![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
Abstract
Following up on a previous impact study of Early Colleges (EC) based on retrospective admission lotteries, this study assessed longer-term impacts on students’ postsecondary outcomes with 4 more years of data. The study found that students who won EC admission lotteries were significantly more likely to enroll in college, enroll in 2-year colleges, complete a college degree, complete associate’s degrees or certificates, and complete bachelor’s degrees within 6 years after expected high school graduation than control students. Moreover, it found that treatment students completed postsecondary degrees earlier and faster than control students. Consistent with EC’s focus on college exposure during high school, the EC impacts on college enrollment and the completion of associate’s degrees largely occurred within high school. The study also found that EC impacts did not vary significantly by students’ demographic characteristics; however, some impacts were significantly stronger for students with higher levels of prior achievement.
Disclosure Statement
The opinions expressed in this research are those of the authors and do not represent the views of the Institute or the U.S. Department of Education.
Notes
1 ECs employ a variety of approaches to cover students’ tuition costs, including tuition payments from the state or district or a college’s decision to waive tuition for EC students. Funding availability and local/state policy contexts, however, require that students in some states absorb some or all of this expense (Berger et al., Citation2009).
2 This requirement for instructor qualifications applied to all ECs included in the sample of this study.
3 Miller and Corritore (Citation2013) suggest that one potential explanation for the lack of effect on science pipeline progression is the difficulty that small ECs face in employing a sufficient number of highly qualified staff to teach multiple science courses, whereas a single highly qualified teacher may teach multiple math courses based on North Carolina certification policies.
4 To simplify the text, we refer to the impact of ECs in this article; however, EC impacts in this article refer to the effect of receiving the offer to enroll in an EC.
5 The study team examined lottery records to ensure that random assignment occurred as planned. Students who were admitted to ECs outside of the lottery process (e.g., siblings of students already admitted to ECs) were excluded from the study sample.
6 Procedures for imputing missing background data for this site are identical to the methods applied to participants with missing background data from other study sites. We provide a description of multiple imputation methods in the Analytic Approach section.
7 Results of these sensitivity analyses largely resemble the results presented in this article, with one exception noted in the later Results section. In addition, for outcome measures that are common between the original study and this follow-up study, results from the original study (where we had actual data for this site) and the follow-up study (where we had imputed data for this site) were generally similar.
8 This is a reasonable assumption given the NSC’s almost universal coverage of postsecondary enrollments in the nation. Even though there is a very slight chance that some students’ postsecondary records may be missing from the NSC database, there is no reason to expect the missingness to be related to treatment status or bias the impact estimates.
9 Some readers may have the concern that college enrollment by Year 4 might be overaligned with the treatment of this study. However, as we show in this article, many control students also had the opportunity to enroll in college by Year 4, thus college enrollment by Year 4 is not overaligned with the treatment according to Version 4.1 of the What Works Clearinghouse (Citation2020) group design standards. Further, while findings about the EC impacts on shorter-term outcomes may certainly need to be contextualized, our assessment of the EC impacts on longer-term outcomes helps overcome that concern.
10 The study team also collected student-level data about English learner (EL) status and Individualized Education Program (IEP) status before entering high school. Because only a small percentage of students in our sample were ELs (less than 1%) or had IEPs (7%), we did not include these two variables as covariates in the impact analyses.
11 Given the presence of noncompliance with treatment assignment (i.e., no-shows and crossovers), we supplemented the ITT analyses with complier average treatment effect (CATE) analyses to estimate the effects of actually attending an EC—as opposed to the effects of being offered admission to an EC through a lottery—for students who complied with their treatment assignment (i.e., compliers). Across the 23 lotteries (including sublotteries) included in this study, no-shows occurred in 18 lotteries, with an overall no-show rate of 20.9% among treatment students. Crossovers occurred in only three lotteries, with an overall crossover rate of 2.0% among control students. Technical details and results of the CATE analyses are available upon request.
12 For a given lottery with m sublotteries, SUBLOTmij was coded -1 for students in the omitted reference sublottery (i.e., if m = 1), 1 for students in sublottery m within the given lottery, and 0 for all other students. Given the effect coding, the treatment effect for such a lottery represents the equally weighted effect across the m sublotteries within the lottery. There is one set of sublottery indicators for each lottery with sublotteries in the level-1 equation, although only one set is shown for simplicity.
13 In two of the 23 lotteries, all EC students were minorities; therefore, these two lotteries were excluded from the analysis of differential impact by minority status. In addition, one lottery was excluded from the analysis of differential impact by low-income status because no students in this lottery were from low-income families.
14 Our findings for college enrollment “between Year 5 and Year 10” may overestimate the EC impact on “college enrollment after high school” because, while Year 5 is intended to represent the first year after expected high school graduation, students at three ECs offering 5-year programs may have still been enrolled in the ECs in Year 5. However, in two of these ECs (accounting for 11% of the study sample), the majority (78%) of the EC students actually graduated within 4 years. At only one EC site, the majority of treatment students graduated from high school within 5 years (60%) rather than within 4 years (19%). This site, however, accounted for only 6% of the overall study sample, thus the 5-year high school program attended by some EC students should not substantially affect the interpretation of findings for college enrollment between Year 5 and Year 10.
15 The EC impact on bachelor’s degree completion within 6 years after expected high school graduation was slightly smaller and marginally significant (p = 0.096) after removing one site where student background data were imputed for all students.
16 It is possible that the lack of significant differential impacts may be due to insufficient statistical power. In particular, analyses focusing on differential impacts by race/ethnicity were based on a substantially reduced sample size because two lotteries with over 600 students in total were excluded from these analyses given that 100% of the students in these lotteries were minority students.
17 Although one may assume that only selective, high-performing high schools would have the opportunity to initiate a lottery system due to oversubscription, this was not always the case. In fact, some of the ECs in this study implemented admission lotteries due to local or state education policies, or as part of citywide high school application processes.