The Birth/Death Model

Posted by Ray on July 10, 2010 under Economy | Be the First to Comment

I just read John Mauldlin’s weekly newsletter and he apparently got into a discussion over the much misunderstood and hated, by myself and many others, birth/death model used by the BLS. I have not been receiving the newsletter on a regular basis, some type of server issue I guess, but someone did not like what he had to say about it and I guess John misspoke about it. The whole thing about the birth/death model is it is meant to be a smoothing mechanism, I know that, everyone knows that, but it stinks and is not accurate which is what John ultimately said it was in his letter or at least it has not been accurate the last few years.

Now, I may have misspoke or led many to misunderstand what the birth/death model is and does. It is this little provision that helps the BLS make up for data they either do not receive back by survey participants or never receive, but it leaves room for interpretation and is never fixed in real time. In fact, they wait until February of the following year to correct any errors in the birth/death model that it may have had on the unemployment rate, fantastic, right? The model is not seasonally adjusted so when the BLS says 83,000 jobs were created it is not as if they added in the 147,000 (June’s B/D adjustment) figure to come up with that figure, that would be lying and a government agency would never do that.

Instead, what they do is add in the not seasonally adjusted B/D figure to the not seasonally adjusted employment figure and THEN seasonally adjust it. Now, you are thinking, big deal that shouldn’t make a big difference. Well, you may be right and you may be wrong. If you are talking 1M as a figure and the B/D adjustment is 50K it is no big deal, but if you are talking about a headline figure of 800K and the B/D adjustment is 147K (June figure) or 241K (May figure) well, you tell me, would that impact the seasonally adjusted figure? I would say yes it would. I have history on my side on this as well.

You see, in the fall of 2008 when Lehman collapsed and the world came to an end we all saw unemployment shoot to the moon, remember? Well, the BLS thought since so many people were losing their jobs and the business environment was so good that must be why so many survey respondents did not get back to them, they were busy making money! So, they added in hundreds of thousands of jobs from September of 2008 until the end of 2009. They were so aggressive in their B/D modeling they underestimated unemployment by 880,000 people, that is a pretty large underestimation by anyone’s standards considering the total ‘official’ unemployment total is 14M people and the underestimate for that time frame was about 10% of the total of the newly unemployed.

One could say, well, that is within the margin of error, but I don’t buy that since the government is the one who processes unemployment benefits and receives the initial claims data. In other words, it is pretty easy to correlate the data within a reasonable time frame, in my mind at least, but I am not a bureaucrat, so what do I know. Basically, if one removes the B/D figures from the non-seasonally figures and seasonally adjusts them you would have a bit of a difference in the monthly numbers. The series would be much more volatile, but it would also, in my mind, be more accurate and real time which seems to be something no one wants anymore with this figure which is why Santelli and Liesman get into screaming matches about it every first Friday of the month.

The bottom line is the adjustments matter, they boost the jobs number every month and they don’t come clean about any adjustments until the next year. That does not help anyone except for politicians and when more and more people are saying employment is now a leading indicator we need a better way to report unemployment. At the very least the BLS can correlate with the state data bases along with the household survey and that might give us a better view of what is going on. I think it is pretty much a proven fact that when we have the government guessing at any figure it is pretty much going to be wrong so why anyone would defend the B/D model is beyond me. The idea is fine, I guess, but how the BLS does it and how no one questions it, especially when it creeps up month after month when it really shouldn’t be, is very odd as, again, the only people who truly benefit from it is the political class.