Do More Selective Colleges Actually Deliver Higher Earnings?

Bloem, Hu, and Hurwitz (2026) ask how post-college earnings vary across colleges when using the full earnings distribution and how college selectivity relates to those outcomes. They analyze U.S. Department of Education College Scorecard data covering nearly all four-year colleges, including earnings percentiles 6, 8, and 10 years after entry. They find that rankings change substantially across percentiles, with about 40% of colleges shifting more than 150 places. Earnings overlap widely across institutions, and selectivity only strongly predicts earnings among highly selective colleges, where top-end earnings can be nearly twice as high and gender gaps exceed $16,000.

Why This Article Was Selected for The Policy Scientist

Bloem, Hu, and Hurwitz (2026) ask how post-college earnings vary across colleges when using the full earnings distribution and how college selectivity relates to those outcomes. They analyze U.S. Department of Education College Scorecard data covering nearly all four-year colleges, including earnings percentiles 6, 8, and 10 years after entry. They find that rankings change substantially across percentiles, with about 40% of colleges shifting more than 150 places. Earnings overlap widely across institutions, and selectivity only strongly predicts earnings among highly selective colleges, where top-end earnings can be nearly twice as high and gender gaps exceed $16,000.

Full Citation and Link to Article

Bloem, M. D., Hu, X., & Hurwitz, M. (2026). Understanding variation in post-college earnings: Evidence from the U.S. Department of Education’s College Scorecard. Education Finance and Policy. https://doi.org/10.1162/EDFP.a.445

Central Research QuestionThis study investigates two closely related questions: first, what additional insights emerge when post-college earnings are evaluated using the full distribution of outcomes rather than relying solely on median earnings; and second, how college selectivity is associated with earnings outcomes across that distribution, including variation by gender and field of study. The authors are motivated by the growing prominence of earnings metrics in college choice and rankings, particularly through the U.S. Department of Education’s College Scorecard. They seek to determine whether commonly used summary statistics—especially median earnings—misrepresent institutional differences and whether the relationship between institutional selectivity and earnings is uniform or heterogeneous across contexts.

Previous LiteratureThe paper builds on several strands of research. First, it contributes to the literature examining the College Scorecard as a policy tool. Prior work has shown that the introduction of earnings data had limited effects on enrollment behavior and that simplified presentation of these data can lead to misleading comparisons. The authors extend this work by leveraging more recent and more detailed Scorecard releases, particularly those that include multiple points in the earnings distribution and longitudinal cohort data.

Second, the study engages with a large literature on the returns to college selectivity. Earlier causal studies, using quasi-experimental designs such as regression discontinuity or matching frameworks, generally find positive earnings returns to attending more selective institutions, though with important heterogeneity. In contrast, descriptive studies often reveal more ambiguous relationships once selection bias is considered. The present study contributes to the descriptive side of this literature by examining a near-universe of U.S. four-year colleges and by focusing on distributional variation rather than mean or median effects.

Third, the authors connect to research on heterogeneity in returns by gender and field of study. Existing evidence shows persistent gender earnings gaps and substantial variation in returns across majors, with high-paying fields such as economics and engineering exhibiting different dynamics than regulated or credential-based professions like nursing or teaching. The paper extends this literature by examining how these patterns interact with institutional selectivity.

DataThe analysis relies on publicly available data from the U.S. Department of Education’s College Scorecard, which links administrative records on college attendance with earnings data derived from tax records. The dataset covers approximately 1,500 public and private non-profit four-year institutions, representing a near-universe of such colleges in the United States.

The key outcome variables are earnings percentiles—typically the 10th, 25th, 50th (median), 75th, and 90th percentiles—measured at multiple time horizons (6, 8, and 10 years after college entry). These data allow for a detailed examination of the full earnings distribution within each institution. Additional variables include institutional characteristics such as admissions rates (used as a proxy for selectivity), as well as disaggregated earnings by gender and by field of study.

A notable limitation of the data is that earnings are calculated only for students who received federal financial aid (Title IV recipients), which may underrepresent higher-income students at elite institutions. However, prior validation studies suggest that these earnings measures are highly predictive of overall institutional earnings patterns, mitigating concerns about external validity.

MethodsThe authors employ a descriptive empirical strategy, focusing on systematic comparisons across institutions and within distributions rather than causal identification. The analysis proceeds in several stages.

First, they examine the sensitivity of college rankings to different points in the earnings distribution. By re-ranking institutions using various percentiles, they assess how reliance on median earnings may distort comparative evaluations.

Second, they analyze the degree of overlap in earnings distributions across institutions by plotting percentile ranges for colleges grouped by median earnings. This approach highlights whether institutions with different central tendencies nonetheless produce similar outcomes for subsets of students.

Third, they construct a panel dataset combining multiple Scorecard releases to examine variation in earnings across cohorts and over time. This allows them to assess volatility and the influence of macroeconomic conditions on observed earnings outcomes.

Fourth, they estimate descriptive relationships between admissions selectivity and earnings outcomes, including separate analyses for different percentiles, genders, and fields of study. These relationships are visualized using scatterplots and stratified comparisons rather than regression-based causal inference.

Finally, they decompose institutional earnings differences by constructing “expected” earnings based on field-of-study composition, enabling them to assess how much of the selectivity–earnings relationship is driven by differences in academic program mix.

Findings/Size EffectsThe results demonstrate that median earnings are an incomplete and potentially misleading summary of institutional performance. Rankings based on alternative percentiles differ dramatically: approximately 40 percent of colleges shift by more than 150 positions when ranked by 90th percentile earnings instead of median earnings, and more than 200 institutions shift by over 300 places. These discrepancies indicate substantial heterogeneity within institutions that is obscured by central tendency measures.

The authors also find extensive overlap in earnings distributions across colleges. Students at the 75th percentile of lower-ranked institutions often earn more than those at the 25th percentile of higher-ranked institutions, suggesting that institutional averages conceal meaningful within-college variation.

Temporal analyses reveal significant volatility in earnings metrics across cohorts, particularly for smaller institutions. For example, around 16 percent of colleges exhibit standard deviations in median earnings exceeding $4,000 across cohorts, with even greater variability at higher percentiles. Earnings trajectories are also sensitive to macroeconomic conditions, with cohorts entering the labor market during recessions experiencing persistently lower earnings.

The relationship between selectivity and earnings is highly nonlinear. Among institutions with acceptance rates above roughly 40 percent, there is little to no association between selectivity and earnings outcomes. In contrast, among more selective institutions, the relationship becomes steep. Median earnings at colleges admitting fewer than 10 percent of applicants are nearly double those at institutions admitting around 40 percent. At the upper tail of the distribution, effects are even larger, with 90th percentile earnings frequently exceeding $200,000 at highly selective colleges.

Gender disparities are substantial and increase with selectivity. Median earnings gaps between men and women are approximately $5,000 six years after entry and $9,000 after ten years, but exceed $16,000 at the most selective institutions.

Field-of-study analyses reveal pronounced heterogeneity. In fields such as nursing, social work, and teacher education, there is little relationship between selectivity and earnings, consistent with regulated labor markets. In contrast, fields like economics, finance, and computer science exhibit strong positive relationships, particularly at highly selective institutions. Differences in program composition explain part—but not all—of the selectivity premium.

ConclusionThe study demonstrates that reliance on median earnings as a summary measure of college outcomes is analytically insufficient and potentially misleading for both policymakers and prospective students. A more complete understanding of institutional performance requires attention to the full distribution of outcomes, as well as to temporal variation and contextual factors such as field of study.

The findings also complicate conventional narratives about the value of college selectivity. While highly selective institutions are associated with substantially higher earnings at the upper tail, this relationship is not uniform across the distribution or across institutional tiers. For the majority of colleges, selectivity appears to have little association with earnings outcomes.

From a policy perspective, the results suggest that the College Scorecard’s consumer-facing interface could be improved by incorporating distributional metrics, longitudinal data, and disaggregated outcomes. Such enhancements would provide a more accurate and nuanced basis for decision-making. More broadly, the study underscores the importance of moving beyond simplistic metrics toward richer, multidimensional evaluations of higher education outcomes.