Occupational Licensing Rules Need Evidence on Quality, Access, and Costs

Kyle Rozema framed occupational licensing as a law-and-economics problem in which the central question is rarely whether regulation is good or bad in the abstract. The problem is the trade-off: protecting the public on one side, and raising costs or reducing access to services on the other.

He separated public protection into two possible channels. Licensing can work through selection, by excluding people who pose higher risk from entering an occupation. It can also work through training, if education or examination requirements actually make practitioners better. Those two rationales matter because different licensing rules lean on different mechanisms. A background check is mostly selection. A degree requirement may be training, screening, or both. A recurring exam may update knowledge, deter low-quality practice, or simply push older workers out.

Rozema used that premise to force a kind of judgment economists often avoid making in public. He said lawyers often make policy recommendations while acknowledging limited evidence, while economists often have evidence but avoid giving policymakers a direct answer. His questions therefore took the form policymakers commonly use: should a state do this or not?

The useful pattern in the answers was not a general pro- or anti-licensing position. The panelists tended to defend some licensing functions while attacking particular pathways, bottlenecks, opaque screens, and disciplinary practices. Their common test was whether a rule plausibly protects the public through the mechanism offered to justify it, whether reform is aimed at the right margin, and whether there is enough evidence about quality or access to justify the burden.

None of the three panelists endorsed full abolition of the requirement that lawyers be licensed to practice law. The defense was narrower than that result makes it sound: they supported some entry standard for lawyers, while leaving room to revise portability rules, support-role licensing, training routes, and the structure of legal education.

Alicia Plemmons said she would focus on changing the license rather than eliminating it. Her first target was interstate portability: if someone has already satisfied bar requirements elsewhere, she suggested, a state might require only a state-specific exam covering local legal rules rather than make the lawyer repeat the full process. She also distinguished lawyers from support roles. Some states license paralegals, but most do not; Plemmons saw paralegal work as better suited to learning under someone else on the job, and therefore less obviously justified as a licensed occupation.

Her position was not anti-licensing. She said she “love[s] licensing for lawyers” but wants an easier way to move lawyers where they are needed, while respecting state-specific legal concerns. For support services, by contrast, she thought licensing may no longer serve safety and may instead sacrifice access.

Jason Hicks agreed with the bottom line and emphasized political economy. Repealing law licenses outright may be an intellectually available option, he said, but it is not a politically realistic one. A legislator has little to gain and a lot to lose from supporting such a repeal. Hicks therefore urged attention to margins that facilitate movement or reduce unnecessary burdens. He mentioned British Columbia’s move, as he described it, to eliminate a requirement under which new lawyers had to work for low wages for roughly a year and a half, treating that as the kind of barrier worth scrutinizing even if licensing itself remains.

Yun Oh also leaned no, speaking from the perspective of older workers and career changers. Longer training duration, he said, is often a larger barrier than fees. Lower barriers could help older workers start new careers, but he did not think eliminating standards was acceptable in law. Even during training or early practice, some minimum standard is needed to avoid harming consumers.

The result was a defense of lawyer licensing, but not a defense of every component of the existing legal licensing structure. The panel defended some form of entry standard for lawyers while opening room for portability, narrower licensing of support roles, and alternative ways of satisfying requirements.

When the occupation changed from lawyers to economists, the answer changed. The panelists rejected licensing economists, and their reasons treated economics as an example of substitute forms of quality control.

Jason Hicks invoked Morris Kleiner’s work on public-interest justifications for occupational regulation. As Hicks described the framework, a policymaker should weigh the consequences of harm from the occupation’s tasks, the degree of direct interaction with the public, and task complexity. Economists score high on complexity, he said, but on aggregate fall lower on the public-interest rationale for licensing. The marginal value of regulating them did not seem sufficient.

Alicia Plemmons made the same point through an analogy to psychic licensing, which she said exists in some municipalities. Her point was that economists, like psychics in the licensing examples she has been studying, make predictions. If economists were to be licensed, she suggested, the analogy would be to a local prediction-regulation regime. But she said the California Supreme Court had treated psychic licensing as a First Amendment issue because psychics were expressing predictions, opinions, and inferences. Plemmons said she would defer to that logic and reject licensing economists.

Yun Oh initially said that for himself he might say yes, but for the general public he would say no. His reason was that economics has developed standards without licensing. Economists have argued with one another for roughly a century, competing to show which theories are right or wrong and generating new theories in the process. A license would be intended to ensure a minimum quality level, but he was not sure it would improve quality in a field that already has strong internal competition.

Hicks added that academic economics already has something like a de facto education requirement: to be a professor at a university, one typically needs a PhD. That requirement is not a state license, but it operates as a standard in the relevant labor market.

Plemmons used that point to move from occupational licensing to accreditation. At her business school, she said, AACSB accreditation imposes standards about how much teaching must be produced at a scholastically academic level. Faculty may need a PhD or significant contribution to literature and practice. That raised a broader alternative: instead of licensing individual workers, institutions or businesses might choose whether to be accredited under particular standards.

Oh added that economists also perform continuing education without a legal mandate. In many licensed occupations, continuing education requires ten to fifteen hours of coursework without necessarily requiring an exam. Economists, he said, study, produce new research, and update themselves without formal licensing or continuing education rules.

The point was not that quality control is unnecessary. It was that licensing is only one form of quality control, and often not the right one. In economics, credentialing, accreditation, professional competition, publication, and self-updating were treated as substitutes for state licensure.

Competence can be tested once at entry, repeatedly throughout a career, or in response to evidence of a problem. The panel’s resistance to a recurring bar exam was not a rejection of competence testing. It was a concern that periodic exams can become a costly, weakly targeted barrier unless they measure the right thing and solve an identified problem.

Rozema’s law hypothetical assumed a shift from a one-time bar exam after graduation to a boards-style system: lawyers would take an exam every five years, with a 95% pass rate. He contrasted that with a given bar exam pass rate of roughly 70%, while later clarifying that eventual bar passage within two years is closer to 95%. On his framing, the existing system may delay entry more than permanently exclude law graduates.

Jason Hicks resisted focusing on the pass rate alone. The key question, he said, is what is being tested. If 95% of lawyers pass because the exam measures competencies directly connected to successful practice and almost all practitioners meet them, then a high pass rate is not a problem. If the exam is not tied to practice quality, the number itself tells policymakers little. Hicks also said that before adding later exams, he would need to know what problem the current system fails to solve.

Alicia Plemmons proposed a more targeted model: retesting triggered by observed quality problems. She drew on an Iowa bill she had heard about that would create a board to license autonomous medical-practice systems, effectively healthcare AI. As she described it, the system would not be retested every year simply because time had passed. It would be retested if safety ratings fell too low, too many mistakes occurred, or monitoring showed quality concerns. Plemmons suggested a similar logic for lawyers: if 95% will pass, administering an expensive exam to everyone may waste resources. Monitoring, screening, additional continuing education, or retesting the concerning minority may be more directed.

After Rozema clarified the eventual pass-rate comparison, Plemmons voted no on the recurring bar proposal: keep the bar as is, while considering future pathways. Hicks agreed.

Yun Oh also leaned toward the current system. A higher pass rate could increase regional labor supply, he said, but repeated testing may become a barrier for incumbent practitioners. He cited a 2024 study by Shakia and colleagues on maintenance-of-board exams for physicians, assigned every ten years. According to Oh, the exam created a barrier for older physicians: someone facing a large exam at age 62, with perhaps three or four more years of work ahead, may simply retire. He said some states reduced those barriers by offering alternative ways to maintain certification, increasing retention of older workers by 1.4 to 1.9 percentage points biannually.

Oh’s conclusion was that exams imposed late in a career can unintentionally reduce supply by pushing experienced workers out. He hoped that one entry exam, followed by market sorting, would be sufficient to distinguish good from bad practitioners in law.

Medicine pushed the same design problem in the opposite direction. The panel was more receptive to reducing recurring exam burdens for physicians, but not to eliminating ongoing competence requirements altogether. The reason was not simply that doctors are different from lawyers; it was that continuing education, pace of knowledge change, and the visibility of harm change the regulatory calculus.

Plemmons said she would support a one-time physician exam only because she believed medicine already has strict continuing education requirements with competency pieces built into them. If physicians are completing continuing education and demonstrating competence through that process, that might substitute for recurring large exams. She still wanted more research.

Hicks was also receptive to reforms that would reduce burdens and keep healthcare practitioners practicing, especially given sensitivity in Canada about the supply of doctors and other healthcare providers. But he pushed back on Rozema’s “old lawyer, young physician” formulation. Hicks characterized some research as pointing instead toward a middle-aged doctor: experienced, not right out of school, and still up to date with the literature. Continuing education is what preserves that balance.

Oh emphasized the pace of change in medicine. Medical knowledge changes quickly, he said, using shifting advice about carbohydrates, fats, and protein as an example. Because medical mistakes can produce serious harm, physicians need continuing education and some mechanism to ensure they meet minimum standards.

Hicks added that the consequence of error is more visible in medicine than in law. It may be difficult to trace a bad legal outcome to insufficient training, but when a physician makes a serious mistake, the public can more readily connect the error to training or standards, whether or not that connection is empirically correct. That political visibility makes reform harder.

Alternative pathways drew support when they could lower entry costs without abandoning training structure. The ABA-accredited law degree requirement was the central example. Rozema noted that most states require lawyers to graduate from an ABA-accredited law school, while some allow apprenticeship models. In California, he said, someone can study law under a judge or practicing lawyer, but still faces examination requirements, including the “baby bar” before taking the regular bar. He also said the Texas Supreme Court had recently voted to remove the ABA degree requirement.

Alicia Plemmons was strongly favorable toward apprenticeship pathways. She drew first from a non-law example: West Virginia manicurists can go through an apprenticeship. They still take in-person safety and training classes, then learn on the job. Plemmons said she has seen that model create a route for people who left high school but were not ready for college, allowing them to learn from experienced workers while building practical skills.

She acknowledged that manicurists are not lawyers, but argued that society should think more seriously about what workers can learn from existing experts. Classroom education is often one expert teaching students. Apprenticeship can mean learning under multiple experts in real workplaces. She voted yes on removing the ABA requirement, with caveats: apprenticeship mentors might need ABA approval, additional training, or specific educational obligations before supervising apprentices. Her underlying claim was that not everyone needs a four-year college route, and that much professional learning happens under real practitioners rather than through formal schooling alone.

Jason Hicks agreed with the promise of apprenticeships but wanted more detail before voting. He asked what structured education would occur inside firms. Would apprentices gain real legal knowledge, or merely push paper while that counted as training? The design of the apprenticeship would determine whether it worked.

He also connected apprenticeship to rural workforce development. In skilled trades, he said, apprenticeship models can be valuable in rural areas because training institutions may not be nearby. A similar model might help legal access in rural Wisconsin, rural Washington, or other places where lawyers are scarce. If apprentices can shadow and work under lawyers in those communities, the pathway might increase the local supply of lawyers.

But Hicks still questioned whether the bar exam is enough to guarantee quality. The existence of an exam matters only if the exam reflects actual competence and reduces the downside risk the licensing system is meant to prevent.

Yun Oh brought in evidence from plumbing apprenticeships. He described a 2022 paper by Warner and Bonato on Washington plumbers, comparing formal apprenticeships with fixed syllabi and training against an alternative pathway of working for other plumbers for 5,000 to 8,000 hours before obtaining a journeyman-level license. A measurable quality outcome was injury. According to Oh, workers who went through formal apprenticeships had lower injury rates and lower likelihood of injury-related workers’ compensation claims.

The apprenticeship discussion therefore did not become a simple argument against formal education. Plemmons’s strongest case was for alternative pathways. Hicks’s concern was whether the alternative contains actual training. Oh’s evidence suggested that formality inside apprenticeship can matter: learning on the job may work better when it includes a syllabus, supervision, and defined standards.

Character-and-fitness review in law is a selection device, not a training device. Rozema described the current process as extensive and opaque: applicants submit employers, traffic tickets, criminal justice involvement, and other background information. He said denial rates appear small, perhaps around 1% or less, but characterized the process as a black box.

His hypothetical made the trade-off sharper. A machine could be trained to predict the likelihood that an applicant would later produce a bad outcome. The state would deny the same number of applicants as before, but use the algorithm instead of individualized portfolio review. Assume the algorithm is more successful at reducing bad outcomes. Should the state adopt it?

Jason Hicks said he would support it only with oversight. He would not hand the decision entirely to a machine, and he doubted the public would accept that. As a public official concerned with reelection, he would need to justify the system and show human correction, checking, and verification. The magnitude of the gains would matter: is the algorithm 10% better, 30% better, or only marginally better? The answer would affect his vote. His broader point was that policymakers often make decisions under imperfect evidence, and the political context is part of the policy environment.

Alicia Plemmons rejected the premise most sharply. She compared it to Minority Report: predicting that someone may become a bad actor rather than responding to something actually done. She worried that algorithmic screening could become “thought crime,” denying people work based on predicted future misconduct and undermining fairness and the possibility of growth.

But she did not reject data analysis entirely. Plemmons said AI could help identify which prior crimes are actually relevant to a profession. Many states still have good-moral-character requirements under which certain crimes can block licensure. That can affect occupations one might not expect, such as barbers, and can also affect substance-use-disorder counselors who have prior substance-use convictions, completed recovery, and may be especially well suited to help others. She described reform efforts in which states require boards to report denials based on crimes and require the crime to be related to the licensed occupation.

Her example was direct: an arson conviction may be relevant to licensing a firefighter; a bounced check at age 20 should not be. AI, in her view, might help analyze real data to determine which crimes are substantially related to later licensing problems. But she opposed a black-box system that predicts whether someone will be bad at a job before they have the chance.

Yun Oh framed the issue as a welfare comparison. Algorithmic screening might increase consumer welfare by reducing bad outcomes. But it could reduce worker welfare by denying people the chance to work. The question is whether the consumer-welfare gain is large enough to compensate for the worker-welfare loss. Without comparing those magnitudes, he would not decide.

Hicks then shifted the discussion toward transparency in criminal-record restrictions. He referred to work by Peter and Bobby on screening, felony convictions, and the Black-white wage gap, and to his own experience collecting data on criminal-record restrictions during graduate school. The laws were difficult to interpret: caveats, offense lists, clauses, and ambiguities made it hard to tell who was restricted. Even lawyers sometimes could not give clear interpretations without case-specific analysis. Hicks argued that making these rules more transparent and focusing restrictions on crimes related to the occupation would go a long way.

The current black box did not receive much defense, but replacing it with another black box did not satisfy the panelists either. Relatedness, explainability, oversight, and the magnitude of welfare trade-offs would determine whether a mechanical filter is defensible.

Professional discipline after licensure raised a different question: whether disclosure should be designed for deterrence, consumer information, research, or punishment beyond the formal sanction.

Rozema distinguished three channels. Suspension incapacitates a practitioner: they cannot practice during the suspension. Punishment may also create specific deterrence, changing that person’s future behavior. But making discipline public adds general deterrence: other practitioners may behave better if they know punishment will be visible.

Law has private discipline in many states, Rozema said. Some disciplinary action is not publicly announced, and researchers cannot collect data on privately disciplined lawyers. By contrast, California publicizes suspensions and disbarments, including through its State Bar account on X. Rozema asked whether a state should post every suspension or disbarment on X.

Alicia Plemmons voted no on mandatory posting to X. She acknowledged that public shaming can change behavior, citing a study she could not name in which publishing pictures of people delinquent on child-support payments in a local newspaper increased payment. Behavioral economics, she said, shows that people respond to public exposure.

But she did not think regulators should require social-media shaming. What she did want was usable disciplinary data. Plemmons works with malpractice and adverse-action data, and said private discipline creates a serious measurement problem. A state may appear not to have a quality or safety problem simply because private discipline is invisible. She would support mandatory internal records or data availability for research, so analysts can study quality and safety concerns. The data need not be turned into a social-media penalty to be useful; it could be collected in a form that allows boards, researchers, and policymakers to see whether discipline is concentrated in particular occupations, firms, violation types, or regulatory structures.

Jason Hicks also voted no. He questioned why X specifically: if the policy is social-media disclosure, why not Threads or every platform? He did, however, support more transparent disciplinary processes. He described British Columbia’s recent re-regulation of many occupations: as he characterized it, the government would no longer allow practitioners to handle the disciplinary process by themselves, would add public members and government-appointed participants, and would make disciplinary actions publicly available. For Hicks, posting on a website or making discipline accessible may be reasonable; tweeting every case did not have clear added value.

Yun Oh also opposed the X policy. Discipline, he said, should punish but also allow correction. If a practitioner is suspended, the aim is not only to remove them temporarily but to encourage better conduct when they return. Online publication can become permanent. A name posted today may still be searchable one, two, or five years later. He compared it to a lasting one-star Yelp review: even after returning to the profession, the practitioner may be marked indefinitely, making rehabilitation harder.

The policy distinction was between durable, structured transparency and performative public exposure. The panelists were skeptical of mandatory social-media posting as a tool of general deterrence. They were more open to records that would let regulators track discipline, researchers study quality and safety, and consumers or public members understand serious disciplinary histories without turning every sanction into an indefinitely amplified reputational penalty.

An audience member challenged the panel from the demand side of the market. The speaker argued that many examples discussed in the United States have already been tried in other countries, and that policymakers should look more systematically at those experiences. In Europe, the speaker said, doctors are not regularly tested in the same way; whether they are worse than American doctors is an empirical question worth examining. In the United Kingdom, the speaker said, legal qualification has recently allowed different educational backgrounds or apprenticeships, followed by a common exam.

The same audience member argued that the core question is not whether lawyers should be licensed, but what type of license is needed. There is a spectrum of risk, externalities, and information asymmetry. Entry barriers may be justified, but conduct barriers can duplicate the same protections. For example, lawyers may face restrictions on limited liability, ownership of legal firms, outside investment, or multidisciplinary practice, sometimes on safety grounds already addressed by training.

The audience member’s strongest challenge concerned quality data. Malpractice data, they argued, is limited because professional associations may receive reports and have little interest in disclosure. Instead, there should be an independent system, public data, and analysis that empowers consumers. In a market with high information asymmetry, regulating supply alone is insufficient. Demand should be empowered through transparency: if a consumer hires a lawyer, they should be able to see the practitioner’s history, what kind of malpractice occurred, and how serious it was.

Hicks responded that this was precisely why he had raised British Columbia. He said BC’s reorganization of regulatory colleges followed that rationale: changing who controls discipline and making disciplinary information public. He also noted that knowledge of such reforms is limited even across borders; Plemmons had not known about the BC change, despite its proximity to the United States.

The line that emerged was not between secrecy and disclosure. The speakers were skeptical of social-media shaming as a regulatory tool, but more open to disciplinary transparency for research, oversight, and consumer information. The hard question was what should be public, at what level of detail, through what institution, and with what protection against permanent punishment after the formal sanction has ended.

The final exchange returned to Rozema’s original complaint about expert advice. An audience participant argued that policymakers often ask researchers for a snap yes-or-no answer after the policy has already been designed. Experts may respond that they should have been brought in weeks, months, or years earlier. If the goal is a real outcome rather than reelection, the speaker argued, researchers and policymakers need ongoing collaboration early in the process, not one-off requests at the end.

Jason Hicks agreed and connected the problem to risk aversion. If a policymaker is forced to vote on the spot under uncertainty, the default will often be no. The downside risk is difficult to quantify, and supporting reform can be politically costly. That is why long-term relationships between researchers and politicians matter: they create a channel for evidence to inform design before the decision is locked in.

Alicia Plemmons noted that she regularly brings legislators together to talk with them, while another participant said such “fly-ins” happen all the time in some circles. The implication was not that economists should stop being careful. It was that the institutional relationship between evidence and policy is poorly timed. Researchers want to analyze design choices before they harden; policymakers often ask for a vote only after the choices have been made.