Economist Writing Every Day

Yesterday I showed that BLS jobs reports from the CES aren’t getting worse over time, if we judge them by how much they are later revised. In fact, they are much better than decades past, with the last 20 years or so standing out as much better than the past.

Today I want to address a related but separate topic: are the initial jobs reports good at telling us when a downturn in the labor market is beginning? This is actually the strongest argument for releasing this survey data in a timely manner, even though the data often goes through significant revisions later. The report typically comes out the first Friday of a new month, so it is very current data. Given that the likely new BLS Commissioner has signaled he prefers the more accurate quarterly release, even though it is 7-9 months after the fact, it is useful to ask if these initial reports have any value in telling us when labor market declines (and recessions) are beginning.

That’s right: you are getting two posts from me this week, on essentially the same topic. Because it’s very important right now.

The short answer: the report is very good for the purpose of identifying downturns, especially the start of the downturns. Let’s walk through the past few recessions.

COVID Recession, 2020

As we all know, this was an unusual recession in many ways. It was caused directly by government orders to shut down businesses and consumers suddenly drastically changing their behavior in the spring of 2020. It was also very short, in terms of the time the economy was contracting, even though the recovery took years.

According to the NBER dating, the peak in activity was February 2020, and the trough was just two months later in April 2020.

The initial jobs report nailed this one perfectly. It showed the first monthly job loss in March 2020, and the labor market was starting to recover in May 2020. So the dates from the initial jobs report exactly match the NBER dates, and perhaps more importantly, the months of decline in the initial jobs report exactly match the revised numbers, even though the exact numbers weren’t right. The job losses in March 2020 were actually twice as large as initially reported (1.4 million rather than 700,000), but the initial April report was almost exactly correct (both the initial and revised numbers are 20.5 million).

Great Recession, 2008-2009

This recession was arguably the worst since the Great Depression of the 1930s, so it’s a really good test of whether the initial jobs report got it right.

The NBER dates say that the peak was in December 2007. The initial jobs report tells us the first month of employment decline was January 2008, as does the revised jobs numbers. Nailed it perfectly again!

The job market keeps declining for all of 2008, all of 2009, and through the first 2 months of 2010. Yes, it was a long downturn! Both the initial and revised numbers show the exact same pattern, with March 2010 as the first month of job growth. The NBER dates actually show that the recession ended in June 2009, as other indicators began to improve, even though the labor market kept declining. In fact, there are 4 later months of decline in jobs from June through September 2010, and both the initial and later revised jobs numbers agree on this (though the revised numbers show one month of growth in jobs in the midst of those 4 months, but only 1,000 jobs).

I think we can say that the initial jobs numbers did very well at getting the timing of this labor market recession perfectly. Note: there were 2 months of job declines in mid-2007 (July and August) that aren’t considered to be part of the recession. The initial jobs report missed one of those months. So not exactly a perfect record. But that ended up being irrelevant for the start of the recession.

Post-Dot Com Recession, 2001

The US experienced a mild recession in 2001. According to NBER, the business cycle peaked in March 2001 and was over by November 2001, so a very short recession by historical standards, just 8 months.

According to the initial jobs report, the first month of job losses were in March 2001. This agrees with the revised jobs numbers too, though there was one month of job decline in January 2001 (though not February) in the revised numbers.

The initial jobs report showed that employment started to grow in February 2002, though the later revised numbers say that didn’t start until May 2002. So not a perfect match on the ending.

Also, it is worth noting that both the initial and revised jobs numbers showed job losses in the second half of 2002 and through most of 2003, though they don’t agree on every single month. But 2002-03 is not considered by the NBER to be part of a new recession, nor a continuation of the 2001 recession. Consistent job growth doesn’t actually start until September 2003 — and both the initial jobs report and the later revisions agree on this point.

1990-1991 Recession

This recession was another short, 8 month recession. It’s worth pointing out here that the quarterly QCEW jobs numbers, which are certainly much more accurate, aren’t available until about 6-8 months after the fact. For example, the report for the first quarter of 2025 will be released in September 2025 — so the January 2025 figures aren’t available until 8 months later. Do you see the problem here? This recession and the previous one were only 8 months! So if we are relying on the QCEW release, you won’t even know there is a downturn in the labor market until the recession was already over. You will get data from the other NBER indicators (they mostly come from BEA), but without the initial jobs reports, policymakers would be flying blind on the labor market.

NBER dates this recession as starting in July 1990. The first month of job declines in the initial report: also July 1990. The later revised figures agree with this month for the start of the downturn.

The recovery began in March 1991 according to NBER, even though job losses continued consistently until August 1991. The initial and revised jobs figures agree on this August 1991 date, though they each have one month of growth before that, but not the same month (May in the initial report, June in the revised report). Once again, there are sporadic job losses over the next year or so, and the initial and revised reports don’t always agree on the months, but 1992 is not considered by NBER to be part of the recession (even though they don’t make the announcement until December 1992 — conveniently one month after an election when the recession was one of the main issues).

False Positives?

We can see that in each of the last 4 recessions, the initial jobs report performed extremely well. It always agreed with the revised numbers on the start of consistent employment declines, and it usually agreed on the end date of those job declines (usually within a month or two). The initial jobs report declines also almost always coincided with the start of the NBER-defined recession. This is in contrast to what you sometimes hear about employment being a “lagging indicator.” This might be true of the unemployment rate, but it is not true of job growth: declining job growth (from the initial jobs report!) tends to coincide with the downturn in other NBER indicators (industrial production, income, spending, etc.), even though job losses sometimes continue for several months after those other indicators also improve.

But does the initial jobs report ever give us a false positive? Does it ever signal a recession might be starting, but also turn out to be wrong? In other words, do the revised jobs numbers ever tell us it was a false alarm?

There are a few candidates for this criteria, but in general I would say the initial jobs report is not giving us false positives on the start of a recession.

Let’s work backwards in time.

September 2017 is one candidate. The initial jobs report said that there were 33,000 jobs lost. The later revisions, including the one released just the next month, do not show a decline. So… false positive?

Probably not in a serious way. First, it was a small job loss. Second, one month later we already saw that not only was this number revised up, but that October 2017 also was not a month of job declines. In fact, media reports at the time attributed this one-month blip to the effects of hurricanes. In other words, nobody freaked out and started preparing for a recession. That initial report turned out to be wrong, but really no harm here.

August 2011 is another month to examine. Here is an interesting one: the initial report showed exactly zero job growth. So if revised, it would be wrong in either direction, whether it was revised up or down! But by the following month, it was already revised up to 57,000 jobs, and later revisions would actually peg it at 132,000.

Was this a false positive for a recession? The key here is don’t freak out about one month of job losses or no growth, unless they are very large (see March 2020). There was some freakout at the time, and blaming of the President (apparently some Republicans called Obama “President Zero”), but this was perhaps warranted because the labor market was still recovering from the 2008-09 recession: unemployment was over 9%, even though jobs had been growing consistently for almost a year.

As it turns out, this report was overblown: and we already knew that within one month, as the numbers were revised in the September 2011 report.

I’ve already discussed the August 2007 false positive above, though that decline showed up in the revised job numbers too, so it wasn’t actually wrong, it just didn’t signal the beginning of a recession (and anyway, it was a single report of 4,000 job losses).

September 2005 was the only other false positive worth mentioning in the past 20 years. The initial report showed 35,000 job losses, as did the revision the next month, but by the second revision (November 2005 report) it was a positive report. And there were no subsequent declines in any other months after September 2005.

Once again, no freakout. Why? Again, media reports attributed it to a hurricane. In fact, estimates were for even larger job losses, possibly of 150,000 jobs! So this was clearly not the start of a recession, and everyone knew it at the time.

Summing It All Up

So we can see two things from my post.

First, in the past 4 recessions, the initial jobs reports nailed the start of recessions. If you look for 2 or 3 months of consistent employment declines, without them being revised up, it’s a really strong indication that you are entering a recession (especially if you ignore hurricane-related declines). And these initial reports were perfectly consistent with the timing of the later revised reports.

Second, in the past 20 years, there are maybe 3-4 months where the initial jobs report gave a false positive, but all of these are pretty weak candidates (again, two of them were hurricane related!). Once again, if you look for 2-3 months of consistent employment declines, you won’t be fooled by these initial job reports.

And most importantly, waiting for the quarterly QCEW jobs data would have been too late in 3 of the 4 most recent recessions. And even for the Great Recession, you wouldn’t be seeing the job declines until 8 months after the fact, as opposed to 1 month with the initial CES jobs data.

Conclusion: the initial jobs report is good, and worth both continuing to do and continuing to release it to the public. If it can be improved, do it! But delaying the release is not helpful for economic purposes (though it may be helpful for political purposes).