Anyone who is paying attention to labor markets in the United States knows that there has been a dramatic reduction in private-sector unionization in the past 50 years. In 1973, 25% of private-sector workers were members of a union. By 2023, this rate had dropped to 6%. Many economists attribute this precipitous decline to what is termed “skill-biased technological change” (SBTC). SBTC refers to the fact that US economic growth has shifted the economy towards high-skill service sectors and away from areas like manufacturing that historically had high unionization rates. The argument is that SBTC will raise the value of worker skills, and workers therefore will find unionization to be less desirable since unions tend to compress the wage structure.
A recent paper of ours calls into question this established wisdom. By tracking the skill composition of occupations by union status over time, we produce new evidence which shows that, concurrent with the dramatic reduction in private-sector unionization, the types of jobs, and in particular the types of skills covered by private-sector unions have shifted to become more skilled rather than less skilled. These findings are directly at odds with SBTC-based models of unionization.
To conduct this analysis, we combine data from three sources. The first is information on worker occupation, demographic characteristics, union coverage, and wages from the Current Population Survey (CPS). The CPS is a nationally-representative survey of the US population that is conducted by the Bureau of Labor Statistics. It is the source of much of what we know about unions, since it has asked about whether workers were covered by a union or a member of a union since 1973.
The second two sources provide information on the skills required by each occupation. The Dictionary of Occupational Titles (DoT) and the Occupational Information Network (O*NET) datasets contain information from surveys of employers on a range of specific skills required for the occupations at their firm. For example, the surveys collect information on the extent to which occupations require mathematics ability, finger dexterity, and spatial orientation/coordination. Labor economists have increasingly used these data to develop “task” or “skill” based measures of the labor market, characterizing jobs by their skills requirements rather than by their industry or the occupation title.
We consider four types of worker skill that are derived directly from the O*NET and DoT survey responses: routine manual, routine cognitive, non-routine manual, and non-routine cognitive. Our focus on these skills is motivated by the historically strong union coverage among heavily routine and manual occupations combined with SBTC-based shifts in the US economy over the past half century toward occupations that require more cognitive and non-routine skills.
Photo by Aurelien Romain on Unsplash.
Our results show that the skill content of private-sector union jobs has changed in ways that mirror overall changes to the US economy. Unionized professions experienced a large increase in their coverage of non-routine cognitive skills and saw large reductions in coverage of routine cognitive and routine manual skills. Non-routine cognitive skill coverage changed little over time, however. Interestingly, skills coverage of union jobs has changed relative to non-union jobs, such that the skills required by unionized professions have grown relative to non-unionized professions. For example, non-routine cognitive skills were more prevalent in non- unionized jobs, but we find that over time 35% of the union-non-union gap in skills was closed. In addition, we find that routine cognitive and routine manual skills declined in the union sector, but they declined more in the non-union sector. These results show that unionized jobs became more skill-intensive relative to non-union jobs in three of the four categories of skill we consider.
We then examine whether these trends differ between men and women, since men and women on average choose very different occupations. The increase in non-routine cognitive skill coverage is evident for men and women, with larger increases for women. Coverage of this skill increased for both men and women relative to the non-unionized sector, while the relative increase in routine cognitive and routine manual skills is more concentrated among women.
Why have these changes occurred? We answer this question by decomposing the changes in skill content of unionized jobs into three parts: (1) changes in skill requirements within an occupation, (2) changes in worker concentration across existing occupations, and (3) changes to the occupational mix from both entry and exit over time. The first component measures the extent to which occupations themselves have changed, while the second two components measure whether workers are sorting differently across occupations either because of changes in occupation choice or because the occupations available to workers change over time.
For men, (2) and (3) explain most of the increase in non-routine cognitive skill intensity and decreased intensity of the other three skill types. The fact that the other three skill types have changed little is due to an offsetting effect from within-occupation changes to skill requirements. Among women, non-routine cognitive skill increases also are driven by (2) and (3), while all three components are relevant for explaining the decline in the other three skill types. These decompositions show that changes to the types of occupations workers select are responsive to changes to the skill content of unionized jobs over time.
Our findings are at odds with SBTC-based models of unionization. Broadly, those models predict that SBTC will increase the return to skill, which will make unionization less desirable because unions compress wages and thus will reduce the return to skill.
There are two additional implications of the SBTC-based model that we examine in our study. The first is that unions should reduce the return to skill over time relative to non-union jobs. We show that this is not what happened: the return to the types of skills we examine in union jobs relative to non-union jobs have increased or remained constant over the period of our analysis. This finding is important because it calls into question the idea that skill-based shifts to the US economy will reduce the desire of workers to unionize. The second implication is that the union wage premium, which is how much unions increase worker wages, should decline over time. Our paper joins with some other recent research to show that the union wage premium has held constant over time at around 20%. The evidence is clear that unionized workers continue to earn more than their non-union counterparts and that unions are able to compensate workers for the increasingly higher skill demands of their professions.
How can we square these empirical findings with the fact that the unionization rate has declined and that the economy has shifted towards higher-skilled jobs? We propose two changes to unionization models that help to reconcile theory and evidence. First, we account for the fact that collective bargaining is done by firm- and occupation-specific bargaining units that can be quite homogenous. This allows unions to negotiate for compensation that reflects the market-wide return to skills covered by their occupation(s). While unions may compress the wage structure within bargaining units, as long as they are negotiating based on the median skill level in the unit workers will continue to want to be unionized.
The second model change we propose is to account for changing costs of unionization and de- unionization. Using data from the National Labor Relations Board, we show that the costs of unionization likely rose considerably in the 1980s due to policies of the Reagan Administration that made it more difficult to organize and continue in many ways to today. This makes it harder for new occupations to unionize, as well as making it difficult to organize workers in occupations that previously were not unionized and are growing due to the SBTC-based shifts in the US economy. The types of occupations workers select has changed over time due to changes to the economy’s occupation mix and large shifts in the US industrial base. These new and growing occupations are less likely to be unionized, and workers face high costs in unionizing in these occupations. Thus, unionization has declined not because workers desire it less, but because economic growth has occurred in occupations that are not unionized, and workers face high costs of unionization.
Our research highlights that the reduction in private-sector unionization has been accompanied by a change in the type of worker who is unionized. Unions are playing a different role in the labor market today than they did 50 years ago because of the change in the skill composition of the workers covered. These changes are likely to intensify as the US economy continues to shift away from manufacturing and toward service-intensive sectors, “gig” occupations, and the knowledge economy. A recent ruling by the National Labor Relations Board makes it easier for “gig” workers to unionize, which could lead to another large change in the types of skills covered by unionized jobs. We expect that as the US economy continues to evolve, the skill content of union occupations will change along with the role unions play in the labor market. Our research shows that focusing on the skill content of unionized work is an important lens through which to understand these ongoing changes and their impacts on labor markets and workers.
To read the full research findings, which are currently available as a NBER working paper, click here.
