analytical discussion o f poverty developments and should be closely linked with the results o f the poverty assessment. - 99 - ANNEX 1: BURKINA FASOAT A GLANCE 3/4/05 I Sub. POVERTYand SOCIAL Bulxlna Saharan Low- Faso Ml.3 income Developmentdiamond. 2004 Population,md-year (millons) 12.4 703 2,310 GNI per capta (Aflas method, US$) 350 490 450 Lifeexpectancy GNI (Aflas method, US$ blfons) 4.4 347 1,038 Average annual growth, 199844 Population (%) 2.4 2.3 1.9 Laborforce (%) 2.0 2.4 2.3 ON1 Gross Most recentestimate (latestyear available, 199844) 3er primary apita enrollment Poverty (% ofpopulabn belownafonalpovertyline) 45 Urbanpopulabon (% oftotalpqpulat~n) 18 36 30 Lifeexpectancy at birth (pan) 43 46 58 Infant"dtity (per 1,000/we births) 107 103 82 Childmalnutritron (% of drildrenunder 5) - 34 44 Access to improvedwater source ~ccessto an improvedwater source (% ofpepulafon) 42 58 75 I Illiteracy (% olpepulabn age 15+) 35 39 Gross pnmaryenrollment (% of school-agepepulaton) 46 87 92 Burkina Faso Male 94 99 Low-incomeamuo Female 80 85 KEY ECONOMICRATIOSand LONG-TERMTRENDS 1984 1994 2003 2004 GDP (US$ billons) 1.3 2.2 4.3 4.9 Economicratios. Gross domesticinvestmenVGDP 20.4 18.3 20.5 Exportsof goodsand seMces/GDP 12.4 11.5 9.1 9.7 Trade Grossdomesticsawngs/GDP -2.0 9.0 6.8 8.0 Gross nationalsawngs/GDP 18.8 8.5 9.4 T Currentaccountbalance/GDP -1.6 -10.3 -11.3 InterestpaymentsiGDP 0.5 0.7 Domestic Investment Total debVGDP 31.7 51.1 39.3 35.9 savings Total debt seMce/exports 8.0 12.2 15.4 13.0 Presentvalue of debtlGDP 1 Presentvalueof debtlexports Indebtedness 1984-94 199444 2003 2004 2004.08 (average annualgrowth) GDP 4.8 5.9 8.0 4.6 4.7 - Burkina Faso GDP per capita 2.4 3.3 5.5 2.4 __ I nw-inmmr nmrm Exportsof goods and se~ces 2.0 5.8 9.8 15.3 6.5 I STRUCTURE of the ECONOMY (% or GDP~ Agnculture 32.9 31.1 31.7 30.6 I 5 Industry 21.9 19.3 20.9 21.5 ?a Manufactunng 3 Services 45.2 49.6 47.4 47.9 a Privateconsumption 873 76 1 609 793 gs w a i 02 03 01 General governmentconsumption 147 149 122 127 Importsof goods and seMces 303 229 205 222 ----GOi - O - G D P 1984.94 199444 2003 2004 (average annualgrowth) Growthof exportsand Imports (%) Agriculture 4.7 6.3 10.8 1.7 eu T Industry Manufactunng Services Pnvateconsumption 3.5 6.0 6.1 General government consumption 2.6 7.5 1.2 Gross domestic investment 0.11.6 11.73.5 4.0 9.9 10.98.5 -EXPOW -1mpons Importsof goodsand seMces ~ Note 2004data are preliminaryestimates Group data are for 2003 * The diamonds show four key indicatorsin the country (in bold)comparedwth its income-groupaverage If data are missing.the diamnd wlll be incomplete - 100- BURKINA FASOAT A GLANCE (CONT'D) Burkina Faso PRICESand GOVERNMENT FINANCE 1984 1994 2003 2004 inflation (*A) Domesticprices (Om change) Consumerprices 4.1 24.8 2.0 -0.2 ImplicitGDP deflator 25.5 0.9 1.3 Government flnance (?4 of GDP, includescurrent grants) 1 Current revenue 9.9 13.0 15.0 15.0 Current budgetbalance -0.9 1.9 4.6 3.7 -GDP deflator -0-CPI Overallsurplus/deficit -5.8 -4.1 -5.6 TRADE 1984 1994 2003 2004 (US$ millions) ExportandImportlevels(US$mill.) Total exports(fob) 191 321 434 1 Cotton 58 206 306 Livestockproducts 61 43 47 Manufactures Total imports(ci0 0 561 737 Food 0 83 95 Fuel and energy 0 134 192 Capital goods 0 222 289 Exportpriceindex (1995=7001 93 93 96 Importpriceindex (1995=700) 100 124 131 Tens of trade (1995=1001 93 75 73 BALANCE of PAYMENTS 1984 1994 2003 2004 (US$ millions) Pnentacfountbalanceto GDP (%) I Exportsof goodsand services 179 248 381 503 0 Importsof goodsand services 397 495 893 1,131 2 Resourcebalance -218 -247 -512 626 4 Net income N/A -14 1 3 8 Netcurrenttransfers 95 227 71 66 8 Currentaccount balance N/A -35 -440 -559 -io Financingitems (net) 153 552 578 -12 Changesin net reserves N/A -119 -112 -20 .14 Memo: Reservesincludinggold (US$ millions1 238 331 336 Conversionrate (DEC, locaWSfj 437.0 524.6 581.2 534.4 EXTERNALDEBT and RESOURCEFLOWS 1884 1994 2003 2004 (US$ mil/ions) ompositionof 2003 debt (US9 mill.) Total debt Outstandingand disbursed 410 1,129 1,675 1.768 IBRD 0 0 0 0 IDA 124 518 760 Error Total debt service 22 44 66 73 lBRD 0 0 0 IDA 1 6 15 Compositionof net resourceflows Officialgrants 81 212 283 158 Officialcreditors 41 80 91 Error Privatecreditors 0 0 Foreigndirect investment 28 0 Portfolioequity 0 0 0 0 World Bank program Commitments 7 91 160 - iBRD E Bilateral - Disbursements 13 79 74 15 D. Other mltilateral F Pnvate - Principalrepayments 0 3 10 .-IDA IMF G ~Shoft-term Netflows 13 76 65 interest payments 1 3 5 Nettransfers 12 73 59 DevelopmentEconomics 3/4/05 - 101- ANNEX 2: STATISTICAL ANNEX ON POVERTYMEASUREMENT This annex expands the technical discussion o f chapter 2 on the measurement o f poverty as applied in the poverty assessment and presents fbrther statistical information intabular form. As noted in chapter 2, constructing a comparable welfare measure, in this case aggregate consumption, requires makingchoices as regards coverage, normalization, price adjustment, and weighing of household members Coverage of the indicator (i) Few consumption items were excluded from the comparable consumption aggregate for theoretical reasons, notably the purchase o f durables, other household investment^^^ and bulky expenditures associated with ceremonies. These items represent investments made inthe current (recall) period that generate utility for the household duringtheir lifetime; bulky expenditures made at one moment in time, but used for a longer period o f time. The standard practice i s to include only their rental value as component o f current consumption, or - as in our case - to exclude them ifthe information needed to estimate this value is not available (Deaton and Zaidi, 1999 and Deaton and Grosh, 2000). The exclusion o f these items does not have a large impact on reported poverty, but reduces inequality. (ii) The components of the consumption aggregate that were not collected identically in the 1998 and 2003 surveys were also excluded. The following items fall inthis category: fish and fish products; other meats than poultry; vegetables; pr&ta porter; textiles (see Table 7). By using only a subset o f items recorded similarly inthe two rounds there i s little loss in precision, as this abbreviated consumption indicator still covers 84% o f the food consumption and 88% o f total consumption reported in 2003 (Table 8). The coverage o f the partial indicator is higher in 1998 (93% o f total consumption and 92% o f food consumption), indicating that the 1998 questionnaire captured, ceteris paribus, relatively less consumption than in 2003. There are two main reasons for the underestimation o f consumption in 1998. The 1998 questionnaire (a) does not record the consumption o f some items, such as school uniforms; and (b) includes more aggregated items which trigger higher recall bias, and hence less consumption being reported, in the case of fish, meat, vegetables, pr&t aporter and textile consumption. (iii) The remaining items were aggregated into 9 commodity groups (separately for purchases and autoconsommation or gifts): (a) food, beverages and tobacco; (b) clothes and shoes; (c) rent and utilities; (d) house maintenance; (e) health; ( f ) transport; (g) leisure; (h) health; and (i) items. These groups match the CPI sub-indices collected by INSD other Price Department for Ouagadougou city (see Table 9), used in the next steps to correct the differences inthe cost o f livingover time. 48Such as (i) purchases of cars, motorbikes or bikes (code 23 in 1998 and codes 225-227 in2003); (ii)purchases of television sets, refrigerators or freezers (code 25 in 1998 and codes 229-231 in2003); (iii) purchases of furniture or sanitary installations (code 27 in 1998 and 233 in 2003) and expenditures associated with ceremonies (code 22 in 1998 and 224 in2003). - 102- Table A2.1: Items recordeddifferently in 1998and 2003 (Excluded from the consumption aggregate) PS I11998 code CWQ 2003 code Fishand other see products I O 110,111 Meat(except poultry) 11 112, 113, 114 Vegetables 18 121,122, 123, 124 Pret aporter 15 215,216 Textiles 17 218,219 Schoolunformes not collected 1.3 Table A2.2: Coverage of the partial but comparableconsumption aggregate in 1998 and2003 inpercentoffull/non-comparable aggregate Welfare FulVNon- Partiall Aggregate Comparable Comparable Ratio Year (1) (2) -- 1 Percapitaconsumption 1998 109,827 102,568 93% 2003 130,734 114,846 88% Per capita food consumption 1998 59,753 54,992 92% 2003 70,700 59,613 84% Note: The full/non-comparable aggregate is constructedusingthe same methodology as the partiakomparable one, without excludingany of the consumptionitems which were either added or disaggregatedin 2003. Bothindicators do not include durablesor other householdinvestments. Normalization. Each component of household consumption is normalized into annual consumption. Keeping with INSD practice, we multiply by a factor o f 24 food consumption (purchases, autoconsommation and from gifts, with 15 days recall period), and by 12 all other elements (with 30 days recall period) except schooling (whose recall period i s the current school year). Tobacco consumption was added to the food component after it was annualized, as we use the same CPI for the whole commodity group (food, beveragesand tobacco). Price adjustments. As in many other countries, in Burkina Faso there are substantial differences in the cost of living both across space and time. Two adjustments are used to correct for such differences inthe purchasing power: (i) First, the regional price index constructed by INSD for each survey is used to transform household consumption from local to constant Ouagadougou prices (Annex 1 presents the value of the regional price indices usedin each o f the surveys), by dividing the value o f consumption expressed (reported) in local prices by this index. The index was constructedusing abasket o f 10 staple goods, o fwhich 6 foods and 4 non-foods. (ii) Second, each o f the 9 components o f household consumption estimated under A.(iii) are expressed in Jun-2003 constant prices using the corresponding price indices from Table 9. This method performs a more accurate adjustment than, for instance, the use o f a unique CPI for total consumption, as it implicitly uses the observed, household-specific consumption shares instead o f the CPI weights. The latter are computed from an older - 103 - consumption survey, and are based on the shares o f these commodity groups in total household purchases, not consumption. The most important implication is that the weight o f food is lower in the CPI (as it includes only purchases, without taking into account autoconsommation and gifts), compared to our method. As food prices are a volatile element of household consumption, the use o f commodity-group-specific price indices increases the accuracy of the adjustment. The two adjustments described above implicitly assume that prices are changing, over time, proportionally with the changes observed (only) inOuagadougou. TableA2.3: Ouagadougouconsumerpriceinflators,by commoditygroup (June 2003=100) EP I1 EP I11 CommodityGroup May-98 Jun-98 Jul-98 Aug-98 Apr-03 May-03 Jun-03 Jul-03 Food, beveragesand tobacco 103.5% 98.2% 103.2% 101.7% 117.4% 106.5% 100.0% 108.6% Clothesandshoes 113.6% 113.8% 113.8% 113.7% 99.8% 100.0% 100.0% 100.9% Rent and utilities 111.3% 112.4% 112.5% 111.4% 101.3% 99.7% 100.0% 101.3% Housemaintenance 101.6% 101.1% 101.0% 100.4% 99.9% 99.9% 100.0?/0 99.9% Health 108.2% 108.0% 110.5% 111.3% 100.9% 100.9% 100.0% 100.0% Transport 118.8% 119.7% 119.4% 119.7% 97.2% 98.5% 100.O?h 100.4% Leisure 103.2% 103.2% 103.2% 103.2% 100.0% 100.0% 100.0% 100.0% Education 110.7% 110.7% 110.6% 110.6% 100.0% 100.0% 100.0% 100.0% Other goods 113.6% 113.5% 113.5% 113.2% 100.0% 100.0% 100.0% 100.1% Total 109.2% 107.2% 109.3% 108.6% 105.2% 102.0% 100.0% 102.7% ALZjustments for household size. Keeping with INSD practice, there i s no adjustment for household size, i.e. for economies o f scale in consumption or for differences in the needs of children versus adults. The welfare aggregate i s expressed inper capita terms. The formula usedto construct the comparableconsumptionaggregate for 1998 and 2003 is: Realper capita consumption= [24 *F*Fcpi+l2 *(C*Ccpi+R *Rcpi+M*M,pi+H*H~pi-t-*Tcpi+L *Lcpi+O*OcpJ+S*ScpJ/ /[(Regional price deflator)/(Household size)] where: F=food, beverages and tobacco consumption, and Fcpi is the corresponding CPI index; C =consumption o f clothes and shoes, and Ccpi is the corresponding CPI index; R=rent and utilitiespayments, andimputedrents, andRcpiisthe corresponding CPIindex; M=articles for housingmaintenance, andMcpiis the correspondingCPIindex; H=consumptionofhealthservices andmedicines, andHcpiisthe correspondingCPIindex; T =transport expenditures, and Tcpi is the corresponding CPIindex; L=expenditures for leisureactivities, andLcpiisthe correspondingCPIindex; S = consumption o f schooling, and Scpi is the corresponding CPI index; 0=consumptionofothergoods or services, andOcpiisthe correspondingCPIindex. Poverty line. The poverty line is determined endogenously from the 2003 survey using the comparable per capita consumption aggregate, to generate the same poverty headcount as estimated by INSDusing the 2003 CWIQ survey and the full consumption aggregate. A poverty - 104- line o f 72,110 CFA Francs in June 2003 Ouagadougou prices replicates the poverty headcount o f 46.4% reported by INSD for 2003 (INSD, 2003). Additional statistical annexes Table A2.4: Assignment of the EPI1and I11non-food codes into the 9 commodity groups CommodityGroup PS I1 1998 PS I11(CWIQ) 2003 Included Excluded Included Excluded Food,beverages and tobacco 1!%food 22!+ food Clothes and shoes c 15-18 215-220 Rentand utilities c 1-8 201-208 Housingmaintenance c 9-14.26,28,29 25,21 209-214, 232,234,235 229-231, 233 Health health health Transport 20 222 Leisure 21 22 223 224 Schooling school school Other goods and services 24, 30,31 23 228,236,231 225-221 Table A2.5: RegionalPrice Deflators used in 1994 by INSD Regional Price Deflator Weights for the regional price deflator strate Total Food Non-Food Food Non-food Rent Total ouest 0.842 0.804 0.936 0.54 0.39 0.07 1.oo sud et sud ouest 0.863 0.825 0.947 0.54 0.40 0.07 1.oo 1.oo centre nord 0.888 0.890 0.927 0.55 0.34 0.11 centre sud 0.863 0.854 0.921 0.54 0.36 0.10 1.oo nord 0.979 1.004 0.973 0.66 0.30 0.04 1.oo 1.oo autres villes 0.970 0.962 1.010 0.40 0.50 0.10 ouaga bobo 1.000 1.000 1.000 0.36 0.51 0.13 1.oo Table A2.6: Regional Price Deflators used in 1998 by INSD Region Price Index ouest 0.83 nord-ouest 0.75 sahel 0.87 est 0.79 sud-ouest 0.69 centre nord 0.85 centre -ouest 0.74 centre 0.89 nord 0.78 centre-est 0.56 Table A2. 7: Regional Price Deflators used in 2003 by INSD Region Price Index Hauts Bassins 0.96 Boucle du Mouhoun 0.86 Sahel 1.01 Est 0.71 Sud Ouest 0.84 Centre Nord 0.80 Centre Ouest 0.82 Plateaucentral 0.97 Nord 0.91 Centre Est 0.84 Centre 1.oo Cascades 0.96 Centre Sud 0.89 - 105 - Table A2.8 Correspondenceof the codesof the consumptionitemsin the three surveys PSI1994 PS n 1998 PS 1112003 depensesscolaires I'anneescolairepassee I'annee scolaireen cours I'annee scolaire en cours h i s de scolarite qlOall(?) N N livres et foumitures qlOa21(6) N N h i s transport scolaire qIOa31(6) N N diversescontributionsscolaires qlOa416 _- cotisationsdes parentsd'eleves __ N N autrescontributionsscolaires N N uniformesscolaires N depensesdesante 30 demiersjours 30 demiersjours 30 demiersjours frais consultations qlOa5l (6) N N medicaments q10a61 (6) N N hospitalisation q10a71(6) N N autresservicesmedicaw q10a81(6) N N h i s &analysesmedicales _- N N depensesalimentaires 30 demienjours 15 demiersjours I 5 demiersjours 1 Riz 1 1 101 2 Milet sorgho 2 2,3 102, IO3 3 Mais 3 4 I04 4 Niebbt 4 5 105 5 Farines 5 6 106 6 Igname,tuber.et plan 6 9 109 7 Poissonet prodmer 7 10 110,111 8 Viandes et oeufs 8 11,12,26 112,113,114,115,132 9 Huiles 9 13 116 I O Arach. et plte dara I O 14.15 117,118 I 1 Tomate en pot I 1 16 119 12 Fruits 12 17 120 13 Legumes 13 18 121,I22,123,124 14Condiments et assais 14 19,20,2 1,22 125,126,127.1 28 15 Pain,galettes,confis 15 7 3 107,108 16Sucre 16 23 129 I 7Cafe,the.cacao 17 24 130 18 Produitslaitiers 18 25 131 19 Boisson non alcoolis 19 27 133, 136 20 Boissonsalcoolisees 20 28.29,30 134,135 21 Eau 21 31 137 22 Cola 22 32 138 23 Autres dep.aliment. 23 33 I39 24 Ensemblealiment. 24 (Ne peut pas detailler) 34 140 depenses non-alimenfaires 30 demiersjours 30 demiersjours 30 demiersjours 1 Charbonde bois I 1 201 2 Bois 2 2 202 3Gaz 3 3 203 4 Electricite 4 5 205 5 Bougie, petrole 5 6.7 206,207 6 Loyer 6 8 208 7 Telbhone 7 9 209 8 Domestique 8 10 210 9 Equipementde menage 9 27,28 233,234 I O Savon,produits dent I O 11,12 211,212 11Cosmetiquesoincorp 11 13 213 12 Habillem.cout.chauss 12 14,15,16,17,18 214-220 13 Cigarette-tabac 13 19 221 14Voyagestransport 14 20 222 15 Loisirs 15 21 223 16Ceremoniesdiverses 16 22 224 17 Materielroulant 17 23 225-227 18 Essence,lubrifiant, I 8 24 228 19 Investissement 19 25 229,230,231 20 Transfertsverses 20 30 236 21 sante 21 _-_ ___ 22 Autres depenses 22 31 237 4 Eauachett (fontaine, facture et frais debranchement) voire depensealimentaire 21 204 26 Rbarationd'equipementde menage: radio, televiseur, refrigerateur,congelateur 26 232,235 31 Autres depenses 237 - 106- ANNEX 3: POVERTY DETERMINANTSSTATISTICAL ANNEX TableA3.1: Deternunan&ofthelogarithmofpercapitaeqmditwqpercapitamcomeandassetscores ' Expenditure Expenditure Income Income Assets Assets Urban Rural Urban Rural Urban Rural -0.22 -0.19 -0.87 -1.39 -0.10 -0.29 [6.15]*** [2.81]*** 10.11]*** [12.55]*** :4.43] *** [7.451*** Sahel 0.07 -0.07 0.72 ' -0.16 -0.34 -0.38 /Est [1.591 [0.60] [7.84]*** [1.041 13.381 * * * [6.20] *** 0.11 0.32 -0.17 -0.37 -0.22 -0.20 [2.69]*** [3.40]*** [1.72]* [2.70]*** :8.88]*** [3.64]*** -0.16 0.05 -0.15 0.34 -0.34 -0.03 [3.85] *** [0.36] [1.65]* [1.631 13.161 *** [0.35] 0.20 0.52 0.40 0.19 -0.26 0.07 [5.25]*** [6.60]*** [4.75]*** [1.71]* 1I.14]*** [1.681 * 0.08 0.26 0.00 -0.46 -0.12 -0.12 [1.81]* [4.23]*** [O.Ol] [5.15]*** [4.82]*** [3.35]*** -0.10 0.10 -0.10 [2.26]** [1.021 [3.88]*** -0.26 -0.15 -0.51 -0.32 -0.12 -0.18 [6.71]*** [2.14] ** [5.42]*** [3.2 11*** [4.85]*** [4.79]*** -0.I 4 0.03 -0.09 -0.18 -0.10 -0.07 [3.5 81*** [0.38] [1.08] [1.81]* [4.06]*** [1.76]* -0.06 0.29 -0.27 -0.21 -0.04 0.01 [0.88] [8.24]*** [1.89]* [4.25]*** [0.87] [0.47] -0.06 0.07 -0.50 -0.46 0.01 -0.14 [1.091 [0.99] [3.90]*** [4.35]*** E0.201 [3.48]*** -0.29 -0.96 -0.10 [6.361** * [9.19]* ** [3.5 81*** ectaresof landowned 3.43 3.24 2.06 0.45 I.85 -0.07 [12.26]*** [2.86]*** [3.31]*** [0.27] 10.79]*** [O. 121 umber of young children -0.I 4 -0.30 -0.24 -0.36 0.03 -0.05 [12.79]*** [9.80]*** [8.70]*** [8.15]*** [4.12]*** [2.68]*** umberof young children squared 0.01 0.05 0.02 0.08 0.00 0.01 [7.16]*** [6.20]*** [4.77]*** [6.10]*** [1.781* [2.73] *** umber ofchildren -0.14 -0.18 -0.19 -0.20 0.01 0.03 [13.871 ** * [9.291 *** [8.15]*** [6.95]*** [1.66]* [2.52]** umber ofchildren squared 0.01 0.02 0.01 0.02 0.00 0.00 [7.43]*** [5.68]*** [4.47]*** [3.26]*** [1.OO] [0.91] umber ofadults -0.07 -0.11 -0.07 -0.09 0.05 0.05 [5.94]*** [7.21]*** [2.69] *** [4.32] * * * [7.91]*** [5.51]**# umber ofadults squared 0.00 0.01 0.00 0.00 0.00 0.00 [I.79]* [4.72]*** [1.15] [2.57]** [2.84]*** [1.33] 0.08 0.26 0.06 0.36 0.00 -0.02 [1.81]* [5.05]*** [0.61] [4.92]*** [O.Ol] [0.68] -0.26 -0.21 -0.45 -0.34 -0.21 -0.02 - 107- (Table A3.1) [5.43] * * * [4.11]*** [4.10]*** [4.69]*** ,7.1I]***[0.39] grationtemporaire(6 mois) 0.I6 0.21 0.10 0.16 0.09 0.10 [6.89]** * [6.65] *** [1.96]** [3.55] *** :6.28] * ** [5.613 *** grationtemporaire(12 mois) 0.11 0.19 -0.02 0.17 -0.06 0.04 [1.65]* [1.731* [O. 1I] [1.071 [1.461 [0.69] grationin Cote d'Ivoire 0.13 -0.09 0.3 1 0.01 0.03 -0.03 [2.86]* ** [1.13] [2.94]*** [0.06] [0.95] [0.63] -0.09 0.08 -0.28 0.21 -0.11 -0.11 [1.61] [0.97] [2.34]** [1.741* :3.49]*** [2.38]** ad has lowerprimary education 0.I4 0.04 0.07 0.15 0.15 0.07 [2.56]** [0.61] [OS71 [1.571 :4.44]*** [1.73]* eadhashigherprimary education 0.13 0.22 -0.10 0.25 0.21 0.17 [3.03]*** [5 SO] *** [1.10] [4.41]*** [8.25] *** [7.49] *** eadhas low secondaryeducation 0.24 0.30 0.34 0.32 0.41 0.32 [3.73]* ** [6.86]*** [2.42] ** [5.07] *** 10.471*** [12.821** * eadhas high secondaryeducation 0.53 0.41 0.27 0.46 0.61 0.40 [5.10]*** [7.60] *** [1.21] [5.86] *** [9.46]*** [12.94]*** eadhas technical education 0.44 0.71 0.69 0.93 1.27 0.61 [1.221 [7.50]* ** [0.79] [6.84]*** [4.93]*** [11.13]*** eadhas superioreducation 0.94 0.77 1.03 0.88 0.96 0.57 [5.51]*** [12.73]*** [2.78]*** [10.07]*** [9.34]*** [16.63]*** pouse has lower primaryeducation 0.20 -0.08 0.15 0.02 0.11 -0.06 [2.56]** [1.001 [0.92] [0.22] [2.17]** [1.44] Spousehas higherprimaryeducation 0.06 0.19 0.22 0.16 0.16 0.16 [1.041 [3.94] * ** [1.751* [2.27]** [4.37]** * [5.70]*** Spousehas low secondaryeducation 0.06 0.24 0.28 0.25 0.28 0.18 [0.63] [4.54]*** [1.301 [3.32]*** [4.49]*** [6.00]*** Spouse has high secondaryeducation -0.28 0.28 -0.07 0.28 0.16 0.18 [1.301 [3.501 * ** [O. 151 [2.42]** [1.24] [3.95]*** Spouse has technicaleducation -1.33 0.59 0.16 0.53 -0.15 0.24 [1.97]** [4.90]*** [O. 111 [3.08] *** [0.38] [3.41]*** Spouse has superioreducation -0.06 0.73 1.74 0.72 0.31 0.31 [0.09] [5.99] *** [1.271 [4.15]*** C0.761 [4.57]*** 0.18 -0.42 0.11 -0.58 0.15 -0.27 [1.641 [6.01]*** [0.48] [5.82]*** [2.31]** [6.74]*** ead is looking forjob 0.28 0.01 -0.36 -0.23 -0.20 -0.07 [1.88]* [O. 131 [l.lO] [1.881* [2.19]** [1.44] ead has asecondaryoccupation 0.01 0.11 0.29 0.27 0.08 0.04 [0.60] [2.32]** [6.04] *** [3.90] *** [6.06]*** [1.42] ead is underemployed -0.04 0.14 -0.19 0.12 -0.06 0.05 [1.201 [2.50]** [2.411 ** [1.41] [2.63] *** [1.531 eadis readyto work next month 0.13 -0.12 0.00 -0.18 0.05 -0.08 [3.521** * [1.84]* [O.Ol] [1.971** [2.36]** [2.15]** ead is employed in public sector 0.23 0.19 0.29 0.15 0.24 0.17 [1.971** [3.96]*** [1.19] [2.17]** [3.31]*** [6.21]**' 0.04 -0.12 -0.18 -0.06 -0.20 -0.09 - 108 - (Table A3.1) ~~ ~ [0.27] [1.82]* [OS21 [0.60] [1.97]** [2.33]** Headis afamily worker -0.19 0.04 -0.24 -0.22 -0.24 -0.05 [1.561 [0.29] [0.88] [1.231 [3.16]*** E0.731 Headis independent -0.16 0.02 -0.29 0.09 -0.27 0.03 [1.641 [OS51 [1.471 [1.37] [4.58]*** [1.26] Headis in industry 0.17 0.40 0.24 0.62 0.22 0.24 [2.64]*** [6.72]*** [1.72]* [7.27] *** [5.5 I]*** [7.18]*** Headis in services 0.31 0.44 0.68 0.63 0.38 0.28 [5.93]*** [8.571** * [6.27] *** [8.34]*** 11.98]*** [9.47]*** Spouseis working -0.3 1 0.03 -0.06 0.11 -0.43 -0.02 [1.14] 10.261 [O. lo] [0.70] [2.64]*** [0.39] Spouse is looking forjob 0.07 -0.04 -0.02 -0.18 -0.11 -0.04 [0.47] [0.45] [0.07] [1.591 [1.141 r0.781 Spouse hasa secondaryoccupation -0.01 -0.15 0.06 -0.02 0.00 -0.02 [0.41] [2.20]** [1.01] [0.20] [0.07] [0.60] Spouse is underemployed 0.03 0.05 0.00 0.01 0.00 0.01 [0.76] [OS71 [0.01] [0.06] [0.11] [0.27] Spouseis readyto work next month -0.02 -0.09 -0.01 -0.10 -0.04 -0.08 [O.Sl] [0.94] [0.06] [0.74] [1.20] [IS31 Spouseis employed in public sector 0.62 -0.11 -0.09 -0.07 0.39 -0.06 [1.761* [1.041 [O. 121 [0.47] [1.76]* [1.08] Spouse is paidby the task 0.89 -0.08 0.18 0.08 0.34 -0.20 [2.60]*** [OS21 [0.23] [0.36] [1.63] [2.18]** Spouse is a family worker 0.24 -0.02 -0.16 -0.27 0.45 -0.04 [0.86] [O. 181 [0.25] [1.791* [2.72]*** [0.66] Spouseis independent 0.18 -0.15 -0.01 -0.19 0.40 -0.09 [0.64] [1.581 [0.02] [1.451 [2.43]** [1.80]* Spouse is in industry 0.13 -0.02 0.18 0.05 -0.01 0.01 [1.471 I [0.17] [1.01] [0.37] [0.26] [0.25] Spouseis in services 0.14 0.08 0.16 0.16 0.03 0.07 [2.79]*** [1.17] [1.551 [1.541 [1.09] [1.79]* Constant 11.92 12.35 11.90 12.59 0.76 1.15 [180.75]*** [182.10]**' [78.83]*** [128.33]**' [18.641*** [29.84]**' Observations 2528 5599 2439 4616 2400 5277 R-squared 0.53 0.33 0.45 0.28 0.56 0.43 Absolute value o f t statistics in brackets. significant at 10%; * ** significant 5%; *** significant at at 1% - 109 - ANNEX 4: TECHNICAL ANNEX ONINEQUALITY AND PRO-POOR GROWTHMEASURES Notation Po is the poverty headcount; 6 isthepovertygapratio; P2 is the squarepoverty gap ratio; a isthe parameter ofinequality aversion. The higherthe value of a ,the greater the weight given to the poorest ; p is the meanincomeper capita income ; z is the poverty line ; Gis the Giniindex (the Excel version ofPAMS evalue only the inter-group Gini index) ; B is a poverty measure that is characterized by the poverty line, the mean income, andthe Lorenz curve (which i s a general measure of inequality): 0 =Q(Z,P,UP)) p corresponds to the bottomppercent ofthe population; p varies from 0 to 1. L(p)is the Lorenz function, it corresponds to the shareo fincome enjoyedby the bottom p percent o f the population. The Foster, Greer and Thorbecke (1984) proposedthe following class o f poverty measures: P, = J(-la z z - x f(x)dx O Z Kakwani (2000) totally decomposed the percentage change inpoverty as: 4=(A?,)m+(A?), hp, = Growtheffect +Inequality effect Poverty decomposition (Kakwani, 1993) Poverty elasticity for a 0 + 7 =- a[',-, -Pa1 ,weevaluatefor (i)a=1;and(ii)a=2 . Pa Interpretation: 7, is the pure growth elasticity. It measures the impact o f economic growth on poverty reduction when the inequality income (measured by the Lorenz curve) does not change. Inequality elasticity for a # 0 - 110- E, =??a + zp, ,we evaluatefor (i) =1;and(ii)a=2. a Interpretation: E,is the effect o f inequality on poverty. That is, ifthe economic growth leads to an increase inthe Gini index by 1percent, the poverty index will change by E, percent. The first term is always negative and while the second is always positive. Inequality growth trade-off (Kakwani, 2000) Kakwani (2000) proposed the Inequality-growth trade-off index (IGTI) as: IGTI=4, =--77, &a This concept was introducedby Kakwani (2000). The concept (i) Economic growth increase meanincome, which intumhas an impact o freducing poverty, (ii) Change inincome leads to new distribution o f income, that is the income inequality (iii) Mean income andincome inequality, each intum affect poverty. (iv) Now, the question is: What is the trade-off between inequality and growth (or mean income)?, Ifthe Gini index increases by 1percent, how much shouldbe the growth rate inorder that poverty does not increase? Interpretations (proposed by Kakwani, 2000): (i) If,for example, IGTIis equalto 3.0, itmeansthat a 1percent increase inthe Gini index will require a growth rate o f 3 percent inorder to offset the adverse impact o f increase ininequality, (ii) By following pro-poor policies, ifwe can reduce the Gini index by 1percent, then this policy is equivalent to having an additional 3 percent growth rate. This suggests that the larger the IGT, the greater will be benefits o f following pro-poor policies that would reduce inequality, (iii) IGTIis by definition the marginal proportional rate o f substitution (MPRS) between mean income andincome inequality. Derivation From the following decomposition o f the proportionate change inpoverty: -=77, dPa dp - dG - + E , pa P G Keeping the poverty at the same level (iso-poverty), that is -- -0=77, -+E, -,(or 7, -=-E, -)themarginalproportionalrateofsubstitution dP, dp dG dP dG pa P G P G (MPRS) between mean income and income inequality is: - 111- Sectoral Poverty Decomposition(Son, 2003) -- -WithingroupgrowthEffect + WithingroupinequalityEffect +PopulationShift pa pff (i) The first term measuresthe effect of growth on the percentagechange inpoverty, assuming all groups enjoy the same uniformgrowth rates. (ii) The secondterm on the righthand side takes into accountthe fact that the actual growth rates vary from one group to the other. (iii) The third term is the population shift effect. The shift inpopulationis deemedpro- poor ifits (corresponding) term is negative. Son (2003) pointedout that such a situation could happenifthe migrationflow is from a poorer to a richer region. Pro-poor growth (Kakwani & Pernia, 2000) (xiii) Formula used:Pro-poor growth index @(a) =-(a=lor 2 for 4and 7, (a) Pzrespectively), Where: (9 117w= 117g(a)+117, (a) (ii) ~ ( ais)the proportional changeintotal povertywhenthere is apositive growthrateof 1 percent (between period 1and 2) (iii) qg(a)is theproportional change inpovertywhen there is apositive growth rate o f 1 percent (between period 1and 2) provided that the relative inequality does not change. (iv) 117, (a)i s the proportional change in poverty when inequality changes but the the real mean income does not change (between period 1and 2). Kakwani & Pemia (2000) proposed Korea, Lao PDR, and Thailand case study that: a. If @(a)0 ,growth i s anti-poor; < b. If 0 4 @(a)0.33, growth is weakly pro-poor; I c. If 0.33 +4(a)I ,growth is 0.66 moderately pro-poor; d. If 0.66 4 @(a)1.O ,growth is pro-poor; and 4 - 112- e. If 4(a)21,growthishighlypro-poor. PAMS Estimation: I. q g ( a ) = q a (4=3 A 1 2+XI From: 2= q, - E, - dP dp + dG Pa P G Dividingbothside by -andthe expression interms o fdiscrete changes, one gets: dP P where: (i) AI, is the proportional change inthe Gini index follow. As mentioned above, A12i s the h estimate o fthe inter-group Gini index, therefore ~ ( a# )q(a) (ii) g,,isthepositivegrowthratebetween1and2 (iii) Let's define E,X = q(a) q(a)- Then, by substitution we get: q(a)= q, +E,(Al2 +x) g12 We therefore made the following estimates: tri,(a)=-34,+XI Pro-poor growth (Ravallion & Chen, 2003) Formula g,(p)= ([-""I -1)thegrowthrateofthepercapitaincomeatthepthpercentile Yt-1(P) InterDretation Ravallion and Chen state that: Ifg,(p)lies above zero for all p, then there is a first-order dominance o f the distribution at date t over t-1.Thehigher is g,(p)for all for all p, the greater is the reduction of poverty. Ifg,(p)> g,, then growth i s pro-poor. - 113 - Derivation This approach is based on the first-order dominance Some limitations Pro-poor growth (Son, 2003) Formula (following the above notations): y(a) = g,,#(a) y(a) i s called the "Poverty Equivalent Growth" or (PEGR) The author inferred that, if: y(a) > g,, then the "Growth is pro-poor" 0 0 thenthe growth is "Immiserizing" the "Poverty Equivalent Growth" is the growth rate that wouldresult inthe same level ofpoverty reduction as the actual growth rate ifthe growth process was not accompanied by any change in inequality. - 114- ANNEX 5: SYNTHESIS OF SIMULATIONEXERCICES 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Real GDP growth Baselinescenario 8.0 4.8 5.3 5.2 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Simulation 1, cottonexport = +20% 8.0 8.9 5.9 4.9 5.0 5.2 5.2 5.2 5.2 5.2 5.2 5.2 5.2 Simulation2, export price=+20% 8.0 5.4 4.7 4.8 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Simulation3, Non-exp.goods prod= +20% 8.0 10.4 7.3 6.1 5.1 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Simulation4, Public wages= +4% 8.0 4.8 5.3 5.2 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Simulation5, Adv weather= - 10%prim sect. 8.0 4.8 4.9 5.0 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Simulation6, change inprodstructure 8.0 4.8 5.3 5.2 5.2 5.2 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Primary sector Baselinescenario 10.8 1.8 4.5 4.5 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation I,cottonexport = +20% 10.8 21.8 4.9 4.4 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation2, export price=+20% 10.8 7.4 10.1 4.6 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation3, Non-exp.goods prod= +20% 10.8 21.8 5.2 5.0 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation4, Public wages= +4% 10.8 1.8 5.2 5.0 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation5, Adv weather = -10%primsect. 10.8 1.8 1.3 4.6 4.5 4.5 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Simulation6, change in prod structure 10.8 3.8 3.3 6.6 6.5 6.5 6.0 6.0 6.0 6.0 6.0 6.0 6.0 Secondary sector Baselinescenario 10.4 6.3 6.7 6.6 6.6 6.6 6.0 6.6 6.6 6.6 6.7 6.7 6.7 Simulation 1, cotton export = +20% 10.4 6.6 7.0 6.6 6.6 6.6 6.0 6.6 6.6 6.6 6.7 6.7 6.7 Simulation2, export price=+20% 10.4 6.8 6.9 6.6 6.6 6.6 6.0, 6.6 6.6 6.6 6.7 6.7 6.7 Simulation3, Non-exp. goods prod= +20% 10.4 6.3 6.7 6.6 6.6 6.6 6.0 6.6 6.6 6.6 6.7 6.7 6.7 Simulation4, Publicwages= +4% 10.4 6.3 6.7 6.6 6.6 6.6 6.0 6.6 6.6 6.6 6.7 6.7 6.7 Simulation5, Adv weather= -I0% primsect. 10.4 6.3 6.1 6.4 6.6 6.6 6.0 6.6 6.6 6.6 6.7 6.7 6.7 Simulation6, change in prod structure 10.4 6.2 6.6 6.5 6.5 6.5 5.9 6.5 6.5 6.5 6.6 6.6 6.6 Tertiary sector Baselinescenario 5.5 6.1 5.3 6.8 6.8 6.8 6.5 6.5 6.5 6.5 6.5 6.5 6.5 Simulation 1, cotton export = +20% 5.5 19.1 5.9 7.1 6.5 6.7 6.5 6.5 6.5 6.5 6.5 6.5 6.5 Simulation2, export price=+20% 5.5 8.0 5.6 6.8 7.0 6.9 7.0 6.6 4.8 6.5 6.5 6.5 6.5 Simulation3, Non-exp.goodsprod= +20% 5.5 -6.6 5.2 6.4 7.6 7.3 6.2 6.5 6.5 6.5 6.5 6.5 6.5 Simulation4, Publicwages= +4% 5.5 6.1 5.3 6.8 6.8 6.8 6.5 6.5 6.5 6.5 6.5 6.5 6.5 Simulation5, Adv weather = 10%prim sect. - 5.5 6.1 2.0 6.9 6.5 6.6 6.6 6.5 6.5 6.5 6.5 6.5 6.5 Simulation6, change in prodstructure 5.5 4.9 4.1 5.6 5.6 5.6 5.3 5.3 5.3 5.3 5.3 5.3 5.3 Poverty headcount Baselinescenario 46.4 44.1 42.4 40.3 39.2 38.0 36.0 34.8 33.8 32.4 30.7 29.7 28.7 Simulation 1, cottonexport =+20% 46.4 43.8 42.2 40.1 39.1 37.9 35.8 34.6 33.7 32.3 30.6 29.6 28.7 Simulation2, export price=+20% 46.4 44.1 42.4 40.3 39.2 38.0 36.0 34.8 33.8 32.4 30.7 29.6 28.7 Simulation3, Non-exp.goods prod= +20% 46.4 43.3 40.9 37.9 36.6 35.0 32.3 30.2 28.5 26.3 23.4 21.5 19.9 Simulation4, Publicwages= +4% 46.4 44.1 42.4 40.3 39.3 38.0 36.0 34.8 33.8 32.3 30.7 29.6 28.7 Simulation5, Adv weather = -10% primsect. 46.4 44.1 42.8 41.2 40.6 39.8 37.4 36.4 35.8 35.0 33.4 32.8 32.2 Simulation6, change inprod. Structure 46.4 43.9 `42.1 39.4 38.0 37.0 34.5 32.7 31.5 29.8 27.3 25.4 23.9 - 115- ANNEX 6: MEASURINGWELFARE Our measure of social welfare is drawn from Newbery and Stem (1987), Deaton(2000), and Son (2003). The Social welfare W i s specified as the average weighted sum o f the individual household indirect utility function uh = ~ ( x(p), , p) over Nhouseholds. (1) w = cukg(xk (p)),where g(xk is the relative size of the household k.It allows to evaluate N k=l the different welfares measures at the whole population level. g(x) is the number o f individual in N the household times the household weight relative to the p o p u l a t i ~ n ~ ~therefore ~ a n d g(xk) =1 k=l The idea is to define a welfare measure that is sensitive to income and distributional changes in order to capture the impact of fiscal policies. Following Atkinson (1970) and Son (2003), let x*be the equally distributed equivalent level o f income. x' is the level o f income that would result inthe same social welfare W if receivedby all households, that is, N (2) u(x*)= =c u C x k )g(xk (p)) k=I Additionally, let Ibe inequality index, and let ,ube mean income. N (3)p= x x k g ( x k ) a n d k=l (4)1=l--orX- x' =p(l-I) P Finally, in order to get a scale-independent measure o f inequality, Son (2003) proposed the following homothetic utility functions: =A+B- 1-E Xi-" for E # 1 u(xk)= A+ Blog(x,) for E =1 E is the degree of inequality aversion, E 2 0 The greater is a E ,the larger is the weight associated to the bottom class o f the poor household in the distribution5'. This definition with alternative values of the inequality aversion parameter, provides insight on the extent to which poor are favored by policies or exogenous shocks that impact individual income or e~penditure"'~~.Substituting (5) into (3) yields, 49This representation is consistent with the household data record for BurkinaFaso. 50The larger is E, the more policy is concemed about the poor o f the society. When tends to infinity, social preference is maximin, that is the policy maximizes the minimum level o f welfare. When E is zero, there is no aversion to inequality, inthat case, the social welfare is measured by the mean income o f society. 5'The g(x) are the weights associated to each household in order to extrapolate sample size level welfare measures to the household level. The sample level and the population level measures o f welfare are different since the g(x)'s are different across household. - 116- where Exp and log stand respectivelyfor the exponential and the natural logarithm functions53. Let vi( x )be the ith component of a representative household with a per capita income income equal to x. Let f,and pibe respectively the share of the ith income component relative to x and the mean income o fthe ith component. income By definition, one has the following relationships: M (7) x = vi( x ) where Mis the number of income components i=l N Utilizing the properties o f the Lorenz curve and the concentration curves, one can define the elasticity vi(and 77) o f X *with respect to pi(and ,u) as follow: Ifviisgreater(less)than1,thentheithincomecomponentissaidwelfaresuperior(inferior). The higher i s vi, the greater is the effect o f an increase o f the ith income component on the social welfare. Let g(xk)= akand from (9,(6) and (10)one has: for E = 2 for E =1 52 E is notestimated fromthe householddata, we will exogenously give the alternativevalues of0,l and 2. 53 (5) shows that the welfare measure invariant to any transformation o f the utility functions since x * is independent o f A and B. - 117- Welfare Reform Index a. Household revenue-based measure On the household income side, Son (2003) defines the welfare reform index o f the ith income component o f a representative household by, (12)4 =--17. 4i measures the social marginal benefit o f an increase in the ith component by 1unit o f income currency. When it is positive (negative) it implies that raising that income component has a larger (smaller) impact at the margin on raising social welfare than its share in total income, if 4i + 4j for i + j then raising the ith component is more favorable to social welfare than income the jthincome component. b. Household expenditure-based measure The welfare reform index assesses the marginal reform o f government's tax and expenditure policies in terms o f welfare. Indirect taxes and government subsidies are two important instruments o f public policy, that affect the prices faced by the households. This sub-section suggests an index of price reform measure. Assuming that uh = y(xh(p),p)is the indirect utility function and that M (13)x(p) = CPi4i(x)+s(x) i=l Where piand qi are respectively the price and the quantity o f the ithcommodity, and s(x) is the saving of the household.54 Using Roy's identity we get the elasticity viof X* with respect to Piby: and the price reform index i s defined by: (15)4i =-Pi4i ViP +I The index measures the marginal social benefit o f raising expenditure for the ith expenditure item. The index allows ranking price reforms (and thus taxation o f goods) as regards to their social welfare impact by identifying whether changing the price o f a good affects social welfare more or less than its share inhousehold expenditure. 54Note that s(x) does note depend on pi. -118 - REFERENCES Atkinson, T. (1970), "On the Measurement of Inequality,". Journal of Economic The0y,Vol. 2. Bardasi, E. and Q. Wodon (2004), Crop Diversijkation and Risk-Adjusted Farm Income in Burkina Faso, mimeo. Bigsten, A. and Shimeles, A. (2003), Prospects for `Pro-Poor' Growth in Africa, Economic Commission for Africa Economic PolicyResearchCenter, ESPD/NRP/2003/4. Bonnet, D. (2001), Malnutrition: A Subject-matterfor Anthropology. InKolsteren, P., T. HoerCe, and A. Perez-Cuerto ed., "Promoting the Growth and Development of Under Fives." Proceedingso fthe International Colloquium. Antwerp (Belgium). Deaton, A. (2000) Analysis of Household Surveys: A Microeconometric Approach to Development Policy. Handbook. The World Bank and The Johns Hopkins University Press. Deaton, A. (2003), How to Monitor Poverty for the Millennium Development Goals, Princeton University, mimeo. Deaton, A. (2003), Measuring Poverty in a Growing World (or "Measuring Growth in a Poor World,M E R Working Paper 9822; RPDS Working Paper 222, Princeton University. Deaton A. and M. Grosh (2000), "Consumption," in Grosh and Glewwe (eds.), Designing Household Survey Questionnairesfor Developing Countries: Lessonsfrom Ten Years of LSMS Experience, Chapter 5,2000, pp. 91-133. Deaton A. and S. Zaidi (1999), Guidelines for Constructing Consumption Aggregates for WelfareAnalysis, RPDS Working Paper 192, PrincetonUniversity. Dercon S., and P. Krishnan (2000), "Vulnerability, Seasonality and Poverty in Ethiopia," The Journal of Development Studies,Vo1.36, No. 6. Donvad A., J. Kydd, J. Morrison, I.Urey (2004), A Policy Agenda for Pro-Poor Agricultural Growth, WorldDevelopment, Vol. 32, No.: 73-89. Dow, W., P. Gertler, R. Schoeni, J. Strauss, and D. Thomas (1997) Health careprices, health and labor outcomes: Experimental evidence. Labor and Population Program Working Papers Series 97-01, RAND, SantaMonica, California. Edmond, J., A. Comfort, and C. Leighton (2002) Maternal Health Financing Profile: Burkina Faso. The Partnersfor HealthReformplus Project, Abt Associates Inc., Bethesda, MD. Fofack, H. (2002), The Nature and Dynamics of Poverty in Burkina Faso in the I990s, Policy ResearchWorking Paper No. 2846, The World Bank, Washington, D.C. Grimm, M. and I.Giinther (2004), How to achievepro-poor growth in a poor economy - The Case of Burkina Faso, mimeo. - 119- INSD (1996), Le Profil de Pauvretk au Burkina Faso - Etude Statistique Nationale, Ouagadougou. INSD(2000), Profil et Evolution de la Pauvrete`au Burkina Faso -Etude Statistique Nationale, Ouagadougou. INSD(2002), Survey on social risks and vulnerability, Ouagadougou. INSD(2003), BurkinaFaso: La Pauvrete`en 2003, preliminaryreport, Ouagadougou. INSD (2004), Burkina Faso :EnquZte De`mographique et de Sante` 2003, preliminary report, Ouagadougou. Kabore, T.H. (2002) Secteur de Sante`: RisquesPaludisme et VIH-SIDA, mimeo, Ouagadougou. Kakwani, N.(2000) "On MeasuringGrowth and Inequality Componentsof Poverty with application to Thailand", Journal of QuantitativeEconomics,Vol. Xx, No. xx. Kakwani, N. and Pernia, E. (2000), "What is pro-poor growth?," Asian Development Review, Vol. 18, No. 1. Kakwani, N and Son, H. (2002), Pro-poor Growth and Poverty Reduction: The Asian Experience, The PovertyCenter, Office ofExecutiveSecretary, ESCAP, Bangkok. Kakwani, N., Khandker, Sh., and Son, H. (2003), Poverty equivalent growth rate: with applications to Korea and Thailand,mimeo. Lachaud, J.-P. (2003), Pauvrete` et ine`galite` au Burkina Faso :Profil et dynamique, mimeo. Lanjouw, J.O. and P. Lanjouw (2000), How to Compare Apples and Oranges: Poverty Measurement, mimeo. Ministry of Economy and Development and UNDP (2003), Rapport Pays Suivi des Objectgs du Mille`nairepour le De`veloppement,Ouagadougou. Ministry of Health (2002), Annuaire Statistique/Sante`-Annie 2001. Direction des Etudes et de la Planification. Ministkre de la Sante. Ouagadougou. Newbury, D. andN.Stern (1987), The Theory of taxationfor Developing Countries, Oxford. Patemostro, S., J. Razafindravonona and D. Stifel (2001), Changes in Poverty in Madagascar: 1993-1999, Africa RegionWorking Paper Series No. 19, The World Bank. Pereira da Silva, L., Essama-Nssah, B. and SamakC, I.(2002), A Poverty Analysis Macroeconomic Simulator (PAMS): Linking Household Surveys with Macro-models, Policy ResearchWorking PaperNo. 2888, The World Bank, Washington, D.C. Ravallion, M. (1998) Poverty Lines in Theory and Practice, LSMS Working Paper 133, The World Bank, Washington D.C. Ravallion, M. and Chen, S. (2003), Measuring Pro-poor Growth, Policy Research Working Paperno. 2666, The World Bank. Siaens, C. and Q. Wodon (2004), International Migration, Conflict, and Remittances: The Impact of the Crisis in C6te d'Ivoire on Burkina Faso, mimeo. - 120- Son, H. (2003), "On Assessing the Equity of Governments' Fiscal Policies with application to the Philippines," Public Finance, Vol. 53, No. 3-4. Son, H.(2004), "A NoteonPro-PoorGrowth," EconomicsLetters, Vol. 82. Timmer, (2003), Agriculture and Pro-Poor Growth, Pro-Poor Economic Growth Research Studies, mimeo. Wagstaff, A,, and E. van Doorslaer (2001) Paying for Health Care: Quantzbing fairness, Catastrophe, and Impoverishment, with Applications to Vietnam, 1993-1998. Policy Research Working Paper No.2715. Washington, D.C. The World Bank. World Bank (200l), Protection Sector Strategy: From Safety Nets to Springboard, Social Washington,D.C. World Bank (2002), A Sourcebookfor Poverty Reduction Strategies, Washington, D.C. World Bank (2003), Santk et Pauvrete` au Burkina Faso: Progresser vers les objectqs internationam duns le cadre de la stratkgie de Iutte contre la pauvrete`, Human DevelopmentUnit,Africa Region, Washington, D.C. World Bank (2004a), Burkina Faso: Risk and Vulnerability Assessment, ReportNo. 28144-BUR, Washington,D.C. World Bank (2004b), Burkina Faso: TheBudget as Centerpieceof PRSP Implementation, Public ExpenditureReview, ReportNo. 29154-BURYWashington, D.C. - 121-