Table 6.

Additional robustness and validity tests for the log of deaths.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Log COVID-19 Deaths
	Baseline Model	Limiting to Comparable Counties with 50%–70% Essential Workers	Limited to Counties with Population Density <2000 in the Treatment Group	Dropping States Near the Mexican Border With Possible Smoke Exposure	Limiting to Counties in States with Both Treated and Control Counties	Limiting to Only Essential TRIs	Limiting to Only TRIs Emitting Air Pollution
Treated Counties After the Rollback	0.1055∗∗∗ (0.0272)	0.1371∗∗∗ (0.0377)	0.0429∗ (0.0238)	0.1098∗∗∗ (0.0298)	0.0891∗∗∗ (0.0270)	0.2307∗∗∗ (0.0293)	0.1208∗∗∗ (0.0244)
With County, Month, and Day of Week Fixed Effects	X	X	X	X	X	X	X
Observations	84126	45236	79162	73155	82452	84126	84126

Notes: Columns 1–7 present the results for being in a county with 6 or more TRI sites after the EPA’s rollback with the log of COVID-19 deaths as the outcome. Column 1 replicates our results from Table 3. Column 2 presents the results when limiting to counties with similar percentages of essential workers. Column 3 presents estimates in which we drop treated counties with population densities of more than 2000. Column 4 presents results when we drop counties with potential seasonal smoke exposure. Column 5 presents results when the sample is limited to states with both treated and control counties. The results in column 6 limit the analysis to essential TRI sites. Column 7 limits the sample to include only TRI sites that emit air pollution. All models control for being after the rollback, social distancing, stay at home orders, re-openings, mask mandates, days since the first COVID death, total tests administered, daily number of confirmed COVID-19 cases, weather, and day of the week, county and month fixed effects. Standard errors are clustered at the county level and are in parenthesis. Coefficients labeled as ∗∗∗, ∗∗, and ∗ are statistically significant at the 1, 5, and 10 percent levels, respectively.