Lots of measures are telling us that the United States labor market is doing well. In some cases, very well.
Most prominently, the unemployment rate has fallen steadily over the last nine years. It dipped to 3.7 percent in September, and it has averaged 4 percent over the past year, the same as it did at the economic peak just before the 2001 recession.
So in many ways, the labor market looks like the economic heyday of 2000 and 2001, but in one important way it does not: wage growth.
The government has several wage metrics that take different approaches, each with pros and cons. But virtually all of these agree that wage growth today is much slower now than it was on the eve of the 2001 recession, with its identical unemployment rate.
This is, to put it mildly, a mystery. If workers are as scarce as the unemployment rate and many other measures suggest, employers should be raising wages to compete for them.
The most widely followed wage measures grew at an average of around 2.9 percent over the past year. Wage growth is certainly stronger than in the depths of the recession in 2009 and 2010, and has been gradually increasing since then. But it is still well below the 4.2 percent average right before the 2001 recession.
Inflation and benefits can’t explain the decline
So why has wage growth slowed since 2001, across many different measures, when unemployment is so low?
We can rule out two possible reasons immediately. It wasn’t because of a decline in inflation, and it wasn’t because benefits like health insurance and hiring bonuses crowded out wages.
As with wages, the government has many different inflation measures. But none that we analyzed have slowed by nearly enough since 2001 to explain the weakness in wage growth; some have even increased a bit.
And government data suggests that measures of pay growth that include nonwage benefits are also below their 2001 levels. Moreover, the fastest growth in nonwage benefits was before 1994; in recent years, the nonwage share of compensation has grown more slowly.
Very few demographic groups are back to 2001 levels
Another possibility is that slower wage growth is simply a side effect of the changing composition of the American work force since 2001 — for example, a work force that is older than it used to be.
But this hypothesis is also mostly wrong. Using the Current Population Survey to calculate wages across a variety of demographic and labor market categories — including age, sex, education and race/ethnicity — we find that wage growth is consistently slower now than in 2001.
It’s only when we begin to split up the data by geography that we find a group for whom wages have recovered to their 2001 pace — and in fact, now exceed it: nonmetro workers in the West South Central census division (Texas, Oklahoma, Arkansas and Louisiana).
Why might rural workers in these particular states be seeing such strong wage growth? One possibility is the boom in American oil and natural gas production driven by the shale revolution. Several shale basins traverse these four states.
Supporting this theory: If we look at wage growth by the industry of the worker, mining and extraction stands out as being the only one where typical wage growth is above 2001 levels.
Slack, productivity and labor bargaining power
Economists do not agree on what’s driving lower wage growth, but they have several hypotheses.
One is that there’s more slack in the labor market today than the unemployment rate suggests. Slack is essentially the shortfall between the amount of work the economy could be supporting and the amount it actually is.
The unemployment rate counts only people either employed or actively looking for work — not those who give up looking. A rise in what economists call labor force nonparticipation — whether because of discouragement, school enrollment, disability or retirement — was a distinguishing feature of the Great Recession.
One measure that accounts for both the unemployed and nonparticipants is the employment-to-population ratio. Unlike the unemployment rate, it has not returned to its 2001 level. But when adjusted to account for demographic changes, it shows a tight relationship with wage growth. This is what we might expect to see if slack not captured by the unemployment rate were still a drag on earnings.
Another idea is that weaker wage growth primarily reflects a slowdown in productivity gains. Janet Yellen, the former Fed chair, espoused this view in a speech last year. If the value of what workers produce is growing more slowly than in the past, we may expect this to be reflected in smaller raises. And while productivity data is noisy, analysis based on the same approach Ms. Yellen used shows that the trend in productivity growth is down more than a percentage point from its 2001 pace.
This slowdown is its own economic mystery, particularly since research shows that the phenomenon is present globally across advanced economies.
The Current Population Survey does not track worker productivity, but it does show that occupations of all skill requirements — even high-skilled ones — are seeing wage growth down from 2001.
In fact, when we rank different jobs by their skill level using data from the economists David Autor and David Dorn, we find that the highest-skilled occupations — which we would expect to be more productive — typically saw the highest wage growth in 2001. But in 2018, that is no longer the case.
This suggests that if the productivity slowdown is a factor in slow wage growth, it’s a pervasive one across occupations.
A third hypothesis is that weaker wage growth is connected to inequality and lower labor bargaining power.
Inequality has been rising over the long term, and union membership has been on a persistent decline, but neither solely explains slower wage growth since 2001. Wage growth is down since 2001 across all five wage quintile groups and across union membership.
However, rising inequality and falling unionization may be affecting wage growth in more indirect ways, such as in lower upward mobility and weaker bargaining power. Also, some of the same economic forces feeding into them may also be influencing wage growth.
For example, new research finds that employer concentration — when a few firms dominate the market — is an increasing concern in some segments of the labor market, and that it’s associated with weaker wage growth.
Moreover, all three of these hypotheses — and other possibilities — need not be mutually exclusive. Slack labor markets, for example, may be feeding into weak productivity growth, and rising employer concentration may be widening the gap between wage gains and productivity growth.
One thing is clear: Although the unemployment rate may look the way it did in boom times in 2000, for many Americans wage growth has much further to go.
Ernie Tedeschi is an economist and head of fiscal analysis at Evercore ISI. He worked previously at the U.S. Treasury Department. The analysis here is solely his own.
Note on methods: Median wage growth is calculated from the Current Population Survey (C.P.S.) using a methodology originally developed by Mary Daly, Bart Hobijn and Theodore Wiles, similar to the one used for the Atlanta Fed Wage Growth Tracker. The C.P.S. tracks wages and salaries, but not benefits. All of its earnings and hours worked data are self-reported by the household, and so are subject to survey error.
