If you have "no place to go," come here!

BLS Jobs Report Covering January 2012: Not so Smooth Operators

The first thing to say, as I noted last month, is that the Bureau of Labor Statistics (BLS) jobs report covering January 2012 reflects both yearly revisions in the Establishment (business/jobs) survey and the first incorporation of data from the 2010 Census into the Household (people/employment) survey. This means that comparisons with previous months need to be explained and caveated more than usual and in some cases may not be possible.

The most obvious example of this is in the civilian non-institutional population over 16 (the potential labor force which I abbreviate as NIP). For December 2011, this was 240.584 million. The January 2012 number is 242.269 million. This is a difference of 1.685 million. Now the population did not grow this much between December and January. The BLS simply didn't update the 2011 data. If it had, as per the BLS, the December 2011-January 2012 increase would have been 175,000.

Before beginning my own analysis, let me lay out the basics of the BLS report, the stuff that will hit the news. 243,000 jobs were created in January. The November 2011 number was revised upward by 57,000 to 157,000. Unemployment dove another 0.2% to 8.3%.

So what is really going on behind all these changes? For the Household data, we need to keep 3 numbers in mind: the new numbers (seasonally adjusted), what the old numbers would have been if the BLS had officially incorporated the new Census data into them (which the BLS unofficially provides), and the seasonally unadjusted numbers (which can be compared to the seasonally adjusted numbers and less well with older unadjusted data).

Perhaps the best way to see what is really going on is to work down through the numbers largest to smallest. NIP > labor force > employed > unemployed > the BLS undercount.

I have already discussed the NIP. So let's take a look at the labor force, the sum of the employed and those the BLS counts as the unemployed. In January 2012, the seasonally adjusted labor force numbered 154.395 million. In December 2011, it was 153.887 million. This is a difference of 508,000. The BLS attributes 258,000 of this to Census effects and 250,000 to the actual month over month change. But the real story here is in the unadjusted numbers. In December, it was 153.373 million and in January, 153.485 million, a difference of only 112,000. However the difference between the seasonally adjusted and unadjusted numbers is 910,000. Essentially, the BLS is smoothing the numbers because January is the month where the greatest number of seasonal jobs are lost (just as June and July are the months where the greatest number are gained). There are two things to note here. The first is that 910,000 represents a statistical process, not real people, and presupposes that there will be a big pick up in seasonal employment in the summer to compensate for it. The second is that the difference between the seasonally adjusted and unadjusted January 2011 labor force was 714,000 so the BLS increased substantially its smoothing factor this year and I am not sure such an increase was warranted. While the population estimate increased due to the Census, the labor force participation rate was higher in January 2011 than in January 2012.

Along these lines, the labor force participation rate (actual labor force divided by potential labor force or NIP) dropped 0.3% seasonally adjusted from 64.0% to 63.7%. The BLS ascribes all this decline to Census changes. That is the lowest participation rate since May 1983 (unadjusted 63.4%, lowest since February 1984). This is the first of the big red flashing lights to what on the surface is an excellent jobs report. It indicates that one reason unemployment is falling is because the BLS is defining more and more Americans out of the labor force.

Looking next at employment, seasonally adjusted employment rose 847,000 from 140.790 million in December 2011 to 141.637 million in January 2012. The BLS ascribes 216,000 of this to Census effects so 631,000 for the actual month over month change. But this "actual" change is all a smoothing effect. The unadjusted numbers show a 737,000 decrease from 140.681 million in December 2011 to 139.944 million in January 2012. Aren't numbers wonderful? This is how the BLS can tell us that employment is improving at the very same time that ordinary Americans see the opposite.

We see the same effect, of course, in the employment-population ratio (employed divided by NIP). Seasonally adjusted this rate was unchanged December to January at 58.5%. Unadjusted it dropped 0.7% from 58.5% in December to 57.8% in January.

Seasonally adjusted unemployment decreased 339,000 from 13.097 million in December 2011 to 12.758 million in January 2012. The BLS ascribes an increase in unemployment of 42,000 to Census effects and a 381,000 decrease to the month over month change, another smoothing effect. The unadjusted numbers are again the reverse. Unadjusted unemployment grew 849,000 going from 12.692 million to 13.541 million.

And as we would expect, the same trend occurs in the U-3 unemployment rate. Seasonally adjusted, the unemployment rate decreased 0.2% from 8.5% to 8.3% while unadjusted it jumped 0.5% from 8.3% to 8.8%.

What I take to be the BLS' own measure of its undercount is its category: Not in labor force, Want a job now. Again more inversion of the adjusted and unadjusted numbers. Seasonally adjusted December to January, it decreased 66,000 from 6.385 million to 6.319 million. Unadjusted, it increased 360,000 from 6.135 million to 6.495 million.

Turning to the BLS' broader measure of un- and under employment, the U-6, seasonally adjusted decreased 0.1% December to January from 15.2% to 15.1%. Unadjusted it spiked a full percentage point from 15.2% to 16.2%.

The U-6 has three components: the U-3 unemployed, involuntary part time workers, and the unadjusted marginally attached. The marginally attached are a subset of the BLS Not in labor force Want a job now category.

Let's look for a moment at involuntary part time workers, that is those workers who work part time but would work full time if they could find it. Seasonally adjusted these increased 132,000 December-January from 8.098 million to 8.230 million. Unadjusted, they increased 490,000 from 8.428 million to 8.918 million.

The marginally attached were 2.809 million in January. This was an increase of 269,000 from December's 2.809 million.

Summing the components that form the U-6 rate gives 23.797 million, the number that the rate represents.

Now if you have read my previous job posts, you know I calculate alternate measures taking into account my own estimate of the BLS undercount. I begin by calculating what we would expect the labor force to be if the economy was in a good expansion: .67(NIP) or 162.320 million.

The difference between this and the BLS' labor force for January (154.395 million) is my measure of the undercount or 7.925 million. This is a substantial (621,000) increase from my number for December 2011 of 7.304 million. I would attribute most of this to the new Census figure for the NIP. This is a major difference from the BLS unadjusted Not in labor force Want a job now category of 6.495 million. The reason I started calculating an alternative to the BLS number is because I noticed the BLS number did not track well with changes in the economy, such as the December 2007 recession. My measure of the undercount and theirs have been diverging for months.

Once we have the undercount number though we can calculate other things, such as the real unemployed and unemployment rate. We do this by adding the undercount to the U-3. Because of the difference between the seasonally adjusted and unadjusted U-3, I will give both for the real unemployment rate. Seasonally adjusted, we have 20.683 million unemployed and a real unemployment rate of 12.7%. Unadjusted, we have 21.466 million and a real unemployment rate of 13.2%

Now if we add in the involuntary unemployed, we come up with disemployment, the analog to the U-6, and we can look at both seasonally adjusted and unadjusted numbers and rates. Seasonally adjusted there are 28.913 million disemployed and a disemployment rate of 17.8%. Compare this to the seasonally adjusted U-6 of 15.1% and the 23.797 million it represents. Also for comparison although using different bases, my calculation of disemployment last month was 28.499 million and a 17.7% rate.

Unadjusted, there are 30.384 million disemployed currently and a disemployment rate of 18.7%.

The seasonal smoothing I noted continued in unemployment by race and age. The unemployment among whites was 7.4%, a decrease of 0.1%; African Americans 13.6%, a decrease of 2.2%; and African Americans 16 to 19 year olds 38.5%, a 3.6% improvement.

Finally, using the 175,000 month over month population increase I cited early on in this post and an employment-population ratio of 58.5%, the economy needed 102,000 in January to keep up with population growth.

Turning now to the Establishment data, it did considerably better than that. The private sector created 257,000 jobs, government lost 14,000, netting 243,000 jobs gained. What struck me though was that most of the jobs created were poorly paying ones. Professional and business services added 70,000 jobs but 33,000 of these were in employment services. 44,000 were gained in leisure and hospitality. 31,000 in healthcare. Growth in wholesale and retail trade was anemic at 14,000 and 10,500, respectively. If we were looking at the beginnings of a consumer led recovery, we would expect to see it here, and we are not. Manufacturing added 50,000. 44,000 of these were in durable goods. The average work week in durable goods manufacturing increased 0.3 hours to 41.3 hours and overtime 0.1 hour to 3.4 hours which is good but average hourly wages remained unchanged December-January in this sector at $25.30.

Overall, the average private hourly work week remained unchanged at 34.5 hours and average hourly earnings increased 0.17% or 4 cents to $23.29 (a 2% annual rate). In 2011, wages increased by 2%, the CPI by 3% so workers fell behind. The CPI is not released until later this month and we will know then if this trend is continuing. What I can say though is that workers are not making gains even on a very low inflation rate.

So what is the take home in all this. In the Household survey (employment/unemployment), we have a lot of revisions clouding the data but all of the positive numbers are the result of seasonal smoothing. This would be defensible if we could depend on this year being like other years, but we live in interesting times. Smoothing the data to fit an assumed annual cycle doesn't create jobs now. It simply assumes that the jobs deficit now will be made good later in the year and that this will all balance out.

In this regard the 243,000 jobs from the Establishment survey are more problematic for the smooth operators at the BLS than they are for me. You see their unadjusted Household numbers aren't showing it. And even when we look at the Establishment data, the jobs being created aren't good jobs and in terms of wages, workers are making no headway and are even falling behind. Try spinning a recovery out of that.

No votes yet


Submitted by Hugh on

Apologies to all for the lateness of this post. It took me most of the day to wade through the report.

nihil obstet's picture
Submitted by nihil obstet on

for spending your day on this insightful analysis for us.

Submitted by jawbone on

squirrely due to change from using 2010 census numbers, I figured you would be doing yeoman's work today...or into the night.

Time will tell about the smooth operators at BLS and their smoothing, soothing numbers. But...if people get the rah-rah and keep seeing no improvement, I figure that will really, really hurt Obama.

Better to be up front, BO. But that's not how he rolls.

(I really appreciate having access to your work, an analysis that I can trust.)

Submitted by lambert on

I know this is a n00b question--

You've uncooked the BLS books, and you mention that your figures and the BLS figures have been diverging for months.

That's very interesting. Are there any proxy measures we (and by we I think I mean you ;-) could use as proxies to confirm your metrics from a second source.

For example, if every American had to purchase a pair of new shoes when they got a job, those figures would track new hires. (This is like Greenspan soaking in his tub and thinking about ammonia shipments.) So shoes would be a proxy...

Also, what I would really like is what the people who are kicked out of the labor force are doing. IIRC the figures are unevenly distributed: West Coast and South Atlantic. System D?

Submitted by Hugh on

No proxies that I know of. There are a couple of things to keep in mind. The BLS category Not in Labor Force Want a Job Now is an extension of its focus from job seeking to job wanting. Both are in a lot of ways psychological in nature. Are applicants fired up or motivated enough to actively seek work (the unemployed) or would they accept a job if it was offered to them (Not in Labor Force Want a Job Now)?

Here is what this category looks like from the year before the recession to now.

Sorry that it's not centered. Anyway it's pretty cyclical in nature with spikes in spring and at Christmas. There is an underlying trend upwards but it is fairly linear. Now consider what the U-3 unemployment rate has been doing during this period.

What is interesting about this graph is that unemployment shoots up as the recession takes hold in 2008, plateaus in late 2009 and most of 2010, and then begins to gradually fall off. We know that falloff is not due to job creation but erosion of the labor force has more and more people are defined by the BLS out of the labor force. But we aren't seeing anything like a corresponding spike in that nice linear rise in the Not in Labor Force Want a Job Now graph.

My stance is forget the psychology. Look at the data from the fairly recent past. From October 1996 to June 2000 (with one exception in February 1997: 66.9%), that's 45 months, the labor force participation rate was 67% or higher, topping out at 67.3% in January-April 2000. My take is if the jobs are there, how many people will join the labor force? The answer from our last experience with good employment was at least 67%. So that's the number I use. I can't show you a graph like these others because this is my calculation not the BLS'. But the erosion we see of unemployment out of the labor force which we see in that second graph is essentially the divergence which I am talking about between the BLS data and mine.

Submitted by lambert on

... isn't working. I centered and changed the widths, but I just checked the URLs and they don't work. Sorry if I banged something.

To bad about the proxies. When I said "confirm" I was sloppy. I wanted the tell for a talking point (not questioning your interpretation, just looking for an easy way to explain it).

Submitted by Hugh on

That was the first time I tried to export a graph I had made at the BLS site. Sorry, that it didn't work. I wish I could just take a picture and then paste the picture, no URLs involved.

Anyway, this is the link for the BLS A- or Household data tables:

If you click on A-1, you'll go to a long list that you can check off. The U-3 is near the top and the Not in Labor Force Want a Job Now is at the bottom. It is also in A-16. As long as I am saying this, the U-6 is in A-15, part time workers is in A-8, and workers by age and race is in A-2. These are the principal tables I look at when I analyze Household data.

So if you are at the A-1 table list, you can check unemployed or unemployment rate seasonally adjusted and then hit retrieve at the bottom of the list. This will take you to the last ten years of the data you requested.

Once there you can select a range of years at the top of the page and hit graph just below this and go. This will graph the data for you and produce the table containing it.

The tell I suppose is that the BLS has highly specific and restrictive definitions for unemployed and if you don't fit that definition you are defined out of the labor force. Most people would think that unemployed means unemployed, that you don't have a paying job but that you would take one if one came along. The Clinton expansion showed that more Americans would work if the jobs were there then is measured by either the unemployed or the Not in Labor Force Want a Job Now as measuered by the BLS.

DCblogger's picture
Submitted by DCblogger on

directly I heard the report, I was waiting to see what you would say about it.

Submitted by lambert on


letsgetitdone's picture
Submitted by letsgetitdone on

Whenever a report comes out, I never believe it. I always wait for your obviously more truthful analysis to evaluate what really happened. In this particular case I've been amply rewarded for my patience by your analysis.

Submitted by lambert on


Submitted by Hugh on

That's a nice snap shot of the participation rate but Ed buys too much into that people being forced out of the labor force. They are being forced out of jobs. It is the BLS that is defining them out of the labor force. The BLS' definitions of unemployed vs. not in labor force do not capture what is happening now and, conveniently from a political point of view, understate the severity of unemployment and disemployment.

The participation rate is useful because while the BLS can massage its jobs and labor force analysis, the participation rate compares the manipulated number of the labor force to the working age population (the non-institutional population over 16), what I call the NIP. Now the NIP is a much larger number. It has some caveats to it but being so large it can't really be manipulated by the BLS. It comes from the Census anyway.

So while the BLS can tell us that unemployment is declining, we can compare this to a number the BLS doesn't control, the working age population, and see through the participation rate that this statement doesn't correspond to the larger reality: fewer Americans have jobs. Put more simply, the BLS is understating real unemployment.

lookin206's picture
Submitted by lookin206 on

I'm always glad to see you reworking of the numbers.

I often look at Gallup's daily numbers. Today they have unemployed at 8.7, underemployed at 18.7 and participation at 64, numbers much closer to yours than the gov's. Do you have any information on how Gallup tracks these numbers?

Like others, i look forward to this monthly post of must take a bit of time to work through the data and write it up.

Submitted by Hugh on

I was meaning to get back to you on this. At the bottom of this article, Gallup lists differences between its survey and that of the BLS.

It looks like Gallup uses definitions similar to those the BLS uses, except I think for legal reasons, it polls 18 year olds and older whereas the BLS Household survey covers those 16 and up.

If you want to see their questions and definitions, go here:

and click on the "read more" link just under their comparison of their methodology to that of the BLS (same comparison as in previous link above).

Submitted by Alcuin on

Like many others, I was astonished when I read the news - it sure didn't reflect what I see going on around me in the work place! Thanks for the detailed analysis, Hugh. I wonder if all the speculators on Wall Street had access to the reports' conclusions before the opening bell on Friday. Maybe the congresscritters shared their executive summaries with their favored speculators?