Table 3 Policy effects derived from the seven methods based on education-stratified analysis and the inclusion of interaction terms
Method | Specification | Low-educated [95% CI] | High-educated [95% CI] | Interaction term [95% CI] |
|---|---|---|---|---|
Regression adjustment | Logistic regression, adjusted for gender | 0.647 [0.570, 0.734] (odds ratio) | 0.679 [0.550, 0.839] (odds ratio) | 0.953 [0.745, 1.218] (odds ratio) |
Propensity score matching | Matched on gender | −0.048 [-0.065, -0.031] (probability difference) | −0.020 [-0.031, -0.009] (probability difference) | Not applicable |
Difference-in-differences | Logistic regression | 0.666 [0.574, 0.773] (odds ratio) | 0.687 [0.530, 0.890] (odds ratio) | 0.970 [0.719, 1.307] (odds ratio) |
Fixed effects | Linear regression, adjusted for time | −0.044 [-0.051, -0.037] (probability difference) | −0.016 [-0.023, -0.009] (probability difference) | −0.029 [-0.039, -0.019] (probability difference) |
Instrumental variable | Probit regression | −0.050 [-0.063, -0.037] (probability difference) | −0.020 [-0.029, -0.011] (probability difference) | −0.036 [-0.057, -0.015] (probability difference) |
Regression discontinuity | Logistic regression around the income threshold | 0.678 [0.495, 0.929] (odds ratio) | 0.687 [0.483, 0.977] (odds ratio) | 0.987 [0.615, 1.583] (odds ratio) |
Interrupted time-series | Linear regression | −0.023 [-0.027,-0.020] (probability difference) | −0.005 [-0.008, -0.002] (probability difference) | −0.019 [-0.023, -0.014] (probability difference) |