Sensitivity of regression results for effect of ozone on respiratory illnesses to weather and copollutants
| 1 | 2 | 3 | 4 | 5 | |
|---|---|---|---|---|---|
| A. OLS | |||||
| Eight-hour ozone | 0.113** [0.023] | 0.130** [0.020] | 0.118** [0.023] | 0.097** [0.022] | 0.105** [0.027] |
| B. IV—First stage | |||||
| Boat traffic / 100,000 | 4.608** [0.029] | 4.733** [0.035] | 4.959** [0.030] | 4.037** [0.031] | 24.463** [0.351] |
| (Boat traffic /100,000)*distance | −0.198** [0.001] | −0.192** [0.001] | −0.207** [0.001] | −0.187** [0.001] | −0.571** [0.003] |
| C. IV—Second stage | |||||
| Eight-hour ozone | 0.454** [0.162] | 0.465** [0.168] | 0.435** [0.155] | 0.451** [0.171] | 0.543** [0.210] |
| Controls for weather | Y | N | Y | N | Y |
| Controls for copollutants | Y | N | N | Y | Y |
| Interactions with ozone | N | N | N | N | Y |
Notes:
↵* significant at 5 percent,
↵** significant at 1 percent. N = 1,927,187 in all regressions. Robust standard errors clustered by date in brackets. Dependent variable is the number of respiratory related hospital admissions per day, zip code, and age category. All regressions include age dummies, year-month dummies, day of week dummies, cubic day trend, and zip code fixed effects. “Weather controls” include maximum temperature, minimum temperature, precipitation, wind speed, humidity, and cloud cover. “Co-pollutant controls” includes carbon monoxide and nitrogen dioxide. For Column 5, ozone is interacted with weather and copollutant variables, and the coefficient shows the marginal effect of ozone evaluated at the mean of the weather and copollutant variables. All interactions are instrumented by boat arrivals and departures interacted with weather and copollutant variables, but only coefficients for boat arrivals and departures and distance from first stage are shown.