BLS Birth Death Conundrum ?

July Non Farm Payrolls is out tomorrow, making today a good day to review one of our favorite whipping boys: BLS Birth Death model.

We’ve discussed this ad nauseum, but a brief explanation for the new BP readers: The idea behind the B/D model is to make up for the lag between new firms forming and when they eventually get counted into the Labor statistics. B/D has two major impacts over the course of an economic cycle: It makes the reporting of job creation more accurate in the early recovery phases, but at the expense of overstating employment at the end of the cycle.

I’ve noticed that quite a few defenders of the Birth/Death Model have become emboldened by the market rally. I don’t want to name names, but as a group, the B/D defenders were horrifically wrong across the board about nearly everything — about the housing crisis, the credit collapse, the recession, the market crash, and of course, the massive loss of jobs since hiring peaked late 2007-08.

Their track record precludes taking their rhetoric about B/D very seriously.

For those of you who relish an intellectual challenge, consider the following: Since reaching peak employment in January 2008, small-size businesses have shed nearly 2.5 million jobs. Construction employment dropped for the thirty-first consecutive month in August. The total decline in construction jobs since the peak in January 2007 to 1,562,000. ADP counts 2.5 million small biz job losses since January 2008.

Over the same period, the BLS’s Net Birth/Death Model, which is supposed to account for small biz job losses and gains, shows 1.507 million job gains. The difference is over 4 million jobs — about 3% of the total employed.

A quick look at the historical data tracks the changes to the B/D model, proposed in 2001 ans implemented a few years later (try to guess where):

Historical Net Birth/Death Adjustments *
1999:  30k
2000: 193k
2001: 91k
2002: 470k
2003: 695k
2004: 827k
2005: 817k
2006: 1002k
2007: 1130k
2008: 904k
2009: 555k  (January – July)


* 1999 to 2002 are April thought March measures; 2004 forward is calendar year; 2003 data is for April-December. Data for 2007-09 is preliminary, subject to future benchmarking


Is there a legitimate way to defend this? Or, is the B/D model so wildly wrong as to be indefensible?


Current Employment Statistics – CES (National)
Historical Net Birth/Death Adjustments

CES Net Birth/Death Model
Current Employment Statistics – CES (National)

What's been said:

Discussions found on the web:

Posted Under