Application of Difference-in-Difference Strategies in Finance: The Case of Natural Disasters and Bank Responses

Abstract

Natural disasters are negative shocks that can severely disrupt the communities in which they occur. Disasters like hurricanes, tornados, floods, wildfires, and earthquakes, moreover, can cause severe property damage, including damages to homes, businesses, and automobiles. An important issue that arises is whether branches of banks in communities affected by natural disasters raise deposit rates to attract additional deposits in response to any deposit withdrawals and to meet any increase in loan demand for the rebuilding that takes place. Studies of the responses of banks to natural disasters increasingly find it useful to rely on a difference-in-differences (DID) identification strategy. The contribution of our paper is to examine how different choices that can be made affect the empirical results. Importantly, as our empirical results indicate, the discretion that a researcher uses in terms of the choices made at different stages do indeed produce different conclusions about the impact of natural disasters on bank deposit rates. If we do not match branches affected by natural disasters with those in adjacent communities not affected the results indicate that natural disasters have a statistically negative effect on deposit rates without matching and with a low degree of matching. However, when we use a medium or a high degree of matching there is no statistically significant effect. Moreover, when we use two different matching methods, the results differ. In the case of PSM, we find a statistically significant effect, but no effect in the case of CEM.