Association for Public Policy Analysis & Management

Households underinsure against natural disasters. In the last decade, about 70% of disaster losses were uninsured. Post-disaster, low-interest loans are an attractive policy alternative to direct grants. Because the loans must be repaid, they can potentially ease liquidity constraints and fund households' recovery at a lower taxpayer expense. We use comprehensive data on 1.2 million household applications to the U.S. federal disaster loan program, which offers low-interest loans to affected households, to consider two research questions.

First, we examine how a household's income affects the likelihood that it applies for and receives a disaster loan. We compare the incomes of applicants to other households in their ZIP codes. The median applicant is at the 55th income percentile for their ZIP code, those approved are at the 63rd percentile, and those declined are at the 46th percentile. Thus, while households throughout the income distribution apply, it is higher income households who receive loans.

Second, we examine the program for evidence of charity hazard in insurance demand. We show theoretically that insurance and post-disaster borrowing are partial substitutes – households who are more likely to receive a low-interest loan should insure less. We check this empirically by assessing the percent of applicants' losses that was insured. Counter-intuitively, we find borrowers from the program were more insured than declined applicants. Also, higher income applicants, who are more likely to be approved, insured more. Thus, we do not find evidence among applicants that the program reduces their insurance demand.

We show that borrowing after a severe loss is a strategy mostly available to higher income populations to supplement their insurance payments. Households with below-median income were less insured and more likely to be deemed un-creditworthy by the program. Our findings are poignant in the current environment of rising income inequality and increasing disaster severity.