Gender wage gap in China: a large meta-analysis

Journal for Labour Market Research

Table 5 Meta-regression analysis of publication selection bias

(a) FAT-PET test (Equation: t = γ₀ + γ₁(1/SE) + v)
Estimator	OLS	Cluster-robust OLS	Multi-level mixed-effects RML	Cluster-robust random-effects panel GLS	Cluster-robust fixed-effects panel LSDV
Model	[1]	[2]	[3]	[4]^a	[5]^b
Intercept (FAT: H₀: γ₀ = 0)	− 2.8691^*** (0.480)	− 2.8691^*** (0.870)	− 2.7968^*** (0.820)	− 2.7870^*** (0.832)	− 2.7526^*** (1.013)
1/SE (PET: H₀: γ₁ = 0)	− 0.0721^*** (0.008)	− 0.0721^*** (0.015)	− 0.0721^*** (0.013)	− 0.0722^*** (0.013)	− 0.0737^*** (0.014)
K	1472	1472	1472	1472	1472
R²	0.335	0.335	–	0.335	0.335

(b) PEESE approach (Equation: t =γ₀SE + γ₁(1/SE) + v)
Estimator	OLS	Cluster-robust OLS	Multi-level mixed-effects RML	Random-effects panel ML	Population-averaged panel GEE
Model	[6]	[7]	[8]	[9]	[10]
SE	− 35.4277^*** (6.747)	− 35.4277^** (14.107)	2.2418 (7.122)	2.2418 (10.274)	− 16.4359^*** (6.397)
1/SE (H₀: γ₁ = 0)	− 0.0885^*** (0.006)	− 0.0885^*** (0.012)	− 0.0797^*** (0.012)	− 0.0797^*** (0.003)	− 0.0843^*** (0.012)
K	1472	1472	1472	1472	1472
R²	0.605	0.605	–	–	–

^aBreusch-Pagan test: χ² = 1301.60, p = 0.0000
^bHausman test: χ² = 0.55, p = 0.4598
Figures in parentheses beneath the regression coefficients are standard errors. Except for Model [9], robust standard errors are estimated. *** and ** denote statistical significance at the 1% and 5% levels, respectively