There’s a reason millions of American workers are still feeling left out from what on the surface looks like a fairly strong economy: a distinct absence of consistently strong wage growth.

The unemployment rate has stayed at or below 4.0 percent since March 2018. But, nominal wage growth continues to be weaker than expected and, in fact, appears to be decelerating this year so far. In our nominal wage tracker that measures year-over-year changes, wage growth has flat-lined in recent months and has yet to reach the Federal Reserve’s target zone (given inflation targets and productivity potential). Looking at more-recent trends—wage growth between the first and second quarters of this year—there has actually been a deceleration in wage growth this year. The Employment Cost Index, released last month, also shows a marked deceleration in private sector wage growth.

Last month, the Bureau of Labor Statistics (BLS) also released preliminary benchmark revisions to payroll employment for April 2018 through March 2019. Each year, the BLS benchmarks total nonfarm payroll employment to state unemployment insurance tax records. While revisions in most years tend to be relatively small and don’t get officially incorporated into the historical numbers until the final revisions are released in February, this year’s revisions came in much higher. The preliminary estimate of the benchmark revision indicates a downward adjustment to March 2019 total nonfarm employment of -501,000. This means that, between April 2018 and March of 2019, there were a half million fewer jobs created than initially reported. Over the last ten years, preliminary revisions averaged about -92,000, so -501,000 is large in comparison. And, usually the difference between the preliminary revision and final is plus or minus 40,000. Therefore, it’s likely the final revisions will also be around 500,000 fewer jobs in that period.

The figure below illustrates what this means for job growth over the last two years. Here, I’m comparing April 2017 through March 2019, linearly interpolating the 501,000 losses equally over the 12-month period. Initially, it appeared that payroll employment growth increased between the year ending in March 2018 and March 2019, with monthly employment growth going from an average of 193,000 to 210,000. With these sizable downward revisions, average monthly employment growth actually fell from 193,000 to 168,000 over those two periods.

While these revisions are concerning in themselves because they mean the economy wasn’t creating jobs as quickly as initially thought, large downward revisions are sometimes associated with early signs of a recession because it means the BLS’s model for predicting the birth and death of firms is off, often accompanied by a turning point in the economy. These revisions aren’t a sky–is-falling moment, but certainly something to keep in mind in coming months. Coupled with other trends (decelerating wage growth, for example), they certainly boost concerns over the economy slowing.

Other things to consider in today’s economy and as the latest jobs report comes out tomorrow are what other indicators are missing in the top line stories of payroll employment growth and the unemployment rate. In honor of Labor Day, the Economic Policy Institute released a series of reports on how well the economy is working for working people. All the short essays are worth a close, but I’m going to highlight two here.

My colleague Valerie Wilson and I illustrate that wage growth remains weak for a labor market with such a low overall unemployment rate, and we also show that the pace of wage growth is uneven across race and gender. What is particularly striking in our findings was that wage growth was stronger in the late 1990s than the last five years, two periods with similar unemployment rates. Not only was wage growth stronger overall, but wage growth was stronger for workers with a college degree in the former period compared to the latter (10.3 percent versus 5.5 percent).

And, a striking pattern emerged. The figure below shows that the 10.3 percent growth of the late 1990s was shared relatively evenly among men and women, and among black and white workers. However, the much lower 5.5 percent average growth rate over the most recent five-year period was not shared evenly at all. Over the last five years the racial wage gap among college graduates has widened as a result of declining wages among black college graduates (-0.3 percent) and rising wages among white college graduates (6.6 percent). The gender wage gap also grew as men with college degrees experienced wage growth more than twice as fast as women college graduates (7.8 percent versus 3.0 percent).

Another essay, authored by Jhacova Williams and Valerie Wilson, highlights how black workers endure persistent racial disparities in employment outcomes. They find that not only are black workers twice as likely to be unemployed as white workers at almost every level of education, but also black workers with a college degree are more likely to have their credentials underutilized in the workplace. This means that black workers are less likely than white workers to be employed in a job that is consistent with their level of education, and are less likely to reap the rewards of those wages and benefits associated with those degrees.

In light of this research, when the jobs report comes out tomorrow, I’ll be watching headline payroll employment and unemployment figures—but also paying close attention to nominal wage growth and how the economic expansion is or is not reaching workers across the economy.