It’s Labor Day, so let’s talk spillovers.

I’m not referring to somebody knocking over your beer at the picnic later this afternoon. I’m talking about an economic phenomenon in which a policy or intervention that directly affects people in one part of the economy spills over and affects others. A common example, sticking with Labor Day themes, is the minimum wage, which directly lifts the wages of the lowest-wage workers but has also been found to spill over and give a wage bump to workers who earn above the minimum.

Now, think about why real wages have stagnated for so many working Americans. Most people tend to blame globalization, which has tilted against blue-collar production workers, and the shift from manufacturing to lower-paying services; technological changes that have favored those with skills in IT and apps; and/or slack labor markets that have zapped the bargaining power of those who depend on strong labor demand to get a fair slice of the pie.

Those are all legitimate reasons, but a new paper by Jake Rosenfeld, Patrick Denice and Jennifer Laird, published by the Economic Policy Institute (EPI), argues that there’s another big factor in play: negative spillovers to non-union workers from the loss of union jobs. The paper’s logic is compelling and its evidence rigorous. I won’t vouch for the exact numbers, as many moving parts are in play, some of which are hard to control for in this sort of study. But I’m sure they’re in the ballpark.

Though we’ve never come near European levels of union coverage, there was a time, in the 1950s, when about a third of the U.S. workforce was covered. Today, union membership is down to about 11 percent (that’s a weighted average of 7 percent in the private sector and 35 percent in the public sector). That decline has meant a direct loss of bargaining clout for unionized workers, of course, but according to Rosenfeld et al, it has also been a consequential factor in the wage decline of non-union workers.

When union coverage is relatively high, “unions set pay and benefits standards that nonunion employers follow.” Such employers may be motivated by trying to avoid an organizing drive in their own workplaces, or they simply might have to meet the wage set by the union to avoid losing workers. In their heyday, in industries where they were dominant, such as autos and steel, unions roughly set the pay for members and non-members alike.

To gauge the magnitude of this effect, the authors build a statistical model to predict wages of non-union workers, with the key variable being union density for those workers’ industries and parts of the country. Thus, if a non-union trucker in Georgia got a wage bump from the contracts negotiated by trucking unions in the state, their model would register that impact. They then ask: suppose, holding other changes constant, we plug in the 1979 union membership rate for Georgia truckers today. How much would that help today’s non-union truckers in the region?

Their answer (for the nation) is a 5 percent increase for 40 million non-union men, meaning an additional $2,700 in earnings for year-round workers. For non-college-educated men, the loss in union coverage was greater, so the simulated impact of recovering that loss is greater as well: 8 percent, or $3,000 per year. Union coverage was lower and fell less for women, so their gains are smaller, about 2 percent to 3 percent.

These estimates could be biased up, as much more has changed than the loss of union coverage since the late 1970s (the authors themselves raise these caveats). Globalization has put downward pressure on the bargaining clout of today’s union workers, as have the shifts to lower value-added industries, lower economy-wide productivity, and persistent labor market slack. But even the authors’ low-range estimates come to about $2,000 for non-college-educated men.

So while we can quibble about point estimates, I’m sure the thrust of their argument is sound. What troubles me is the plausibility of tapping this source to help reverse wage stagnation. What are the odds of what the authors call a “unionization revival?”

The blue line in the figure below shows the long-term trend in union coverage in the United States (I’ll get to the other line in a moment). Contemplating the long slide, one could conclude that a reversal is highly unlikely; we’d be lucky to stop the slide.

But it’s not that simple. Damon Silvers, director of policy for the AFL-CIO, notes that the extent of union membership is partly a function of big, lasting shocks to the system, both positive and negative. The 1936 Flint sit-down strike against General Motors represented a turning point in the legitimacy and growth of the United Auto Workers, for example. Conversely, President Ronald Reagan’s firing of 11,000 air-traffic controllers whose union, PATCO, was on strike signaled a level of unbridled conservative opposition that persists to this day.

What could precipitate a positive shock that could lastingly reverse the negative trend in union density? Silvers and other to whom I posed this question had the same answer: centralized bargaining. Especially given the depth of opposition and the existence of the “gig economy” (where the “workplace” hardly exists), organizing one establishment at a time is a recipe for further stagnation. As EPI’s president, Larry Mishel, put it, “We need a design where people have collective bargaining rights as restaurant workers, as opposed to one where they gain those rights one restaurant at a time.”

There’s historical precedent: Sectors such as autos, trucking, airlines and steel once used pattern bargaining to set compensation norms for their industries, and such pattern bargaining will seriously amp up spillovers of the type that Rosenfeld et al squeeze out of data. The other line in the above figure is the share of income going to the top 10 percent, showing the solid correlation between falling unionization rates and higher income inequality.

Correlation is not causation, and all the moving parts noted above are also in play, as the authors themselves note. But in another type of far-reaching spillover, a strong union movement has the potential to progressively impact each one of those other determinants. The expansion of international trade, for example, plays out in a way that’s much more beneficial to production workers in Germany, where half the workforce is covered by union contracts, than it does here. And even in its diminished state today, there is no single, larger institutional force pushing for progressive economic outcomes in our politics than the labor movement.

Wouldn’t it require a change in labor law to facilitate this more sweeping approach to union coverage, and isn’t that implausible? Yes and no. The way to think about this is not “How do we get this Congress to support progressive labor law reform?” Won’t happen.

Instead, the right question is: “How can we engender the regime change that will fundamentally change the political power dynamics that preclude needed reforms?”

That sounds very abstract, but let me try to make it more concrete. When I was a member of President Obama’s economics team in his first term, when Democrats were the majority in both the Senate and House, there was pressure on us to try to pass the pro-union Employee Free Choice Act. Both Obama and Vice President Biden recognized the extent to which the unions that helped their campaign were willing partners. But there was never a filibuster-proof majority in the Senate, and it wasn’t just Republicans who blocked the bill. The legislative problem unions face today is that the Republicans hate them more than the Democrats love them.

In other words, power-shifting regime change won’t happen until bottom-up campaigns enforce a new, truly progressive politics. I view groups like the Fight for $15, the Black Lives Matter movement and the Fed Up coalition in this light, but if we are to get the change we need, groups like these must persist, proliferate, and garner support not just from the left, but from the center-left as well.

That’s not going to happen next week, next year, or by next Labor Day. It may not happen at all, and the two lines in the figure above may just continue to diverge. But given the impact of that divergence on our current politics and the deep, social divisiveness that is so painfully evident in today’s America, we should do everything we can to avoid that fate.