Selective High Schools, Major Intentions, and Academic Achievement

• In Event: Promising Scholarship in Education: Dissertation Fellows and Their Research

Abstract

An enduring topic for debate in the US education landscape is how best to serve the nation’s high achieving students. A sizable proportion of these students enroll in exam or magnet schools. Despite the abundance of resources channeled into these institutions, research that evaluates their causal effects on student achievement is still limited. What evidence exists in the US context finds that elite schools have null or only small effects in promoting student achievement and postsecondary outcomes.

This paper investigates the effectiveness of selective high schools on student achievement and postsecondary intentions. A prevailing challenge to identification is the non-random selection of students into selective schools based on individual preferences and school attributes that are unobserved by the econometrician. Recent literature circumvents this issue by using a regression discontinuity (RD) approach or comparing students who were chosen via a lottery process among over-subscribed schools. I elect the former method. Using new application data from a selective public high school in North Carolina that comprises continuous admissions scores, I construct credible treatment and control groups on either side of the cutoff. The regression discontinuity framework enables identification of the effect of attending a selective high school for students on the margin of treatment.

A particular strength of this dataset derives from legislative rules governing the admissions procedure. The institution must equitably serve all thirteen congressional districts in North Carolina. This implies that students applying from across the state are evaluated against peers from the same residential congressional district. The within-district nature of the application process generates substantial student background variation. The resulting range of applicant demographics and baseline abilities affords an opportunity to examine how outcomes differ by the extent to which the new peer environment departs from that of the sending school.

Another contribution of the present paper lies in its focus on science- and math-related outcomes. The selective institution is STEM- (science, technology, engineering, and math) oriented, enabling me to directly test for the effect of attending a science- and math-intensive school on students’ predilections for these fields, using major intentions and (eventually) actual major choice. The elite school literature explores dimensions of academic performance and postsecondary success, but does not examine major or occupational choice outcomes. The only study focused solely on the causal impact of STEM institutions finds insignificant effects under the assumption of selection on observables. This study then adds to this nascent literature by estimating the causal impact of STEM high schools on student achievement and major orientation. These findings are well suited to comment on one of the primary goals of STEM institutions.

To carry out these analyses, I create a novel dataset by merging admissions records with administrative files from the K-12 public education system in North Carolina. The resulting longitudinal dataset includes demographic, academic performance, course choice, and family background variables. I initially examine outcomes in standardized test scores, major intentions in the second half of high school, and college application behavior. A lingering question is whether students who express an interest in STEM fields persist in these majors until the end of their collegiate careers. I am in the process of gathering this set of longer-term postsecondary outcomes from the University of North Carolina system microdata and National Student Clearinghouse files.

Findings show that students above the admissions cutoff perform better in both SAT math and critical reading relative to those predicted to have not enrolled. Estimates document an increase of 20 points, the equivalent of approximately 0.2 standard deviations, for both subjects. Meanwhile, I find limited evidence that treated students are more likely to indicate an interest in STEM majors.

Next I examine the effect of attending a selective high school on college application behavior. While treated students do not appear to change the number of schools or the share of very selective institutions applied to, their college portfolios now comprise institutions with a greater STEM focus, as measured by the share of undergraduate degrees awarded in STEM.

I rely on substantial variation in student and sending school attributes to evaluate heterogeneous treatment effects. Gains in SAT math scores are most prominent among students with lower baseline performance, namely female students, students from families with less parental education, or students from lower-performing or rural schools. SAT verbal gains are more broad-based, although large gains do accrue to disadvantaged groups such as those coming from lower performing schools.

Increases in the STEM-intensity of college applications, on the other hand, are most pronounced among groups with higher baseline levels of STEM orientation. Male students and those coming from non-rural schools, for instance, record larger gains in magnitude.

Author

• Ying Shi, Duke University