Skip to main content
. 2021 Sep 7;118(37):e2026386118. doi: 10.1073/pnas.2026386118

Table 2.

Reduced-form RD effects on high school persistence, by year

Dependent variable (1) (2) (3) (4) (5) (6)
Full sample Bandwidth |Grade 8 GPAi| ≤ 0.67
Estimate n Mean Estimate n Mean
Enrolled in district, year 2 −0.00987 1,405 0.974 0.0261 424 0.960
(0.0419) (0.050)
Attendance, year 2 5.741* 1,369 94.7 9.672*** 407 90.0
(3.065) (2.432)
Credits earned, year 2 8.014* 1,369 114.8 16.62*** 407 101.8
(4.794) (5.739)
Enrolled in district, year 3 0.00203 1,405 0.949 0.0511 424 0.910
(0.0516) (0.060)
Attendance, year 3 6.323** 1,333 93.1 9.677*** 386 87.3
(2.989) (3.540)
Credits earned, year 3 10.56 1,333 174.2 24.66*** 386 156.1
(6.570) (7.913)
Enrolled in district, year 4 0.0374 1,405 0.928 0.0822 424 0.868
(0.0606) (0.068)
Attendance, year 4 7.158*** 1,304 91.3 11.56*** 368 84.7
(1.799) (3.728)
Credits earned, year 4 15.29** 1,304 231.7 31.46*** 368 212.7
(7.724) (9.067)

Each cell contains the result of a separate regression of the effect of I(Grade 8 GPAi< 2.0) on high school persistence measures (i.e., district enrollment, attendance, and credits earned) from administrative data. All models include linear splines and a full set of demographic controls (i.e., student sex and race/ethnicity indicators) and grade 8 (i.e., prior to treatment status) special education identification, English learner status, attendance, and suspension history. Grade 8 GPA is centered at 2.0. Robust SEs are reported in parentheses. *P < 0.10, **P < 0.05, and ***P < 0.01.