When it comes to stagnant wage trends, I yield to no one (except maybe the Economic Policy Institute’s Larry Mishel) in my efforts to elevate the issue and tie it to deep-seeded structural changes that have been zapping worker bargaining power for decades. I’ve tried to be particularly vigilant in ringing this lack-of-real-wage-growth alarm bell in recent months, as the tightening job market has led to threatening chatter about the need for the Federal Reserve to ratchet up rates sooner than later.
So when I tell you I’m a little surprised to see almost no movement in wage growth despite the improving employment situation, I hope you’ll give me a listen. To be clear, that’s “a little surprised.” There’s still considerable slack in the job market, and, like I said, workers’ ability to bargain for a bigger slice of the pie has taken a real beating over the years.
But given the extent to which the job market has tightened up in recent months, I would expect a bit more wage pressure than I’ve seen (“tightening,” “improving,” “less slack” are all econo-mese for stronger labor demand leading to faster job growth and lower unemployment). So let’s look at the evidence for these claims and think about why the wage dog is not barking. While I offer a number of credible hypotheses, the one I favor is pretty straightforward: Raising pay is simply not part of the business model of American employers. They will not do so until they’re absolutely forced to by a labor market that’s much tighter than what we’ve got today.
Exhibit A in the absence of wage pressures is below. The two wage measures are year-over-year percentage changes in nominal average hourly wages and hourly wages of lower-paid, non-supervisory workers (“non-sup” in the figure). The slack line is a comprehensive measure derived by economist Andrew Levin that combines numerous dimensions of labor market slack, including involuntary part-time work and folks who’ve dropped out of the labor force but would come back in if the jobs were there.
As slack has come down, the “non-sup” wage has gained a little traction, but the average wage has been stuck at 2 percent, just about the rate of inflation and clearly unresponsive to tightening.
A more formal statistical test of this relationship is shown in the next figure. See the data note for details, but it basically tracks the negative correlation between wage growth and job market slack over the current economic recovery, which began in mid-2009. As you see, that correlation has become smaller (in absolute value) and increasingly flat.
That’s the what. Now for the “why.”
A phenomenon called “downward nominal wage rigidity” could be in play here. Employers would actually have liked to reduce workers’ pay during the recession/weak recovery, but they couldn’t because outright pay cuts are actually pretty rare. Even for the demoralized American worker, that’s a step too far.
So instead of lowering the nominal wage, they let inflation erode the real wage. But the fact that inflation’s been so low for years has blocked that path. So now that things are tightening up, employers are happy to leave wages where they are for a while as their lower — but unattainable — target wage catches up to the real wage.
Let me make this confusing scenario concrete. Your employer is paying you $20 an hour. Because of weak demand, she’d like to pay you $18 an hour, but she can’t because it would piss you off too much to get a smaller paycheck (it’s not your fault the banks blew up the economy!). Now that demand is coming back, she’d like to pay you $19, then $19.50, and finally $20. But, of course, you’ve been at $20 the whole time, so that’s why you’re still there.
That dynamic may be in play, but note that it’s purely an employer’s-got-the-upper-hand story. It assumes all the bargaining chips are in her hands, none in yours.
And that’s what I think is really going on here. Worker bargaining power may be so diminished in this country that it’s become what you may recall from Algebra II as a “non-continuous function.” That is, wage growth doesn’t continuously accelerate as slack is diminished. It plods along until the job market is at truly full employment, which is the only thing that we can count on to budge wage growth from its entrenched moorings.
That’s just a hypothesis, but here are some things you’d expect to see if I’m right:
— Employers complaining about not finding the workers they need but not raising pay to get them. Check.
— High levels of income inequality. with most of the growth flowing into profits, not wages. Check.
— Less movement in the job market by workers seeking to upgrade to a better job. Check.
— A flattening of the slack/wage-growth curve, as shown above. Check.
The U.S. business model has devolved to a point where raising pay is antithetical to sound practice. If you’re a successful employer, it’s the very last thing you do, and you do it only if you’re pushed to the wall by such a tight labor market that you’ll literally lose workers and, thus, profits if you don’t.
That, in turn, has two implications. First, policymakers seeking to raise workers’ pay need to be mindful of this dynamic and push for chock-full employment and other countervailing measures, like a higher minimum wage. Second, the Federal Reserve must factor severely diminished worker bargaining power into its calculus, a factor that militates against preemptive tightening.
Longer term, workers need a lot more bargaining clout. But that won’t happen until there’s a politics that reacts to all this technical analysis with the urgency it deserves. I know, the working class isn’t the donor class. But they are, or at least they should be, the voter class, and last I checked, at least on paper, this is still a democracy.
Data note for figure 2: The figure plots the coefficient from a Phillips wage curve rolling regression. The model regresses the wage gap — the difference between actual and long-term wage growth of the non-supervisory series — on the Levin slack variable and six lags of the dependent variable. Long-term wage growth is derived using an HP filter applied to the actual series. I run the model using monthly data (year-over-year percent changes on the wage variable) from January 1990 to June 2009 and then add one month at a time to the sample, pulling off the gap coefficient in each iteration.