Making Sense of the Monthly Job Numbers (and a little reminder about how blurry is our vision of reality)

Since 1970, there has been only one year that ended with a lower unemployment rate than 2021. That was 2019. In this century, the unemployment rate has been higher than December's 3.9% 90% of the time. So last month's unemployment rate is really good news.

Job creation of 199,000 was absurdly low for December and while the 6.4 million for 2021 is the most jobs created for any year on record, it still leaves us down 3 million jobs from December of 2019. So that's mixed news. We needed a big number for job creation and we got a fairly normal one (by the standards of this century.)

The variability in monthly job creation during the last 2 years has been incredible. We've had a month in which we lost 21 million jobs and another month in which we gained nearly 5 million jobs. In one month. Annual job creation jumps all over the place but 180,000 is pretty typical for a healthy, normal month. Since COVID hit, half the monthly jobs numbers represented a swing of more than 650,000 in one direction or another, so monthly volatility has been about 3 to 4 times what it normally is.

On top of that, there seems to be a lot of noise in the monthly measures.

How many new jobs were created in December?
The consensus expectation was for 450k.
ADP is a private company that tracks monthly changes in private employment. Their count for December was 807k.
BLS's household survey that - well, surveys households - arrived at an estimate of new jobs of 651k.
And the official number from BLS - which surveys employers - shows a gain of just 199k. 

This 199k is the headline number.

It's worth remembering that during the long run of uninterrupted job creation in the aftermath of the 2008 Great Recession, the jobs created in a a normal month was about 200,000. What is often forgotten is that this is a net number. Every month about 2 million people were fired, laid off, quit or retired. And about 2.2 million were hired. So, the net for the month was 200,000 "new jobs" but the fact was that there were millions of new jobs that offset the millions of jobs ended. If your measurement error on the 2.2 million new and 2 million ended jobs is just 5%, and the real net for the month was 200,000, you could double the reported new jobs or erase them. Let me repeat that. With an underlying reality of 200,000 new jobs and a measurement error of 5%, it is possible to report that as anywhere from 0 to 400,000 new jobs. The last 22 months have been so volatile that it's easy to imagine that measurement error has gone up. I don't know what actual measurement error is. I do know that it exists and that's just one reason to be cautious in interpreting numbers from one month - particularly before they've been revised. 

The BLS numbers will be revised twice more over the next two months, as they always are. Pandemic volatility makes it tough to track what is going in this labor force of 162 million Americans. The good news is that we're moving in the right direction. By one measure - the unemployment rate of 3.9% - the job market is already healthy. By another measure - the millions of jobs that we've lost over the last couple of years - the economy is still weak.

As with so much in life, you can choose your narrative. Job creation is strong? Cheer about that or complain about inflation. Unemployment rate is low in December? Cheer about that or complain about the rate of job creation in December.

Lots of early retirements and some long COVID disabilities are possibly depressing the number of folks looking for work, which would explain how the unemployment rate could be so low even while we are still millions of jobs short of where we were 2 years ago. Additionally, there seem to be lots of folks who simply haven't worked out issues like childcare or elder care to be able to work again. And while we're buying more goods than ever before, the service sector still hasn't fully recovered.  You might have better or more interesting theories about what is going on. The most robust theories still await more data, though, data that smooths out the monthly variations that sometimes seem nearly as big as the phenomenon they're measuring.