The Impact of Hurricane Harvey on Psychosocial Outcomes: A Quasi-Experimental Investigation

Background: Extreme weather events are occurring more frequently and with greater intensity. These events entail acute stressors, exposure to these events are associated with negative psychosocial consequences for individuals who experience them, including poor mental health outcomes, increased health risk behaviors, and disruption to social networks. However, this literature is limited by methodological factors, including convenience samples, lack of baseline data, retrospective data, and lack of appropriate comparator groups. Hurricane Harvey made landfall in Texas in August 2017, and to date remains tied with Katrina as the most expensive storm in U.S. history. Guided by the Social Vulnerability Model, we use a quasi-experimental approach to more robustly estimate the population-level psychosocial effects of the storm.

Methods: We use data from the Behavior Risk Factor Surveillance Survey (BRFSS), a dataset that contains representative samples of adults (18+) at the Metropolitan Statistical Area (MSA) level. We included the data from the 2012-2019 waves of the BRFSS in eight MSAs in Texas. The exposed group were five MSAs that had a major disaster declaration from FEMA for Hurricane Harvey (Austin, College Station, Corpus Christi, Houston, and San Antonio). The control group were three MSAs that did not (Dallas, El Paso, and Fort Worth), for a total analytic sample of 66,631. Four self-reported outcomes were examined: 14+ days of poor mental health, poor self-rated health, binge drinking, and cigarette use. A difference-in-differences approach was used to assess the average marginal effects of residing in an exposed area as compared to a non-exposed area. Triple interactions were used to assess for differential impacts by sociodemographic factors.

Results: All results are presented as percentage point changes. Adults exposed to the hurricane had larger increases in poor mental health (1.62), compared to those in non-exposed areas (1.02) (interaction p=0.04). A significant triple interaction (p=0.03) revealed White adults exposed to Hurricane Harvey had larger increases in poor mental health (0.023) compared to those in non-exposed areas (-0.006). In contrast, no differences in changes in poor mental health were found for non-White adults. While adults with some college or more in exposed areas had larger decreases in cigarette use (-0.023) than those in non-exposed areas (-0.014) (triple interaction p=0.008), there were no differences in changes in smoking among adults with a high school degree or less.

Discussion: The use of more robust estimation methods showed significant, although modest, adverse mental health effects for areas exposed to Hurricane Harvey. The use of a quasi-experimental design allows for more confidence in a causal effect than the use of a purely observational design. However, there were no significant changes to self-rated health or alcohol/tobacco use, suggesting the importance of highlighting the role of resilience following disasters. The results largely did not support the Social Vulnerability Model, with most impacts appearing to affect individuals equally across sociodemographic lines, although education may be protective for the effects on health risk behaviors.