Report No. 30381-PK Pakistan Rural Factor Markets Policy Reforms for Growth and Equity November 19, 2004 Rural Development Unit South Asia Region Document of the World Bank 11 RURALFACTORMARKETSINPAKISTAN: PolicyReformsfor Growth and Equity TABLE OF CONTENTS Pakistanat a Glance............................................................................................................vu .. Acknowledgements .............................................................................................................. ix ExecutiveSummary ............................................................................................................. x I Introduction . .............................................................................................................. 1 Plan of the Report ....................................................... 1. .................................................. 3 I1 . AgriculturalPolicyAnd Sector Performancein Pakistan:An Overview ..........4 Agricultural Markets and Prices .......................................................................... 6 I11 . RuralLandMarkets: Institutionsand Constraints .............................................. Land Ownership and Utilization.......................................................................... 10 10 LandDistribution, LandPolicy and LandReform .............................................. 14 LandReforms...................................................................................................... -14 Land Administration, Land Titles and Contract Enforcement............................. 16 Patterns o f LandUse and Productivity by Farm Size andTenure Status ............19 IV. Water MarketsIn PakistanAgriculture ................................................................ Water Availability inPakistanAgriculture.......................................................... 24 24 Access to Water and Agricultural Productivity ................................................... 27 29 Implications for Agricultural Productivity........................................................... Groundwater ........................................................................................................ 30 Laws andRegulatory Framework for Water ....................................................... 31 V. RuralLabor Markets .............................................................................................. Structure o f Labor Use......................................................................................... 33 33 Sectoral Composition o f Labor Use..................................................................... 35 Trends inWages .................................................................................................. 36 Labor Mobility and Migration............................................................................. 38 Gender Issues....................................................................................................... 38 Child Labor .......................................................................................................... 40 Bonded Labor....................................................................................................... 42 GovernmentLabor Policies ................................................................................. 43 VI. RuralCreditMarkets 46 Structure o f Rural Credit Markets........................................................................ .............................................................................................. 46 Historical Trends.................................................................................................. 47 Repayment o f Loans ............................................................................................ 50 Structure of Interest Rates.................................................................................... 50 ... 111 Formal Sector Credit and Mechanization............................................................ 51 Farmer HouseholdAccess to Credit ................................................................... 52 Access to Credit ................................................................................................... 54 Micro-credit Programs and Savings Mobilization............................................... 54 Implications o f Credit Constraints for Productivity and Income......................... 59 VI1. ImplicationsFor IncomeDistributionAnd PovertyReduction The Puzzle o fPersistentRural Poverty and HighAgricultural Growth..............60 .......................... 60 Agricultural GrowthandRural Income Distribution........................................... 65 Factor Market Reforms........................................................................................ 68 Implications for Growth and Poverty Reduction................................................. 69 VI11 ConclusionsAnd PolicyImplications . ..................................................................... 70 Land..................................................................................................................... 70 Water.................................................................................................................... 75 Labor.................................................................................................................... 79 81 Implicationsfor Rural PovertyReduction........................................................... Credit.................................................................................................................... 84 Conclusions.......................................................................................................... 85 LISTOFTABLES Table 1.1, Poverty Estimates for Pakistan................................................................................. 1 Table 2.1. Agricultural GrowthinPakistan. 1959-60 to 2001-02 ............................................. Table 2.2. Area. Yield and Productiono f Major Crops inPalustan. 1999-00 to 2002-03 ........ 45 Table 2.3 Ratio o f Wholesale Prices (Lahore) to Support Prices. 1995-96 to 2001-02 .......... 7 Table 2.4. Subsidies inPalustan Agriculture: 1996-97 to 1999-00 (mnRs.)........................... 9 Table 3.1. LandUse inPakistan. 1990-91 and 2000-01............................................................ 10 Table 3.2. Pakistan: Distribution o f Land Ownership.............................................................. 11 Table 3.3 Table 3.4. Pakistan: Distribution o f Farms and Area by Farm Size (operated area) ................11 GiniCoefficient for OwnershipHoldings byProvince............................................ 13 Table 3.5. Share o f different crops intotal cropped area by size o f farm (1990 and 2000) ......21 Table 3.6. Recent Findingson LandProductivity and LandMarkets....................................... 23 Table 4.1. Average Water Balance o f the Indus River System................................................. 25 Table 4.2. Gross Revenue. Total Cost and Returns to Irrigation o fMajor Crops in Various Regions of Pakistan 2002-2003 ............................................................... 31 Table 5.1. Pakistan Rural Labor Force. 1998 (thousands) ........................................................ Table 5.2. Labor Force Participation, Unemployment and Underemployment Rates...............34 34 Table 5.3. Percentage Distribution ofRural EmployedPersons by Major Sectors and Provinces (1999-00) .............................................................................................. 35 Table 5.4. UnderemploymentRates by Occupation (1999-00) ................................................. 35 Table 5.5. Growth inReal Wage Rates inPalustan, 1984-2002 ............................................... 36 Table 5.6. Female Labor ForceParticipation andUnemployment inPahstan, 1999-2000......39 Table 6.1. Formal Sector Banking (bn2000-01 Rupees) .......................................................... 46 Table 6.2. Sources o f Credit inPalustan, 1995-96 .................................................................... Table 6.3. Zarai Taraqiati Bank Ltd. (ADBP) Lending, 2002-03 ............................................. 47 49 Table 6.4. Palustan: Interest Rates inRural Credit Markets by Source .................................... 51 Table 6.5. Outstanding Loans by Tenure Categories (Formal and Informal) ........................... 53 Table 6.6. Average Size o f Loans to FarmHouseholds, 2002-03 ............................................. 53 iv Table 6.7. Credit Rationing inRural Pakistan2001.................................................................. 55 Table 6.8. Major Micro-credit Programs inPakistan, 2003 ...................................................... 55 Table 7.1. Real Prices of Major FoodGrains inPakistan, 1970-2003 ...................................... 61 Table 7.2. Rural Poverty Across HouseholdGroups, 2002-2002 ............................................. 62 Table 7.3. Rural HouseholdIncome by Source and ExpenditureQuintile (2001-02)...............63 Table 7.4. Palustan Rural Agricultural Incomes........................................................................ 65 Table 7.5. Implications of Changes inEconomic Structure for Agricultural Multiplier Effects.................................................................................................................... 67 Table 7.6. Simulated Distributional Impact of Factor Market Reforms.................................... 68 Table 8.1. Characteristics of Factor Markets inPalustan.......................................................... 71 Table 8.2. Major Linkages of Key Factor Market Distortions .................................................. 72 Table 8.3 Palustan Rural Factor Markets: Current Government Policies andPrograms and Possible Reforms............................................................................................ 86 LISTOF FIGURES Figure 2.1. Wheat Prices inPalustan, 1995-2003.................................................................... 7 Figure2.2. NominalRates of Protectionof Major Crops........................................................ 8 Figure3.1. Pakistan: Percentageof Farms and Area Cultivatedby LandTenure Status.......13 Figure 3.2. Pakistan: FarmIncome per CultivatedArea inPunjab, 2000/01......................... 20 Figure 4.1. Area Irrigatedby Source of Irrigation, Palustan 1982-2002................................. 25 Figure 5.1. Real Wages inPalustan: 1984-2002..................................................................... 37 Figure 5.2 Child Labor inPakistan, 1999-2000 ..................................................................... 40 Figure 5.3. Child Labor Force Participation Ratesby Province, Palustan 1999-2000............41 Figure 6.1. Growth of InstitutionalAgricultural Credit (mn2000-01 Rps.) ........................... 49 Figure6.2. ADBP/ZTBL Tractor Loans land Domestic Tractor Production, 1986-87 -2002103 ............................................................................................. 52 Figure6.3. Formal and Informal Credit by Tenure Categories, 2001-02................................ 53 Figure6.4. Credit Rationinginthe Formal Market................................................................. 56 Figure6.5. Credit Rationinginthe Informal Market............................................................... 56 Figure7.1. Real Prices of Major FoodGrains inPakistan, 1970-2003................................... 61 Figure7.2. Rural HouseholdIncome Sources by Quintile, Palustan, 2001-02 ....................... 63 Figure7.3. Palustan: Per Capita Incomes and Expenditures 2001-02.................................... 64 Listof Text Boxes Box 3.1 Land Price Determination inPakistan: LandRents and Asset Effects......12 Box 3.2. Historical Rural LandTenure Systems inPakistan' ............................................. 15 Box 3.3. LandPolicies and LandReform inPalustan ......................................................... 16 Box 3.4. LandRightsinPunjab and Sindh.......................................................................... 18 Box 4.1. The IndusBasinIrrigation System........................................................................ 26 Box 4.2. Administrative Reforms inCanal Water Management......................................... 28 Box 5.1. Child Labor inRural Palustan............................................................................... 41 Box 5.2. BondedLabor inPakistan..................................................................................... 42 Box 6.1. Recent Reforms inFormal Sector Agricultural LendinginPalustan.................... 48 Box 6.2. Pakistan Poverty Alleviation Fund(PPAF)' .......................................................... 57 Box 6.3. Aga Khan Rural Support Programme (AKRSP) ................................................... 58 Box 7.1. Agricultural Growth and Rural Poverty................................................................ 66 Box 8.1 Improving the Enforcementof Contracts inLandAdministration in Palustan.............................................................................................................. 73 Box 8.2. Examples o f LandReform and LandTitling Projects........................................... 74 V Box 8.3. Important Issues Regarding Tradable Water Rights............................................. Box 8.4. InfrastructureConstraints for Development o f Water markets inPakistan ..........77 78 Box 8.5. Factor Market Reforms and Agricultural Diversification inChile ....................... 79 Box 8.6. Factor Market Reforms and Labor Mobility inPoland......................................... 80 Box 8.7. Bonded Labor and Reforms inNepal.................................................................... 81 Box 8.8 Micro-Finance for Agriculture .............................................................................. 83 Box 8.9 Suggested Guidelines for ProvidingMatching Grants for Income-generating Activities ............................................................................................................ 84 ANNEXES Annex 1:Credit Access andAgriculturalProductivity .................................................... 88 Annex 2: Determinantsof FarmRevenueIncomeinPakistan-Revenue Functionsfor PakistaniFarmHouseholds 98 Econometric Specification.............................................................................................. ....................................................... 101 Main Findings................................................................................................................. 102 Human Capital ................................................................................................................ 102 FarmOperation Size ....................................................................................................... 105 Land Ownership.............................................................................................................. 106 Family Size ..................................................................................................................... 107 Surface Water ................................................................................................................. 108 Credit .............................................................................................................................. 109 Other Results................................................................................................................... 111 The Effects o f Tenancy Type.......................................................................................... 111 Annex 3: DistributionalImpactsofAgriculturalGrowthandFactorMarket Reforms ................................................................................................................. 113 The 2001-02 Pakistan S A M............................................................................................ 113 Modeling Framework...................................................................................................... 118 Simulation 1: Output Shocks to Crop and Livestock Agriculture ................................. 120 Simulation 2: Shift from Share-Cropping 122 Simulation 3: Removal o f Credit Constraints ................................................................ to FixedRents........................................... 124 127 Conclusion ...................................................................................................................... Simulation 4: ImprovedDistribution o fWater .............................................................. 129 Table A1.1 List o f Variables and Descriptive Statistics .......................................................... 89 Table A l . 2 Probit on Access to Formal Credit ........................................................................ 91 Table Al.3 Probit on Access to Informal Credit...................................................................... 92 Table Al.4 Probit on Access to Formal and InformalCredit .................................................. 93 Table A l . 5 Credit Access and FarmProductivity ................................................................... 94 Table A l . 6 Credit Access and FarmProductivity (IV Estimates) ........................................... 96 Table A2.1 Canal Water Usage and Transactions by District.,............................................... -99 Table A2.2 The Distribution o f Farm Size and Land Ownership (Acres) ............................... 100 Table A2.3 Net FarmRevenue RegressionResults (Dependent Variable: Net Farm Revenue (Rs)....................................................................................................... 103 Table A2.4 Elasticitieso f Household FarmRevenue with Respect to Quasi-fixed Factors................................................................................................................ 105 v1 Table A2.5 Return on Credit (Elasticities) ............................................................................... 110 Table A3.1 Pakistan: HouseholdIncomes and Expenditures, 2001/02 .................................. 116 Table A3.2 Pakistan: Factor Income Shares by HouseholdType 2001/02............................. 117 Table A3.3 Disaggregated Factor Incomes Shares by Household Group, Pakistan 2001/02............................................................................................................... 117 Table A3.4 Pakistan Rural Agricultural Incomes .................................................................... 118 Table A3.5 Simulation Results: 10Percent Output Shock to Major Crops and Livestock............................................................................................................ 121 Table A3.6 Share o f HouseholdGroup Are Cultivated by Tenure Status, 2001/02 ................123 Table A3.7 Share o f Area Cultivatedby Tenure Status and Household Group 2001/02............................................................................................................... 123 Table A3.8. SimulationResults: Percentage Change inHousehold Incomes.......................... 124 Table A3.9 Estimates o f Productivity Gains from Removal o f Credit Constraints .................125 Table A3.10 Distribution o f Estimated Returns to Small FarmIrrigatedLand (Major Crops), 2001/02.................................................................................................. 126 Table A3.11 Crop Yields by Location Along Canals (tons/hectare) ......................................... 127 Table A3.12 Simulation Results: Household Incomes.............................................................. 128 FiguresinAnnexes Figure Ala1 Credit Rationinginthe Formal Market................................................................. 89 Figure Al.2. Credit Rationinginthe Informal Market............................................................... 90 Figure Al.3 Operated vs.Owned Area by Province................................................................. 97 BoxinAnnex Box A3.1 Structure ofthe 2001-02Pakistan S A M ............................................................... 114 References ............................................................................................................................. 130 Background Papers for the Pakistan Rural Factor Market Study ................................. 137 vii Pakistan at a glance 9117/04 POVERNand SOCIAL South Low- Pakistan Asia income levelopmentdiamond' 2003 Population,mid-year (millions) 148.4 1,425 2,310 Life expectancy GNI per capita (Atlas method, US$) 430 510 450 GNI (Atlas method, US$ billions) 64.0 726 1,038 Average annual growth, 1997-03 T Population(%) 2.4 1.8 1.9 Labor force (%) 3.2 2.3 2.3 ;NI ier Most recent estimate (latest year available, 1997-03) :apita Poverty(% ofpopulation belownational poverty line) 33 Urban population(% of totalpopulation) 34 28 30 Life expectancyat birth (years) 64 63 58 I Infant mortality(per 7,000live births) 76 68 82 Child malnutrition (% of children under 5) 48 44 Access to improvedwater source Accessto an improvedwater source (% ofpopulation) 90 84 75 illiteracy(% of population age 75+J 59 41 39 - Gross primaryenrollment (% of school-agepopulation) 73 95 92 Pakistan Male 84 103 99 Low-incomegroup Female 62 88 85 KEY ECONOMIC RATIOS and LONG-TERM TRENDS 1983 1993 2002 2003 Economic ratlos' GDP (US$ billions) 28.7 51.5 58.9 68.6 Gross domestic investmenffGDP 18.8 20.8 14.7 15.5 Exportsof goods and servicesiGDP 11.9 16.3 18.7 20.5 Trade Gross domestic savingsiGDP 7.7 14.7 14.4 15.6 Gross nationalsavings/GDP 17.0 20.6 18.1 22.2 T Currentaccount balance/GDP 0.6 5 . 5 2.7 6.1 Interest paymentsiGDP 1.1 1.5 1.3 1.1 Total debffGDP 41.9 47.7 57.1 52.7 Total debt serviceiexports 20.9 23.9 21.0 16.7 Presentvalue of debffGDP 44.9 Presentvalue of debffexports 195.2 indebtedness 1983-93 1993-03 2002 2003 200347 (average annual growth) GDP 5.8 3.4 2.8 5.8 6.1 -* Pakistan GDP per capita 3.1 0.9 0.4 3.3 3.7 Low-income(ITOUR Exportsof goods and services 10.4 3.3 10.3 32.9 1.5 STRUCTUREof the ECONOMY 1983 1993 2002 2003 Growth of investmentand GDP (%) (% of GDPJ I II Agriculture 30.3 25.0 23.2 23.3 " T Industry 221 247 233 235 lo Manufacturing 153 167 161 164 o ~ Services 477 503 535 532 .,o Privateconsumption 80.8 72.2 74.4 72.7 - n o 1 General governmentconsumption 11.4 13.1 11.3 11.7 imports of goods and services 23.0 22.4 19.0 20.4 -GDI +GDP 1983-93 199343 2002 2003 (average annual growth) Agriculture 4.4 3.5 -0.1 4.1 I lndiistrv 7.1 3.6 5.4 5.4 Manufacturing 6.6 4.1 5.0 Services 5.7 4.1 4.1 Privateconsumption 4.3 3.5 1.4 -0.6 Generalgovernmentconsumption 6.6 1.3 13.5 10.4 Gross domestic investment 4.9 0.7 -3.0 10.5 -Exports -imports Importsof goods and services 3.2 0.0 4.5 13.9 Note: 2003 data are preliminaryestimates. * The diamonds show four key indicators Inthe country (in bold) comparedwith its income-group average. If data are missing,the diamondwill be incomplete. viii PRICES and GOVERNMENT FINANCE 1983 1993 2002 2003 1 I Domestic prices Inflation (Oh) I (% change) Consumer prices .. 9.8 2.7 3.1 Implicit GDP deflator 5.3 8.7 3.1 4.6 Governmentfinance (% of GDP, includes current grants) Current revenue 18.1 19.5 20.8 Current budget balance .... -2.4 0.2 1.1 Overall surplusldeflcit .. -8.1 -6.7 4 . 5 TRADE 1983 1993 2002 2003 (US$ miilions) Export and Import levels (US$ mill.) Total exports (fob) 2,694 6,782 9,140 10,889 I 12,500 cotton 271 18 49 Rice .... 317 448 555 10 000 Manufactures 1,881 3,723 5,368 6,653 7 500 Total imports (ci0 .. 10,049 9,432 11,333 Food .. 1,290 413 587 5 000 Fuel and energy .. 1,578 2,664 3,097 2 500 Capital goods .. 3,409 2,594 3,392 1 I 0 Export price index (1995=100) 88 80 89 97 98 99 00 01 02 Import price index (1995-100) 96 90 101 Exports imports O3 Terms of trade (1995=100J 92 89 88 BALANCE of PAYMENTS 1983 1993 2002 2003 1 (US$ millions) ;;nt account balanceto GDP (%) Exports of goods and services 3,420 8,339 11,056 13,686 Imports of goods and services 6,593 12,856 11,646 14,047 Resource balance -3,173 -4,517 -590 -361 Net income -421 -1,498 -2,319 -2,210 Net current transfers 3,416 2,688 4,500 6,775 Current account balance -178 -3,326 1,591 4,204 Financing items (net) 2,682 1,685 1,475 Changes Innet reserves .... 644 -3,276 -5,679 1-10 1 Memo: Reserves including gold (US$ millionsJ 2,758 1,369 4,997 10,243 Conversion rate (DEC, local/US$) 12.7 25.9 61.6 58.6 EXTERNAL DEBT and RESOURCE FLOWS 1983 1993 2002 2003 (US$ millions) Composition of 2003 debt (US$ mill.) Total debt outstanding and disbursed 12,026 24,546 33,672 36,132 IBRD 351 2,624 2,749 2,695 G: 1.245 A 2,695 IDA 1,145 2,683 5,394 5,869 F 2,265 Total debt sewice 1,343 2,383 2,850 3,028 IBRD 63 343 367 373 IDA 14 45 111 130 Composition of net resource flows Official grants 277 361 1,495 1,038 Official creditors 350 1,011 781 -133 E 14342 Private creditors -87 710 -515 -530 Foreign direct investment 30 306 368 612 Portfolio equity 0 270 -491 -394 World Bank program Commitments 306 625 736 498 A - iBRD E. Bilateral Disbursements 144 598 961 145 B IDA D. Othermultilaterai F Private - Principal repayments 34 182 318 357 C IMF -- G* Short-term Net flows 110 416 643 -212 Interest payments 42 206 160 147 Net transfers 68 210 483 -358 DeveloDmentEconomics 9117104 i x ACKNOWLEDGEMENTS This study was managedby Sakwa Bunyasi andPaulDorosh, (co-task managers). PaulDorosh is the principal author o f the final report. Albert0Valdes designedthe study, provided overall guidance for the work and contributedto the analysis and synthesis. Sohail J. Malik playeda major role inthe work on agricultural credit and poverty, the synthesis of the results, and inorganizing seminars and workshops in Palustan. The report draws from a set o f backgroundpapers prepared for this study listed at the end o f this report. Gustavo Anriquez, Hanan Jacoby, Ghazala Mansuri, Hina Nazli, Muhammad Khan Niazi and Sarfraz Qureshimade major contributions, bothas authors ofbackground papers, and inthe synthesis o fthe results. Backgroundpapers were also prepared by Munir Ahmad, Zafar Altaf, Alejandra Cox-Edwards, Zakir Hussainand A.R.Kemal. Manuel Contijoch contributed to the discussion o f irrigation and water markets. Adolfo Brizzi, Derek Byerlee, Tekola Dejene, RashidFaruqee, Steven Haggblade, Abed Hasan, Isabel Lavadenz, John Mellor, Stephen Mink,Abdul Salaam and Dina Umali-Deininger provided useful comments and suggestions on various aspects o f the report. The report has also benefited from comments from the Pakistan Ministryo fFood, Agriculture and Livestock (MINFAL) and discussions with participants at seminars at the Agricultural Prices Commission, the Lahore University o f Management Sciences, the Pakistan Institute o f Development Economics and the InternationalWater Management Institute inLahore, and a seminar in September 2004 inIslamabad hostedby MINFAL. X EXECUTIVE SUMMARY Introduction 1. InPakistan, public andprivateinvestments inirrigation, adoptionofgreen revolutiontechnology and trade and output pricing reforms have spurred substantial agricultural growth over the last three decades. Agricultural GDP growth slowed substantially since the early 1990s, particularly inthe crop sector (2.4 percent average annual growth from 1990to 2002, compared with 3.5 percent per year growth from 1960to 2002), however, raising concerns about the sustainability o f agricultural growth inthe mediumterm. Moreover, the sharp declines inrural poverty that accompanied agricultural growth inthe 1970sand 1980shave not continued. The latest available poverty estimates (38.9 percent in2001-02), may be inflated by the effects o f the drought, but estimated rural poverty in 1998-99(35.9 percent) i s essentially the same as in 1990-91(36.9 percent). 2. Pakistan's Poverty Reduction Strategy Paper (PRSP) written in2003 (World Bank, 2003b) gives a centralrole to the rural sector inaccelerating growth and reducing rural poverty. However, although the PRSP places a major emphasis on employment-generating growth inagriculture, small and medium enterprises, housing and construction, information technology, telecommunications and exports, it lacks an explicit rural development strategy beyond that for the agricultural sector. Acceleratingrural economic growth and reducingrural poverty i s complex, and requires a comprehensive strategy built on sound analysis covering all major aspects o f the rural economy including agricultural output markets, inputmarkets (seeds, fertilizer, extension), factor markets (land, water, labor and credit), the ruralnon- farm economy, and targeted interventions. 3. Thisreport focuses on one aspect o fthis complex puzzle -rural factor markets. Itbuilds on earlier work on output and inputmarkets, and anticipates future analytical work on the rural non-farm economy and other key aspects o f the rural economy. The report i s designed to identify the main factor market inefficiencies inrural Pakistan, provide estimates o f their impact interms o f efficiency and equity, and suggest policy measures to improve the functioning o f these markets, increase rural incomes and help reduce rural poverty. 4. Inpresenting the results, each factor is discussedintum, drawingon descriptions ofinstitutional constraints and statistical analysis o f recent householdsurvey data. Key new findings include quantitative estimates o f the importance o f linkages across factor markets and the impact o f factor market inefficiencies on agricultural output and revenues. Inaddition, the report presents analysis showing that for the 61 percent o frural poor households that do not cultivate crops (agricultural laborers and non-farm households), expansion inoutput o f major crops may not be sufficient to significantly raise their incomes, even with substantial multiplier effects on the ruralnon-agricultural economy. PromotingEfficiencyandEquity:The RoleofFactor Markets 5. Trade and exchange rate policy reforms undertaken inthe late 1980s greatly reduced the divergence between domestic and internationalprices for most major agricultural commodities in Pakistan, and in general increased incentives for production and farmer incomes. Although some trade taxes and marketingrestrictions on agricultural outputs still remain, lack o f access to land, water and capital, and inefficiencies infactor markets for land, water, labor and capital are now among the greatest impediments to agricultural and rural income growth, and reductions inrural poverty. 6. Well-functioning factor markets can help ensure that the available supply o f these factors i s allocated to its most productive use, thus maximizing value added and incomes. Inefficiencies inthese xi factor markets, whether due to structural characteristics o f the factors or administrative procedures and policies, increase costs o f transactions or even prevent transactions from occurring, however, and thereby lead to a misallocationo f factor resources that reduces output and economic growth. This report does not attempt to directly measure the relative contributions o f structural characteristics and policy distortions, but insteadpresents statistical evidence ofthe existence of factor marketdistortions andtheir effects (e.g. productivity effects arising from lack o f access to credit for small farmers). To the extent that inefficiencies infactor markets are due to policy distortions, administrative and policy reforms can improve productivity. Where the inefficiencies are due to the very structure o f the factor market, (e.g. uneven distribution o f landholdings or the physical properties o f the Indus river irrigation system that limit storage and control ofwater flows), administrative andpolicy reforms alone will notnecessarily result inlarge efficiency or equity gains. 7. Moreover, major linkages between factor markets, most importantly betweenland and water markets, and between land and credit markets, mean that structural andpolicy-induced inefficiencies in one factor market have significant implications for hnctioning o f other factor markets. These inefficiencies in factor markets can have particularly adverse effects for the poor, who often lack both employment opportunities and the access to the land, water and capital that would enable them to earn higher returns on their labor in own-farm or own-enterprise activities. Land 8, Land i s at the heart o f Pakistan agriculture and the rural economy, and returns to land are estimated to be about halfo f incomes (value added) fi-om crop agriculture. Distribution o f land i s highly skewed, however, a major cause of income inequality inrural Pakistan.. According to the 2000 Agricultural Census, only 37 percent o frural households owned land, and 61 percent o f these land- owning households owned less than 5 acres, (15 percent o f total land). Two percent o f households owned 50 acres or more (accounting for 30 percent o f total land). The overall Gini coefficient o f land ownership in2000 inPakistanwas 0.66 includingrural landlesshouseholds, the Ginicoefficient was 0.86. By comparison, Gini coefficients for land ownership are 0.71 inIndia, 0.42 inBangladesh, and 0.85 in Brazil. 9. Landi s rarely bought and sold due to hightransactions costs, potential disputes about accuracy o f land records, landprices inexcess o fthe discounted value o f potential agricultural earnings from the land, and lack of access to credit by those without land. The low frequency o f land transactions helps perpetuate the highly unequal distribution o f land, hamper labor mobility, and reduce returns to family labor. Highly skewed land distribution and inefficiencies inland markets also limit access to credit for many households, since land i s the major acceptable form o f collateral in formal credit markets. Land rentals invarious forms are more common, however 33 percent o f area cultivated in2000 was under some form o f tenancy arrangement according to 2001-02 HIES data (18 percent under share-cropping). 10. Most empirical evidence indicates that productivity o f land on large farms inPakistani s less than that of small farms, holding other factors constant. For example, plot-level regressions o f productivity, correcting for plot characteristics and some household characteristics (such as tractor ownership and number o f household workers) suggests a doubling o f operated area leads to 10percent lower wheat yields (and 13 percent lower rice yields), (World Bank, 2002). Farmhouseholdincome data also show that small farmers have higher net returns per hectare than do large farms. These findings suggest that increases inthe share o f land cultivated by small-holders would tend to increase overall farm productivity inPakistan. 11. Econometric evidence also suggests that tenure arrangements have significant impacts on productivity. Productivity of share-croppers i s about 20 percent lower than productivity o f landowners, xii holding other factors constant, because o f lower incentives for own-labor inputs. Though share-cropping i s declining as a form o f tenancy, the benefits to share-croppers (shared financial risks o f crop failure and credit for purchase o f inputs) and owners (greater owner inputinto management decisions) likely explain i t s persistence as a tenancy arrangement. 12. Three major attempts at redistributivelandreforminPakistan have failed (most recently in 1977), and land reformhas neither political support, nor the backing o f Islamic religious authorities. Because o f these severe institutional constraints to traditional redistributiveland reform, more feasible options to increase access to land by poor households involve a win-win approach. Land purchase schemes that include grant components for the poorest landless households are one example, though fiscal costs could limit the scale o f suchprograms. Policymeasures to increase access to land also include increased access to credit to enable poor households to purchase land, land taxation to reduce incentives for holding land for speculative purposes, and measures to improve the efficiency o f land sales and rental markets. 13. Administrative reforms inthe enforcement o f existing norms and regulations, including enforcement o f contracts and the rights o f tenants, and liftingrestrictions for rentals would also help improve the functioning o f land markets. Simplification o f complicatedlegal procedures that result in long delays inproceedings and a backlog o f cases would reduce land transactions costs. Poor households are at a particular disadvantage incasesbrought before civil courts, since they often cannot afford to pursue these cases. Possible steps include limits on the number o f appeals, procedural reforms, alternative conflict resolution mechanisms, and barring o f the jurisdiction o f civil courts inlandrevenue matters. Improved dissemination o f information on landprices and transactions would also enhance efficiency o f land markets. 14. Improvementsinlandrecords' could also enhance the efficiency o f land markets and increase security o f tenure. Without secure rights to their land, farmers not only have less incentives for investment, but also devote resources to defending their rights. Lack o f secure title also reduces access to land by landless households through rentals by increasing the risk o f landowners who might otherwise rent out more land. Inaddition, banks are reluctant to lend money using land as collateral because they do not trust the current recording system. Questions regarding the validity and enforceability o f titles also make it harder to buy and sell land. Computerizationo f land records, as tested ina recent pilot project in Punjab, i s one option for establishing clearer landrecords and reducing fraudulent dealings by land administrationofficials. Water 15. Given low levels o f annual rainfall concentrated over only a few months o f the year and periodic droughts, water availability at the farm level i s a major determinant o f farm productivity inPakistan agriculture. Nearly 80 percent o f cropped area i s irrigated, and agriculture i s by far the largest user o f available water sources inthe country, consuming on average about 95% o f available water resources. In 2001-02, out o f 18.0 million hectares o f irrigated land, 38 percent was irrigated solely with canal water; 39 percent with canal and tube well water, and 19 percent solely with tube well water. Efficiency o f water use in surface irrigation i s low, however: conveyance losses (part o f which adds to groundwater recharge) are 55 percent o f total canal diversions. 16. Inprinciple, accessto canal water is determinedbyphysicallocationalongthe canal andthe warabandi systemo f allocation o f water through administratively set rotations. Access to canal water i s InPunjab, landdisputes are primarily due to fraudulent dealing because the records are not clear andpatwars (land administration officials) often accept informal payments. xiii thus explicitly tied to access to land. Ownership o faplot withina watercourse commandarea confers access to irrigation water, but does not guarantee canal water availability. Inparticular, water availability decreases significantly ifthe watercourse i s locatednear the tail o f the distributary or minor, and/or ifthe plot i s locatednear the tail end o fthe watercourse (World Bank, 2002). Water theft by farmers upstream i s a major reason for these water shortages faced by these tail end farmers. Recent administrative reforms insome parts of Sindhgiving greater control to farmer organizations havereducedwater theft, as well as improved maintenance o f water courses and collection o f fees.2 These could be expanded to other regions, as well. Other alternatives for improving delivery o f surface irrigation water services to farmers also exist, including private professional canal management. 17. Availability o f groundwater has an even higher impact on yields than does canal water, but only 8.1 percent o f cultivating households owned tube wells in2001/02. Informal groundwater markets significantly improve access to groundwater, particularly for small farmers, landless tenants and younger households who often lack the resources (or land and water rights) to install their own tube well. However, water purchasers do not have full access rights to the water, and are frequently denied access when water or energy supplies are scarce. Moreover, access to tube well water depends on the proximity to an existing tube well and conveyance mechanisms (e.g. channels) to distribute the water. 18. Informalwater markets for surface irrigation (mainly barter) and groundwater exist, but establishment o f water markets that would permit trading o f water rights across the canal system (such as those inChile and Mexico) would require major institutional changes and investments. Inparticular, any system o f tradable water rights would needto account for the minimal degree o f water control above the water course level inthe current Indus basin system, because o fthe absence o f sufficient storage capacity and control structures (gates) to regulate water flows. 19. Several institutional changes could bringabout major improvements inthe efficiency and equity of water use inPakistan agriculture, with or without the establishment o f water markets, however. First, inorder to increaseaccountabilityfor Operations andMaintenance, management ofdistributaries and/or minors could be transferred to Farmer Organizations (FO's), permitting assessment, collection and retention o f abiana (water charges) by the FO's. Second, to increase transparencyand enable more informed decisions on area irrigated and crop choice, a) periodic measurements o f the actual rate o f water flow at various points inthe water course could be made; and b) open access to the written rules o f the water rotation could be provided to farmers. Third, reforms inthe institutional arrangements for allocation o f water could be made to create a more demand-drivensystem. Options include: a) Water Course Associations, Water Users' Associations and Area Water Boards (AWB's); b) Professional canal management. 20. Development o f effective water markets would require two steps beyond those outlined above: i) establishing legal tradable water rights at the national and provincial levels; ii)strengthening local institutions (such as Water Course Associations and Water Users' Associations or professional canal management agencies) to enable them to serve as brokers inwater markets, helping to match buyers and sellers o f water. Once effective water markets were developed at the water course level, trades between water courses could be brokeredthrough these local institutions at the minor level. 21. The experience o f Chile shows that improved allocation o f water, (inthe Chilean case through the establishment o f water markets), combined with appropriate trade and macro- policies, can facilitate Inthe commandarea coveredbythe LeftBankArea Water BoardinSindh, where tail endfarmers hadnot received water inthree years, 48 illegal direct outlets serving an area o f about 25 thousand hectares were closed in early 2004. xiv diversification o f agriculture into high-value products for urban and export products, increased labor demand and reductions inrural poverty. Given the infrastructure constraints inPakistan, the potential gains from establishing water markets may be smaller than inother countries. Nonetheless, benefits o f more efficient allocation o f water through institutional reforms with or without water markets could still be substantial ifthey are combinedwith other measures to promote agricultural diversification, including strengthening security o f land titles, public investments inrural infrastructure (roads and electricity), private investments inprocessing and storage, and establishment o f market links to export markets. Labor Markets 22. Unlikethe markets for landandwater, rurallabormarkets are generally characterized bydemand constraints, rather than supply constraints, as evidenced by substantial underemployment. There i s considerable labor mobility, in spite o f hightransactions costs involved injob search and especially in migration, However, formal sector employment inrural areas i s minimal, and over 80 percent o f agricultural labor and most o f rural non-agriculturallabor i s self-employed. Female participation inlabor markets remains limited, though: only 16 percent for females age 15 and over (not including home-based activities), compared with 85 percent for males. Job skill levels for bothmale and female labor are generally low. 23, Education i s an important determinant o f female labor force participation, rural-urban migration, and labor productivity inrural non-fann sectors. Econometric analysis shows that education i s also a key determinant o f agriculturalproductivity. For the average producinghousehold, another year o f education translates into a 13% increase inyearly net revenues. Effective and more wide-spread vocational training programs could provide needed skills for non-agricultural labor. Women's access to educational and training institutions could be increasedthrough more reservation o f spaces for women. 24. Real wage rates o fbothregular and casual agricultural workers increased by an average o f 1.O percent per year from 1984 to 1994, indicating long term tightening o f the labor market. Likewise, real wage rates o f construction laborers increased by 1.1percent over the same period. Most o f the increases inreal wage rates took place inthe 1980s, however. The trends inthe 1990sindicatelittle gains inreal wages. From 1991-97, real wage rates o f casual agricultural laborers grew by an average o f 0.8 percent per year, while realwage rates o fregular agricultural workers fell by 2.4 percent per year. Likewise, real wage rates o f construction laborers grew by only 0.4 percent per year from 1991-97 and for the 1991- 2002 as a whole, fell by 0.4 percent per year. 25, The decline inrealwages inthe agricultural sector inthe 1990s coincides with a period o f slower but stillpositiveper capita agricultural growth. This suggests rather weak linkages betweenagncultural growth and rural labor demand. The smaller rate o f decline o f real wages for constructionlaborers suggests that lack o f labor mobility between sectors and betweenrural and urbanregions hinder integration o fnon-agricultural and agricultural labor markets 26. A better functioning landmarket could also promote labor mobility, helping farmers who want to leave farming to sell or rent out their farm land. This would require greater security o f property rights, and facilitation o f land transactions that are presently stifledby the absence o f a valid title, risks o f court challenges, and highregulatory and transactions costs (for example, stamp dutyplus registration fee plus capital tax amounted to 16% o f the average market price of land in 1995). 27. A concertedeffort to enforce existing legislationregarding the rights o fwomen and children is also needed. Inaddition, measures to assist households trapped inbonded labor arrangements, including xv legal protection, provision o f credit and debt cancellation programs should be urgently ~onsidered.~ Although these programs often face considerable opposition from powerful vested interests, such programs could have a major impact on some o f the poorest rural households. Credit 28. Inprinciple, creditmarketsoffer awide scope forpoliciestoincreaseproductivityandtoreduce ruralpoverty, and can help overcome inefficienciesinother factor markets. InPakistan, however, access to formal credit markets inrural areas i s generally limited to landowners, since land i s the main form o f acceptable collateral for loans. According to Pakistan Rural Household survey data, only 11percent o f farmers obtained formal sector loans in2001-02. Access to informal credit markets, particularly supplier's and consumer's credit i s more widespread, (75 percent o f farm households receivedloans), but approximately 40 percent o f rural (farm) households are credit constrained, not being able to obtain as much credit as they would like at existing interest rates. 29. Econometric analysis o f plot-level data suggests that lack o f access to credit significantly reduces farm productivity (measured as gross value o f output per unit o f land). Households who faced credit rationing inboththe formal and informal sectors (approximately 17 percent o f all households) had a 23 percent reduction invalue o f yields. Furtherstreamlining o f lending procedures, development o f alternative forms o f collateral, and other lending innovationsto increase access o f small farmers to commercial credit could help relieve these credit constraints for small farmers. 30. Removal o f subsidies on credit (implicit inthe continuing low repayment / highdefault rates for loans inthe formal sector) could eliminate these policy-induced incentives toward capital-intensive technologies that reduce labor demand. Efforts at promoting agricultural diversification should thus avoid subsidies on credit that would encourage capital-intensive technologies inproduction, processing and marketing. Instead, agricultural diversification, the development o f the rural non-farm sector and poverty reduction may be better achieved throughpublic investments in electricity, transport and other infrastructure. Furtherwork on particular constraints to rural non-farm investment, employment and growth i s needed, however. 31. Production credit and micro-credit programs targeted to landless and small farmers that required other forms o f collateral (besides land) could also help promote rural employment and income generation. Rural credit from formal sources i s currently available for a narrow range o f agriculturalproduction activities and does not serve the needs o f the non-farm sector. Inparticular, availability o fmedium-term credit i s extremely limited. Increased access to credit by small scale enterprises inrural areas, for example, would promote farm to non-farm linkages, develop the rural non-agricultural sector and increase labor demand. Targeted matching grants to poor households for income-generating activities are one option. 32. The Government o fPakistan inrecent years has, in fact, placed considerable emphasis on developingmicro-finance as part o f its poverty alleviation strategy. In2003, Rs. 1,442 million (2.5 % o f total institutional credit) was disbursed as small loans through the Pakistan Poverty Alleviation Fund (PPAF), Khushali Bank and the Zarai Taraqiati Bank Limited (ZTBL, formerly called ADBP). N o Nearly 200 thousand rural households inSindh and Punjab (equivalent to about 2 percent o f Pakistan's rural population), were estimated to be indebt bondage based on 1990 Agricultural Census data. Analysis o f 2000 Agricultural Census data suggests that inPunjab alone, 50 thousand sharecroppers were indebt bondage, equivalent to about 9 percent o f total landless tenant farmers inPunjab. xvi comprehensive study o f the overall impact o fthese micro-credit schemes has yet been done, and there remains a major concern regarding financial sustainability o f these programs. Promotion o f savings mobilization schemes and implementationo f an adequate regulatory fkamework for deposit-based institutions could help provide sustainable resources for micro-credit. IncomeDistributionand PovertyReduction 33. Increasedagriculturalproduction has been a major force for reducingrural poverty over much o f South Asia over the past four decades. Rural poverty inPakistan didnot decline inthe 199Os, however, inspite ofagricultural growth. Overestimation oftrue sectoral growthratesbecause ofabrupt changes in estimates o f livestock population i s part o f the explanation for this lack o f correlationbetween agricultural growth and poverty reduction. Stagnation inthe real consumer prices o frice and wheat (incontrast to declines inearly decades) i s another major factor. 34. Two other major factors are the changing structure o f the Pakistan economy, inparticular the declining share o f agriculture intotal GDP (39 percent in 1970, but only 24 percent in2000), and the persistent unequal distribution o f land (and land revenue). Moreover, since more than 60 percent o f rural poor households are not farm households (according to HIES 2001-02 data), the magnitude o f these linkage effects with the non-agricultural sector i s crucial to the poverty impacts o f agricultural growth. Ruralnon-farm households account for 46 percent o f the rural poor; agricultural laborer households comprise 15 percent o f the rural poor. Though there i s substantial poverty among small landowning farmers (38 percent are poor), this group accounts for only 24 percent o f total rural poor households. Landless tenant farmers (61 percent o f whom are poor) account for another 13 percent o f the rural poor. 35. Model simulationresults o f the direct and multiplier effects o f a 10 percent increase inthe output o f all major crops (wheat, basmati and IRRIrice, cotton and sugar cane, which together account for about 35 percent o ftotal agricultural value-added) suggest that the largest gains o f increased production o f these major crops accrue to large and medium land owners, whose incomes rise by 7 percent. Incomes o f small farm owners and pure tenants also rise by about 5 percent. Butthe poorest rural household groups (agricultural laborers and rural non-farm poor -- 29percent o f the rural population) reap only 7 percent o f the total income gains, and their incomes rise by only 3-4 percent. Gains from a 10percent increase in livestockproduction (mainly cattle, goats and dairy products) are more evenly spread out given the distribution o f ownership o f livestock, suggesting that measures to increase livestock productivity may have more positive effects on equity. 36. Similarly, inspite o fmultiplier effects on agricultural and non-agricultural output, simulations suggest that a shift from share-cropping to fixed rents would remove labor disincentive effects and raise the productivity o f former share-croppers to equal that o fland owners (affecting 18percent o farea cultivated inPakistan) would raise average incomes o f tenant farmers by 3 to 4 percent (and incomes o f farmers directly impacted by the reforms by about 10percent), but are likely to have only small overall impacts on rural incomes inaggregate. Releasing constraints on access to credit has similar impacts on small farmers, as well as inducing an estimated 2 percent gain inaverage incomes o f the rural non-farm poor, Raising productivity o f farmers at the tail-end o f watercourses to matchthat o f head-end farmers through increased access to water, ifit could be achieved without reducingthe productivity o f head-end farmers, would have even larger impacts on small farmers (4-6 percent average income gains) and could raise incomes o f the rural non-farm poor by 3 percent. 37. Thus, although agriculturalgrowth can have a major impact on a sizeable segment o fthe rural poor through combined direct and multiplier effects, it i s unlikely to be sufficient to rapidly raise incomes o f the 30 percent o f the rural population (and 61 percent o f the rural poor) who are landless agricultural laborers andnon-agricultural rural poor households. Inthe absence o f a change inthe structure o frural xvii incomes and employment or significant gains inthe rural non-farm economy apart from agricultural growth-induced linkage effects, targeted interventions to agricultural laborers and the rural non-farm poor will beneeded. 38. The Poverty Reduction Strategy Paper (PRSP) o f Pakistan issued inDecember 2003 (World Bank 2003b) recognizes this explicitly. Inaddition to advocating sustained and broadbased economic growth emanating from the rural areas the strategy also emphasizes the simultaneous provision o f social and economic services and infrastructure for the poor, creatingjob opportunities and improving governance as essential elements. A more detailed rural development strategy i s still needed, however. 39. Significant structural changes have taken place inPakistan's economy over the past several decades that make factor market reforms critical to increasing agriculturalproductivity and raising incomes o f the ruralpoor. Ruralpopulationhas steadily increased, gradually reducing average farm sizes inthe face o fconstraints on the overall supply o farable landandwater. The size o fthe agricultural sector relative to the overall economy has steadily shrunk, as the output o f the rural non-farm economy has expanded, tending to lessen the impact o f agriculture on the overall rural economy. 40. Even inthe face o f increasingly severe constraints on land and water resources, agricultural growth can play a major role inincreasing incomes and reducing poverty among farm households. Experience o f other countries suggests that factor market reforms, coupledwith outward-oriented trade and exchange rate policy can lead to rapid agricultural growth andrural poverty reductionthrough diversification o f agricultural productioninto high-value products (e.g. fruits and vegetables) with labor- intensive production and processing. Reforms infactor markets that promote the allocation o f scarce land and water resources to their most efficient uses, and increase returns to unskilled labor, are essential for replicating this success inPakistan. 41. However, because many rural poor households lack access to land, agricultural growth alone i s not sufficient to significantly raise incomes of a large fraction o f the rural non-farm poor, even with its substantial multiplier effects on the non-agricultural economy. Policies and programs to increase access to credit and enhance worker and entrepreneurial skills can directly support growth inthe ruralnon- agricultural economy and further raise incomes o f the rural non-farm poor. Targeted interventionsare also neededto reach the poorest households including expansion o f micro-credit, enforcement o f laws on bonded labor, and training inmarketable skills. I.RURALFACTORMARKETSINPAKISTAN: POLICYREFORMS FORGROWTHAND EQUITY I.INTRODUCTION 1.1 Palustan's agricultural sector has enjoyed steady growth for most o f the last three decades, with agricultural Gross Domestic Product (GDP) increasing by an average of4.1 percent per year from 1975 to 2000. Green revolution technology o f improved seeds, irrigation, and increased fertilizer use spurred rapid growth incrop agriculture (especially wheat and rice) beginning inthe later 1960s. Livestock production outpacedthe rapid growth incrop agriculture, with value added increasing by 5.3 percent per year from 1975 to 2000. 1.2 Long-term agricultural growth has also been accompaniedby reductions inrural poverty, particularly inthe 1980s. Rural poverty fell from 49.3 percent in 1984-85 to 33.4 percent in 1993-94, contributing to an overall decline o f poverty inPakistan from 46.0 percent in 1984-85 to 28.6 percent. Table 1.1. Poverty Estimatesfor Pakistan 1984-85 1987-88 1990-91 1993-94 1998-99 2001-02 Urban 38.2 30.7 28.0 17.2 24.2 22.7 Rural 49.3 40.2 36.9 33.4 35.9 38.9 Overall 46.0 37.4 34.0 28.6 32.6 32.1 Note: 1998-99 data fromPIHS; all other years HIES. PIHS and HIES combined since 1998- 99 Source: World Bank (2002), p. 20. For 2001-02, Government o f PakistanPakistan Economic Survey (2002-03) 1.3 More recent trends of agricultural incomesand of rural poverty are far less encouraging. However, agricultural GDP fell by 1percent between2000 and 2002, inlarge part because o f drought, highlightingthe vulnerability o fPakistanagriculture and the importance o f availability o f water as a major constraint for many farmers. Likewise, the long-termdownward trends inrural poverty have not continued: estimates o fruralpoverty for 2001-02 o f 38.9 percent suggest that for the 1990s as a whole, rural poverty has not declined inspite o f agncultural GDP growth. 1.4 The deceleration o f agricultural growth and the rate o fpoverty reductionhas occurred in spite o f generally favorable policies affecting markets and prices o f agricultural outputs and non-factor inputs (e.g. seeds, fertilizer and pesticides). Major macro-trade policy and output market reforms in 1980s reduced the bias against apcultural productioninPalstanreflectedinprices o f agncultural outputs and inputsrelative to those inthe non-agricultural sectors. Though some trade taxes and marketing restrictions remain, the incentive structure for production o f major crops (wheat, cotton, basmati and IRRIrice, and sugar cane) i s more favorable today than inthe early 1980s when substantial growth occurred. 1.5 Long-term agricultural growth and rural poverty reduction are still constrained by the structure and performance o f key rural factor markets (land, water, labor and credit), however. These four factor inputsinto production, and the markets that facilitate their efficient allocation across alternative uses, are key determinants o f agriculturalproduction and growth. Moreover, retums to these factors o fproduction, and their structure o f ownership across households, are major determinants o f rural household incomes and poverty. 2 1.6 In a perfectly competitivemarket, economic theory suggests that returns to each factor o f production would be the same for all activities. Profit maximizing households and firms would utilize factors upto the point where marginalreturns to each factor equaled its cost (the market price o f the factor). Any excess demand for a factor could be purchased at the market price; similarly any excess supply o f a factor could be sold at the market price. The economic reality inrural Palustan i s far different from the economic theory o f perfectly functioning factor markets, however, as inefficiencies inmarkets for land, water, labor and capital result inlower production and incomes. 1.7 These inefficiencies, whether due to structural characteristics o f the factors (e.g. the pattern of land ownership or physicalproperties o f the Indusriver irrigation system that limit storage and control o f water flows) or administrative procedures and policies (e.g. lack o f clear land titles) can increase costs o f transactions or even preventtransactions from occurring, however, and thereby lead to reduced output and economic growth. However, to the extent that inefficiencies infactor markets are due to the very structure o f the factor market, (e.g. uneven distribution o f land holdings or the physicalproperties o f the Indusriver irrigation system that limit storage and control o f water flows), administrative andpolicy reforms alone will not necessarily result inlarge efficiency or equity gains.4 1.8 Land is the most immobile and most unequally distributed of the factors, and land sales markets inPakistan are very thin. The absence o f well-functioning land markets hinders access to land for the landless or small farmers, and perpetuates a highly skewed distribution of land ownership. Landrentals invarious forms (share-cropping, cash rentals) are more common, however, 20 percent o f area cultivatedin2000 was under some form of tenancy arrangement. Most empirical evidence suggests that productivity of land on large farms inPakistan i s less than that o f small farms, holding other factors constant. Moreover, both economic theory and empirical evidence suggest that productivity o f land under tenancy arrangements will be less than productivity o f owned land. Thus, the absence o f a well-functioning land market results inlower overall output, while the skewed ness o f land ownership contributes to widespread rural poverty. 1.9 Access to water inPakistan is closely tied to access to land, and water markets are nearly as thin as those of land. Throughout the Indusbasin, surface water is allocated through the warabandi system inwhich water flows are regulated administratively on a rotating basis. Trading o f canal water i s common (although sales o f canal water are illegal), but i s limited to trades with farmers inthe same distributary. Groundwater irrigation (mainly bytubewells) supplements surface irrigation for many farmers, increasing the reliability o f water supply and crop yields. Purchases o f groundwater increase access to water and yields o f small holders and tenants, but likewise are limited bydistance the water musttravel (often through unlinedwatercourses). 1.10 Rurallabor markets for male labor generally function well, though there i s substantial rural underemployment. Lack o f access to land limits self-employment opportunities for small farmers and landless households, while subsidized formal sector credit encourages labor-displacing mechanization. Informal labor markets account for essentially all employment inrural areas, but agricultural labor demand remains highlyseasonal. Ruralnon-farm employment i s mainly self- employment, and returns to labor are constrained inpart by lack o f skills and physical capital. Female labor force participation i s low: inrural areas only 32 percent, even when 14householdlabor activities are included. Thisreport does not attempt to directly measure the relative contributions o f structural characteristics, but instead presents statistical evidence o f the existence and effects o f factor market distortions (e.g. productivity effects arising from lack o f access to credit for small farmers). 3 1.11 Credit markets are potentially the most flexible o f all factor markets since overall supply o f credit i s determinedby government monetary and bankingpolicy. There are substantial inefficiencies informal creditmarkets, however, where credit isrationed andthe landless andsmall farmers lack the collateral to access credit. These groups get almost all o f their credit through informal markets, or through tied arrangements with landlords, employers or traders. Nevertheless, access to informal sector credit for poor households can be problematic, as evidenced by instances o fbonded labor arrangements. 1.12 Because o f the strong linkages across rural factor markets inPalustan, inefficiencies inone market can adversely affect the performance of other markets, magnifying the efficiency losses and consequences for income distribution. For example, ownership o f land i s a major determinant o f access to formal credit; water rights also are tied to ownership or user rightsto land. The skewed distribution of land thus results ina skewed access to water and credit, and the earnings derived from these factors. Better functioning water and credit markets, however, could help ease the adverse efficiency and distributional consequences o f skewed land ownership. Plan of the report 1.13 Section 2 o fthis report presents an overview o f Palustan's agriculture and agricultural policies. Sections 3 through 6 then analyze the four factor markets (land, water, labor and credit) in turn. Section 7 discusses the implications ofagricultural growthandinefficiencies infactor markets for rural poverty inPalustan, drawing on analysis o f farm-level data and model simulations. Section 8 concludes with a summary o f the major findings, lessons from other countries, and policy implications. 11. AGRICULTURAL POLICYAND SECTOR PERF'ORMANCEIN PAKISTAN: AN OVERVIEW 2.1, The agricultural sector i s at the heart o f Pakistan's rural economy, accounting for 70 percent o f rural household incomes and about 25 percent o f national GDP. Both crop and livestock sub-sectors have enjoyed substantial growth over four decades, but Pakistan's agriculture remains highly susceptible to droughts and i s heavily reliant on irrigation. 2.2. The agriculturesector hasmaintained a long-term annual growth rate of 3.7 percent over the periodof 1959-60 to 2001-2002 with wide year-to-year variations. Apart from a periodof slow growth inthe first half o fthe 1970s, average agricultural growth exceeded 3.2 percent per year ineach quinquennium from 1960 to 2000. However, the performance o f the agriculture sector (particularly the crops sub-sector), has suffered inrecent years because o f severe droughts inthe country. Agricultural GDP fell by 2.64 percent in2000-01 and grew by only 1.39 percent in2001-02 (table 2.1). Table 2.1 Agricultural Growth inPakistan, 1959-60 to 2001-02 Agriculture Crops Livestock Wheat Rice Cotton SCane Maize 1960-65 3.78 4.76 1.90 3.27 4.88 5.32 11.85 1.30 1965-70 6.26 8.18 1.97 9.70 12.20 7.24 7.15 4.82 1970-75 0.78 0.49 2.01 1.02 -0.74 2.64 -4.23 2.26 1975-80 3.91 4.15 3.07 7.19 6.80 3.61 5.29 3.21 1980-85 3.28 2.63 4.77 1.51 0.61 6.73 3.17 3.27 1985-90 4.57 3.70 6.12 4.11 -0.58 7.62 2.01 2.79 1990-95 3.93 3.02 5.70 3.50 1.37 0.32 5.85 2.25 1995-00 4.81 3.19 6.57 4.39 8.39 5.26 -0.36 4.62 2000-01 -2.64 -7.11 4.86 -3.33 -5.49 -1.88 -4.85 -1.87 2001-02 1.39 -0.09 3.42 -2.42 -11.10 6.46 4.10 -0.64 1960-2002 3.67 3.50 3.98 3.77 3.28 4.41 3.65 2.99 1990-2003 3.46 2.40 5.79 2.44 3.31 0.83 2.66 3.59 Source: M.Ahmad, (2003). 2.3. Agricultural incomesinPakistan are nearly evenly split between crop and livestock agriculture. The crops sub-sector currently contributes about 54 percent o f the value added, with major crops (wheat, cotton, rice, sugar cane) accounting for 38 percent o f total agricultural GDP and minor crops contributing 16 percent. Wheat, cotton andrice together account for 60 percent area cultivated nationally (table 2.2). The livestock sub-sector, dominatedby dairy, sheep and poultry, has increased its share over time, from only 25 percent in 1982 to its present level o f 46 percent. Fishand forestry are minor sectors inPakistan, accounting for 2.2 and 0.3 percent o f agricultural GDP, respectively. 5 Table 2.2 Area, Yield and Production of Major Crops inPakistan, 1999-00 to 2002-03 Area Area Area Yield Yield Production Production Notes: Oilseeds includes rapeseed, mustard and sesamum. Other includes fruits, vegetables and all other crops. Growth rates are average growth rates, 1990-91to 2002-03, 2002103 data are preliminary. Source: Economic Survey of Pakistan, 2003, (tables 2.1 and 2.4). 2.4. Over 80 percent of arable land inPakistanis irrigated, and availability of water is a major constraint on agricultural production at the farm level. 82 percent o f area cultivated o f Pakistan's major crops i s grown on irrigated land (both surface and groundwater), with cotton, rice and sugar cane grown exclusively on irrigated land. 2.5. Dominant cropping patternsvary across Pakistan according to soil type andwater availability. Inthe northernirrigated areas o fPunjab, basmatirice is cultivated inthe monsoon season (kharif), followed by wheat inthe winter season (rabi). Insouthern Punjab and northern Sindh, cotton i s the major kharif crop, with the wheat crop immediately following5 Insouthern Sindh where drainage problems inhibitcultivation o f cotton and warm night temperatures reduce wheat yields, non-aromatic rice (typically called IRRIrice) i s the major crop. Indryland (barani) areas o f northern Punjab and NWFP, wheat is cultivatedalong with pulses and other minor crops. 2.6. Adoption of green revolutiontechnology, (improved seeds, increased fertilizer use, and irrigation) enabled the crops sub-sector to grow at an average rate o f 3.5 percent per year over the last 40 years. Within the crops sector, cotton production grew mostrapidly (4.4 percent per year). Wheat, rice, sugar cane and maize also grew at rates inexcess o f 3 percent per year (3.8, 3.3, 3.6 and 3.0 percent per year, respectively). 2.7. Annual growth rates o fwheat (2.36%), cotton (0.53%) and sugar cane (2.46%) have been much lower since 1990. Total area cultivated has increasedby only 0.18 percent per year as expansion Here, farmers face a tradeoff between leaving cotton crop inthe field for an additional harvest and delays inwheat planting that result inlower wheat yields. Although area planted with cotton i s only about 113 o f wheat area, the two crops each account for about 30 percent agricultural value added at 1980-81base prices. 6 inmultiple-croppinghas slowed. Severalyears ofdrought have also limited areaexpansion, as well as diminished yields. Much o f the gains from the initial introduction o f new technologies for major crops hadbeen realizedby the early 1990s. Average wheat yields inPakistan (2.12 tonsha) were 16percent below average wheat yields in India (2.52 tonsha) over the 1989-90 to 2002-03 period. Average rice yields were essentially the same as those inIndia (2.74 and 2.85 tons/ha, respectively).6 Sugar cane yields inPakistan are only two-thirds those of India (45.5 and 67.7 tonsha, respectively). 2.8. Recent analysis suggests that total factor productivity in crop agriculture in Punjab grew by only 1.26 percentper year from 1966 to 1994, and that deteriorating soil and water quality are reducing productivity growth by an average of 0.22 percentper year (Ali and Byerlee, 2002). Although land and labor productivity bothincreasedby about 2.5 percent per year from 1966 to 1994, rising costs o f inputslimited growth intotal factor productivity incrop agriculture. Labor use incrop agriculture inPunjab declined from 98.7 daydhectare to only 71.7 dayshectare due to increased mechanization (tractors, harvesters and thre~hers).~Total factor productivity growth inthe rice-wheat cropping system has actually declined, inpart due to overuse o fpoor-quality tube well water, leading to soil salinity. 2.9. Livestock production dominates the rural economies in pastoralareas of Balochistan and NorthWest Frontier Province (N.W.F.P), and is a major source of farm earnings inother parts of Pakistan, as well. Production o f sheep and goats accounts for 17 percent o flivestock value added. Dairy production (milkand ghee) i s a major source o f incomes for low income households inthe Punjab and Sindh, with production of milk increasing by 80 percent between 1990-91 and 2002-03. Poultry productionhas increased even more rapidly, with the number o f birds and egg production increasing by 135 and 145 percent, respectively, from 1990-91 to 2002-03, but the sub-sector still accounted for only 3 percent o f livestock value added in2002-03. 2.10. Livestock and agricultural GDP growth inthe 199Os, though substantial, are overstatedbecause o f a change inthe base for cattle production, According to official GDP statistics, the value added o f livestock grew by 26% in 1995-96. This largejump i s the result o f usingfigures from the 1995-96 Livestock Census on the stock o f cattle to calculate value added inthat year without making adjustments to livestock figures inprevious years. Using the 10-year average livestock growth rate o f 5% for 1995-96 (instead o f 26%) reduces agricultural growth rate inthat year from 12%to 4.7%. With these adjustments, agricultural growth duringthe 1990s i s reduced from 4.5% to only 3.1%. Agriculturalmarketsandprices 2.11. The Government o f Pakistan intervenes inboth output and inputmarkets inorder to stabilize prices and improve incentives for agricultural production. Support prices are set for all major crops (wheat, rice, sugar cane and cotton), though government procurementtakes place only for wheat and cotton, The Palustan Agricultural Storage and Supplies Corporation (PASSCO) and provincial Food Departments procure 25-40 percent o f total productiono f wheat, and private market wholesale prices in Lahore have closely tracked the support price for wheat inmost years (table 2.3 and figure 2.1). There has been no procurement o fbasmati or InternationalRice ResearchInstitute (IRRI) rice since 1995-96, , even in2000-01when the support price (inrice equivalent) was near the wholesale price inLahore. The Trading Corporation o f Pakistan (TCP) procures cotton, and minimumcotton export prices are announced Invalue terms, riceproductivity inPakistanexceedsthat ofIndiasince a higher share ofPakistan's production derives from hgh-value basmatirice. Ali andByerlee, (2002), Table 2, p. 847. 7 daily which effectively act as ceilingprices. Market prices fell substantially below support prices in 1999 and 2001, however. The TCP also exports rice, sugar and wheat. Table 2.3 Ratio of Wholesale Prices (Lahore) to Support Prices, 1995-96 to 2001-02 Basmati IRRl Sugar Wheat Rice Rice Cane 1995-96 1.16 1.46 1.74 1.19 1996-97 0.99 1.59 1.99 1.48 1997-98 1.25 1.28 1.54 0.83 1998-99 1.14 1.67 1.55 0.82 1999-00 0.98 1.47 1.69 0.90 2000-01 1.07 1.01 1.13 1.20 2001-02 1.03 1.47 1.28 0.85 Average 1.09 1.42 1.56 1.04 Notes: Missing data for wheat prices from January-April 1999 i s interpolated. Support price of basmati and IRRIpaddy i s converted to rice equivalent using a milling ratio o f 0.67. Source: Agriculture Statistics of Pakistan(2001-02) Figure 2.1 Wheat Prices in Pakistan, 1995-2003 18.0 I 1 16.0 14.0 12.0 10.0 8.0 6.0 4.0 1-Wholesale Price (Lahore) +i+Procurement Price +Import Parity Lahore 1 Source: Calculated from Federal Bureau of Statistics data and M.Ahmad (2003). 2.12. Up untilthe mid-l980s, negativeindirecteffects of overall trade and exchangerate policy exacerbated the effects of agricultural domestic price interventions, and most major agricultural commoditieswere implicitly taxed. Inthe 1983-87 period, direct effects o f domestic price and trade policies reduced wheat and basmati rice prices by an average o f 25 and 72 percent, respectively, below border prices, The indirect effects o f trade and exchange rate policies that ledto an appreciation o f the 8 real exchange rate further reducedtheses prices to an average o f 42 and 78 percent below border prices. For cotton, this real exchange rate appreciation reduced the protectionaffordedto domestic cotton production from 177percent down to 27 percent. These price disincentives reduced wheat and rice production by an estimated 24 and 25 percent (relative to what they would have been duringthis period), and reduced farm incomes by an estimated 29 percent (Dorosh and Valdes, 1990).8 2.13. As aresultof policy reforms,divergencesbetweendomesticandinternationalpricesof majoragriculturalproductsare muchsmaller nowthan inthe mid-1980s(figure2.2). Wheat support prices were on average only 10 percent below import parity levels (measured at procurement centers) in 1999-00 and thoughprices since thenhave been below import parity levels, wheat import demand by Pakistanat world prices was essentially zero, (so that wheat was essentially a non-traded good).g Support prices for basmati rice were on average 22 percent below border prices (measured at procurement centers); IRRIrice prices were on average 15 percent above border prices. Cotton prices were within 5 percent o f estimated export parity in2000-01 and 2001-02. Moreover, trade and exchange rate reforms since the late 1980s have to a large extent eliminatedmajor trade and foreign exchange restrictions and greatly reduced the appreciation o f the real exchange rate relative to its mediumterm equilibrium levels. Figure2.2 NominalRates of Protectionof Major Crops 0.40 0.30 0.20 0.10 0.00 B3.I 1999-00 2000-01 -0.10 02001-02 -0.20 -0.30 -0.40 -0.50 ' Source: Based on calculations inM.Ahmad (2003). 2.14. Some informalandformalrestrictionsoninternalmarketsthat raisetransactions costsand reducemarketefficienciesremain,however. Econometric evidence suggests that wholesale wheat prices invillage markets are determined by the procurement price and the distance o f the village to the procurement center (Kurosaki, 1996), but movement o f grain across district boundaries has been banned for a few months after the wheat harvest in some recent years. Rent seeking activities occur, inmarkets withandwithout government intervention, as well. Inmango markets inthe Sindh, rents are involved in Hamid, Nabi and Nasim (1990) give similar figures for nominal rates o f protection, as well as a detailed account o f the policy interventions from 1960 to the mid-1980s. In2001, the GOPprovidedasubsidyonwheat exports that totaled about $1million. In2002-03, the wheat export subsidy on wheat was Rs 3250/ton, equivalent to 30 percent o fthe estimated C&F price o fU S wheat at Karachi. 9 the collection o f market fees, issuance o f trader licenses, and allotment o f shops (Smith, Khushk,and Stockbridge, 1999). 2.15. There have been no direct subsidies on pesticides, seeds and fertilizer since 1996, but subsidies on surface irrigation water and electricity (inBaluchistan only) remain. The pesticide and seed subsidies were eliminated in 1981-82 and 1982-83, respectively. The subsidy on locally produced fertilizer was eliminated in 1993; a small subsidy on imported fertilizer was eliminated in 1996." Likewise, a subsidy on tube wells o f Rs 16,000 to 20,000 per unit was abolished in 1994-95. Subsidies on canal irrigation (calculated as the difference between revenues collected from farmers as water charges and the cost incurred by the government for operation and maintenance, excluding capital costs) averaged Rs 4.6 billion from 1996-97 to 1999-00, equivalent to 0.8 percent o f agricultural GDP (table 2.4). Table 2.4 Subsidies inPakistanAgriculture: 1996-97 to 1999-00 (mn Rs) ~ lrrigation Credit Electricity Total 1996-97 4,550 1,000 604 6,154 1997-98 5,111 83 972 6,166 1998-99 4,237 437 1,336 6,010 1999-00 4,608 279 652 5,539 Average 4,627 450 891 5,967 Source: M. Ahmad (2003); Statistical Supplement (2002b); Pakistan notifications to the WTO. lo fertilizersnowfacea15percentsalesduty,andimportedfertilizersalsofacea5percentcustomduty. All 10 111. RURALLANDMARKETS:INSTITUTIONSAND CONSTRAINTS 3.1 Land, particularly irrigatedland, i s the most important productive asset inrural Pakistan and a major determinant of social status, political influence and income. Land ownership provides access to political power and public incomes for large landowners, and has considerable pecuniary and non pecuniary, often invisible, associated returns. An effective land market has failed to develop because o f legal impediments to the sale and purchase o f land such as the right o f shufa (first purchase) and excessive bureaucratic intervention. While the market for landrental i s somewhat more developed, inadequate information flow and other rigidities (including lack o f access to credit) are responsible for an economically and socially sub-optimal allocation o f this basic resource. Land Ownership and Utilization 3.2 Cultivated area inPakistan has increasedby about 50 percent since Independence in 1948 due mainly to increases inwater availability at the farm level. Most o f this gain in area cultivated was achieved prior to 1990; between 1990/91 and 2000/0 1, cultivated area increasedby only 3.2 percent, from 20.96 to 22.13 millionhectares, as cropping intensity increased from 1.35 to 1.43 (table 3.1). Table 3.1 LandUse in Pakistan, 1990-91 and 2000-01 7990-97 2000-07 Forest Area 3.46 3.77 (6.0) (6.3) Not Available for Cultivation 24.34 24.31 (42.2) (40.0) Culturable Waste 8.85 8.95 (15.4) (15.4) Cultivated Area 20.96 22.27 (36.4) (37.2) a. Net Area Sown 16.11 15.4 (28.0) (25.9) b. Current Fallow 4.85 6.73 (8.4) (11.3) Total Cropped Area 21.82 22.12 (37.9) (37.1) Total Area 57.61 59.44 (100.0) (100.0) a. Net Area Sown 16.11 15.4 b. Area Sown More than once 5.71 6.64 c. Cropping Intensity 1.35 1.43 Note: Figures inparentheses are the percentages o f total reported area. Source: Agricultural Census (1990 and 2000). 3.3 Distribution o f owned land inPakistan i s highly skewed and the number o f small farms i s increasing over time. According to Agricultural Census data for 1990, 54.4 percent o f farm households owned less than 5 acres (accounting for only 11.4 percent o f total area), while 2.8 percent o f households owned 50 acres or more (accounting for 34.0 percent o f total land). By 2000,61.2 percent o f farm 11 households owned less than 5 acres o f land, (accounting for 14.8 percent o f total farm area), and 2.0 percent farmers owned 50 or more acres, but these farms accounted for 29.7 percent o ftotal area. 3.4 The overall Gini coefficient o f land-holdings in2000 inPakistan was 0.66. Including the 63 percent o f rural households with no land, the all Pakistan Gini coefficient i s 0.86. By comparison, Gini coefficients for land ownership are 0.71 (NSS, 1991-92) for India as whole" (0.76 for Punjab, 0.63 in Uttar Pradesh), 0.42 inBangladesh, 0.37 inThailand, 0.85 inBrazil.12 Inspite o f the increase inthe number o f small farms, the overall Gini coefficient for land ownership inPakistanremained approximately constant from 1972 to 2000 (table 3.2). Table 3.2 Pakistan: DistributionofLandOwnership 1972 1980 I990 2000 Gini Coefficient 0.66 0.65 0.66 0.66 % of landless households 62.0 63.3 % Share of Holdings <5 acres a. Households 47.3 n.a. 54.4 61.2 b. Land 5.4 n.a. 11.4 14.8 % Share of Holdings 50+ acres a. Households 3.3 n.a. 2.8 2.0 b. Land 22.4 n.a. 34.0 29.7 Source: Agriculture Census (various issues) 3.5 Similarly, inequality o f land holding withinprovinces didnot change significantly between 1972 and 2000. The Gini coefficient for Punjabremained almost constant at about 0.62. Inequality in Sindh declined fiom 1972 (Gini coefficient o f 0.69) to 1980 (Gini coefficient o f 0.62), but has remained the same since then. InBalochistan, the Gini coefficient declined slightly from 0.70 in 1990 to 0.66 in2000, indicating an increasedequality o f land distribution (table 3.3). Table 3.3 GiniCoefficient for Ownership Holdingsby Province 1972 1980 1990 2000 Punjab 0.63 0.62 0.62 0.62 NWFP 0.68 0.69 0.65 0.70 Sindh 0.69 0.63 0.63 0.63 Balochistan 0.69 0.68 0.70 0.66 Source: Agricultural Census (Various Issues) 0.59 (1985 FA0 data). l2These landownership datalikelyunderstate the concentrationofland-holdings inPakistan(andother countries) since a household's land may be registered inthe names o f several household members. 12 3.6 Land sales markets inPakistan are extremely thin: according to Pakistan RuralHousehold Survey data, in2001-02 less than 0.2 percent o f total agricultural land was ~ 0 l d .Over l ~ time, however, land prices have risen substantially relative to landrents, making land less accessible for landless and small holders since agricultural incomes from purchased land may be insufficient to repay money borrowed for a landpurchase, even if credit were available (See Box 3.1). Box 3.1 LandPriceDeterminationinPakistan:LandRents andAsset Effects Conventional rent theory suggests that the price o f land i s determined by the discounted value o f the land's potential earnings (rents). Data on long-term trends inrental values and land prices incolonial and post-independence Punjab show a long term decline inthe rental value o f land relative to the price o f land, suggesting that agricultural earnings are not the main determinant o f land prices (Hirashima, 1996). Instead, asset effects, likely linked to the use o fland as a symbol o f prestige and power, and private capital formation innon-agricultural sectors, appear to be the major determinant o f landprices inPunjab inthe longrun. Duringperiods o f major technological change (the peak ofthe greenrevolution, 1976-86), however, rents rose faster than landprices inrural Punjab villages. Nonetheless, from 1960 to 1989, the ratio o f annual rent to land asset price (the WP ratio) declined from 3.93 percent to 2.59 percent inthe irrigated villages and from 2.85 percent in 1960 to 1.98 percent in 1989 innon-irrigated villages (Renkow, 1991). The large asset component inlandprices makes it very difficult for landlesshouseholds to purchase land, Highland prices increase the financial capital required, and low returns to land relative to these landprices meanthat agricultural incomes alone may be insufficient to repay money borrowed for a landpurchase. Only those who can afford to wait for the asset effects to be captured inthe long-run, without expectingmuchreturn inthe short-run, can participate inland market. Inthis situation, the income from outside the agricultural sector (including through remittances) are needed for most landless or small farm households to purchase land. Moreover, because increases inland prices from asset value effects vary according to the level o f infrastructure and the dynamism o f the local rural non- farm economy, uneven regional economic growth i s likely to increase wealth disparities across farmers inPakistan. Source: Hirashima, 1996 ;Renkow, 1991 ;Qureshi, 2003. 3.7 Landrentals are much more common than land sales, but still account for a relatively small share of area cultivated. In2000, 78 percent o f farms inPakistan, accounting for 73 percent o f area cultivated, were cultivated solely by the owners o f the land. Fourteenpercent of farms, with 12 percent o f total area cultivated, were cultivated solely by tenants. The remaining 8 percent o f farms (with 15 percent o f area cultivated) were farms consisting o f both owned andrented-in land (figure 3.1). Tenancy i s declining in Pahstan, however, between 1990 and 2000 the percentage o f farms with at least some rented-in land fell from 31percent to 22 percent. 3.8 According to the 2000 Agricultural Census, about two-thirds o f tenant-cultivated farms involve share-cropping arrangements where the landlord i s responsible for most input costs (such as seed, l3 Pakistan RuralHousehold Survey (PRHS) 2001-02. By comparison, annual tumovers o f land inLatinAmerican are often 10 to 20 times larger, averaging 5 percent inColombia, 2 to 3.5 percent inVenezuela, and 1.4 to 2 percent inEcuador (Jaramillo 2001). Incontrast, no formal landsalesmarkets inCambodia, China and the Lao People's Democratic Republic, where the state or the collective still owns the land. 13 fertilizer and pesticides), the tenants share insome other expenditures (such as water and rental o f machinery), and the output i s shared equally between landlord and tenant. Most o f the remaining one- third o ftenant-cultivated farms involve fixedrents paid incash. Figure3.1 Pakistan:Percentageof FarmsandArea Cultivatedby LandTenure Status 90 80 70 60 3 50 2 S 40 0 30 20 10 0 Owners OwnlRent Tenant 1 % of Farms E% of Area Source: 2000 Agricultural Census. 3.9 There i s substantial variance inthe incidence o f share-cropping across provinces, ranging from 11.4 percent o f farm area inBalochistanto 17.2 percent o f farm area inSindh. Although share-cropping declined from 1990 to 2000, the relative importance o f share-cropping across provinces has not changed. 3.10 The distribution o f operational holdings i s nearly as skewed as distribution o f land ownership. In 2000, 58 percent o f farms inPakistan were smaller than 5 acres in size. These farms occupy 16 percent o f total farm area. Incontrast, only 5 percent farms were 25 acres and above and total area under these farms was 38 percent in2000 (table 3.4). Trends indistribution o f land cultivated are likewise similar to those of land ownership, The number o f small farms (under 5 acres) has increased from 46 percent in 1960 to 58 percent in2000. Area under these farms has also increased considerably - from3 percent in 1960 to 16 percent in2000. The number o f farms inall other farm category shows a decline since 1960, except for the medium farms (5 to 12.5 acres). Table 3.4 Pakistan:Distributionof Farms andArea by FarmSize (operated area) Source: Agriculture Census (various issues) 14 3.11 Across provinces o f Pakistan, both owned land and operated land are most unequally distributed inthe NWFPandBaluchistan. Overthe pastthree decades, however, the differences across provinces have narrowed, as distribution in Sindh and Punjab have worsened, rising from 0.43 and 0.49, respectively in 1972 to 0.56 and 0.57, respectively in2000. Over this period, the gini coefficients for operated land area have remained almost constant inthe NWFP and Balochistan, varying between0.63 and 0.65. Land Distribution,Land Policy and LandReform 3.12 The current distribution o fland inPakistanreflects the government landpolicies under British rule and successive governments o f independent Pakistan. As part o f its strategy to "win friends amongst enemies", landrightswere granted to rural elites as a reward for cooperation and to encourage future assistance (Naqvi et. al., 1987).14 Land policies varied over time and across province, and included four concurrent major land administration systems prior to independence -- including the zamindari (landlord tenant), ryotwari (peasant proprietorship) systems, that established or encouraged an unequal distribution o f land (Box 3.2). LandReforms 3.13 After Independence in 1947, various tenancy acts were passed inPunjab and Sindh in 1950, followed by three national landreform acts (1959, 1972, 1977). Implementation o f these reforms was weak, however, due inpart to a general lack o fpolitical will and insufficient political strength o f tenants to counter the oppositionby landlords (Box 3.3). 3.14 Under the 1959 and 1972 land reform acts,15 the government acquired 1.6 million hectares (about 8 percent o f the country's cultivated area) from large landowners. About 1.3 millionhectares was redistributedto about 280,000 beneficiaries (an average o f 4.6 hectares per beneficiary). Much o f this redistributedland was not o f highquality, however, (less than three-quarters o f the distributedarea was undercultivation). Not allbeneficiaries were sharecroppers: a highproportion o frecipients were already small-landowners. 3.15 Despite continuing highlevels o f land concentration, further landreforms are currently o f f the agenda o f policy-makers inPakistan. The Shariat Court has given a rulingthat imposing a ceiling on land ownership i s in accordance with Islam, and that the land reform acts must be rescinded. The Government o f Pakistan has appealed this ruling, and the case i s still pending. l4NaqviN.H.,KhanM.H.andChaudhuryM.G.(1987) LandReformsinPakistan: AHistoricalPerspective, PIDE, Islamabad The 1977 land reform act was never implemented. 15 Box 3.2 Historical Rural Land Tenure Systems in Pakistan Both the Britishandthe Moguls before them granted large rent-free tracts o f land calledjagirs to individuals who had helped them inconquering parts o f Pakistan. Inthe Sindh, jagirdars often collected part af the produce from occupancy tenants, i.e. tenants who had permanent, heritable and transferable rightsto Eultivate the lands.' Inmany areas these occupancy tenants, called Mukhadims, did not cultivate the land themselves but engaged tenants-at-will (tenants with few, if any, legal rights, including occupancy rights), called haris. Similar to the jagirdari system o f land administration, the Pattadari system was operated inthe northem parts of Sukkur and Shikarpur districts inthe Sindh for Afghan settlers. These settlers collectively paid a nominal rent to the government, but an unequal distribution o f land holdings prevailed among individuals as these were based o n ancestral and customary rights. The Britishofficials also recognized the proprietary rights o f landlords (zamindars) who had acquired large estates during the periodo f political instability immediately before the extension o f British rule to the Punjab and other areas constituting present-day Pakistan. The construction o f the Sukkur Barrage in 1932and perennial canals inthe following years, much o f the newly irrigated land was purchased by zamindars at high prices and subsequently rented out to poor cultivators. InSindh, the Britishintroducedthe ryotwarisystem, under whichthe state kept the proprietary rightsto the land and leased the land to tenants who paid landrevenues only for the years they ploughed the fields. Likewise, the Mahalwari system (a refinement o f the ryotwari system) was established inthe easternpart o f the Punjab, by which the peasants o f a village were responsible collectively and individually for the payment o f land revenue to the British. & Ownership System Region Ri hts Mahalwari Eastern Punjab Individual Peasants responsible individually and collectively for payment of land revenue to British Jagirdari Punjab, Sindh Individual Tax-free tracts of land allocated to favored individuals Occupancy-tenants (Mukhadims) with permanent, heritable and transferable rights Mukhadims engaged tenants-at-will with few, if any legal rights Zamindari Pakistan State Large landlord-holdings with heritable, divisible and transferable rights provided land revenues were paid to State Ryotwari Pakistan State Peasant-holdings with heritable, divisible and transferable rights provided land revenues were paid to State Pattadari Sindh State Nominal rent paid to State by Afghan settlers collectively; unequal distribution of land holdings based on ancestral Source: Basedon Qureshi, Sarfraz. 2003. RuralLand Markets inPakistan: Institutions and Constraints. Background paper for PakistanRural Factor Markets Study. 16 Box 3.3 LandPoliciesandLandReforminPakistan Numerous landpolicies have been enactedinPakistan since 1947, both at the provincial and federal level The PunjabTenancy Acts (1887 to 1950) includedseveral important provisions limitingland ownership and conditions o f tenancy: i) Alandlordholdingmorethan100acreswastokeeponly50acresforcultivatinglandhimselfand the rest o f the land ownedby himwas to be given to the tenants. ii)Thelandlord'sshareoftheproducewasfixedat40percentandhewastopaygovernmentduesin the same proportion. iii)AllJagirswereabolishedexceptmilitaryJagirsandthoseconnectedwithreligiousandcharitable institutions. iv) Tenants couldbe ejected only ifthey failed to pay tax intime, didnot cultivate the land, or the landlordwanted the land for his own cultivation. v) Occupancy tenancy was abolished and the creation ofnew tenants was prohibited. In 1950, the Sindh Tenancy Act gave permanent rightsto the tenants (Hark) who hadcultivated at least four acres of land continuously for three years. It abolished `Beaar' (free services) and various illegal charges. Underthe 1959 Land Reformenactedunder MartialLaw, i)ceilings onholdings were fixed at 500 acres of irrigated land and 1000 acres of un-irrigated land; ii)tenants hadthe first claim to purchase the land resumed bythe government; iii)landlords were givencompensation through interest -bearing bonds; iv)jagirdari of all types was abolished without compensation; v) security o f tenure o f tenants was guaranteed; and vi) division o f land into uneconomic holdings was prohibited. Inaddition, a plan for consolidation of holdings was adopted, a comprehensive plan for land utilization was formulated, and credit facilities were envisaged for the new landowners. In 1972, more radicallandreforms (under MLR-115) includedthe following provisions: i)the ceilingon land ownership was reduced from 500 acres o f irrigated land to 150 acres and from 1000 acres o f unirrigated land to 300 acres; ii)land owned above the permissible limit was to be surrenderedto the government without compensation; iii)all the resumedlandwas to be distributed among the tillers free o f cost; iv) Shikaraahs (huntinggrounds), except those already ownedby government, were to beresumed;v) landover 100acres acquiredby a government servant during his term inoffice or two years after his retirement was to be confiscated; vi) all state land was to be givento landless cultivators or those having smaller than subsistence holdings on easy installments; vii) tenants could be ejected only ifthey failed to give the crop share or rent; and viii) the water rate and cost of seed was to be paidby the landlord. The 1977 landreforms were designedto be evenmore stringentinterms ofthe ceilings on landownership. However, these were never implementeddue to opposition by large land owners. LandAdministration,Land Titles and Contract Enforcement 3.16 Pakistan has an elaborate system o f land administration inherited from the British, involving the establishment of rules and regulationregarding sale, purchase and use o f landresources and the collection of land tax. This systemhas decayed inmany respects due to inadequate resource allocation for routine land administration, and the additional administrative burdens o f development work placed on revenue officials after independence in 1947. 3.17 The original primary intent o f agrarian laws inPakistanwas to facilitate collection o f land taxes and not to provide conclusive titles to land. As a result, neither the agrarian laws for providing titles to land nor the actual land records provide a firm foundation for efficient landmarkets or secure land tenure. Instead, the ambiguity inlaws has ledto incessant litigation and clogging o fthe judicial machinery inthe country. 17 3.18 The agrarian laws inPakistan dealing with land administration are patterned on the laws prevailing duringthe Britishperiod. Under those laws, the Kingo f Englandwas the eventual proprietor o f all lands. The rights of any individual to have any kindo f right for the land were subject to payment to himo flandrevenue unless a special waiver was granted fromthe payment o flandrevenues as for jagirdars 3.19 The major legal provisionsrelated to land inPakistan were codified inthe West PakistanLand Revenue Act, 1967, which unified the provisions of the Bombay (Sindh) LandRevenue Code, 1879, and the Punjab Land Revenue Act, 1887." Sections 42-45 o f this unified Act enumerate the rights (of landowners), and all land sales or transfers o f land are requiredto be recorded inthe records-of-rights according to specified procedures. Section 52 o f the Act attaches the presumptiono f correctness to the records-of-rights prepared at the time o f the settlement, as also to any entry made inthe records-of-rights or inthe periodical records inaccordance with the provisions laid down in sections 42-45 and the rules there under. 3.20 The present law does not profess to provide for a State certificate oftitle to land under the aegis of a public authority. The records-of-rights and other documents based on the land records, by virtue o f provisions inthe land laws are presumed to be accurate. These entries only providepresumptive status o f rightsunder land laws. Many court rulings have also maintainedthat entries inthe land records are not sacrosanct, that the revenue records are not the documents o f title, and that it i s permissible to challenge the entries for determiningthe title to land.I8This gap inthe law has been the basis o f litigation inland matter~.'~ 3.21 Preparation, maintenance and updating o f land records inPakistan are administered at the provincial level by a Board o fRevenue, which i s responsible for policy matters, and i s the ultimate forum for the final appeal o f all land disputes. The Board also organizes full-scale settlement operations for the periodic updating o f landrecords. At the district and sub-district levels, Directorates o f Land Records and field-level landmanagement officials (tehsildurs andputwaris) under the District Coordination Officer (DCO) are responsible for bothmaintenance and continuous updating o f landrecords.20 These basic land records at the field level consist of: i)the record o frightdperiodical record (along with the pedigree table of the owners); ii)the Register o f Mutations; iii)The Register o f Crop Inspections (Register Girduwari); and iv) the Cadastral Map (Shujru Kishtwur). l6TheGovernmentsofPunjabandNWFPenactedlegislationtoabolishJugirsin1950,butasimilaractionofthe Sindh government was successfully challenged inthe High Court in 1955. l7SincetheLandRevenueActwasunifiedin1967,thelandrevenueactsinallprovincesareidentical.Thereare, however, minor variations inthe administrative set-up o f the provinces for land management. '* The Lahore HighCourt (PLD 1966) held that "entries inrevenue record are not sacrosanct. They are certainly an important piece o f evidence but like all other evidence they can be countered". l9 The RegistrationAct, 1908, (the model Act on which the present Act inPakistan i s based), provides for the registration o f documents but not for the registration of titles. Registration o f documents i s compulsory only where some provision inthe Transfer o f Property Act (for example, Section 543 inthe case o f an outright sale o f an immovable property) or some provision inthe Registration Act (for example, Section 17 dealing with various transactions concerning immovable property) provides for compulsory registration. Inall other cases, unless providedby a special law, registration o f documents i s optional, particularly inthe case o f wills. The registering officer i s not supposed to concern himself with the validity o f the document. 2o The central survey and settlement organizations (within the Boards o f Revenue ) and the directorates o f land records are a small setup with the main mandate o f conducting training and providing technical advice to field officials and providing support to the Board o f Revenue. 18 3.22 Contract enforcement o f landrental contracts result innumerous legal disputes over illegal possession o f land, restoration of tenancy, incorporationo f transfer o f property (by oral transactions, registered sale deeds, or decree o fthe court) inthe revenue record, eviction o f tenants, land leases and grants, (particularly o f landjointly owned), and recovery o f rent. Although the Revenue courts have a legal mandate to adjudicate land disputes, many o f these cases are heard in civil courts.21Ingeneral, small land holders and tenants, tend to prefer to come to the revenue courts, however, because these courts are more accessible, cheaper and less time - consuming than the civil courts. Nonetheless, there i s a widespread perception that implementationand enforcement of decisions ofrevenue courts are often biased in favor o f large andpowerful land owners. 3.23 The ambiguity o fagrarianlaw regardingrecords of landrightsis particularly harmful to poor, who cannot afford protracted land disputes. The insecurityo f land tenure also serves as a disincentive for investment, reducing growth o f the agricultural sector. Unambiguous and authoritative laws, together with an effective landrecording system, would promote the smoothtransfer o ftitles and an efficient land market (Box 3.4). Box 3.4: LandRights inPunjab and Sindh InPunjab, oral mutationhas ledto fraudulent transactions entered intheRegister ofMutationsbythe patwari and hence to serious disputes inrural areas. Accordingly, the relevant sections o f the Transfer o f Property Act, 1882, were amended.However, the old system of oral mutation was revivedin 1978 under pressure from vested interest groups. According to the Punjab Land RevenueAct, 1967 andthe Punjab Tenancy Act, 1887, thejurisdiction o f the civil courts i s barred inrevenue cases, but the civil courts may take cognizance indisputes relatedto title. The common causes of dispute are the demarcation ofboundaries, recovery ofrent, restorationoftenancy, eviction of tenants, land leases and inheritance disputes.According to the Muslimlaw of inheritance, daughters, sisters, mothers, grandmother are also entitled to inherit property, but inthe agricultural community there is a tendency to deprive them o f their share o f inheritance. Women have become more aware o f their rightsnow becauseo f an increase ineducation and general awareness. InSindh, any personwho acquires aright intheproperty has to report orally or inwritingthe nature ofhis right to the specially set up cell at Taluka (sub-district) level within three months. The Mukhtiarkar makes necessarycorrections within 15 days. Ifobjections are received, the Mukhtiarkar inan open katchery disposes these of, There are serious defects in the system o f maintenance o f Records o f Rights in Sind. Most importantly, there are cases of corruption, public harassment and inefficiency. Moreover, the Mukhtiarkar or other officials manning the Taluka cells, may not be available for either receiving statements or for hearing objections, given their busy schedule and pre-occupation with many other administrative tasks. Source: Qureshi, Sarfraz. 2003. 21 Under the new systemof devolution introduced by the present government, the functioning of the Revenue courts remains muchthe same as before, with the work of Deputy Commissioner, Assistant Commissioner and Commissionernow transferredto District Officer (DO) (Revenue), DeputyDistrict Officer (DDO), and Executive District Officers [EDO] (Revenue). The LandRevenue Laws and operational capacity of the revenue courts remain the same, however. 19 Patterns of Land Use and Productivity by FarmSize and Tenure Status 3.24 Size o f farm and tenure status have clear implications for farm incomes, but their relationship with productivity has beenthe subject of considerable debate inPakistan. Economies of scale (especially with regardto mechanization), less aversionto risk, andbetter access to credit, purchased inputs, extension services and markets could potentially make large farms more productive than small farms. But more intensive use of own family labor and lower management costs per unit o f area farmed could result inaninverserelationshipbetweenfarm size andproductivity.For farms ofequal size, owner-operated farms might be expected to be more productive than tenant-operated farms, because o f their relatively greater incentives to invest inland and tube wells which they themselves own. Moreover, under share- cropping arrangements where the costs o f some inputsare borne by the tenant, a profit-maximizing tenant will use less inputs andproduce less output per hectare than he would if owned the land (or paida fixed rental amount).*' 3.25 According to agricultural census data, cropping patterns vary little by farm size and tenure, except for very small farms (less than 0.5 hectares) (See table 3.5). Wheat i s the predominantcrop across all farm sizes, accounting for 36 to 43 percent of area. Among the major field crops, only for maize and fodders, and to a lesser extent, rice and cotton, do cropping patterns vary significantly by farm size. Maize i s the second most important crop (accounting for 21 percent o f area cultivated) for the smallest farms (less than 0.5 hectares), but accounts for an average o f only 4 percent o f area cultivated o f all farms. By contrast, the share ofrice and cottonintotal area cultivated o fvery small farms (5 and 6 percent, respectively, in2000) i s less than halfthe average area cultivated for farms o f all sizes (12 and 14 percent, respectively). These differences to a large extent reflect distribution o f irrigated land inthe main areas o f cultivation o f these crops. Similarly, vegetables and orchards occupy a significantly larger proportion o f the cultivated acreage o f the largest farms than o fthe smallest farms. Agricultural Census data show no distinct patterns across tenurial status interms o f the crops grown, however. 3.26 Data on farm incomez3per cultivated area across ago-climatic region in the Punjab for 2000/0 1 suggests that small farms are more productive interms o f net revenue per unit of land cultivated than are larger farms, particularly on irrigated land. Farm incomes per cultivated area o f farms less than 5.0 acres insize inthe partialbaranizone (dependent onrainfall andwells for water) andthe irrigatedzone (with canal, well and rain water) were 3.1 and 1.6times farm incomes per cultivated area o f farms greater than 25 acres inthe respective zones. There are only small differences inincomes across farm size inbarani (rainfed) zones, however, and the estimated differences inincome per acre among farms greater than 5 acres i s only 8 to 16 percent inirrigated zones (figure 3-2). 22 See Nabi, Hamid and Zahid (1986) for a detailed discussion of farm size and productivity, and economic incentives under tenancy. 23 Farmincome includes both crop and livestock income. 20 Figure3.2: Pakistan:FarmIncomeper CultivatedArea inPunjab,2000/01 35,000 30,000 $ 25,000 Less than 5 20,000 5 to 12.5 u) 0 012.5to 25 Q, 15,000 a Q D25andabove 10,000 5,000 Barani Partial Irrigated Average Barani Source; PEN (2001) Note; (1) Barani zone consists o f four districts o fPunjab namely Attock, Rawalpindi, Chakwal and Jehlum districts. (2) Partial Baranythal zone: consists o f Gujarat, Gujranwala, Sialkot, Sheikhupura, Mandi Bhaudin, Lahore, Narowal, Kasur and Hafizabad districts. (3) Irrigatedarea: consists o f remaining 25 districts inthe province o f Punjab. N miN N N b N b N 3 9. m' N n B 0 0 v 0 (d 22 3.27 However, most recent econometric evidence suggests relatively small diseconomies o f scale inphysical production (table 3.6). Plot-level regressions o fproductivity, correcting for plot characteristics and some household characteristics (such as tractor ownership andnumber o f householdworkers) suggests a doubling o f operated area leads to 10 percent lower wheat yields (and 13 percent lower rice yields), (World Bank, 2002). Similarly, controlling for access to credit, Jacoby and Mansuri (2004b) findthat a doubling o f plot area leads to a 12 percent reduction in gross productivity (controlling here for access to credit). 3.28 Estimates o f scale effects on netrevenue per acre (Le. gross revenue less the value o f inputs)differ widely though. World Bank (2002) estimates that a 100percent increase inplot area results ina 22 percent reduction inkharif season and 38 percent inrabi season net revenues. Incontrast, correctingfor landowned, Anriquez andValdes (2004) findanoverall elasticity of net revenues with respect to total operational size o f a farm o f 0.33. Intheory, if land rental markets worked efficiently, landrental rates would equal marginalreturns to land, and net revenues would exhibit constant returns to scale. 3.29 Econometric evidence on the effects of tenancy suggest that the productivity o f share- croppers who are relatively un~upervised~~ i s 18 percent less than productivity o f supervised share-croppers and farmers who cultivate their own land." Since only 19percent o f land in Pakistani s share-cropped, and an estimated 35 percent of tenants are relatively unsupervised, "unsupervised" tenants account for only about 7 percent o ftotal area cultivated and the effect on total production o f major crops i s small: only 1.4 percent (Jacoby and Mansuri, 2004a). 3.30 Lack o f secure tenancy arrangements also has negative implications for productivity through disincentives on long-term investments by the tenant inland quality. Econometric analysis o f PRHS 2001-02 data, (Jacoby and Mansuri, 2003), shows that farmers used less manure on leasedplots than on their own plots, reducingyields by about one percent inthe first year and a cumulative total o f 2-2.5 percent in future years.26 These effects on yields are relatively small, but disincentive effects o f tenancy contracts may be larger for investments in canal lining and tube wells. 24 Inthis study, tenants were considered to be unsupervised, if during the cropping season there were less than ten meetings between the tenant and landlord to discuss management o f the plot. 25 Share-cropping arrangements benefit share-croppers by enabling them to share the risk o f output fluctuations with the landlord, and also to share the costs o f purchased inputs. However, they provide less incentives for labor input by the tenant, since the output i s shared, as well. 26 Application o f manure increases yields, by providing additional soil nutrients and by improving soil structure, aeration, water retention, and ability to retain nutrients. m c.l a E: cl ed 8 IV. WATER MARKETSINPAKISTANAGRICULTURE 4.1 Like the supply o f cultivable land, the supply o f water inPalustan i s essentially fixed within a year, though with large year-to-year fluctuations. Large-scale movements o f water from one location to another take place through the extensive Indusbasin canal system inPakistan, notwithstanding substantial conveyance losses. Small storage capacity relative to water flow and lack o f control structures (such as gates) greatly limit the control o f water flows within the system, however. Distribution o f this canal water at the water course level i s determined by physical location along the canal and the wavabandi system o f established, administratively-set rotations. 4.2 Inpartbecauseofunreliabilityofirrigationbycanalwater, tubewellirrigationhasgreatly expanded since the 1960s, and has ledto increased agricultural productivity by providing greater water control at the farm level. Groundwater markets have increased productivity o f farmers who lack their own tube wells, but surface water markets are limited to informal transactions within water courses due to both insufficient water storage and control structures, as well as the absence o f tradable water rights. Water Availability inPakistanAgriculture 4.3 Given low levels o f annual rainfall concentrated over only a few months o f the year and period droughts, water availability at the farm level i s a major determinant of farm productivity inPakistan agriculture. Nearly 80 percent o f cropped area i s irrigated, and agriculture i s by far the largest user o f available water sources inthe country, consuming on average about 95% o f available water resources. 4.4 Increases inwater resource availability for irrigation and expansion o f irrigated area have played a central role inagricultural growth since the 1960s. Total irrigated area increased by 80 percent between 1960 and 2000, from 10.4 to 18.0 million hectares, mainly due to an expansion intube well irrigation. In2001-02, out o f 18.0 million hectares o f irrigated land, 38 percent was irrigated solely with canal water; 39 percent with canal andtube well water, and 19 percent solely with tube well water. (Other forms o f irrigation, including wells, canals with wells and tanks) accounted for the remaining 4 percent). (figure 4.1). About 60 percent o f irrigatedwater available at farm head i s provided by canal water; the remaining 40 percent i s supplied by ground~ater.~' 4.5 Nearly 97% o f this irrigated area i s inthe IndusBasin, the largest contiguous irrigation system inthe world (Box 4.1). Onaverage, approximatelytwo-thirds oftotalnet water inthe Indusbasin (103.8 out o f 151.6 MAF) i s divertedto canals. Of this total, about 45 percent goes for consumptive use, with the remainder as conveyance losses (including groundwater recharge). Surplus flows to the estuary (29.4 MAF);system losses (9.9 MAF) and IndusTreaty Allocations to India (8.4 MAF) account for the remaining water. (Table 4.1) ~ 27 GOP, Water Vision 2025 Country Report, cited inWorld Bank, 2003a, Pakistan Public Expenditure Management, Accelerated Development o f Water Resources and Irrigated Agriculture, vol. 11. 25 Figure 4.1 Area irrigatedby source of irrigation, Pakistan 1982-2001 I I 9 1 I Source: Economic Survey 1996-97 and 2001-2 Table 4.1 Average Water Balance of the IndusRiver System Item Annual Volume MAF (millionacre feet) BCM(bn cubic meters) INFLOW Westem Rivers 143.18 176.54 Eastern Rivers 8.40 10.36 Total Inflow 151.58 186.90 OUTFLOW Indus Treaty Allocation to India 8.40 10.36 Average Canal Diversions 103.84 128.03 System Losses 9.90 12.21 Surplus flow to lower River and Estuary 29.44 36.30 Total Outflow 151.58 186.90 Source: World Bank (2003a). 26 Box 4.1: The Indus BasinIrrigation System The Indusbasin irrigation systemconsists of 3 major reservoirs (Tarbela, Mangla and Chasma dams), with 19 barrages or headworks, 12 link canals and 43 canal commands and about 107,000 water courses. Within each canal command there are typically 4-6 branch canals, 2 distributaries per branch canal, 20-30 minors per distributary, 20-30 water courses per minor. A typical water course i s 2-3 kilometers inlength, irrigating200-700 acres (80-280 hectares) for 20-30 farmers. Inthe Indussystem,river water is divertedbybarragesandweirs intomaincanals and subsequently into branch canals, distributaries and minors. The flow to the farm is deliveredby the watercourses which are suppliedthrough outlets (moghas) from the distributaries and minors. The mogha is designedto allow a constant discharge that self adjusts to variations inthe parent canal. Within the watercourse command, farmers receive water proportional to their landholding. The entire discharge o f the watercourse i s givento one farm for a specified period on a seven or ten day rotation, called "warabandi". Flat topography and lack o f well-definednatural drainage inthe IndusPlaincreate a surface drainage problem which has been compounded by the construction of roads, railways, flood embankments, and irrigation systems that obstruct natural drainage flows. Inspite of investments insurface drains and deep tube wells, (the latter installedas part of SCARPS-- Salinity Control and Reclamation Projects), drainage remains a major problem. Source:World Bank (1994). PakistanIrrigationandDrainage: Issues and Options. 4.6 Conveyance efficiency inthe system i s low, however (45 percent). Inadequate maintenance results incanal losses (up to canal outlets) o f 25 percent. Unlinedwatercourses result inanother 40 percent loss from canal outlet to farm gate. Though these conveyance losses reduce availability o f surface water, they nonetheless contribute to groundwater resources.28 4.7 Surface water availability at the farm head inthe IndusBasin System has changed little since the late 1970swhen the Tarbela dam was completed, remaining at about 62 MAF per year. The major increase inavailability has been due to the growth ingroundwater use - from 4.2 MAF (8 percent o f water availability) in 1960-61 to over 40% inthe late 1980s, after which time it appears to have changed very little. This large growth ingroundwater use was stimulated inpart by the use o f tube wells for vertical drainage and water table control under the Salinity Control and Reclamation Projects (SCARPS),which are now being gradually turned over to communities and Water Users Associations (WAS),as well as bythe large spurt infarm mechanizationinthe 1980sandthe availability of inexpensive tube well technology inthe market. However, one critical consequence o f this growth i s that increasing amounts o f salt are circulating vertically inthe system, progressively poisoning the soil because drainage and evacuation o f salt from the basin i s inadequate. (World Bank, 2003a). 28 Infreshgroundwater areas, canal andfield seepage are anincreasingly important source ofirrigationwater (thoughthey are indeedlosses insaline groundwater areas). This seepage, however, essentially reallocateswater from those downstreamwho only have access to canal water to those with access to tube wells. Moreover, since conjunctive use of canal water and groundwater to control water logging andsalinity has been akey principle underlying the overall irrigation development strategy since the 1960s, one cannot look at either canal water or groundwater inisolation inthe Indus Basin system. (World Bank, 2003a) 27 4.8 L o w prices o f energy (electricity and fuel), which remainedinplace untilpricingreforms inthe late 1990s), also contributed to the growth intube well irrigation. Removal o f these subsidies have increasedthe cost o f groundwater for tube well owners and purchasers ingroundwater markets. The l o w efficiency o f pumps also contributes to highoperating costs. 4.9 Moreover, inpart because o f successive droughts, water tables inPunjab, NWFP, and Baluchistan have receded, adding to costs o fpumpingand contributing to reduced quality. Ali and Byerlee's2' decomposition o f crop sector productivity growth inPunjab from 1971-94 showed that on average about half of the positive contribution of technological change (cropping intensity and new crop varieties) and public investment (roads and literacy) was offset by soil and water quality degradation due to water-logging and salinity. Accessto Water and Agricultural Productivity 4.10 Surface Water. The development o f surface irrigation inPakistan permitted the extension o f cultivation into areas and seasons that lack sufficient rainfall for agriculture. However, the public irrigation systems do not provide farmers with adequate water or enough control over irrigation deliveries to meet the demands o f more input-intensive green revolution technology and increasing population pressure. Moreover, problems with operation and maintenance o f the canal systems have led to increased water losses inthe channels, but no allowances for these water losses are made inthe water rotation system. 4.11 Since the time o f the BritishRaj inPakistan, access to surface water has been tied with land ownership. Canal water i s distributedto each acre on a water course according to rotation system (warabandi),with each land owner or tiller o f land allotted a turnto use water flow inthe canal at a specified time every seven to ten days (depending on the geography o f the water course). The access to water i s limited to a farmer who owns land ina water course command area and the time allotted to each turn i s directly proportional to size o f holding, irrespective o f the cropping pattern. The duration varies from 20 to 30 minutesper acre. This distribution system was designed for cropping intensity o f 70 percent, but the existing cropping intensity is126 percent (Government o f Pakistan 2002). As a result, tail-end distributaries and watercourses do not receive enough canal water for the current cropping intensity, and delivery schedules are unreliable inmany areas.30 4.12 Water charges are assessedon the bases o f crop land, the relative burden o f payable water rates per canal command area i s essentially determined by the intensity o f cropping. Water rates vary by crop, and there i s systematic relationship between water rates and water requirements for various crops. For example, water charges for rice, wheat and sugar cane are Rs. 2.0, Rs. 3.0 and Rs 2.0 per acre inch. The present level and structure o f water charges constitute only a small fraction o f cash production cost and net farm incomes (varying from 8 percent to 26 percent o f total cash costs).31 Thus there i s considerable scope for increases inwater charges to farmers to cover the operation and maintenance o f the irrigation systems. 29 Productivity Growth and Resource Degradation in Pakistan's Punjab: A Decomposition Analysis; Mubarik Ali and DerekByerlee; World Bank Policy ResearchWorking Paper No. 2480; November 2000. 30 Monthly canal withdrawals vary from month to month depending upon the rainfall and decision o f the Indus River System Authority. The average monthly flow i s 4.7 MAF with a coefficient variation (CV) as 32 percent in Punjab, 3.9 MAFand C V as 5 1percent inSindh, 0.055 MAFand CV as 66 percent inNWFP, and 0.26 MAF and CV as 67 percent inBalochistan. (Monthly flow data for 1999;WB 2003, p.xx) 31 Z. Hussain, 2003. 28 4.13 Ownership of a plot within a watercourse command area confers access to irrigation water, but does not guarantee canal water availability. Inparticular, water availability decreases significantly if the watercourse i s located near the tail o f the distributary or minor, and/or ifthe plot i s located near the tail end o f the watercourse (World Bank, 2002). 4.14 Water theft by farmers upstream i s a major reason for water shortages faced by tail end farmers. For example, inthe Nara canal, an estimated 30-40 percent o f the water i s diverted through illegal direct o~tlets.~'Administrative reforms giving greater control to farmers as part o f the National Drainage Program (box 4.2) may helpreduce these misappropriations o f water, however. Inthe command area coveredby the Left Bank Area Water Board, where tail end farmers hadnot received water inthree years, 48 illegal direct outlets serving an area o f about 25 thousand hectares were closed inearly 2004. Box 4.2: AdministrativeReformsin CanalWater Management Within the framework o f the NationalDrainage Program, a package o f administrativereforms were to be implemented inone canal command in each province. These reforms consist primarily o f decentralization and management transfer o f the irrigation and drainage system to water users. In this comprehensive approachto CanalManagement: i)federal agencies, notably the Water and Power Development Authority's (WAPDA's) Water Wing, are strengthenedto better implement their federal responsibilities; ii)Provincial Irrigation and Drainage Authorities (PIDAs) are established with a certain degree o f autonomy to oversee and provide technical assistance to the transfer process as well as to agree on water allocation and monitoring; iii)Area Water Boards (AWB) are establishedat the level o f a canal commandto manage the transfer process at the canal and distributary level with the participation o f Farmers' Organizations; iv) Farmer Organizations (FOs) are promotedto take over operations andmaintenance at the watercourse level bypermitting assessment, collection andretention o f the abiana; and v) development o f water markets and individual water property rights i s fostered. The degree o f implementation o f these administrative reforms has varied considerably across province. Altogether, only the Sindh Province has implementedthe reform package and applied it to the Nara canal command area. While a thorough impact assessment o f the reforms inthe Nara canal i s needed, initial evidence suggest that this participatory irrigationmanagement approach has resultedin increased efficiency, effective participation o f farmers inoperations and maintenance as well as indecision making over the allocation of water, andreduction inwater theft. Source:World Bank 1997. Staff Appraisal Report: PakistanNational Drainage Program Project, April. ReportNo. 15310-PAK. 4.15 Eveninthe absence ofwater theft, there is a lack o ftransparency inwater allocation that derives largely from the lack o fparticipation o f farmers inwater management decisions and the tendency o f the departments to centralize management decisions such as water allocation. Some farmers attempt to influence these water allocations at the local level through informal payments to officials and political pressure, however. Regression analysis shows that when the data are controlled for plot location, the availability o f canal water increases significantly where farmers reported informal 32 Manuel Cantijoch, Senior Water Resources Specialist, World Bank. 29 payments for water; where there i s less inequality o f land ownership, and where the land holdings o f the three largest cultivators on the watercourse is high.33 4.16 Sales of canal water are technically illegal, but limitedsales and exchanges do take place. According to PRHS (200 1-02) data, average farm sales/purchases o f canal water were only 5.5 hours, equal to only 2.7 percent o f canal water endowments, though inone district (N.Shah in Sindh), average farm sales/purchases were 25.6 hourdfarm, equivalent to 39 percent o f canal water endowments. Average non-market transactions were even smaller (5.2 hourdfarm, only 2.5 percent o f canal water endowment^).^^ The PRHS data apparently do no capture informal canal water trading, however. According to a 2003 survey incentral Punjab, few farmers buy or sell canal water (less than 6 percent) for cash, with the price of canal water ranging from Rs 125 to 240 per Incontrast, morethan 80 percent o f the farmers at the headand middle o f distributaries, (and 50 percent o f farms at the tail) traded water. Groundwater 4.17 Availability of canal water at the farm level varies significantly, not only because o f weather conditions that affect the amount o f water inthe Indus river, but because o f the uncertainties in distribution relatedto water misappropriationdiscussed above. Groundwater supply, incontrast, i s more reliable, and thus helps overcome water shortages, as well as improve irrigation timing. Largely for these reasons, groundwater use has expandedrapidly inPakistan (World Bank, 2003a). Inaddition, untilpricingreforms inthe late 1990s, tube well irrigation was subsidizedthroughlowprices o f energy (electricity and fuel). 4.18 The l o w efficiency o f pumps contributes to highoperating costs. Moreover, inpart because o f successive droughts, water tables inPunjab, NWFP, and Balochistan have receded, adding to costs o f pumping and contributing to reduced quality. Water quality i s also increasinglya problem in Sindh. 4.19 PRHS (2001/02) data indicate that only 8.1 percent o f cultivating households owned tube wells, with little systematic variation by income (proxied as total consumption). Evenamong the poorest 20 percent o f cultivating households, 5.1 percent owned tube wells; 11.O percent o f the richest 20 percent of cultivating households owned tube wells. (p.86 WB 2002). 4.20 Groundwater markets significantly improve access to groundwater, particularly for small farmers, landless tenants and younger households who often lack the resources (or land and water rights) to installtheir own tube well. However, water purchasers do not have full access rightsto the water, and are frequently denied access when water or energy supplies are scarce. Unreliability o f access to purchased water was a problem for more than half of the water buyers in an IFPRI sample survey in 1993, limitingthe productivity o f groundwater for purcha~ers.~~ 4.21 However, current informal water markets inPakistan are fragmented and localized, and are usually supply-driven. Farmers offer water for sale only after satisfying water needs on their own fields. According to a 2003 survey incentral Punjab, few farmers buy or sell canal water (less than 6 percent) 33 One reason for the positive correlation between larger landowners and water availability i s that these cultivators are likely to be able to lobby more effectively with irrigation officials and politicians. World Bank 2002, p. 90). 34 Anriquez and Valdes, 2004, p.8 35 Z. Hossain (2003). 36 Meinzen-Dick (1996). 30 for cash, with the price o f canal water ranging from Rs 125 to 240 per hour. Incontrast, more than 80 percent o f the farmers at the head and middle o f distributaries, (and 50 percent o f farms at the tail) traded water. Tube well water sales are muchmore common: more than 90 percent o f farmers with tube wells sold tube well water, with the mean quantity sales of 5 hours. 4.22 A flat charge per hour o fpumpingis the most common form of groundwater market contract in most o f the groundwater market areas. This type o f arrangements occurs under all types o f tube wells. Water from diesel pump sets i s commonly sold under an arrangement whereby the buyer supplies the diesel and motor oil for the pump and pays an additional fee per hour to the well owner to cover the wear and tear o n the engines. Share cropping contracts for water are usedunder both diesel and electric tube wells. 4.23 Prices under the hourly charge system depend on the pump type, capacity and location. The higher prices o f water from PTO tube wells reflect the higher cost o f operating this type o fpump.The average price of water under the hourly charge system i s approximately the same for diesel and electric tube wells, although the former are usually more expensive to operate. The meanhourly cost o fwater to the purchaser from diesel tube wells i s slightly higher under the buyer-brings-fuel system than under the flat hourly charge. Water sellers with diesel pumps are apparently only recovering their own operation and maintenance costs under either type o f contract. The seller's transaction costs inacquiring the fuel and operating or supervising the operation o f the pump are presumably higher under the hourly charge contracts but there may be reluctance to let some purchasers operate the pumps themselves under the buyer-brings-fuel system. Implications for AgriculturalProductivity 4.24 Data on water use and crop incomes suggest that large gains inaverage water productivity may be possible through shifts incropping patterns inducedby increasing the opportunity cost o f water to farmers. The average net income o f water for IRRIrice inPunjab i s only Rs 47 /acre-inch o f water compared to Rs. 215 for cotton (table 4.2). The corresponding figures for Sindh are Rs 66 for IRFUrice and Rs 361 for cotton.37Increases inthe opportunity cost o f water for IRRIrice farmers through the opportunity to sell water rights could promote more efficient use o f water inIRRIrice cultivation, as well as increase incentives and water availability for cultivation o f less water-intensive crops. Similarly, better allocation o fwater could even bring about an increase inarea cultivated insome regions. For example, because o f uncontrolled direct outlets inthe Nara Canal in Sindh, the command area has increased by 30 percent, leaving large areas inthe tail portion o f the canal without sufficient water. 4.25 Increasing water availability to tail-end farmers could also have large effects on productivity. Crop yields o frice and wheat are 1.3 to 1.5 times higher at the tail o f canals than at the head; sugar cane yields are 1.9 times higher at the tail than at the head3*,39 37 These data need to be interpreted cautiously since water quality and availability o f supplemental tube well irrigation vary within provinces. Moreover, during the kharif season, there are few alternatives to rice on water- logged land. 38 National Agricultural Commission Report (1987), p. 289. 39 Incontrast, econometric analysis of survey data fromthe early 1990s (Jacoby andMurgai, 1998)suggests that at a micro-level (withinthe boundaries o f a water course), there were only minor productivity losses to tail end farmers, Moreover, the expansion of private tube wells may have even further reduced the productivity differential between head and tail farmers. 31 Table 4.2 Returnsto Irrigationof Major Crops, Pakistan2002-2003 Gross ProvinceJCrops Revenue Total Cost Net Income Irrigation Water Net IncomebVater Use Rs Rs Rs UseAcre Inches Rs/acre inch Punjab Cotton 15,093 11,658 3,434 16 215 Sugarcane 25,425 15,622 9,802 64 153 Wheat 9,000 6,130 2,569 16 161 Rice Basmati Rice 13,395 5,722 7,622 64 119 IRRIRice 9,387 6,405 2,98 1 64 47 Sindh Cotton 13,554 6,333 7,220 20 361 Sugarcane 37,181 21,375 15,806 64 246 Wheat 8,325 6,913 1,412 16 88 IRRIRice 10,374 6,140 4,234 64 66 NWFP Sugarcane 32,199 211,163 11,036 64 172 Note: Net income excludes irrigation cost. Source: Agricultural Prices Commission (APCOM), Ministryo f Food, Agriculture and Livestock, Government of Pakistan, Cost o f Production Studies, series 2002-2003 4.26 Increased availability o f groundwater could have even greater effects. Meinzen-Dick (1996) showed that groundwater has a higher impact on wheat yields than canal water, and that water markets that increased groundwater available to farmers who do not own tube wells clearly increasedthe level of agricultural productivity. Water from own wells was even more productive, most likely because o f the less regular availability o f purchased water.40 4.27 Analysis o f PRHS 2002-03 data on groundwater use per acre at the plot level suggests that markets for tube well water markets do not function efficiently. Controlling for tube well availability, plots cultivatedby tube well-owning households have a 15 to 17 percent higher utilization rate than do households without tube wells. Iftube well markets functioned efficiently, water sales would result ina smaller difference inwater use between owners o f tube wells andbuyers o f tube well water. The estimated impact on productivity i s very large and highly significant inthe rabi season: ownership o f a tube well, controlling for access o f a plot to a tube well (whether owned or not) raises wheat yields by 30 percent and rabi seasonnet revenues by 83 percent.41 Laws and Regulatory Framework for Water 4.28 Efficient water markets, tradable water rights and physical infrastructureto transfer water and measure water flows would enable farmers to maximize their incomes by allocating water according to 40 Meinzen-Dick (1996). 41 Pakistan Poverty Report, 2002, pp. 99, 101. 32 its most optimal use (whether on their own fields or sales to or purchases from other farmers). The current regulatory framework inPakistan, however, impedes the development o fwater markets. 4.29 Pakistani Law treats all surface water as provincial property. Under the Pakistan Constitution, while the Federal government i s responsible for regulation and development o f inter-state rivers and river valleys, provincial govemments are responsible for water supplies, irrigation and canals, drainage and embankments, and water storage .Nonetheless, the power o f the provinces has emerged as pre- eminent and water has come to be perceived as a provincial subject. 4.30 This has serious ramifications for inter-provincial water development and allocation, since basins (including the Indusbasin) extend across province boundaries. Inthe absence o f legal clarity on what individual provincial shares are, eachprovince has argued for as large a share as possible. It i s therefore important to introduce necessary legal arrangements to facilitate the management o f surface water on a river basinbasis, before formal water markets can be introduced. Government has constituted the IndusRiver System Authority (IRSA) which decides inter-provincial allocation water based on the Water Accord o f 1991. But, disputes are quite common among the provinces, and allocations are still made on the basis o f historical distributions o f water. 4.3 1 Individual usufructuary rights for surface water are not clearly defined, however, as the legislation has failed to devise a system for providing secure, defensible and enforceable surface water rights.Unless surface water rightsare better clarified and infavor o f individuals, conflict and litigation will grow inthe future and formal water markets will not be possible. 4.32 Separating rightsto groundwater from rights to land i s another necessary step for groundwater market developments. According to the riparian allocation o f water inPakistan, ownership o f groundwater belongs to the owner o f the land above. By virtue o f these laws, groundwater i s "attached landproperty" and cannot be transferred separately from the land to which it i s attached. This has limited the potential for inter-sectoral allocation. For instance, a local municipal committee or council i s not authorized to have access to groundwater from the neighboring farms without buyingtheir land. To establish an active formal water market, rightsto water use must be separated from land for alternative use. 4.33 Finally, existing laws place no quantitative limits on groundwater withdrawal by individual users. The lack of these limits, coupled withtying landrights with water rights, has serious equity implications, because it allows larger farmers with higher pumpingcapacity and deeper tube wells to have a disproportionate claim over water than others. Establishment o f withdrawal limits would also promote efficient water use. Inaddition, an efficient market requires the prior existence o f an effective legal institution of property rights, establishing the initial resource endowments o f individuals. There i s therefore a need to specify water withdrawal limitsby individuals involumetric terms. Finally, ecological sustainability requires that collective withdrawal limits keeping inview the annual recharge, V. RURALLABORMARKETS 5.1 Unlike the markets for land and water, rural labor markets are generally characterized by demand constraints, rather than supply constraints, as evidenced by substantial underemployment. Comparedto land and water, labor i s much more mobile across sectors, and labor migration between regions and betweenrural and urban areas helps link labor markets across the country. Incomes from labor are a major source o f incomes for the rural poor, who often lack sufficient access to land and water. Nonetheless, low levels o f educational attainment, lack o f marketable skills, low female participationrates, and ineffective enforcement o f laws regarding child and bonded labor, all contribute to l o w productivity o f labor and reduced labor earnings. Structure of L a b o r Use 5.2 Two-thirds o f Palustan's populationlive inrural areas (100.6 million out o f 149.0 million, in 2002-03). Of this total, 29.82 million (29 percent) participatedinthe labor force inwork other than householdwork. Counting only population age 15 and above, the labor force participation rate was 50 percent (table 5.1). Reported unemployment i s low (5.7 percent in 1997-98), reflecting the predominance o f informal sector and self-employment inthe labor market. Estimated underemployment (defined as employment less than 35 hours per week) i s considerably higher, however (16 percent, according to the 1999-2000 Labor Force Survey). 5.3 Aggregate labor force participation rates have changed little over time, though official figures for unemployment rates doubled from 3-4 percent inthe mid-1980s to 7-8 percent in2002-03 (table 5.2). Reportedunderemployment rates have also increased, but at a lesser pace. Underemployment in rural areas rose from 12percent in 1985-86 to 16 percent in 1999-2000. The highestrates were found inNWFP(21 percent), andthe lowest inBalochistanand Sindh(12 and 11percent, respectively). 5.4 Participationrates vary sharply by gender: 85 percent o f men age 15 years or greater are employed or actively seeking employment, as comparedto only 16 percent o f women inthe same age group, Including housekeeping and other related activities as employmentraises the calculated participation rate o f women age 15 years or greater to 50 percent. These low female participationrates inpartreflectculturalnorms andindividualpreferences. Nonetheless, various studies have found that females' decisions to participate inthe labor market are significantly influencedby household income, family size, educational attainment, and the market wage rate. 34 Table 5.1 Pakistan RuralLabor Force, 1998 (thousands) Male Female Total Population 43,830 41,233 85,063 51.5% 48.5% 100.0% Working Age Population 29,640 28,099 57,739 51.3% 48.7% 100.0% Labor Force (age 10-14 years) 1,015 392 1,407 72.2% 27.8% 100.0% (age 15+ years) 19,733 3,407 23,141 85.3% 14.7% 100.0% Total (age IO+ years) 20,748 3,799 24,548 84.5% 15.5% 100.0% Participation Rate (age 10-14 years) 17.2% 7.6% 12.7% (age 15+ years) 83.1% 14.8% 49.6% (age 1O+ years) 70.0% 13.5% 42.5% Total Population 47.3% 9.2% 28.9% Employed Labor Force 19,812 3,349 23,161 85.5% 14.5% 100.0% Unemployed Labor Force 936 451 1,387 Unemployment Rate (97-98) 4.5% 11.9% 5.7% Source: Pakistan PopulationCensus and Labor Force Surveys. Total unemployment rate from Table 12.9Pakistan Economic Survey 1999-2000. 1999-00 I 39.84 28.03 I 28.97 29.82 I 7.82 6.94 13.64 1 16.08 42.38 II 28.97 II 7.82 II 6.94 II I2002-03 I 29.82 I 29.82 I I Source: Labor force survey (various issues), Government o f Pakistan 5.5 Educationlevels o f the rural labor force are low, but have improved over time. Inthe late 1960s, 75 percent of employedpeople inPakistan were illiterate. By 2001-02 this percentage had declined to 53 percent overall, and 62 percent inrural areas [HIES (2001-02)]. Econometric analysis indicates that an education i s positively correlated with the probability o f rural-urban migration, labor productivity inthe non-farm sector, and female labor force participation. Higher education i s also positively correlated with unemployment, as skilled workers who cannot findjobs intheir fields often 35 choose to remain unemployedrather than take unskilledlabor positions. For example, in 1999-00, among total unemployed inrural areas, 61 percent were literate. Moreover, higher education i s correlated with increasedrural household incomes. For the average producing farm household, another year o f education translates into a 13% increase inyearly net revenues (Annex 2; Anriquez and Valdes, 2004). SectoralComposition ofLabor Use 5.6 The agricultural sector accounts for two-thirds of rural employment (excluding household activities), mainly own-family labor. There are significant variations across provinces, however. Agriculture1accounts for 74 percent o frural employment inSindh, but only 54 percent o f rural employment inNWFP, where the shares o f trade, transport and other services are higher than the national average (table 5.3). Table 5.3 Percentagedistributionof rural employedpersonsby major sectors andprovinces (1999-00) Source: Labor force survey (1999-00), Government o f Pakistan 5.7 Overall, 15 percent o f people employed in agriculture were underemployed, accounting for 63 percent o fthe total number o f underemployed persons (table 5.4). Nearly half of the people employed inconstruction were underemployed. Underemployment inservice sectors (not including trade, restaurants and hotels) was 24 percent. Table 5.4 Underemploymentratesby Occupation(1999-00) Proportion in total Proportion in total employed underemployed Agriculture, forestry, hunting and fishing 63 15 Manufacturing 8 14 Construction 12 47 Wholesale, retailitradelrestaurantsihotels 4 10 Community, social andpersonal services 11 24 Total 100 17 ~Source: Labor Force Survey (1999-00), Government o f Pakistan 36 5.8 Econometric analysis o f the 2001 Labor Force Survey data (Annex 5.1) suggests that the more educated members o f a household are more likely to take paid farmjobs, although marriage tilts this choice towards marriedmen and away from married women. However, as the individual level o f schooling increases, heishe becomes less likely to work inpaid farm employment and more likely to work innon-farm jobs. Remittances ingeneral tend to reduce participationinpaid-farmjobs but have no significant effect on non-farm employment. 5.9 Foreign remittances (which are typically large relative to local wages) tend to reduce participation innon-farm employment. This finding i s consistent with Adams (1998) report on the use of foreign remittances by rural households. H e finds that foreign remittances tend to be investedin physical assets, suggesting that households that receive foreign remittances can increase labor productivity in own-farm or family enterprises. Trends InWages 5.10 Real wage rates o fbothregular and casual agricultural workers increased by an average o f 1.O percent per year from 1984 to 1994, indicating long term tightening o f the labor market. Likewise, real wage rates of construction laborers increased by 1.1percent over the same period. Most o f the increases inreal wage rates took place inthe 1980s, however. The trends inthe 1990s, however, indicate little gains inreal wages. From 1991-97, real wage rates o f casual agricultural laborers grew by an average o f 0.8 percent per year, while real wage rates o fregular agriculturalworkers fell by 2.4 percent per year. Likewise, real wage rates o f construction laborers grew by only 0.4 percent per year from 1991-97 and for the 1991-2002 as a whole, fell by 0.4 percent per year. 5.11 The decline inreal wages inthe agricultural sector inthe 1990s coincides with a period o f slower but still positive per capita agricultural growth. As i s discussed inchapter 7, the structure o f land distribution and total agricultural earnings may partly explain why agricultural growth didnot lead to increased or at least approximately constant real wages. The smaller rate o f decline o f real wages for constructionlaborers suggeststhat lack o f labor mobility between sectors and betweenrural and urban regions hinder integrationo f non-agricultural and agricultural labor markets (table 5.5 and figure 5.1). Table 5.5 Growth inReal Wage Rates inPakistan, 1984-2002 Agricultural Construction Regular Casual Carpenter Laborer 1984-91 3.27 1.46 0.62 1.64 1991-97 -2.39 0.84 -1.70 0.37 1984-97 0.95 0.99 -0.34 1.12 1991-2002 n.a. n.a. -1.49 -0.44 Note: Real wages are calculated usingthe national Consumer Price Index. Sources: Government o f Pakistan (various issues). Monthly Statistical Bulletin. Federal Bureau o f Statistics. Pakistan Economic Survey (2002-03). 37 Figure5.1 RealWages inPakistan: 1984-2002 EmploymentContractsInAgriculturalAndNon-AgriculturalSectors 2 :::; - 2 110 - N 2 9 0 - 0 0 C5 70- 2 50 - 5.12 According to the 1990 Agricultural Census, family workers accounted for 72 percent o f the agricultural labor force; tenant farmers accounted for another 25 percent. The shares o f permanent and casual hired workers were only 2 and 0.8 percent, respectively. Casual labor i s usually employed only inperiods ofpeak labor demand for seasonal activities such as transplanting (paddy) andhar~esting.~~ Moreover, the use of casual labor has declined across farm size as well as different tenurial categories. 5.13 Inthe ruralnon-farm sector, contractual arrangements take many forms: permanent, fixedterm, casual, piece rate, unpaid family helpers and self-employed. Wage workers are the dominant type o f laborers (65 percent of rural non-farm employment inthe 1999-2000 Labor Force Survey); self- employed and unpaid family labor account for only 33 percent o f rural non-farm employment. 5.14 Most o f the employment i s inthe informal sector (68 percent) and involves low-skilled jobs. The technology usedi s generally highly labor-intensive (Sayeed and Ali, 1998). Only 7 percent o f these informal sector employees in 1999-2000 were female. 5.15 The other thirdofthe non-farm labor force is employedbythe formal sector (registered firms), under either a permanent or fixed-term contract basis43,All labor laws are applicable to permanent labor. Contract workers, on the other hand, are recruitedthrough a contractor for a specific time period, and are usuallypaidper amount o f time worked (monthly, weekly or daily) by the contractor. The contractor, inturnhas entered into an informal arrangement with the employers for provision o f labor 42 Fafchamps and Quisumbing(1989) reported that 91 percent o f kharif(monsoon season) farmers and 89 percent o f rabi (winter season) farmers inthe districts sampled did not use any hired labor. Hired labor, employed mainly at harvest time, accounts for 21.5 percent and 23.6 percent o f total labor for kharif and rabi, respectively. 43 Sayeed and Ali (1998) found that 25 percent o f the total employees were permanent workers inthe manufacturing sector. A large majority (68 percent was concentrated inthe large scale manufacturing. The proportion o f contractual workers in large scale manufacturing as percent o f total manufacturing employment was 32 percent. This indicates that informal sector manly consists o f contractual and casual labor. 38 for a pre-determined lump sum.44These workers are not eligible for any o f the employment benefits that permanent workers have. Labor Mobility andMigration 5.16 Considerable rural-urban migration has taken place inPakistan inthe past several decades, in spite o fheavy transactions costs involved that limit the capability o fpoor households to migrate. According to HIES (1998-99), 30 percent o f current urban residents are migrants. Among them, 41 percent moved from other urban areas and 59 percent moved from rural areas. Migrant households are also found inrural areas, though to a lesser extent. Nearly 14percent o f rural households are migrants. Rural to rural migration (8 1percent o frural migrants) far outnumber urbanto rural migrants (19 percent o f migrants). 5.17 The available literature on migration indicates that age and education are important determinants o f migrati~n.~'45 percent o frural to urbanmigrant heads o f households migrated inorder to find jobs and 9 percent for the purpose o f business. The majority o f these migrant workers are employed as low- or unshlled wage workers. Age, education, and employment status are major determining factors o f rural-urbanmigration (Akram, 2000). Inparticular, unpaid family helpers are more likely to move from one rural area to another rural area, whereas, employers and employees move towards urban centers. 5.18 InruralareasofPakistan, domestic andforeignremittancesareimportant source ofincome, accounting for 9 percent o f total rural household incomes (HIES, 2001-02 data). Foreign remittances tend to be used for the accumulation o f land, and domestic remittances have a positive and significant effect on the accumulationo f agricultural capital (Adams, 1996). 5.19 Since migration incurs some cost, poorer households may not be able to afford to migrate, even incase ofextreme distress. They howeverhavethe opportunity to reallocatetheir labor towards other sources o f income, such as, non-farm activities, particularly ifthey acquire education. According to Fafchamps and Quisumbing (1998), one additional year of schooling for an adult males raises household incomes by 4.5 percent. One-fifth o fthis additional income was achieved by reallocating labor away from farming and toward non-farm work. Kurosaki and Khan (2003) found similar results for NWFP, Palustan. Wages and productivity inthe non-farm sector rise at an increasing rate with education, whereas non-agricultural workers were found responsive only to primary education. Gender Issues 5.20 Female labor force participation inPakistan for activities outside the home i s the lowest in South Asia: 11percent in 1999-00 (Table 5.6). Participation rates vary substantially by province from only 3 percent inBalochistanto 15 percent inPunjab. Including unpaid family labor, however, participation rates increase to 34 percent nationally, and are much more uniform across province, ranging from 3 1percent inBalochistan to 41 percent inNWFP. Female labor force participationrates are generally lower for younger women, because o f householdresponsibilities and cultural and family norms (Arifand Sheikh, 2002).Reported unemployment rates are very highinNWFP and Balochistan (3 1and 44 percent, respectively). 44 See SayeedandAli (1998) and FES (1993). 45 See Adams (1996) andQureshiandArif (1999). 39 Table 5.6 FemaleLabor ForceParticipationandUnemploymentin Pakistan, 1999-2000 Laborforce Improvedparticipation Unemployment Underemployment participation rate rate* Pakistan 10.7 34.3 14.0 6.9 Punjab 13.2 32.7 10.9 8.5 Sindh 6.0 34.4 11.0 2.6 NWFP 8.4 40.9 31.2 6.5 Balochistan 3.O 3 1.4 44.2 3.1 Source; Labor Force Survey (1999-00) Notes: *According to the old methodology persons 10 years o f age and above reporting housekeeping and other related activities are considered out of labor force. However, as per improved methodology, they are identifiedas employed ifthey have spent time on one o ffourteen specified home-based agricultural and non-agricultural activities. 5.21 InPakistan, women's decisions to participate inthe labor force are greatly influencedbythe strong social and cultural values adopted by the household. Economic factors play a significant, though smaller role, Numerous studies have shown that women's decisions to participate inthe labor market are significantly determined by family structure, relationship with the head o f the households, family size, educational attainment, household income and the market wage rate.46Chaudhry and Khan (1987) observed a negative association o f income class and presence o fmale earner, and positive effect o f educational attainment on the probability o f participating in income generating activities. Sultana, Nazli, and Malik (1994) get similar results when they divide total time into market production leisure and home production. These studies stress the need o f expandingthe opportunities and access to educational and training institutionsinaddition to opening up more vacancies particularly for women. 5.22 According to the 1999-2000 Labor Force Survey, 87 percent o f working females were engaged inagriculture. Major agriculturalactivities includelandpreparation, seedpreparation, collecting farm yard manure, weeding and harvesting, as well as cleaning, drying, and storage o f grains. Rural women also usually take care of household livestock -- collectingfodder, cleaning sheds and processing animal product^.^'. 5.23 Limited access to health, education, vocational training and credit facilities constrain women's opportunities for non-agriculturalwage labor and especially white collarjobs. Community, social and personal services (6 percent o f total employment) and manufacturing (5 percent o f total employment) accounted for most o f the remaining employment. Out o f total working women, only 1percent were employed as professionals and 0.27 percent worked inadministrative and managerialpositions. This lack o f opportunity for non-agricultural employment, together with the very low labor force participation rate, imply a considerable loss o f productivity for the Pakistan economy. 46 See Cox-Edwards (2004), Alderman and Chishti (1989); Kozal and Alderman (1990); Hamid (1991); Kazi and Raza (1991); Kazi and Bilquees (1992); Sathar and Lloyds (1993); and Sathar and Desai (1996). 47 See Khan and Bilquees (1976); Ahmad, Asghar and Khan (1993); and Sarwar and Saleem (1993). 40 ChildLabor 5.24 Child labor i s widespread inrural Pakistan.48According to the 1999-2000 Labor Force Survey, 10.9 percent o f children age 10-14 participatedinthe labor force. Labor force participation was much higher for boys (18.3 percent) than for girls (2.8 percent). Inrural areas, 21.7 percent ofboys age 10-14 were involved inthe labor force, compared with only 2.7 percent o f girls (figure 5.2). Figure5.2 ChildLabor in Pakistan, 1999-2000 25 .= 20 '0 2 + f0 15 E0al 10 Boys Bg 5 0 Total Rural Urban I Note: Data are for childrenages 10-14. Source: Labor Force Survey, 1999-2000. 5.25 Data from the 1996 Federal Bureau o f StatisticsiILO-IPEC nation-wide survey on child labor present a similar picture. An estimated 8.3 percent o f the 40 million children 5-14 years old were working. More than 500 thousand o f these children were 5 to 9 years old (2.7 percent of this age group). Inrural areas, 10.3 percent o f children were working, including 3.4 percent o f children from 5- 9 years old. Most children that work inPakistan do not attend school: only 28 percent o f the working children were enrolled in schools (equal to about 7 percent o f the total rural population age 5-14). 5-26 Child labor force participation rates varied substantially across province, however (figure 5.3). Participationrates for rural boys 5-14 ranged from 1.2percent inBalochistan to 23.0 percent inNWFP. InruralNWFP, 9.4 percent ofboys aged 5-9 worked duringthe week previous to the survey; compared to a national rural average o f 3.7 percent o f boys. 5.27 Among rural children across Pakistan, 74 percent of those who participated inthe labor force worked infamily enterprises. Many farm households prefer to use own family labor, even at the cost o f their children's education (Box 5.1). 48 Grootaert and Kanbur (1995) identify two extremes to define child labor. One inwhich all non-educational, non-leisure time of children i s counted as child labor, and the other which takes into account only the fill-time employment ineconomic activities of the children. 41 Box 5.1: Child Labor in RuralPakistan Contrary to conventional wisdom, child labor is not confined only to very poor households. Using household survey data for Pakistan and Ghana, Bhalotra and Heady (2001), show that the children o f landowners are less likely to be in school than the children o f poor-landless households. They compared the landless with three different categories o f farmers: marginal (with less than 0.4 hectares), small (0.4 to 1.2 hectares) and large (above 1.2 hectares). They found a positive and significant association between size of land andthe likelihood o f children working. This "paradox" arises because farm owners may feel it difficult to supervise the hired labor, especially inapculture, so they prefer to employ their own children, even at the cost of the children's education. Yaqub (2002) confirms these findings by describing the tale o f two farmers. "Ghulam Ahmed is a wealthy farmer with 2.8 hectares inKarimo Habib Hajivillage inthe Charsadda district, some 60 kilometers from Peshawar inNorth West Frontier Province (NWFP). H e grows tobacco, wheat, maize and sugarcane. Tobacco harvesting, though, is hard work and his children are vital inhelping out. H i s elder sons stay at home and help him. H e says that it is always good to have family members work o n the land. Bydoing so one cankeep away from the hassle o fhiringworkers each season, spending extra money and supervising them. Incontrast, Ajab Khan from nearby Mandani village, i s a landless farmlaborer and a father o ffour. Beinglandless, he can forgo his children's labor infavor o f school. H e says that education i s the only wealth that he can give to his children for better future" Sources: Yaqub, Nadeem (2002). "Child Labor - Not Only a Problem of Poverty", Development and Cooperation (No. 1,JanuaryiFebruary 2002, p. 28). Bhalotra, Sonia and Christopher Heady (2001). "Child Farm Labor: The Wealth Paradox". Social Protection Discussion Paper Series. No. 125. September Figure 5.3 ChildLabor Force* ParticipationRates by Province, Pakistan 1999-2000 20 2 15 h 20 8 L 10 v 5 0 I * Children o age 5-14 years. f Source: FederalBureau of Statistics / ILO-IPEC 1996 C h l d Labor Survey, 42 BondedLabor 5.28 Bonded labor is labor that i s tied to an employer in an exploitative relationship under which the worker i s required to provide work at very low wages or must work for the employer inorder to pay back debts at very highinterest rates. Bonded labor persists inPakistan because o f poverty, lack o f education, and lack o f access to alternative sources o f credit o n the part o f the bonded laborers, as well as the absence ofpolitical will o f local govemment officials inthe enforcement of existing legislation banning the practice (See box 5.2). Box 5.2 BondedLabor in Pakistan Bonded labor exists in various sectors and industries; for example, agriculture, carpet weaving, brick- kilns, stone quarrying, hotels and restaurants and construction. Debtbondage is the worst form ofbonded labor, in which workers are bound to employers by debts owed to them as well as through Because o f , the lack o f assets, extremely low incomes and lack o f funds the poor who are generally landless tenants or laborers dependent on the landlord or employer for credit. Due to low repayment capacity, many o f these households continue to suffer from mounting debt and continuing bondage inthe form o f free labor for the landlord or employer. Labor bondage continues over generations. When adults grow old or are disabled, their children become liable for the debt. Similarly, when husbands die, the wives must join other families to repay these debts. Inits worst form, the bonded laborers are not allowed to work at other places without the permission o f the landlord/employer. In some cases, women and children must remain resident on the farm as `collateral' for good behavior. Religion, ethnicity and caste play major roles in determining the burden o f bonded labor. Non-Muslims belonging to lower caste are the worst o f f followed by Muslimsbelonging to lower castekribe. According to a report carried out by Australian consultants for the Government o f Sindh and the Asian Development Bank under the Sindh Rural Development Project: "There are some 1.7 million landless agricultural workers (haris) and sharecroppers in five districts o f Sindh Province (Thatta, Dadu, Badin, Mirpurkhas and Umerkot). Most o f these people are in debt bondage. While bonded labor exists throughout Sindh, the majority o f those bonded in the north belong to the Muslim majority, while most of the bonded agricultural laborers in southern Sindh belong to dalit (untouchable) and to tribal communities who have migrated from the drought-prone area o f Tharparkar desert. Poverty and starvation have forced these communities to accept the landlords' cash advances, and to be available for work from dawn to dusk. Bonded laborers may be detained or guarded to stop them escaping and in these situations o f total ownership rape o f women is not uncommon" [Agrodev (2000)]. Many o f the bonded laborers work for no wages. The cost o f the food and clothing that they are provided i s added to their accumulated debt along with interest. Thus their debt, infact, increases on a daily basis. Most o f the indebted poor are forced to provide begar (unpaid labor services). Trafficking in bonded laborers is a common practice. These bonded laborers are sold by one landlord to another, usually for a price higher than the debt they had with their previous landlord thereby increasingthe overall debt. Escaping from bondage has serious reprisals, including threats and kidnapping. Despite the Bondea Labor System (Abolition) Act, 1992 and the serious penalties it prescribes the political and financial strength o f the landlords, especially in Sindh allows them to continue using bonded laborers with impunity. Sources:Nazli (2003) andAgrodev (2000). 43 5.29 Data on bonded labor i s scarce. Extreme cases o f exploitation o f laborers are reportedby humanitarian and non-government organizations. Statistical evidence i s difficult to obtain because o f the lack o f accurate data on household debts and terms o f repayment and employment innational surveys. Moreover, there i s no standard operational definition o f what constitutes bonded labor. Using data from the 1990 Agricultural Census, Ercelawn and Nauman (2000) calculated that 722 thousand share cropper households inPunjab and Sindhwere inbondage, inthe sense that they owed provided unpaid labor (begar) to their landlords. Of these households, 197 thousand households also owed debts to their landlords and were thus counted as being under debt bondage (equivalent to 17 percent o f the 1.13 million landless tenant farmers inthe two provinces and 1.5 percent o f the 13.27 million rural hou~eholds~~).Debt bondage, so defined, was concentrated in Southem Punjab andnorthem Sindh, with six districts of Hyderabaddivision (Sindh) accounting for around 55,000 debt bonded sharecropper households. 5.30 Estimates o fbonded labor inPunjab province usingdata from the Agricultural Census 2000 are similar to those for 1990 (Ercelawn and Sohnia Ali, 2003). About 7 millionpersons had been requiredto undertake some form o f forced labor, and 50 thousand sharecroppers were indebt bondage (comparedto anestimated 58 thousandinPunjab in1990), withMultan, D.G.Khan and Bahawalpur divisions in southem Punjab accounting for half o f sharecroppers inbondage. Thus, sharecroppers indebt bondage represented about 9 percent o f total landless tenant farmers inPunjab. 5.3 1 These estimates o f the number o f bonded laborer households likely overstate the extent o f exploitative labor and credit arrangements since some o f the debts incurredmay simplyreflect noma1 extension o f credit by landlords to sharecroppers at approximate market interest rates. Further work i s neededto quantify the extent o f the problem. 5.32 Bonded labor i s not confined only to agriculture. Ercelawn and Nauman (2000) estimated that 1million brick kilnworkers inPunjabwere also under debt bondage. A 2002 survey o f 300 landless laborers (haris) in Sindh and 300 brick kilnworkers inPunjab conducted by the FederalBureau o f Statistics found that 73 percent o f the brick kilnworker households are indebted, with 88 percent o f these getting loans from the employers. Moreover, the majority o f the indebtedhouseholds (nearly 83 percent) have inheritedtheir debt; 11percent o f the indebtedhouseholds pay both incash and with labor and 10 percent o f the indebted households repay with labor only. One-third o f brick kilnworkers reportedemployer threats and violence. Workers with debts to the employer are often required prohibited from taking other jobs until the debt was repaid." GovernmentLaborPolicies 5.33 Pakistan's labor legislation i s based on an elaborate institutional framework inherited from the British.This legislationrecognized the worker's rightto form trade unions, butwith only with highly restricted rights to strike. According to the constitution, labor i s a `concurrent subject', which means that it is the responsibility o f both the federal and the provincial governments. Labor legislation i s usually enacted at the federal level, but the responsibility for enforcing it belongs to the provinces. 5.34 The legislative framework for worker's protection exists under four categories: the Factories Act 1934; the Shops and Establishment Ordinance 1968; the West PakistanIndustrial and Commercial 49 The figure for the number o f landless tenant households i s calculated from HIES 2001-02 data. 50 Similar results were found ina rapid assessment survey by an NGO (PILER)of over 100brick kilns around the urbanandperi-urban areas o fthe districts o fHyderabad, Multan, Lahore, Rawalpindi, Peshawar andHaripur duringOctober to December 2002. 44 Employment (Standing Orders) Ordinance 1968; and the IndustrialRelations Ordinance 1969. These laws are, however, not applicable on the agricultural sector51. 5.35 There are, however, various tenancy laws that are applicable only to the `title tenants'52,In reality, occupancy rightsdo not exist and all the tenants are `tenants at will' without any occupancy rights-- they canbe evicted by the landlordat anytime. Even among the titledtenants, the land occupancy rights essentially belong to the male head o f household. 5.36 Pakistanhas a federal minimumwage for unskilled workers. It applies only to industrial and commercial establishments employing 50 or more workers and i s determined by the National Minimum Wage Commission and requires approval by the Parliament. The minimumwage for skilled workers, however, i s set by the provinces. Minimumwages practically apply only to workers inthe formal sector and do not affect agricultural workers. Labor law, applicable nationally, but only to firms with 10 or more workers, provides for a maximum work week o f 54 hours for seasonal factories and a maximum work week o f 48 hours for other industrial workers, as well as for rest periods duringthe workday. Some non-wage benefits include official government holidays, overtime pay, annual and sick leave, health and safety standards inthe workplace, health care, workers' children education, social security, employees old age benefits, and a workers welfare fund. 5.37 Labor regulations (that mainly apply only to urban areas) do not directly alter employment conditions inrural areas, but they do have a significant effect inthe dynamics o f employment creation inurbanareas andthus influence incentives for rural-urbanmigration andrurallabor markets. Employment growth inurban areas has been slow, inpart because o f an atmosphere of mutual hostility and distrustthat typify industrialrelations. Inparticular, the unusually highdegree o f discretion over conditions o f employment o fregular workers in specific establishments heldby provincial authorities i s not conducive to cooperation between employers and workers. Replacing the discretion o f authorities on the specific nature o f labor contracts by a few rules on the most common aspects o f these contracts (minimumage, minimumwage, overtimepay) andthe establishment o fprocedures to help deal with possible disputes arising from voluntarily agreed contracts regarding more detailed aspects o f these contracts would help buildtrust inthe industrial relation process, and encourage employers to create jobs and invest inhuman capital. 5.38 Wage labor inagriculture sector, however, i s not covered by any legal framework. Agricultural wage labor i s often recruitedindirectly through contractors and denied even basic labor rightsrelating to working hours, minimumwages and working conditions. 5.39 Although Article 11o f the Constitution o f Pakistanprohibits slavery and forced labor, and there exists a law, "the Bonded Labor Abolition Act 1992" that prohibits the use o f bonded labor, the practicenot only continues but i s even defended by landlords.53 The Government o f Pakistan adopted a National Policy and Plan o f Action for the Abolition o f Bonded Labor and Rehabilitationo f Freed Bonded Laborers in2001. The ILO Headquarters in Geneva has also developed a programme on 51 Agriculture does not come under the classification o f industrial, commercial or services sector. therefore none of these laws are applied to this sector. 52 Those with legally recognized occupancy rights. 53 The situation o fbonded laborers inSindh has deteriorated further, especially when the HighCourt o f Sindh dismissed 94 petitions for the release o fbonded laborers o n 9 January 2002. the court declares that these were disputes between landlords and haris over debts and should be settled under the Sindh TenancyAct 1950. However, the Court made no reference to the Bonded Labor Abolition Act 1992 which should take precedence in this matter andwould require the loans to be nullified andthe bondedlaborers to be freed. 45 bonded labor for Pakistan under their Special Action Programme to Combat ForcedLabor. This programme i s to be implementedto support the National Policy and Plan o fAction for Bonded Labor, The preparatory phase o f the ILOprogramme was initiatedinJuly 2002. VI. RURALCREDITMARKETS 6.1 The extent to which markets for land, water and labor can allocate these factors to their most productive uses i s limited by the immobility or hightransport costs o f the factors, themselves. In contrast, financial flows through credit markets face extremely low physical transactions costs, (though once invested, the capital goods themselves, such as tractors and storage facilities, may be less mobile). InruralPakistanhowever, poorhouseholds are oftencredit-constrained. Informal markets, accessto credit i s generally limitedto landowners, since land i s the main form o f acceptable collateral for loans. Access to credit ininformal markets, particularly supplier's and consumer's credit i s more widespread, yet approximately 40 percent o f rural households are credit constrained. Structure of RuralCreditMarkets 6.2 Formal sector banking inPakistan i s largely urbanbased. Advances to the rural sector accounted for only 3.5 per cent o f advances to the urban sector in2001-02. Similarly, rural deposits for the one year (1999-00) for which comparable numbers are available are about 28 percent o f urban deposits (table 6.1). The data imply a net outflow o f financial funds from the rural formal financial sector o f Rs (2001-02) 128 billion in 1999-2000, though investments inrural non-farm activities may not be consistently shown inthe rural advances figures. Table 6.1 FormalSector Banking (bn2000-01 Rupees) ~~ Rural Urban Year Deposits Advances Deposits Advances 1996-97 24.18 482.86 1997-98 38.37 566.89 1998-99 46.52 582.66 1999-00 169.87 41-92 614.22 645.93 2000-01 44.79 628.72 2001-02 50.12 1426.13 Source: Economic Surveys o f Pakistan(2002-03); values deflated with GDP deflator. 6.3 The Zarai TaraqiatiBank Ltd.(ZTBL), (formerly the Agriculture Development Bank of Pakistans4),commercial banks, the Federal Banks for Cooperatives (FBC's) and other financial institutions are the major sources o f formal credit inPakistan. Informalcredit, (including loans incash or kind) derives from friends and relatives, commission agents, input dealers, professional moneylenders and landlords. Survey data from 1995-96 (the most recent year for which such data are available) indicate that informal sector credit i s nearly three times as large as formal credit inrural Palustan, accounting for about 72 percent of total rural credit (table 6.2). Assuming that formal credit still accounted for 28 percent o f total rural credit, lending by the ADBP/ZTBL (equal to nearly 60 percent o f formal rural credit disbursed), accounted for only about 15 percent o f total rural credit in 2001-02. Commercial banks accounted for only about 9 percent o frural credit in2001-02; but has increased rapidly since that time and exceeded ADBP/ZTBL lending inthe third quarter o f 2003-04. Cooperatives and private banks, together, accounted for less than 3 percent o f total rural credit in2001 02. 54 The Agricultural DevelopmentBank o f Pakistan (ADBP) was renamed the Zarai Taraqiati Bank in2002. 47 Table 6.2 Sources of CreditinPakistan,2001-02 _ _ _ _ ~ Amount Type of credit (bn Rs) Shareformal Share rural Urban(formal) 1,492.4 96.6% n.a. Rural (formal) 52.4 3.4% 28.0% ADBP/ZTBL 29.1 1.9% 15.5% Domestic Private Banks 0.6 .., 0.3% Cooperatives 5.3 0.3% 2.8% Commercial Banks 17.5 1.1% 9.3% Rural (informal)a 134.9 n.a 72.0% Total (formal) 1,544.8 100.0% n.a. Total (rural) 187.3 n.a. 100.0% a Calculatedusing the share of informal credit intotal rural credit (72%) in 1995-96. n.a. denotes not available. ... denotes less than 0.1%. Source: Economic Survey o f Pakistan2002-03 and Rural Credit Survey 1996. 6.4 Thus, the rural credit market inPakistan continues to be dominated by informal sources o f lending. Borrowingfrom informal sources i s comparatively easy interms o f access, procedures and collateral requirements. This translates into low transaction costs. Furthermore, unlike institutional credit, for which credit i s available i s for only limitedpurposes, non-institutional credit i s available for a wide variety o f purposes including consumption, social ceremonies and other non- productive purposes Historical Trends 6.5 Prior to 1947, institutional credit inrural areas was mainly provided as taccavi (distress) loans by the government and as co-operative credit by the co-operative societies. Inorder to reduce the dependence on informal sources, two specialized institutions i.e. Agricultural Development Finance Corporation and Agricultural Bank were established inthe 1950s. These were later merged to form the ADBP in 1961. 6.6 Withthe increasinguse o f fertilizer, pesticides, improved seeds, and mechanization, credit requirements o f the agricultural sector have increased considerably over time. From 1972 to 2001, institutional agricultural credit grew at an annual rate o f 15 percent. Formal ADBP loans, the major source o f formal credit, grew at over 17 per cent per annum duringthis period. Credit from commercial banks and cooperatives grew at 11and 15 percent per year, respectively. 6.7 Growth of formal sector credit has varied considerable over the last three decades, however, as a result o f major policy shifts. Formal agricultural lending quadrupled inreal termss5between 1978-79 (Rs 12.3 billion) and 1986-87 (Rs 51.7 billion), due mainly to a large expansion ingovemment- supported, subsidized productionlendingby the commercial banks. At its height in 1986-87, subsidized credit through commercialbanks was the major source o f rural formal sector lending (figure 6.1). Commercial bank loans declined dramatically thereafter, after the discontinuation o f this subsidized credit scheme, however, falling inreal terms from 23.8 to 8.6 (2000-01) billionrupees between 1986-87 and 1988-89. 55 Billions o f 2000-01 rupees. . . 48 6.8 Total formal sector then stagnated untilreforms inADBP policies inthe late 1990s. The One Window Operation initiated in 1997 streamlined the application, processing and disbursal o f loans, particularly for small farmers. Access to ADBP/ZTBL loans was also increased through a reduction in the minimumrequirements o f land from 12.5 acres to 5.0 acres in2000. Territorial restrictions on commercial bank lending operations were removed, as well. 6.9 As a result o f these policy reforms, and in spite o f an end to subsidized credit to the ADBP/ZTBL and Federal Banks for Cooperatives from the State Bank o f Pakistan, rural formal sector credit increased by 107 percent inreal terms from 1996-97 to 2001-02. The agricultural credit to agricultural GDP (at current factor cost) rose from 4.9 percent duringthe 1990s to 7 percent in2001-02. (See Box 6.1. and figure 6.1) Box 6.1 RecentReformsinFormal SectorAgriculturalLendinginPakistan ~~ Zarai Taraqiati BankLtd. (formerlyAgriculturalDeveloDmentBankof Pakistan) One Window Operation. The ADBP initiated the One Window Operation in 1997, simplifylng proceduresfor loans, inorder to make credit facilities more easily accessibleto small farmers. Under this operation, undertaken in collaboration with the Provincial Governments, Revenue Officials and Postal Authorities, focal point officials provide clients with Agnculture Pass Books on the spot and note land recordsinthe pass books. The ADBP/ZTBL focalpoint officials are responsible for sanctioning loans (up to amaximumo f Rs 50,000), which are disbursedthe following day from the concernedbankbranch. Micro-credit. The ADBP initiated a micro credit scheme inJuly 2000, which targeteddrought-affected farmers. This scheme encouragedthe rural poor to engage inprofitable commercial activities. Under this scheme potential borrowers could obtain a maximum o f Rs. 25,000 against a personal guarantee and a viable security. The range of activities for which loans could be givenwas expanded to include 136 loan able items andfundingwas extendedfor 18months at 16percentper annum. Landrequirements. Inorderto enhance the access of small farmers to formal credit, ADBP reducedthe requirementofminimum12.5 acres of landto 5 acres of land.Moreover, inview of small land holdings inNWFP the loan limit of Rs. 50,000 for productionloanwas increasedto Rs. 100,000 againstpersonal security. ADBP also simplifiedthe loan appraisalas well as the application forms. StateBankofPakistan(SBP) Credit for agro-processing and marketing. The agricultural credit scheme has been extended to encompass the complete value chain of agnculture ranging from inputs, production, storage and marketingto transport, processingand distribution. End to subsidized credit for agriculturallending. InOctober 2000, the provision of subsidized credit facilities to ADBP/ZTBL and FBC was ended. The minimumaverage rate on T-bills has beenset as the pricing criteria for all new credit lines to the ADBP/ZTBL and other financial institutions have been encouragedto lendtheir funds to ADBP/ZTBL. Relaxation of territorial limits to commercial banks. Prior to the reforms, SBP assigned territorial boundaries to commercial banks for lending purposes. Repealing this rule was undertaken to increase competition amongbanksand give farmers awider choice interms o f selecting the potential lender. Revolvingcredit facility. Revolving credit was introduced to allow farmers to obtain fresh credit even if they have not beenable to repay outstanding loans. Formerly, banks could not sanction revolving credit facility for more than three years, particularly not to those farmers who had been unable to repay prior debts. Bankswould also not demandfresh documents at eachrenewal. Source: Malik, Sohail J. 2003b. Rural Credit Markets inPakistan: Institutions and Constraints. Backgroundpaper 49 Figure 6.1 Growthof InstitutionalAgriculturalCredit (mn 2000-01 Rps) 60,000 50,000 40,000 30,000 20,000 10,000 0 - - ADBP CooperativedFBC --+s-. Commercial Banks Total (Rs. Million) Source: Malik (1996) and Economic Surveys (various issues) 6.10 Nearly 80 percent o f total lendingby the ADBP/ZTBL in2002-03 was for production loans (for fertilizer, pesticides, seeds and other productionneeds). Average size o fthe 376 thousand product loans was Rs 61 thousand. The average development loans was nearly twice as large (Rs 120 thousand), with the 10,200 loans for tractors averaging Rs 275 thousand per loan, and accounting for 9.6 percent o f total lending (table 6.3). Table 6.3 Zarai Taraqiati BankLtd. (ADBPa)Lending,2002-03 Average Number ofValue Loans Loan Size of Loans (mn Rs) Value Share ('000 Rs) Total Disbursement 427471 29,270 100.0% 68 Production loans 376,307 23,111 79.0% 61 Fertilizer 170,215 11,121 38.0% 65 Pesticides 59,460 4,596 15.7% 77 Seeds 139,174 6,652 22.7% 48 Other production loans 7,458 742 2.5% 99 Development loans 51,164 6,160 21.O% 120 Tractors 10,203 2,802 9.6% 275 Tube wells 5,602 970 3.3% 173 Dairy farming 26,527 1,162 4.0% 44 Other development loans 8,832 1,225 4.2% 139 Micro-credit 1,965 44 0.1% 22 a ADBP was renamed as Zarai Taraqiati Bank Ltd. in 2002. Source; Zarai Taraqiati Bank Ltd., viww.adb.or& (March 2004). 50 6.11 Institutional credit i s unequally distributed across provinces: in 1999-00, nearly 74 percent of the institutional credit was disbursed inPunjab, which accounts for only 59 percent o f the total rural population. The province o fNWFP with 16percent o f the rural populationreceivedonly 5 percent o f the institutional credit. Balochistan, which accounts for 6 percent o frural population and where agriculture i s the major source o f livelihood receivedonly 1.5 percent o f total institutional credit56. Sindh i s the only province where the share o f credit i s approximately equal to its share o f total rural population. On a per hectare basis, however, Sindh has the highest average disbursement (Rs 3,978 / hectare), followed by Punjab with Rs 2,890, NWFP (Rs 1458 / hectare) and Balochistan (Rs 937 / hectare). 6.12 The wide variation inlending across provinces i s mainly attributable to: (1) the financial position o f farmers inthe respective province and their ability to provide acceptable collateral, (2) variations inthe distribution o f land, (3) the track record or repayments, (4) traditional reliance on non-institutional sources, and (5) concentration o f agricultural activities [SBP (2002)l. Repayment of Loans 6.13 Loan recovery rates inthe formal banking system are low, though substantially higher than achieved inthe 1980s and 1990s. In2001-02, nearly 71 percent o f the current dues were reportedly recovered (table 6.5). Recoveryrates of commercial banks (86 percent) were higher than those of the ADBPiZTBL (68 percent). Recoveryrates ofpast due loans are substantially less (36 percent in2000- 01 and 28 percent in2001-02). 6.14 According to the State Bank o f Pakistan (SBP, 2002), net agriculture credit (disbursements less recovery) declinedby Rs 1.0 billion during 2001-02, but the outstanding amount o f credit increased from Rs 93.4 billion at end-June FYOl to Rs 101.5 billion inFY02,because o f accrued interest on past- due loans. With low recovery rates o f current and past-due loans, the outstanding stock o f agricultural debt will continue to grow, and will "ultimately restrict the banks' capacity to meet the increasing credit demand from the rural sector". 6.15 Given its low deposit base, ADBP/ZTBLi s increasingly dependent on subsidies from the State Bank o f Pakistan. The relatively better performance o f commercial banks i s due mainly to their lending beingbased on their own deposits and higher recovery rates on current and past due loans. Structure of InterestRates 6.16 Like other developing countries the informal credit market inPakistan also exhibits a large degree o fheterogeneity. The structure and operations differ from region to region. Furthermore, cultural and religious values prohibit the public admission o f charging an explicit interest rate. Much of the zero interest rate lending i s explainedby implicit mark-ups (for purchases o f goods and services), other tied transactions and co-insurance. These tied credit transactions account for a very large share o f total rural credit: data from the last national Rural Credit Survey (1985), suggest that 90 percent of rural informal credit carriedno explicit interest; the interest rate on other loans averaged 22 percent, and the overall interest rate (both zero-interest and loans with interest) was 2.2 percent (Malik, 1999). 6.17 Average interest rates inthe informal sector have varied between 7 and 11percent higher than those inthe formal sector (table 6.4), though peak interest rates ininformal credit markets can be extremely high. For example, Irfan et a1(1999) reported interest charges o f moneylenders inPunjab 56 The coverage o f commercial banks in Balochistan i s extremely low. 51 ranging from 48 to 120 percent per year." Such very highinterest rates are often tiedto the marketing o f crop. Table 6.4 Pakistan:InterestRatesinRuralCreditMarketsby Source Year Formal Credit lnterest Rate Informal Credit lnterest Rate Average 8.2 Average 15.3 ADBP 7.2 Money Lenders 18.5 Commercial banks 8.9 Land Owners 10.3 1985 Average 12.2 (5.7) Average 21.9 (2.2) ADBP 12.3 (9.6) Money Lenders 33.4 (315) Commercial banks 12.3 (2.6) Land Owners 22.8 (1.4) 1996 ADBP 14 Average 25 2004 ADBP/ZTBL 14 Micro-credit 16 Averages are for all sources, weighted by volume o f credit. Figures inparentheses are average o f interest-free and regular loans. Sources; 1973 Pakistan Rural Credit Survey; 1985 Pakistan Rural Credit Survey Malik (1999), Irfan et. al. (1999), Zarai Taraqiati Bank Ltd.,www.adb.org.pk (March 2004). 6.18 Higher interest rates inthe informal market do not necessarily reflect absence o f competition, however. Irfan et. al. (1999) found some evidence o f competition inthe informal credit market, as indicatedby the presence o f more than one lender ina village -- contraryto the earlier findings by Aleem (1990). The higher interest rates inthe informal sector are due to the opportunity, administrative and risk costs, as well as the additional cost o f generating funds. Borrowing, especially from formal sources, i s the main source of funds for the informal lenders. Nearly one-third o f the total funds utilized inthe informal credit transactions originate from formal credit sources, (70 percent o fthe funds for agriculturalprocessing units derive from institutional sources). On average, informal lenders paid about 19 percent annual interest rate on their funds borrowed from institutional sources (5 percent greater than the ADBP/ZTBL rate o f 14percent). Interestrates charged for inputsvaried by product (35 percent for pesticides and 8 percent for seeds), with an average interest rate o f 25 percent. FormalSector Credit andMechanization 6.19 The subsidy on formal sector credit has encouraged mechanizationinPakistan. From 1979-80 to 2002-03, the ADBP/ZTBL financed purchases o f 399 thousand tractors, an average o f 16,630 tractors per year. As shown infigure 6.2, the number o f tractors purchased with ADBP/ZTBL financing closely mirrored domestic productionof tractors from 1986-87 to 1997-98, pointing to the linkbetweenavailability ofADBP/ZTBL credit andtractor sales. 6.20 Since 1998-99, however, non-ADBP/ZTBL financing o f tractors has become nearly as important as a source o f credit as ADBP/ZTBL loans. Although the average annual number o f 57 Similarly, World Bank (2002) reported rural informa1 sector interest rates ranging from 80 to 150 percent. 52 ADBP/ZTBL tractor loans fell by 10percent from the 1986-87 to 1997-98periodto the 1998-99 to 2001-02 period, average annual productiono f tractors increasedby 67 percent, and domestic production o f tractors exceeded the number ADBPiZTBL financed tractor purchases by 85 percent. Figure6.2: ADBP/ZTBLTractorLoansandDomesticTractor Production, 1986-87-2002103 40 j 0 i 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 86- 87- 88- 89- 90- 91- 92- 93- 94- 95- 96- 97- 98- 99- 00- 01- 02- 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 Tractor production --t Tractors - ADBP loans 1 Farmer HouseholdAccess to Credit 6.21 According to PRHS 2001-02 data, nearly 80 percent o f cultivator households participate inthe credit market. Access to institutional credit i s highlyrestricted, however. Few farmers inthe sample fully cover their financial needs through formal credit, and those without land have almost no access to formal sector credit. 6.22 Overall, 11percent o f farmers had obtained formal sector loans, 14 percent o f land owners, but only 2 percent o fnon-landowners (table 6.5). Three-fourths o f formal sector credit goes to land- owners; owner cum tenants receive 24 percent o f formal sector credit. Less than 1percent of formal credit goes to tenants (figure 6.3). Interms o f operated farm size, large farms and medium farms receive two-thirds o f total formal sector credit.58 '*Numerous earlier studies have also found that small farmers and poor households rely predominantly on non- institutional sources for their credit needs, inspite o f policies and subsidies designed to enhance their access to institutional credit. See Punjab Economic ResearchInstitute (1986); Applied Economics Research Center (1986); Scott and Redding (1988); Malik (1989, 1990, 1992 and 1999); and Qureshi and Shah (1992). 53 Table 6.5 OutstandingLoans By Tenure Categories (Formal and Informal) 1Sample IHouseholds with Informal 1 Formal 1Both I Total 1 I I 1 1 1 1 1 I ~~ Sample 1590 100 1239 (78) 100 1072 67 I 4 3 2.7 124 8 Source; PRHS 2001-02 data; Jacoby andMansuri,(2004b). Figure6.3 FormalandInformalCreditby Tenure Categories,2001-02 Formal Credit Informal Credit (Distribution across tenure categories) (Distribution across tenure categories) Sharecropping Fixed rent Owners cum Fixed rent Sharecropping tenants tenants 22% Owners 44% Owners cum 75% tenants 27% I Source: PRHS 2001-02 data; Jacoby andMansuri, (2004b). 6.23 Informal sector credit i s much more widespread. 75 percent o f farmers had obtained informal sector loans, including 72 percent o f land owners and 84 percent o f non-landnon-landowners. Moreover, even landless farmers (sharecropping tenants and fixed rent tenants) have access to informal sector credit, and account for 29 percent of the value o finformal sector lending. The average value o f informal sector loans to land owners (Rs 24,400) i s only 17 percent greater than the average informal sector borrowingby tenants (Rs 20,900). Total average borrowing by land-owners (including both borrowers and non-borrowers) i s Rs 29,500, 65 percent greater than total average borrowingby landless farmers (Rs 17,900). See table 6.6. Table 6.6 Average Size of Loans to FarmHouseholds, 2002-03 Formal Sector Informal Sector All Loans All Loans* No. of mean No. of mean No. of mean No. of hhs (RS) hhs (RS) hhs (Rs) hhs mean(Rs) Owners 128 81,595 609 20,043 641 35,336 871 26,005 Owners cum tenants 30 100,677 196 37,925 204 51,243 252 41,483 Share tenants 6 12,800 305 19,790 308 19,847 365 16,747 Fixed rent tenants 3 37,500 86 24,678 86 25,986 102 21,910 Total 167 81,759 1,196 23,242 1,239 33,456 1,590 26,070 Farmerswith land 158 85,218 805 24,397 845 39,176 1,123 29,478 Landlessfarmers 9 21,033 391 20,865 394 21,187 467 17,875 "Includes bothborrowers and non-borrowers. All other data i s for borrowinghouseholdsonly. Source; PRHS 2001-02 data; Jacoby andMansuri, (2004b). 54 6.24 The average duration of informal sector loans i s only about one-fifth that o f formal sector loans (9 versus 47 months). About two-thirds o f informal loans (mainly from relatives and friends or a neighbor) have n o specific payment date. Overall, relatives are the main source o f informal credit, accounting for 3 6 percent o f the total number o f informal loans (and having among the highest average loan size across the different credit sources). Other personal relations, "friends and neighbors", are also an important informal credit source (15 percent). 6.25 Almost 80 percent of informal loans are incash. The remaining are in-kindloans coming mainly from traders, shopkeepers, and landlords of sharecropping tenants. Typically, these would take the form o ffertilizer or other inputsadvancedat the beginningo f an agricultural season. 6.26 All formal loans require collateral, with about athirdof loans usingagriculturalland for this purpose. Other agricultural equipment and animals are used as collateral in 14 and 8 percent o f the loans, respectively. Inmore than a third o f the cases, collateral i s not inthe form o f agricultural assets; residential or commercial property accounts for 10 per cent and personal guarantees from relatives friends and employers for 17 per cent. Thus, access to formal credit typically requires ownership o f visible assets, mainly land and farm equipment, or at least a strong personal connection with someone who probably owns such assets. By contrast, the vast majority o f informal loans (90 percent) require no collateral. Access to Credit 6.27 Although only 11.3 percent o f households had loans inthe formal sector; another 50.6 percent o f households reportedthat they didnot need credit or found credit to be too expensive. The remaining 38.1 percent o f cultivators were access rationed inthe formal market, inthe sense that they wanted formal sector credit at the terms offered, but were unable to obtain it. However, among the 11.3 percent of households that had formal sector loans, 23 percent (i.e. 2.6 percent o f the entire sample) faced credit limitson the size o fthe loan at the given interestrate. Thus, a total of40.7 percent o f households were either access or quantity rationed (table 6.7 and figure 6.4). 6.28 Ingeneral, accessto creditinthe informal sector was greater: 76 percent offarmers hadsome type o f loan (consumption, production or other loan) from informal sector sources, and 18.1 percent o f farmers were able to borrow but chose not to. Only the remaining 6.2 percent were access rationed in the informal market. Another 11.7 percent o f households faced bindingnon-zero limits on borrowing inthe informal sector, sothat atotal of 17.9percent ofruralhouseholds were access or quantity rationed inthe informal market (figure 6.5). Micro-credit Programs and Savings Mobilization 6.29 Inrecentyears, theGovernment ofPakistanhasplacedconsiderable emphasis ondeveloping micro-finance as part o f its poverty alleviation strategy. Micro-finance i s disbursed as small loans - through the ADBP/ZTBL, the Khushali Bank and the PakistanPoverty Alleviation Fund(PPAF). In 2003, micro-finance credit by these three institutions equaled Rs. 1,442 million (2.5 percent o f total institutional credit). Two-thirds o f this micro-credit was disbursed through the PPAF, and 31percent through the Khushali Bank; ADBP/ZTBL lending accounting for only 3 percent o f the total (table 6.8). Inall, 57 thousand micro-credit loans were advance, with an average loan size ofRs 25.3 and23.7 thousand per loan through the PPAF and Khushali Bank. The average loan size in2003 through ZTBL was considerably larger (Rs 41.6 thousand), though this i s considerably less than their average loan size in2002 (Rs 126.3 thousand for only 135 loans). 55 Table 6.7 Credit Rationing inRuralPakistan 2001 Access or Total %of Total Non- Access Quantity Sample Sample Borrowers Rationed Rationed 'ormalCredit iorrowers 180 11.3% 0 0 41 Not Quantity Rationed 139 8.7% 0 0 0 Quantity Rationed 41 2.6% 0 0 41 lon-borrowers 1410 88.7% 1410 606 606 Able to borrow 804 50.6% 804 0 0 Not able to borrow 606 38.1% 606 606 606 'otal 1590 100.0% 1410 606 647 'ercentage of Total 100.0% --- 88.7% 38.1% 40.7% iformal Credit lorrowers 1205 75.8% 0 0 186 Not Quantity Rationed 1019 64.1% 0 0 0 Quantity Rationed 186 11.7% 0 0 186 Jon-borrowers 385 24.2% 385 98 98 Able to borrow 287 18.1% 287 0 0 Not able to borrow 98 6.2% 98 98 98 -otal 1590 100.0% 385 98 284 'ercentage of Total 100.0% --- 24.2% 6.2% 17.9%1 iource; WB-PIDE 2001RuralHousehold Survey; Jacoby and Mansuri(2004a). Table 6.8 Major Micro-credit Programs inPakistan, 2003 PPAF* KhusaliBank ZTBL** Total Number of Loans (thousands) 37.7 18.7 1.I 57.5 Value of Loans (mn Rs) 951.4 443.0 47.7 1,442.0 Average Size (thousand Rs) 25.3 23.7 41.6 25.1 ***PakistanPoverty Alleviation Fund. Zarai Taraqiati Bank Ltd., formerly Agricultural Development Bank of Pakistan. 56 Households Households with with b outstandingloans whoface Households outstanding no quanitty restrictions with loans (139) outstanding (180) ' loans Households that are NOT (181) Households access rationed intheformal Farmers with no loans credit sector ad who do not face any quantity restrictions in the that are NOT sample access (804) rationed inthe 1590 formal credit sector (804) Households that have formal Households with Households restrictions (41) no outstanding b rationed that are access b Households that are access Specification I Specification 2 Specification 3 Figure6.5 CreditRationinginthe InformalMarket Households with Households Households with outstanding loans who face with outstanding loans no credit limits (1019) outstanding (1205) \ b Households that are Households that are not not current current borrowers but are Farmers borrowers but are not access rationed inthe not access rationed formal credit sector (287) Households that are informal loans but face Households with + access rationed quantity restrictions (186) no outstanding (98) b loans (335) Households that are 57 6.30 Most o f the PPAF loans were given for livestock and for enterprise development and trade and commerce, through a network o f development existingNGOs (Box 6.2). A recent evaluation o f the PPAF noted that its operation now covered 72 districts and a large percent of the loan recipients (44%) were women. Box 6.2 PakistanPovertyAlleviationFund(PPAF) The Government of Pakistan set upPakistan Poverty Alleviation Fund(PPAF) inFebruary 1997 in order to "enhance the access o f the poorer households and communities to socio-economic services". PPAF was set up as aprivate, non-profit, limitedcompany, with the aim o f reachingthe poor and disadvantaged communities inbothrural and urban areas throughNGOs and the Community Based Organizations (CBOs). PPAF funds for income generation activities and improved community physical infrastructure are disbursedthrough its three main units 1) Credit and Enterprise Development Unit; 2) Community Physical InfrastructureUnit; and 3) Human and Institutional Development Unit. Micro credit loans have increased from Rs. 35.6 million to Rs. 2,814 million between FY2000and FY2003. The bulk of the micro credit i s disbursedunderthe Credit and Enterprise Development Unit with disbursements inFY2003 o f Rs 1,314 million, 47 percent of the total PPAF disbursement.This amount was equivalent to 4.8 percent of the total institutional loans disbursed during that year. In 2003, the 119,196 borrowers receivedan average loan o f Rs. 8,816. Nearly 44 percent o f loans went to women. Most of the loans (38%) were disbursedfor livestock, followed by agriculture (32%), and enterprise development and commerce and trade (30%). According to the PPAFAnnual Report (2003), by the end o f FY2003, a total o f 218,702 individuals had availed o f Rs 2,8 14 million inPPAF financing since the inception of the program, in72 districts through 37 partner organizations. Gallup Pakistan(2003) evaluated the functioning of PPAF through a survey o f 1700 borrower and nonborrower households in 140 community organizations o f 17 districts inall four provinces o f Pakistan. This study found that borrowinghouseholds were better off than those that hadnot borrowed. On the average their incomes and consumptionhad increased and there was an improvement intheir personal and business assets, housing facilities and other amenitieshervices, and intheir social status (especially for women borrowers). Borrowers were not able to obtain as much credit as they wanted, however. Inthe survey sample, the meanvalue o f loans receivedby the borrower group was Rs.9,138, 53 percent o f their desired amount (Rs. 7,136). According to study, the meanreturn on investment for the loans was approximately 30% per annum. 6.3 1 No comprehensive study of the overall impact o f these micro-credit schemes has yet been done. Apart from questions related to the effectiveness o f targeting and coverage of the programs, there remains a major concern regarding sustainability o f these programs. By setting ambitious overall targets and usingthe local contacts of the NGOs, the programs have achieved an impressive growth to date (though the share inthe total institutional disbursements i s still quite small - less than 5 percent). However, rural credit programs inPakistan have historically had a poor record o f repayments. As long as large inputso f donor and other funds are available for on- lendingby these programs, and the possibility o f turning over loans exists, it will be difficult to assess the actual default rates on these loans. 58 6.32 Inadditionto these micro-credit schemes, rural development projects implemented through the Aga Khan Rural Support Programme (AKRSP) also include savings and credit programmes based on a system of collective savings and loans intermediated through village organizations (VOs) (Box 6.3). This system minimizes the transaction costs for borrowers. D u e to peer group pressure, the transparent nature o f the credit operations, and group guarantees for loans, repayment rates for loans are nearly 100percent. The employment and income impact, especially for the poor, has been quite substantial. Given the success o f the AKRSP, the National Rural Support Programme (NRSP) was started in 1991 and since 1992 has had direct support from the Government o f Pakistan. Box 6.3 Aga KhanRuralSupportProgramme(AKRSP) The Aga Khan Rural Support Programme (AKRSP) was initiated in 1982 for the purpose o f implementing a rural development programme in the Northern Areas o f Pakistan. The essential elements o f the AKRSP are the establishment o f effective Village Organizations (VOs). Each VO starts its operation with one grant-assisted productive physical infrastructure project. The women's VOs focus on various aspects of raising women's productivity and welfare. Examples o f the activities in which these organizations are engaged are: (i)introduction and dissemination o f improved varieties and cultivation techniques for crops; (ii) improved livestock husbandry practices; and (iii)new productivity-raising technology for apricot-drying tents, nut-cracking machines, and butter-chumers. AKRSP helps the village organizations to obtain access to institutional credit by acting as a loan guarantor through the promotion o f savings by members, which serve both as a form of loan collateral for credit and provide supportive marketing services for different products. The savings and credit programme o f the AKRSP i s based on a system o f collective savings and borrowings. Each member deposits hisiher savings with the Manager o f the village organization. The collective savings of the Village Organization are deposited in a single account held at a commercial bank. Credit facilities are extended ina similar manner. The village organization takes out a single collective loan and disburses sub-loans to members according to individual's credit need. This system minimizes the transaction costs for borrowers inboth savings and borrowings. Due to the peer group pressure, the transparent nature o f the credit operations, and the group guarantees for loans, the repayment rates for loans are almost 100 percent. The employment and income impact, especially for the poor, has been quite substantial. Basedon substantial positive results from group-based savings mobilizations and credit programmes by the AKRSP, aNational Rural Support Programme (NRSP) was started in 1991and expandedin 1992 with a grant from the Government o f Pakistan. NRSP is building a network o f grassroots organizations which plan and undertake their own development. It uses a development strategy which mobilizes communities and maximizes the use o f local resources, similar to that o f the Aga KhanRural Support Programme (AKRSP) andthe Orangi Pilot Project (OPP). The evaluations o f the AKRSP and OPP show that the poor have been successfully helpedby the credit and other relatedinterventions ina financially sustainable manner. Nonetheless, poor households participating intheprogram still lack enough credit and other financial resources to purchase their own land. Source: Malik, Sohail J. 2003b. Rural Credit Markets in Pakistan: Institutions and Constraints. Background paper for Pakistan RuralFactor Markets Study. 59 Implications of Credit Constraintsfor Productivity andIncomes 6.33 Econometric analysis o f plot-level data suggests that lack o f access to credit significantly reduces farm productivity (measured as gross value o f output per unit o f land). Controlling for various plot-level characteristics including plot size, land value, topography, irrigation and soil type, as well as crop mix,59the value o f yields o f the 41 percent o f households who faced credit rationing inthe formal sector was 9 percent lower than the value o f yields o f non-credit rationed households. Households who faced credit rationing inboth the formal and informal sectors (approximately 17 percent o f all households) had a 23 percent reduction invalue o f yields6' 6.34 This reductioninfarm productivity for credit-constrained farmers is likely due to lower inputuse (seed, fertilizer andtractor rentals), as well as fewer long-term investments inland or equipment. Inaddition to this effect on yields, lack o f access to credit may prevent some farmers from cultivating more land (through rental markets). The net effect on area cultivated i s zero, however, unless the land rentedinwould otherwise be fallow. 59 The regressionincludes dummy variables for the type o f crop cultivated (cotton, sugarcane, wheat and rice). 6o The figures cited are ordinary least squares estimates; correcting for endogeneity o f access to credit, the estimated effect o frationing rises to 26 and 45 percent, respectively, though the latter figure i s not statistically significant. See Annex 1. VII. IMPLICATIONSFORINCOMEDISTRIBUTIONAND POVERTY REDUCTION 7.1 The effects o f inefficienciesinfactor markets on income distribution and poverty reduction depend not only on increases inagricultural and non-agricultural output, but on who receives the additional income, as determined by ownership or access to factors o f production. Given the skewedness o f the distribution o f land and the predominance o f own-family labor in agriculture, the direct gains o f agriculturalproduction accrue mainly to large (and to a lesser extent) small farmers. Gains to rural landless households depend on the extent to which the rural non-agricultural sector spurs employment and informal sector capital eamings. 7.2 Empirical analysis o f India, Chile and other countries suggest that these growth linkage effects, combined with reductions inthe real price o f food, have led to significant reductions in ruralpoverty. Trends ingrowth and rural poverty inthe 1970s and 1980s inPakistan show similar patterns. However, rural poverty inPakistan didnot decline inthe 1990s inspite o f high rates o f agricultural growth, calling into question the extent to which rural factor market reforms will leadto major reductions inrural poverty, particularly through the pathway o f agricultural growth. This chapter explores possible reasons for the lack o f correlationbetween agricultural growth and rural poverty reduction inthe 1990s, examines the structure o f factor incomes o f rural households, and presenting simulation results o fthe effects o f broad-based agricultural growth and specific factor market reforms on household incomes. The Puzzleof PersistentRuralPovertyandHighAgriculturalGrowth 7.3 Numerous studies have shown a linkbetween agricultural productivity growth and poverty reduction (Ahluwalia 1978, Mellor 1976 and 1978, Hazel1and Ramasamy, 1991) through reductions infood prices and growth inemployment. Moreover, econometric evidence suggests that agricultural growth i s more effective inreducingrural poverty than i s industrial growth, though this relationshipmay not hold when ownership o f land i s highly skewed (Ravallion and Datt, 2002, Timmer 1997 and Thirtle 2001; see Box 1). 7.4 The experience o f Pakistan inthe 1980s fits this general pattem. Average real agricultural GDP growth was 3.9 percent per year duringthis period, (3.2 percent for crops). Rural poverty (headcount) declined steadily fiom 49.3 percent in 1984-85 to 36.9 percent in 1990-91 to 33.4 percent in 1993-94. Similarly the poverty gap declined from 11.9 percent in 1984-85 to 7.8 percent in 1990-91 to 6.4 percent in 1993-94. 7.5 For the decade o f the 1990s overall, however, rural poverty rates stagnatedinspite o f continued agricultural growth. Real agricultural GDP rose by 4.6 percent per annum, yet the percentage o f rural poor livingbelow the poverty line remained essentially unchanged between 1990-91 (36.9 percent) and 1998-99 (35.9 percent), and even rose to 38.9 percent in 2001-02, a drought year.61 61 Survey evidence suggests that incomes and the level o f poverty (headcount) vary substantially across regions of Pakistan, though problems of sample size and sample design prevent definitive conclusions. World Bank (2002) reports rural poverty headcounts for 1998-99 that vary little betweenPunjab (34.7 percent) and Sindh (37.1 percent), but are higher inNWFP (46.5 percent). 61 Basmati Wheat Wheat flour Rice Average 1970-74 1.04 1.02 1.09 1.05 1975-79 1.05 1.05 1. I O 1.06 1980-84 0.87 0.95 1.I4 0.98 1985-89 0.86 0.95 1.08 0.96 1990-94 0.82 0.94 1.01 0.93 1995-99 0.91 0.99 1.04 0.98 2000-03 0.94 1.05 1.08 1.02 Figure 7.1 Real Prices of Major Food Grains inPakistan, 1970-2003 1.30 I 1 0 1.10 0 II 1.00 0.90 Q) 1 7 0.80 0.70 Note: The three commodity index consists o f an average o f wheat, wheat flour andbasmati rice prices. 62 Datt and Ravallion (1998) showed that a decline inthe real price o f major food staples was a major factor inreductions inrural poverty inIndia inthe 1970s and 1980s. 62 7.8 Slow actual overall agricultural growth and a rise inthe prices o f major staples may be sufficient to explain the low correlationbetween agricultural growth and rural poverty reduction inPakistaninthe 1990s. Inaddition, the skewed structureof ownership and accessto factors of production inrural Pakistan, which are inpart due to distortions infactor markets, may have also contributedto the disconnect between agricultural growth and poverty reduction inthe 1990s. Moreover, these structural factors have important implications for future efforts to reduce rural poverty. 7.9 According to HIES 2001-02 data, more than 60 percent o fruralpoor households are not farm households (table 7.2). Rural non-farm households account for 46 percent o f the rural poor; agricultural laborer households comprise 15 percent o f the rural poor. Though there i s substantial poverty among small landowning farmers (38 percent are poor), this group accounts for only 24 percent o f total rural poor households. Landless tenant farmers (61 percent o f whom are poor) account for another 13 percent o f the rural poor. Farmers without irrigatedland account for only 7 percent o f rural poor households. Total Poor Poor Poor HHs HHs (mns) HHs (percent) HHs (mns) % of totalRural Poor Mediumand LargeFarmers 0.59 24.1% 0.14 2.3% ofwhich farmers w/o irrig. 0.11 31.9% 0.04 0.6% SmallFarmers 3.83 38.3% 1.47 23.7% ofwhich farmers wlo irrig. 1.06 32.0% 0.34 5.5% LandlessFarmers 1.33 60.5% 0.81 13.0% ofwhich farmers w/o irrig. 0.17 49.9% 0.08 1.3% RuralAgric Laborers 1.32 69.0% 0.91 14.7% Ruralnon-farmnon-poor 3.33 0.0% 0.00 0.0% Ruralnon-farmpoor 2.87 100.0% 2.87 46.3% TotalRural 13.27 46.7% 6.20 100.0% ofwhich farmers wlo irrig. 1.34 34.2% 0.46 7.4%( Source: HIES2001-02. 7.10 HIES(2001-02) data onrural incomes also highlight the importanceo fthe ruralnon-farm economy and labor income for the poor (table 7.3 and figure 7.2). Crop income (own-farm) accounted for only 23 o f household income, and only 18 percent o f income for the poorest 20 percent o f households. Livestock accounted for 14percent o f total rural household incomes, with little variation across expenditure quintiles. Thus, agricultural farm income was only 37 percent o f total rural incomes. Wages and salaries accounted for 26 percent o f total reported rural household income across all rural households; for the poorest two quintiles, the labor income shares were 37 and 34 percent, respectively. Foreignremittances accounted for 6 percent o f incomes o f the top expenditure quintile, but only 0.5 percent o f incomes o f the poorest quintile. 63 Table7.3 RuralHouseholdIncomeBy SourceandExpenditureQuintile(2001-02) I Total I Ql 1 Q2 I Q3 1 Q4 1 Q5 Average monthly income (Rs) 6031 4259 4967 5582 6269 8914 Wages and Salaries 26.33 37.45 34.37 29.21 22.95 17.72 Crop Production 22.94 17.11 18.30 21.61 25.95 26.65 Livestock 13.53 12.34 12.35 13.74 13.93 14.29 Non-farm activities 13.49 15.30 14.20 13.44 13.34 12.44 Property income 9.14 6.26 7.08 7.64 9.07 12.55 Social InsuranceBenefits including 1.44 0.72 0.99 1.45 1.78 1.75 Pension Gift Assistance 3.16 3.98 3.24 2.91 2.77 3.22 Foreign Remittances 3.13 0.48 1.38 2.68 2.97 5.67 Domestic Remittance 5.43 4.14 6.19 5.94 5.91 4.89 Other Sources 1.40 2.23 1.91 1.39 1.33 0.83 Figure7.2 RuralHouseholdIncome Sourcesby Quintile,Pakistan2001-02 100% 90% E 80% 70% EOther income y 0 60% 0 Gifts, transfers, pensions b 0Wages & Salaries f! 40% E Livestock 30% t5 20% 10% 0% 1 2 3 4 5 Expenditure Quintile 7.11 Comparedwith the national accounts data, however, total household expenditures and incomes inthe HIESare substantially understated. Total annual private per capita consumption inthe HIESdata is only Rs 11,400, 62percent ofthe nationalaccounts figure (Rs 18,500), (figure 7.3). Inorder to obtain an estimate of the level and structure of household incomes that was more consistent with agriculturalproduction and national accounts data, a social accounting matrix ( S A M ) was constructed using data from these and other secondary sources (see Annex 3). The results suggest that farm incomes are even more highly skewed than indicatedinthe HIES or PRHS, and that the importance o fnon-farm incomes intotal ruralhouseholdincomes i s generally understated inthe household survey data. 64 Figure 7.3 Pakistan: Per Capita Incomes and Expenditures 2001-02 35 L e 30 25 L > & 20 8 15 P 2 10 5 0 Rural Urban Total PRHSIncome HIESExpend. MNationalSAMPnv. Cons. MNationalAccounts Pnv. Cons. 7.12 Based on data from the Agricultural Census, Agricultural Price Commission Cost o f Production, national accounts, and MINFALthe estimates from the S A M indicate that returns to land account for 55 percent o f value added incrop production and 27 percent o f total agricultural income (including livestock, fishing and forestry). Returns to operated land on medium and large farms alone account for 21 percent of total crop value added. Total returns to labor are 27 percent o f value added inthe crop sector, but the share o fhired labor inthe sector value added i s only 5 percent. 7.13 Given this large share of land incomes in agricultural value added and the skewed distribution o f land, medium and large land owners (those with 12.5 acres or more), who account for 10 percent o f agricultural households, receive an estimated 32 percent o f agricultural incomes 7.14 However, agricultural incomes (including livestock) generally account for only 56 percent of total incomes for agricultural households (41 percent o f total population inPakistan). Includingruralnon-agriculturalhouseholds (3 1percent o fnational population), the share o f agricultural incomes intotal rural incomes i s only 37 percent. Thus, incomes from rural non- agricultural activities, includingprocessing and trade o f agricultural products, small industry, construction and general trade and transport services account for 63 percent o f total rural incomes, and essentially all o f non-transfer incomes for about 40 percent o f the rural population (table 7.4). 65 Table 7.4 Pakistan Rural Agricultural Incomes PRHS PRHS SAM SAM Agric Inc Agric Inc Agric Inc Agric Inc Per Capita Share Per Capita Share ('000 Rs) (percent) ('000 Rs) (percent) Medium and Large Farms 15.7 83.5 29.9 57.2 Small Farms 6.1 67.9 8.6 54.8 Landless Farmers 7.2 87.7 5.3 59.7 Rural Agric Workers 2.2 53.1 5.5 53.1 Rural Non-Farm Non-Poor 0.3 1.9 0.1 0.4 Rural Non-Farm Poor 0.2 6.3 0.2 4.5 Total Rural 6.1 69.7 5.6 37.2 Rural Agric Households 7.1 74.8 9.6 55.9 Source: PakistanS A M 2001/02; Pakistan Rural Household Survey 2001/02. Agricultural Growth and Rural Income Distribution 7.15 The contribution o f agricultural growth to increasedrural incomes i s not limited to farm incomes, however. Increases inagricultural productiongenerally involve increased demand for agricultural inputs,processing and marketing services. Also, as householdincomes rise, consumer demand for bothurban andruralproducts and services increases. To the extent that the supply o fgoods and services is elastic, these increases indemand can spur increases in production and further increases indemand. (See box 7.1.) 7.16 These multiplier effects o f agricultural growth and productivity gains from factor market reforms can be estimated usinga semi-input-output model, based on the PakistanS A M described above. The model assumes that output o f tradable goods (major crops, cattle products, most industrialproducts, i s inelastic, and does not expand due to increases indemand. For these products, increased demand results inincreasednet imports. For elastically supplied products, (minor crops, poultry, and services), increased demand i s assumed to induce increases inoutput. 7.17 Simulation results o f a 10 percent increase inthe output o f all major crops (wheat, basmati and IRRIrice, cotton and sugar cane) suggest that the largest gains o f increased agricultural crop production accrue to large and medium land owners, whose incomes rise by 7.2 percent. Incomes of small farm owners and pure tenants also rise by about 4.6 percent. Due to multiplier effects, incomes o f non-farm rural household groups also rise by 3.4 percent, on average. The poorest rural household groups (agricultural laborers and rural non-fann poor (29 percent o f the rural population) reap only 6.7 percent o fthe total income gains, and their incomes rise by only 2.6-4.1 percent. Much o f the gains inincomes accrue to owners o f capital inboth rural and urban areas. 7.18 Gains from a 10 percent increase inlarge livestock production (mainly cattle, goats and dairy products) are more evenly spread out given the distribution o f ownership o f livestock. Overall incomes rise by 4.5 percent, and small farmers and landless agricultural households (who also have significant livestock income) rise by 9.7 to 10.9 percent. 66 Box 7.1 AgriculturalGrowthand Rural Poverty Tamil Nadu, India Despite concerns that the green revolution technology for rice (improved seeds, increased fertilizer use and irrigation) would lead to increased income inequality, most rural household groups benefited from increased rice production inthe 1970s and early 1980s, (Hazel1 and Ramasamy eds., 1991). Household surveys o f farmers indicated that, although the first adopters o f the rice technology tended to be larger farmers, by the early 1980s small farmers invillages with access to irrigation had also adopted the new technologies at the same rate as large farmers. Land distribution remained essentially unchanged, and agricultural wage earnings for agricultural laborers increased (though real agricultural wage rates didnot increase). Moreover, increasedrice production led to substantial rural growth linkages, adding an additional 0.87 rupees o f value added for every 1 rupee of value added from additional rice production. One-half o f these multiplier effects were due to increased demand for agricultural inputs, marketing and processing services; the remainder derived from increased consumer demands as household incomes rose. Survey results showed that real incomes o f small paddy farmers and landless laborers rose by 90 and 125 percent, respectively, between 1973174 to 1983/84. Incomes o f non-agricultural households also rose by 55 percent, due inpart to the growth linkage effects emanating from agricultural growth. Real income gains o f large paddy farmers were only 18 percent, due to increased costs o f fertilizer and labor. Notably, non-paddy farmers with no access to irrigated land saw real income gains of only 17 percent. Similarly, analysis of state-level data on growth and poverty inIndia (Datt and Ravallion, 1998) shows that agricultural productivity growth from 1958-1994 played a major role inreducing poverty. Duringthis period, agriculturaloutputper acre o fnet sown area grew by2.91 percentper year, contributing to higher real wages for agtlcultural laborers (which grew by 2.84 percent per year) and a decline inthe real price of food. (The price o f food relative to agricultural laborers' consumer price index fell by 0.15 percent per year from 1976-94, after having increased by 0.62 percent per year from 1958-75.) Higher farm yields for small producers also contributed directly to their incomes, both directly through increased value of output and indirectly through increased employment. 7.19 Thus, while agricultural growth has substantial benefits for low income farmers, expansion in output (and processing) o f major crops results in smaller gains to rural non-farm households on average, in spite o f significant multiplier effects. These linkage effects o f agriculture are not large enough to substantially raise incomes of the rural non-farm poor for two major reasons: 1) the size o f Pakistan's agricultural sector relative to total GDP and the rural non- farm economy has declined over time, 2) Pakistan's agricultural labor market i s segmented, with much farm labor deriving from own-family labor. 7.20 Over time, the share o f major crops intotal GDP inPakistanhas fallen dramatically, from 0.234 in 1970 to 0.091 in2000 (table 7.5). As a result, assuming an unchanged value-added multiplier, the direct effects of a 10percent gain inmajor crop production is smaller: equivalent to 2.3% of GDP in 1970, but only 0.9% o f GDP in2000. Likewise, the relatively smaller size o f the initial shock and the larger share o f impact o f the multiplier effects o n the rest o f the economy is also smaller (4.6% of GDP in 1970compared with 1.5% of GDP in2000). 67 Table 7.5 Implications of ChangesinEconomic Structurefor AgriculturalMultiplier Effects 1960 1970 1980 I990 2000 2002 Agriculture Share of GDP 0.458 0.389 0.306 0.246 0.240 0.232 Share of major crops in GDP 0.231 0.234 0.176 0.130 0.091 0.078 10% shock to agric (as YOof GDP) 4.6% 3.9% 3.1% 2.5% 2.4% 2.3% Value added multiplier 1.5 1.5 1.5 1.5 1.5 2.5 % change in total GDP 11.5% 9.7% 7.6% 6.1% 6.0% 8.1% % change in non-agriculture 12.7% 9.5% 6.6% 4.9% 4.7% 7.6% 10% shock to major crops (as % of GDP) 2.3% 2.3% 1.8% 1.3% 0.9% 0.8% Value added multiplier 1.5 1.5 1.5 1.5 1.5 3.3 % change in total GDP 5.8% 5.8% 4.4% 3.2% 2.3% 2.7% Yo change in non-major crops 4.5% 4.6% 3.2% 2.2% 1.5% 2.8% Note: The value added multiplier i s defined as the additional change invalue added generated divided by the value added o f the initial production gain (shock). The 2002 value added multiplier is derived from SIO-model simulations; multipliers for all other years are assumed to be equal to 1.5 based o n Hazel1and Roe11(1983), Mellor (1992) and Mellor and Gavian (1999). Source: Dorosh, Niazi and Nazli (2004). 7.21 Moreover, even ifthe value added multiplier has increased substantially over time to 2.5, (the average agriculturalmultiplierimplicit inthe SI0 model), a 10percent output increase in production o f major crops in2002 results inonly a 2.8 percent gain inoutput of other sectors, about 40percent less than the estimated change in non-major crops assuming a multiplier of 1.5 in 1970. 7.22 The implication i s that growth linkage effects have a smaller impact on GDP growth and rural poverty reduction today than in 1970, (unless the gains innon-agricultural incomes go disproportionately to the rural non-farm poor). Agricultural growth still has positive impacts on incomes o f the rural poor, but these effects are not as large, relative to the size o f the economy, as they were three decades ago. Even with agricultural growth o f 3.9 percent per year (the historical average from 1975 to 1990, a rapid period o f real agricultural GDP growth) and the higher growth multipliero f2.5, non-agricultural income growth in2002 would be about 2.9 percent, implyingaper capita growthrate o fless than 1.O percent per year. 7.23 The second reason for the relatively small size o f the multiplier effects on rural non-farm incomes i s that the simulations assume that unskilled labor markets are segmented into own-farm labor, agricultural wage labor and non-agricultural unskilled labor markets. Thus, inthe simulations, most of the gains intotal agricultural labor incomes accrue to farm households (reflecting the average distribution o f agricultural labor incomes inthe PRHS surveys). Ifthis pattern still holds for marginal increases inagricultural output inthe future, increased agricultural output will result inrelatively little increase indemand for landless agricultural labor. The simulations also assume rural and urban markets for unskilled non-agricultural labor are 68 integrated. Thus the gains innon-agriculturalunskilled labor incomes are split across both urban and rural household^.^^ 7.24 Finally, it should be noted that the simulations show the average gains inincomes for the various household groups. Ifnon-agricultural labor incomes are concentrated among only a segment o f rural non-farm poor households, the percentage gains inincome o f this segment o f households will be larger, (and the gains to other ruralnon-farm poor households correspondingly smaller). Factor Market Reforms 7.25 The impact o f factor market reforms on rural poverty depends inlarge measure on the size o f impact on overall productivity inthe agricultural sector (table 7.6). 7.26 Improvements inproductivity arising from a shift from share-cropping to fixed rents could lead to substantial benefits for share-croppers, butrelatively small benefits for most o f the rural poor. Assuming that replacing share-cropping with fixed rent tenancy leads to an increase inproductivity of the former share-croppers by 18percent, (because o fincreasedlabor effort, as suggested by the econometric analysis inchapter 3), would only increase average incomes o f small-farm renters (11percent of the rural population) by 3.0 percent. Since the size o f this productivity shock is small relative to the rural economy, there are only minimal effects on average incomes o f other household groups, even accounting for linkage effects to the rural non- farm economy. Table 7.6 SimulatedDistributionalImpactof FactorMarket Reforms Base Tenancy Credit Increased Water Incomea Reform Reforms Productivity Large Farmers-Sindh 19.1 0.2 5.3 15.4 Large Farmers-Punjab 64.1 0.2 5.0 10.7 Large Farmers-0th Pak 10.8 0.2 3.1 7.6 Med Farmers-Sindh 44.6 0.2 6.4 9.9 Med Farmers-Punjab 146.0 0.3 5.6 8.4 Med Farmers-0th Pak 35.6 0.3 5.8 9.0 SmFarmers-Sindh 57.6 0.4 3.9 6.1 SmFarmers-Punjab 318.9 0.4 4.4 6.6 SmFarmers-0th Pak 125.0 0.3 2.7 4.2 SmFarmRenters-Sindh 43.7 4.0 4.0 6.2 SmFarmRenters-Punjab 46.0 2.8 4.3 6.3 SmFarmRenters-0thPak 15.0 2.7 2.8 4.4 Agric Workers-Sindh 20.8 0.2 2.2 3.3 Agric Workers-Punjab 68.2 0.2 2.4 3.6 Agric Workers-0th Pak 9.5 0.1 1.o 1.5 Non-farmNon-poor 400.8 0.3 2.3 3.4 63 PRHS data indicate that rural non-farm households have a lot o f underemployed labor (per capita non- agricultural labor incomes are five times higher for rural non-farm non-poor households compared with rural non-farm poor households). Given that the marginal shares used to allocate labor incomes are the average shares from the SAM, the income gains to rural non-farm poor households may be underestimated. 69 ~ ~~ ~~ Base Tenancy Credit Increased Water Income" Reform Reforms Productivity Non-farmPoor 134.4 0.3 2.3 3.3 UrbanNon-poor 1744.0 0.2 1.1 1.7 7.27 Easing credit constraints on small farms, assuming a 35 percent productivity gain o f formerly credit-constrained small farmers, has larger effects, raising small farm incomes on average by about 3.8 percent. Because o f positive linkage effects, the non-agriculturalrural economy also expands and other households benefit, as well. Incomes o f most other poor rural household groups rise by about 2.3 percent. 7.28 Under optimistic assumptions regarding increases innational crop productivity (14 percent) as a result o f provision o f water to tailenders incanal systems (through physical investments or establishment of water markets), incomes o f small farmers and tenants rise by 5.8 percent, and incomes o f agricultural laborers and rural non-farm poor households rise by 3.1 and 3.3 percent, respectively. Implicationsfor GrowthandPovertyReduction 7.29 Because o f data uncertainties and simplifyingassumptions used, the above analysis o f the effects o f increases in agricultural output on incomes inPakistan agriculture illustrates only the broad order o f magnitude o f the effects. Moreover, these simulations show only the static results o f factor market changes, not dynamic effects o f improved incentives on private investments and market development. Nonetheless, the broad structure o f rural production, distribution o f land and other factors o f production, and structure o f householdincomes are reflectedinthe analysis. 7.30 Two broad conclusions emerge. First,the level and distribution of the benefits o f improvements infactor market efficiency depend crucially on the magnitude o f the induced productivity shock and the ownership of the assets involved. Inspite o f multiplier effects, reforms relatedto share-cropping are likely to have only small overall impacts on rural incomes inaggregate inthe short run,thoughthe benefits to tenants couldbe significant. Creditmarket reforms and improvements inwater availability have potentially larger effects on the overall rural economy. Second, the linkage effects o f expansion intraditional crop agriculture, though important, may not be adequate to substantially raise incomes o f agricultural laborers and the rural non-farm poor who lack significant capital resources. Thus, inthe absence o f a change in the structure o frural incomes and employment or significant gains inthe rural non-farm economy apart fi-om agricultural growth-inducedlinkage effects, targeted interventions to agricultural laborers and the rural non-farm poor will likely be needed to rapidly reduce the poverty o f these groups. VIII. CONCLUSIONSAND POLICYIMPLICATIONS 8.1 Well-functioning factor markets can help maximize productivity by providing a mechanism to allocate all factors o f productionto their most productive use. Significant inefficiencies infactor markets inPakistan, however, limit agriculturalproductivity and rural income growth, (especially lack o f water markets to allocate water to farmers and land without sufficient access to water, and underemployedlabor). Experience inother countries suggests that the long-term gains in productivity from improvements inthe functioning o f factor markets could be large, particularly given the inter-linkages across factor markets, resulting in significant potential gains for the rural poor. 8.2 Markets for land, water, labor and credit vary considerably intheir performance, however. Rural factors inPakistan that are most rigid in supply (land, water) and involve the highest transport costs, have the thinnest markets, and face the greatest political constraints to reforms. These contrasting characteristics to a large extent determine limits on factor market efficiency and feasibility o f policy changes (table 8.1). Land 8.3 Land i s at the heart o f Pakistan agriculture and the rural economy. Arable land i s essentially fixed in supply inthe absence o f major new investments inirrigation and drainage. As a fixed factor inarelatively labor abundant agriculture, returns to land are estimatedto be about halfofincomes (value added) from crop agriculture. The unequal distribution o f land, therefore, i s a major reason behind the highincome inequality inrural Pakistan. Moreover, inpart because o f inefficiencies in land administrationand land markets, land i s rarely bought and sold, helpingto perpetuate the highly unequal distribution o f land, hamper labor mobility o f farmers who want to sell land, and reduce returns to family labor. The skewed land distribution and inefficienciesinland markets also reduce access to credit for non-landowning households, since land i s the major acceptable form o f collateral informal credit markets (table 8.2). 8.4 Evidence on the impacts o f land market imperfections on efficiency of land use are mixed, however. Agricultural census data indicate that crop choice does not vary substantially by farm size. Similarly, there i s little variation incropping pattemby tenure status. Nonetheless, survey data on crop income per cultivated area inPunjab suggest that incomes per cultivated acre o f farms less than 5 acres in size are nearly double those o f farms 25 acres or more. More recent econometric evidence of productivity by plot that correct for some land quality characteristics, however, suggest relatively small negative economies o f scale inproduction, i.e. that a doubling o f operated area leads to a 10-13 percent reduction inproductivity. 8.5 Three major attempts at redistributive landreform inPalustanhave failed (most recently in 1977), and landreform has neither political support, nor the backing o f Islamic religious authorities. Because o f these severe institutional constraints to traditional redistributive land reform, more feasible options to increase access to land by poor households involve a win-win approach. Land purchase schemes that include grant components for the poorest landless households are one example, though fiscal costs could limit the scale o f such programs. Policy measures to increase access to land also include increased access to credit to enable poor households to purchase land, landtaxation to reduce incentives for holding land for speculative purposes, 64 and measures to improve the efficiency o f land sales and rental markets. 64 Increases inland taxes (and water charges) would reduce incentives for holding o f land for speculative purposes, and encourage greater land sales. E 73 8.6 Administrative reforms inthe enforcement o f existingnorms and regulations, including enforcement o f contracts and the rights o f tenants, and liftingrestrictions for rentals would also help improve the functioning o f land markets. Simplification o f complicated legal procedures that result in long delays inproceedings and a backlog o f cases would reduce land transactions costs. Poor households are at a particular disadvantage in cases brought before civil courts, since they often cannot afford to pursuethese cases. Possible steps include limitson the number o f appeals, proceduralreforms, alternative conflict resolutionmechanisms, and barring o f the jurisdiction o f civil courts in landrevenue matters. Improveddissemination o f information on land prices and transactions would also enhance efficiency o f land markets (box 8.1). Box 8.1 ImprovingtheEnforcementofContracts inLandAdministration inPakistan Enforcement o f land contracts remains a major probleminPakistan. Complicated legalprocedures result inlong delays inproceedings and a backlog o f cases. The poor are often losers inthis process since they have less resources. Some possible remedial measures to improve the enforcement o f contracts inthe LandAdministration are as follows: On-the-spot inspections to establish facts on the ground and summary methods o f disposal to avoid longprotractedlitigation. Adequate safeguardsneedto be maintained so that the process i s open and transparent. Reduction inthe number o f tiers inwhich a case can be heard. For most revenue cases, the last court o f appeal shouldbe the Deputy District Office (DDO) and not the Board o f Revenue(as is the casepresently). Barring thejurisdictiono f civil courts inrevenue matters. Frequentinterventionby the civil courts on the pretext o f enforcing legal procedures(and determininglegal rights o f the contesting parties) works to the disadvantage o f the poor who lack resources to pursuethese cases. Fixedtime limits for each category o f land dispute and imposing a law on the postponement o fhearingby revenue courts on administrative grounds to help expedite the settlement o f dispute. Improvements inthe poor quality o f landrecords and clarification o f ambiguous laws. Establishment of a systemof open courts and grievance hearing days by senior officials to reduce the transaction costs o f obtaining legal redress. AdaDted from S. Oureshi (2003). 8.7 Improvements inlandrecords could also enhance the efficiency o f land markets and increase security o f tenure. Without secure rights to their land, farmers not only have less incentives for investment, but also devote resources to defending their rights. Lack o f secure title also reduces access to landby landless households through rentals by increasing the risk o f landowners who might otherwise rent out more land. Inaddition, banks are reluctant to lend money usingland as collateral because they do not trust the current recording system. Questions regarding the validity and enforceability o f titles also make it harder to buy and sell land. Computerization o f land records, as tested in a recent pilot project inPunjab, i s one option for establishing clearer landrecords and reducingfraudulent dealings by land administration Officials (box 8.2) 74 Box 8.2 Examples of LandReform and Land Titling Projects Brazil: Participatory Negotiations and Market-Assisted Land Reform The Northeast, accommodating one-half o f all Brazilian poor people and two-thirds o f all rural poor people, i s vulnerable to drought and has a relatively poor resource base. Additional constraints to improved investment and productivity inagriculture include ineffective financial systems and distorted land distribution. Studies have shown that family farms inBrazil are more productive and labor intensive than large farms. The highlydistorted land ownership pattem is the result o f economic distortions, including subsidized agricultural credit, highinflation, and inappropriate tax provisions. Past approaches to land reform, based on centralized government-administered expropriation and redistribution, have had limited success. As a result, the government was interested inexperimenting with faster, cheaper, and less conflictive approaches to landreform. What's innovative? A market-based approach to landreformwith negotiations made directly between willing buyers (poor beneficiaries) and willing sellers, with financing for purchases made available through a government fund. PROJECTOBJECTIVES AND DESCRIPTION The objective of the Land Reform and Poverty Alleviation Project was to raise agricultural output and increase poor family incomes by providing improved access to land and funds for complementary investment subprojects planned and implementedby community associations. The project experimented with a program o f market-assisted land reform inwhich beneficiaries are given access to financing for the purchase o f suitable lands. The purchases are negotiated between willing sellers and willing buyers. Five states were selected for participation inthe pilot project based on the severity of the landlessness problem and conditions for successful implementation (that is, the immediate availability o f land inthe market and the capacity o f the State agencies to implement the project). The project had five components: A landpurchase fund to finance landpurchases. Community subprojects (small matching grants to communities for investment projects, technical assistance, and start-up). Institutional strengthening (technical assistance and training at the state level). Project administration, supervision, and monitoring. Impact evaluation and dissemination by the federal government. Community associations consisting o f landless rural workers or rural workers owning land sufficient only for subsistence farming selected suitable land and negotiated its purchase with willing sellers. Following negotiations, the associations consulted with the State Land Institute to confirm that the title was clean and that the negotiated purchase price was consistent with market conditions. Communities then presented their project to the State Technical Unit (STU), which verified the eligibility (based on agricultural skills) o f the beneficiaries. At this stage, beneficiaries were eligible for credit from the Land Purchase Fund with credit initially given for a 10-year period with a three- year grace period at the government long-term interest rate. Lao PDR: Second Land Titling Project. The Laos Land Titling Project (LTP II), co-financed by AusAID, is the secondphase o fthe long-term landtitling program aiming at the development of the land administration capacity to support economic development and poverty reduction goals. The project is aligned to the government's National Growth and Poverty Eradication Strategy Program Box 8.2. Cont'd. 75 (NGPSEP), approved bythe National Assembly inOctober 2003. The project responds to the findings ofthe I-PRSP that highlightedthe importance of landpolicy and landmanagement to provide the basis for poverty reduction and economic growth. The project will buildupon the achievements and experiences o f the first LandTitling Project (LTP). LTP I1will provide secure land titles inperi- urban (mostly agricultural land) and urbanareas, providing incentives for land holders to invest in land development and insustainable agricultural practices and enabling greater access to credit for agriculture inputs and landimprovement. The project will help buildgovemment capacity to coordinate the development of the national landpolicy and regulatory framework, and will support the formulation and adoption ofkey landpolicies for landadministration and management. Punjab:Computerizationoflandrecordsproject. InPunjab, landdisputes are primarily due to fraudulent dealing becausethe records are not clear and patwars (land administration officials) often accept informal payments. Moreover, banks (which are currently capital rich), are reluctant to lend money usingland as collateral because they do not trust the current land recording system. To address these problems, the govemment of Punjab initiated a landrecords project as apilot inone district (Kasur) in 1996/97, butbecauseo f administrative problems, major project activities didnot begin until 1988 whenthe Director ofLandRecords ofthe BoardofRevenuewas putincharge. Progress has beenslow, however: ina five year period from 1998 to 2003, landrecords hadbeencomputerized inonly 601 ofthe 641revenueestates inthe district. InAugust 2003, the pilotwas transferred to the Kasur district government. Inlate 2003, the provincial government decidedthat a new model, and therefore a new pilot (inLahore city), would be requiredto accelerate the process o f computerization of landrecords. Progress with the newmodel has beenquite rapid, and as a result, the provincial govemment has decided to scale up the project to other districts. Source; World Bank.2004b. ImplementingReaching the Poor: A progressreport onRural Strategy Implementation for the World BankBoardof Executive Directors; HanidMukhtar, World Bank Pakistan office. Water 8.8 Although surface water, unlike land, i s transportable, there are often major conveyance losses, and annual supply depends mainly on weather (and long-term investments inwater storage). Transactions costs, including the management costs involved inadjusting flows o f canal water, have limited water trading and market transactions to local areas to date. Informalwater markets for groundwater, however, have been shown to significantly increase agricultural productivity who lack access to their groundwater sources. Development o f water markets for surface irrigation (which are currently highly localized, with most transactions on a barter basis), could similarly increase productivity o f canal water by allowing markets to help allocate water to its most productive use. 8.9 Several institutional changes could bringabout major improvements inthe efficiency and equity of water use inPakistan agriculture, with or without the establishment o f water markets, however. First,inorder to increase accountabilityfor Operations and Maintenance, management of distributaries andor minors couldbe transferred to Farmer Organizations (FO's), permitting assessment, collection and retention o f abiana (water charges) by the FO's. Recent administrative reforms giving greater control to farmer organizations (Water Users' Associations at the minor level and Water Course Associations at the water course level) in some parts o f Sindh have shown promise inreducingwater theft byclosingillegal outlets andthereby increasing accessto water for tail-end farmers. This suggests that increased local involvement in canal water management can improve the 76 equity o f water d i ~ t r i b u t i o n .In~addition, improved local maintenance o f water channels by ~ Farmers' Organizations has contributed to increased efficiency o f water use. 8.10 Second, to increase transparency and enable more informed decisions on area irrigated and crop choice, periodic measurements o f the actual rate o f water flow at various points inthe water course could be made. Having accurate information on water flows would enable farmers to make more informed decisions regarding crop choice and area irrigated. For example, given a shortage o f available water, a farmer might choose to maximize his overall revenues through reducingtotal area cultivated, so that crops grown on the remaining area receive adequate water. 66 Open access to the writtenrules o f the water rotation for farmers would also improve transparency. 8.11 Third, the current system is supply-drivenand mostly managedbyprovincial Departments of Irrigation. Reforms inthe institutional arrangements for allocation of water could be made to create a more demand-drivensystem that would better respond to the needs o f the farmers. Options include a) a system o f Water Course Associations, Water Users' Associations and Area Water Boards (AWB's), as is currentlybeing implementedinfour o fPakistan's 43 canal command areas ,and b) professional canal management. 8.12 Development o f effective water markets would require two steps beyond those outlined above: i)establishing legal tradable water rightsat the national and provincial levels; ii)strengthening local institutions (such as Water Course Associations and Water Users' Associations, or professional canal management agencies) to enable them to serve as brokers inwater markets, helping to match buyers and sellers o f water. 8.13 Legal establishment o f tradable water rights would formalize these water markets, defining the initial allocation o f rightsalready implicit inthe warabandi systemo f water r~tation.~'These water rights would need to include an explicit formula for allocation o f water rights inyears o f shortage, as well as address environmental issues (see box 8.3) 65 Inthe command area covered by the Left Bank Area Water Board in Sindh, where tail end farmers had not received water inthree years, 48 illegal direct outlets serving an area o f about 25 thousand hectares were closed in early 2004. 66 This principle may apply to Pakistanagriculture as a whole: higher agricultural incomes mightbe achieved through a contraction intotal irrigated area so that crops grown on the remaining irrigated area received optimal amounts o f water. 67 Pakistani Law treats all surface water as provincial property. Under the Pakistan Constitution, while the Federal government i s responsible for regulation and development o f inter-state rivers and river valleys, provincial governments are responsible for water supplies, irrigation and canals, drainage and embankments, water storage etc. Under the de facto interpretation o f these responsibilities, the power o f the provinces has emerged as pre-eminent and water has come to be perceived as a provincial subject. Establishment o f tradable water rightswould thus come under the purview o f the provinces, though federal legislation might be required, including amendment o f the Canal and Drainage Act (1873) which ties water allocations to specific plots and also ties land sales with water sales, and the LandReforms Act (1972) which makes landowners responsible for water charges. 77 Box 8.3 Important issues regarding tradable water rights 1. Method for initial allocation of water rights: historical water use (e.g. Mexico) or historical use combined with redistribution o f large historical shares (e.g. Chile) 2. Prior o r proportional water rights: appropriative rights systemthat allocates water first to those who have heldthe rights the longest period o f time (e.g. California) or shares over a variable flow or quantity o f water (e.g. Mexico and Chile) 3. Consumptive use and return flows: tradable water rights are full diversion rights with no rightsto returnflows (Mexico and Chile) or limitedtransferability ofwater rights maintaining third-party rightso f retum flows (California) 4. Protection against negative indirect economic effects: negative indirect economic effects from water trading have beensmall or non-existent in California, Chile and Mexico 5. Protection of the environment: laws prohibitingwater transfers with unreasonable impacts on wildlife (California); quality o f discharge for non-agricultural uses mustbe specified in water rights(Mexico) 6. Role of water user associations: inChile, W A Sown and manage the physical infrastructure, monitor allocation o f water, approve water transfers, provide a forum for conflict resolution; inMexico, water rights are generally initially provided to WUA which thengrant water rightsto their members 7. Infrastructure required for tradable water rights: InChile, water is usually only measured in the mainchannels; elsewhere, simple proportional division devices are usedto divide the water inshares proportionalto canal flow. Source Rosegrant, Schleyer and Yadav, 1995. Rosegrant, MarkW., Renato Gazmuri Schleyer and Satya N.Yadav. 1995. "Water policy for efficient agricultural diversification: market-basedapproaches". FoodPolicy, Vol. 20, No. 3, pp. 203-223. 8.14 Any system o ftradable water rights would needto account for the minimal degree o f water control above the water course level inthe current Indusbasin system, however. Because there i s essentially no storage capacity within a water course (and little control o f water flow into the water course), individual farmers are technically not permitted to refuse to take their turns in the distributionsystem. Likewise, water control at the distributary and higher levels i s also greatly limitedby the absence o f storage capacity and adequate number o f control structures (gates) to regulate water levels. 8.15 Thus, institutional mechanisms to enforce limitsonnet water transactions or new investments in storage and control structures would be needed to insure adequate drainage (e.g. that farmer sales o f water rights did not leadto excessive water flows downstream) and minimum water flow velocity inthe canals (to avoid siltation). (box 8.4) Overcoming these physical constraints o f the canal system to permittrading beyond the boundaries o f a water course through investments inwater storage capacity and control structures or coordination o f the volume o f water trades becomes progressively more difficult and expensive as the level at which water transactions are permitted (i.e. water course, distributary, minor) increases. 78 ~~ ~ Box 8.4 InfrastructureConstraintsfor Developmentof Water MarketsinPakistan The physical layout (geography) ofthe IndusRiver Basin canal systemand existing infrastructurelimit the degree o f water control and the scope for water markets outside a given water course (thetail end o f the irrigationdeliverysystem). Inthis run ofthe river system, storage capacity is only about 13 percent o f annual averagewater flow. (By comparison, storage capacity inthe controlled storage system o f the Colorado Riverbasin inthe UnitedStates is about five times annual average water flow.) Thus, there is very limitedcapacity to control total annual water flow inthe IndusRiverbasin. Moreover, there are very few control structures (gates) to regulate water levels and water flow velocity mustbe maintained aboveaminimumrate to avoid siltation of canals. Inthe absenceof storagecapacity and control structures, water trading is largely limitedto within a water course where the warabandi system o f water rotation is inplace. Becausethere i s essentially no storage capacity within the water course (and little control o f water flow into the water course), individual farmers are technically not permittedto refuse to take their turns inthe distribution system. Informal trades and sales within water courses do occur, but these exchanges do not affect the total inflow of water or water use inthe water course. Incontrast withthe IndusRiverbasin, the geography and infrastructure ofthe riverbasins incountries with widespread markets enable far greater water control. InMexico and the US.riverbasins with developed water markets are control storage systems. In Chile, there are numerous relatively small river basins with their own water markets, lessening theproblems of coordination andwater control. Establishing water markets for surface water would require both changes inlaws to establish water rights separate from ownership of land, reliable measurement system (including water flow measurement devices/ personnel to measure/systems o f monitoring / farmers organizations - water users associations), and the capability o f controlling water flows. Developing water markets above the water course level inPakistan would require investments in storage (dams) and control structures (gates) so that water flows could be divertedinaccordance with water transactions. Such investments at the minor level may be feasible, permittingwater transactions within aminor. Establishingeffective water control at distributary or higher levels, however, would require sizeable investmentsinstorage and control structures. 8.16 As shownby the experience o f Chile, increasing the efficiency o f water use (inthe Chilean case through the establishment o f water markets) can be an integral part o f pro-poor agricultural growth. Trade and exchange rate policy reform (including reductions intariffs, removal o fnon-tariff barriers, and liberalization o f foreign exchange markets), together with other macro-economic reforms, provided incentives for agricultural exports. Development o f highvalue agricultural products (fruits, vegetables, andwine) for exportto Northernhemisphere (off season) markets was facilitatedby establishing security o f land titles (that encouraged investmentsinplanting orchards, forestry, and irrigation) and water markets that helpedto allocate available water to highvalue crops. Processing o f the agricultural exports was labor- intensive, increasingwages o f unskilled labor and contributing to reducedpoverty. (box 8.5). 8.17 Given the infrastructure constraints inPakistan, the potential gains from establishing water markets may be smaller than inother countries, unless an institutional mechanism for insuringthat adequate drainage andwater flow levels can be devised. Nonetheless, benefits o f more efficient allocationo f water through institutionalreforms, with or without water markets, 79 could still be substantial, ifthey are combined with other measures to promote agricultural diversification, including strengthening security o f land titles, public investments inrural infrastructure (roads and electricity), private investments inprocessing and storage, and establishment o f market links to export markets. Box 8.5 FactorMarket ReformsandAgriculturalDiversificationin Chile InChile, the establishment oftradable waters was linkedto reformsinthe landmarkets in 1975, underwhich landwhich hadbeencollectivizedin1966was re-privatized. Former landowners received 40 percent o f their original landalong with the proportional right to water historically used on this land. The remaining 60 percent o f the land and the water rightswere allocated to former works on the lands. These reforms, along with a broad economic reformincludingtrade and exchange rate liberalization, ledto the creation ofwater markets inthe early 1980s, increased water use efficiency, and rapid crop diversification. Reforms inwater markets resultedinincreased scarcity value o f water, and a large increase inarea plantedto fruits and vegetables, whichrequire less water per gross value o f output than do field crops. Aggregate water use efficiency inagriculture increased by an estimated 22-26 percent between 1976 and 1992 (Frias, 1992; Munita, 1994). Because o f the intensiveuse o f labor inprocessing o f fruits and vegetables, particularly for the export market, agricultural diversification has ledto substantial ruralpoverty reduction. Using data from 1990 to 1996, Lopez and Anriquez (2003) estimated that a 4.5 percent inagricultural output in Chile leads to a 8-11percent decreaseinthe poverty headcount, mainlyby raising real wages o f unskilledworkers. Source: Rosegrant,Schleyer andYadav, 1995; pp. 208, 210. Frias, J.L. 1992. Evolucion Reciente de la IndustriadelAgua enInglaterra. Francia y Chile, McKinsey andCompany, Mexico. Munita, J. 1994.Aumento de Eficiencia en e l Us0 del Agua por Incorporacion de NuevasTecnicas y ArrendamientosTemporales. Universidad de Chile, Santiago. Lopez, Ramonand GustavoAnriquez. 2003. Poverty andAgricultural Growth: Chile inthe 1990s. Paper preparedfor the Roles of agriculture International Conference, 20-22 October, 2003. Rome, Italy. Labor 8.18 Unlike the markets for land and water, rural labor markets are generally characterizedby demand constraints, rather than supply constraints, as evidenced by substantial underemployment. There i s considerable labor mobility, inspite o f hightransactions costs involved injob search and especially inmigration. However, formal sector employment inrural areas i s minimal, and over 80 percent o fagricultural labor and most o frural non-agricultural labor is self-employed. Female participation inlabor markets remains limited, though, andjob skill levels for both male and female labor are generally low. 8.19 A better functioning landmarket could promote labor mobility, helping farmers who want to leave farming to sell or rent out their farm land. This would require greater security o f property rights,and facilitation o f land transactions that are presently stifled by.the absences o f a valid title, risks o f court challenges, and highregulatory and transactions costs (for example, stamp duty plus registration fee plus capital tax amounted to 16%o f the average market price o f 80 land in 1995).@ The case o fPoland suggests that removal o f impediments to landtransactions can facilitate labor market mobility. (box 8.6). Box 8.6 Factor Market Reforms and Labor Mobility in Poland A major concern as Polandpreparedtojoin the EuropeanUnion was that a large part ofthe country's agricultural sector was uncompetitive with imports from other EUcountries. Moreover, a decline infarm incomes could have especially severe impacts on the poor, given the predominance o f small farms and the highincidence o f poverty inrural areas incentral and southern Poland. Background studies of land, labor and credit markets, combined with analysis of rural household survey data showed low mobility out of farming, however, particularly among small farmers (owning 2 hectares o f less). One reason for this lack o f mobility was that households were requiredto hold land inorder to qualify for the highly subsidized agricultural pensionscheme, which was more remunerative than the non-agriculture pension scheme. A secondreason for the lack of mobility out of agriculture was a lack o f ruralnon-farm employment opportunities. Taxes on labor, regulations requiringhighseverancepayments, social security tax for employers, and minimumwages madehiringofworkers expensive for ruralnon-farm employers. These interactions o f labor and landmarket restrictions inhibitedfarm restructuring. Although farmers lobbied for more farm subsidies, little could be done on output markets inlight o f trade and fiscal policy constraints linked to Poland's impending accession to the EU. Reforms infactor markets offered an alternative mechanism for smoothing the economic transition. Source: World Bank (2001). 8.20 Increasingthe marketable skills o f rural labor, though effective and wide-spread vocational training programs would spur growth inthe non-agricultural rural economy and rural incomes. Given the low participation rates o f women inthe labor force, slulls training for women i s especially important: one option for increasing women's access to educational and training institutions i s to reserve more spaces for women. 8.21 Labor regulations (that mainly apply only to urban areas) do not directly alter employment conditions inrural areas, but they do have a significant effect inthe dynamics o f employment creation inurban areas and thus influence incentives for rural-urban migration and rural labor markets. Employment growth inurban areas has been slow, inpart because o f an atmosphere o f mutual hostility and distrustthat typify industrialrelations. Inparticular, the unusually highdegree of discretion over conditions o f employment o f regular workers in specific establishments heldby provincial authorities i s not conducive to cooperation betweenemployers and workers. Replacingthe discretiono f authorities on the specific nature o f labor contracts by a few rules on the most common aspects o f these contracts (minimumage, minimumwage, overtime pay) and the establishment o fprocedures to help deal with possible disputes arising from voluntarily agreed contracts regarding more detailed aspects o f these contracts would help buildtrust inthe industrialrelationprocess, and encourage employers to createjobs and invest in human capital. 68 Qureshi, 2003. 81 8.22 A concerted effort to enforce existing legislationregardingthe rights o fwomen and children, and to abolish bonded labor i s needed. Measures to assist households trapped inbonded labor arrangements, including legal protection, provision o f credit and debt cancellationprograms shouldbe urgently considered. Although these programs often face considerable opposition from powerful vested interests (box 8.7), such programs could have a major impact on some o f the poorest rural households. Box 8.7 BondedLabor and Reformsin Nepal [nNepal, government policy reforms in2000 aimedat raising agriculturalwages and endingbonded labor (Kamaiya) encountered considerable opposition. InJanuary 2000, the Local Self-Governance Act established a minimumwage for agncultural laborers o f Rs. 74, just over US$1 per 8-hour workday, Later that year, bonded-labor families filed a petitionagainst their master, an ex- government minister, demanding minimumwages in compliance with the new regulation. These workers, were beingpaid the equivalent o f about Rs 14-16 a day inrice, far below the newly Zstablishedminimumwage. When the former minister refused, the bonded laborers attempted to register a complaint with the ChiefDistrict Officer. Thereafter followed strikes and protests by human rights activists and others inaneffort to endthe practice ofbondedlabor, which culminatedwith a July 17, 2000 national government decree declaring the practice of bonded labor to be illegal. An estimated 16,000 to 40,000 bondedlaborers (mostly inwestern Nepal) were declared free o f their ancestral debts. The decree also mandated aprison sentence o f up to ten years for anyone violating the ban onbonded labor. Many landlords strongly resistedthese reforms, however. Thousands of liberated Kamaiyas were drivenout o f former homes, and their possessionsconfiscated by landlords. Many former Kamaiyas movedto refugee settlement camps infive districts inWestern Nepal andhave continued to press for the enforcement ofthe minimumagricultural wage; 10kuttus of landfor each displaced former bondedlaborer family; identitycards for former bonded laborers; protection from police and former landlord harassment; food, health services, and education for people inthe temporary settlements. Inresponseto the needs ofthe former bondedlabors, the InternationalLabor Organization(ILO) began a $3.5 million project, fundedby the U.S. Department of Labor, that provides vocational training, education, legal and counseling services, micro credit, and the establishment o f revolving savings and credit funds. Sources: Sushma Joshi, "Bonded to Labor: The Contemporary Situation inNepal", South Asian Magazine for Action andReflection (SAMAR) 16: FalUWinter, 2003, (ww~~~.samaniiiarazine.orq). www.us.ilo.orelarchivelilofocusl200 llwinter Credit 8.23 Credit markets are potentially the most flexible o f factor markets, involving essentially no transport costs, though considerable transactions costs for small farmers inrural formal credit markets. Nonetheless, formal credit has often been rationed, and access to credit i s a significant problem. Approximately 17 percent o f all rural households face credit rationing inboth the formal and informal sectors, resulting ina 23 percent reduction inthe value o f crop yields. 8.24 Removal o f subsidies on credit (implicit inthe continuing low repayment / highdefault rates for loans inthe formal sector) could eliminate these policy-induced incentives toward 82 capital-intensive technologies that reduce labor demand. Efforts at promoting agricultural diversification shouldthus avoid subsidies on credit that would encourage capital-intensive technologies inproduction, processing and marketing. Instead, agricultural diversification, the development o f the rural non-farm sector and poverty reduction may be better achieved through public investments inelectricity, transport and other infrastructure. Further work on particular constraints to rural non-farm investment, employment and growth i s needed, however. 8.25 Production credit and micro-credit programs targeted to landless and small farmers that required other forms o f collateral (besides land) could also helppromote rural employment and income generation. Rural credit from formal sources i s currently available for a narrow range o f agricultural production activities and does not serve the needs o fthe non-farm sector. In particular, availability o f medium-term credit i s extremely limited. Adaptations to conventional lendingproducts and delivery mechanisms, such as matching disbursement and repayment to agriculturalproduction cycles, reducing costs through use o f appropriate modem technology, and riskmanagementthroughportfolio diversification canhelp make micro-finance for agricultural more viable (box 8.8). Increased access to credit by small scale enterprises inrural areas, for example, would promote farm to non-farm linkages, develop the ruralnon-agricultural sector and increase labor demand. Targeted matching grants to poor households for income-generating activities are one option (box 8.9). 8.26 The Government o fPalustan inrecent years has, infact, placed considerable emphasis on developingmicro-finance as part of its poverty alleviation strategy. In2003, Rs. 1,442 million (2.5 % o f total institutional credit) was disbursed as small loans through the Pakistan Poverty Alleviation Fund(PPAF), Khushali Bank and the Zarai Taraqiati Bank Limited (ZTBL, formerly called ADBP). Comprehensive analysis o f existing micro-finance programs i s needed to assess the effectiveness and financial sustainability ofthese programs. Promotion o f savings mobilization schemes and implementation o f an adequate regulatory framework for deposit-based institutions could help provide sustainable resources for micro-credit. 83 Box 8.8 Micro-finance for Agriculture Microfinance institutions (MFIs) have tended to avoid less densely populated or diversified rural areas, and financing o f seasonal or longer-term crop and livestock activities. Conventional micro credit relies heavily on short-term loans with frequent, regular repayments, which does not fit well with seasonal crop productionor livestock production (except for poultry). Adaptations to conventional financial products and delivery mechanisms can make micro-finance for apculture more viable: Matching disbursement and repayment to agricultural production cycles. Flexibility in loan disbursement and repayment is needed, with finance made available when farmers need it and repayments matching income from produce sales. PRODEM, a rural MFIinBolivia, adopted a customized repayment scheme for small farmers, with differing repayment schedules even for members within a solidarity group to fit harvest calendars. Individual agricultural loans were also introduced, with collateral at a ratio o f 1.5:l to the loan amount. PRODEMfurther minimizes risk by restricting final loan payments to a maximum of 60 percent of the loan amount, and by limitingeach office's portfolio ineach economic sector to 30 percent. Agricultural lending now accounts for about one-fifth o f PRODEM's loan portfolio. Flexibility in collateral requirements. Land may hold little value as collateral, as land-use rights may be difficult to prove, clients may not own land, land markets may be weak, or cost o f registering land as collateral may be high. Financial service providers therefore need to be more flexible in terms o f the collateral required, even if the value of nomortgage guarantees is sometimes more significant as a repayment incentive than its real resale value. Personal guarantees, movable assets, and group guarantees can all be adequate alternatives. Using technology. Technological innovations can increase operational efficiency and lower costs o f operating inrural areas, while improving financial services available to rural clients. There has been a tendency to overestimate the short-term benefits and underestimate the up-front implementation costs. For example, inIndia, the Swayam Krishi Sangam's experiment combines smart cards with hand-held computers to streamline meetings between clients and borrowers inremote areas. However, this did not produce the dramatic time savings expected, and expansion o f the project has beenput on hold. Flexible delivery mechanisms. Making use o f existing delivery outlets, rather than investing in expensive new branch networks, can significantly lower the cost o f providing financial services, and can also allow a wider range o f services to be provided. Options include: working with/through rural post offices, retail stores, rented offices inschools and hospitals, or shared offices with other financial institutions. Mobile staff can help reduce operating costs and improve access inmore remote areas. Riskmanagement techniques. A principal factor discouragingMFIlending to small farmers is the systemic risk inherent inmuch smallholder agriculture. Most MFIs that have successfully moved into agricultural lending have used a diversification strategy to reduce lending risk,both intheir portfolio and at the householdlevel. Portfolio diversification helps ensure that a loan portfolio be diversified across sectors and regions/communities, and that repayments do not fall due at the same time. This increases the stability o f the portfolio and reduces lendingrisk from weather events and from price fluctuations incertain crops. Confianza, a Peruvian MFI, which developed from a purely agricultural portfolio, has now set a target percentage for agricultural lending o f 30 percent of its overall portfolio. 1Source: World Bank. 2004. Agricultural InvestmentSourcebook. 84 Box 8.9 Suggested guidelinesfor providingmatchinggrantsfor income- generatingactivities Target matching grant assistanceto very poor communities or groups that are beyond the current reach o f micro-finance institutions. One way to make the grant amount large enough to be usehl to the poorest, but too small to be worthwhile for better off groups. IGAsfinanced mustbe cost-effective ways of reducingpoverty. This canbe demonstrated by impact and cost-effectiveness indicators. The investment shouldnot compete directly with private investment (existing or likely). Beneficiaries should always contribute at least partly incash (no less than 5 YOo f the total project cost), which will demonstrate commitment and can serve toward the initial working capital to operate the activity. Over generous subsidies may create false positive demand (even where beneficiaries contribute inform o f labor or cash). This mayjeopardize future maintenance o f the investmentby beneficiaries. Subsidies usedwithout tied commitments for repayment, user fees, or maintenance, may create perverse incentives for greater rent-seekingby beneficiaries. Grants should be combined with training and support to establish local savings and credit associations to capture increased income flows and finance future economic activities (group or individual) on a sustainable basis. Subsidies basedon percentage cost-sharing o f subprojects creates an unknown contingent liability for government. This can be avoidedbybasing these on per capita amounts, or fixing a ceiling for government contributions with excess beingmetby beneficiaries. Source World Bank, Agricultural Sourcebook, 2004a. Implicationsfor RuralPovertyReduction 8.27 Increased agriculturalproductionhas been a major force for reducingrural poverty over much of South Asia over the past four decades. Increasedproduction o frice and wheat6' through adoption o f green revolution technology o f improved seeds, irrigation and fertilizer, directly raised incomes of small farmers, led to reductions inthe real price o f major food staples for consumers, and spurred employment and increases inreal wages through demand for agricultural labor, as well as demand for labor inthe ruralnon-agricultural economy. 8.28 Rural poverty inPakistan did not decline inthe 1990s, however, in spite o f agricultural growth. Overestimation o f true sectoral growth rates because o f abrupt changes inestimates o f livestock population i s part o f the explanation for this lack o f correlationbetween agricultural growth and poverty reduction. Stagnation inthe real consumer prices o f rice and wheat (in contrast to declines in early decades) i s another major factor. Two other major factors are the changing structure o f the Pakistan economy, inparticular the declining share o f agriculture in total GDP (39 percent in 1970, but only 24 percent in2000), and the persistent highdegree o f concentration o f land (and land revenues) amongst medium and large farmers. As a result o f these structural characteristics, growth inagricultural incomes, though they lead to substantial multiplier effects on the rural non-farm economy, has a smaller impact on rural incomes (interms 69 InPakistan, increased yields o f cotton also contributed to rural income growth, particularly inthe 1990s. 85 of the percentage increase innon-agriculturalhouseholds' incomes) than agricultural growth in Pakistan had inpast decades. 8.29 Rural agricultural growthremains a major engine o f growth for both agricultural and non-agricultural rural households. However, landless agricultural laborers and non-agricultural rural poor households account for 61 percent o f the rural poor (and nearly 30 percent o f the rural population). Although agricultural growth can have a major impact on a sizeable segment o f the rural poor through combined direct and multipliereffects, it i s unlikely to be sufficient to rapidly raise incomes o f this entire segment o f the rural population. Thus, inthe absence o f a change in the structure o frural incomes andemploymentor significantgains inthe ruralnon-farm economy apart from agricultural growth-induced linkage effects, targeted interventions to agricultural laborers and the rural non-farm poor will be needed. 8.30 The Poverty Reduction Strategy Paper (PRSP) o fPakistan issued inDecember 2003 (World Bank, 2003b) recognizes this explicitly. Inaddition to advocating sustained and broad based economic growth emanating from the rural areas the strategy also emphasizes the simultaneous provision o f social and economic services and infrastructure for the poor, creating job opportunities and improving governance as essential elements. Conclusions 8.3 1 Significant structural changes have taken place inPakistan's economy over the past several decades that make factor market reforms critical to increasing agricultural productivity and raising incomes o f the rural poor. Ruralpopulationhas steadily increased, gradually reducing average farm sizes inthe face o f constraints on the overall supply o f arable land and water. The size o f the agricultural sector relative to the overall economy has steadily shrunk, as the output o f the rural non-farm economy has expanded. Rural poverty which declined steadily (as a percentage o f total rural population) inthe 1980s and early 199Os, has increased marginally in recent years. 8.32 Even inthe face o f increasingly severe constraints on land and water resources, however, agricultural growth can play a central role inincreasing rural incomes and reducingpoverty. Experience o f other countries suggests that factor market reforms, coupled with outward-oriented trade and exchange rate policy can lead to rapid agricultural growth and rural poverty reduction through diversification o f agricultural production into high-value products (e.g. fruits and vegetables) with labor-intensive production and processing. Reforms infactor markets that promote the allocation o f scarce land and water resources to their most efficient uses, and increase returns to unskilled labor, are essential for replicating this success inPakistan (see table 8.3). 8.33 Because many rural poor households lack access to land, agricultural growth alone i s not sufficient to significantly raise incomes o f a large fraction o f the rural non-farm poor, even with its substantial multiplier effects on the non-agricultural economy. Policies and programs to increase access to credit and enhance worker and entrepreneurial skills can directly support growth in the rural non-agricultural economy and further raise incomes o f the ruralnon-farm poor. Targeted interventions are also needed to reach the poorest households including micro- credit, enforcement o f laws on bonded labor, and training inmarketable skills. Inthese ways, rural factor market policy reforms have the potential to contribute to a rapid reduction inrural poverty through increases inboth agricultural and non-agricultural productivity, as well as through direct interventions at the household level. 8 b a2 a 9 a a a a a a a a a a a c, a, x P B +cd .i .i Ye .3 8a, Px 3 Y802 a, ,-c % 3 0 Y 88 Annex 1:Credit Access andAgricultural Productivity' A 1.1. This annex examines the linkbetween access to credit andgross farm productivity, as measured by per acre crop yields. We are not interestedhere inthe relationshipbetween use o f credit per se andproductivity. A farmer who does not take out a loan is not necessarily credit rationed; he may simply not need credit, and consequently may bejust as productive as a farmer who does take out a loan, The hypothesis that we are interested intesting i s whether the inability to obtain credit (or enough credit) when desired reduces a farmer's productivity vis a vis a farmer who obtains as much credit as he wants or one who does not want any credit inthe first place. Al.2. The main channel by which credit i s likely to influence farm productivity i s through input use. Pakistani farmers typically borrow at the beginning o f the agricultural season to finance the three principal planting inputs: seed, fertilizer, and tractor rental. Access to credit can thus ensure that key inputsare available insufficient quantity and ina timely fashion. Farmcredit is also used for long- term investments inland or equipment; the relatively long duration o f formal loans suggests that they are often used for such purposes. The resulting capital improvements would presumably also enhance agricultural productivity once they are put into place. We do not explore the precise channel through which credit access impactsproductivity at this stage, and it i s unclear whether productivity gains from long-term investment could ever be detected indata such as ours. A1-3. Our unitof analysis here is the plot. We have informationonthe value o f crop production on each plot duringthe kharifand rabi seasons o f 2000-01, along with input use and expenditures. The advantage o f usingplot-level data i s that one can control for a wealth o f plot characteristics, such as topography, irrigation, and soil-type, that also influence productivity. Credit access, o f course, i s defined only at the level o f the household, and for this reason it i s important to correct the standard errors for any correlation o f the errors across plots within the same household. Al.4. Inadditionto plot characteristics, alltheregressionsinclude dummy variables for eachofthe 134 villages, which should capture variation inprices as well as inthe structure o f factor and credit markets more generally. We also include a tractor ownership dummy inall the regressions to allow for the possibility o f imperfectionsinthe tractor rentalmarket. Inaddition, we include some household level characteristics that may directly affect productivity such as the education level o f the household headand the number o f adult males inthe household (table A1-1). A1.5. We use four measures o f credit access: (1) the credit access dummy takes the value 1 ifahouseholdisrationed(ineithersense)intheformalmarketand0otherwise.(This corresponds to the 3`d specification intable Al.2; (2) the credit access dummy takes the value 1ifahouseholdis rationed(ineither sense) inthe informalmarket, and 0 otherwise. (This corresponds to the 3rd specification intable A1.3; (3) the credit access dummy takes the value 1ifa household is rationedinthe formal market but is not rationedinthe informal market, and 0 otherwise. (This corresponds to the lSf specification intable Al.4; and (4) the credit access dummytakes the value 1if a householdi s rationed inboth the formal and the informal markets, and 0 otherwise. (This corresponds to the 2nd specification intable Al.4). Table This annex is taken from Jacoby and Mansuri (2004b), "Agricultural Credit and FarmPerformance: Evidence from Pakistan". 89 Al.5 presents the OLS regressions for logyields under these four specifications respectively.2 Table Al.l List of variables and descriptive statistics List of variablesanddescriptivestafistcs Descvivtion Unitsof measurament Mean Stand.Deviation Min Max landowned acres 6.09 14.91 250 dummy for landowners - 0.71 0.46 1 dummy for tractor owners - 0.08 0.27 1 numberof animals owned (not includingpoultry) - 2.84 3.57 53 highest level ofeducationofadult malesmembers years 5.69 4.98 21 numberofadultmalesinthe household persons 5.17 3.19 26 householdsize persons 8.74 4.68 42 distance from closestbank l=20h incomefromnonagriculturalstable employment Thousandsof Rupees 8.2 28.6 393.2 Figure Al.l Credit Rationinv inthe Formal Market Households Householdswith with ' outstandingloans who face Households outstanding no quantityrestrictions with b loans (139) outstanding (180) ~~ Intaking logs, we drop cases withzero values. Eventhough we lose observations (about 4 percent) inthis procedure we actually gain inprecision o f the estimates. At the same time, selection bias does not appear to be a problem, as otherwise equivalent specifications using all the data with the dependent variable defined as log(expenditures +I) give very similar results. 90 FigureA1.2 CreditRationinginthe InformalMarket Householdswith Households Householdswith outstandingloans who face with outstandingloans no credit limits (1019) Householdsthat are Householdsthat are not not current currentborrowers but are Farmers borrowersbut are not accessrationedinthe in the notaccess rationed forma[ credit sector (287) sample 1590 Households that have informalloansbut face Householdswith no accessrationed quantityrestrictions (186) outstandingloans u Householdsthat are accessrationed(98) Specification 3 Al.6. The regression results indicate that rationedhouseholds realize significantly lower yields on their plots than their credit-rationed neighbors. The size o f the effect i s roughly the same whenwe look at the two markets separately. However, as we discuss above, households can substitute one source o f credit for another-although the two sources may not beperfect substitutes. The third and fourth specifications tackle this question. A1.7. Inspecification 3, we examine the impact o frationing inthe formal market for households who can obtain all the credit they needfrom informal sources. Interestingly, the productivity effect disappears entirely. This suggeststhat the impact o f formal sector rationing i s dampened considerably inrural Pakistanbecause o f a vibrant informalmarket which absorbs mucho f the residual demand from the formal market-recall that 75% o f those rationed inthe formal market are able to obtain all the credit they want inthe informal market. We turn finally to the remaining 25% o f households who are unable to get adequate access to credit ineither market. These households constitute some 10percent o f farm households. We find that these households realize yield losses o f 23 percent on their plots as compared to their non-rationed neighbors. A1.8. One issue that we have ignored, so far, is the endogeneity o f credit access. We know from our analysis inthe previous section that access to credit, especially formal credit, is strongly dependent on land ownership. Access to credit may also be affected by unmeasured factors that influence agricultural productivity and input use directly, such as the ability o f the farmer or the fertility o f the land. The presence o f such unobservables will render the OLS estimates intable Al.5 biased. Moreover, under any reasonable modelo f credit rationing, anunobservable that raises land productivity (such as farmer ability or soil fertility) should also improve access to credit-so the OLS estimates o f the credit access coefficients are likely to be biasedupwards. We attempt to correct for this, by usingan IV strategy. 91 Table A1.2 ProbitOn Access To FormalCredit Explanatoy Variables (I) (2) (3) I Log Land Owned' 0.19 0.14 0.12 (0.023)***(0.013)***(0.013)** Tractor Owner 0.41 0.52 0.38 (0.148)*** (0.17)*** (0.15)** lLog of Livestock Value -0.03 -0.001 -0.00 (0.012)** (0.009) (0.009) Highest Level o f Education 0.02 0.03 0.02 (0.011)* (0.008)***(0.008)** Number o f Adults -0.01 -0.03 -0.02 (0.022) (0.18) (0.017) Household Size 0.01 0.012 0.011 (0.014) (0.02) (0.02) Distance From Bank 0.056 -0.04 -0.08 (0.041) (0.031) (0.03)** Dummy for Stable Income -0.35 -0.043 -0.00 (0.114)*** -0.084 -0.005 Log o f Durable Assets 0.18 0.01 0.02 (.007)** (.005) (.005) Observations 1574 1574 1574 Absolute value of z statistics inparentheses * significant at 10%; ** significant at 5% ;***significant at 1% 'log(acres of land owned + .01) 'log(non ag income+l) All specifications contain regional dummies. The dummy for Balochistanis dropped Dependentvariable: (1) 1ifsome formal credit, 0 ifno formal credit;(2) 1ifnot access rationed, 0 ifaccess rationed; (3) ifneitheraccessnorquantityrationed,0ifrationedaccessorquantityrationed. 92 Table A1.3 Probit On Access To Informal Credit Explanatory Variables (1) ' (2) (3) L o g Land Owned' -.26 .o1 .02 (0.014)" (0.020) (0.O14) Tractor Owner -.27 -.02 -.23 (0.138)* (0.214) (0.15) L o g o f Livestock Value -.01 -.04 -.02 (0.009)** (0.016)*** (0.o1O)* * Highest Level o f Education -.02 .oo .01 (0.009)*** (0.o12) (0.009) Number of Adults -.02 -.03 -.03 (0.019) (0.03) (0.019) Household Size -.oo .oo .01 (0.012) (0.o18) (0.01) Distance From Bank .04 .06 -.02 (0.034) (0.049) (0.035) Dummyfor Stable Income -.13 -0.09 -.03 (0.085)*** (0.131) (0.090) L o g of Durable Assets .003 0.03 0.02 (.006)*** (.008)*** (.006)*** Observations 1574 1574 1574 ~~ Absolute value of z statistics inparentheses * significant at 10%;** significant at 5% ; *** significant at 1% of land owned + .01) regional dummies. The dummy for Balochistani s dropped ependent variable: (1) 1 if some informal credit, 0 ifno informal credit;(2) 1 ifnot access rationedinthe informal (3) 1 ifneither access nor quantity rationed inthe informal market ,0 ifrationedaccesi 93 TableA1.4 Probit On Access To FormalandInformalCredit Explanatory Variables (1) (2) Log LandOwned' -.09 -.06 (0.0 13)*** (0.0 17)""" ITractor Owner -.40 -.11 (0.161)** (0.217) Log of Livestock Value -0.01 .03 (0.0009) (0.013)** HighestLevel of Education -.14 -.02 (0.008)* (0.011) Number ofAdults 0.02 0.01 (0.018) (0.024) Household Size -0.01 0.01 (0.012) (0.016) Distance From Bank 0.05" 0.07 (0.032) (0.042) Stable Income -0.06 0.12 (0.083) (0.109) Log of Durable Assets 0.00 -0.01 (.005) (-007) Observations 1574 1574 Absolute value of z statistics inparentheses * ***significant at 1% 'log(acres of land owned + .Ol) significant at 10%; ** significant at 5% ; 'log(non ag income+ 1) 3 All specifications containregional dummies.The dummy for Balochistani s dropped Dependentvariable: (1) 1ifrationedinformal butnot ininformal, 0 otherwise;(2) 1ifrationed informal andininformal, 0 otherwise. 94 TableA1.5 CreditAccess and FarmProductivity Rationed in Rationed in Rationed in Rationed Credit Access - 0.09 ** (0.04) (0.05) (0.038) (0.074) Tractor ownership 0.07 0.07 0.07 0.07 Educationo f head 0.01** 0.01"" 0.01** 0.01* (0.006) (0.006) (0.006) (0.006) No. o f adult males 0.02"" 0.02"" 0.02** 0.01** (0.006) (0.006) (0.006) (0.006) Log o fplot area -0.12*** -0.12" ** -0.12*** -0.12*** (0.02) (0.02) (0.02) (0.02) L o g value per acre 0.13*** 0.13*** 0.12*** 0.13*** (0.03) (0.03) (0.03) (0.03) Plot outside village -0.04 -0.04 -0.04 -0.04 Flat topography 0.10* 0.11* 0.11" 0.11* (0.058) (0.058) (0.059) (0.058) Perennial canal 0.16** 0.17** 0.17** 0.15* (0.077) (0.077) (0.076) (0.078) Non-perennial canal 0.12 0.12 0.12 0.10 Good groundwater -0.05 -0.05 -0.05 -0.05 Brackish groundwater 0.18"" 0.19** 0.19** 0.19** (0.092) (0.091) (0.092) (0.091) V.brackish groundwater -0.10 -0.11 -0.11 -0.10 Sandy soil -0.05 -0.05 -0.05 -0.05 Maira soil 0.03 0.03 0.03 0.04 Chikni soil 0.01 0.01 0.01 0.01 Cotton area 0.49** 0.49"" (0.197) 0.48** 0.50** (0.197) (0.198) (0.197) Sugarcane area 0.26 0.28 0.28 0.27 Wheat area -0.22 -0.20 -0.21 -0.21 Rice area -0.11 -0.09 -0.10 -0.10 No. Observations 2030 (40%) 2030(18%) 2030(30%) 2030(10%) Adj. R2 .JI r? .57 .56 .57 r'otes: Standard errors, adjusted for clustering by household, are inparentheses (". **, and *** denote significance at the 0.1, 0.05, and 0.01 levels, respectively). All regressions include village fixed effects. Al.9. The results are reportedintable AI.6.We report the results for the two specifications o fmain interest. The first and the fourth. Our instrument set includes a number o f variables that capture household wealth (such as land holdings, the proportion o f owned land for which the householdhas a legal title, the value of livestock, and the household's annual expenditures) and demographics (whether the household has a female head and whether the headwas born inthe village). Inprincipal, ifcreditmarketsfunctionwell,householdwealthshouldhavenoimpactonproductivity.Soonecan reasonably argue that to the extent that wealth matters it i s due to imperfections inthe credit market. If imperfectionsexist,however,householdwealthmayalsoreflectunobserveddifferencesin such skillor ability that also impactproductivity. To control for this, we includethe household head's level of education inall o f our specifications. It i s certainly unlikely that any unobserved ability that i s reflected inhousehold wealth shouldnot be captured inthe household's observed level o f education. As our results in section 4 have already indicated, we are better able to identify the determinants o f 95 access to credit inthe formal market. Thus our IV strategy also works better when we look at rationing inthe formal market. The F test for excluded instruments i s 26 inthe first case, but only 5 in the second. Inboth cases, however our IV estimates have the right sign and are much larger. Standard errors also increase inboth cases-as expected. However, the yield loss for households who are rationed i s now 26 percent and significant at the 95% level. A1,lo. While we do not explore the precise channel throughwhich accessto creditraises productivity inthis paper. One such channel may be that it facilitates long-term investments inland and equipment. Giventhe `time-to-build' for some o f these investments,however, it may be difficult to detect the long-run productivity effects o f current credit access status inour data even ifthey are important. Nevertheless, to give this hypothesis a `fair' test, we re-runthe log yield regression dropping all variables that are potentially investments or mightbe influencedby past investments; in particular, the canal and tube well irrigation variables, the tractor ownership dummy, and the log per acre value o f land. This specification allows the credit access variable to pick up the full yield impact of any recently completed agricultural investment. The resulting coefficient on credit access i s somewhat larger (-.lo inthe 1'` specification and -.24 inthe second) and the SE i s smaller inboth cases, This increases the yield loss to 11percent for farmers rationed inthe formal market, and to 26 percent for those rationedinbothmarkets. While this suggests a credit-investment-productivity channel, it i s by no means conclusive. A more complete analysis would require data on past investments along with a credit history to account for the time lag between the taking on o f debt and the completion o f the intended project. A1.l. Finally, giventhe positiverelationshipbetweencredit access and land ownership, it is 1 possible that access to credit also affects the amount o f land that households can operate-particularly iflandleasingmarketsarefunctioningreasonablywell.Specifically, itispossiblethatbigger landowners have more cash on hand to finance production inputs, and therefore can operate more land, Figure Al.3 presents results for a bivariate regression (by province) o f land operated on land owned. Indeed, the results are quite suggestive o f a strong positive relationshipbetween owned and operated holding^.^ Thus, it i s likely that inaddition to any impact o f credit access onper acre land productivity, it may also influence the scale o f farming, which ultimately affects householdwealth. This positive relationship between operated and owned area also holds up strongly ina linear regression with village fixed effects. 96 Table A1.6 Credit Access and FarmProductivity (TVEstimates) IVEstimates Rationed in Rationed in Formal Both Yield (') Credit Access -Yield (a) 0.26 ** -0.45 (0.126) (0.479) Tractor ownership 0.05 0.06 Education of head 0.01** 0.01"" (0.004) (0.004) No. of adult males 0.02""" 0.01** (0.005) (0.005) Log of plot area -0.12* ** - O . l l * * * (0.02) (0.02) Logvalue per acre 0.12"* 0.11""" (0.03) (0.03) Plot outside village -0.03 -0.04 Flat topography 0.12** 0.14** (0.047) (0.047) Perennial canal 0.16* 0.16* (0.085) (0,091) Non-perennial canal 0.09 0.09 Good groundwater -0.05 -0.05 Brackish groundwater 0.16" 0.15* (0.085) (0.o89) V.brackish groundwater -0.08 -0.08 Sandy soil -0.05 -0.06 Maira soil 0.02 0.04 Chikni soil 0.5 1*** 0.02 0.02 Cotton area 0.52*** (0.167) (0.167) Sugarcane area 0.26* 0.29** (0.139) (0.141) Wheat area -0.21 * -0.18 (0.122) Rice area -0.09 -0.05 No. Observations 1961(40%) 1961(10%) gtes: Standard errors are inparentheses (*, **, and *** denote significance at the 0.1, 0.05, and 0.01 levels, respectively). All regressions include village fixed effects. a) Sig. test o f excluded IVs infirst stage credit access equation: F( 6, 1885) = 26.19 b) Sig. test o f excluded IVs infirst stage credit access equation: F( 6, 1885) = 5.2 97 FipureA1.3 Operatedvs. OwnedArea by Province Operated vs. Own II I I I I 0 50 acres100 150 200 owned -.-.-.- Punjab ----. Sind _._.. NWFP -Balochistan I 98 Annex 2: Determinantsof FarmRevenueIncome in Pakistan4 RevenueFunctionsfor PakistaniFarmHouseholds A2.1. The emphasis o fthis analysis is on the ability o fhouseholds to generate net income from agricultural operations i.e. livestock and farming activities; that is, the focus i s on the net returns to householdfarm-related assets, which include human and physical capital associated with the farm operations. Household income as defined here i s farm-related income and excludes labor income from off-farm employment as well as pensions and transfers. For landowners, returns may be generated from their own cultivation o f their land, and from income derived from renting or inthe form o f share-cropping arrangements. For tenant households (both cash leasing and sharecropping), income i s generatedby their share inthe farm operation, which yields returns to their on-farm labor and capital (off-farm income and remittances are excluded). For "landless" farm-operating households, income corresponds to that generated with own cattle andpoultry production on common-property and owned non-cultivatedland; and that generated by leasing out all o f their agricultural land. A2.2. Net farm-relatedhousehold income i s defined as gross farm revenues less the cost o fpurchased inputs.To analyze the determinants o fnet household income, we make use o f the restricted revenuefunction. Algebraically, the gross revenue from agricultural production activity can be represented as usingtwo inputs: a short-term fixed input,z ,and a variable input, x : The restricted revenue function i s definedas gross revenues less purchased (variable) inputs,x : R(p;z) =R(p;x,z)-wx (2) Assumingthat the household selects the variable inputsto maximize short-term net revenue, one can estimate the relationship between observed net revenues and levels o f prices and fixed inputs. In the case ofPakistan, netrevenue derives from many crops, andwe assume that farmers choose an optimal combination o f outputs. Crops include field crops (cotton, sugar, wheat and rice), other crops (including other cash crops, h i t s ) , and livestock (which includes poultry and dairy revenues). Quasi-fixed inputs consist o f land, canal water, non-land capital, education, own labor, and agricultural credit.6To calculate net revenues, we subtract from gross revenues the variable inputcosts associated with hiredlabor, seeds, fertilizers (organic and chemical), pesticides, purchased tube-well water, and machinery and equipment rental; and addsubtract the revenue/cost o f land leasing arrangements. This annex is taken from Gustavo Anriquez and Albert0 ValdCs (2004), "Determinants o f FarmRevenue in Pakistan", BackgroundPaper for the Pakistan RuralFactor Markets Study. An interesting property o fthe restrictedrevenue function(2) is that it is no longer a function o fthe variable input.Partially differentiating withrespect to x we obtain: at?(p; x,z) I ax = dR(p;x,z) I ax -w = 0. Ifthe marginal retums of the variable input, dR(p;x, z) / ax, are higher than the cost of acquiring it, then the farmer will employ more o fthat input, untilthe retums on the last unit employed equalize the cost o fusingit, w.As per a regular revenue function, k ( p ;z ) i s positively homogenous o f degree 1, continuous, increasing and convex inoutput pricesp and i s continuous, increasing and concave infixed inputs z. A more thorough discussion on the possible endogeneity o f quasi-fixed inputsi s given inAppendix D. 99 A2.3. Iti s important to notethat for Pakistan canal water canbe considered a quasi-fixed input, because canal water i s tied to landholdings and the market for water i s limited. Normally, farms are endowed with one canal turn per week per acre o f land ina rotational system, called warabandi, with turns usually lasting20 minutes. These water endowments depend on landholdings, not crop area, and are independent of any other demand consideration, such as the type o f crop under cultivation, terrain seepage, or seasonality. Farmers make some adjustments by trading their canal turnsto adjust for peak crop demand, butthese trades usually occur betweenneighboringplots. Sales o f canal turns are uncommon, and inany case not necessarily legal, a fact which may lead to some underreporting. table A2.1 reports canal water usage and market transactions inthe different districts o frural Pakistan. Except for the districts o f Badin, N.Shah and M.P. Khas, there are very few market transactions o f canal water. A2.4. Land operated, once the farm i s inoperation, i s a fixed inputinthe short term, i.e. the agricultural season. Because o fthe abundance o f share-cropping and landrental agreements we purposely differentiate land owned from land operated. The land operated includes owned land, as well as land operated through a fixed rental or sharecropping agreement. 'The cost o f usage o fnot- owned operated land i s discounted for the households. That i s for fixed rental arrangements we subtract the cost o f rentingland. While for tenant sharecroppers we account only their agreed upon share o fthe value of output (revenues); inother words we are implicitly discounting their cost o f use o f land. The returns to household assets include revenues generated by the household's operated land use, irrespective o fthe land's ownership. The returns to land specifically are distributed between the owner and the operator, who could be the same agent. Revenues and input costs and usage are calculated at the operated farm level. For landlords with tenants operated land excludes that which i s leased-out, but we include the revenues from those lease agreements that i s fixed rent paid incash or inhnd, and the corresponding net value o f output from their share-cropping contracts . Table A2.1 Canal Water Usage and Transactions by District I ,verage Hours Y Farm) Given or Sold or Surveyed Farms with District Endowed Received Purchased Canal Water Faisalabad 525 6.30 1.99 56 Attock 0 0 0 0 Hafizabad 24 0.11 0.03 21 Vehari 37 0.03 1.75 52 Muzafar Garh 16 0.04 0.02 21 Bahawalpur 16 1.10 0.04 54 Badin 382 19.31 9.45 80 NShah 65 1.78 25.56 44 M.P.Khas 73 0.00 4.57 46 Larkana 143 8.00 0 6 ' Dir 0 0 0 0 Mardan 391 2.00 0.00 36 Total 204 5.19 5.52 416 Source: PlDE2001 The degree o f involvement o f the landlord ina sharecropping operation varies tkom an absentee landlord to a landlordthat will make all the decisions regarding the agricultural operation treating the sharecropper almost like a wage laborer paid inkind. 100 A2.5. Agricultural land distribution i s notably unequal inrural Pakistan. Inthe case o f east and south Asia, the average Gini for operationalholdings o f agricultural landranges between 0.52 and 0.56. Pakistan has Gini coefficients similar to the highest regional average, this honor belonging to LatinAmerica. Table A2.2 shows the disparitybetween the mean and the median of farm sizes (owned and operational), reflecting the skewness of the distribution. Inthe case of owned land, for example, the mean is 10.17 acres incontrast to the median o f 3 acres. Not surprisingly, the GINI i s high (0.82). Table A2.2 also shows that land leasing mitigates the inequality inlandownership, lowering the relative disparity between the mean and the median, the coefficient of variability andthe GINI index for operational landholdings. Table A2.2. The distributionof farm size and landownership (Acres). Mean Median Coefficient of Variability GINI 1Land Ownership 10.17 3 4.62 0.824 Leased(0ut) Land 1 1.93 3.71 I 4.40 I 0.894 1 A2.6. Other important determinants o fhousehold revenue that can be treated as fixed inputsin the short term include household head education, the household size (a determinant o f own family labor allocation) and capital stock. Agricultural credit (formal and informal) may be thought o f as a quasi-fixed input, fixed within a givenyear, although adjustable over a longer term. Treating credit as a separate fixed input serves to measure the marginal effect on profitability o f the distribution o f financial funds across farms. The restrictedrevenue function we proceed to estimate may transparently be described by: R(p;A,X ,E,K,W ,D;L)=p F(A,X ,E,K,W,D)- C(purchased inputs) + a r(L - A) . Where F(.) i s the productionvalue from the farming operation- income from farming directly. A i s land utilized inthe farm operation, r i s the rental rate per acre, L i s total land owned, and r(L-A) i s earnings from leasing out or costs from leasing in land used inthe farm operation. C(.)represents the costs of all purchasedvariable inputs.The quasi-fixed inputsthat determine farm value are: X, which i s family labor; E i s education, K capital, W i s water, and D i s credit. Costs o f purchased inputsandproductivity could vary by district, and dummy variables are included inthe estimation to account for such differences. A2.7. The reader shouldnote what income sources are excluded inthis measure o f income from farming: off-farm wage and non-farm enterprise income and income frompensions, government transfers and remittances. This would affect the interpretation o fthe results, specifically the effects on farm income due to changes inhousehold labor (reported here interms o f elasticities). The estimated marginal effect o f household labor, X, on farm income shouldreflect, inequilibrium, the opportunity costs o f working off the farm, which includes income money earnings net o f transactions costs (likely non-trivial). This is in contrast to the land operated input infarm production, the (money component) opportunity cost o f which i s included explicitly inthe expression above; and thus the marginal effect o f operated land on farm income should be, in equilibrium, equal to zero. Ifthe elasticity o f farm income with respect to land i s not equal to zero, 101 this would suggest that there is a sub-optimal use o fthe landinput,whichwouldbe due to some constraints not incorporated inthe analysis. A2.8. The estimated coefficients for operated land, A, should leadto marginal value product equaling the rental rate, -= dR 13, +pAxX+etc. dF dA = dA -r=0,ifthereareno ~ constraints on land trades, the elasticity o f net revenues with respect to the operational size of the farm should be equal to zero. Inthe case o f land owned, L, the marginal effect on revenues should be equal to the land rental rate -= dR pL+pLxX+etc. dL =r >0;and, ifthere are no non-rental benefits or costs to land ownership not accounted for inthe revenue function and without constraints on land trades, the elasticity should be equal to the rental income relative to total net revenues - -= -2 0, positive but less than 1.There is, dR L rL dL R R however, a possibility o f a negative elasticity with respect to land owned ifthe owner perceives other benefits to owning more land, not included inthe revenue variable, and there are constraints on rentingout land. Such additional benefits might arise from expected land appreciation, land being a means o f wealth accumulation (savings), easier credit terms, and non-pecuniary returns to social status. But constraints on landtrades would be necessary for the farmer not to rent out an additional acre o f land at the rental rate. Such forgone income could be explained by the perceivedrisk o f expropriation o f land, and search costs o f findingtrustworthy renters. ECONOMETRIC SPECIFICATION A2.9. To estimate the restrictedrevenue function, we choose a generalized quadratic specification. This flexible functional form does not impose any restriction on the underlyingtechnology interms of elasticity o f substitution, and, inprinciple, all theoretical properties can be tested.8 The underlying restricted revenue function can be approximated linearly: where indexes i andj refer to output prices p= {1.field crops, 2. livestock, and 3. other crops} and indexes k, 1refer to quasi fixed inputsZ={1. Capital, 2. Education, 3. Household Size, 4. Land, 5. Water, and 6. Credit}. To reduce the number o f estimated parameters, we use the property o f linear homogeneity inprices o f the function and divide by the price of other crops: * This functional form also has the advantage that it does not make logarithmic transformations that cannot be applied to variables with zeros like in our data set. 102 wherepl i s the price o f other crops, and cl and j 2are price o f field crops divided by price o f other crops and price o f livestock divided by price o f other crops, respe~tively.~District dummies capture region-specific differences between farms, such as land fertility and other unobservable farm characteristics correlated with district inwhich the household lives. Additionally, we include type-of-tenancy dummies (i.e. owner operated, sharecropping tenant operated, etc.) to test for tenancy specific differences inrevenues. The results are presentedinterms o f elasticities. As an example, consider the elasticity o fthe revenue function withrespect to education: where the bar over the variable indicates sample means, and the fijand Sij are the regression n estimated coefficients o f equation (4). Ifinstead o f sample means we use group means o f the variable, we can approximate the elasticities for the different farm g r o ~ p s . ' ~ MAINFINDINGS A2.10. The detailed results o f estimating equation (4) are presented intable A2.3.In general, the results are satisfactory, with a reasonable fit (R2= 66%)." The elasticities of the revenue function with respect to the quasi-fixed inputs are reported intable A2.4. Below the results are discussed interms o f individual factors o f production. HUMAN CAPITAL A2.11. Usingthe entire sample, we find that the input with the highest elasticity i s human capital (education), with a value o f 0.8. l2 This estimate implies that ifthe education level of the household head were to rise by 10% then overall net revenues would rise by about 8%. Using the average education level of 6.43 years, another year of education would represent an increase o f approximately 16% inthe education level, which would translate into a 13% increase innet W e assume that price variability i s the result o f exogenous conditions (such as distance to markets) and not correlated with farm characteristics or product quality. loFurthermore, to test ifthe elasticity i s significantly different from zero, we divide the elasticity by its standard error, which i s equal to the square root of: k=l i=O j t i \ i=O i=O k=l All the first order and second order theoretical curvature properties o f a revenue function are met. Hampering the precisiono f the estimates o f returns to education, and for that matter the reliability o f the education figures inTable 1, is the unsatisfactory level o f response to the question o f education level o f the head of the household inthe survey. W e opted to use average education level at the district level for these missingvalues, which allows us to use these observations, butreduces the reliability o fthe results. 103 revenues. Interms o f present value, usinga 10% discount rate, the benefit o f an additional year o f education would result in an increase inwealth that i s 130% o f yearly revenues. More realistically, for younger generations, for which the opportunity cost o f one more year o f schooling i s the foregone returns associated with family labor, this elasticity suggests notably highpotential rates o f returns in farming, perhaps even higher ifthey migrate to off-farm employment. VARIABLE CO-EFFICIENT STANDARD ERROR 104 VARIABLE CO-EFFICIENT STANDARDERROR Land Owned -6.980 4.664 Land Operated 0.719 7.936 Land Operated x Land Operated 0.003 0.039 105 Table A2.4 Elasticities of householdfarm revenuewith respectto Quasi-fixedfactors Landless Medium Large TotalSample Workers Small Farms Farms Farms PO-4Acres) p4 -20Acres) f>20Acres) Capital 0.267""" 0.220""" 0.133"" 0.345""" 0.268" (0.104) (0.089) (0.055) 1.124"" * (0.139) 0.845"** (0.151) Education 0.814""" 0.873" 0.242 (0.236) (0.458) (0.225) (0.331) (0.250) OwnLabor 0.320 1.200""" 0.716""" 0.047 0.204 (0.206) (0.430) (0.233) (0.271) (0.313) Land -0.214 -0.247 -0.070 -0.123 -0.632 (Owned) (0.131) (0.158) (0.046) (0.076) (0.432) Land 0.325"" 0.052 0.373""' 1.135""" (Operated) (0.135) N/A (0.O52) (0.134) (0.420) Water 0.082"" 0,071"" 0.077"" 0.110 (0.036) N/A (0.033) (0.036) (0.219) Credit -0.014 -0.018"" -0.036" -0.042 0.204 (informal) (0.080) (0.008) (0.021) (0.165) (0.139) Note: Standard errors inparenthesis. Implies significant at the 10% level, * ** implies significant at the 5% level, and *** implies significance at the 1%level. A2.12. But it is notable that the elasticity with respect to education i s much smaller for small farms (0.2), and greatest for medium sized farms (1.1).This suggests an important and positive interactionbetweenhuman capital and operational scale. Constraints to enlarging farm size would be a bottleneck to improving farm-generated incomes via improvements ineducational levels, which inturn would tend to reduce the demand by small farm households for education o f their children. A factor that might counteract the negative effect o f scale i s the opportunity to make use o fhigher educationlevels inthe generation o f off-farm wages and income from non-farm enterprises. Supporting this possibility i s the muchhigher elasticity (0.873 intable A2.4) o f revenuewith respect to education that i s estimated for householdwith no operational land (some o f which are landless). FARMOPERATIONSIZE A2.13, The elasticity o f operated farm size is increasing with farm size. Because net revenues account for the cost o fthe use o f land (and so the expected elasticity with respect to land inan efficiently working land market shouldbe zero) the results suggest a greater deviation from the efficient allocation o f land inthe case of larger farms. The elasticity of revenues with respect to operational size for small farm i s 0.05, while medium farms have an elasticity o f 0.37 and large farms of 1.13.Ina well-functioning market, a priori one would expect that increasing the farmer's and sharecropper's operational land by one acre (by leasing in, not getting an acre free) would increase income by its marginal value product and decrease income by its rental rate. It would not affect income directly attributable to farming - because the marginal product of land should equal 106 the rental rate. Therefore, the measuredelasticity should be zero, and that i s what the elasticity estimate suggests for small farmers. This i s gratifying result, implyingthat, on average, small farmers are rationally allocating land, given the levels o f other variable inputs. A2.14. By contrast, the positive and higher elasticities for medium and larger farms suggest that for those farms the marginalbenefits from scale expansion are higher than the observed marginal cost (the rental rate), and those farms should expand. This raises the question o f what mightexplain this apparent maladjustment of land use inmedium and larger farms, incontrast to the case o f smaller farms. Ina well-functioning land market, the net revenue elasticity with respect to land shouldbe zero for all farm sizes, although this does not implythat rental rates shouldbe the same across farm sizes. Ifthe elasticities are not zero, hypothetically if small farmers are side by side with larger farmers, the higher elasticity for larger farmers would imply an incentive for larger farms to expand, biddingup rentalrates and inducing a movement o f land use from the small to the larger operations. A2.15. For areas where small and larger farmers are not side by side, the differences inelasticities by farm size could also be due to differences inoutput mix (and different marginal revenue curves with respect to land) and inthe efficiency inthe functioning o f landmarkets. Ifsmall farmers are concentrated inareas where their smaller size i s more appropriate for the most profitable product mix for those areas (likely associated withhigher per-acre valuedcrops, for example horticulture, being produced closer to towns), landtransfers would be between small farmers and not between small and larger operations outside those areas. Land markets mightbe better developed inthose areas where small farmers concentrate. Finally, rentalrates might differ between smaller and larger operations, but this should not bereflectedin different elasticities, ifmarkets were functioning to set rental rates to equilibrate across farms the marginal value product o f land. Infact, average observed land rental rates per acre for small farms (15,400 rupees) are 2.2 times as great as rental rates for medium farms (4,800 r~pees).'~ Why mightthisbe so?The concentration o f small farms in areas producing high-valued (per-acre) products would explain the differences inrental rates, but not the positive elasticities for larger farmers. Consistent both with large rentalrate difference and with non-zero revenue-land elasticities is the hypothesis that land owners set rentalrates for larger operations below those for smaller farmers, and below the larger renter's short-run marginal product. This would require a rationing o f total land rented out at that lower rate. Land owners mightration landinthis way ina similar manner as employers offer so-called efficiency wages, extractingrenter loyalty and more efficient land and water use, and reducing supervision costs. Efficiency rentalrates would also tend to induce a larger pool o f potential renters from which to select, including tenants with more working capital. A2.16. Insummary, there are two hypothesis (not necessarily exclusive) for these results regarding operation size: the first i s that small farmers are geographically concentrated in areas with higher-valued products andwell-functioning landmarkets; and the second i s that landlords ration landto mediumand large tenants. LAND OWNERSHIP A2.17. The marginal effect o f owned land on householdnet revenues deriving from farm production and landrental (positive for land owners, negative for renters) appears to be negative although not statistically significant. This result, which might appear surprising at first given that the marginal value product o f owned land ought to equal the rentalrate, suggests that owning land (incontrast to the landinputinfarmproduction) has avalue that is not fully accountedbythe l3 largefarms (morethan For 20 acres) there are very few observations on rental rates. 107 agricultural or rental income it generates. That i s ifland has a use beyond its employment as crop production input (say as an instrument for access to credit and markets, as a store o f wealth, as a speculative investment, and as a source o f social power), then agriculturalrevenues would not fully account the returns to land. Landowners would perceive a marginal utility associated with an additional owned acre that i s greater than the marginalrevenues it generates incultivation or inthe rental market. And owners would have the incentive to expand ownership size beyond the optimal point based only on yearly agricultural and rental income. A2.18. Butthe possibilityo frentingout land should, ina perfectly working market, offer constant returns to scale interms o fnet revenue generation. Logically, therefore, some constraints on land rental (interms o f the optimization o f the farm revenues we are measuring) must exist for the farmer simply not to rent out an additional acre at the rental rate. As inthe case o fthe results for operational size above, these results for scale o f ownership suggest some imperfections inthe land rental market. The owner might not rent out due to a perceived risk o f expropriation o f land, and/or to the search costs o f finding trustworthy renters. For medium and large farms, the estimates suggestthat tenants would be willing to pay more than the observed rental rate, while land owners would be willing to accept less. Other benefits and costs are involved inthe land market that are not reflected bythe observed rentalrate.14 FAMILYSIZE A2.19. Bothusingall observations, and inthe case ofthe specific estimates for mediumand large farms, the effect on household revenue o f family size i s not statistically different from zero. To the extent that family size i s a proxy for family labor employed inthe farm operation, this result suggests that own labor i s employed up to the point where marginal gross revenue i s zero. But the reader should note that the absence o f data on the quantities o f family labor effectively employed in the farm operation i s a major weakness. (Quantities o fhired labor are also unavailable. but the cost o f hired labor i s subtracted to obtain net householdrevenues). Ifall relevant inputs are accounted for inthe net revenue measure, and iffamily size is a good proxy for family labor employed in production, then the result suggests a near-zero opportunity cost o f family labor inmedium and large farms; but inthe case o f small farms the revenue elasticity with respect to family size i s relatively muchhigher and significantly different from zero, suggesting a non-zero, positive opportunity cost of family labor. More likely, however, for medium and large farms, family size i s weakly correlated with family labor actually employed inproduction; that i s for smaller farms, family size i s a better indicator o f household agricultural labor actually applied than for larger farms. An increase inthe number o f family members for larger farms i s perhaps directedto leisure or to non-farm-productionwork (either non-farm-production work or off-farm work, the revenues fi-om which are excluded from the net revenue mea~ure).'~ l4 Note that neitherthe elasticity o f revenues with respect to owned landnor the elasticity with respect to operational land are measuring the marginal products o f land incultivation, as would be obtained from a production function. Using net revenues at the level o f individual plot, another study by the World Bank (2002) based on a preliminary version o f the same data set, found decreasingreturns with respect to farm operational size. '' Furthermore, in the case ofthese Pakistani data, family size is positively correlated with wealth and wealth with farm size. The dependency ratio (family dependents relative to family workers) will likely increase with family size. T o the extent that family size i s positively correlated with both children and aged grandparents (and other relatives), there would be a lower correlationbetween family size and the family labor input effectively employed o n the farm. 108 A2.20. By contrast, inthe case of small farms and landless households, the estimates of the household income elasticity with respect to of family size labor are positive and significant. The effect o f family labor is particularly high for the landless group. It i s highly likely that such families have less opportunity o f employing non-family labor and less income to afford leisure. For the small farmer the positive elasticity i s also likely related to the discrete nature of labor allocation associated with minimal requirements o f family labor to even maintain production and with an absence of a market for small amounts o f off-farm labor.16 SURFACEWATER A2.21, Turningto the water input, the revenue elasticity with respect to surface water is increasing with farm size, althoughthe estimates are significant only for small and mediumsized farms. There i s no out-of-pocket expense to the use o f surface water, and farmers are endowed with a given number o f canal turns per week, which depends on farm size. Therefore the positive and significant elasticities for small and medium size farms can be directly interpreted as reflecting the marginal value o frelaxing the constraint to canal water use. This result supports the idea that larger farms face less constraint to water use; and it could be socially efficient to induce transactions that result inwater transfers from larger farms to smaller. More developed water markets would facilitate such transactions. A2.22. Our analysis provides an idea to the marginalreturns to assets, but it i s only a snapshot in time, Ali and Byerlee (2000) analyze the productivity o f assets through time with a long time series data from the Punjab region. They find that there has been a measurable deterioration both inwater and land resources. Organic matter has been decreasing at 2.3% annual rate, at the same time that sodicity o f the land has also been increasing. The quality o f water has also been inthe decline with increasing carbonate residuals intube well water. These observations explain their alarming estimation o f a negative growth o f the productivity o f water. Further evidence i s provided by World Bank (1997), that estimates that salinity causes a 25% reduction inthe yield o f major crops, and as much as 40 to 60% inSindh. Furthermore, water logging i s an increasing hindrance to agricultural growth inthe Indusbasin. This study also highlightsa deteriorating infrastructure, increasing canal seepage (decrease indelivery efficiency and results many times inincreased water logging), and reduction in storage capacity due to sedimentation. A2.23. What the Ali and Byerlee study suggests is that the marginal physical product o f water could be low and perhaps declining over time. This declining trend could lower the growth rate of the sector, although this could be partially compensated by increasingother factors or improvements incrop varieties. Butthe returns to these other improvements i s lower. Under the current circumstances, o f deteriorating infrastructure and quality o f the resource, the social costs o f inefficiencies inthe use o f surface water are high.The question is, to what extent does the absence o f a working water market and the mismanagement o f the irrigation system exacerbate - or even contribute to create -these inefficiencies? A2.24. The deterioration inthe "resource" factors can be partially reversed, but this requires public and private investments, and institutional changes. Fromthe perspective o f the farmer, the creation of a surface water market, which would make the allocation o f the resource more efficient and less constraining, would also create incentives for on-farm investments. Investments such as l6 The small farm household's immobility for part-time employment off-farm couldbe influenced by lack o f transport and communicationfacility to find and travel to part-time jobs beyond their immediate neighbors. 109 conservation measures, precision land leveling, zero-tillage, bed furrowing, and improved irrigation techniques, could all improve water efficiency. A2.25. Groundwater markets, on the other hand, which are thought to be monopolized by the large landowners with access to the water tables and the capital to fundtube wells, are also subject to static and inter-temporal inefficiencies. The static inefficiency i s generated by the monopoly power o f these colorfully labeled "waterlords". However, Jacoby et al. (200 1) estimate that the social losses of these monopolies are rather low." The inter-temporalinefficienciesare due to the open access nature of the groundwater resource. Tube well owners only consider their private costs in their decision of extracting groundwater; this decision, however, does not incorporate negative externalities, such as lowering the water table and thus raising the cost o f extraction to all water users. The increase inthe use of tube-wells will likely contribute to further over-extraction, defined interms of social costs. This is aproblemthat private agents are unlikely to resolve individually, butrequires coordination, which inmany countries is provided by a government agency to enforce conditions for the expansion o f tube well extraction. These institutional issues are discussed inthe chapter on water. CREDIT A2.26. InPakistan, previous authors have noted that small farmers have limited access to institutional credit, relying on non-institutional sources; larger operations are more favored by formal credit lenders (Malik, 2003b). Farmers who are poorer and small rely primarily on informal credit markets. As shown intable A2.4, the elasticity with respect to informal credit i s insignificantly different from zero for the entire sample and negative (and statistically significant) for small farms. The cost o f credit i s not explicitly accounted for inthe net revenue measure, but to the extent that informal credit costs are reflected inpurchased input costs, the low elasticity o f informal credit to recipients implies less significant constraints to credit use. At the margin, farmers on average have adjusted informal credit use such that the addition to gross revenues fiom an increase ininformal credit i s offset by the addition to variable costs o f purchased inputs; and thus an increase ininformal credit would produce little interms o f additionalnet revenues, and one would expect an elasticity o f zero. Inthe case of small farmers, the negative and statistically significant elasticity may imply over-borrowing, with negative impacts on production value, which i s difficult to justify. Or more likely the result implies an endogeneity inthe credit variable, where ceteris paribus less productive farmers have less access to formal credit and substitute informal credit. The direction o f causality may run fiom production revenues to borrowing ininformal markets. A2.27. Intable A2.5 we compare this base result usinginformal credit to estimates usingboth formal credit alone and the sum o f formal and informal credit. One should note first that the estimated elasticities for the other inputs, usingthe alternative indicators o f credit use, are not statistically different from the base case estimates, which indicates some robustness inthe estimated elasticities. The elasticity withrespect to formal credit i s not statistically significant, but using the total credit available yields a positive and statistically significant elasticity (0.059). It i s worth noting that formal credit i s a larger source o f fundingo f agricultural borrowing. The insignificance o f the two estimated elasticities (for the entire sample) o f both informal credit alone (negative) and "Note,however,thatMeinzen-Dick(1996)concludesthatheranalysisoffarm-level surveydataimplies that, "water markets do improve the productivity o f agriculture, particularly for small and medium farmers, butpoliciesthat expand tube well ownership are likely to provide greater welfare gains thanthose which encourage groundwater sales from tube wells owned by a few farmers," (p.2). 110 formal credit alone (positive), and the positive and statistically significant estimate usingthe sum o f two sources, suggests that the two credit types are substitutes. A2.28. It should be emphasized, however, that these results reflect an average for farmers o f all sizes, and smaller farmers receive a small share o f formal credit. Small farmers apparently enjoy on average little gain to a marginal increase ininformal credit, but informal credit i s usually more expensive; so this evidence does not contradict the possibility that small farmers mightbenefit from greater access to formal credit, which i s usually offered at lower interest rates. Table A2.5 Returnon Credit (Elasticities) Base Case I nform a1 Formal Informal andor Credit Credit Formal Credit Capital 0.267""" 0.198" 0.207"" (0.104) (0.105) (0.105) Education 0.814""" 0.831*** 0.777""" (0.236) (0.224) (0.219) Own Labor 0.320 0.271 0.247 (0.206) (0.192) (0.195) Land -0.214" -0.290" -0.282" (Owned) (0.131) (0.149) (0.148) Land 0.325"" 0.334"" 0.340"" (Operated) (0.135) (0.140) (0.139) Water 0.082"" 0.091""" 0.095""" (0.03 6) (0.036) (0.037) Credit -0.014 0.031 0.059"" (0.080) (0.046) (0.026) Note: Standard errors inparenthesis. Implies significant at the 10% level, *** implies * ** implies significant at the 5% level, and significance at the 1% level. None o f the elasticities are different from their base case values at the 10% significance level. A2.29. Khandker and Faruqee (2000) show that the returns to lending from the Agricultural Development Bank o f PakistanADBP (which accounts for 55% o f formal lending inrural areas) i s much higher for smaller farms than for larger farms, but nonetheless lending i s highlybiasedtowards larger farms. Furthermore, the recoveryrate o floans is higher for small farmers than for larger ones. This suggests that the ADBP as o f 1995 (whenthe survey usedby the authors was implemented) could have beenmore efficient inits targeting o f loans. Private financial institutions, on the other hand, are reluctant to participate inrural financial markets. Although they accrue 30% o f their deposits from rural areas, they only allocate 5% o f their lendingportfolio to rural areas.18 18Khandker and Faruqee (2002) 111 OTHER RESULTS A2.30. Contrary to what was expected, there are few statistically significant interactions between output prices and factor returns. Perhaps the only strong interaction we find i s between the relative prices o f output and education. An increase inthe prices of field crops and livestock, relative to the price o f other crops, would increase the elasticity o fnet revenues with respect to education. A deeper interpretation of this result ina cross-section analysis i s problematic, because it depends on the mix o f "other crops,'' and also could be a result o f the distribution o f prices across regions. In whatever case, this i s a noteworthy deserving more attention in future research. We also findthat the elasticity with respect to capital increases with the price o f field crops and decreases with the price o f other crops. To the extent that present policies offer protectionto field crops relative to other crops, farmers have an increasedincentive to invest, most likely inthe protected activities. Field crops are usually more intensive inthe use o f physical capital. THEEFFECTSOFTENANCYTYPE A2.3 1. The coefficients on the dummy variables representing different tenancy types are all positive, implyinggreater revenues due to access to agricultural land compared to those without operational farmland (some landless). We find, that, controlling for farm size, education and the other variables, landlords who are also tenants leasing in, and fixed rentaltenants earn more revenues than sharecropping tenants, who eam only slightly more (6,365 rupees) than those without operational farmland.'g For a small farmer, controlling for all other variables inthe revenue function, simplybeing a fixed-rent tenant raises per-acre revenues by 35,000 rupees (75% o f an average small farm's total farm income, or 49% o f the household's total income). A2.32. But the difference between fix-renter income and share-cropper income varies with family and farm sizes, andwith water use. The negative coefficients on the interaction variables share-cropper/household-size and share-cropper/operation-size suggest that the difference inrevenue between fixed renters (or owner-operators) and share-croppers increases with families and farm size. Thus for smaller farms the advantage of the non- sharecropper would be smaller; the advantage o f fix rents increases with farm size. (The reader should note that the p-values of the t w o interaction coefficients are 0.15 for family size and 0.12 for farm size, higher than the standard significance levels o f 5% and lo%.) The point estimates of the coefficients associated with sharecropping also suggest that there could b e sufficiently small share-cropper family and farm sizes that the share-cropper income would equal or even exceed owner-operator or fixed renter income. A2.33. Does this implythat share-cropping is an inefficient contract compared to fixed rental? Possible inefficiencies associated with share-cropping arise due to the sub-optimal provision of hidden or non-contractible inputs (not subject to monitoringby the landlord). The incentive to under-provide variable inputs arises because the share-cropper only earns a fraction of the marginal value product of the operation. Where monitoring of input use by landowners i s effective, one would not expect under-provision. Inour study possible "hidden" inputs (non-monitorable) are own family labor (measured by family size) and l9The effect onrevenues ofbeinga sharecropper (relative to those without operational farmland) is calculated using the average values o f small farms for household size, operational farm size and hours o f water use, and i s 299.6 + (8.51)*(-27.77) + (2)*(-31.73) + (56)*(0.51) = 13.3 (although not highly statistically significant). To get the value inRupees we multiply by the price o f "other crops" yielding 13.3* 480 = 6,365 RP* 112 surface water use. We find that this type o f inefficiency exists inthe provision o f surface water as indicatedby the positive coefficient on the interaction o f sharecropper dummy with water. Since the marginalreturns o f canal water are greater for sharecropper tenants, this implies that inreality there i s an effective under-provision of this input.The marginal product o f an additional unit of surface water is, ceteris paribus, greater for the sharecropper than for other tenancy types. Increases infarm size andthe number o f family members employed inthe farm operation may also increase monitoring costs per farm, giving incentive to have small operations for sharecropping. (But this may be off-set by increasing the monitoring costs o fhaving more operational units to monitor.) A2.34. Land leasingarrangements as they prevail inPakistan apparently do not provide land tenancy security or a mechanismto foment longer-term investments by the tenant. Present land leasing arrangements do not provide incentives to undertake on-farm investments attached to the land. Such investments mightbe insoil-improvement, land leveling, fixed structures and orchards, and water-saving projects. These are projects that provide longer-termbenefitsbut over which the tenant (sharecropper or fixed rent) has no ownership rights.This i s an area where Pakistancould examine altemative rental systems, as for example operate inEurope, where land rentals are common. 113 Appendix 3: DistributionalImpactsof Agricultural Growth and Factor -- Market Reforms'" A3.1. Rural factor market reforms have the potential to raise agriculturalproductivity through improving the allocation o fland, water, labor and capital across agricultural activities. This increased agricultural output, inhun,increases demand for inputsinto agricultural production, as well as demand for processing o f agricultural products (e.g. wheat millingor cotton ginning and spinning). Inaddition, as household incomes increase, final demand for bothagricultural and non-agricultural goods and services increases. To the extent that domestic supply i s elastic, the increased demand can spur growth inproduction and incomes throughout the economy. H o w these gains inincome are distributedacross households i s determined largely, however, by the ownership and access to factors o f production. A3.2. This appendix presents an analysis o fthe impacts o f agricultural growth and selected factor market reforms on agricultural and non-agricultural growth, and household incomes inPakistanusing a fixed-price multiplierframework. The data base for the analysis i s a 2001-02 social accountingmatrix (SAM), a consistent set o f accounts o f the economic flows betweenproduction activities, earnings o f factors o fproduction, incomes o f various types o fhouseholds, and demand for goods and services. THE2001-02 PAKISTANA M S A3.3. The 2001-02 Pakistan S A M includes 34 activities, each producing a single commodity, except for irrigatedwheat and non-irrigated wheat, both o f which produce a single commodity (wheat), (Box A3.1). For the twelve agricultural activities, returns to land and own-family labor are disaggregated by region (Punjab, Sindh, and Other Pakistan) and by size o f farm (small (O- 12.5 acre), medium(12.5-50 acre) and large (50 acre plus) farms (defined according to operated area, not land ownership). Out o f the 27 factors o f production that are specified, 23 involve only agriculturalproduction: 8 types o f agricultural labor, 12 types o f land, and 3 other factors: water, livestock capital and other agricultural capital. A3.4. This detailed treatment o frural factors and agriculture inthe S A M reflects the primary objective o f constructing the S A M : to better understand the relationshipbetween agricultural performance and rural income growth inthe context o f imperfect rural factor markets. Fifteen o f the nineteen household categories are rural agricultural households, split according to amount o f land cultivated (large farm, small farm, landless) and region (Sindh, Punjab and Other Pakistan). Non-farm households, bothrural and urban, are split into poor and non-poor according to their 2000-01 per capita household expenditures, with poor households defined as those with a per capita expenditure o f less than 748 rupeeslmonth per capita. By this definition, 23.5 percent o f the ruralpopulation (19.8 percent o f total population) are classified as non-farm ruralpoor households21and 22.7 percent o f urban households (15.3 percent o f total national population) are poor. 2oThis annex is basedonmultiplier analysis by Paul Doroshusing a Social Accounting Matrix for Pakistan, documented inDorosh, Niazi andNazli (2004), "A Social Accounting Matrix for Pakistan 2001-02: Methodology and Results", background paper for the PakistanRural Factor Markets study. 21Non-farmruralhouseholds are defied as rural households for which the mainoccupation o f the head o f household i s not crop or livestock farming. 114 BOX A3.1: STRUCTUREOF THE 2001-02 PAKISTAN S A M Activities(34) Agriculture (12): Wheat irrigated*, Wheat non-irrigated*, Rice-IRRI*, Rice-basmati*, Cotton*, Sugarcane*, Other major crops, Fruits and vegetables, Livestock-cattle/dairy*, Poultry, Forestry, Fishing Industry (16): Mining*,Vegetable oils*, Wheat milling*, Rice milling-IFW*, Rice milling-basmati*, Sugar*, Other food, Cotton lint/yam*, Textiles*, Leather*, Wood products, Chemicals*, Cementhricks, Petroleumrefining*, Other manufacturing*, Energy* Services (6): Construction, Commerce, Transport, Housing, Private Services, Public services Commodities(33) Same as activities with Wheat irrigated and Wheat non-irrigated aggregated as one commodity (Wheat). Factors(27) Labor (10): Own-farm (Large farm, Mediumfarm Sindh, Medium farm Punjab, Mediumfarm Other Pakistan, Small farm Sindh, Small farm Punjab, Small farm Other Pakistan), Agricultural wage, Non- agricultural unskilled, skilled Land(12): Large farm (Sindh, Punjab, Other Pakistan), Irrigated mediumfarm (Sindh, Punjab, Other Pakistan), Irrigated small farm (Sindh, Punjab, Other Pakistan), Non-irrigated small farm (Sindh, Punjab, Other Pakistan) Other factors (5): Water, Capital livestock, Capital other-agriculture, Capital formal, Capital informal A3.5. Data used inconstruction o f the S A M derive mainly from the 1990-91 Input-Output Table (97 sectors), 2001-02 National Accounts (value added for 15 sectors), 2001-02 Pakistan Integrated Household Survey (consumption disaggregation), the 2001Pakistan Rural Household Survey (household income disaggregation), and the 2001-02 PakistanEconomic Survey (sector/commodity data on production, prices, trade).22 A3.6. Household incomes and expenditures relative to those of other household groups follow similar patterns as inthe PRHS and HIES surveys, though absolute levels o f household incomes and expenditures are substantially higher given the apparent substantial under-reporting o f expenditures (particularly on services) and informal sector incomes inthe surveys. A3.7. On average, household incomes inthe S A M are 2.1 times greater thanhousehold expenditures inthe HIES Survey, andrural household incomes inthe S A M are 1.7 times greater than rural household incomes inthe PRHS survey (table A3.1). These ratios vary byhousehold group, however. The S A M calculations suggest that expenditures oflarge andmedium farmers are seriously understated inthe HIES -by a factor o f 3.8, on average. Compared to total incomes inthe PRHS survey results, however, factor incomes of medium and large farmers inthe S A M are only 2.8 times higher. N o t e that m u c h of the difference inincomes is due to estimated retums to cultivated land. F o r only t w o household groups, rural non-farm poor andrural agricultural labor in Sindh, are the S A M household incomes less than household expenditures inthe HIES.23 ~~ 22See Dorosh, Niazi and Nazli (2004), "A Social Accounting Matrix for Pakistan, 2001-02: Methodology and Results" for details o f the construction o f the SAM. 23Note that, because of a lack of reliable data, the S A M does not contain householdtransfers receivedfrom other households. Including transfer income would likely bringthe S A M householdper capita income 115 A3.8. Agricultural factor incomes account for only 23 percent o ftotal factor incomes in Pakistan,but 60 percent o f total factor incomes for agricultural households (77 percent for agricultural households inSindh), (table A3.2). 91percent o f agricultural incomes derive from land, water, own-farm labor, or livestock, so that without access to land or livestock; earnings from hiredlabor and (non-livestock) agricultural capital account for only 9 percent o f agricultural incomes. A3.9. Landaccounts for morethan 30 percent ofhousehold incomes, only for mediumand large farm owners (table A3.3). Agricultural labor accounts for 32 percent o f incomes for agricultural wage laborer households, but only 9 percent o f incomes o f tenants and 4 percent of incomes o f small fanners. Overall, agricultural incomes account for an average o f 56 percent o f estimated incomes of agricultural households inthe S A M , compared with 75 percent inthe PRHS (table A3.4). As described above, the higher figures inthe S A M arises mainly because o f allocation o f income from non-agricultural labor and informal capital to rural households. estimates for these groups to approximately the same level as the HIES household per capita expenditure estimates. " " 9 1 ? I I 00 e; 3 117 TableA3.2 Pakistan: factor incomeshares by householdtype, 2001/02 Rur Rur Rur Rur AgHHs AgHHs AgHHs AgHHs All Rural Other All HHs HHs Sindh Punjab Pak Pakistan Labor 0.522 0.298 0.248 0.240 0.190 0.232 Land 0.081 0.144 0.242 0.220 0.190 0.218 Water 0.003 0.005 0.014 0.009 ----- 0.008 Capital 0.394 0.553 0.497 . 0.531 0.620 0.542 Of which Livestk &Ag Capital 0.115 0.205 0.408 0.249 0.379 0.302 Total 1.ooo 1.ooo 1.ooo 1.ooo 1.000 1.ooo Own-Farm&Agric / Total Inc 0.231 0.411 0.766 0.560 0.652 0.615 Own Farm/ Agric Income 0.906 0.906 0.890 0.896 0.949 0.905 Source: Pakistan S A M 2001102. Table A3.3 DisaggregatedFactor Income Sharesby HouseholdGroup,Pakistan2001/02 Agric Agric TotalNon-Agr Non-Agr Total Land Capital Labor Agric Capital Labor Other Incom Large Farm Owners (50+ acres) 0.34 0.36 0.00 0.70 0.15 0.10 0.05 1.00 Medium Farm Owners (12.5-50) 0.31 0.21 0.00 0.52 0.27 0.16 0.05 1.00 Small FarmOwners (42.5 acres) 0.18 0.32 0.04 0.55 0.23 0.16 0.06 1.00 PureTenants 0.16 0.35 0.09 0.60 0.15 0.20 0.06 1.00 Agricultural Laborers 0.00 0.21 0.32 0.53 0.27 0.14 0.06 1.00 Rural non-farm non-poor 0.00 0.00 0.00 0.00 0.50 0.43 0.07 1.00 Rural non-farm poor 0.00 0.05 0.00 0.05 0.59 0.30 0.07 1.00 Urban non-poor 0.00 0.00 0.00 0.00 0.11 0.45 0.44 1.00 Urbanpoor 0.00 0.00 0.00 0.00 0.18 0.76 0.06 1.00 Rural Subtotal 0.13 0.20 0.04 0.37 0.33 0.24 0.06 1.00 Rural non-agric households 0.21 0.29 0.06 0.56 0.23 0.16 0.06 1.00 Urban subtotal 0.00 0.00 0.00 0.00 0.11 0.48 0.41 1.00 Ip.11 Pakistan 0.06 0.09 0.02 0.17 0.21 0.37 0.25 1.04 Source: Pakistan S A M 2001/02. 118 Table A3.4 PakistanRuralAgricultural Incomes PRHS PRHS SAM SAM Agric Inc Agric Inc Agric Inc Agric Inc Per Capita Share Per Capita Share (`000Rs) (percent) (`000 Rs) (percent) Medium and Large Farms 15.7 83.5 29.9 57.2 Small Farms 6.1 67.9 8.6 54.8 Landless Farmers 7.2 87.7 5.3 59.7 Rural Agric Workers 2.2 53.1 5.5 53.1 Rural Non-Farm Non-Poor 0.3 I.9 0.1 0.4 Rural Non-Farm Poor 0.2 6.3 0.2 4.5 Total Rural 6.1 69.7 5.6 37.2 Rural Agric Households 7.1 74.8 9.6 55.9 Source: Pakistan S A M 2001/02; Pakistan Rural Household Survey 2001102. ModelingFramework A3.10. This analysis of growth linkages inPakistan uses a variant o f the fixed-price, linear input- output (IO) model, the semi-input-output (SIO) A3.11. The most basic input-output models use fixed coefficients to simulate inter-industry production and consumption linkages, assuming fixed prices and perfectly elastic supply in all sectors. Given the assumption o f perfectly elastic supply, any increase indemand leads only to higher output, with no change inprice. A3.12. Total supply ineach sector (Z) i s modeledas the sum o f inter-industry input demand (AZ) and final demand (F), where final demand includes consumptionby households ( Y) and p exogenous sources o f demand such as exports (E). Income (Y) i s related to production through a fixed value added share (v) in gross commodity output (Z). Z = A Z + F (1) =AZ +pY +E =AZ +pVZ +E A3.13. Intheir unconstrained versions, these models assume supply to be perfectly elastic inall sectors. Presumedexcess capacityinall sectors ofthe economy implies no investment constraint to increases inoutput. Instead, total output and incomes are determined purelyby the level of exogenous demand (E) and a set o f behavioral parameters ((3 (2) z =(I- c yE 24This presentationof the S I 0 model closely follows Dorosh and Haggblade (2003), "Growth Linkages, Price Effects and Income Distribution inSub-Saharan Africa". Journal o f African Economies, 12(2). 119 A3.14. Perfectly elastic supply is, o f course, an unrealistic assumption inmost developing countries, especially for agriculture, where land, labor, rainfall and technology frequently limit output. Industrial output i s also often constrained by lack o f capital. By ignoring supply constraints altogether, the unconstrained input-output (IO) models typically exaggerate the size o f the inter-sectoral linkages. Comparisons from three different rural economies suggest that classic, unconstrained input-output models overstate agricultural growth multipliers by a factor o ftwo to ten (Haggblade, Hammer and Hazell, 1991). A3.15. To better simulate real-world supply rigidities, semi-input-output (SIO) models disaggregate sectors into those which are either supply-constrained (Z,) or perfectly elastic in supply (Z,) (Bell and Hazell, 1980). Insupply-constrained sectors (Z,), firms operate at full capacity, and output cannot increase without additional capital investment or introduction o f new, more productive technology. As indicated inequations (3) and (4), the S I 0 model permits output responses only inthose sectors with excess capacity (Z,). Perfect substitutabilitybetween domestic andimportdexports inthe supply constrained sectors (Zl) guarantees that prices are fixed for all tradeable goods. Thus, for these models to produce a reasonable approximation o f reality, the supply constrained sectors must correspondto tradeable goods with fixed domestic supply at the given fixed price, and the perfectly elastic sectors must correspond to non-tradeable goods. In supply-constrained sectors (Zl), increases indomestic demand merely reduces net exports (E& which then become endogenous to the system. A3.16. Investmentinadditionalproductive capacity or the introduction o fnew technology will trigger expansion inthe productiono f tradeable goods (Z1) such as agricultural cereals, export crops and manufactures. Therefore, the key shocks initiating growth are those that release production constraints inthese tradeable sectors. New investment inproductive equipment - induced by government policies or incentives -will increaseproductive capacity o f tradeable goods. Public investment in transportation infi-astructure or irrigation facilities opens up new regions to external markets. Public investments inapcultural research generate new technology that improve productivity o f cereals and other tradeable agricultural products. Equation 4 provides the means of quantifyinghow much impact these investment shocks will trigger, inthe initial sector as well as throughout the economy. A3.17. The specification of which sectors are considered elastic insupply is crucial to these S I 0 multiplier estimates. Inthis Pakistan analysis, the major agricultural traded commodities (wheat, IRRIrice, basmatirice, cotton, and sugar cane) are treated as inelastically suppliedinboththeir raw and processedforms. Livestock (cattle and other large animals) i s also modeled as inelastic in supply, thoughpoultry is assumed to be constrainedby demand. Most industrial sectors (mining, vegetable oils, leather, chemicals, petroleumrefining, other manufacturingand energy) are modeled as inelastic in supply; services are modeled as elastic insupply. (See box A3.1). 120 A3.18. Parameters for the model are ingeneral, are derived directly from the S A M . However, consumption parameters were adjusted to reflect marginal household budget shares and savings rates different from the average shares and savings rates reflectedinthe SAM.*' SIMULATION 1: OUTPUT SHOCKS TO CROPAND LIVESTOCK AGRICULTURE The first set o f simulations model a 10percent increase in output o f tradable crops, together with a 10 percent increase inprocessing o f these activities.26 as shown inappendix table A3.5, a simultaneous 10 percent increase inwheat, basmati and irririce, cotton and sugar cane production and processing, increases incomes o f large and medium farmers by 7.2 percent, and incomes o f small owners and tenants by 4.6 percent. Incomes o f agricultural workers rise by only 2.9 percent, however, as much o fthe gain in labor incomes accrue to farm owners and tenants. Growth linkage effects on non-agricultural output result ina 3.3 percent increase inincomes o f the non-farm rural poor and a similar increase in the incomes o f the non-farm rural non-poor and the urban poor. A 10percentincrease inthe output oflarge livestock has an even greater effect, raising overall incomes by 4.5 percent (compared to 3.1 percent for a corresponding increase inoutput o f the major crops). Due to the distribution o f livestock assets, the gains for small farmers and landless agricultural households are especially large: 9.7 to 10.9percent. Combining 10percent increases inmajor crops and livestock, raises national incomes by an average o f 7.6 percent, with gains to the rural non-farm poor (8.2 percent) approximately halfthe size o f gains to farm households and agricultural laborers (13.8 to 14.7 percent). Thus, linkage effects o f agricultural growth on rural non-agricultural incomes are significant, though agricultural growth linkages alone are not sufficient to rapidly raise rural non-agricultural householdincomes. 25 Specifically, marginal savings rates o f households were raisedby 0.10, andmarginal budget shares o f manufactured goods, transport and private services were changed by 0.10, -0.5 and -0.5, respectively. 26 Since both the agricultural production and processing activities o f these crops are modeled as inelastic in supply, it is necessary to exogenously increase both activities inthe simulation. 122 SIMULATION 2: SHIFTFROMSHARE-CROPPINGTO FIXED RENTS Intheory, productivity of share-croppers is expectedto be lower than o f owners or those with fixed-rent tenants, because for some inputs (such as own-labor), share-croppers pay the full (implicit) cost, but reap only a part (generally one-half) o f the benefits interms of value o f output. Usingdata from hies 2001/02, jacoby and mansuri (2004a) findno significant difference between productivity o f share-croppers and owners for cultivation o f major crops. However, share-croppers who are not supervised (those who meet with their landlords fewer than ten times in a season to discuss a given plot) are found to be 18 percent less productive than land-owners or share-croppers who are supervised. Simulation 2 models the effects of a 22 percent (=l/( 1-.18)) increaseinproductivity for major crops as a result o f a hypothetical shift from share-cropped to either fixed rent or owned land. Given that 18.8 percent o f land inPakistan i s share-cropped (table A3.6), and o f this, an estimated 35 percent o f tenants are not supervised (based on prhs 2001/02 data), the effect on total production of major crops i s only 1.4 percent2' inthe simulation, all the marginal increase inreturns to land are allocated to share-croppers?* according to their shares o f total share-cropped land inthe hies 2001/02.29 Note that 39 and 27 percent o f share-cropped area i s farmed by pure tenants in sindh and Punjab, respectively (table A3.7). Results are given intable A3.8.30 though the gains on a national level are small, the impacts on tenants are significant. Incomes o f tenants in sindhrise by 4.0 percent; incomes o f tenants inPunjab rise by 2.8 and 2.7 percent, respectively. However, even with a significant (and likely over-optimistic) value-added multiplier o f 2.5, the initial shock i s small relative to the size o f the Pakistan economy, and the multiplier income gains are spread widely across household groups, so incomes o f non-farm groups rise by only 0.3 percent at most. 27 (share-croppedland/ total land) = 0.22 * 0.35 * 0.188 = 1.4%. (Productivity effect on unsupervisedshare-croppedland) * (unsupervisedtotal share-croppedland) * 28Inthe case ofa shift fromshare-croppingto fixed rents, this implicitlyassumes that the fixed rentinrupee terms i s set onthe basis of the value ofproduction less purchasedinputs for the landwhen it was share- cropped. Thus, the simulation models an upper-boundonthe direct benefits to former share-croppers of the productivityincrease. 29 Inthe model simulations,this value addedis paiddirectlyto households,bypassingthe factor accounts. 30Note that inthis simulation (andthe ones that follow), we model a production shock insteadof a simple value added shock, under the assumptionthat the productivity gain requires an increase inkey inputs like labor and fertilizer. 123 Table A3.6 share of householdgroup area cultivated by tenure status, 2001/02 Source: Authors' calculationsfrom HIES 2001/02 data. Table A3.7 share of area cultivated by tenure status and household group, 2001/02 Source: authors' calculations fromhies 2001/02 data. 124 Table A3.8 simulation results: percentagechangeinhouseholdincomes r- Increased Source: Model Simulations. SIMULATION 3: REMOVAL OF CREDIT CONSTRAINTS UsingPRHS 2001/02 data, Mansuri and Jacoby (2004b) show that farmers that are credit- rationed have lower productivity incrop production than do non-rationedfarmers. Simulation 2 models a relaxation o f the credit constraint for small farmers inPakistan by increasingtheir crop output according the econometric estimates o f the productivity effects. Usingthe simplifying assumption that only small farmers are credit constrained, the magnitude o f the productivity gain on total output o f major crops i s calculated usingthe estimated share of small farms that are credit rationed. Inestimate 1, the 11.1percent of farmers that are constrained inboth formal and informal credit markets have 23 percent lower yields, suggesting that total output o f all small farmers would rise by 3.3 percent inthe absence o f credit constraints. Similar, usingthe alternate estimate o f a 26 percent decline inyields for farmers who are constrained informal credit markets (45.7 percent o f small farmers), output o f small farmers would rise by 16.0 percent (table A3.9). This 16percent increase insmall farm output (split across small farmers according to their shares inarea o f each crop as reflected inthe S A M ) results inmoderate gains inaverage total incomes for small farmers: 3-4 percent for small farm owners and pure tenants in Sindh and Punjab. 125 The increase in small farm output results ina bigger percentage gain inincomes from large farmers than small farmers inlarge part because landrents received account for a large share o f income o f large and (especially) medium farmers (table A3.10). For example, returns to land on small farms (mainly landrents) account for 23 percent o f income for mediumland-owners inSindh, while returns to land on small farms account for only 12 percent of incomes of owners (and owner cum tenants) o f small farms. CreditConstraint Formal and Formal Informal Only (1) Share o f farm households that are credit rationed 0.1 0.41 (2) Number o f small owner and owner cum tenant farms as % o ftotal farms 66.6% 66.6% (3) Number o f pure tenant farms as % o f total farms 23.1% 23.1% (4) Number o f small farms (all types) as % of total farms, (2)+(3) 89.8% 89.8% (5) Percentage of small farms that are credit rationed, (1)/(4) 11.1% 45.7% (6) Impact o f credit rationingon productivity (regression results) -0.23 -0.26 (7) Outputgain from relaxing credit constraint, [1/(1+(6)] - 1 29.9% 35.1% (8) Percentage gain in output o f small farmers from relaxing credit constraint, (5 1*(7) 3.3% 16.0% (9) Share o f small farmers intotal land value added (major crops) 53.8% 53.8% (10) % increase intotal value added (major crops) 1.8% 8.6% 126 TableA3.10 Distributionof EstimatedReturnsto SmallFarmIrrigatedLand(Major Crops), 2001/02 SmallFarmer Punjab 41,746 34.8% 13.1% SmallFarmer OthPak 9,676 8.1% 7.7% LandlessFarmerSindh 4,099 3.4% 9.4% LandlessFarmerPunjab 4,774 4.0% 10.4% LandlessFarmerOthPak 1,365 1.1% 9.1% All Farmers 119,902 100.0% 12.9% LargeFarmers 7,004 5.8% 7.5% MediumFarmers 44,134 36.8% 19.5% SmallFarmers 58,525 48.8% 11.7% LandlessFarmers 10,238 8.5% 9.8% All Farmers 119,9021 100.0% 12.9% 127 SIMULATION 4: IMPROVED DISTRIBUTIONOF WATER Excess use o f water by farmers at the head o f canals, and water supply shortages inthe canal itselfresult inlower productivity for farmers at the tail end o f canals. Simulation 3 models the effects on productivity and incomes o f easing the water constraint by providing more water to farm plots at the tail endo f canals. This gain in supply o f water to tail end farmers could be achieved through better water distribution or increased availability o f water achieved through measures such as canal lining. Inthe simulation, this increase inwater at the tail end i s assumed to have no effect on productivity o f farmers at the head end, however. Detailed data on the share o f land affected by water shortages at the tail end o f canals i s unavailable, but limited survey evidence (University o f Faisalabad 2003 survey, Z. Hussain, 2003) suggests that approximately 27 percent o f farmers are located at the tail end o f canal systems. In the simulation, productivity o f crop production at the tail end i s increasedby 30 to 90 percent, based on the productivity differentialsbetweenhead and tail farms reportedinTable 4.6.31 Tail end land i s assumed to be distributedby farm size inthe same proportions as the average distribution in each region (Sindh, Punjab, Other Pakistan). As shown intable A3.7, large farmers enjoy the largest percentage increase inincomes as a result o f this productivity gain, with incomes o f large and medium farmers in Sindh risingby 15.0 and 9.9 percent, respectively. Incomes o f large and mediumfarmers inPunjab rise by 10.7 and 8.4 percent, respectively. Small farm owners and pure tenants inthese two provinces benefit as well, as their incomes rise by 6.1 to 6.6 percent. However, in spite o f the significant (and perhaps over-optimistic) gains in output modeled here, incomes o f agricultural workers inthe two provincesrise by only 3-3-3.6percent and incomes of the non-farm rural poor rise by only 3.3 percent. Thus, in spite o f a 14percent increase inthe value o fproductiono f major crops and large GDP multipliers, the gains to the poorest rural household groups (agricultural workers and non-farm ruralpoor) are relatively small. TableA3.11: Crop Yields by LocationAlong Canals (tondhectare) Ratio Head Tail Head: Tail Wheat 2.4 1.7 1.4 IRFURice 2.9 1.9 1.5 BasmatiRice 2.2 1.7 1.3 Sugar Cane 54.7 29 1.9 31Cotton production o ftail end farmers i s increasedby 50 percent (the arithmetic average o f the productivity differentials inAppendix Table 3.10). 128 Table A3.12: Simulation Results:HouseholdIncomes I lTota' II I I I I 34851 0.331 2.31 I 3.541 Shares of total income gains (percent) aInbillionrupees. Source: Model simulations 129 Conclusions Because o f data uncertainties and simplifyingassumptions used, the above analysis o f the effects of increases inagricultural output on incomes inPakistan agriculture illustrates only the broad order o f magnitude o f the effects. Moreover, these simulations show only the static results of factor marketchanges, not dynamic effectso fimproved incentives onprivate investments and market development. Nonetheless, the broad structure o f rural production, distribution of land and other factors o f production, and structure o f household incomes are reflectedinthe analysis. Two broad conclusions emerge. First,the level and distribution o f the benefits ofremoval of factor market distortions depend crucially on the magnitude o f the inducedproductivity shock and the ownership of the assets involved. Inspite o fmultiplier effects, reforms related to share- cropping are likely to have only small overall impacts on rural incomes inaggregate inthe short run,thoughthe benefits to tenants couldbe significant. Credit market reforms and improvements in water availability have potentially larger effects on the overall rural economy. Second, the linkage effects o f expansion intraditional crop agriculture, though important, may not be adequate to substantially raise incomes o f agricultural laborers and the rural non-farmpoor who lack significant capital resources. Thus, inthe absence o f a change inthe structure o f rural incomes and employment or significant gains inthe rural non-farm economy apart from agricultural growth- induced linkage effects, targeted interventions to agricultural laborers and the rural non-farm poor will likely be neededto rapidly reduce the poverty o fthese groups. 130 References Agrodev (2000). Sindh Rural Development Project-Final Development - the Government o f Sindh and the Asian Development Bank, October 2000. Report prepared for Planning and Agrodev Canada Inc Aleem Irfan (1990). Imperfect Information, Screening and the Cost o f Informal Lending. The WorldBank Economic Review 4:3. Alderman, H. and S. Chishti (1989). Simultaneous Determination o f Household and Market Oriented Activities o f Women in Rural Pakistan. Research in Population Economics. Volume 17. Ahluwalia, Montek S. (1978). Rural Poverty and Agricultural Performance in India, Journal of Development Studies. 298-323. Adams, Richard H Jr. (1996). Remittances, Income Distribution, and Rural Asset Accumulation, FCND Discussion Paper No. 17. Food Consumption And Nutrition Division. International Food Policy ResearchInstitute. Washington, D.C. Ahmad, Munir, Chand Asghar, and Nasir Alam Khan. (1993). Participation o f Rural Women in Agricultural and Household Activities: A Micro-level Analysis, In A. S. Haider, Z. Hussain, R. McConnen, and S. J. Malik (ed.) Agriculture Strategies in the 1990s: Issues and Policies. Islamabad: Pakistan Association o f Agricultural Social Scientists. Applied Economic Research Centre (1986). Specific Constraints Facing Small Farmers in Baluchistanand Sindh, AERC, Karachi. Anriquez G., and A. Valdes. (2004). Determinants o f FarmRevenue inPakistan, Background paper prepared for the Pakistan Rural Factor Markets study (manuscript). Ali, M.and D.Byerlee, (2000). Productivity Growth and Resource DegradationinPakistan's Punjab: A Decomposition Analysis, World Bank Policy Research Working Paper N o 2480. ,(2002). Productivity Growth Differential inPunjab, Pakistan: A District-Level Analysis, The Pakistan Development Review, Vol. 40, No. 1(1-25). kif, G. M.f and Khalid Hameed Sheik (2002). Women's Employment Concerns and Working Conditions inPakistan, PIDE Workshop Paper. Bhalotra, Sonia and Christopher Heady (2001). ChildFarmLabor: The Wealth Paradox, Social ProtectionDiscussion Paper, The World Bank, Series. No. 125. September. 2001 Chaudhry, M.Ghaffar and Zubeda Khan (1987). Female Labor Force Participation Rates inRural Pakistan: Some Fundamental Explanations and Policy Implications, ThePakistan Development Review. Vol. XXVI, No. 4, Part 11. Cox-Edwards, A. (2004). Labor Markets inRuralPakistan, Backgroundpaper prepared for the Pakistan Rural Factor Markets study (manuscript). 131 Datt, Gaurav and Martin Ravallion (1998). Farmproductivity and rural poverty inIndia, Journal of Development Studies. No. 34 (April). Dorosh, Paul and Steven Haggblade (2003). Growth Linkages, Price Effects and Income Distribution in Sub-Saharan Africa, Journal ofAfrican Economies. 12(2). Dorosh, Paul, Md. KhanNiazi and HinaNazli (2004). A Social Accounting Matrix for Pakistan 2001-02: Methodology and Results, Background paper prepared for the Pakistan Rural Factor Markets study (manuscript). Dorosh and Valdes (1990). Effect o f Exchange Rate and Trade Policies on Agriculture in Pakistan. International Food Policy ResearchInstitute Research Report 84. Washington, D.C. Ercelawn, Aly and Sohnia Ali (2003). Forced Labor in the Punjab Agricultural Sector, Daily Dawn. January 6th. Fafchamps, Marcel and Agnes R. Quisumbing(1998). Human Capital, Productivity, and Labor Allocation in Rural Pakistan. FCND Discussion Paper No. 48. Food Consumption and Nutrition Division. International Food Policy ResearchInstitute.Washington, D.C. FA0 (1998). Women in Agriculture, Environment and Rural Production-Fact Sheet Pakistan. FAO-Bangkok. Federal Bureau o f Statistics (2001). Poverty inthe 1990s. FBS. Islamabad. FederalBureau o f Statistics / ILO-IPEC 1996 Child Labor Survey. Frias, J.L. (1992) Evolucion Reciente de la Industria del Agua en Inglaterra. Fraincia y Chile, McKinsey and Company, Mexico. Government o f Pakistan (various issues), Labor Force Survey. Federal Bureau o f Statistics. Statistics Division. Islamabad. Government o f Pakistan, (1990), Agriculture Census of Pakistan-1990, Federal Bureau o f Statistics. Islamabad. Government o f Pakistan, (various issues), HouseholdIncome and Expenditure Survey. Federal Bureau o f Statistics, Islamabad. Government o f Pakistan, (various issues), Pakistan Rural Household Survey. Government o f Pakistan, (various issues), Pakistan Economic Survey. Economic Advisor's Wing, MinistryofFinance, Islamabad, http://www.finance,gov.pk. Government o f Pakistan, (various issues) Monthly Statistical Bulletin, Islamabad: FederalBureau o f Statistics. Government o f Pakistan (1985). Rural Credit Survey o f Pakistan, Lahore, Agricultural Census Organization. 132 Government o f Pakistan (2003). Cost o f Production series data for various crops, Agricultural Prices Commissions(APCOM), Ministryo f Food, Agriculture and Livestock. Government of Pakistan (1999). Agricultural Statistics o f Pakistan, Ministry o f Food, Agricultural and Livestock, Islamabad Grootaert, C. and Ravi Kanbur (1995), Child Labor: A Review, Policy Research Working Paper No. 1454. The World Bank. Washington D.C. Hamid, Shahnaz. (1991). Determinants o f the Supply o f Women inthe Labor Market: A Micro- Analysis, ThePakistan Development Review. Volume 30, No. 4. Hamid, Nabi and Nasim (1990). Trade, Exchange Rate, and Agricultural PricingPolicies in Pakistan, The World Bank. Hazell, P.B.R.and C. Ramasamy (eds.). (1991). The Green Revolution Reconsidered: The Impact of High-Yielding Rice Varieties in South India. Baltimore: The Johns Hopkins University Press. Hazell, P.B.R. and A. Roell. 1983.Rural Growth Linkages: Household Expenditure Patterns In Malaysia and Nigeria. ResearchReport 41. Washington, D.C.: InternationalFood Policy Research Institute. Hirashima, S. (1996). Asset Effects inlandPrice Formation inAgriculture: The Experience from South Asia, ThePakistan Development Review. Islamabad. Ifran, M., G.M. Arif, Mubashir Alo and Hina Nazli (1999). The Structure o f Informal Credit Market inPakistan. Pakistan Institute o f Development Economics Research Report 168. Jacoby, Hanan G. and Ghazala Mansuri. (2003). Incomplete Contracts and Investment: A Study of LandTenancy inPakistan, DEDCRG, World Bank, (manuscript). ,(2004a). The (In)-Efficiency o f Share-Tenancy Revisited: Evidence from Pakistan. Draft manuscript. . (2004b). Agricultural Credit and FarmPerformance: Evidence from Pakistan, Backgroundpaper prepared for the Pakistan RuralFactor Markets Study (manuscript). Jacoby, H.and R.Murgai (2001). Monopoly Power and Distribution inFragmented Markets: The Case o f Groundwater, World Bank, Policy Research Working Paper 2628. Jaramillo, C. F. 2001. "Liberalization, Crisis, and Change: Colombian Agriculture inthe 1990s." Economic Development and Cultural Change 49(4): 821-46. Kazi, Shahnaz and Faiz Bilquees (1992). Final Report of the Sub-committee on Women and Development, Reportprepared by the Committee on Economic and Social Well-being for the EighthFiveYear Plan. Pakistan Institute o f Development Economics.Islamabad. Kazi, S, and B.Raza (1991). Duality of Female Employment inPakistan. ThePakistan Development Review. Volume 30. No. 4. 133 Khan, S. A. and F. Bilquees (1976). The Attitude, Environment and Activities o f Rural Economy: Study of Jhok Sayal, Pakistan Institute o f Development Economics Research Report No. 98. Islamabad. Kozel, Valerie and Harold Alderman (1990). "Factors Determining Work Participation and Labor Supply Decisions in Pakistan's Urban Areas", The Pakistan Development Review. Volume 29, 11-18. Kurosaki Takashi and Humayun Khan (2003). Human Capital, Productivity, and Stratification in RuralPakistan April 2003. www.ier.hit-u.ac.ip/-kurosaki/edu0304.pdf Kurosaki, T, (1996). "Government Interventions, Market Integration, and Price Risk in Pakistan's Punjab", ThePakistan Development Review. Vol35. No. 2. Lopez, Ramon and Gustavo Anriquez. (2003). Poverty and Agricultural Growth: Chile inthe 1990s. Paper prepared for the Roles o f agriculture IntemationalConference, 20-22 October, 2003. Rome, Italy. Malik, Sohail J. (1990). Report on the Benchmark Credit Survey, Washington D.C., Intemational Food Policy Research Institute, (Report). .(1992). Credit Constraints andHouseholdBorrowingBehavior: Impact onHousehold Consumption and Fertilizer Use in Selected Areas o f RuralPakistan. Intemational Food Policy Research Institute, Washington, D. C. September. Mimeo. . (1999). Poverty and Rural Credit: The Case o f Pakistan, Pakistan Institute o f Development Economics, Islamabad. , (2003a). Agricultural Growth and Rural Poverty in Pakistan: A Review o f Evidence. Report to the Asian Development Bank, Palustan Country Office (manuscript). , (2003b). Rural Credit Markets in Pakistan: Institutions and Constraints, Background paper for Pakistan Rural Factor Markets Study, Islamabad. Malik, S.J. et al, (1989). ((Differential Access and the Rural Credit Market in Pahstan: Some Recent Evidence", ThePakistan Development Review. 28:4 Part 11. Malik S., Aftab, S;, And Sultana, N.(1993). Pakistan's Economic Performance 1947-1993:A Descriptive Analysis. Lahore, Pakistan: Sure Publishers. Malik, S. J. and Hina Nazli (1999). "Rural Poverty and Credit Use: Evidence from Pakistan". Fifteenth Annual General Meeting and Conference, Pakistan Society o f Development Economists. Meinzen-Dick R. (1996). Groundwater Markets in Pakistan : Participation and Productivity, Intemational Food Policy Research Institute Research Report, 105, Washington, D.C. Mellor, John W. (1976). TheNew Economics ofGrowth. Come11University Press, Ithaca, NY, . (1978). "Food Price Policy and Income Distribution in Low-Income Countries", Economic Development and Cultural Change. 1-26. 134 , (1992). Agriculture on the Roadto Industrialization. Baltimore: Johns Hopkins University Press. Mellor, John W. and Sarah Gavian. 1999. "Determinants o f Employment Growth inEgypt:The Dominant Role o f Agriculture and the Rural Small-scale Sector". Monitoring, Verification and EvaulationUnit, Agricultural Policy ReformProgram. Impact Assessment Report No. 7. Abt Associates, Inc. (manuscript). Munita J. (1994) Aumento de Eficiencia en el Us0 del Agua por Incorporacion de Nuevas Tecnicas y Arrendamientos Temporales, Universidadde Chile, Santiago. Nabi, Hamid and Zahid (1986). TheAgrarian Economy of Pakistan: Issues and Policy. Karachi: Oxford University Press. Naqvi N.H.,Khan M.H. and Chaudhury M.G. (1987). Land Reforms in Pakistan: A Historical Perspective. Pakistan Institute o f Development Economics. Islamabad Nazli, Hina (2003). Effect o f Education, Experience and Occupation on Earnings: Evidence From Pakistan. Innovative Development Strategies. Islamabad. (manuscript). Punjab Economic Research Institute (1986). Constraints Facing Small Farmers in Punjab. PEW, Lahore. PakistanPoverty Alleviation Fund(PPAF) (2003). Annual Report. World Bank. Qureshi, S. K and A Shah (1992). "A Critical Review o f Rural Credit Policy in Pakistan". The Pakistan Development Review 31:4. Qureshi, Sarfraz K., and G. M. k i f (1999). Profile o f Poverty in Pakistan, 1998-99, Pakistan Institute o f Development Economics, Islamabad. Qureshi, Sarfraz (2003). Rural Land Markets in Pakistan: Institutions and Constraints. Backgroundpaper for Pakistan Rural Factor Markets Study, Islamabad Ravallion, Martin and Gaurav Datt (2002). "Why Has Economic Growth Been More Pro-Poor in Some States o f India than Others?", Journal of Development Economics. 381-400. Renkow, Mitch (1991) Land Price, Land Rents, and Technological Change: Evidence from Pakistan. Mexico: C I M M Y T (Economic Working Paper 90/01). Rosegrant, Mark W., Renato Gazmuri Schleyer and Satya N. Yadav (1995). "Water policy for efficient agricultural diversification: market-based approaches, Food Policy. 20(3), pp. 203-223. Sathar, Zeba. A. and Cynthia B. Lloyd. (1993). "Who Gets Primary Schooling in Pakistan: Inequalities Among and Within Families", Working Paper No. 52. Research Division. The Population Council. New York. Sathar, Zeba. A. and Sonalde Desai (1996). Work Pattem in Pakistan: Intersections Between Gender, Family and Class, Working Paper No. 90. Research Division. The Population Council. NewYork. 135 Sarwar, M. and M. A. Saleem (1993). Baseline Study o f Punjab Small Holders Dairy Development Project inGujranwala. Punjab Economic ResearchInstitute. Lahore. Sayeed, Asad and Karammat Ali (1998). Labor Market Policies and Institutions: A Framework for Social Dialogue, PILER. ResearchReport No. 3. Scott, W. E.and David A. Redding (1988). Agricultural Credit inPakistan. USAID, Islamabad. Smith, Khushk, and Stockbridge (1999). Case Studies o f Corruption inAgricultural Markets in Sindh Province, Pakistan, and Implications for Market Liberalization, Journal of International Food and Agribusiness Marketing. 11(l), pp. 19-41. Sultana, Nargis, Hina Nazli, and Sohail J. Malik (1994). "Determinants o f Female Time Allocation in SelectedDistricts o f Pakistan", ThePakistan Development Review. 33(4). Sushma Joshi (2003). Bonded to Labor: The Contemporary SituationinNepal, South Asian Magazine for Action and Reflection ( S A M A R ) 16. (m.samarmaqazine.orgL Thirtle, Colin (200 1). Relationship Between Changes inAgricultural Productivity and the Incidence o f Poverty inDeveloping Countries, London, DIVDReport No. 7946. Timmer, C. Peter (1997). H o w Well do the Poor Connect to the Growth Process?, CAER DiscussionPaper No. 178, Harvard Institute for International Development, Cambridge. World Bank (1994). Pakistan Irrigation and Drainage: Issues and Options, Report No. 11884- PAK. Washington, D.C. (1997). PakistanNational Drainage Program Project, Staff Appraisal Report, Report N o 15310-PAK, November. Washington, D.C. (2002). Pakistan Poverty Assessment: Poverty inPakistan: Vulnerabilities, Social Gaps, and Rural Dynamics. Report No. 24296-PAK. Poverty Reductionand Economic Management Sector Unit. South Asia Region. Washington, D.C. (2003a). Pakistan Public Expenditure Management, Accelerated Development o f Water Resources and IrrigatedAgriculture, Vol. 11. Washington, D.C. (2003b) Poverty Reduction Strategy Paper on Pakistan, Report No. 25463-PAK. Washington, D.C. (2004a). Agriculture Investment Sourcebook, Agriculture and Rural Development. Washington, D.C. (2004b). ImplementingReachingthe Poor: A progress report on Rural Strategy Implementation for the World Bank Board o f Executive Directors; Hanid Mukhtar, Pakistan Office. Yaqub, Nadeem (2002). "Child Labor - Not only a Problem o f Poverty", Development and Cooperation (No. 1, JanuarylFebruary 2002, p. 28). 136 Z. Hussain (2003). Rural Water Markets in Pakistan: Institutions and Constraints, Background paper for Pakistan Rural Factor Markets Study, Islamabad Websites consulted wmv.pak.gov.pk WUTV. fa0.org wwwr.financc.gov.pk www.us.ilo.ordarchivdilofocusl200 1lwinter \Y\?rw.\?iorldbank.org W- 137 BACKGROUND PAPERS FOR THE PAKISTANRURAL FACTOR MARKET STUDY Ahmad, Munir.2003, Agricultural ProductMarkets inPalustan. Altaf, Zafar. 2004. Livestock Sector Development: Constraints and Opportunities. Anriquez, Gustavo and Albert0 Valdes. 2004. An Analysis o fthe Determinants o f Farm Revenue inPakistan. Dorosh, Paul, Muhammad KhanNiazi and Hina Nazli. 2004. A Social Accounting Matrix for Palustan, 2001-02: Methodology and Results. Cox-Edwards, Alejandra. 2004. Labor Markets inRuralPakistan. Hussain, Zakir. 2003. RuralWater Markets inPakistan: Institutions and Constraints. Jacoby, Hanan G. and Ghazala Mansuri. 2004. The (In)-Efficiency o f Share-Tenancy Revisited: Evidence from Pakistan. Kemal, A.R. 2003. Agricultural Growth and Rural Poverty inPakistan: A Regional Analysis. Malik, Sohail. 2003. Rural Credit Markets inPakistan: Institutions and Constraints. Malik, Sohail. 2004. A Note on Sources of Income and Poverty Dynamics inRural Pakistan. Nazli, Hina. 2003. Rural Labor Markets inPakistan: Institution and Constraints. Qureshi, Safraz. 2003. Pakistan Rural Factor Markets Study: RuralLand Markets inPakistan: Institutions and Constraints.