Endogeneity testing using the Heckman treatment effect using a two-stage least square (2SLS) regression model: The relationship between politically-connected boards (PCBs) and over-(under-)investment in labor (OVER_LAB or UNDER_LAB)
| Panel A. 1st stage regression | Panel B. 2nd stage regression | |||||
|---|---|---|---|---|---|---|
| Variable | PCBs | OVER_LAB | UNDER_LAB | |||
| Coefficient | t-values | Coefficient | t-values | Coefficient | t-values | |
| #1 | #2 | #3 | #4 | #5 | #6 | |
| PCBsit | −0.189*** | (−3.87) | −0.007 | (−1.32) | ||
| AVG_BOARD_AGEit | 0.057*** | (4.82) | −0.036 | (−0.72) | 0.008 | (1.05) |
| AVG_BOARD_EDUit | 0.928*** | (5.32) | −0.080 | (−1.40) | 0.007 | (0.97) |
| TOP5it | −0.515 | (−1.31) | −0.379*** | (−3.67) | −0.015 | (−0.89) |
| LEVit | 0.514 | (1.39) | −0.016 | (−1.05) | −0.002 | (−0.60) |
| ROAit | 0.571 | (0.81) | 0.073** | (2.52) | −0.006* | (−1.84) |
| DPRit | 0.194 | (1.57) | 0.029** | (2.57) | −0.004** | (−2.20) |
| TANGit | 0.064 | (0.33) | 0.016** | (2.08) | 0.001 | (1.17) |
| OPER_CCit | 0.020 | (0.25) | 0.103*** | (3.14) | 0.004 | (1.28) |
| Qit | 0.102** | (2.50) | 0.302*** | (2.95) | 0.103*** | (7.54) |
| IMRit | −0.189*** | (−3.87) | −0.007 | (−1.32) | ||
| Constant | −5.618*** | (−5.15) | −0.036 | (−0.72) | 0.008 | (1.05) |
| Industry FE | Yes | Yes | Yes | |||
| Year FE | Yes | Yes | Yes | |||
| Observations | 2,442 | 1,429 | 1,013 | |||
| Wald Joint | 87.47*** | |||||
| Pseudo R2 | 0.176 | |||||
| F | 5.49*** | 11.78*** | ||||
| R2 | 0.293 | 0.424 | ||||
| Variable | PCBs | OVER_LAB | UNDER_LAB | |||
|---|---|---|---|---|---|---|
| Coefficient | Coefficient | Coefficient | ||||
| #1 | #2 | #3 | #4 | #5 | #6 | |
| −0.189*** | (−3.87) | −0.007 | (−1.32) | |||
| 0.057*** | (4.82) | −0.036 | (−0.72) | 0.008 | (1.05) | |
| 0.928*** | (5.32) | −0.080 | (−1.40) | 0.007 | (0.97) | |
| −0.515 | (−1.31) | −0.379*** | (−3.67) | −0.015 | (−0.89) | |
| 0.514 | (1.39) | −0.016 | (−1.05) | −0.002 | (−0.60) | |
| 0.571 | (0.81) | 0.073** | (2.52) | −0.006* | (−1.84) | |
| 0.194 | (1.57) | 0.029** | (2.57) | −0.004** | (−2.20) | |
| 0.064 | (0.33) | 0.016** | (2.08) | 0.001 | (1.17) | |
| 0.020 | (0.25) | 0.103*** | (3.14) | 0.004 | (1.28) | |
| 0.102** | (2.50) | 0.302*** | (2.95) | 0.103*** | (7.54) | |
| −0.189*** | (−3.87) | −0.007 | (−1.32) | |||
| −5.618*** | (−5.15) | −0.036 | (−0.72) | 0.008 | (1.05) | |
| Yes | Yes | Yes | ||||
| Yes | Yes | Yes | ||||
| 2,442 | 1,429 | 1,013 | ||||
| 87.47*** | ||||||
| 0.176 | ||||||
| 5.49*** | 11.78*** | |||||
| 0.293 | 0.424 | |||||
Note(s): Heckman two-stage model, t-statistics calculated based on the robust standard errors clustered at firm-level. Columns 1 to 4 report the regression coefficients and t-statistic values in parentheses. Panel A reports the first-stage probit regression between PCBsit with instrumental variables (AVG_BOARD_AGEit and AVG_BOARD_EDUit) and the control variables used in the second-stage regression. Panel B reports the second-stage regression results. Labor investment inefficiency is measured as over-investment in labor (OVER_LABit) and under-investment in labor (UNDER_LABit). The presence of PCBsit is a dummy variable set to 1 if firm i in year t has politically-connected board member(s) and zero (0) otherwise. Columns 3–4 report the OVER_LAB regression coefficients and t-statistic values in parentheses, columns 5–6 report the UNDER_LAB regression coefficients and t-statistic values in parentheses. *, **, and *** indicate statistical significance at the 10, 5 and 1% levels, respectively (two-tailed). The definitions of variables are presented in Appendix
Source(s): Table 6 by authors