Abstract
Catastrophic events and the damages they cause have recently been paid more attention due to their increased severity and frequency. For society to better prepare for and ameliorate the consequences of these events, it is important to understand how people make decisions when faced with choices involving disaster mitigation strategies. This thesis examines the expected utility theory of insurance demand, insurance being a widely-used disaster mitigation tool, and uses an empirical demand estimation to test the predictions it makes concerning how the purchase of earthquake insurance varies with price, income, and risk. Expected utility theory is one of the most widely used decision theories of the past century, it can and has been used to model people’s choices under the type of uncertainty catastrophic events create, and its predictive validity is still uncertain. Three years of zip code level California Earthquake Authority (CEA) data are used to estimate income, risk, and price elasticities for residential housing earthquake insurance demand. In order to better deal with the disadvantages created by the time invariant risk data used, several models are estimated to study the robustness of the results across specification. A total of six models are estimated, including two pooled models, two time fixed effects models, and two earthquake zone fixed effects models. The first four models estimate elasticities for price, income, and risk. The two earthquake zone fixed effects models estimate elasticities for price, and income. The estimated elasticities are then compared to theoretical predictions of the elasticities. Both the price and income elasticities are predicted to be either positive or negative, the empirical results showing them negative and positive respectively. The predicted risk elasticity is positive, and the empirical results agree with the prediction. The empirical results find the signs of the elasticities of price, income, and risk robust over specification, but the magnitudes change from elastic to inelastic across some of the different models. The estimated risk elasticity changes significance depending on specification. One concern with the analysis was in how risk was measured. Every effort was made to characterize risk so inferences could be drawn from the analysis. In the future, researchers should try to obtain risk data that changes over time and across entity in order to more accurately test predictions about how risk plays a part in individual’s insurance purchasing decisions.