Specific Objectives of the Project
We will predict the probability of crop disasters and large crop insurance payouts in response to weather shocks and show how global climate change might amplify such conditions. Virtually the entire existing literature examines how weather affects mean yield. Mean yield is a key outcome measure for some questions such as the effect of weather on global food prices. However, farmers and policy makers concerned about crop insurance and disaster payment are more interested in predictions of the frequency with which yields fall below the critical levels that trigger crop insurance or disaster payments or cause financial disaster. That is, they are not concerned about how weather affects the mean (average) outcome; they care about how it alters the lower part of the distribution of yield outcomes. We will use a novel approach to estimate these tail events reliably. This approach has not previously been used to analyze agricultural outcomes.

Project Report/Summary of Results
Corn and soy crop yields vary nonlinearly with both temperature and precipitation. Using county-level data over many decades, we estimate the joint distribution of yield, temperature, and precipitation conditional on other factors that affect yield and then infer the conditional yield distribution given temperature and precipitation. This nonparametric approach allows us to avoid arbitrarily imposing rigid structural relationships as in traditional regression analyses and fits the data much better both within sample and out of sample. Moreover, this approach is much more likely to correctly identify the likelihood of catastrophic crop yields in response to adverse weather conditions.