Table 5 Alternative instrument—Jinshi density effects on contemporaneous per capita wage.

	(1)	(2)	(3)	(4)	(5)
Panel A: First stage
Dependent variable:	Jinshi density in Ming-Qing(logged)
Distance to the Great Canal (logged)	−0.152*** (0.009)	−0.125*** (0.007)	−0.069*** (0.016)	−0.069* (0.035)	−0.082*** (0.016)
The number of Ming Courier station (logged)	0.586*** (0.022)	0.605*** (0.024)	0.495*** (0.018)	0.495*** (0.037)	0.418*** (0.121)
KP-F statistics	348.010	347.859	519.718	125.309	15.317
Panel B: Second stage
Dependent variable:	Wage per worker (logged)
Jinshi density in Ming-Qing(logged)	0.100*** (0.025)	0.085*** (0.027)	0.087** (0.035)	0.087** (0.039)	0.060*** (0.014)
Cluster	County	County	County	City	Province
Observations	250,059	250,059	250,059	250,059	241
Firm controls	No	Yes	Yes	Yes	Yes
Other controls	No	No	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes
Industrial fixed effects	Yes	Yes	Yes	Yes	Yes

Each Column in each panel represents a separate cross-sectional 2SLS regression. Panel A displays first-stage results of the 2SLS estimate, showing river distance to the bamboo/pine forest effects on the regional density. In Panel B, Column (1) reports the effects of Jinshi density effects to wage per worker without any covariates. We add firm controls and other controls stepwise in Columns (2) and (3). In Column (4), we adjust the cluster to the city level. In Columns (5), we report results of city-level 2SLS estimates. Covariates include rainfall and air pollution, nightlight in 2004, population density in Ming-Qing, the urbanization rate in Ming-Qing, distance to coast, agricultural sustainability, and terrain ruggedness. All 2SLS regressions are clustered at the county level. These estimated Jinshi density effects can be interpreted as the percentage changes in wages per worker with a percent change in Jinshi density. The KP F-statistic is the Kleibergen-Paap Wald rk F-statistic for weak identification in the first stage (Kleibergen and Paap, 2006). *** denotes significant at 1% level, ** denotes significant at 5% level, * denotes significant at 10% level.

Search