Do University Budget Incentives Lead to More Master's Degree Programs?
When public universities let colleges and departments keep more of the tuition they bring in, do those units start more master's degree programs? Kelchen (2026) examines this using federal data from 2004–05 to 2023–24 on 154 public research universities, 58 of which adopted responsibility center management (RCM) budget models. The answer appears to be no. After adopting RCM, a university added about one-fifth of a program on average, on a base of roughly 63, and its master's graduates rose about 1.9 percent, changes too small to rule out chance. Universities did not close more programs either, and the mix of fields did not shift.
Why This Article Was Selected for The Policy Scientist
Public universities face sustained financial pressure. Tuition now funds a far larger share of instruction than it did a generation ago, undergraduate enrollment has softened, and federal limits on graduate lending, research funding, and international enrollment add new uncertainty. In response, many campuses have adopted budget models that reward academic units for generating revenue, on the theory that deans and departments will act more entrepreneurially. Master's programs, which are self-contained and have grown rapidly, are the clearest test of that theory, and the answer matters for students who carry the debt and for taxpayers who subsidize these institutions. Robert Kelchen has published extensively on higher education finance, and this study contributes the first national, hand-collected dataset on RCM adoption, extending earlier case-based work such as Jaquette et al. (2018). The heterogeneity-robust event-study design with inverse probability weighting is a credible causal approach, though adoption was not random and the results apply to public research universities. The article appears in Education Finance and Policy, a leading peer-reviewed journal in the field.
Full Citation and Link to Article
Kelchen, R. (2026). Do responsibility center management budget models encourage entrepreneurship? Evidence from master's programs at public universities. Education Finance and Policy. Advance online publication.
Central Research Question
More than 50 public research universities have replaced incremental budgets, in which units receive roughly what they received the year before, with responsibility center management (RCM) models. Under RCM, colleges and departments receive revenue tied largely to the tuition their students generate and pay a share of central administrative costs. The study asks whether this shift changes behavior where the incentive should be strongest: master's programs, whose curricula usually sit within one unit, so more of the revenue stays there. It asks whether adopters change the number of master's programs or graduates at different rates than non-adopters, and whether the types of programs and their return on investment for students change. Because RCM also lets units keep cost savings, closures are examined alongside launches.
Previous Literature
RCM draws on new public management, which urged agencies to behave more like businesses (Hood, 1991), and fits the theory of academic capitalism, in which universities pursue revenue and status through entrepreneurial activity (Slaughter & Leslie, 1997). Practitioner guides describe its intended benefits (Curry et al., 2013), but rigorous evidence is thin. Jaquette et al. (2018), using synthetic control methods, found that three of four early public adopters saw higher tuition revenue within five years. Rutherford and Rabovsky (2018), studying 32 research universities, linked RCM to higher bachelor's graduation rates for White students and more STEM degrees. Choi (2026) found that RCM increased master's degree production while reducing doctoral production, but did not examine whether universities created new programs or simply enrolled more students in existing ones.
Data
The sample begins with the 160 public universities classified as research or doctoral institutions in the 2000 Carnegie classification, minus six that pursued RCM but never implemented it, leaving 154: 58 adopters and 96 non-adopters. Because no national dataset existed, the author and a research assistant built one from budget websites, board minutes, financial documents, and archived pages on the Wayback Machine, recording both the announcement date and the first year RCM actually allocated funds. Adopters rose from 13 in 2004–05 to 23 in 2013–14 and 58 by 2023–24, more than one-third of public research universities.
Program counts come from federal IPEDS completions data grouped at the four-digit Classification of Instructional Programs level, with launch dates proxied by the year before a program's first graduate. About one-third of apparent closures in the raw data were reporting gaps and were recoded as continuing programs. Return on investment uses College Scorecard data for cohorts graduating between 2014–15 and 2019–20: whether median debt was below 50 percent of first-year earnings, and whether median earnings reached $66,600, the median for full-time bachelor's degree holders ages 25–34. Roughly one-third of programs met each standard. Controls cover enrollment, student composition, finances, state funding policies, state economic conditions, and partisan control.
Methods
Because universities adopted RCM in different years, conventional two-way fixed-effects regressions can be biased when effects vary across institutions or over time. The study instead uses the heterogeneity-robust difference-in-differences event-study estimator of de Chaisemartin and D'Haultfoeuille, comparing adopters with universities not yet or never treated for up to ten years after adoption, enough time for program approval and recruitment. Observations are weighted by the inverse probability of adopting RCM, estimated from 2004–05 characteristics, and models include institution and year fixed effects with state-clustered standard errors.
The key assumption is parallel trends. Event studies using institutional characteristics as outcomes show no consistent pre-adoption trends, which supports but does not prove it. Weighting narrowed but did not eliminate group differences: adopters still offered more programs (63.2 versus 56.3) and received less state funding per student ($10,503 versus $11,754). Incentive strength also varies across universities in mostly unobservable ways, so the estimate is an average over very different treatments, and some comparison universities use their own revenue incentives, which may narrow the contrast.
Findings/Size Effects
The central finding is a null result. Averaged across post-adoption years, RCM increased the number of master's programs by 0.19, about 0.3 percent of the 63.2 programs at a typical adopter, and master's graduates by 1.9 percent; neither is statistically significant. The design could detect effects of roughly two programs, or five percentage points in graduates, in most years. Using the announcement date instead yields a marginally significant 3.9 percent increase in graduates but still no change in programs.
Launches and closures show the same pattern. Adopters added about 1.4 programs and dropped about 0.6 per year, versus 1.2 and 0.5 at non-adopters, but event-study estimates show no consistent change after adoption, despite power to detect about one program per year. Results by field are flat: total effects range from −0.33 programs in STEM to 0.40 in "other" fields, with business at 0.05; none is significant.
The return-on-investment results raise a tentative concern. RCM had no detectable effect on programs with debt below 50 percent of earnings (0.02) or earnings above $66,600 (−0.13). Estimates for lower-value programs were larger, 0.38 and 0.43 additional programs, but not statistically significant, and are best read as patterns to monitor rather than findings.
Conclusion
The study finds no evidence that RCM adoption changed the number, mix, or student returns of master's programs at public research universities, or the number of graduates, in contrast with smaller studies that found revenue gains at some early adopters. Two explanations are offered: new programs require upfront investment in faculty, recruitment, and approvals that units may be unable to finance, and the faculty who design curricula may see little direct benefit from a budget formula, while deans may doubt promised rewards will arrive.
These results do not establish that RCM has no effects; it may still shape enrollment in existing programs, costs, or specializations finer than the program codes capture, and strong versions of RCM could matter even if the average effect is small. The findings may not extend to private universities, whose budget details are rarely public. Future research should examine tuition-sharing rates and central taxes, online program management firms, faculty and administrator views, and whether universities that abandon RCM, as Temple University announced it would in 2025, achieve their aims.
Comments