Monday, February 14, 2011

The January Jobs report: the Unemployment rate II.

- by New Deal democrat

This is the third and last post in which I dissect the January jobs report. As you recall, there was a lot of head scratching initially because employment only rose 36,000, and yet the unemployment rate fell 0.4% to 9.0%. Among the more ignorant or conspiracy minded observers, this was deemed impossible.

In my first post, I explained that the 36,000 employment number is likely to be revised significantly higher. In my second, I showed that, based on past relationships of initial jobless claims vs. unemployment going back almost half a century, the drop in the unemployment rate was no fluke. In this installment I am going to argue that the substantive reasons for the drop in the unemployment rate were:

1. population effect.
2. seasonality effects.
3. aging boomers.

1. Population effect

The important thing to remember is that the EMployment number comes from the establishment survey, in which businesses are polled. The UNemployment RATE comes from the entirely separate household survey, in which a smaller sample of households are polled. While over time the two surveys correlate very well, in any given month they can give wildly different results. After taking into account the size of the sample, in this case, businesses told one set of surveyors that they had only hired 36.000 new workers . But households told the other set of surveyors that 589,000 of their members had started new jobs in January!

So, without the annual population adjustment, here is what the BLS report on the Household Survey Data would have read like:
The civilian noninstitutional population grew to 239,051,000 in January. The civilian labor force remained constant at 153,690,000. Civilian employment increased sharply by 589,000, and the number of unemployed declined by 590,000. The unemployment rate declined to 9.07% (rounded to 9.1%) and so declined by 0.3%.
But the BLS does one, annual population adjustment for this survey, in January, and thus the entire effect is captured in one month rather than spread throughout the year. Because of that, the official report instead showed that population had declined by 185,000, employment had only gone up 117,000 (which is still considerably higher than the establishment survey), unemployment had decreased by 622,000, and so the unemployment rate was 9.04%, rounded to 9.0%.

All of this information is contained in the very simple, easy to read Table C of the Employment Situation Report for January. (BTW, in recessions there is less immigration, and couples put off having children, hence population tends not to increase as much as otherwise).

2. Seasonality effects

There is a lot of temporary holiday season hiring in October through December, and generally those people are then let go in January. You may recall that in 2008 and 2009, holiday season hiring - and subsequent firing - was abysmal. I suspect that there was a different holiday season hiring and firing pattern in 2010 than in prior years, which resolves most of the conundrum as to why unemployment rose so much in November and dropped so much in January.

Why do I believe this was so? First of all, because the surprise decrease of 0.4% in the unemployment rate in January is the mirror image of the similarly surprising INcrease in the unemployment rate from 9.4% to 9.8% in November. (BTW, I did a little search, and the people who are sure that the sudden drop in the unemployment rate last month was a canard, had no problem at all with the sudden increase in November. Fancy that.).

To get a better look at what I am discussing in this part and part 3, consider the following graphs, from the BLS report:

This is the same information, broken down by gender as well as age:

First, compare the steepness of the November-January decline with the corresponding steepness of the increase from October to November. With a few exceptions, age group by age group, and comparing by gender as well, the comparisons are mirror images.

Second, notice that teenagers didn't participate in the conundrum at all. In fact their unemployment rate generally increased.

Third, age group 35-44 was relatively unaffected by the conundrum. Their unemployment rate remained much more stable.

Fourth, notice that age group 45-54's experience was similar to that of age 55 and up. That means it isn't simply a matter of early retirement applications for social security (more on that in part 3).

Fifth, notice that age group 20-24 had the steepest decline, followed by age group 25-34. That certainly opens up the idea that the conundrum is explainable by younger workers giving up and moving back home. Paul Krugman put up an excellent post this weekend on the plight of this group, entitled Failure to Launch.

Sixth, notice that the conundrum was much more evident among men than women. There is a large difference in the change in the unemployment rates by sex. In fact, married men almost totally explain the conundrum.

For some reason the Census Bureau's seasonality adjustment seems to have undercounted hires in November, and undercounted fires in January. I think there were two primary drivers: first, the Census Bureau overestimated its seasonal adjustment generally. When holiday hiring didn't live up to the adjustment in November, there was a spike in the unemployment rate. When those workers were let go in January, the seasonal adjustment made the opposite error. It expected a lot more fires than actually happened. Further, if the age and gender of holiday workers was different this season than the seasonal adjustment anticipated, it would undercount that group in one month, and overcount in another. Remember, this was the first relatively robust holiday hiring season in 3 years. In short, the seasonal adjustment may have been fooled in November into thinking that some categories of workers for some reason weren't hired for holiday season jobs in 2010 - and then didn't get fired in January.

3. Aging Boomers

[I wish to thank Fladem and SilverOz for the analysis that went into drafting this part of the post]

Two subgroups where there isn't a mirror image between November and January is in age groups 45-54 and 55-64. In fact age group 55-64 is the place where the participation rate is falling most steeply. It appears that aging boomers, especially men, are putting in disability claims if they have a case, and are filing for early retirement at an accelerated rate. At least some of these people may have been on unemployment benefits and decided to switch over to disability or retirement if prospects of returning to the work force before expiration of their benefits were grim.

The Social Security Administration publishes statistics for annual social security disability and retirement claims. These statistics show thatin this recession as in prior ones there is an increase in such claims, that abates when the economy improves. Comparing the numbers from the mid-90's to now shows a dramatic increase in such recipients since the onset of the "Great Recession." There were 1.5 million more beneficiaries in 2010 than in 2009.

Here are the averages for SS disability and retirement for recent years:
2005 277,497
2006 291,598
2007 331,582
2008 546,029
2009 1,020,929
2010 867,978

Prediction if growth rate from 2005 to 2007 had held for 2008 to 2010:
2008 335,213
2009 340,780
2010 351,888

Difference: 1,407,978

Of this, only 2/3's of the increase is retirees. The remaining 1/3 is a dramatic increase in filings for SS disability.

In summary, there are double the amount of SS disability recipients now vs. less than 20 years ago. That can't be explained by population increase, it has to be either looser standards or filings by boomers whose bodies have been beaten up by blue collar labor. That appears to also explain some of the 45 and up increase in people leaving the labor force (and not just in the 55 and up metric, as noted in part 2 above), and also explain why it seems to be disproportionately men.

On the similar note, over the weekend Calculated Risk passed on a research note by Sven Jari Stehn at Goldman Sachs, and concluded that
the key point is most of the recent decline in the participation rate is due to demographics and not because of cyclical effects - although there will probably be some small bounce back of the next couple of years.
Whether aging boomers are filing in part *because* of poor prospects for re-entering the work force, or simply because they have reached the age where they can, the simple fact is that aging boomers (and secondarily 20-somethings who have "failed to launch")- along with the annual population adjustment and unusual seasonal effects - explain the unemployment rate conundrum in the January jobs report.