Private costs for environm ental goods provision in a
Private costs for environmental goods provision in a development context: lab and field tests of a novel cost-revealing mechanism Prepared for CBEAR-MAAP 2017 Samuel D. Bell Oregon State University October 15, 2017
Roadmap 1. Motivation • Setting • Research Objectives
2. Random Quantity Mechanism (RQM) • Features • Mechanics
3. Experimental Test of the RQM • Experimental design • Results
4. Smallholder Supply of Environmental Goods: Evidence from Agroforestry • Setting and experimental design • Results
Setting & Motivation • Payments for environmental services, with: • Heterogeneous private costs • Information asymmetries
• Three broad options to improve efficiency
• Target on characteristics correlated with private costs of provision • Procurement auctions for allocating contracts • Offer screening contracts tailored to distribution of private costs
• Better information on private costs can improve efficiency • Random quantity mechanism (RQM): • Truthfully reveals private costs • Can estimate heterogeneous treatment effects
Research objectives
1. Test performance of the RQM using induced costs in experimental lab; 2. Implement the RQM in a field setting; • smallholder agroforestry in Zambia
3. Assess the potential of the RQM as a contracting mechanism and/or research tool.
What is the RQM?
Cost-revealing extension to the Becker-DeGroot-Marschak mechanism
• Provides quasi-experimental variation in treatment (contract allocation) • Provides random variation in contract terms • Enables direct estimation of willingness to accept (WTA) across intensive margins using a repeat elicitation format Ø Allows estimation of cost structures and supply
• Enables estimation of heterogeneous treatment effects
RQM Procedure
Three steps: A. Transfer value -> B. Quantity offer -> C. Random quantity draw • Contract iff C ≤ B
Ø Optimal strategy is to offer quantity for which total cost of production equals transfer
• Similar decision task to the BDM (in ’reverse’). • Many of BDM characteristics apply to RQM
• Profit = transfer value - private cost of drawn quantity
Induced cost lab experiment
• Procurement setting - participants make production decisions based on induced costs of production and contract value/transfer • Repeat elicitation across five firms • 5 different cost structures, hypothetical homogeneous good.
• Participants made quantity offers in response to 3 randomly-drawn transfer payments within each firm • Each participant made 15 offers in total (3 per each of 5 firms)
• Market clearing once all offer rounds complete, to limit learning • One contract per firm implemented (random draw of transfer and Q) • 20 participants, $34 average earnings
Lab experiment results
Efficiency measures: aggregate and individual expected payoffs • 17 out of the 20 participant’s offers resulted in optimal expected payoffs • 79% of the pooled offers resulted in optimal expected payoffs
Conclusions: • Experimental test supports incentive compatibility • Cost revealing and efficient in expected payments • Caveat: sample size quite small
RQM field experiment
• Smallholder farmers (n = 223) in Zambia • Agroforestry tree planting - environmental good with public and private benefits • Test new mechanism in the field – does the RQM work? • Construct aggregate supply curves • Compare RQM against (simulated) procurement auction • Explore determinants of WTA
Field implementation
• Smallholder is informed of the contract payment and asked how many trees he/she would plant and maintain • Repeated for 5 different transfers [20, 40, 70, 100, 140] in ‘000 ZMK • One transfer randomly drawn, then quantity drawn from [12, 25, 37, 50, 75] • Quasi-random variation in contract terms
• Payment structure
• 50% of payment made for transplanting all seedlings (activity-based) • 50% paid pro-rata on survival 1 yr following planting (outcome-based)
• Partner NGO monitored transplanting and outcomes. • 60% of participants received contracts
Determinants of WTA • • • •
Regress quantity offers on survival expectations and observables, controlling for transfer values Survival expectations interesting - may affect effort and outcomes (direction ambiguous ex-ante) Higher survival expectations reduces WTA – i.e. lower cost farmers have higher expectations Possibility of targeting low-cost farmers on observables, if correlations strong
• Selling foods and crafts, Length of food insecurity (months) have significant positive effects. • Female headed household strong positive effect (lower WTA), female is negatively correlated and marginally significant. Intra-household decision making effect?
Comparison of:
• RQM aggregate supply (by transfer) • RQM aggregate supply (pooled) • Single-bid multi-unit auction aggregate supply
What if these supply curves are used to inform posted price? • Illustrative budget constraint $600:
RQM vs. Generalized Vickrey Quantity target: Matching the field experiment (3,677)
RQM vs. Generalized Vickrey
• RQM is a useful research tool, less effective as an allocation mechanism:
• Provides precise measure of WTA allowing direct, non-parametric estimation of supply (like auctions) • Allocates contracts with positive probability, allowing otherwise out-ofsample predictions and provides exogenous variation in contract allocation and terms, allowing estimation of heterogeneous treatment effects • Random variation in price paid, condition on meeting participation constraints • Helps in separating selection effects from causal impact of payments
Determinants of contract performance
Results – contract outcomes
• Largest transfer payment results in 1.2 to 3.4 extra trees surviving compared to smallest transfer • Largest quantity draw results in 2 to 3.9 fewer trees surviving
• Survival expectations have (small) negative effect on survival outcomes • Low cost farmers have lower survival. • “Premium” constructed to explore the impact of higher payments conditional on a given WTA. • Causal effect of payment: payments above minimum WTA significantly positively impacts outcomes.
Private costs for environmental goods provision in a development context: lab and field tests of a novel cost-revealing mechanism Prepared for CBEAR-MAAP 2017 Samuel D. Bell Oregon State University October 15, 2017
Roadmap 1. Motivation • Setting • Research Objectives
2. Random Quantity Mechanism (RQM) • Features • Mechanics
3. Experimental Test of the RQM • Experimental design • Results
4. Smallholder Supply of Environmental Goods: Evidence from Agroforestry • Setting and experimental design • Results
Setting & Motivation • Payments for environmental services, with: • Heterogeneous private costs • Information asymmetries
• Three broad options to improve efficiency
• Target on characteristics correlated with private costs of provision • Procurement auctions for allocating contracts • Offer screening contracts tailored to distribution of private costs
• Better information on private costs can improve efficiency • Random quantity mechanism (RQM): • Truthfully reveals private costs • Can estimate heterogeneous treatment effects
Research objectives
1. Test performance of the RQM using induced costs in experimental lab; 2. Implement the RQM in a field setting; • smallholder agroforestry in Zambia
3. Assess the potential of the RQM as a contracting mechanism and/or research tool.
What is the RQM?
Cost-revealing extension to the Becker-DeGroot-Marschak mechanism
• Provides quasi-experimental variation in treatment (contract allocation) • Provides random variation in contract terms • Enables direct estimation of willingness to accept (WTA) across intensive margins using a repeat elicitation format Ø Allows estimation of cost structures and supply
• Enables estimation of heterogeneous treatment effects
RQM Procedure
Three steps: A. Transfer value -> B. Quantity offer -> C. Random quantity draw • Contract iff C ≤ B
Ø Optimal strategy is to offer quantity for which total cost of production equals transfer
• Similar decision task to the BDM (in ’reverse’). • Many of BDM characteristics apply to RQM
• Profit = transfer value - private cost of drawn quantity
Induced cost lab experiment
• Procurement setting - participants make production decisions based on induced costs of production and contract value/transfer • Repeat elicitation across five firms • 5 different cost structures, hypothetical homogeneous good.
• Participants made quantity offers in response to 3 randomly-drawn transfer payments within each firm • Each participant made 15 offers in total (3 per each of 5 firms)
• Market clearing once all offer rounds complete, to limit learning • One contract per firm implemented (random draw of transfer and Q) • 20 participants, $34 average earnings
Lab experiment results
Efficiency measures: aggregate and individual expected payoffs • 17 out of the 20 participant’s offers resulted in optimal expected payoffs • 79% of the pooled offers resulted in optimal expected payoffs
Conclusions: • Experimental test supports incentive compatibility • Cost revealing and efficient in expected payments • Caveat: sample size quite small
RQM field experiment
• Smallholder farmers (n = 223) in Zambia • Agroforestry tree planting - environmental good with public and private benefits • Test new mechanism in the field – does the RQM work? • Construct aggregate supply curves • Compare RQM against (simulated) procurement auction • Explore determinants of WTA
Field implementation
• Smallholder is informed of the contract payment and asked how many trees he/she would plant and maintain • Repeated for 5 different transfers [20, 40, 70, 100, 140] in ‘000 ZMK • One transfer randomly drawn, then quantity drawn from [12, 25, 37, 50, 75] • Quasi-random variation in contract terms
• Payment structure
• 50% of payment made for transplanting all seedlings (activity-based) • 50% paid pro-rata on survival 1 yr following planting (outcome-based)
• Partner NGO monitored transplanting and outcomes. • 60% of participants received contracts
Determinants of WTA • • • •
Regress quantity offers on survival expectations and observables, controlling for transfer values Survival expectations interesting - may affect effort and outcomes (direction ambiguous ex-ante) Higher survival expectations reduces WTA – i.e. lower cost farmers have higher expectations Possibility of targeting low-cost farmers on observables, if correlations strong
• Selling foods and crafts, Length of food insecurity (months) have significant positive effects. • Female headed household strong positive effect (lower WTA), female is negatively correlated and marginally significant. Intra-household decision making effect?
Comparison of:
• RQM aggregate supply (by transfer) • RQM aggregate supply (pooled) • Single-bid multi-unit auction aggregate supply
What if these supply curves are used to inform posted price? • Illustrative budget constraint $600:
RQM vs. Generalized Vickrey Quantity target: Matching the field experiment (3,677)
RQM vs. Generalized Vickrey
• RQM is a useful research tool, less effective as an allocation mechanism:
• Provides precise measure of WTA allowing direct, non-parametric estimation of supply (like auctions) • Allocates contracts with positive probability, allowing otherwise out-ofsample predictions and provides exogenous variation in contract allocation and terms, allowing estimation of heterogeneous treatment effects • Random variation in price paid, condition on meeting participation constraints • Helps in separating selection effects from causal impact of payments
Determinants of contract performance
Results – contract outcomes
• Largest transfer payment results in 1.2 to 3.4 extra trees surviving compared to smallest transfer • Largest quantity draw results in 2 to 3.9 fewer trees surviving
• Survival expectations have (small) negative effect on survival outcomes • Low cost farmers have lower survival. • “Premium” constructed to explore the impact of higher payments conditional on a given WTA. • Causal effect of payment: payments above minimum WTA significantly positively impacts outcomes.
Private costs for environmental goods provision in a development context: lab and field tests of a novel cost-revealing mechanism Prepared for CBEAR-MAAP 2017 Samuel D. Bell Oregon State University October 15, 2017
