Table 3

Number of Test-Takers and Enrollment

	PSLE Data—Exam-Sitters		EMIS Data—Enrollment
	All years	2015 and 2016	Grades 4–6	Grade 6	Grade 7	Grade 7/Grade 6
	(1)	(2)	(3)	(4)	(5)	(6)
0–10th percentile in previous year	−2.039***	−1.676***	−0.773	−0.329	−1.646**	−0.028**
	(0.763)	(0.589)	(1.787)	(0.882)	(0.737)	(0.013)
10–20th percentile in previous year	−1.846***	−1.068**	1.929	0.865	−0.479	−0.008
	(0.635)	(0.423)	(1.334)	(0.659)	(0.480)	(0.010)
Diff-diff	Yes	No	No	No	No	No
Control lagged exam score	Yes	Yes	Yes	Yes	Yes	Yes
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Control mean	52.01	50.07	197.14	64.10	51.55	0.84
Observations	77,431	31,520	31,150	31,150	31,150	15,828
R²	0.910	0.953	0.968	0.930	0.931	0.127

Notes: Each column represents a separate regression. All specifications include flexible controls for lagged test scores and school and district-by-year fixed effects. Column 1 is estimated on outcomes from 2012–2016, including indicators for prereform associations between district-rank deciles and subsequent outcomes. Reported coefficients in that column correspond to the differential effect of being ranked in the associated decile of within-district performance in the post- (vs. pre-) reform period, compared to the middle six deciles. In Columns 2–6, data are restricted to postreform years 2015–2016 in which EMIS data are available; this does not allow for a difference-in-difference specification. In Columns 1 and 2 the outcome variable is the number students sitting the exam, using PSLE data. In Columns 3–6 outcomes are different constructions of enrollment, based on EMIS data. The outcome in Columns 3–5 is the number of students enrolled in the corresponding grade(s). In Column 6 the outcome indicator is Grade 7 enrollment in 2016 divided by Grade 6 enrollment in 2015. Standard errors are clustered at the district level.