Dependent Variable: Dropout (Not Taking ENLACE Exam) | (1) | (2) | (3) | (4) | (5) | (6) | (7) |
---|---|---|---|---|---|---|---|
Score ≥ cutoff | 0.092*** | 0.095*** | 0.105*** | 0.107*** | 0.093*** | 0.108*** | 0.106*** |
(0.0169) | (0.0164) | (0.0173) | (0.0165) | (0.0167) | (0.0178) | (0.0171) | |
[0.00] | [0.00] | [0.00] | [0.00] | [0.00] | [0.00] | [0.00] | |
(Score ≥ cutoff) * (Middle school GPA) | −0.048*** | −0.064** | −0.052** | ||||
(0.0179) | (0.0266) | (0.0212) | |||||
[0.03] | [0.00] | [0.00] | |||||
(Score ≥ cutoff) * (Parental education) | −0.005 | −0.008 | |||||
(0.0044) | (0.0052) | ||||||
[0.21] | [0.16] | ||||||
(Score ‡ cutoff) * (Change in commute due to admission) | 0.005*** | 0.007*** | 0.005*** | ||||
(0.0020) | (0.0022) | (0.0020) | |||||
[0.03] | [0.04] | [0.02] | |||||
(Score ‡ cutoff) * change in mean HS peer COMIPEMS exam score due to admission) | 0.003 | 0.003* | |||||
(0.0019) | (0.0019) | ||||||
[0.03] | [0.00] | ||||||
(Score ‡ cutoff) * (Student resides in Federal District) | −0.026 | 0.004 | |||||
(0.0323) | (0.0421) | ||||||
[0.36] | [0.94] | ||||||
(Score ≥ cutoff) * (Middle school GPA) * (Change in commute due to admission) | −0.010*** | ||||||
(0.0029) | |||||||
[0.00] | |||||||
Observations | 20,259 | 20,614 | 13,567 | 13,589 | 20,281 | 11,435 | 13,550 |
Adjusted R-squared | 0.098 | 0.016 | 0.022 | 0.019 | 0.019 | 0.110 | 0.108 |
Mean of dependent variable 1 point below cutoff | 0.415 | 0.408 | 0.403 | 0.402 | 0.416 | 0.397 | 0.401 |
Bandwidth | 15.5 | 16.5 | 13.4 | 12.4 | 15.3 | 12.2 | 13.0 |
Notes: Estimates are from local linear regressions, including separate linear terms for each of the 16 IPN cutoff schools, cutoff school-year fixed effects, and covariates whose interaction terms are included in the regression. Piecewise-linear terms in centered COMIPEMS score are interacted with the corresponding demeaned covariate in each column. Column 7 includes the interaction of the demeaned middle school GPA and change in commuting distance variables, and this measure’s interaction with piecewiselinear terms in centered COMIPEMS score. The edge kernel is used in each regression and in computation of the corresponding optimal Imbens–Kalyanaraman bandwidths. Standard errors clustered at the admitted high school level are in parentheses. Wild cluster bootstrapped p-values, clustered at the centered COMIPEMS score level, are in brackets.
↵* p < 0.10,
↵** p < 0.05,
↵*** p < 0.01.