Occupational licensing is sold as consumer protection that screens out incompetent or dangerous workers. Today, about 22 percent of the American workforce needs a government license to do their jobs, and that safety and quality argument is what state and local lawmakers use to justify the entry barriers, licensing fees, and training hours that typically accompany it.
The trouble is that the claimed benefits have historically been difficult to test or measure, so the debate over whether licenses really protect anyone was fueled mostly by anecdotes. That’s why a new NBER working paper by Jonathan Hall, Jason Hicks, Morris M. Kleiner, and Yun taek Oh is so interesting. It leverages Uber data and finds no consistent evidence that a government occupational license improves riders’ ratings or drivers’ behavior on the road. The results suggest that, once a platform is already screening and monitoring its drivers, licensing adds no detectable, additional consumer protection.
Why Uber Is a Good Testing Ground
Uber’s vast datasets from trip tracking and consumer ratings have proven a goldmine for researchers on an array of economic questions.
One of us has written before about how data on female and male drivers’ earnings gave good insights into non-discriminatory causes of gender pay gaps. Now similar data can be used to examine the effects of occupational licensing laws for rideshare drivers.
Uber dispatches drivers mainly by proximity. So when a licensed and an unlicensed driver are both near a rider, which one is dispatched is essentially a coin flip. The researchers use that quasi-random assignment along with data Uber collects about the ride quality, such as the rider’s star rating and telematics data on how carefully the car was driven (by measuring things like hard braking and hard acceleration, which are tied to greater crash risk).
The authors exploit two settings that allow a test of licensing’s effects. First, New York City licenses rideshare drivers through the Taxi and Limousine Commission, which, at the time covered by the study, required a fingerprint background check, a defensive-driving course, and a 24-hour vehicle for hire course and exam. Neighboring New Jersey does not have these requirements, yet drivers from both areas serve the same New Jersey riders.
Second, in 2017, Texas preempted local occupational licensing of rideshare drivers, abolishing Houston’s restrictive licensing requirements. That let the researchers compare previously licensed drivers with those who entered the market after the entry restrictions were lifted.
The researchers sampled 213,000 trips in the New York area from April to August 2017 and nearly 497,000 in Houston from September 2017 to January 2018 to look for differences between drivers with and without an occupational license. They checked for differences across seven metrics:
- Star rating (1–5)
- Proportion receiving ratings less than 5 stars
- Proportion receiving 1‑star ratings
- Prevalence of hard braking
- Whether more than 20 percent of brakes in a trip were hard brakes
- Prevalence of hard acceleration
- Whether more than 20 percent of accelerations in a trip were hard accelerations.
No Sign Licensing Helps Anyone
Across both metro areas, the paper finds no consistent evidence that licensing improved consumer outcomes.
In the New York vs. New Jersey comparison, licensed drivers actually earned slightly lower ratings than their unlicensed counterparts—0.0228 fewer stars on Uber’s five-star scale, or less than half a percent off the unlicensed mean score of 4.78.
Just one of the seven metrics compared came out in licensing’s favor. Licensed drivers logged fewer trips with a high share (20 percent or more) of hard-braking events, yet even that result didn’t survive robustness checks the authors undertook. It faded once the authors widened their sample window and swung wildly depending on whether they controlled for vehicle model and year.
The results were similar in Houston. The researchers could not detect a statistical difference between the drivers who entered after deregulation and previously licensed drivers on all seven outcomes, despite the two groups differing sharply in experience (288 prior trips versus roughly 2,584 for the licensed) and age. This is strong suggestive evidence against the common fear that ending entry requirements would flood a market with worse providers.
When previously licensed drivers were instead compared with unlicensed drivers who had already been on the platform, the only differences ran against licensing: on three of the four driving-behavior measures, the previously licensed drivers braked and accelerated harder, with roughly 15 percent more of the high-hard-braking trips that flag risky driving.
Licensing Is a Costly Regime
A large body of research has already established the costs associated with occupational licensing. A strict licensing regime restricts entry into a market, thins the supply of workers, and thus tends to push prices up.
As was summarized earlier this year in Cato’s Handbook on Affordability, stringent nurse-practitioner scope-of-practice rules raise child checkup costs between 3 and 16 percent, home improvement jobs are between 15 and 50 percent more expensive (depending on the job type) in states with the strictest home technician licensing regimes, and in one Virginia case study of hair braiding, the number of beauty shops grew 7 percent faster than in bordering states after Virginia deregulated in 2012.
So, although this paper doesn’t examine how occupational licensure affects fares, it’s reasonable to think based on existing literature that licensing rideshare drivers simply means fewer drivers and (absent substitute platforms outside the license regime) higher fares. This paper supplies evidence that despite that added cost, licensure often buys nothing in the way of improved quality or safety.
Rideshare markets already have methods that go a long way to vetting drivers’ quality—reputational feedback, ratings, reviews, or the threat of account deactivation. That makes further government licensure redundant. The new paper is a strong complement to what Cato argued in our handbook earlier this year: the cheapest way to lower the price of a service is often just to let people provide more of it.
