duke university, sanford school of public policy

environmental & natural resource economist

(919) 613-9240


New Drafts

Land Use & Climate Mitigation
Water & Climate Adaptation
Air Quality & Energy Use
Information & Audits

Land Use & Climate Mitigation

Global & Comparative

Pfaff, A.*, E.O. Sills, G.S. Amacher, M.J. Coren, K. Lawlor, C. Streck (2009). “Policy Impacts on Deforestation: lessons from past experiences to inform new initiatives”. Copenhagen Draft with Executive Summary Reviews literature on international and domestic conservation and development policy impacts on forest loss (for REDD but more evidence on deforestation than on degradation).

Joppa, L.* and A. Pfaff*. “Global Park Impacts: how could protected areas avoid more deforestation?” . Protected areas are the leading conservation tool and are likely to play a role in climate policy too. Almost all prior assessments of PA impacts do not sufficiently address the fact that PA networks are not randomly distributed across their national landscapes. Specifically, parks tend toward land that if unprotected is less likely than average to be cleared. Thus prior analyses overstate average PA impact. We show this across all 147 countries with >100 km2 of PA area by using ‘matching’ to control for effects of land characteristics. This global extension confirms the few smaller-scale studies that control for land characteristics, finding that this reduces park-impact estimates by half or more.

Delgado*, C., D. A. Conde, J. O. Sexton, F. Colchero, J.J. Swenson and A. Pfaff. “Deforestation dynamics in response to the evolution of the Western Amazonian Inter-Oceanic Highway” The Inter-Oceanic Highway (IOH) has grown from a dirt road to the critical first corridor through western Amazonia. While roads have long been linked to tropical deforestation, evaluating their effects requires understanding large variations in local impacts. In the tri-border region of Brazil, Peru and Bolivia, satellite images reveal rising clearing rates over 1989-2007. In Bolivia, clearing has accelerated as urban centers sprawl toward the Brazilian IOH. In Brazil, where development of the IOH began in the1980's, recent paving in cleared areas had no significant influence and growing secondary roads were the new drivers. Peruvian deforestation has increased slightly, yet the IOH is not yet complete. Our disaggregation of impacts in Brazil indicate that paving the IOH within Peru will accelerate deforestation significantly.

Conde, D.A.* and A. Pfaff*.  “Sequenced Road Investments and Clearing of the Mayan Forest”. Examines deforestation following road investments in the Mayan Forest in southern Mexico, Guatemala and Belize. Finds that impacts of new roads investments vary according to prior levels of development, with lower impacts far from previous development investments, though in these locations the new roads most strongly shape where the clearing occurs.

Brazilian Amazon

Pfaff A.*, R. Walker, S. Perz, W. Laurance, C. Bohrer, J.A. Robalino, S. Aldrich, E. Arima, M. Caldas and K. Kirby. “Roads & Deforestation in the Brazilian Amazon”.A recent study suggests that road expansion in areas of Amazonia that had already experienced significant development actually reduced deforestation rates. Our more spatially detailed data for early road investments and deforestation show not only that in the great majority of Amazonia new roads significantly accelerated deforestation but also that only in highly deforested areas were the impacts of new roads not significant (and they never diminish the rate of forest loss). The durable concept is that road impacts vary with local conditions. However, short-run impacts appear to be lower not only where prior development and clearing are high but also were they are very low. Short-run impact is actually highest in between and we discuss why.

Pfaff, A.* , J.A. Robalino*, S. Perz, R. Walker et al.  “Development, Roads and Deforestation in the Braziliam Amazon: an empirical view across decades”.Greatly extending suggestive early period findings, using spatially detailed data across multiple periods of land-use responses we show that roads investments clearly accelerate deforestation. Further, we show that the magnitude of that effect varies with prior development and clearing in ways that differ from those suggested in the literature.

Pfaff, A.*  and J.A. Robalino* et al.  “REDD Roads? Spatial dynamics and spatial variation in causal impacts” We use matching on much more spatially detailed data than in past work to examine new roads’ impacts on deforestation given the likelihood of partial adjustment and endogenous development. Prior development affects the baseline deforestation rates, consistent with other investments responding to early access and then driving development. Failure to control for this history yields spurious results. Using exact matching to control for this, prior development affects new road impacts. New-road impacts in the short run were relatively low for low prior development and high prior development. In between these bounds, road investments immediately raise deforestation significantly.

Pfaff, A.* , J.A. Robalino* , D. Herrera et al.  “Amazon Parks ” Analyzes protected area impacts on deforestation using matching to improve greatly the 'apples to apples' comparison for broad average results. They suggest the potential increase in deforestation impact from targeting lands under higher pressure.

Delgado*, C., A. Pfaff* et al..  “Will Nearby Protected Areas Constrain Road Impacts On Deforestation?” Examines the impact upon deforestation rates of the Chico Mendes Extractive Reserve in comparison with other protected areas in the state of Acre in the western Brazilian Amazon. The latter suffer less clearing but prevent less deforestation since they face little pressure, while Mendes has significant impact.

Costa Rica

Robalino J.A.* and A. Pfaff. “Contagious Development: neighbors’ interactions in deforestation”.working paper We estimate neighbor interactions in deforestation in Costa Rica. To address simultaneity and the presence of spatially correlated unobservables, we instrument for neighbors' deforestation using the slopes of neighbors' and neighbors' neighbors' parcels. We find that neighboring deforestation signicantly raises the probability of deforestation. Correct estimation of the interaction reverses the naive estimate's prediction of multiple equilibria while still implying significant effects on the amount and spatial pattern of deforestation. Policies for agricultural development or forest conservation in one area will affect deforestation rates in non-targeted neighboring areas.

Pfaff, A.* , J.A. Robalino* and G.A. Sanchez-Azofeifa.  “Payments for Environmental Services: empirical analysis for Costa Rica 1997-2000”.working paper Evaluating its impact using the deforestation observed in matched untreated areas, we find that Costa Rica’s PSA program of payments for environmental services had little effect on 1997-2000 forest clearing. Reasons include: a low national rate of deforestation; no targeting of those locations more likely to change land usage; and a goal of transferring surplus to landowners. This pioneering effort could save much of its budget, or greatly increase forest impact from current funds, if it could avoid enrolling lands which would remain forested even without such payments.

Robalino, J. *, A. Pfaff *, A. Sanchez, F. Alpizar, C. Leon, and C.M. Rodriguez. “Deforestation Impacts of Environmental Services Payments: Costa Rica’s PSA Program 2000-2005”.  working paper Costa Rica’s PSA environmental services payments were the pioneers and evolved over time. We evaluate whether a change in implementation over time changes impacts on deforestation. For forest-protection contracts during 2000 and 2005, we find that less than 5 in 1000 (about 0.4%) of the parcels enrolled would have been deforested annually without payments. This is slightly higher than in 1997-2000. The shifts in PSA implementation eliminated the bias of the PSA payments towards lands that are relatively unprofitable. Thus we see that within the same country for the same policy, the details of implementation matter.

Robalino, J.A.*  and A. Pfaff*  et al. “Spatial Spillovers from Parks in Costa Rica”. Mimeo, CATIE. Given concerns about 'leakage', or displaced deforestation which would lower the total impact of a conservation policy such as protected areas, we add to the very limited empirical literature on spatial spillovers.

Kerr S.* & A. Pfaff* . “Development and Deforestation: evidence from Costa Rica”.In light of significant focus upon 'additionality' of observed land use (with implications for greenhouse gas emissions), we examine deforestation over time in Costa Rica, adding to the limited analyses of land-use change over decades within a given country.

Pfaff A.*  and S. Kerr. “What Would Have Happened? reviewing and improving estimated baselines for tropical forests and sequestered carbon”. Following the ever increasing focus on rigorous evaluation of the impacts of forest conservation actions, we review what has been learned concerning small-scale projects and discuss the potential for larger-scale alternatives.

Kerr S.*, J. Hendy*, S. Liu* and A. Pfaff. “Tropical Forest Protection, Uncertainty, and Environmental Integrity of Carbon Mitigation Policies”. Many policy makers and negotiators fear that if credits for avoided deforestation are allowed to count for compliance, the environment will suffer because the credits will not reflect truly additional carbon storage. This paper considers uncertainties involved in estimating carbon stocks and predicting deforestation. We build an empirically based, stochastic, model that combines data from field ecology, GIS data from satellite imagery, economic analysis and ecological process modeling to simulate the effects of these uncertainties on the environmental integrity of credits for avoided deforestation.

Water & Climate Adaptation

(still under construction)


Air Quality & Energy Use


Mueller, V. *, A. Pfaff*, J. Peabody, Y. Liu, T. Riddell and K.R. Smith. “Evaluating Impacts of Improved Stoves in China on Self Reported Health Using Matching”. (presented at ASSA 1/08, AERE 6/08) Demonstrates that typical stove-impact evaluations can steer policy decisions wrong by generating the wrong rankings of impact across options.


Information & Audits


Yin, H. *, A. Pfaff and H.C. Kunreuther (2010). “Can Environmental Insurance Succeed Where Others Fail? the case of underground storage tanks”. Mimeo, University of Pennsylvania. Environmental insurance can achieve social efficiency even when two traditional policy instruments – ex post fines and risk-management mandates with ex ante fines – do not. The evolution of Underground Storage Tank programs suggests that despite the hurdles for novel insurance products, mandating environmental insurance can succeed. Insurance can avoid bankruptcy problems and the burden on a regulator is low if all it has to do is confirm that the firm has insurance.