The union wage premium: Difficult to calculate, likely overblown

What is the union wage premium?

The “union wage premium” — the amount a union worker makes in wages and salary above a nonunion worker — is often used to highlight one of the values of joining a union. Workers are often told that if they join a union their wages will increase because the average union worker makes more than the average nonunion worker. If this were unilaterally true, it might be a compelling argument for enrolling in a union. But the decline in union membership rates over the last several decades shows that an increasing number of workers have not been persuaded to join or organize unions. This suggests that workers are not convinced that unionizing will automatically boost their pay.

Unions still maintain, however, that their members earn significantly more, on average, than nonunion workers. On a webpage titled “The Union Difference,” the AFL-CIO mentions the union wage premium and says “union workers’ wages are 27 percent higher than their nonunion counterparts.” [1] The U.S. Secretary of Labor, Thomas Perez, also claims that there is a 27 percent union wage premium. On a morning talk show in 2014, he said that union workers make $950 per week while nonunion workers make only $750 per week. [2]

So, where does the truth lie? Is comparing the average weekly earnings of union and nonunion workers, as the AFL-CIO and Secretary Perez did, give a true apples-to-apples comparison of union and nonunion wages? Or are there statistical controls that need to be made to make sure the comparison is valid? Is the union wage premium the same across all sectors of the U.S. economy and has it changed over time? Will all workers benefit if they choose to unionize? Most importantly, if workers really can earn almost 30 percent more by simply joining a union, why has the unionization rate steadily gone down rather than up over the last several decades?

Calculating the Union Wage Premium

The problem with the 27 percent figure the AFL-CIO and others cite is that it does not take into account inherent differences between union and nonunion workers. For instance, union workers are more likely to work in occupations that already pay a higher wage on average. Also, unionized workers tend to be located in certain regions in the U.S. where the cost of living and wages naturally tend to be higher — in the Northeast, for example.

A better analysis of the union wage premium will take these factors into account, and previously published academic research has done just that. In a 2002 study, economists Barry Hirsch and Edward Schumacher controlled for the factors that affect a person’s wage, such as age, education, experience, industry and geographical region. They estimated that the average union wage premium for private sector workers was 20 percent in 2001, down from 26 percent in 1984. [3] David Blanchflower and Alex Bryson used a similar statistical model in their 2003 study. They calculated the union wage premium for private sector workers to be just 16.5 percent in 2002. [4]

What about a more up-to-date estimate of the union wage premium? To answer that question, I used data from the U.S. Census Bureau’s Current Population Survey and employed a regression analysis similar to the ones used in the aforementioned studies. These regressions analyses allow a comparison of the wages of two groups of workers who are similar in every way except one: one worker belongs to a union and the other does not. The results suggest that the union wage premium has continued to decline since 2002. According to this CPS data, the premium was just 10.2 percent in 2014, an average for both private and public sectors.

The Problem with Government Data

But that’s not the end of the story because biases built into the CPS data understate this figure. In short, the Census Bureau imputes the wages for some survey respondents without considering whether these workers are unionized. So even if these workers are unionized, their wages are categorized as nonunion wages by the Census Bureau.

Fortunately, the records for survey respondents with imputed wages can be removed from the data set. Though this reduces the sample size, it provides a more accurate picture of reality. When this correction is made, the CPS data show the union wage premium in 2014 is 14.7 percent on average, almost half of what it is routinely touted by unions.

This 14.7 percent figure is an average across all sectors of the U.S. economy, and a more detailed look at the CPS data suggest that there are important departures from this across-the-board estimate. For instance, in two entire sectors — nondurable goods manufacturing and wholesale trade — the union wage premium is zero, or, technically, not statistically different than zero. This means that union and nonunion workers earn about the same in these sectors, which add up to about 8 percent of the U.S. economy. And in two other sectors — durable goods manufacturing and retail trade — the union wage premium is below 10 percent. Meanwhile, the largest union wage premiums are in the construction, transportation and warehousing, and education and health care sectors.

While regression analyses of wage data from the CPS continue to show a union wage premium, it is worth noting that it has steadily declined over time. This is only possible if nonunion wages are growing faster (or decreasing less) than union wages. And this is, in fact, the case in nearly every sector of the economy that has unionized employees.

For instance, the inflation-adjusted nonunion weekly wage in construction increased by 12 percent from 1985 to 2014, but wages of unionized employees only increased by 1 percent. Similar trends were found in nondurable goods manufacturing, durable goods manufacturing, wholesale trade, retail trade and transportation and warehousing. In fact, in wholesale trade and retail trade, the real average weekly union wage decreased from 1985 to 2014 (by 9 and 14 percent, respectively). But the weekly wage for nonunion employees in those sectors rose significantly (18 and 35 percent, respectively).

Young workers considering whether to join a unionized workplace should consider these trends. Though nonunion workers may not start out at the pay level as their unionized counterparts, they might enjoy faster wage growth over time. They may also end up, later in their careers, at a higher pay level than they would have enjoyed in a unionized shop.

The Firm-Size Bias

These more nuanced views of official government data, however, still may not be telling the whole story. For instance, there may be other inherent biases in the data that researchers cannot control for. For example, it could be that unions simply tend to organize workers in firms that tend to pay higher-than-normal wages. And there’s some evidence to this effect. Unions are more likely to organize larger firms, since they can collect more in dues there, and larger firms are more likely to pay higher wages. So workers in these larger firms would have already been paid more than their counterparts in smaller firms, even if they weren’t unionized.

Some researchers have recognized this firm-size bias. To get around that, they have attempted to compare workers’ wages at firms that recently and narrowly voted to unionize with workers’ wages at firms that narrowly voted not to unionize. By tracking the firms and workers over time and comparing their histories, they are able to mitigate the firm-size bias. The findings of these studies suggest that a newly formed union does not raise wages for workers compared to their nonunion counterparts.

One recent study that uses this technique was conducted by economist Brigham Frandsen in 2013. Frandsen used a regression discontinuity research design and matched individual worker earnings to employers, based on the results of close union elections. By comparing the wages and employment security of workers who had just voted to unionize and those who had just narrowly voted not to unionize, Frandsen was able to isolate (as best as possible) the impact that unionization had on real, individual workers. As the title of his paper states, the findings were “surprising.” [5]

Frandsen found that recently unionized firms actually reduced payroll, paid lower wages on average, hired fewer workers and were more likely to go out of business than firms that almost but did not unionize. The reduction in payroll and average lower wages at recently unionized firms was primarily caused by the more experienced and higher paid employees leaving the firms after unionization. But even workers who remain employed by newly unionized firms “are little affected on average,” according to Frandsen’s research. [6]

If the older employees who left these newly unionized firms were relatively well paid because of the value they generated for the firm, it stands to reason that the firms that pushed these workers out through unionization also were more likely to go out of business. This dynamic creates a conundrum for workers who see unionization as a means to better wages and working conditions, because if they are successful in unionizing, they are more likely to have no wages, due to the firm closing. Presumably, for most employees, a nonunion job is better than no job at all.

Frandsen’s research is probably the best look we have of the real impacts of unionization on individual workers. All other efforts, many of them outlined above, must rely on aggregated government data. These data, no matter how reliable, have inherent shortcomings and biases that researchers can only attempt to mitigate through statistical controls. Frandsen tracks the conditions of real workers and firms over time and isolates the unionization variable as best as possible. His findings help provide a better overall picture of the effects of unionization and should be of particular interest to employees who are considering organizing a new union in a workplace.

Conclusion

Calculating the union wage premium is not an easy task, and as demonstrated, it is a surprisingly complicated issue. The various approaches researchers have taken have produced a variety of results, all with their strengths and weaknesses. According to imperfect government data, unionized workers do earn more on average than nonunion workers, but the difference is much smaller than the number produced by the simplistic approach some unions tout. In short, there are many factors that may explain why some workers make a different wage than others.

Finally, it’s important to remember that correlation is not causation. If current unionized employees earn more on average than nonunion workers, this does not mean that workers will automatically give themselves a wage boost by unionizing. Nonunion wages are rising faster than union wages, on average, and newly unionized firms tend to do worse, suggesting that it might not be in the best interest of every worker, from a wage perspective, to unionize.

1 “The Union Diference” (AFL-CIO, 2016), https://perma.cc/RLZ5-UVT7.

2 Louis Jacobson, “Labor Secretary Thomas Perez Says Union Members Earn $200 More a Week than Those Not in Unions” (PolitiFact, Feb. 7, 2014), https://perma.cc/QG69-DA6N.

3 Barry T. Hirsch and Edward J. Schumacher, “Match Bias in Wage Gap Estimates Due to Earnings Imputation,” Journal of Labor Economics 22, no. 3 (2004): 689–722, https://perma.cc/T83Y-FQKD.

4 David G Blanchflower and Alex Bryson, “What Effect Do Unions Have on Wages Now and Would Freeman and Medoff Be Surprised?,” Journal of Labor Research XXV, no. 3 (2004): 383–414, https://perma.cc/9CJ2-PX8E.

5 Brigham R. Frandsen, “The Surprising Impacts of Unionization: Evidence From Matched Employer-Employee Data” (Brigham Young University, Dec. 24, 2014), https://perma.cc/F78H-NU6F.

6 Brigham R. Frandsen, “The Surprising Impacts of Unionization: Evidence From Matched Employer-Employee Data” (Brigham Young University, Dec. 24, 2014), https://perma.cc/F78H-NU6F.