This special weekly wrap on the US employment numbers has become a regular rendezvous. On the first Friday of each month I sit still in front of the BLS website waiting 2.30 pm with excitement. Today is even more special, as the Labor Department has finally come up with the yearly revision of the Establishment survey figures. The survey and more precisely the birth death model has again (as in the previous recovery periods) put too little weight on the tough conditions faced by smaller companies. As I pointed before, the expectation was 800,000 additional job losses over 2009. The number came out as 620,000. Good news! Now, the bad news: overall, since the beginning of the recession, the economy lost 8.6 million jobs instead of the 7.2 millions previously reported. Ouch… More bad news, the BLS reported that total nonfarm payroll employment declined again by 25,000; so much for the positive figure expected by the market. Now, what about the unemployment rate? Down to 9.7%: isn’t it a great news? At first sight it may be, and the US administration will surely take this as a victory. Unfortunately, that’s another statistical trick; stick with me till the end and you’ll discover the inconvenient truth.

 

Establishment Survey

According to the BLS website, Construction employment declined by 75,000 in January, with nonresidential specialty trade contractors (-48,000) accounting for the majority of the decline. Since December 2007, employment in construction has fallen by 1.9 million. Housing still the black sheep…

On the other hand, temporary help services added 52,000 jobs and retail trade employment rose by 42,000. The federal government added 33,000 jobs, including 9,000 temporary positions for Census 2010 whereas employment in state and local governments, excluding education, continued to trend down

As for the revision, as I told you before, a figure of 800,000 additional job losses was expected. The US economy finally lost “only” 620,000 additional jobs in 2009 and 1.4M since December 2007.  An upward revision in November mostly explained the gap. Indeed, the nonfarm payroll in November showed an increase of 64,000 (a green shoot that might be only due to statistical error).  

The biggest downward revision occurred in March (-101K to 753 K total losses) and in November (-97 K for a total of 221K losses). I must say that I’m rather pleased with the latter. I was forecasting back then more than -200K, a bad call it looked, and worse, the revisions kept proving me wrong each time. I was having doubts on my forecasting abilities… Actually, I gave up forecasting after that; but now given the final number, I might also revise this.

So bottom line from the ES, the recession was worse than we thought and real improvement in the unemployment rate (i.e. outright job growth of more than 100,000) is still far way.  

But, wait a minute, the unemployment rate? It is coming down to 9.7 % from 10% according to the Household Survey! So, where is the catch? How do you square one with the other?

 

Household survey

So, first of all, one point does not make a trend and the US administration ought to take this number with a grain of salt.

Looking at the broad picture, employment increased by 540K and unemployment decreased by 440K, hence the substantial decrease from 10 to 9.7 %! But, the devil is in the details.

Last month, the weak point was the labor force participation rate which fell to 64.6%. As I noticed previously, it implied that 660K people left the labor force in one month, which is equivalent to what we observed in one year! (from Nov 08 to Nov 09 ). Now, in January, this rate was little changed at 64.7 percent. However, no change is still bad news, since those “desperate” people will come back and search for a job at some point. So, expect a higher unemployment rate down the road.

But now, without further ado, here goes today’s devil. Again, this is nothing but a statistical trick. Check the non seasonally adjusted numbers, and you’ll find that the unemployed people from December to January  increased by more than 1.3 millions, from 14.7 to 16.1 millions (employment increasing only by 1 million). If you redo the calculation given these non seasonally adjusted numbers, the unemployment rate is actually 10.6 %!   

But, for some reason, January is one of the months with the largest seasonal adjustment up in jobs. That’s to say that the seasonal adjustment factor decreased the unemployment figure from 16.1M to 14.8M while increasing the employment numbers by 500K. You understood me well, with a simple tap on a keyboard, the BLS erased the 1.3 people who lost their job from December to January !

Now, I agree on the fact that seasonal adjustment is indeed necessary to get rid of fluctuations due to weather changes, holidays, and schools opening/closing. But, to trust those numbers, one should really understand the BLS methodology, and I don’t (I’ll try again at some point). But if you know please, can you explain to me how this is really done, and how you can decrease a monthly figure by 1.3 millions? Plus, is January so different from December regarding those conditions?

Moreover, if I plug the 1.4 million revisions from the Establishment survey into the seasonally adjusted unemployment rate, the figure is more in line with the unadjusted figure (even reaching 10.7 %). I tend to trust more this type of manipulation…    

The market seems puzzled by the discrepancy between the Establishment survey and the Household survey. I already pointed out some seemingly conflicting results. But when you dig into the numbers, these surveys are actually both pointing toward the same direction: North.

I’ll be still sitting in front of my PC next month, watching the inconvenient truth officially revealed.