There are two primary measures of employment in the US. The first is Total Nonfarm Payroll, and it was up 272,000 in May. The second is the Employment Level, and it was down 408,000.
These two metrics measure different things, but their growth rates generally move in tandem. Over the past couple of years, however, their growth trajectories have diverged considerably:
Back at the start of 2022, both measures showed growth rates of about 5% year-over-year. Since then they've declined, but nonfarm payroll is still growing about 2% per year. The employment level, by contrast, is close to 0%.
This difference has been especially stark over the past 12 months. Nonfarm payroll has continued chugging along at its usual rate, gaining 3 million jobs. The employment level, by contrast, has been completely flat, gaining almost no jobs:
So what's the employment situation really like right now? I'm not sure anyone knows for sure. But it's possible that it's a little less cheery than it seems.
As far as I can tell, the only way these numbers can be reconciled is if the farm payroll is declining rapidly. This seems unlikely.
Farm employment has been in constant decline for more than a century. But it has long since reached the point that it is so small as to be negligible. The real answer is that the employment level survey hasn't been worth the paper it's printed on since the Reagan administration. Back then everyone led with the unemployment rate. Now it's always at the end of the article if mentioned at all since everyone knows it's a outright joke.
This divergence is nothing new. It can seem that way because it is strongly pro-cyclical (i.e., it often increases sharply in times of rapid economic expansion and contraction).
Check out this contemporaneous paper from the NY Fed, trying to explain a similar divergence that emerged in the late stages of the long 1990s expansion (which incidentally makes clear that the payroll survey is almost certainly providing the more accurate estimates):
https://www.newyorkfed.org/medialibrary/media/research/current_issues/ci5-16.pdf
You can also see the same pattern emerging multiple times during the long boom of the 1960s:
https://fred.stlouisfed.org/graph/?g=1oGGP
Basically, the household survey generally lags the payroll survey when things are getting quickly better (or worse).
Couldn't this also be explained by the baby boom retiring which is really kicking into high gear now.