On Those Downward Jobs Revisions

The Bureau of Labor Statistics reported that its benchmark revision to its job numbers shows that the economy created 501,000 fewer jobs between March of 2018 and March of 2019 than previously reported. There are a few points to be made about this number.

First, there is nothing fishy here. Trump has zero to do with the data that comes from the Bureau of Labor Statistics (BLS). BLS is staff by committed professionals who would sure raise a big stink if Trump tried to tamper with the data.

I should also point out that it would be exceedingly difficult for someone to change the data if they did not have a very good idea what they were doing, and even then they would almost certainly have to bring dozens of people in on the scheme. If someone did something like just 100,000 to the monthly job growth number, they would be nailed in a minute. Other numbers would not fit and it would be easy to see that the fake number was out of line.

Anyhow, the revision is based on state unemployment insurance filings, which give a virtual census of payroll employment in the United States. The original data come from the BLS’ monthly Current Employment Situation survey. This is a large survey of businesses, but it is a survey, so that means there will be some error.

The next issue is why the survey would be so far off. (The 501,000 reduction is much larger than a normal revision.) In addition to the survey results, BLS imputes figures for “births” and “deaths” of firms. Births refers to new firms, which could not be included sample because they are new. Deaths are the firms that go out of business and aren’t so polite as to answer the survey before they shut their doors.

BLS imputes numbers for births and deaths using a model that estimates these data based on growth in output and related factors. It usually is reasonably accurate, but in this case it clearly was not. (I’ll make a small criticism of BLS here. They typically show the error as a percent of total employment, which makes it look small. It came to 0.3 percent last year. But what we really are measuring with the survey is the change in employment, which was just over 2 million, which means the error was 25 percent. That is a big deal.)

Anyhow, what this revision means is that we saw more firms die or fewer new firms formed than the model projected. (Actually, the issue is jobs, not firms, but presumably these go together.) That may mean nothing or could suggest that either or both, more firms are going out of business or fewer firms are being started than we should expect, given other factors in the economy.

That brings us to my last point, when we have large downward revisions, it usually is associated with a recession. The downward revision in 2009 was over 900,000 and 2002 it was over 300,000. It is unlikely that we will find that we were actually in a recession between March of 2018 and March of 2019, but add this to the list of worrying data points. It seems that something is not right with the economy.

This article first appeared on Dean Baker’s Beat the Press blog.

Dean Baker is the senior economist at the Center for Economic and Policy Research in Washington, DC.