Notes: Table reports coefficients and standard errors from regression of 2005–2009 and 2001–2005 changes in log firm gross output on 2004–2008 and 2000–2004 changes in the air transport access index. Air transport access is sum of inverse travel time weighted census 2000 county populations. Travel times imputed as described in text. Data from Annual Survey of Industrial Firms 2001, 2005 and 2009, aggregated to county or two-digit industry-by-county cells. Standard errors clustered on nearest airport (approximately 106 clusters). All regressions include controls for firm log employment, log employment squared, log fixed assets, log fixed assets squared and log employment × log assets. All regressions include controls for initial nearest airport × time period fixed, effects based on year 2000 airports (approximately 222 in full sample, approximately 107 in restricted sample). Initial access trend control is the air transport access index in the first year of each period. Census and geographic controls are: 2000 census proportion high school educated proportion unemployed, proportion disabled; minority ethnic county dummy; dryness, days above 10°C, mean temperature, elevation, precipitation, straight line distance to coast, distance to provincial capital, distance to national border, urban district dummy. Pre-2000 population trend dummies are for 40 quantiles of the distribution of the 1990–2000 census change in population. Province dummies: dummies for each Chinese province (up to 31 in restricted sample). Incidentally treated geographical subset is subset of counties for which absolute value of difference between nearest airport distance at beginning of period (2000, 2004) and distance to new airports constructed during time period (2000–2004 or 2004–2008) is <60% of the distance at beginning of period. Gross output is nominal. Values deflated using province-year consumer price indices yield almost identical results. Significance: ***1%, **5%, *10%.