Introduction Conceptual framework Context and design Results The leakage and livelihood impacts of payments for environmental services A targeting experiment in Malawi B. Kelsey Jack Elsa Cardona Santos September 2013 Conclusion Introduction Conceptual framework Context and design Results Conclusion Payments for environmental services Land use decisions affect the environment � When private and social costs diverge: environmental externalities Payments for environmental services: Conditional cash transfer to align private and social costs � In practice, takes on many different forms � � Large national programs Small private programs Introduction Conceptual framework Context and design Results PES impacts Direct impacts: � Contracted land use activity (e.g., Pfaff et al. 2008) Conclusion Introduction Conceptual framework Context and design Results Conclusion PES impacts Direct impacts: � Contracted land use activity (e.g., Pfaff et al. 2008) Indirect impacts: � Environmental outcome(s) associated with the contract (Pagiola et al. 2007) � Environmental outcomes not contracted by the program (Alix-Garcia et al. 2012) � Socioeconomic impacts (Alix-Garcia et al. 2013, Uchida et al. 2009) Introduction Conceptual framework Context and design Results Leakage and livelihood impacts in Malawi Follow up on a randomized experiment in Malawi (Jack 2013) � Afforestation contracts with piece rate incentives � 3 year contracts; 4 payments based on performance Conclusion Introduction Conceptual framework Context and design Results Leakage and livelihood impacts in Malawi Follow up on a randomized experiment in Malawi (Jack 2013) � Afforestation contracts with piece rate incentives � 3 year contracts; 4 payments based on performance Answer two questions 1. Do PES contracts have indirect impacts on leakage and livelihoods? � Use random contract assignment (lottery), conditional on accepting the contract offer 2. Does targeting to low opportunity cost landholders mitigate those impacts? � Use random assignment to incentive compatible auction or lottery Conclusion Introduction Conceptual framework Context and design Results Conclusion An agricultural household model Households are endowed with land and labor � Land markets are missing, so unprofitable land remains idle � Off-farm labor is the only source of cash income through which a variable input (e.g. fertilizer) can be purchased In the absence of PES, households allocate land and labor to agricultural production until the marginal value of doing so equals the opportunity cost Introduction Conceptual framework Context and design Results Conclusion The effect of PES on land use Tree planting requires land and labor, but no variable input � Reallocate land toward tree production as long as the marginal benefit > marginal cost � � � Idle land: marginal cost is the cost of clearing Productive land: marginal cost is foregone crop income Prediction: idle land decreases / clearing increases under PES Introduction Conceptual framework Context and design Results Conclusion The effect of PES on livelihoods Livelihood impacts depend on the size of the payment relative to the opportunity cost of the contract � p > c makes households wealthier PES also affects labor allocation � Ambiguous predictions � � Depend on marginal value of labor allocated to trees, own production, leisure and the off-farm wage Liquidity constraints affect labor allocation if variable input is a substitute for on-farm labor Introduction Conceptual framework Context and design Results Conclusion The impact of contract allocation Household willingness to accept the PES contract is determined by the cost of allocating land and labor toward tree production � Those with surplus land and labor likely to have lower WTA � Targeting contracts to those with low WTA � � is less likely to distort land and labor allocations transfers greater surplus to households → less leakage, better livelihood impacts Introduction Conceptual framework Context and design Results Context Implemented by The World Agroforestry Centre (ICRAF) with funding from Ecobank Malawi � Objective: Increase tree planting by smallholder farmers Pilot program: 433 households in 27 villages in central Malawi � Surplus land (missing land market) � Imperfect labor and capital markets � Average landholding size: 5 acres � Average income: ~40,000 MWK Conclusion Introduction Conceptual framework Context and design Results The contract NGO provides: Farmer provides Seedlings and training upfront Monitoring and payments 1/2 acre of land Land preparation and planting Care for trees Payments in four equal installments (3,000 MWK each) Pay per surviving tree 6 months 1 year 2 years 3 years Conclusion Introduction Conceptual framework The trees (after 2 years) Context and design Results Conclusion Introduction Conceptual framework Context and design Results Conclusion Design and evaluation Part 1: Contract impact evaluation Experimental design: 205 landholders offered a PES contract → 204 accepted → 91 randomly chosen to receive the contract � ∼ 5 percent attrition between baseline and endline Analysis 1. Impact of random assignment to contract � � Compare outcome means at the end of the contract Difference in difference analysis: baseline and endline measures 2. Variable treatment intensity (endogenous) � � Some households kept more trees alive, received higher pay-out Estimate effects at different pay-out margins Introduction Conceptual framework Context and design Results Conclusion Design and evaluation Part 2: Targeting effects Experimental design: 433 landholders assigned to contract or auction → 228 bid in auction treatment, 85 receive contracts → auction clearing price offered to lottery treatment � ∼ 5 percent attrition between baseline and endline Analysis � Impact of allocation treatment on contract impacts � � Difference in difference analysis: outcome means at the end of the contract, by treatment group Triple difference analysis: contract vs. no contract, lottery vs. auction, baseline vs. endline Introduction Conceptual framework Context and design Outcomes � Land use outcomes � � � � Socioeconomic outcomes � � � � � Land clearing Tree planting Land acquisition Food expenditures and food shortages Income from crops Borrowing Assets Household labor outcomes � � Casual labor patterns Self-reported labor constraint Results Conclusion Introduction Conceptual framework Context and design Results Conclusion Part 1: Contract impacts Land use Endline cross section: yiv = αi + δCi + βxi + εiv Has cleared Total plots No. of plots Total trees Has acquired land in last 3 cleared in planted with across all new land yrs last 3 yrs trees plots since 2008 Contract (1) 0.200** (0.083) 0.112 (2) 0.543** (0.245) 0.104 (3) 0.526* (0.264) 0.191 R-squared Mean, no 0.311 0.745 1.943 contract Notes: N=205. *p<0.1 **p<0.05 ***p<0.01. (4) 17.955 (14.017) 0.109 (5) -0.041 (0.039) 0.178 26.481 0.179 Introduction Conceptual framework Context and design Results Conclusion Payment marginal effects: Land use By endogenous contract performance: 4 .6 Est'd coeff .2 .4 Has acquired new land since 2008 Est'd coeff 2 3 No. of plots planted with trees 1 0 0 -.2 0 4000 8000 12000 Total payout under contract (MWK) 0 4000 8000 12000 Total payout under contract (MWK) 0 4000 8000 12000 Total payout under contract (MWK) 0 -.2 Est'd coeff .2 .4 .6 .8 Has cleared land in last 3 yrs Introduction Conceptual framework Context and design Results Conclusion Part 1: Contract impacts Wealth Difference in difference (col 3-5): yivt = αi + δCi + αEndt + γCi × Endt + βxi + εivt Contract Per capita spending on food (1) -32.146 (31.872) Total income Past borrowing Months of from crop sales (0/1) food shortage (2) 436.577 (4806.237) End Contract!End R-squared 0.301 Baseline mean, no contract 112.881 N 195 Notes: *p<0.1 **p<0.05 ***p<0.01. 0.309 32000.000 205 (3) 0.039 (0.059) 0.437*** (0.067) -0.034 (0.096) 0.242 0.510 400 (4) -0.054 (0.243) -1.430*** (0.244) 0.243 (0.379) 0.274 3.630 410 Asset index (5) -0.031 (0.147) 0.158 (0.139) 0.073 (0.181) 0.312 0.899 410 Introduction Conceptual framework Context and design Results Conclusion Payment marginal effects: Wealth By endogenous contract performance: Asset index Est'd coeff .5 1 1.5 1 -.5 0 2 Est'd coeff 3 4 2 5 Months of food shortage .4 .6 Est'd coeff .8 1 1.2 1.4 Past borrowing (0/1) 0 4000 8000 12000 Total payout under contract (MWK) 0 4000 8000 12000 Total payout under contract (MWK) 0 4000 8000 12000 Total payout under contract (MWK) Introduction Conceptual framework Context and design Results Part 1: Contract impacts Household labor Difference in difference: Casual labor Stated labor Casual labor is a coping constraint income (0/1) strategy (0/1) (1) (2) (3) Contract 0.041 -0.007 -0.009 (0.038) (0.067) (0.032) End -0.246*** -0.105 -0.035 (0.047) (0.073) (0.028) Contract!End 0.092 0.182** 0.145** (0.071) (0.087) (0.053) R-squared 0.241 0.137 0.074 Notes: N=410. *p<0.1 **p<0.05 ***p<0.01. Conclusion Introduction Conceptual framework Context and design Results Conclusion Payment marginal effects: Household labor By endogenous contract performance: Est'd coeff .6 .8 .4 1 1 Casual labor is a coping strategy Est'd coeff .6 .8 0 4000 8000 12000 Total payout under contract (MWK) Casual labor income (0/1) .4 0 Est'd coeff .1 .2 .3 .4 Stated labor constraint (0/1) 0 4000 8000 12000 Total payout under contract (MWK) 0 4000 8000 12000 Total payout under contract (MWK) Introduction Conceptual framework Context and design Results Leakage through market channels Contracts awarded to up to 68 percent of households in a village Test village-level land and labor market spillovers � Random variation in the share of households that receive a contract � Control for village, household and individual-level characteristics Conclusion Introduction Conceptual framework Context and design Results Leakage through market channels Contracts awarded to up to 68 percent of households in a village Test village-level land and labor market spillovers � Random variation in the share of households that receive a contract � Control for village, household and individual-level characteristics Results: No evidence of impacts on labor market participation, wages or land transfers Conclusion Introduction Conceptual framework Context and design Results Conclusion Part 2: Allocation effects Land use Has cleared Total plots No. of plots Has acquired Total trees land in last 3 cleared in last planted with new land since across all plots yrs 3 yrs trees 2008 (1) (2) (3) (4) (5) Contract 0.215** 0.582** 0.560** 18.217 -0.052 (0.081) (0.237) (0.256) (14.483) (0.044) Auction 0.196** 0.482* 0.162 -2.339 -0.070 (0.078) (0.236) (0.279) (5.586) (0.044) Contract!Auction -0.138 -0.120 -0.024 -0.431 0.064 (0.119) (0.380) (0.331) (15.800) (0.067) Mean, no contract 0.393 0.942 1.895 24.802 0.132 Notes: N=433. *p<0.1 **p<0.05 ***p<0.01. Introduction Conceptual framework Context and design Results Conclusion Part 2: Allocation effects Wealth Contract Auction Contract!Auction Per capita spending on food (1) -32.736 (33.912) 59.988 (40.650) 64.609 (59.417) Total income Past borrowing Months of from crop sales (0/1) food shortage (2) 1095.708 (4668.174) 8608.354* (4826.427) 8853.264 (7405.957) End Contract!End Auction!End Contract!Auction!End Baseline mean, no contract 112.881 N 408 Notes: *p<0.1 **p<0.05 ***p<0.01. 32000 433 (3) 0.036 (0.058) 0.020 (0.041) -0.011 (0.078) 0.436*** (0.065) -0.033 (0.093) 0.057 (0.059) 0.021 (0.110) 0.522 841 (4) -0.066 (0.234) -0.313 (0.189) 0.719** (0.315) -1.430*** (0.240) 0.243 (0.374) -0.228 (0.328) -0.833* (0.448) 3.300 866 Asset index (5) 0.007 (0.144) 0.003 (0.086) 0.095 (0.141) 0.158 (0.137) 0.073 (0.178) 0.143 (0.160) -0.044 (0.208) 0.947 866 Introduction Conceptual framework Context and design Results Part 2: Allocation effects Household labor Casual labor is a coping strategy (1) Contract 0.038 (0.034) Auction -0.050 (0.042) Contract!Auction 0.020 (0.047) End -0.246*** (0.046) Contract!End 0.092 (0.070) Auction!End 0.064 (0.065) Contract!Auction!End -0.181* (0.099) R-squared 0.202 Baseline mean, no contract 0.84 Notes: N=866. *p<0.1 **p<0.05 ***p<0.01. Casual labor income (0/1) (2) -0.011 (0.066) -0.048 (0.062) 0.116 (0.086) -0.105 (0.072) 0.182** (0.086) 0.112 (0.086) -0.319** (0.124) 0.132 0.625 Stated labor constraint (0/1) (3) -0.017 (0.033) 0.009 (0.031) -0.018 (0.044) -0.035 (0.028) 0.145** (0.053) 0.084* (0.041) -0.065 (0.057) 0.056 0.06 Conclusion Introduction Conceptual framework Context and design Results Conclusion Conclusion Field experiment shows evidence of indirect effects of a PES contract � Within-farm leakage through land clearing � Increase stress on household labor supply � No evidence of overall livelihood impacts More efficient allocation (auction) appears to somewhat mitigate adverse indirect effects � Consistent with Jack (2013) result that auction leads to more tree survival Caveats to generalization � Results are specific to the contract and setting studied Introduction Conceptual framework Context and design Results Thanks Thanks to the Harvard Sustainability Science Programme, the International Growth Centre, ICRAF, USAID and the Harvard Weatherhead Center for International Affairs Conclusion