Photograph by Nathaniel St. Clair
Most coverage of the May jobs data said the picture was pretty good, an assessment with which I largely concurred. The economy added 139,000 jobs and the unemployment rate remained at 4.2 percent, a historically low level. Wage growth also continued at a healthy 3.9 percent year-over-year pace, with the annualized rate over the last three months, a nearly identical 3.8 percent. All looks pretty good.
But it is possible to tell a more negative story. To be clear in what follows I am 100 percent guilty of cherry-picking. I am looking for evidence of a weakening labor market and ignoring factors that could go in the other direction.
The reason is that we have seen many data points elsewhere, notably indexes measuring consumer and business expectations, that paint a very negative picture of the economy. The reality and threat of massive tariff hikes has clearly scared people, as have the threatened and actual cuts in federal spending and employment.
It also has changed behavior in measurable ways, notably the surge in the trade deficit following the election and then the plunge reported for April. The deficit went from $74.3 billion in October to a peak of $138.3 billion in March and then fell back to $61.6 billion in April. People stockpiled items they expected to be subject to tariffs in these months and then cut back sharply after they had made their purchases and the tariffs started to take effect. So, the negative story is not based just on people’s speculation about the future, it can be seen very clearly in real world data.
I’ll also make an important point about the sort of recession we should be expecting. It will not be a cataclysmic collapse like our last two recessions. The 2008-09 recession was triggered by a collapse of an unprecedented nationwide housing bubble and then a subsequent financial crisis. The 2020 recession was the result of the economy shutting down due to the Covid epidemic.
The best comparisons for the sort of recession we are likely to see are the 1990-91 recession, which came on gradually, and the 2001 recession, which didn’t even feature the standard two consecutive quarters of negative growth.
Going back to prior recessions may also not be terribly useful since the economy was very different. Specifically, the highly cyclical manufacturing sector accounted for more than 21 percent of employment just before the double-dip 1980-1982 recessions compared to less than 8 percent at present.
The Bad Story for May
The start of the bad story is that the 139,000 jobs number reported for May is not as strong as it appears on its face. The jobs numbers reported for March and April were revised down by a total of 95,000. To be clear, there is nothing nefarious about these downward revisions, as conspiracy-minded folks had claimed during the Biden years. We simply got more complete data as more establishments responded to the survey.
The job growth was also highly concentrated in the healthcare sector. Healthcare has been leading job growth throughout the recovery, but the 62,200 gain reported for May accounted for almost 45 percent of job growth for the month. Given the likely cuts to the sector in the 2026 federal budget, it seems unlikely the rapid pace of job growth in the sector will continue.
Most of the rest of the job gains came from the local government sector (also being hit by cuts) and hotels and restaurants. Construction added just 4,000 jobs and manufacturing lost 8,000.
There is also some evidence of slowing wage growth. As noted above, the standard year-over-year measure shows no slowing, nor does annualizing over the last three months, but we do see a somewhat different picture if we compare the average of the last three months with the average of the prior three months. I had been using this measure for years because it gives us more recent data than the year over year number and also avoids the erratic movements we get from using data from a single month. I eventually gave it up when I realized I was literally the only person using it.
However, if we do look at the old Baker wage measure, we see that the hourly wage grew at a 3.5 percent annual rate, comparing the average of the last three months (March, April, May) with the prior three months (December, January, and February). That is a definite slowing.
If we look at the story with production workers in the low-paying hotel and restaurant industry, the annualized rate of increase was just 3.3 percent. Wages in this sector are highly sensitive to the tightness of the labor market. They rose rapidly in the tight labor market years of 2022 and 2023.
The household survey is also showing some signs of weakness. While I would not complain about the 4.2 percent unemployment rate, it is worth noting that the employment-to-population ratio (EPOP) fell by 0.3 percentage points as 625,000 people reportedly left the workforce.
This was not just baby boomers opting to retire. The EPOP for prime age workers (ages 25 to 54) fell by 0.2 percentage points.
We also saw some weakness looking at specific demographic groups. The unemployment rate for workers between the ages of 20-24, which is highly erratic, remained at the 8.2 percent rate reported for April. This is 2.2 percentage points above the low hit in January of 2024. Their EPOP fell to 65.4 percent in May, 2.8 pp below the peak hit last January.
Young workers tend to be hit hardest and first in a downturn for the simple reason that they are the ones most likely to be looking for jobs. The unemployment rate for workers between the ages of 20-24 had risen from a low of 6.5 percent in 2000 to 8.1 percent in April of 2001, just as the recession was beginning. It eventually peaked at 10.6 percent in 2003, a total rise of 4.1 percentage points
By comparison, the overall unemployment rate had risen from a low of 3.9 percent in 2002 to 4.4 percent in April of 2001 and peaked at 6.3 percent in 2003. The total increase was just 2.4 percentage points. In the 1990-91 recession, the unemployment rate for young people rose by 3.6 percentage points compared to a rise of 2.8 percentage points overall.
There is a similar story with the unemployment rate for Black workers. The unemployment rate for Black workers actually fell slightly in May to 6.0 percent, but that is still 1.2 percentage points above its low of 4.8 percent in April of 2023. The unemployment rate for Black women edged up to 6.2 percent in May. This is the highest unemployment rate since February of 2022.
The unemployment rate for Black workers rose by 4.0 percentage points in the 1990 recession, with the low actually hit just after the official start date for the recession. In the 2001 recession the unemployment rate for Black workers rose by 4.5 percentage points.
Another item that is highly cyclical and clearly pointing in the wrong direction is the percentage of unemployment due to voluntary quits. This measure reflects workers’ confidence in their labor market prospects since it indicates their willingness to quit a job before they have a new job lined up.
This fell to 9.8 percent, the lowest share since May of 2021, when we were still in the early phase of recovery from the Covid recession. By comparison, the share averaged 13.2 percent in 2018-2019, when the unemployment rate was roughly comparable.
The share of unemployment due to quits fell from a peak of 15.2 percent in 2000 to 13.3 percent in March of 2001, eventually bottoming out at 8.8 percent in 2003. The peak before the 1990 recession was 17.2 percent in 1989, falling to 15.3 percent in March of 1990 and eventually bottoming out at 9.5 percent in 1992. (The survey question for this was slightly different before 1994.)
So, does all of this mean a recession is on the horizon? I don’t think we can look at the May data and make any strong claims like this. However, we can say there are some clear warning signs in this report. It is far from unambiguously positive.
