Anthony Fauci recently testified before the Senate about the COVID-19 crisis. If you were watching (I was not), you probably noticed that only some of those Senators present were wearing face masks.

Twitter noticed:

Dr. Anthony Fauci is about to testify before the Senate. Everyone in this camera shot is wearing a mask, except for Susan Collins.

Because wearing a mask to protect yourself and others from a deadly virus is now deemed partisan and anti-Trump. pic.twitter.com/t4K7cMl4YC

— Charlotte Clymer 🏳️‍🌈 (@cmclymer) May 12, 2020

… as did the New York Times. Many have also remarked on President Trump’s refusal to wear masks, even when looking at masks in a place that makes masks. The message seems to be clear: wearing a mask is a partisan decision, even as a bare majority of Republicans report that they wish that Trump would wear a mask when he travels.

Is that true outside of the Senate—do Americans’ decisions to wear masks depend on their partisan identity? As a continuing part my collaborative work on the politics of COVID-19 in the United States with Shana Gadarian and Sara Goodman, we recently asked a random, representative sample of 2400 Americans if they are wearing masks in public. Here is what we found from logistic regressions that adjust for a full set of dummies for age, race, gender, marital status, income, education, urban-rural, and state fixed effects.

Adjusting for those differences, Democrats are more than 20 percentage points more likely than Republicans to (75% versus 53%) to report wearing masks in public.

We can look to explain what’s driving this result by allowing the estimate of partisanship to differ by respondent characteristics: level of education, family income, urban or rural,* or state-level results from the 2016 presidential election.**

These results tell us a couple of things.

1. Partisan differences between Democrats and Republicans are largest among the middle range of incomes. They are smaller and usually statistically insignificant at the highest and lowest income levels.
2. Partisan differences between Democrats and Republicans are largest in urban areas. They are small and statistically insignificant in rural areas.
3. Partisan differences between Democrats and Republicans can be found across education levels, except for among those respondents who have not completed a high school degree.
4. Partisan differences can be found in Trump-supporting states as well as those who voted strongly for Clinton. However, mask-wearing levels are consistently lower across the board in states that voted strongly for Trump. (The five points plotted are the 10th, 25h, 50th, 75th, and 90th percentile of Trump’s 2-party vote share by state.)

Whatever drives these differences, wearing a mask is partisan now.

NOTES

* For education, income, and urban/rural, we allow for nonlinear relationships between each level of each variable and partisanship.
** Here, we model state-level Trump vote share as a continuous variable, and estimate state-level random effects.