One of the well-known benefits of tight labor markets is that they force reluctant employers to share more of their profits with workers. Particularly in the U.S. labor market, where union power is historically low, people who depend on their paychecks, as opposed to their stock portfolios, also depend on very low unemployment to boost their bargaining clout.

Well, we’ve got low unemployment, but we’re not seeing what we’d expect in terms of faster wage growth. The figure below shows year-over-year nominal (before factoring in inflation) wage growth of non-managers and blue-collar workers, the 80 percent of the private workforce that is particularly dependent on the bargaining-power dynamics described above (the six-month moving average smooths out the series a bit). The tighter job market lifted the pace of their nominal wage growth, from about 1.5 to 2.5 percent, but since early 2016, it has stalled out at about 2.5 percent.

Why is that? Here are some explanations, along with my evaluation of their merit.

There’s still slack in the labor market. No question, that’s true. Contrary to many market analysts, I don’t think we’re quite yet at full employment (certainly, weak inflation data support that conclusion). But there’s also no question that the job market is increasingly tightening, so this explanation only gets you so far. The figure below, though noisy, is interesting and instructive. Eight times per year, the regional Federal Reserve banks collect qualitative assessment of local labor markets. Goldman Sachs researchers plot those assessments against wage growth (“GS Wage Tracker”). By this metric, the job market is about as tight as it was in earlier recoveries (a bit less so than the 1990s), though wage growth is notably flatter.

By the way, there’s another important point to note from the above figure. I follow many wage series, and some show more growth than others. Some adjust for changes in the types of people in the workforce; some follow the same workers over time. The GS series plotted above combines a variety of these different series and still comes up flat of late.

Explanatory power: Medium

Slow productivity growth. This one’s important. We got great wage growth in the latter 1990s, across the wage scale (i.e., not just at the top), because of the potent combination of full employment and fast growth of output-per-hour. Intuitively, because their production efficiency was rising at a good clip, firms could maintain profit margins while doling out higher pay and keeping a lid on prices. Today, pricing power remains low, but with low productivity growth, firms trying to maintain profit margins are resisting higher pay.

The figure below shows a model of wage growth (see data note for details) that tracks the series’ nominal growth pretty well. Note the circled bit at the end. The blue line is a prediction that leaves out productivity growth, and it suggests that wages of blue-collar, nonmanagerial workers should now be growing at 3 percent, not 2.5 percent. But when I add productivity growth to the model (red line), the prediction downshifts to about its current growth path. No less than Fed Chair Janet L. Yellen presents a similar result in a recent speech (“… the model estimates that the increasing upward pressure on compensation growth from rising labor utilization is being offset by a declining contribution from productivity growth”).

One more point. I’ve argued, with some evidence (though, admittedly, not a ton), that full employment and faster productivity are causally linked through the dynamics discussed above. Elevated worker bargaining clout forces firms to find production efficiencies to maintain profit margins, something they don’t have to do in a slack labor market.

Explanatory power: High

Other factors in play: Low inflation is in the mix, as the buying power of even a slower-growing paycheck goes further when prices are also growing more slowly, but that’s just part of the story. In 1999, when the jobless rate was as low as it is now, real wages for these workers rose three times as fast as today (about 1.5 percent vs. 0.5 percent).

The Fed’s higher-interest-rate campaign isn’t helping these workers. My model finds that the higher Fed funds rate is shaving a bit off wage growth, though it’s a relatively small factor.

There are, of course, longer-term developments that have long hurt worker bargaining clout, including the decline of unionization and persistent trade deficits. In his important new book, Rick Wartzman identifies the erosion of the social contract between firms and workers, such that suppressing labor costs, versus investing in your workforce, became a much vaunted signal to investors.

Such disinvestment shows up in my final figure, from a must-read new study on the wage problem from the Brookings Institution’s Hamilton Project. The figure shows the persistent fall-off in labor’s share of income, which predated the last recession. Tightening labor markets have helped to stop the fall, but again, if the job market were really that tight, I’d expect to see this metric climbing back up. (Important note to the Fed: Wage gains paid for out of a rising labor share are not inflationary!).

I’m working on a deep dive into what to do about this problem of slower-than-expected wage growth, and I’ll elaborate those policies in future columns. But diagnosis is always the right place to start.