Forests provide ecosystem services at a variety of scales, from local to global. Recent attention has focused on forests' potential to mitigate climate change because of their ability to store carbon. They also provide substantial local benefits, as trees can mitigate pollution, reduce extreme temperatures, and enhance psychological well-being. Unfortunately, threats such as logging and commodity agriculture have led to substantial deforestation of primary forest. Other factors such as drought, fire, and insects also pose a threat, impeding the provision of ecosystem services and undermining climate mitigation potential. In order to design policies that can effectively deliver on the promise of forests, better evidence is needed to 1) quantify the social costs and benefits that forests provide; and 2) understand how policy can be best designed to support both forest ecosystems and local actors.
Advances in earth observation have made more data detailing the dynamics of land cover and land use change available than ever before. In response, a growing body of work has emerged that integrates econometric methods of causal inference with these big data in order to evaluate the effectiveness of various policy designs. This has become particularly true in the context of forest conservation, where these quasiexperimental impact evaluations are increasingly used to inform new policy. However, factors inherent to their structure such as measurement error and irreversibility may affect the performance of common econometric approaches. Further work is needed to help researchers grapple with how these data can be best integrated with econometric methods of causal inference, thereby providing more informative insight into policy design.
This dissertation seeks to answer three distinct but intertwined questions. In the first chapter, my co-author, Robert Heilmayr, and I ask how binary and irreversible data, a structure found in most deforestation datasets, affects the performance of panel econometric estimators. While the application of quasiexperimental impact evaluation to remotely sensed measures of deforestation has yielded important evidence detailing the effectiveness of conservation policies, researchers have paid insufficient attention to structure of these deforestation datasets. We use analytical proofs and simulations to demonstrate that many commonly employed panel econometric models are biased when applied to binary and irreversible outcomes. The significance, magnitude and even direction of estimated effects from many studies are likely incorrect, threatening to undermine the evidence base that underpins conservation policy adoption and design. To address these concerns, we provide guidance and new strategies for the design of panel econometric models that yield more reliable estimates of the impacts of forest conservation policies.
The second chapter asks how policymakers can best design forest policy in order to capitalize on the potential of forest-based climate solutions, while supporting livelihoods. In order to address the intertwined challenges posed by climate change, biodiversity loss, and rural poverty, policymakers throughout the world have begun to adopt policies that pay private landowners to protect or restore forests. However, relatively little evidence exists documenting the impacts these payments have had, and how incentives can be best structured to achieve multiple objectives. I evaluate the land cover impacts of a forest restoration subsidy included in Chile's Native Forest Law, which prioritized the participation of rural smallholders and indigenous communities. I find that 68.12% of landowners who had applied for the subsidy did not comply with their stated forest management commitments. However, verification protocols included as part of the conditional cash transfer program prevented nearly $30 million USD in unconditional transfers to these non-compliant landowners. Compliant landowners who were paid for forest restoration did expand native forests on their properties relative to a robust counterfactual. I find that the program has expanded Chile’s native forests and paid the average landowner an estimated $36.78 USD per tonne CO2 stored in aboveground biomass. In contrast to many studies on avoided deforestation, complying smallholders in regions of high poverty generated the greatest tree cover gains per enrolled hectare. These findings illustrate that, in contrast to payments for avoided deforestation, targeting for social development may enhance the environmental effectiveness of payments for restoration.
In the third and final chapter, I ask whether ecosystem degradation, specifically invasive species induced tree cover loss, has impacts on education outcomes in metropolitan United States. I leverage variation from the introduction of an invasive insect that exclusively targets ash trees, the emerald ash borer, to the Chicago Metropolitan area. Exploiting the staggered and idiosyncratic spread of the borer, I show how tree canopy cover and education outcomes were affected using difference-in-differences methods robust to general treatment effect heterogeneity. My findings indicate that ash borer infestation reduced canopy cover in affected areas by 1.4% on average, stemming from both increases in tree cover loss and declines in tree cover gain. Further, the ash borer reduced standardized test performance at exposed schools. Infestation exposure led to an average 1.86% fewer students meeting or exceeding the state benchmark at the typical school, with impacts concentrated among low-income students. This paper shows that invasive species can substantially impact ecosystem service provision and ultimately, education outcomes, adding to the damages known to be caused by human-induced environmental change.
