Data
In early 2013, the Institute for Justice (IJ), with funding from the John Templeton Foundation, employed Harris Interactive to conduct a state-level survey of all 50 states and the District of Columbia that yielded around 10,000 usable observations. We used those data to perform the first-ever analysis of occupational licensing at the state level. 1 The sample was the largest then available for within- and cross-state analyses, but it was still relatively small. In particular, the sample sizes within some states limited the ability to detect potential effects from licensing.
In this study, we build on our previous analysis by combining the Harris dataset with data from Wave 13 (gathered in late 2012) of the 2008 Survey of Income and Program Participation (SIPP) and analyzing the resulting larger dataset. In combining the two datasets, we did lose some detail: The Harris survey collected more detailed regulatory, income and labor skill data than SIPP. However, SIPP collected data for a much larger population. This tradeoff of less information for more observations was worthwhile because it allowed us to improve the precision of our state-level estimates and increase the statistical power of the tests. Altogether and after all the necessary data filters were applied, the combined dataset comprised 39,808 observations and is representative of the U.S. population at the state and national levels.
The Harris Data
For the Harris survey, IJ provided Harris Interactive with a draft of a questionnaire that was patterned after the Princeton Data Improvement Initiative, which was used in earlier studies of licensing. IJ and Harris collaborated in finalizing the questions’ order and wording. Several questions regarding the respondents’ employers, job activities and demographics were taken from the Current Population Survey. Harris staff pretested the survey with dozens of volunteer respondents from their regular representative sample of the United States.
Harris conducted the survey in early 2013. Individuals aged 18 or older who were in the labor force were eligible for the survey. We have limited our analysis to those who were employed at the time of the survey or who had held a job during the previous 12 months. 2
The SIPP Data
Data for Wave 13 of the 2008 SIPP were collected in 2012 and cover September through December 2012.The survey excludes individuals under 15 years of age and individuals living in institutions and military barracks. Similar to the Harris survey, it collects data about individuals’ licensing status, labor force activity, and demographic and social characteristics.
Combining the Harris and SIPP Data
We combined the Harris and SIPP data in three steps as follows:
Step 1 was to compare the questions the two surveys used to collect data about the licensing status of the population. The key questions in the Harris survey were:
A. “Do you have a license or certification that is required by a federal, state or local government agency to do your job?”
1. Yes
2. No
3. In process/Working on it
B. “Would someone who does not have a license or certificate be legally allowed to do your job?”
1. Yes
2. No
C. “Is everyone who does your job eventually required to have a license or certification by a federal, state or local government agency?”
1. Yes
2. No
The corresponding SIPP questions were very similar:
A. “Did you have a professional certification or state or industry license?”
1. Yes
2. No
3. Refused
4. Don’t know
5. Not answered
B. “Is this certification or license required for your current or most recent job?”
1. Yes
2. No
3. Refused
4. Don’t know
5. Not answered
6. Not applicable (Never worked)
These questions collected very similar information that allowed us to identify and differentiate between individuals who were licensed or certified. Having a dataset that allowed us to distinguish between licensed individuals and certified ones was crucial to ensuring precision of our estimates.
Step 2 was to apply data filters to the datasets to make them more comparable and then check whether both datasets would provide similar state-level estimates of licensing prevalence. Since the Harris and SIPP datasets had slightly different population distributions by demographic and social characteristics correlated with licensing prevalence (e.g., race, age, educational attainment and sector of employment), a simple comparison of state-level licensing prevalence was inappropriate. Instead, we used a logistic regression analysis to compare licensing prevalence across states. This approach allowed us to compare licensing levels between the datasets controlling for differences in the demographic and social variables’ distributions. The functional form of the regression is shown below:
Licensedi = βiHarrisi+ βiXi + ei
The variable Licensed is a dummy variable that indicates whether a person (“i”) is licensed. The dummy variable Harris indicates whether the data come from the Harris dataset or the SIPP dataset. The vector X is a set of individual-level control variables that includes individuals’ gender, race, age, union status, sector of employment and a two-digit Standard Occupational Classification (SOC) code. 3
A statistically insignificant gradient of the Harris variable would indicate that there is no difference in licensing prevalence between the Harris and SIPP datasets and that the existing observable differences in levels, if any, could be explained by differences in the distributions of the explanatory variables. The shortcoming of this approach is that a statistically significant gradient of the Harris variable would not necessarily indicate that there was a difference in licensing prevalence and could instead indicate that we detected some other unobserved differences between the two datasets.
We estimated one regression for each state. The Harris variable gradient was only significant at the 5 percent significance level in three states and at the 10 percent level in another four states. The similarity of the Harris survey and SIPP in both the data they collected and the licensing prevalence estimates they provided indicated the two datasets could be combined successfully.
Step 3 was to have Nielsen Holdings, which acquired Harris Interactive in 2014, reweight the combined dataset to make it representative of the population at the state level. Unless otherwise noted, all analyses were conducted with those weights applied.
The results of the combined dataset showing the percentages of workers licensed in each state and nationally are presented in Tables A1 and A2.
Table A1. State Percentages Licensed or Certified, With Ranks
| State | Licensed† | Rank | Certified†† | Rank |
|---|---|---|---|---|
| Alabama | 18.1% | 38 | 3.4% | 51 |
| Alaska | 18.4% | 34 | 7.2% | 12 |
| Arizona | 19.1% | 24 | 5.4% | 36 |
| Arkansas | 20.1% | 18 | 5.8% | 28 |
| California | 17.2% | 46 | 4.8% | 44 |
| Colorado | 17.6% | 44 | 5.4% | 34 |
| Connecticut | 21.5% | 10 | 6.7% | 17 |
| Delaware | 15.2% | 50 | 8.7% | 4 |
| District of Columbia | 18.9% | 28 | 4.5% | 45 |
| Florida | 21.1% | 14 | 4.4% | 46 |
| Georgia | 14.4% | 51 | 4.2% | 50 |
| Hawaii | 21.3% | 13 | 9.1% | 3 |
| Idaho | 23.6% | 4 | 5.7% | 30 |
| Illinois | 17.7% | 43 | 6.7% | 15 |
| Indiana | 17.9% | 40 | 6.5% | 18 |
| Iowa | 24.3% | 2 | 6.5% | 19 |
| Kansas | 16.0% | 49 | 7.3% | 10 |
| Kentucky | 19.4% | 22 | 5.4% | 35 |
| Louisiana | 22.4% | 7 | 6.2% | 23 |
| Maine | 24.2% | 3 | 5.6% | 31 |
| Maryland | 18.6% | 31 | 5.2% | 39 |
| Massachusetts | 17.8% | 42 | 4.9% | 43 |
| Michigan | 18.6% | 32 | 5.5% | 33 |
| Minnesota | 21.8% | 9 | 5.8% | 26 |
| Mississippi | 18.7% | 30 | 6.7% | 16 |
| Missouri | 21.0% | 15 | 8.1% | 6 |
| Montana | 19.2% | 23 | 7.0% | 13 |
| Nebraska | 18.2% | 36 | 4.2% | 47 |
| Nevada | 26.6% | 1 | 6.1% | 24 |
| New Hampshire | 16.0% | 48 | 7.2% | 11 |
| New Jersey | 19.6% | 21 | 5.7% | 29 |
| New Mexico | 18.4% | 35 | 6.4% | 20 |
| New York | 20.7% | 17 | 5.3% | 38 |
| North Carolina | 18.9% | 27 | 4.2% | 49 |
| North Dakota | 22.6% | 6 | 4.2% | 47 |
| Ohio | 18.1% | 37 | 6.4% | 20 |
| Oklahoma | 19.0% | 26 | 7.3% | 9 |
| Oregon | 19.8% | 20 | 5.8% | 27 |
| Pennsylvania | 19.1% | 25 | 5.6% | 32 |
| Rhode Island | 17.4% | 45 | 11.2% | 1 |
| South Carolina | 17.8% | 41 | 4.9% | 42 |
| South Dakota | 20.9% | 16 | 5.1% | 40 |
| Tennessee | 21.3% | 12 | 5.3% | 37 |
| Texas | 18.9% | 29 | 5.0% | 41 |
| Utah | 16.3% | 47 | 6.7% | 14 |
| Vermont | 18.5% | 33 | 7.8% | 7 |
| Virginia | 20.1% | 19 | 6.0% | 25 |
| Washington | 21.5% | 11 | 7.6% | 8 |
| West Virginia | 22.0% | 8 | 8.4% | 5 |
| Wisconsin | 18.0% | 39 | 6.3% | 22 |
| Wyoming | 22.8% | 5 | 9.3% | 2 |
† Average margin of error is 3.4% at 95% confidence.
|
||||
Table A2. Percentage of Workers Nationally Who Are Licensed, Certified or Neither
| Variable | % | S.D. |
|---|---|---|
| Licensed Workers | 19.09% | 0.213% |
| Certified Workers | 5.57% | 0.124% |
| Workers Neither Licensed Nor Certified | 75.34% | 0.234% |
| Total | 100.00% |
The demographic and economic characteristics of workers nationally who are licensed, certified or neither are presented in Table A3. They reveal that licensing rates increase with educational attainment: Nearly 39 percent of workers with post-college education have licenses compared to less than 6 percent of workers with less than a high school education. We also find that union members (36.5 percent) are more than twice as likely to be licensed as non-union members (16.8 percent). This finding no doubt reflects in part the large number of people working in occupations such as teacher and nurse that are frequently both licensed and unionized. Public-sector workers (31.7 percent) are also more likely to be licensed than private-sector workers (16.6 percent), a finding that likely carries a link to heavy unionization in the public sector. 4 Women (20.7 percent) are slightly more likely to be licensed than men (17.6 percent), and whites (20.9 percent) are more likely to be licensed than Hispanics (12.7 percent), blacks (16.6 percent) or “other” races (18.4 percent). Finally, we find that licensing rises with age before flattening over age 55. The similar state-level results are presented in the State Profiles.
Table A3. Characteristics of Workers Nationally Who Are Licensed, Certified or Neither
| Variable | Licensed | S.D. | Certified | S.D. | Neither Licensed Nor Certified | S.D. | Total % | Obs. | % Obs. |
|---|---|---|---|---|---|---|---|---|---|
| Gender | |||||||||
| Male | 17.6% | 38.1% | 5.5% | 22.9% | 76.9% | 42.2% | 100% | 18,941 | 48% |
| Female | 20.7% | 40.5% | 5.6% | 23.0% | 73.6% | 44.1% | 100% | 20,867 | 52% |
| Education Level | |||||||||
| Less than High School | 5.6% | 23.1% | 1.6% | 12.4% | 92.8% | 25.8% | 100% | 2,219 | 6% |
| High School | 10.2% | 30.3% | 3.1% | 17.4% | 86.6% | 34.0% | 100% | 9,031 | 23% |
| Some College | 18.2% | 38.6% | 6.5% | 24.7% | 75.3% | 43.1% | 100% | 13,902 | 35% |
| College | 22.1% | 41.5% | 6.1% | 23.9% | 71.8% | 45.0% | 100% | 9,382 | 24% |
| College+ | 38.8% | 48.7% | 8.1% | 27.2% | 53.2% | 49.9% | 100% | 5,274 | 13% |
| Earnings | |||||||||
| Average Hourly Earnings | $27.47 | $22.53 | $24.26 | $21.53 | $20.11 | $17.78 | – | – | |
| Race | |||||||||
| White | 20.9% | 40.7% | 6.1% | 23.9% | 73.0% | 44.4% | 100% | 28,463 | 72% |
| Hispanic | 12.7% | 33.3% | 3.9% | 19.4% | 83.4% | 37.2% | 100% | 4,361 | 11% |
| Black | 16.6% | 37.2% | 5.4% | 22.6% | 78.0% | 41.4% | 100% | 4,127 | 10% |
| Other | 18.4% | 38.8% | 4.3% | 20.3% | 77.3% | 41.9% | 100% | 2,857 | 7% |
| Age | |||||||||
| ≤25 | 8.4% | 27.8% | 3.7% | 18.9% | 87.9% | 32.6% | 100% | 5,522 | 14% |
| 26–54 | 20.4% | 40.3% | 5.8% | 23.4% | 73.8% | 44.0% | 100% | 25,180 | 63% |
| 55+ | 22.0% | 41.4% | 6.1% | 23.9% | 71.9% | 44.9% | 100% | 9,106 | 23% |
| Union Status | |||||||||
| Union | 36.5% | 48.2% | 5.6% | 23.1% | 57.8% | 49.4% | 100% | 4,501 | 11% |
| Non-Union | 16.8% | 37.4% | 5.6% | 22.9% | 77.6% | 41.7% | 100% | 35,307 | 89% |
| Sector of Employment | |||||||||
| Private | 16.6% | 37.2% | 5.5% | 22.7% | 77.9% | 41.5% | 100% | 33,006 | 83% |
| Public | 31.7% | 46.5% | 6.2% | 24.1% | 62.1% | 48.5% | 100% | 6,802 | 17% |
Note: The Obs. column shows the actual number of observations in the dataset. Percentages were calculated using those observations with weights applied. |
|||||||||
Analysis
Pre-Analysis Data Quality Screening
Before estimating the effect of licensing on licensed workers’ hourly earnings—that is, the economic returns from licensing or wage premium—at the national level, we probed whether licensing prevalence is correlated with other factors that might influence licensed workers’ earnings, thereby clouding the analysis.
As a check for the presence of regional patterns in occupational licensing, we used information on states’ geographical location and their percentage population of licensed workers to calculate the global Moran’s I statistic. This allowed us to check whether there were any clusters of states with statistically similar levels of licensed populations. The premise being tested, or null hypothesis, was that levels of licensing prevalence were randomly distributed. We used the permutation procedure to estimate the test’s pseudo-significance level. Using 9,999 permutations, we estimated the pseudo p-value to equal 0.46. This p-value did not allow us to reject the null hypothesis. 5 In other words, we found no indication of geographical clustering.
Licensing prevalence is not correlated with geographical location, but it could be correlated with other factors that could affect our results, such as occupational mix. We did not perform a check for this ourselves. However, the U.S. Department of the Treasury’s Office of Economic Policy, the Council of Economic Advisers and the Department of Labor did test for the presence of occupational mix patterns in licensing using the Harris survey estimates of licensing prevalence and data from SIPP. They found that “variation in licensing prevalence appears not to be driven by differences in occupational mix across States.” 6
The results of these checks for data quality issues suggest that the estimated models allow us to make statistically valid inferences about the effects of licensing on licensed workers’ hourly earnings.
Estimating the Economic Returns from Licensing
Tables A4 and A5 provide the results of our ordinary least squares regressions. The dependent variable in all of the regressions is a log of individual-level hourly earnings. The independent variables include a variable of interest—a Licensing dummy variable that is equal to 1 if a practitioner is licensed and to 0 otherwise—and other individual-level and state-level control variables. Some model specifications also include occupation fixed effects (based on SOC) and state fixed effects. In Table A5, we also add a Certification dummy control variable to the regressions. All reported standard errors were robust standard errors clustered at the state level. Tables A4 and A5 show the national-level effects on hourly earnings of, respectively, licensing alone and both licensing and certification. (Because the dependent variable was in logs, we make the appropriate adjustments in the text wherever we discuss the magnitude of the dummy variables’ economic impact. 7 Tables A4, A5 and A6 report unadjusted coefficients.) The estimates suggest that licensing is associated with average economic returns of 13.88 percent even after accounting for human capital, labor market characteristics and two-digit occupation controls. The influence of other variables such as age, education level, union status and race on hourly earnings is consistent with the economic and policy literature.
Table A4. National Estimates of the Influence of Licensing on Hourly Earnings (log)
| (1) | (2) | (3) | (4) | |||||
|---|---|---|---|---|---|---|---|---|
| Variables | Coefficients | S.E. | Coefficients | S.E. | Coefficients | S.E. | Coefficients | S.E. |
| Constant | 2.800*** | 0.018 | -3.709*** | 0.539 | -2.125*** | 0.084 | -0.822*** | 0.077 |
| Licensed | 0.310*** | 0.024 | 0.115*** | 0.009 | 0.118*** | 0.008 | 0.130*** | 0.007 |
| Female | -0.187*** | 0.005 | -0.188*** | 0.005 | -0.161*** | 0.005 | ||
| Hispanic | -0.098*** | 0.022 | -0.109*** | 0.017 | -0.083*** | 0.016 | ||
| Black | -0.109*** | 0.014 | -0.112*** | 0.012 | -0.089*** | 0.010 | ||
| Other | -0.039** | 0.018 | -0.062** | 0.024 | -0.059*** | 0.021 | ||
| Education | 0.090*** | 0.002 | 0.089*** | 0.002 | 0.064*** | 0.002 | ||
| Age | 0.051*** | 0.002 | 0.050*** | 0.002 | 0.042*** | 0.002 | ||
| Age2 | -0.0005*** | 0.000 | -0.0005*** | 0.000 | -0.0004*** | 0.000 | ||
| Union Member | 0.109*** | 0.010 | 0.098*** | 0.010 | 0.162*** | 0.010 | ||
| Public-Sector Worker | 0.024 | 0.015 | 0.025** | 0.015 | 0.045*** | 0.013 | ||
| Self-Employed | 0.240*** | 0.037 | 0.234*** | 0.037 | 0.219*** | 0.036 | ||
| Private-Sector Worker | 0.038*** | 0.010 | 0.040*** | 0.010 | 0.033*** | 0.009 | ||
| Children | 0.023*** | 0.007 | 0.020*** | 0.006 | 0.022*** | 0.005 | ||
| Divorced | 0.033*** | 0.009 | 0.040*** | 0.009 | 0.029*** | 0.009 | ||
| Married | 0.134*** | 0.007 | 0.140*** | 0.007 | 0.110*** | 0.007 | ||
| Log of Real GDP | 0.384*** | 0.050 | 0.234*** | 0.006 | 0.190*** | 0.006 | ||
| Occupation Fixed Effects | No | No | No | Yes | ||||
| State Fixed Effects | No | No | Yes | Yes | ||||
| R2 | 0.039 | 0.350 | 0.358 | 0.440 | ||||
| Observations | 39,808 | 39,808 | 39,808 | 39,808 | ||||
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.10.
|
||||||||
Table A5. National Estimates of the Influence of Licensing and Certification on Hourly Earnings (log)
| (1) | (2) | (3) | (4) | |||||
|---|---|---|---|---|---|---|---|---|
| Variables | Coefficients | S.E. | Coefficients | S.E. | Coefficients | S.E. | Coefficients | S.E. |
| Constant | 2.789*** | 0.018 | -3.708*** | 0.539 | -2.083*** | 0.089 | -0.791*** | 0.081 |
| Licensed | 0.321*** | 0.012 | 0.118*** | 0.008 | 0.121*** | 0.008 | 0.132*** | 0.007 |
| Certified | 0.170*** | 0.016 | 0.030** | 0.013 | 0.030** | 0.012 | 0.023** | 0.011 |
| Female | -0.187*** | 0.005 | -0.188*** | 0.005 | -0.161*** | 0.005 | ||
| Hispanic | -0.097*** | 0.022 | -0.109*** | 0.017 | -0.083*** | 0.015 | ||
| Black | -0.109*** | 0.014 | -0.112*** | 0.012 | -0.089*** | 0.010 | ||
| Other | -0.038** | 0.018 | -0.061** | 0.023 | -0.059*** | 0.021 | ||
| Education | 0.089*** | 0.002 | 0.088*** | 0.002 | 0.064*** | 0.002 | ||
| Age | 0.051*** | 0.002 | 0.050*** | 0.002 | 0.041*** | 0.002 | ||
| Age2 | -0.0005*** | 0.000 | -0.0005*** | 0.000 | -0.0004*** | 0.000 | ||
| Union Member | 0.109*** | 0.010 | 0.098*** | 0.010 | 0.162*** | 0.010 | ||
| Public-Sector Worker | 0.024 | 0.015 | 0.025* | 0.015 | 0.045*** | 0.013 | ||
| Self-Employed | 0.239*** | 0.037 | 0.233*** | 0.037 | 0.218*** | 0.036 | ||
| Private-Sector Worker | 0.039*** | 0.010 | 0.041*** | 0.010 | 0.034*** | 0.009 | ||
| Children | 0.024*** | 0.007 | 0.020*** | 0.006 | 0.022*** | 0.005 | ||
| Divorced | 0.032*** | 0.009 | 0.040*** | 0.009 | 0.029*** | 0.009 | ||
| Married | 0.134*** | 0.007 | 0.140*** | 0.007 | 0.110*** | 0.007 | ||
| Log of Real GDP | 0.384*** | 0.050 | 0.231*** | 0.007 | 0.187*** | 0.007 | ||
| Occupation Fixed Effects | No | No | No | Yes | ||||
| State Fixed Effects | No | No | Yes | Yes | ||||
| R2 | 0.043 | 0.350 | 0.359 | 0.440 | ||||
| Observations | 39,808 | 39,808 | 39,808 | 39,808 | ||||
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.10.
|
||||||||
We estimated human capital models similar to that shown in Table A4 (the models did not include any state-level controls) for all 50 states and the District of Columbia, finding a positive and statistically significant influence from licensing on licensed workers’ hourly earnings for 36 states. The unadjusted results of these regressions are shown in Table A6.
Table A6. State-Level Estimates of the Influence of Licensing on Hourly Earnings (log)
| State | Licensing Coefficient | S.E. | R2 | Observations |
|---|---|---|---|---|
| Alabama | 0.116** | 0.048 | 0.502 | 573 |
| Alaska | 0.113 | 0.083 | 0.503 | 246 |
| Arizona | 0.117** | 0.051 | 0.457 | 872 |
| Arkansas | 0.075 | 0.059 | 0.468 | 424 |
| California | 0.147*** | 0.028 | 0.477 | 3,074 |
| Colorado | 0.249*** | 0.066 | 0.421 | 619 |
| Connecticut | 0.241*** | 0.059 | 0.468 | 549 |
| Delaware | 0.207* | 0.123 | 0.564 | 249 |
| District of Columbia | -0.166 | 0.184 | 0.96 | 72 |
| Florida | 0.153*** | 0.036 | 0.385 | 1,565 |
| Georgia | 0.030 | 0.052 | 0.417 | 1,035 |
| Hawaii | 0.490** | 0.197 | 0.473 | 259 |
| Idaho | 0.138* | 0.075 | 0.451 | 344 |
| Illinois | 0.156*** | 0.039 | 0.451 | 1,529 |
| Indiana | 0.115*** | 0.038 | 0.437 | 1,287 |
| Iowa | 0.234*** | 0.049 | 0.457 | 573 |
| Kansas | 0.240*** | 0.080 | 0.408 | 437 |
| Kentucky | 0.036 | 0.071 | 0.424 | 524 |
| Louisiana | 0.043 | 0.054 | 0.455 | 616 |
| Maine | 0.165** | 0.068 | 0.51 | 306 |
| Maryland | 0.095* | 0.054 | 0.510 | 859 |
| Massachusetts | 0.199*** | 0.049 | 0.446 | 1,223 |
| Michigan | 0.194*** | 0.051 | 0.437 | 906 |
| Minnesota | 0.074 | 0.047 | 0.425 | 808 |
| Mississippi | 0.118* | 0.069 | 0.424 | 483 |
| Missouri | 0.129*** | 0.042 | 0.417 | 1,090 |
| Montana | 0.190** | 0.095 | 0.400 | 286 |
| Nebraska | 0.165* | 0.088 | 0.447 | 368 |
| Nevada | 0.206*** | 0.078 | 0.426 | 338 |
| New Hampshire | 0.147** | 0.067 | 0.544 | 361 |
| New Jersey | 0.192*** | 0.041 | 0.481 | 1,380 |
| New Mexico | 0.201*** | 0.072 | 0.482 | 363 |
| New York | 0.112*** | 0.038 | 0.426 | 1,701 |
| North Carolina | 0.107** | 0.052 | 0.455 | 993 |
| North Dakota | 0.101 | 0.094 | 0.620 | 103 |
| Ohio | 0.135*** | 0.040 | 0.430 | 1,264 |
| Oklahoma | 0.021 | 0.069 | 0.389 | 522 |
| Oregon | 0.100 | 0.069 | 0.462 | 542 |
| Pennsylvania | 0.151*** | 0.042 | 0.454 | 1,340 |
| Rhode Island | 0.159* | 0.081 | 0.408 | 286 |
| South Carolina | 0.098* | 0.059 | 0.461 | 552 |
| South Dakota | 0.098 | 0.089 | 0.461 | 262 |
| Tennessee | 0.148*** | 0.051 | 0.467 | 834 |
| Texas | 0.131*** | 0.027 | 0.462 | 2,574 |
| Utah | 0.177*** | 0.067 | 0.436 | 454 |
| Vermont | 0.152 | 0.095 | 0.369 | 257 |
| Virginia | 0.123*** | 0.039 | 0.484 | 1,430 |
| Washington | 0.042 | 0.038 | 0.475 | 1,177 |
| West Virginia | 0.077 | 0.078 | 0.352 | 388 |
| Wisconsin | 0.138*** | 0.045 | 0.456 | 1,249 |
| Wyoming | 0.041 | 0.096 | 0.354 | 262 |
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.10.
|
||||