Association for Public Policy Analysis & Management

Hurricane Harvey, which submerged 25–30% of Houston in August 2017, is arguably the most generalizable major flooding event – having affected a wide variety of income groups, both inside and outside of the designated 100-year flood plain, across an economically vibrant city. We examine the impacts of this disaster on household debt using a treatment intensity difference-in-difference design, comparing the administrative credit outcomes of Houston residents according to the degree of flooding in their census block.

We find substantial heterogeneity in outcomes among flooded residents according to floodplain status, income, and housing equity. Counter-intuitively, we find that credit scores rise across Houston after the storm, likely due to broad-brush forbearance. However, credit scores rise 1-2 points less among residents in the most flooded blocks. This is due to a temporary rise in delinquencies in lower-income areas outside of the floodplain.

Our most surprising finding is a reduction in student debt balances in flooded areas outside of floodplain during the 5 quarters since Harvey. Notably, we find a roughly $1,000 (or 4%) pay down in student debt among people with pre-existing student loans, lower housing equity, and from higher income areas. A plausible interpretation is that flooded households are using fungible, low-interest, government loans and grants to eliminate their higher-interest non-dischargeable debt rather than to rebuild.

We believe that this is the first paper to evaluate student debt outcomes following a natural disaster. This paper is also unique in that we use floodplain status as a proxy for the likelihood of receiving assistance primarily through flood insurance rather than through government programs. In contrast, those outside of the floodplain are likely to receive more assistance through SBA loans, IRS refunds, and FEMA grants – all of which do not duplicate compensation from flood insurance and are more fungible than flood insurance.