We examine day-to-day fluctuations in worker-level output at two manufacturing sites located in different industrial towns in China. Ambient air pollution in both towns, as proxied alternatively by fine-particle (PM2.5) and sulfur dioxide (SO2) levels, is severe but significantly variable, in part due to exogenous atmospheric ventilation. Across sites, pollution proxies and estimators, and correcting for worker heterogeneity and seasonality, we find a precisely estimated zero response on daily worker output from concurrent (same-shift) variation in air pollution. This is consistent with the stated perceptions of managers during our visits to over 10 firms in four provinces. We then follow the epidemiological literature and allow worker outcomes to respond to day-to-day variation in pollution with up to 30 days of delay. We uncover statistically significant adverse output effects from more prolonged pollution exposure, but these effects are not large. The cumulative effect – scaled for a large +10 μg/m3 PM2.5 variation in exposure on the day of output and on each of the previous 25 days – amounts to -0.5 to -3% of mean daily output. Estimates using SO2 are similar. Consistent with a simple model, more productive workers tend to respond more to pollution.