Union membership and the wage gap between the public and private sectors: evidence from China

Journal for Labour Market Research

Table 4 Robustness checks using different models

	(1) Nation	(2) Public	(3) Private
(1) Model1: Selection bias adjusted model
Union	0.051***	0.036	0.062***
	(2.75)	(1.15)	(2.73)
Public	0.117***
	(5.25)
Sector selection term	− 0.059	− 0.108	0.371**
	(− 0.25)	(− 0.28)	(2.44)
Union selection term	− 0.860***	− 0.646***	-1.366***
	(− 9.83)	(− 3.65)	(-6.46)
Control variables	Yes	Yes	Yes
No. of observations	14,084	5,496	8,588
R-squared	0.264	0.270	0.260
(2) Model2: LVt-2 model
Union	0.088**	0.067	0.098*
	(2.52)	(1.51)	(1.71)
Public	0.013
	(0.54)
Control variables	Yes	Yes	Yes
No. of observations	3997	1801	2176
R-squared	0.255	0.214	0.308
(3) Model3: FE model
Union	0.036	− 0.026	0.086
	(1.22)	(− 0.72)	(1.39)
Public	0.024
	(0.63)
Control variables	Yes	Yes	Yes
No. of observations	14,084	5496	8588
No. of groups	8242	2934	5701
R-squared within	0.158	0.200	0.131
Between	0.075	0.026	0.007
Overall	0.080	0.027	0.004

The t-values are in parentheses. FE: fixed effects model; LVt_2: lagged term (time t-2) of union dummy variable is used. The control variables include demographic factors (years of schooling, year of work experience, health status, sex, ethnicity, urban hukou, married), work factors (CPC member, occupation, industry sector), province-level region and survey year (2012, 2014, 2016, and 2018) dummies, which are not presented in the table. Parents’ work status, parents’ CPC membership dummies were used in first-stage estimation of sample selection bias adjusted model (Dubin and Mc Fadden, 1984). These results are available upon request. The level of standard error clustering is at the individual level in the FE model. Source: Calculated using data from the CFPS of 2010, 2012, 2014, 2016 and 2018.
***p < 0.01.
**p < 0.05.
*p < 0.1