Does Upward Bound Have an Effect on Student Educational Outcomes? A Reanalysis of the Horizons Randomized Controlled Trial Study

The stated goal of Upward Bound (UB) is to increase the rate at which traditionally disadvantaged students graduate from high school and enroll in and graduate from postsecondary educational institutions. Past experimental studies of Upward Bound have found it to be generally ineffective in improving these student educational outcomes. However, my review of the studies revealed apparent methodological and analysis problems. Analyses to date have not fully addressed these issues prior to estimation, and therefore led to the production of potentially inaccurate estimates. After addressing these problems, I found that UB improves the high school and postsecondary outcomes for low-income, first generation students who took part in the experiment. Specifically I found evidence that suggests UB increases high school graduation rates by 4.5 percentage points, post-secondary education enrollment rates by 2.9 percentage points and post-secondary completion rates by 4.6 percentage points for the intent to treat (ITT) population when using a covariate adjusted model with trimmed sampling weights to compare the treatment and control groups. In addition, I found some evidence of effect heterogeneity: students who might be typically declared ineligible for UB participation had post-secondary completion rates 8.4 percentage points higher than typically eligible students; the effects are positive but smaller for typically eligible students. In contrast, researchers of the UB experiment found no evidence of treatment effects on high school graduation or post-secondary enrollment, and did not explore effect heterogeneity according to eligibility. One possible reason for the differences in post-secondary results is that I do not have access to all the post-secondary data sources that were used in the Horizons study. There are two major implications of my findings. First, UB can be used to reduce high school dropout rates. Second, UB eligibility screening processes, such as those that were in place during the time of the Horizons study, should be amended to facilitate the participation of typically ineligible students. While the results from prior analyses are not robust to weighting methods, the covariate-adjusted models are more robust. Overall, there is compelling evidence that UB can narrow attainment gaps between students from low and high-income households.