Table 6.
Weak-Instrument Robust Confidence Bounds for the IV probit Cigarette Smoking Coefficients
| Model | IV-Probit Coefficient (Newey, 1987) | Conditional Likelihood Ratio (CLR) | Anderson Rubin (AR) | Lagrange Multiplier (LM) | LM-J |
|---|---|---|---|---|---|
| Smoking before pregnancy | |||||
| Instrument Specification 1 | 0.105** (0.046) [0.015,0.196] | [0.018,0.28] | [−0.043,0.521] | [0.025,0.263] | [0.02, 0.273] |
| Instrument Specification 2 | 0.121** (0.053) [0.028,0.323] | [0.03,0.319] | [−0.012,0.485] | [0.033,0.308] | [0.028,0.323] |
| Instrument Specification 3 | 0.114** (0.055) [0.006,0.221] | [0.011,0.357] | [−0.041,0.657] | [0.018,0.333] | [0.013,0.352] |
| Instrument Specification 4 | 0.121** (0.058) [0.006,0.235] | [0.019,0.365] | [−0.01,0.512] | [0.021,0.36] | [0.016,0.38] |
| First Trimester Smoking | |||||
| Instrument Specification 1 | 0.19** (0.092) [ 0.01,0.371] | [0.018,0.703] | [−0.068,1.633] | [0.036,0.616] | [0.027,0.66] |
| Instrument Specification 2 | 0.21** (0.103) [0.009,0.411] | [0.038,0.765] | [−0.017,1.361] | [0.044,0.724] | [0.035,0.785] |
Note: The table reports the IV probit smoking coefficients estimated using the minimum chi-square distance estimator (Newey, 1987) and 95% weak-instrument robust confidence bounds based on the CLR, AR, LM and LM-J statistics. The usual standard errors and confidence bounds for the smoking regression coefficients are also reported in parentheses and brackets, respectively before the regression coefficient. The models are based on a sample size of 856 observations.