Correlation does not prove causation. But if it did, the Center for American Progress (CAP) would have proved unions reduce wages. Of course, correlations are flimsy evidence, so the study proves little. But the episode illustrates why it’s hard to blame America’s economic problems on organized labor’s decline.
CAP’s new report piggybacks off a recent Pew study finding the middle class has shrunk over the past generation. The CAP authors argue unions help workers land middle-class jobs. They theorize that falling union membership has thus contributed to the middle-class decline. So they performed a “shift-share” analysis and concluded that the decline in unions explains half of the drop in middle-class membership. That sounds quite troubling.
However, more than three-quarters of the contraction of the middle class occurred because Americans moved into higher-paying jobs. Pew and CAP both found the middle class has shrunk primarily because Americans have gotten richer. Taken at face value, CAPs results show that the decline in unions helped Americans get ahead. The authors did not highlight that conclusion, but their results unequivocally show it.
Why Union Members Come from the Middle
Nonetheless, conservatives won’t (or at least shouldn’t) be touting this report. The CAP study shows nothing about how unions affect pay. It looked only at correlations. The CAP authors note that union members typically belong to the middle class. They assumed union membership causes middle-class earnings and went from there. But causation runs in the opposite direction too: being in the middle class makes workers more likely to fill a union job.
Unionizing changes how firms hire and which employees want to work for them. Union work rules make firing poor performers very difficult. So unionized firms hire very selectively. Knowing they cannot get rid of unproductive workers, they go to great lengths to avoid hiring them in the first place. At the same time, union seniority systems repel top performers. The union contract limits them to a seniority-based raise, no matter how hard they work. When firms unionize, many of their top employees take their skills elsewhere.
Consequently, union members disproportionately come from the middle of the productivity and income distribution. CAP’s finding that union members disproportionately come from the middle class is completely unsurprising. Less-skilled workers can’t get union jobs; more skilled workers don’t want them.
The CAP authors made no effort to disentangle this correlation from a causal effect of unions on wages. Their analysis shows nothing more than: “Over the past 30 years, higher earnings lifted many Americans out of the middle class. At the same time union membership declined.” Perhaps the two are related, but CAP’s report does not attempt to prove it.
Unions Don’t Raise Wages
This type of thinking pervades far too much analysis of how unions affect workers. Private-sector union density has declined steadily since the 1950s. Many analysts will point that out, point to another economic trend, and conclude, “This wouldn’t have happened with stronger unions.” President Obama himself recently made similar arguments.
Looking beyond correlations to causation suggests unions deserve little credit or blame for recent wage changes. Brigham Frandsen, a Massachusetts Institute of Technology-trained economist, recently examined every company whose workers voted in a National Labor Relations Board union election since 1980. He combined this election data with income and employment data from tax records. In total, his data covers 1.7 million private-sector workers who voted on unionizing. This data allowed Frandsen to examine causation, not just correlation. He could compare pay at companies that unionized to similar companies that did not.
Frandsen expected to find unionizing raised average pay. Instead he found the opposite. Average wages fall about 3 percent at newly unionized companies. Looking deeper, Frandsen found this happened because top performers often left unionized firms. That dragged down average wages. Looking only at workers who stick with their firm, Frandsen found unionizing has no effect on pay. On average, private-sector unions do not cause their members’ pay to rise or fall.
How is that possible? Unions certainly want higher pay. But unionized firms compete against non-union businesses. When unions demand above-market wages, employers must either raise prices or reduce quality. That drives away customers, as the Detroit automakers famously demonstrated. Unions understand this and prefer not to eliminate their members’ jobs. So they aim for contracts that keep unionized firms competitive. That generally means paying market wages.
CAP’s report found shrinking union membership has raised wages. Other reports claim the opposite. But these studies only examine correlations. Looking further to causation shows that private-sector unions have little control over pay. Love them or hate them, unions have little effect on the middle class.
