The Heritage Foundation

Backgrounder #3159 on Labor

September 28, 2016

September 28, 2016 | Backgrounder on Labor

Understanding the Data on Occupational Licensing

Data from the 2015 Current Population Survey (CPS) shed light on the extent and patterns of occupational licensure across the United States. Occupational licensure—the requirement that employees obtain government permission to work in certain occupations—is most common among women and high earners. Workers below age 35 are less likely to have a license. The rate of licensure differs little among the states, and much of the variation is explained by demographics. The licensure rate is highest in Maine and Alaska, and lowest in Georgia and North Carolina. The CPS also shows the limitations of data on this subject: Even in a large nationwide survey, few professions are numerous enough to statistically analyze on their own. Furthermore, data from survey respondents are not very reliable: 18 percent of employed physicians and surgeons in the data claim that they do not have a government-issued certification or license. Nonetheless, policymakers need not wait for perfect data on licensure to begin reviewing and reforming occupational regulations.

Key Points

  1. The regulation of professionals by occupational licenses, certifications, and registration has spread over a large share of the U.S. workforce. Experts are recognizing how little is known about the extent of licensure in America.
  2. The Current Population Survey (CPS) is the most extensive survey on occupational licensure, but it leaves much to be desired.
  3. Occupational licensure is more common among women, high earners, and workers in their 30s and beyond. Licensure rates are similar across the states, but Maine and Alaska have the highest rates of licensure, while Georgia and North Carolina have the lowest.
  4. Future data, including the 2016 CPS, will continue to refine knowledge of licensing. For each profession, policymakers should pursue the least-intrusive regulatory approach that is consistent with public safety.

The regulation of professionals by means of occupational licenses, certifications, and registration has spread over a large share of the U.S. workforce in recent decades. Even as policymakers are recommending a reduction in the extent of the regulations—a positive move overall—experts are recognizing how little is known about the extent of licensure in America. New data from the Current Population Survey (CPS), a major survey performed by the Bureau of Labor Statistics (BLS) that is used for determining such statistics as the unemployment rate, indicate that licensure is widespread, with modest differences among the states. Mature workers, women, and those in high-wage occupations are more likely to hold licenses.

The CPS data augment knowledge of the extent and patterns of licensure, complementing previous efforts at measurement. The most commonly cited previous metrics of occupational licensure are the Institute for Justice’s 2012 “License to Work” study and a 2013 Harris Poll survey. While each of these two has its strengths, the new data offer a larger sample size and the ability to connect licensure to other economic and demographic characteristics in the CPS.

As economists work to better measure the extent and effect of licensure, state legislators should begin reviewing and reforming their licensure regulations in order to make them consistent with the public good. Currently, licensure boards are largely unaccountable, and usually dominated by industry insiders whose interests are at odds with those of customers. In each occupation, legislators should replace this system with the least intrusive form of occupational regulation consistent with public safety.

Measuring Occupational Licensing

An occupational license is a government-issued permission slip that is required for work in a given profession. Truck drivers and cosmetologists, for example, typically have state-issued licenses. Some workers, such as taxi drivers, may be licensed by their city. Workers in a few occupations, such as airplane pilots and merchant mariners, must obtain a federal license in order to work. Obtaining a license usually involves a minimum age, certain educational requirements, paying fees, and passing tests. To keep a license valid, most workers have to complete continuing education courses and re-register periodically.

Recently, the CPS added a sequence of questions about occupational licensure to its ongoing polling.[1] Since 2015, CPS respondents have been asked the following sequence of questions:

  1. Do you have a currently active professional certification or a state or industry license? Do not include business licenses, such as a liquor license or vending license.
  2. Were any of your certifications or licenses issued by the federal, state, or local government?
  3. Is your certification or license required for your job?[2]

An error in administering the third question led the BLS not to report it with the 2015 data; users are left with the answers to the first two questions.

This Backgrounder refers to those who answered “yes” to the first two questions as “licensed” workers. That is not precise, but it is concise.

Survey Error. People are not, as it turns out, very good at answering surveys accurately. For example, about half of those who receive food stamps tell the CPS surveyors that they do not, and 10 percent of those who tell the CPS that they receive food stamps do not, in fact, receive them.[3]

There are other sources of error—BLS employees must interpret vague answers and they sometimes make errors in their data entry. There is also classical sampling error, which means that some surveys will randomly include a higher percentage of barbers and a lower percentage of nuclear engineers than the population as a whole. The smaller an occupation or a geographical area, the more sampling error becomes a problem. However, statistical theory gives us a good idea of exactly how much sampling error researchers can expect; with the other sources of error researchers are usually in the dark.

For most occupations, the CPS sample size is too small to compare across states. Table 1 notes the number of observations in the CPS data for each occupation in which at least three-fourths of workers report having a license.[4] With so few people in each occupation surveyed, it is clearly impossible to use the CPS to compare individual occupations across states: Each state would have an average of two physician assistants, three occupational therapists, five dentists, and so on. Thus, the CPS can offer little insight into the specific legal barriers to work that each state has enacted.

The CPS questions on licensure do not inspire confidence in people’s ability to answer surveys accurately. For example, 18 percent of “physicians and surgeons” in the labor force do not have a government-issued certification or license according to the answers given to the CPS. This pattern of underreporting is true across the many occupations in which all states require a license and extensive illegal work seems unlikely.

By contrast, when 41 percent of barbers report that they are unlicensed, some of them may be telling the truth: They earn money cutting hair, but do so without a license. Given the degree of error among physicians, however, there is no way to know how many barbers are working outside the law and how many are merely forgetful.

People interviewed by the CPS are slightly more likely to remember their own license than that of another household member. After taking occupation, age, and sex into account, employed people who were directly interviewed reported a 2 percentage point higher licensure rate than those whose information was given by another member of the household.[5] Expressed another way, those who answer for themselves are 8 percent more likely to report having a license.

Licensure response rates are higher among whites than blacks and Hispanics. For example, in the largest highly licensed profession (registered nurses), 83 percent of whites report having a license, but only 70 percent of blacks and 72 percent of Hispanics do so. In all likelihood, any of these practicing nurses is, in fact, licensed. But recall or reporting appears to be substantially higher among whites.[6] Unsurprisingly, reporting is also lower than average among foreign-born respondents.

Among households with exactly one man and one woman between ages 30 and 65, the reported licensure rate does not differ significantly between male and female respondents.[7]

Licensed Workers. Even with imperfect data, one can learn some things about licensure, particularly on a national scale. The clearest pattern in the data is that licensure rises with age. Among those employed at age 21, just 7 percent are licensed. By age 27 the share rises to 20 percent, and in the mid-30s it levels off around 25 percent through the rest of the working years.

This pattern fits within a broader one: By their mid-30s, most workers have settled into occupations where they will remain for the rest of their careers. Young workers, however, bounce around. One study showed that the typical American man holds seven full-time jobs in his first 10 years in the labor force, and just three or four thereafter.[8] That is healthy: Young workers need to learn what they are good at, and job switches are often accompanied by pay raises. One of licensure’s costs is that it may inhibit young workers from readily switching between occupations as they look for a good fit.

The list of highly licensed occupations in Table 1 confirms that licensed workers are concentrated in health professions. The 22 occupations in Table 1 account for 23 percent of all licensed workers, according to the CPS. However, recall and coding error rates might not be similar for usually licensed and rarely licensed occupations, so this ratio is far from certain.

Licensure is most common among higher earners. Dividing the employed population over age 29 into quantiles according to weekly earnings shows that licensure is increasing across the earnings distribution. Chart 1 shows the reported licensure rates across the 30-plus workforce. About 30 percent of workers in the top third of the earnings distribution are licensed. Notably, the average hourly wage of every occupation listed in Table 1 is in the upper half of the distribution.


Because young workers tend to have low wages and low licensure rates, adding them to the chart would further lower the licensed share among the lower quantiles.

Among all workers, women are 6 percentage points more likely to be licensed than men, which is no surprise given that female-dominated professions in health and education are among the largest licensed occupations. Accounting for occupation and age, the gap disappears completely; sex differences in licensure are due to occupational differences.

There are large racial differences in licensure. Unlike the gender gap, this is not entirely due to occupational differences. Blacks are 23 percent less likely to report holding a license than whites, and Hispanics are 48 percent less likely than whites to report a license.[9] Even accounting for occupation, age, and sex, Hispanics and blacks were both substantially less likely than whites to report licensure.

Licensure Across the States. Since licensure laws are generally set at the state level, one expects to observe significant differences in the rate of occupational licensure across states. However, it is not obvious whether states with more restrictive approaches will have more or fewer licensed workers. States that require licenses in a broader range of occupations will have more licensed workers. But states that have more restrictive requirements for a given license will have fewer licensed workers.

Table 2 shows the licensure rate across states, unadjusted for demographics. These figures should be interpreted with caution, however, and will be revised and refined as future data from the CPS and other sources increase the sample size and improve the accuracy of these estimates. Based on data published so far, Georgia and North Carolina are the least-licensed states, while Maine and Alaska are the most licensed.

To investigate how common licensure is, conditional on the mix of occupations and demographics, I predicted what share of employed workers would be licensed in each state if the licensure rate in each of its occupation/age/race/foreign-born[10] groupings was the same as the national average. This is not quite perfect, since some occupations are grouped together, but it compensates for the concentration of farmers in Nebraska and the older-than-average population of Maine, for example.


The CPS sample is small enough that most states do not differ significantly from their predicted average.[11] However, the patterns are striking even where they are statistically shaky, as displayed in Chart 2.

Almost all the states in the northwestern U.S., from Wisconsin to Alaska, are more licensed than their demographics and occupational mixes predict. In addition, several states in the eastern U.S. are more licensed than predicted. There is no obvious demographic or industrial commonality that links the eastern and western groups.


The lower-than-predicted licensure states are concentrated in the eastern U.S. However, only Delaware, North Carolina, and Georgia have licensure rates significantly lower than predicted.

Conversely, one can abstract from state laws and identify which states are likely to be most licensed purely because of their occupational mixes. The states with higher concentrations of occupations that tend to be licensed are concentrated on the east coast, particularly in states with large shares of retirees and relatively weak economies, such as Maine, West Virginia, and Kentucky. The states with lower concentrations of highly licensed occupations are in the Midwest and West, including California.

Comparison with Other Data Sources. As limited as the CPS data are, they add significantly to the ability to understand the extent of occupational licensing. The two sources that have been most widely used by policy experts to compare state-level differences in occupational licensing are a 2013 Harris Poll and a 2012 report by the Institute for Justice (IJ).


Harris Poll. The CPS starkly contradicts the state-by-state rankings of the Harris Poll, although it finds a very similar national licensure rate.[12] The Harris Poll’s state-by-state licensure rates have been widely cited, including by me.[13] However, the Harris Poll survey was based on 9,850 employed respondents nationwide, compared to the 184,915 employed respondents in the CPS outgoing rotation group sample used in this Backgrounder. The sample size in the Harris Poll survey implies that the results were based on about 42 licensees per state.

Although the sources agree on the national average level of licensure among the employed (about 22 percent in both cases), they are not even on the same page for regional averages, let alone states. In the CPS data the West is the least licensed region and the Northeast is the most licensed. In the Harris Poll survey, by contrast, the West is the most-licensed region and the Northeast is second-least licensed.[14]

As the CPS data show, it takes a very large sample size indeed to accurately rank what appear to be mostly small variations among the state-level data. For national estimates, however, the Harris Poll’s detailed suite of questions about licensure offers a more detailed view than the three questions asked in the CPS.

Institute for Justice. The CPS data also failed to fully corroborate substantial differences in state statutes licensing low-wage and moderate-wage occupations as compiled by the Institute for Justice in its landmark 2012 study, “License to Work.” To compare the CPS data on individuals to the IJ data on statutes, I matched the professions in the IJ data to four-digit Census occupational codes in the CPS. For each matched occupation, I used IJ’s database to define a group of states that licenses the occupation and a group that does not. Then I limited consideration to 18 occupations with at least 20 observations in each group of states.

One would expect that where states require a license for an occupation, the reported licensure rate would be high, and that it would be low where there is no requirement, with allowance for errors and exceptions. But lining the CPS data up against the “License to Work” tabulation of statutory requirements does not yield the expected pattern.

Among 17 of the 18 occupations considered, the CPS data show licensure rates higher in the licensing states than in the non-licensing states. That is consistent with the IJ data. However, most of the differences were quite small and statistically insignificant. Chart 3 shows the pattern: Licensure rates were slightly higher in states that IJ listed as licensed than in those it did not, but the differences are statistically significant for just two occupations: dental assistants and opticians.

In many cases, the samples are small and the definitions are an imperfect match between the two data sources. A typical case was “childcare workers.” In the licensing states, 18.1 percent of child care workers reported having a license; in the non-licensing states, 13.9 percent reported having a license. Given the margins of error, it is possible that there is no difference in the true licensure rates of child care workers in those two groups of states. One reason that so few child care workers appear to be licensed is that the statutes, according to IJ’s summary, apply only to center-based day care, not to child care workers in homes and other settings.


Small sample sizes are also a problem in the CPS. For example, there are only 104 travel agents in the sample. Of these, 34 are in the eight states that license travel agents, according to IJ, while 70 are in the remaining states. In the non-licensing states, just 7 percent reported having a license, while 25 percent did so in the licensing states. Although that is a substantial difference, it is not—with so small a sample—statistically significant. Furthermore, one learns from “License to Work” that travel agents are required only to register with their state and pay a fee in order to practice. Thus, many registered travel agents may not think of themselves as holding a “license or certification.” Still, the registration and fees are unnecessary barriers to work that function economically in the same way as a licensing requirement.

Another reason for the poor match between the datasets is that some occupational categories are too broad in the CPS. One example (which remains in this Backgrounder for illustrative purposes) is the group of occupations titled “weighers, measurers, checkers, and samplers, recordkeeping.” The IJ study looked only at “weighers.” The CPS does not say how many in its broader category are “weighers,” nor what share of those who call themselves weighers would be covered by the licensing statutes that IJ has tabulated.

There are broad reasons that might explain the poor match between “License to Work” and the CPS data. Errors are a primary reason: Sampling error, recall error, and coding error—the last on the part of CPS workers, IJ researchers, and this author—are all possible. Some workers will have a license that is not required for their occupations. A midwife may hold a license as a nurse or emergency medical technician even where it is not required. Other workers might not hold a license that is required, working illegally or in a gray area where it is not clear whether the occupational laws apply. Finally, laws may have changed in the three years between IJ’s survey and the 2015 CPS.

Still, the failure of the CPS to clearly corroborate “License to Work” data means that both datasets resist clear interpretation, even as each one adds to the understanding of the nature and extent of licensure in the U.S. The particular circumstances of child care workers, travel agents, and weighers, each discussed above, show that licensure interacts differently with each profession. Given the data currently available, researchers will be best served by pursuing fine-grained case studies that draw on all available sources rather than attempting to use either IJ or CPS data for broad cross-sectional inference.

Restricting Employment Without Evidence of Health or Safety Risk

Policymakers do not always need accurate data to do the right thing. So far in 2016, Iowa,[15] Kentucky,[16] and Nebraska[17] released ancestral African hair braiders from the requirement of licensure by cosmetology boards, although the braiders have never been enumerated.

Furthermore, the absence of hard data is a rebuke of licensure laws, which have often restricted economic opportunity on the basis of unsubstantiated fears rather than rigorous analysis. As the balance of evidence reveals that licensure is neither a guarantee of ethical behavior nor an effective screen for quality, states should repeal those that are not grounded in firm evidence of consumer protection.

An intermediate step, as states consider which licenses are truly necessary, is for states to make more data concerning licensure publicly available. Simply releasing annual data that is already collected—on the number of licenses granted and renewed by each board, for example—would aid empirical researchers in drawing accurate empirical conclusions about the effect of licensure on consumer protection, wages, and economic mobility.

Principles and Policies. In principle, licenses are intended to protect consumers from unscrupulous or dishonest professionals, who might otherwise misrepresent their expertise. In most occupations, customers, bosses, or colleagues can quickly weed out a pretender, and the harm done by an unqualified professional would be small and reparable.

For licensure to be reasonably applied, four conditions must be met:

  1. Proper preparation must significantly improve the quality of work. Training is sometimes a poor substitute for professional practice. For example, licensed florists perform no better at flower-arranging than unlicensed ones.[18] Likewise, cosmetology schools rarely teach hair braiding or eyebrow threading, but some states have required that practitioners of both first attend cosmetology school and obtain a license unrelated to the service they provide.
  2. The licensing scheme must be able to differentiate between well-prepared and poorly prepared candidates. Some licensure schemes do not screen candidates effectively. This is likely the case when the licensure is predicated on completing a training program or paying a fee rather than on proving one’s ability.
  3. Poor practice of the profession must predictably lead to harm that is serious and irreparable. In many cases, the potential harm done by a poorly prepared worker is either unserious or reparable. An inexperienced barber might give a bad haircut, or even abrade the scalp of an unsuspecting customer. The damage is real, but small and reparable. For serious but reversible harm, customers can seek redress through the justice system.
  4. Customers and co-workers must have substantial difficulty identifying a poorly prepared candidate. When job skills (or the lack thereof) important to safety are immediately obvious, there is no need of a license. A taxi driver or chauffeur, for example, has already proven that he knows how to drive by reaching the pickup location. By contrast, customers and even other professionals might not be able to identify an airplane pilot who is poorly prepared to handle rare emergencies. The heroism of U.S. Airways pilot Chesley Sullenburger, who landed his Airbus A320 in the Hudson River without a single fatality, was a testament to his diligence in training within the context of a rigorous licensure regime.

For each profession, policymakers should pursue the least-intrusive regulatory approach that is consistent with an acceptable level of public safety.

Salim Furth, PhD, is Research Fellow in Macroeconomics in the Center for Data Analysis, of the Institute for Economic Freedom and Opportunity, at The Heritage Foundation.

About the Author

Salim Furth, Ph.D. Research Fellow, Macroeconomics
Center for Data Analysis

Related Issues: Labor

Show references in this report

[1] I matched three separate extracts of the CPS for this Backgrounder: (1) United States Census Bureau, “CPS Certification Items Extract File 2015,” Current Population Survey, (accessed July 13, 2016); (2) National Bureau of Economic Research, “CPS Merged Outgoing Rotation Groups,” (accessed July 13, 2016); and (3) Center for Economic and Policy Research, “CPS Outgoing Rotation Group,” ceprDATA, (accessed July 13, 2016). To imitate the sample design of the CPS, I stratified the data by state and month and used household as primary sampling unit. For data and codes, please contact the author at 202-546-4400.

[2] Bureau of Labor Statistics, “Frequently Asked Questions About Data on Certifications and Licenses,” April 15, 2016, (accessed August 2, 2016). For simplicity, only the most common form of the third question is presented here.

[3] Bruce D. Meyer and Robert M. Goerge, “Errors in Survey Reporting and Imputation and Their Effects on Estimates of Food Stamp Participation,” Center for Economic Studies Working Paper No. 11-14, April 2011, (accessed August 2, 2016).

[4] Table 1 includes occupations in which at least 75 percent of the labor force reports a government-issued certification or license. Due to underreporting, true licensure prevalence in many of these occupations probably approaches 100 percent.

[5] The raw gap among those employed is 4 percent, but half of that can be accounted for by differences between respondents and non-respondents in occupation, age, and sex. More respondents than non-respondents, apparently, are in licensed occupations.

[6] The discrepancy is not due to the higher average age of whites. Among registered nurses ages 40 and above, the same pattern holds.

[7] The woman was the respondent in 47 percent of these households. These households make up 38 percent of all households and include 50 percent of the CPS population.

[8] This statistic is dated, but the broad pattern remains true. Robert H. Topel and Michael P. Ward, “Job Mobility and the Careers of Young Men,” National Bureau of Economic Research Working Paper No. 2649, July 1988, (accessed July 27, 2016).

[9] People in other categories are licensed at an intermediate rate.

[10] I included race and foreign-born status because they appear to affect response rates.

[11] I assign the national sampling error margin to each state’s predicted licensure rate. Since the comparison is over the same set of observations, the over- or underrepresentation of a particular occupation in a state will be present on both sides of the comparison, so the additional error introduced will be very small.

[12] For details on the Harris Poll, see Morris Kleiner, “Reforming Occupational Licensing Policies,” Hamilton Project Discussion Paper No. 2015-1, Chapter 7, March 2015, (accessed August 23, 2016).

[13] See, for example, Department of the Treasury Office of Economic Policy, the Council of Economic Advisers, and the Department of Labor, “Occupational Licensing: A Framework for Policymakers,” July 2015, (accessed August 23, 2016), and Salim Furth, “The Hidden Tax that Costs Households Up to $1,600 a Year,” The Daily Signal, April 15, 2016,

[14] To compute weighted regional averages from the published state-level data, I used Annual Average Employment by state from the Quarterly Census of Employment and Wages. My thanks to Morris Kleiner, Evgeny Vorotnikov, and Gray Kimbrough for pointing out the specific strengths and weaknesses of the Harris Poll.

[15] Sabrina R. Perkins, “Iowa Is the Most Recent State to Fight Hair Braiding Laws,” Essence, July 6, 2016, (accessed August 23, 2016).

[16] News release, “Kentucky Governor Signs Bill Untangling Hair Braiders from Unnecessary Regulations,” Institute for Justice, June 22, 2016, (accessed August 23, 2016).

[17] Natalie Johnson, “Thanks to New Law, Nebraska Hair Braiders Can’t Be Jailed for Practicing Without a License,” The Daily Signal, March 21, 2016,

[18] Dick M. Carpenter II, “Blooming Nonsense: Experiment Reveals Louisiana’s Florist Licensing Scheme as Unnecessary and Anti-Competitive,” Institute for Justice, March 2010, (accessed August 2, 2016).