How Does Faculty Unionization Affect Wages in Higher Education?

Baker, Halberstam, Kroft, Mas, and Messacar (2025) ask whether faculty unionization changes the distribution of wages in Canadian higher education. They analyze longitudinal administrative salary data from 1970–2022 and link it to the staggered rollout of faculty unions. They find that unionization raises wages at the bottom of the distribution by roughly 10 percent while leaving top salaries unchanged, compressing inequality. On average, salaries rise about 2 percent in the first year and about 6 percent after six years. They also show that these gains are largely financed through higher student enrollment rather than cuts to employment or increases in tuition.

Why This Article Was Selected for The Policy Scientist

This article examines the wage effects of faculty unions, a topic with broad relevance given ongoing debates about wage inequality, public-sector bargaining, and higher-education finance. The policy importance is substantial because universities are major public employers, and unionization trends intersect with wider concerns about labor markets and inequality. The topic is timely as union activity has resurfaced across multiple sectors. The authors have written extensively in this domain, and this piece adds to the literature by pairing unusually rich administrative data with a credible causal difference-in-differences design. The dataset is high quality, and the findings may generalize to similar public university systems outside Canada.

Full Citation and Link to Article

Baker, M., Halberstam, Y., Kroft, K., Mas, A., & Messacar, D. (2025). The impact of unions on the wage distribution: Evidence from higher education. American Economic Review: Insights. https://www.aeaweb.org/articles?id=10.1257/aeri.20240722

Central Research Question

The authors ask whether and to what extent faculty unionization alters the distribution of wages within Canadian universities, with particular attention to differential effects across the salary distribution rather than average wage effects alone. This focus extends beyond the traditional inquiry into mean wage premiums by examining whether unions compress intra-institutional wage inequality, shift salary floors, and alter employment margins. The research also investigates ancillary channels that might finance potential wage changes, such as student enrollment, tuition levels, and provincial transfers. In short, the central question is whether faculty unions reshape the wage structure within public higher education, by how much, through what mechanisms, and with what budgetary tradeoffs. 

Previous Literature

Research on unions has long documented average wage premiums and, in some contexts, reduced wage dispersion. Classic work by Freeman and colleagues explored how collective bargaining affects wage structures and inequality, often concluding that unions raise wages at the bottom and compress the distribution. Much of that research, however, rested on cross-sectional data, selection corrections, or job-mover designs that introduce concerns about sorting and unobserved heterogeneity. More recent quasi-experimental studies have tested whether narrow election victories produce causal wage changes but have generally found muted or null effects on average wages. Those studies, however, typically focus on private-sector workplaces in the United States and seldom examine unconditional wage distributions.

In the public sector, the evidence base is smaller despite high union density. Existing work on teachers’ unions provides mixed estimates of wage effects, ranging from modest positive impacts to zero estimates depending on specification and aggregation levels. In higher education specifically, earlier Canadian studies using institutional aggregates found little to no salary effects, but those analyses lacked linked microdata, did not observe first-contract provisions, and could not control for rich worker-level heterogeneity. The article at hand contributes by combining quasi-experimental identification with population administrative microdata, allowing a cleaner estimate of within-institution distributional effects. Its contribution is thus both empirical and methodological, filling a gap between classic inequality studies and modern causal research designs.

Data

The authors assemble a linked administrative dataset that covers the universe of full-time faculty at Canadian universities from 1970 to 2022, drawing from Statistics Canada’s University and College Academic Staff System (UCASS). The dataset includes anonymized individual identifiers that permit within-institution longitudinal tracking and incorporates rank, degree, salary, and limited demographic information. Crucially, salaries are measured as “base pay,” excluding stipends, grant-funded compensation, and leave adjustments, although the authors later verify robustness with alternative measures of “actual salary” available beginning in 1985.

The salary data are merged with newly collected records on union certifications and first contracts acquired through direct contact with faculty bargaining units, union websites, and university newspapers. These records contain certification dates and, when available, details on contractual salary floors by rank or experience. In addition, institution-level data on enrollment, tuition, and provincial transfers provide a basis for analyzing financing channels. Together, these sources produce a panel featuring staggered treatment timing, precise treatment dates, and unusually high measurement fidelity. The dataset’s scope and level of detail allow analysis of both unconditional wage distributions and ancillary employment margins such as hiring, separations, and promotions. Its quality is high given population coverage, objective compensation records, and documented institutional bargaining events. 

Methods

The authors use a causal difference-in-differences (DiD) framework tailored for staggered adoption treatments, relying specifically on Callaway and Sant’Anna’s (2021) estimators that avoid the negative weighting and contamination issues associated with two-way fixed effects regressions in heterogeneous treatment environments. Treated units are faculty working at universities during years when certification occurs; comparison units are “never-treated” institutions, although later robustness checks add “not-yet-treated” controls. The design relies on within-worker changes following exogenous changes in workplace union status, effectively controlling for time-invariant worker and institutional heterogeneity.

To estimate distributional impacts, the authors combine re-centered influence function (RIF) regressions for unconditional quantiles with DiD estimators, following Firpo, Fortin, and Lemieux (2009). To test mechanisms involving salary floors, they apply a bunching-style estimator that bins salaries relative to contractual floors and tests for excess mass above those thresholds after certification, a strategy similar to minimum wage bunching studies (Autor, Manning, and Smith 2016; Cengiz et al. 2019). They also estimate DiD models for employment stocks, hiring flows, promotion flows, and departures. Finally, they estimate institution-level DiD models for enrollment, tuition, and provincial transfers.

The empirical strategy is causally credible for at least three reasons: first, the timing of union certifications is plausibly exogenous to individual wage trajectories; second, pre-treatment trends are flat; and third, staggered timing with doubly-robust DiD reduces reliance on functional form assumptions. While not a randomized controlled trial, the use of modern causal inference tools places the analysis at the stronger end of observational designs. The authors also subject their findings to extensive robustness checks, including alternative estimators, alternative control groups, extended windows, and sensitivity tests for parallel trends deviations.

Findings/Size Effects

The authors uncover four major empirical results. First, unionization compresses faculty wage distributions. Quantile DiD estimates show increases of roughly 10–12 percent at the 10th percentile of the unconditional salary distribution six years post-certification, tapering to near zero at the 75th and 90th percentiles. This confirms that wage gains are concentrated at the bottom rather than evenly distributed. The time pattern suggests these changes accumulate gradually after certification rather than appearing instantaneously. 

Second, mean earnings rise following unionization, but the magnitude varies by period. Average base salaries increase approximately 2 percent in the first year and approximately 6 percent by year six, but these gains are concentrated in certifications occurring between 1970 and 1995. Certifications after 1995 exhibit near-zero average impacts. The authors interpret this temporal heterogeneity as consistent with declining union bargaining power over time or threat effects that raise salaries in non-union universities.

Third, the distributional compression appears to operate through salary floors. Approximately 85 percent of first contracts contain floors that apply across rank or experience groups. Bunching estimates reveal negative excess mass below floors and positive mass just above, especially within $6,000 of the floor. This pattern strongly supports the mechanism that formal wage minima drive compression at the lower tail. 

Fourth, employment responses are minimal. The DiD estimates show no meaningful changes in total faculty employment, hiring, separations, or promotions. This finding is informative because monopsony or downward-sloping labor demand models would predict reduced employment in response to wage increases. Instead, the authors find no evidence of reductions even in the period where wages increased.

Finally, the financing mechanism appears to be increased enrollment rather than higher tuition or increased government transfers. Institution-level DiD estimates indicate approximately 13–14 percent increases in enrollment within six years post-certification during the pre-1995 period. Tuition and transfers remain flat, consistent with constraints on tuition-setting and provincial budgets. The evidence suggests that universities scale instructional loads rather than reduce staff or raise prices to students.

Conclusion

This study provides compelling evidence that faculty unions reshape wage distributions by raising salaries at the lower tail, thereby compressing intra-institutional inequality. Importantly, the analysis shows that these gains are not offset by employment reductions, tuition increases, or supplemental provincial funding. Instead, universities appear to absorb the costs through higher student enrollment, implying that unions indirectly increase teaching loads or class sizes. The findings also show that mean wage effects depend on historical context: bargaining strength and external labor market conditions matter for whether average premiums materialize.

From a scientific perspective, the article contributes three notable advances: first, it brings credible causal inference tools to a public-sector setting where RCTs are infeasible; second, it integrates distributional analysis with institutional records of contractual provisions; and third, it highlights financing channels, an underdeveloped dimension in union research. The data are comprehensive, longitudinal, and administrative, which enhances measurement accuracy and mitigates concerns about reporting bias. The design is well suited for causal inference, though continued research could explore additional mediating channels such as hiring composition, research productivity, or non-wage benefits. Because the dataset covers the universe of Canadian universities, the findings are likely generalizable to other public university systems with similar institutional funding models and collective bargaining laws.

Overall, the article represents a significant contribution to empirical labor research, public-sector union studies, and the economics of higher education. Its combination of rigorous causal methods, rich administrative microdata, and attention to distributional mechanisms distinguishes it from prior work and positions it as a foundational study in understanding how unions affect wage structures rather than merely average wage levels in large public-sector institutions.