The AI Journal

California required Dr JoAnne Cornwell to spend nine months and sixteen hundred hours in a state-approved cosmetology school before she could legally braid hair for pay, none of which taught braiding. The curriculum covered chemical relaxing, permanent waving and razor cutting; braiding appeared nowhere in it. When Cornwell and the American Hairbraiders and Natural Haircare Association sued, a federal court in San Diego agreed in 1999 that the requirement bore no rational relationship to the skill it supposedly certified. The case reads as a clean instance of what sociologists since Weber have called social closure: an occupational group, or the board that regulates on its behalf, restricting entry to protect the position of those already inside. Frank Parkin gave the concept its modern sociological form in 1979, defining closure as the process by which a collective maximises its rewards by limiting access to a circle of eligibles, and licensing boards have supplied the textbook illustration ever since.

The trouble is that the best available evidence on what licensing actually does to entry does not support that picture as a general account. Beth Redbird’s 2017 study in the American Sociological Review, built from a thirty-year panel exploiting how licensure requirements vary across states for the same occupations, found that licensure does not on average restrict who gets in. It standardises how they get in. Before licensure, entry to a trade typically runs through informal channels: apprenticeship, family connection, word of mouth. Licensure replaces that with a codified, publicly known sequence of requirements, and Redbird found that this codification draws in workers who lacked the informal ties the old system rewarded, women and racial minorities among them. Participation rose. Wages, meanwhile, did not: the same standardisation that widened the door also compressed the skill distribution inside it, erasing the premium licensed workers might otherwise have commanded. Morris Kleiner and Alan Krueger’s 2013 survey in the Journal of Labor Economics, the first nationally representative accounting of the practice, had already shown how large the system had grown: 29 percent of the American workforce held a government-issued occupational license by the mid-2000s, up from under 5 percent in the 1950s, with licensing’s wage effect running well below what the closure story would predict once certification and informal credentials are separated out from binding licensure.

Put those two findings together and the standard reform prescription looks aimed at the wrong mechanism. The Hamilton Project brief that Kleiner wrote for Brookings in 2015 proposes, among other measures, reclassifying licensed occupations to certification or no regulation wherever politically feasible, on the reasoning that licensing imposes costs on entrants without corresponding gains in service quality or safety. That case is easy to make for occupations regulated well past any plausible safety rationale. It is much harder to make once Redbird’s finding is taken seriously, because deregulating a codification regime does not simply remove a barrier; it also removes the standardised, legible entry path that drew in the outsiders who benefited from it. An occupation that is exclusionary in Cornwell’s sense, where the licensing curriculum has no relationship to the skill practised, is not the same problem as an occupation that is merely standardised in Redbird’s sense, where the curriculum roughly tracks the work and the main effect of removing it would be to hand entry back to informal networks. Treating both as instances of one mechanism, closure, and prescribing one remedy, deregulation, for both is where the reform argument goes wrong. It is aimed at the average effect of a policy that operates through two distinguishable mechanisms with opposite implications for who should want it changed.

The diagnostic that follows is not exotic; it asks a question the reform literature mostly skips. Does the licensing curriculum bear a demonstrable relationship to the tasks the occupation performs, of the kind a court applying rational basis review can actually locate, as it could not for Californian hairstyling? Where it does not, as with braiding schools required to teach relaxer chemistry to people who do not relax hair, the closure diagnosis holds and deregulation or curriculum reform is the right target, because there is no codification benefit being traded away, only a cost being imposed. Where curriculum and task line up reasonably well, as with residential electrical work, where a wiring error can burn a house down and the licensing exam actually tests wiring, deregulation trades a real, empirically documented gain in access for a saving in training cost that will disproportionately be recaptured by the same informally connected entrants the license had been standardising around, while removing whatever protection the curriculum does provide against the failure mode it was written to prevent. Andrew Abbott’s account of professions competing over jurisdiction, developed the same decade as Parkin’s, supplies the missing piece: occupational groups do not merely exclude, they contest and redraw the boundaries of what counts as their expert task, and a licensing curriculum is one of the instruments by which a boundary gets drawn well or badly. A board’s incentive to draw that boundary self-servingly does not disappear simply because the curriculum happens, in a given trade, to track genuine hazard; incumbents captured plenty of electricians’ boards too. What changes is only whether the capture and the curriculum point the same way or in opposite directions, and that is precisely the distinction a blanket rule cannot see. A reform agenda that asks, occupation by occupation, whether the boundary is drawn accurately is doing something different from one that treats every license as a rent extracted by insiders, and it will recommend deregulation for some trades, curriculum reform for others, and no change at all for a good many.

The obvious objection is that Redbird’s own result undercuts the case for reforming anything at all, since if licensing does not hold down wages or shut out outsiders on average, the policy problem it is supposed to justify may simply not be there in the volume critics assume. That objection has real force against the sweeping version of the closure argument, the one that treats every license as evidence of capture, and it should be conceded rather than argued around: most licensed occupations are probably not costing entrants very much, and campaigns aimed at licensing as such are chasing an effect too small and too average to matter for most of the trades caught in it. But an average effect across three hundred occupations is compatible with real closure at the tail, concentrated exactly where the curriculum diverges furthest from the task, which is where Cornwell’s plaintiffs and the plaintiffs in the string of hair-braiding suits that followed it kept finding themselves. The reform case survives Redbird’s finding once it stops claiming the whole system as its target and starts claiming the cases where the curriculum-task mismatch is demonstrable rather than assumed. That is a narrower case than the deregulation literature usually makes, and a harder one to win occupation by occupation than a single statute reclassifying licensure to certification across the board. It is also the case the evidence actually supports.