The January unemployment report created quite a stir. Many believed the BLS had simply dropped 1,252,000 people out of the labor force, discarded like trash. Is the BLS an evil doer as so many declare, or could the culprit possibly be the 2010 Census?
We already showed how comparisons between December and January cannot be done due to the incorporation of the 2010 Census data and the yearly population controls, benchmarks and seasonal adjustments incorporated into the January unemployment statistics.
While there is no mythical 1.252 million dropping out of the labor force, there are some highly unusual numbers in the BLS population controls.
The BLS starts the January month with revised population estimates, seasonal adjustments and benchmarks. This year the 2010 Census data was also incorporated into the BLS statistics. They do not go backwards in these revisions. The BLS does not backwards adjust December 2011. Here are the BLS population controls for 2012:
To gauge the impact on the labor force data, the Bureau of Labor Statistics uses special tabulations of December data that incorporate the new population controls. When applied to December 2011, the updated controls increased the estimated size of the civilian noninstitutional population 16 years and over by 1,510,000, the civilian labor force by 258,000, employment by 216,000, unemployment by 42,000, and persons not in the labor force by 1,252,000.
Below is the table from the breakdown of population adjustments added to the end of the year for age brackets:
|Civilian noninstitutional population||1,510|
|Civilian labor force||258|
|Not in labor force||1,252|
|16 to 24 years|
|Civilian noninstitutional population||521|
|Civilian labor force||101|
|Not in labor force||420|
|25 to 54 years|
|Civilian noninstitutional population||-299|
|Civilian labor force||-408|
|Not in labor force||109|
|55 years and over|
|Civilian noninstitutional population||1,288|
|Civilian labor force||565|
|Not in labor force||723|
The first thing to notice is how only 17.1%, or 258,000 of the 1,510,000 additional people discovered from the 2010 Census adjustments were actually added to the civilian labor force. The civilian labor force of those people either considered employed or officially unemployed. That's fairly shocking considering the actual ratio of civilian labor force to non-institutional population is 63.7%.
The next thing to notice is how the civilian non-institutional population of people between the ages of 25-54 actually dropped. That too is shocking, although the last of the baby boomers are now 50 years old. Plus we have illegal immigrants counted into the Census and supposedly border hopping is at a 40 year low.
Finally the biggest shocker is the increase to the non-institutional population of people age 55 and older by 1,288,000 with the incorporation of the 2010 Census population adjustments. The civilian labor force of those 55 or older also increased by 565,000 and gives a labor participation rate of 43.9% for just the population controls adjustment data alone. The 55 or older group also added 723,000 to the not in the labor force population group. Civilian non-institutional population means this group of people are not in nursing homes, in hospitals or otherwise incapacitated. It does include the retired.
We drilled down further into the December to January difference in civilian non-institutional population, which includes not just the above population controls adjustments but also monthly growth rates, not seasonally adjusted. That difference was 1,685,000. We did this because the actual population controls age bracket breakdown isn't available.
For people aged 55 through age 64 the December to January civilian non-institutional population change was 768,000. Yet people aged 65 or older of the civilian non-institutional population increased 721,000 from December to January or 42.8% of the total increase.
In other words, one of the reasons the not in the labor force increase is legitimate, was a disproportionate increase in population of those 65 years and older. Blame the boomers.
Additionally the number of people age 55 to 64 years who entered not in the labor force from December to January was 300,000 or 39% of the civilian non-institutional population increase for this age group, which also includes the population controls. Yet the January labor participation rate for ages 55 to 64 is 63.9%, not seasonally adjusted. This implies a 61% labor participation rate for the 55-64 age bracket in the 2010 Census data population control adjustments only. 685,000 of the 721,000 December to January increase in age 65 or older civilian non-institutional population was categorized as not in the labor force, as expected.
The labor participation rate for those 65 years of age and older is 18%. How could possibly the labor participation rate for just the population adjustments be 17.1% then? The answer is due to the sharp decline in employed for ages 25-54. The labor participation rate for just the population controls was -137%*. The non-institutional population declined but those working declined even more and those no longer counted in the labor force increased as well. Without this age group the population controls give a labor participation rate of 36.8%. Considering those over the age of 55 are 71.2% of that rate, and the January, not seasonally adjusted labor participation rate for those 55 and older is 40.1%, we have to assume the population controls of the age 55 and up bracket published must be a majority of people over the age of 65.
That said, don't let anyone claim people are not dropping out of the labor force due to not being able to find a job and aren't being counted any longer. You cannot blame an aging population or education on the below declining labor participation rates for those between 25 and 54 years of age.
The ages 16-24 population control adjustments are also shocking. Almost all of the additional civilian non-institutional population found via the 2010 Census adjustment went into not in the labor force, or 80.6%. This gives a 19.4% labor participation rate just for the population controls adjustment. The labor participation rate for people ages 16-19 is 33.4% (30.8% not seasonally adjusted). For ages 20-24 it's 71.1% (unadjusted for January it was 69.8%). The labor participation rate for 16 to 17 year olds is 21.2% (18.3% not seasonally adjusted). It's hard to believe the 2010 Census simply located a city of juvenile delinquents they must have missed somewhere, in spite of falling high school graduation rates.
So, how did an uber-low labor participation rate happen for ages 16-24 just for the population adjustments as presented? It seems there were a hell a lot more Hispanic and Asian teenagers for the 2010 Census than in 2000. Additionally, there were way more females than males added to the 2010 population control adjustments. Hispanic teenage girls, ages 16-17, have a labor participation rate of 12.6%, not seasonally adjusted. This pulled down the labor participation rates for just the 2010 population controls data. Teenage girls between the ages of 16-19 added 239,000 or, 15.8% of the total population controls adjustment.
Additionally women overall have a much lower labor participation rate than men, 57.7% vs. 70.2%, for people 16 years and older who are in the civilian non-institutional population. The population control data shows -116,000 men were subtracted from the overall civilian non-institutional population while women added 1,626,000.
Voilà, the seemingly bizarre result from incorporation of 2010 Census data is explained and unfortunately for conspiracy theorists, it all appears to be statistically legitimate. If people still want to blame the government, then it must be the Census. The Census are the ones who update the demographics of overall population, who claim there is a whole tribe of Hispanic teenage girls, way more women in the U.S., and a hell of a lot of retired old people lurking around somewhere than the 2000 Census with population growth adjustments would predict.
The overall difference between the 2000 Census and 2010 Census is:
The 2010 Census reported 308.7 million people in the United States, a 9.7 percent increase from the Census 2000 population of 281.4 million.
All data for this article was obtained from the BLS and the Census.
* This post was updated to correct two typos and also to explain the unbelievable, impossible -137% labor participation rate from the population controls for ages 25-54. Remember, population controls are simply adjustments into the civilian non-institutional population and civilian labor force. For the ages of 25-54, the labor participation rate was dropped a 10th of a percentage point to 81.5% overall due to the population controls adjustments. First, here is the complete table from the BLS on this age bracket. The first column is the December 2011 unemployment statistics, the second is the December 2011 unemployment statistics with the population controls applied backwards to the month (which the BLS does not do) and finally the difference between the two numbers so one can get to the population controls listed in the first table of this article.
|December 2011 Unemployment, 25 to 54 years, with 2012 Population Adjustment|
|As Reported||Population Controls Added||Difference|
|Civilian noninstitutional population||124,690||124,391||-299|
|Civilian labor force||101,732||101,323||-408|
|Labor Participation rate||81.6||81.5||-0.1|
|Not in labor force||22,958||23,067||109|
Now let's just do a little math. Let:
Civilian labor force =
Civilian non-institutional population =
Population controls (difference) =
December civilian non-institutional population before population controls =
December civilian non-institutional population after population controls =
December civilian labor force before population controls =
December civilian labor force after population controls =
......do some Algebra.....
Plug in the above table numbers and you'll see the labor participation rate for just the population controls is -137% for ages 25-54.
Nice work. Thank you for
Nice work. Thank you for your time and effort.
For those who do econometric modeling, we have to have time series. So when I was at Global Insight (WEFA) we had to create our own labor force numbers to smooth out the spikes and put the numbers as best we could into the years they belong. WISH the BLS would do this for us...
No kidding! Thank you for the compliment also. I don't get why the BLS doesn't apply some sort of smoothing filter or algorithm to redistribute the yearly adjustments backwards, monthly proportionate.
It makes no sense to me considering we have benchmarks, SA and birth/death to not go ahead and do this also.
Although if they did I might have less to write about. ;)
Simply great work. not just this post but all the hard work you put in here is as good and mostly far better than anything I used to see while in the fold of the IB world. People should be paying you for this stuff. Keep up the good work.
well, there is a donate button ;)
on the upper right corner. Thank you, no kidding I should be paid, this one was a bitch. The Census website is truly an exercise in "Where's Waldo" and I thought the BLS data layout was horrific, pales in comparison to finding data from the Census resources.
about this 1.252 million "not in the labor force" article update
Folks, if you want to question this, point out a mistake, doubt this exercise for the reader article, please leave a comment. This article is making it's viral way around the Internets, but unless someone posts a comment here, I will not see the criticism, mistake (typos!) or confusion or whatever it is that's an issue.
I ended up going through some algebra (and I'm sorry, any more mathematics than this, I'm going to have to create a full bore white paper!) just to give some validity to at least one of the calculations used to conclude the BLS was aok. The reason I updated the article so much is I was informed of it being blasted elsewhere.
Look, math is math and statistics are statistics, so I'm more than happy to discuss, make corrections or whatever, but I need you to post some comment where I will see it.
Happy Algebra (I was not going to write up in LaTex all of those Algebraic steps, just trust me if you don't follow)! ;)
those interest in BLS seasonal adjustments.
Econbrowser, Chinn, went ahead and ran the X-12-ARIMA algorithm on the NSA time series and graphed it all up. So, those claiming seasonal adjustments are this horrific mystery are basically disproved.
Chinn is hard to read but basically he did as an exercise the "by hand" seasonal adjustments via code you can find easily on the Internets to the BLS not seasonally adjusted data and shows how it's spot on.
It's a lot of work to run the code, get the new data series and graph it all up so you might check it out. Chinn is pretty dog gone good and he's a WI-Madison economics professor.