Suresh Naidu, a Stone Center Affiliated Scholar and a professor of economics and international and public affairs at Columbia University, presented some of his recent research on the history of U.S. labor unions at the 2024 Inequality by the Numbers workshop. In this interview, he discusses who is currently benefitting the most from the union premium, the data challenges in historical research on unions and how he overcame them, and the significant barriers U.S. workers still face in forming a union. The interview also draws on Naidu’s 2022 essay in the Journal of Economic Perspectives, “Is There Any Future for a U.S. Labor Movement?”
How has the union premium has changed over the last 80 years? Is it still the case that union membership benefits relatively less-educated workers the most, even though union workers overall are becoming more educated?
Naidu: The premium is pretty stable. A caveat is that for the long horizon of the 20th century, what we can really measure is a union household income premium, which is how much a household with a union member in it earns in income versus a household without a union member. That’s different from what we normally measure, which is an individual union wage premium, which is how much more an individual worker makes in wages relative to a non-union worker. But those two measures are pretty close for a lot of the 20th century, particularly in a period of single-worker households.
And what we found using Gallup data that goes back to the 1930s, together with a number of other sources, is that the overall union premium is pretty stable. It’s between 10 percent and 20 percent over the 20th century, and it’s always higher for less-educated workers. Those two patterns are constant over time: unions lift up the incomes of their members, and do more so for workers with less education.
We’re just finishing up another paper doing a similar exercise in Canada with similar Gallup data going back to the 1950s. And it’s the same pattern there, and it’s stable over the 20th century.
The flip side of this is that even as this pattern of wage effects has remained constant, de-unionization has been differentially affecting the lower-education segment of the labor market: exactly the people who get the highest return from being in unions are less likely to be in them over time. And that’s true in both Canada and the U.S.: de-unionization is concentrated among the relatively less educated. It used to be the case that the average union member had less education than the average non-union member, but in recent decades it’s the case that the average years of schooling of union members and non-union members is similar.
Have those two trends — the benefit of the union premium, and the concentration of de-unionization among the relatively less educated — continued in the last two decades?
Naidu: Yes, the premium has continued and the differential premium for less-educated workers has continued, and the de-unionization of the less-educated has also continued. The joke I used to make, before the pandemic, is that if the rate of union decline keeps going this way, the only people that will be in unions will be graduate Ph.D. students.
And that’s like what’s happened with some other working-class institutions of the 20th century, like social democratic parties — they have been somewhat taken over by the relatively educated. It was very interesting for us to see that this pattern, which has been well-documented in terms of voting patterns for left parties, is also apparent in union membership.
Are unions still helping to reduce inequality, if they’re losing people who would benefit from union membership the most?
Naidu: Unions do still have a small effect on reducing inequality. It’s very small. The traditional effect is that unions have raised wages at the bottom [of the income distribution]. That effect has fallen over time as union workers have become more skilled.
But unions also compress incomes in the middle: within a unionized sector, there’s less pay inequality. To make it concrete, think of public sector workers who are educated but often unionized. The pay dispersion among unionized public sector workers is a lot smaller than among workers with the same levels of education who not working in unionized jobs. Without unions, you have a lot more managerial discretion around pay scales, less wage standardization, and more variation in who gets promoted.
Your paper discusses the difficulty in studying the historical role of unions. What are the data limitations? And what are the best sources of data today?
Naidu: The standard tool for labor economists who study unions is the Current Population Survey [CPS], which only started asking workers about union status in 1973, and then only has detailed state information together with union status starting in 1977. Before that year, the CPS doesn’t tell us where union members are living.
In the 1980s and ’90s, labor economists studied the patterns of union wage premia and compression using the CPS data, but it was really hard to go back before the CPS. That was a limitation on the historical picture. Economists would often have lots of great regressions and graphs about the effect of unions on wages, but it was all coming from a period in which union were in freefall. And it’s exactly as unions are declining very quickly in the early ’80s that we start getting this good data on unions.
What our paper did was say, Well, the Current Population Survey wasn’t asking about unions until the 1970s, but Gallup pollsters were asking about unions all the way back to the 1940s. And the Gallup survey is done every few months — it’s only a thousand people, but once you start pooling all of those surveys together, you get really big samples. And it’s a household-level survey, because Gallup is interested in surveying voters, and so you get education, you get race, you get state, and you sometimes get income.
And we put together all of this Gallup data, and we augmented it with a few other surveys from the early period of surveys in the U.S. A survey I’m particularly proud of finding is the 1936 Bureau of Labor Statistics consumer expenditure survey, which economists have used: it’s a survey of households, of what they’re spending. But one of the spending lines they have in that survey is: Do you pay anything in union dues? And so then you can use that consumption survey to get another measure of union membership.
By pooling all of these surveys, we’re able to make not just a new estimate of how many union members there are in the whole country, but also what states they’re in, how much education they have, what race they are, and often their income. That’s what lets us calculate these patterns of selection and compression in the pre-1970s period.
Your paper also discusses the latent demand for unions among workers. Why is it so hard to unionize?
Naidu: You have to get through the eyes of many needles to get a union in the U.S. I would say Canada has an interestingly similar system. They both could be called Wagner Act systems, and basically workers need to solve a pretty severe collective action problem at the workplace level in order to get union recognition. In lots of other countries, you don’t. You can have a union with only a few workers, or the union contract automatically covers you even if you don’t join the union.
But in the U.S. system, you need 30 percent to sign a petition asking for a union election. And then you have six to eight weeks until there’s the actual election, and if an employer doesn’t want a union, they can mount a big anti-union campaign during this period, and it can be very intense and very effective.
And there’s an enormous amount of trench warfare — the amount of organizer time, the amount of effort, the fact if you lose, everyone gets demoralized. The process of getting a union is extremely costly and time-consuming, and you have to do it for every establishment, every bargaining unit at a time.
So just imagine you have an economy that’s very dynamic in which employers are going in and out of business all the time, particularly in low-wage sectors like restaurants and retail where there’s tons of turnover. Okay, you win a union, but then the business goes under and some new business comes in and takes its place. You would have to spend an enormous amount of resources on new organizing just to keep the level of density constant, let alone grow. These resources have to be paid by existing union members, so once union density falls below a threshold, it might be very hard to build it back.
That’s what keeps union membership and union coverage very low in these Wagner Act or establishment-based systems. In much of Europe, even though they have fairly low union membership — France, for example has lower union membership than the U.S. — they have collective bargaining agreements that are negotiated by employers’ associations and unions, and cover everyone. So even when a business goes under and a new business starts up, that new business is covered by the old contract. The collective bargaining agreement covers all of the employers in that sector.
In Germany, there’s a different system in which employers can opt out of the collective bargaining agreement. And there, what you’ve seen over the last 25 to 30 years is increasingly employers are opting out of the collective bargaining agreements. That’s one of the reasons that Germany in 2015 passed an aggressive minimum wage: the system of collective bargaining agreements was unraveling as employers were avoiding it.
Your paper noted that in 2022 about 70 percent of Americans said they approve of unions. And we did have that enthusiastic wave of unionization in 2021 and 2022. Where do you think we’re going from here?
Naidu: I think the public is broadly supportive of unions, but people are less certain about whether a union would be good for them, specifically — partly due to lack of information. Most Americans probably don’t know anyone in a union! Also, people will say yes to the question of “I would vote for a union if an election was held tomorrow.” But if you actually walk them through all the hard work of persuading their coworkers to sign a petition, having meetings, generating the workplace conflict, and actually doing the organizing, it’s not clear that they’ll continue to say yes. Particularly at the low end of the labor market, and particularly when faced with determined employer opposition.
So I don’t know if we’re laying enough groundwork. Without any changes in labor law, it’s very difficult to imagine you can get enough collective action and interest to win elections, then maintain that long enough to win good first contracts. You can have this enormous interest in unionization, but if unions aren’t able to turn this energy into higher wages and better working conditions, it will probably peter out.
We might see that with Starbucks. They’re negotiating, but it’s slow, and the company was stalling for a long time. While Workers United has won many store-level gains, all the momentum and energy hasn’t really shown up as a durable improvement in the lives of Starbucks workers yet. I suspect future growth is contingent on workers observing that if you do this work, you’ll get a union, you’ll get a collective bargaining agreement, and your life will be better. That tie, I think, still needs to be made. But Starbucks workers just won their 500th union election, so the energy is still there.
Do you have any policy recommendations for how to make that easier?
Naidu: There is an easy policy laundry list: labor law reform and tight labor markets. Pass the PRO Act, which is a laundry list of labor law changes that would make it a lot easier to organize. One of the things that led to the recent increase in unionization efforts was the tight labor market along with Covid, and people rethinking their relationship to work, seeing the possibilities in unions. So I think if you had three things — some really great positive examples of union wins, improvement in labor law, and continuing tight labor markets — I think that would probably push us in the direction of seeing an increase in union density or at least a much slower decline.
But at the end of the day policy can only do so much. One of the reasons I’m a fan of the labor movement is that it is a major vehicle for normal people to exercise democratic power over their conditions of work and pay. Workers vote to have a union, then vote for union leaders, vote over their contracts, vote over the decision to strike, etc. A powerful union isn’t under the control of policymakers. The flip side of this autonomy is that policies alone can’t build a real organic labor movement if the social fabric and networks that sustain collective action and capacity aren’t being built simultaneously. Unions aren’t just a policy to be turned on; they are independent organizations that rely on engagement and activism, for better or for worse.
Read More: