Birth-death model adjustments to payrolls

There's been a lot of talk over the last year about the Labor Department's birth death model and its impact on the monthly nonfarm payrolls report - it seemed like a good time to whip up a table to get a better idea of what's going on.
Recall that these estimates are used to arrive at the "unadjusted" changes to nonfarm payrolls along with data coming from the Labor Department's survey of state employment insurance records for some 400,000 businesses. The birth death model accounts for the "birth" and "death" of businesses that are not reflected in the regular survey.

This total is then subject to seasonal adjustments, which results in the monthly data that is widely reported in the media.

For example, the details of last Friday's labor report for the month of November showed a total of 240,000 new jobs via the establishment survey, of which 51,000 jobs could be attributed to the birth-death estimate (see above). This 240,000 figure was then seasonally adjusted down to 94,000 which is what you probably heard in the news.

So, last month the birth death model accounted for 21 percent of all new jobs reported which sounds reasonable. But, looking back in the BLS data you see that a total of about 1.5 million jobs now exist that weren't there a year ago and, of that total, over 1.1 million jobs are the result of the birth death model.

This seems like quite a bit - more than 70 percent - and this historically high ratio has been the major complaint about the birth-death adjustments.

Looking at the individual categories provides more detail.

In the Leisure and Hospitality category a total of 368,000 new jobs were created and 328,000 of them were from the birth death model - that's 89 percent.

For Trade, Transportation, and Utilities the total is 74 percent, for Professional and Business Services the total is 51 percent, and for Education and Health Services the total is 29 percent. Note that Education and Health Services is, by far, the single biggest source of job growth adding 561,000 jobs alone over the last 12 months.

The first category that appears to be seriously distorted by the birth death adjustment is Construction, where, before adding in the birth death adjustments, about 274,000 fewer jobs exist today than a year ago, but including the adjustment only 134,000 jobs have been lost.

Similarly, in the Financial Activities category, before adding in the birth death adjustments about 126,000 were lost over the last year, but with the adjustments the job loss is only about 12,000.

Both of these last two adjustments are obviously in error and will be subject to massive revisions.

Last year there was an upward revision of some 800,000 jobs due to the recalibration of the birth-death model long after the initial monthly data was released. Don't be surprised if the next downward revision matches this total.

--------------------------------------

UPDATE: Monday 12/10, 2:55 PM

Here's a new chart - note the areas in red.

The Gain/Loss as Reported column is the 12-month total as of November (these are the not-seasonally adjusted numbers but it really doesn't make a difference). The Birth/Death Adjustment is the data from the table above from this page at the BLS. The Survey Data column is column 2 minus column 3 and the Birth/Death Adjustment as a percent of the total gain/loss as reported is column 3 as a percent of column 2.