TABLE 4.
Propensity score matching.
| (A) Matching model | ||||
|
| ||||
| Dependent variable | Expert | |||
| SIZE | 0.119*** | |||
| (8.11) | ||||
| LEV | −0.101 | |||
| (−1.19) | ||||
| ROA | 0.577** | |||
| (2.21) | ||||
| CFO | 0.854*** | |||
| (4.05) | ||||
| AGE | −0.019*** | |||
| (−7.12) | ||||
| SOE | −0.185*** | |||
| (−5.13) | ||||
| TENURE | −0.044*** | |||
| (−8.63) | ||||
| IBig4 | −1.856*** | |||
| (−16.84) | ||||
| Cons | −2.426*** | |||
| (−7.22) | ||||
| Industry fixed effects | YES | |||
| Year fixed effects | YES | |||
| N (firm-years) | 32389 | |||
| Pseudo R2 | 0.129 | |||
|
| ||||
| (B) Comparison result of the mean value of the control variables after matching | ||||
|
| ||||
| Variable | Expert = 1 | Expert = 0 | t | p > |t| |
|
| ||||
| SIZE | 21.933 | 21.949 | −0.77 | 0.441 |
| LEV | 0.420 | 0.420 | −0.11 | 0.91 |
| ROA | 0.041 | 0.040 | 0.74 | 0.457 |
| CFO | 0.047 | 0.047 | −0.12 | 0.905 |
| AGE | 9.555 | 9.673 | −1.04 | 0.297 |
| SOE | 0.315 | 0.323 | −1.05 | 0.293 |
| TENURE | 3.975 | 3.944 | 0.66 | 0.511 |
| IBig4 | 0.013 | 0.015 | −0.91 | 0.363 |
|
| ||||
| (C) Regression result after matching | ||||
|
| ||||
| Dependent variable | (1) Restatement | (2) LAF | ||
|
| ||||
| SEASONED | −0.381*** | 0.098*** | ||
| (−2.95) | (8.40) | |||
| UNSEASONED | 0.424* | −0.052** | ||
| (1.83) | (−2.51) | |||
| SIZE | 0.099* | 0.349*** | ||
| (1.68) | (42.18) | |||
| LEV | 0.533 | 0.121*** | ||
| (1.57) | (3.34) | |||
| ROA | −3.869*** | −0.294*** | ||
| (−4.66) | (−3.35) | |||
| CFO | −1.945** | 0.205*** | ||
| (−2.43) | (3.16) | |||
| AGE | 0.029** | 0.005*** | ||
| (2.41) | (3.76) | |||
| SOE | −0.542*** | −0.048*** | ||
| (−3.49) | (−2.59) | |||
| TENURE | −0.091*** | 0.002 | ||
| (−3.59) | (1.06) | |||
| IBig4 | −2.061** | 0.848*** | ||
| (−2.11) | (10.58) | |||
| Cons | −3.665*** | 5.654*** | ||
| (−2.92) | (32.58) | |||
| Industry fixed effects | YES | YES | ||
| Year fixed effects | YES | YES | ||
| N (firm-years) | 14946 | 14946 | ||
| Adj. R2 (Pseudo R2) | 0.091 | 0.608 | ||
Standard errors are robust to heteroscedasticity and clustered at the firm level. T-statistics are in parentheses. ***, **, and * denote significance at the 1, 5, and 10% level, respectively.