Did 818,000 jobs vanish?

This morning the Bureau of Labor Statistics released the latest quarterly data for their Quarterly Census of Employment and Wages for the first quarter of 2024. Along with this release is the announcement of their preliminary “benchmark estimate” for March 2024, which will eventually (next year) be used to revise employment data for the Current Employment Statistics program. To keep all of the alphabet soup of programs clear in year head, CES is the more familiar “nonfarm jobs” data that is released each month, usually with some media fanfare.

Benchmarking is an important part of the process for many data releases, because the monthly CES data is based on a survey of employers, a subset of the total. But the QCEW data is the universe of employees — at least the universe of the those covered by Unemployment Insurance law, which is something like 97-98% of workers in the US. So the numbers will never match exactly (CES is supposed to be measuring all workers, not just the 97-98% covered by UI), but they should be pretty close. The media reports the CES monthly data more prominently, because it is more timely and usually pretty close to correct — but benchmarking is the process to see just how correct those initial surveys were.

That brings us to the release today, which is the preliminary estimate of the benchmark adjustment for March 2024 (it will be finalized early in 2025). And that preliminary estimate was a big number, with a downward revision projected of 818,000 jobs. To put this in perspective, the current CES data shows 2.9 million jobs were added between March 2023 and March 2024, so this estimate suggest that the job growth was overstated by perhaps 40 percent. That’s a big revision, though large revisions are not unheard of: the same figure for March 2022 was an estimated 468,000 jobs higher, while March 2019 was 501,000 jobs lower. But this year is a big one (largest absolute number since 2009). Here’s a chart summarizing recent years revisions from Bloomberg:

I’ve covered this topic before, such as an April 2024 post where I noted that as of September 2023, there was an 880,000 gap in job growth between the CES and QCEW over the prior year. So this was not unexpected, and in the days leading up to the report, close followers of the data were forecasting that the revision could be up to 1 million jobs.

Some on social media claiming this shows that the Biden administration has been lying about the jobs numbers or cooking the books. This is a strange argument — again, recall that the March 2022 data had to be revised upwards significantly, so if they are cooking the books they are doing so in a very strange way. A lot of that is political posturing without much understanding of the data.

But the more serious question is how much we should trust the monthly jobs number. If we spread the 818,000 jobs revision over 12 months (which is, more less, what BLS will eventually do once they have the final numbers), this amounts to 68,000 fewer jobs per month on average. Over that time frame (March 2023 to March 2024), the average monthly job growth was 242,000 jobs. Crucially, the lowest monthly job growth during that period was 165,000 jobs, in October 2023. Cutting 68,000 from that month would indeed be large, but every single month in the revision period would still be positive.

As I said in both my April 2024 post and an earlier January 2023 post, it seems that the monthly jobs survey is indeed overcounting jobs, but once we have the complete data the job growth is still positive — just not as robust as originally reported. So discount the monthly reports a bit, but they probably get the direction of job growth correct.