Private provision of public goods: Applying matching methods to evaluate payments for ecosystem services in Costa Rica

Payments for environmental services (PES) is a recent policy innovation attracting attention in both developed and developing countries. In developing countries, PES remain poorly tested. Costa Rica was among the first to implement a national PES scheme in which substantial amounts of money have conditionally changed hands, thus representing an excellent laboratory for evaluating the causes and consequences of PES participation.In this dissertation, I estimate the causal impact of the Costa Rican Program of Payments for Environmental Services (PSA) on forest cover at two scales, using state-of-the-art matching methods. First, I estimate the impact of a PSA forest conservation contract on an individual parcel, during the initial years of PSA in the case study region of Sarapiqui. Second, I estimate the impact of PSA contracting at the census tract level for the entire nation of Costa Rica, considering both binary and continuous measures of contracts signed between 1998 and 2004.The parcel level analysis is based on a household survey of participants and nonparticipants in PSA, with the sample drawn from administrative records of PSA and the national land registry, combined with GIS data on biophysical land characteristics and land cover information from aerial photographs and satellite images. The national analysis is based on a combination of administrative data on PSA, census data on socio-economic characteristics, GIS data on biophysical land characteristics from several sources, and various measures of change in forest cover derived from classified satellite images, all organized at the census tract level.While this dissertation is motivated by the broad policy debate over the effectiveness of PES as a conservation policy tool, it aims to answer a relatively narrow empirical question: what has been the causal impact of PSA contracts on forest cover in Costa Rica? The maintained assumption of this analysis, and of the implementing agency in Costa Rica, is that more forest cover will generate more environmental services. Key hypotheses tested in this research are that (1) only landowners whose opportunity cost of participation is low have enrolled in PSA, and (2) enrollment in PSA has generated a net increase in the area of forest.The parcel level analysis suggests that farms that do not have good alternative uses on their land (because of steep slopes or poor soil quality) tend to be enrolled in the program. Matching results provide evidence that during the initial phase of PSA, the program did have a significant but small effect on forest cover, both reducing loss of existing mature forest (gross deforestation as reported by landowners) and increasing total forest cover (net deforestation as determined by remote sensing) on parcels with contracts signed early in the program.In the national analysis, I also find that participation in PSA is more intense in census tracts with worst soil quality and higher steepness. Using a binary definition of treatment (PSA vs. non-PSA), I find that PSA has a positive and significant impact on forest gain and net deforestation between 1997 and 2005 in the census tracts that contain at least one PSA forest conservation contract signed during the first eight years of the program. However, I cannot conclude that PSA contracting has reduced gross deforestation. Results also indicate that the size and significance of PSA's impact varies by region. Using a continuous definition of treatment, the national analysis suggests that protection intensity matters, even for the direction of PSA's impact on forest gain, forest loss and net deforestation. These results represent the first attempt in the conservation literature to estimate the causal impacts of direct payments for conservation with matching methods that recognize a gradient in intensity of protection.