Collaborative Research: Flexible Regression Methods for Climate Economics

Princeton University, Princeton NJ

Investigators

Abstract

Effective climate policy design requires reliable methods to measure the economic impacts of climate change. An interest in climate and environmental economics is the effect of additional exposure to different temperature levels on economic outcomes. Because damage from increased exposure tends to be most severe at extreme temperatures, correctly capturing these effects is important to inform climate and environmental policy. Existing methods use year-to-year changes in temperature and other climatic variables to measure the impact of these climatic variables on various outcomes, thus are not able to account for these extreme temperatures. This award funds research that will develop flexible regression methods that account for the extremes in exposure and thus offer new methods to accurately measure the effects of climate change on economic outcomes. The research will also study other flexible regression methods that build on the new flexible methods developed in this study. The researchers will develop statistical software that will be used to implement the new methods thus making them widely useful to applied researchers. The new methods will improve climate research quality as well as enhance research in other social sciences. The research results will also improve climate and environmental policy making and implementation as well as help to establish the US as a global leader in climate economics. Binned regression models used to estimate and draw causal inference in mixed frequency panel data in climate economics have many drawbacks that have not been addressed despite their ubiquitous use in climate economics research. This award will fund research to develop a comprehensive toolkit for understanding the flexible binned regressions methods with mixed-frequency panel data method, and generalizations thereof. The unique technical and methodological challenges posed in this approach will be addressed as part of this grant. In particular, the research will demonstrate the empirical and theoretical limitations with current approaches that can potentially lead to incorrect conclusions. It will also develop more robust, data-driven estimation and inference methods. The theoretical work will also contribute to the literature on non-parametric estimation and inference for mixed-frequency panel data. The contributions of the research under this award are expected to include several novel mathematical statistics results. Finally, the research will develop general purpose statistical software, train graduate students, and produce general interest review articles. The new methods will improve climate research quality and policy as well as enhance research in other social sciences. The research results will also help to establish the US as a global leader in climate economics. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.