Do Stricter Licensing Requirements Change Who Becomes a Teacher?
Does making it harder to earn a teaching license change who becomes a teacher? Larsen et al. (2026) studied 37 state licensing rules from 1991 to 2007 and tracked 26,280 new teachers, using the average SAT scores at each teacher's college as a measure of academic background. They find that tougher rules, mainly requirements to take college courses in math, English, science, and other subjects, screened out some candidates from the least selective colleges. Stricter licensing reduced the share of new middle and high school teachers from those colleges by about 2 percentage points, but it did not attract more teachers from top colleges. Certification exams made no measurable difference.
Why This Article Was Selected for The Policy Scientist
Occupational licensing now governs a large share of the American workforce, and teaching is among the largest and most consequential licensed professions. Each state decides who may teach in its public schools, and those rules shape the academic preparation of more than three million teachers who instruct nearly every child in the country. The question is timely: in response to persistent shortages, many states have recently relaxed or waived certification requirements, usually with little evidence about how such changes alter the composition of the teaching workforce. This study shows that the consequential variation in licensing during the 1990s and 2000s came from academic coursework requirements rather than the certification exams emphasized in earlier research. Lead author Bradley Larsen has published extensively on occupational licensing. The national data span 17 years and all states, which supports generalizability, though the period ends in 2007. The difference-in-differences and modern event-study designs provide credible causal identification. The article is forthcoming in American Economic Journal: Economic Policy, one of the leading field journals in economics.
Full Citation and Link to Article
Larsen, B. J., Ju, Z., Kapor, A., & Yu, C. (2026). The effects of occupational licensing stringency on the market for public school teachers. American Economic Journal: Economic Policy. Forthcoming.
Central Research Question
The study asks how the stringency of state teacher licensing requirements affects the academic backgrounds of people who enter public school teaching. In particular, it asks whether stricter requirements raise the lower end of the distribution by screening out candidates from less selective colleges, lower the upper end by discouraging candidates from highly selective colleges who have attractive options outside teaching, or both. It also asks which licensing requirements actually drove changes in stringency, whether effects differ in high-poverty and high-minority districts, and whether stricter licensing reduces the number of teachers.
Previous Literature
Concern about the academic preparation of teachers gained national attention with A Nation at Risk (Gardner, 1983), which argued that too many teachers came from the bottom of their graduating classes. Later research documented a negative relationship between academic ability and the likelihood of entering teaching (Hanushek & Pace, 1995; Hoxby & Leigh, 2004; Bacolod, 2007). How to respond was contested. Some researchers argued that stricter certification would keep less qualified candidates out of classrooms (Darling-Hammond et al., 2001), while others argued that certification imposes costs that deter capable candidates (Ballou & Podgursky, 1998).
The most closely related empirical work, Angrist and Guryan (2004, 2008), examined state requirements for basic skills and subject-matter exams and found no significant effect on the average college selectivity of teachers. A theoretical model by Larsen et al. (2020) showed that stricter licensing can simultaneously raise the lower tail and lower the upper tail of teacher academic ability, leaving the average unchanged. Related studies have found that high-stakes teacher evaluation raised the lower tail of new teachers' college selectivity (Kraft et al., 2020), that performance-based licensing tests reduced the number of teacher-preparation graduates (Chung & Zou, 2025), and that certification exams can have disparate effects on minority candidates (Deneault et al., 2026).
Data
The study assembles a new panel of 37 teacher licensing requirements in every state and Washington, D.C., from 1991 to 2007, drawn primarily from state certification manuals. The requirements include academic coursework, certification exams, student teaching, background checks, fees, citizenship rules, and specialized training. Each is coded as present or absent in each state and year.
Teacher data come from the restricted-use Schools and Staffing Survey, a nationally representative survey conducted in 1993, 1999, 2003, and 2007. The sample includes 26,280 teachers who began teaching between 1991 and 2007, linked to the licensing rules in place in their state when they started. Each teacher's academic background is measured by the average SAT score of entering freshmen at the teacher's undergraduate institution, standardized so that results are expressed in standard deviations. After restricting to state-year cells with at least six teachers, the analysis includes 857 state-year cells. The study adds state-level controls for school and student characteristics, teacher and non-teacher labor markets, unemployment, school finance reform, desegregation orders, collective bargaining, charter schools, and state politics.
Methods
The authors first use principal factor analysis to reduce the 37 requirements to a small number of underlying dimensions. The first factor explains far more variation than any other, with an eigenvalue nearly three times that of the second. It is driven almost entirely by academic coursework requirements outside education schools, with correlations above 0.9 for social science, natural science, English, math, and humanities. Certification exams correlate at 0.32 or less. The authors treat this first factor as their measure of licensing stringency. Several states changed it substantially: Kentucky and North Dakota became stricter, while Arkansas, Florida, and Georgia became less strict.
To estimate effects, the authors use two complementary designs. First, they classify states as moving from low to high stringency or from high to low and estimate event studies using several recent estimators that address staggered adoption and heterogeneous effects, alongside a conventional two-way fixed effects model. Second, they estimate difference-in-differences regressions with the continuous stringency index, state and year fixed effects, progressively larger sets of controls, and, in one specification, state-specific time trends. Outcomes include the mean, 10th percentile, and 90th percentile of college selectivity in each state and year. Standard errors are clustered by state.
The design rests on the assumption that, absent changes in licensing, states that changed their rules would have followed trends similar to those that did not. Some pre-treatment estimates in the event studies are statistically significant, although confidence intervals largely include zero. College selectivity is an imperfect proxy for individual academic ability and does not directly measure classroom effectiveness, a limitation the authors acknowledge.
Findings/Size Effects
The central finding is that stricter licensing raised the lower tail of teachers' college selectivity, particularly among secondary school teachers. In the preferred specification, a one-standard-deviation increase in stringency raised the 10th percentile of college selectivity for secondary teachers by 0.14 standard deviations, with estimates ranging from about 0.13 to 0.17 across specifications. The effect is statistically significant and is not driven by any single state. The effect on the mean was small, positive, and not statistically significant, while the effect on the 90th percentile was negative (-0.037 standard deviations) but also not significant. The gap between the 90th and 10th percentiles narrowed significantly.
The authors translate these estimates into more intuitive terms. The increase in the 10th percentile corresponds to a 2.1 percentage point decrease in the share of secondary teachers who graduated from very low-ranking colleges. The decline at the top corresponds to a 0.45 percentage point decrease in the share from very high-ranking colleges, though this estimate is not significant. The event-study estimates point in the same direction: moving from low to high stringency raised the 10th percentile for secondary teachers by roughly 0.14 to 0.19 standard deviations. For elementary teachers, the study finds no robust effects.
The effects come from coursework requirements. Replicating Angrist and Guryan, the authors find that certification exams affected neither the average nor either tail of the distribution. The second and third factors, which capture specialized training and exams, also show no effect.
The study finds no evidence that stricter licensing harmed high-poverty districts. If anything, the effect on average college selectivity was more positive there. In high-minority districts, however, stricter licensing narrowed the distribution more for secondary teachers: the 10th percentile rose by an additional 0.15 standard deviations and the 90th percentile fell by an additional 0.18 standard deviations relative to low-minority districts. There is suggestive evidence that stricter licensing reduces teacher supply. Moving from high to low stringency increased the number of teachers by about 0.04 to 0.05 log points, roughly 4 to 5 percent, though estimates using the continuous index lose significance once controls are added.
Conclusion
The study shows that teacher licensing changes in the 1990s and 2000s operated mainly through academic coursework requirements rather than certification exams, and that stricter requirements raised the lower tail of new secondary teachers' college selectivity without a significant change in the average. These results suggest that licensing can screen out some candidates from the least selective colleges, consistent with its stated purpose, while offering little evidence that it attracts stronger candidates overall. The narrowing of the distribution in high-minority districts, including a decline at the top, warrants attention from policymakers considering stricter standards.
Several limitations qualify these conclusions. College selectivity is a proxy for academic background rather than a direct measure of teaching effectiveness, and the authors' exploratory analysis of student test scores is inconclusive because of data limitations. The data end in 2007, before the more recent wave of licensing changes prompted by teacher shortages. Future research should link licensing changes to student achievement and long-run outcomes, and should examine whether recent decisions to relax certification requirements have shifted the composition of the teaching workforce in the opposite direction.
Comments