APPENDIX C: ANALYTIC METHODS OF IMPACT ANALYSIS AND RESULTS FROM ALTERNATIVE MODELS
The analytic sample for our primary impact analysis excludes two alternative schools that
were part of the New Heights expansion, Ballou STAY and Roosevelt STAY. We did this
because the impact of New Heights may be different for students in alternative schools, and thus
including them in our main analysis may result in impacts that
are less generalizable to
traditional schools. The estimated impact of New Heights on students in our primary sample and
in STAY schools results in a somewhat smaller impact on unexcused absences—4.2 fewer
unexcused absences per semester compared to 4.5—and somewhat larger impacts on days
attended—4.3 days per semester compared to 3.4 (Table C.4). The impacts on excused absences
per semester and the semester graduation rate are not significant when we
include STAY
schools, whereas the impact on credits earned per year is similar to our primary impact estimate.
Our primary analysis considers a student to have attended a study school if the student was
in a study school for at least one day from October 1 through May 1 of the school year. If a
significant proportion of students attended a study school for a very short period, including them
in the analysis might have biased our estimates down due to their limited exposure to New
Heights. We therefore estimated the impact of New Heights including only students who were in
a study school for at least 7 days, as well as by including only students who were in
a study
school for at least 30 days. As indicated by the sample size of 9 students and 46 students,
respectively, very few students in our primary sample were in a study school for such short
periods of time (Table C.4). The estimated impacts of New Heights based on these restricted
samples are similar to our primary impact estimates.
Although our primary analysis includes both the fall and spring semesters for the attendance
and semester graduation rate outcomes, it may be that schools more accurately track graduation
in
the spring semesters, when most graduations occur. We therefore conducted a sensitivity
analysis restricting the sample to only spring semesters, which, for simplicity, we conducted for
each outcome. The estimated impacts on unexcused absences per semester, days attended per
semester, and credits earned per year remain significant, while the impact on excused absences
per semester and the semester graduation rate are not significant (Table C.4).
Do'stlaringiz bilan baham: