Trump voters were significantly less likely to evacuate before Hurricane Irma devastated Florida

Floridians who voted for Donald Trump in 2016 were less likely than those who voted for Hillary Clinton to have evacuated before Hurricane Irma made landfall, according to a new study that examined GPS location data from more than 2.7 million U.S. smartphone users. The research was published in the journal Science Advances.

As Hurricane Irma approached Florida in 2017, conservative radio talk show host Rush Limbaugh questioned the severity of the category 5 storm, suggesting that media outlets were trying to create “fear and panic” to push a “climate change agenda.” Shortly afterward, conservative pundit Ann Coulter also expressed skepticism about the severity of Hurricane Irma.

The authors of the new research were interested in whether this widely-reported hurricane skepticism had impacted actual evacuation behavior.

“We had followed much of the news around hurricanes in 2017 with Harvey and Irma and had initially been interested in examining evacuation disparities using widescale smartphone data. Until now, most studies had relied on surveys with small samples or questionnaires about a hypothetical hurricane,” said study author Elisa Long, an associate professor at the UCLA Anderson School of Management.

By statistically analyzing anonymized smartphone location data, county-level hurricane alerts, the 2010 U.S. Census, and precinct-level results for the 2016 U.S. presidential election, the researchers were able to find that political partisanship was associated with differences in evacuation behavior among Florida residents in 2017.

They found that about 45% of Clinton voters in Florida evacuated Hurricane Irma compared to only 34% of Trump voters — a gap that was not present during Hurricane Harvey in Texas in August 2017 or during Hurricane Matthew in Florida in October 2016, before conservative media commentators introduced “hurricane skepticism.”

In the 10% of Florida precincts with the highest share of Trump voters, fewer than 29% of residents evacuated during Irma, as opposed to more than 40% in the 10% of precincts with the highest share of Clinton voters. The divide in evacuation behavior was found both in counties that received an official hurricane warning and in counties that received no such alert.

“Hurricanes and other natural disasters used to be largely non-political, but in recent years, there is both a growing mistrust of government-issued warnings, compounded by climate change skepticism. After conservative media outlets widely espoused some ‘hurricane conspiracy’ theories just before Irma, we see a wide gap (about 10-11%) emerge in the likelihood of evacuating an oncoming hurricane. This difference was not found during Harvey or Matthew,” Long told PsyPost.

This was true even after the researchers controlled for sociodemographic variables and even geographic factors such as distance to the coast and elevation above sea level.

“One possible explanation is: Maybe Democrats just tend to live near the coast or in flood-prone areas, and that’s why they evacuated at higher rates? To test this, we include very fine geographic controls (called a geohash-7) that are independent of any political or geographic boundary,” Long explained.

“Essentially, we look at people living in the same area roughly the size of two football fields with similar incomes, employment rates, demographics, etc but one lives in a Trump precinct and one lives in Clinton precinct. Here we find that the likely Clinton voter was 10-12% more likely to evacuate Irma than a nearby likely Trump voter. Given how close they live, the hurricane could not possibly hit the Clinton voter more aggressively, across the entire state of Florida.”

But the study — like all research — includes some limitations.

“Because we use smartphone location data, we of course cannot include everyone in our sample, particularly more vulnerable individuals like nursing home residents. But given that about 80% of Americans own a smartphone, we are able to include over 1 million people in our Florida sample,” Long said.

“Future questions should examine whether this ‘evacuation wedge’ has persisted in 2018-2020 as the media continues to report on hurricane skepticism. And with the wildfires devastating the West Coast, it would be interesting to examine if there are differential evacuations by political leanings.”

“We are also hoping that our general methodology on using smartphone data to track movements before, during, and after natural disasters, like hurricanes, could help first responders and relief workers target their message accordingly, and help with response efforts,” Long added.

The study, “Political storms: Emergent partisan skepticism of hurricane risks“, was authored by Elisa F. Long, M. Keith Chen, and Ryne Rohla.