BLS Jobs Report, January 2011: What a mess!
[Warning: If you thought my previous job posts were dense, then my advice is run from this one. It involves many changes that the BLS has made and a significant correction in my own calculations.]
Otherwise jump in.
The big news in today's job report that will be hyped in this report is that unemployment fell 0.4% from 9.4% to 9.0%. The bad news is that this decrease is meaningless. It is the result of the BLS playing with their numbers for a second straight month, not people finding jobs. Although it comes from a different data series, the weak 36,000 growth in jobs reflects this fact.
In the last two months, the BLS has revised down the U-3 unemployment rate by 0.8%. None of this decrease has come from anyone who was unemployed finding a job. This all happened on paper. Now 0.8% may not seem like much to you, but the US labor force is 150+ million, and this 0.8% represents an error of 1.2 million. Another way of looking at this is the last two months changes in the U-3 rate have reduced that rate by 8.2%. On top of this, these changes are using data from the 2000 Census for their baselines, that is ten year old material.
The question we have to ask ourselves is is such a large error acceptable? And if this is the best the BLS can do, what does this have to say about how we use and view its numbers? Each month we hang upon and are spun stories about whether unemployment has increased or decreased by a tenth of a percent, but the BLS is telling us with these revisions that the unemployment could be off by around a percentage point.
This is so even though the BLS revises its initial figures for a month in the two following months (for example, the numbers for March in the April report will be revised in the May and June reports). Then by stretching its year end revisions over two months, this means you can't really make month to month comparisons for November-December and December-January. As I noted last month, revising December figures screws up evaluating year long (January-December) trends because the December number is not the same as the others.
And finally, as we all know, the U-3 unemployment rate is a woefully inadequate indicator of what workers are experiencing in the labor force. For that, we have to look beyond to the U-6 measure of un- and under employment, and beyond that we need to calculate on our own because the BLS doesn't the disemployment rate (the U-6 plus the BLS undercount of the unemployed). Additionally, except with under employment to some extent, none of this touches on job quality, something the BLS doesn't really measure at all.
I find this all both obfuscating and frustrating. With so much flux and uncertainty in its data, at a minimum, the BLS needs to include what its confidence intervals are in its reports. And it needs to be more up front about the problems its models are having in the extreme conditions of the current economic environment. My impression is that most of the instability in BLS numbers comes from models poorly adapted to what we are experiencing.
What I do not appreciate is "The unemployment rate fell by 0.4 percentage point to 9.0 percent in January" which is how this month's report begins. The unemployment rate did not fall. It was changed, and in such a way as to make impossible any direct comparison with the previous figure.
With that long preamble done, I will now look at the January numbers as if they actually mean something. Any December-January comparisons I mention should not be taken as changes in the real economy but rather how the BLS has changed its numbers.
The U-3 (unemployment rate) was reduced, as I said, from 9.4% to 9.0%. The[no-glossary][/no-glossary] 9.0% represents 13.863 million Americans. This is the seasonally adjusted number. Interestingly, the unadjusted number is 14.937 million (a difference of 1.074 million) and corresponds to a 9.8% unemployment rate.
The U-6 rate for un- and under employment was changed from 17.0% to 16.1%. The U-6 is defined as the unemployed plus the marginally attached plus part time employed for economic reasons (i.e. not their own choice) divided by the labor force (employed and unemployed) plus the marginally attached. The January number for the marginally attached was 2.8 million. This is a little tricky because this number is not seasonally adjusted whereas the others are. Also the marginally attached are not counted as in the labor force. The labor force, according to the BLS, was 153.186 million. So we have .161(153.186 + 2.800) = .161(155.996) = 25.115 million. This is the level of un- and under employment.
Both the non-institutional population over 16 (the potential labor force or NIP) and the labor force (employed plus unemployed) numbers were decreased. The NIP numbers went from 238.889 million in December to 238.704 million in January, a reduction of 185,000. The labor force numbers went from 153.690 million in December to 153.186 million, a decline of 504,000. Because the labor force was reduced considerably more than the NIP, the labor force participation rate which is the ratio of these two declined, from 64.3% to 64.2%.
The BLS seasonally unadjusted number for those not in the labor force but wanting a job, i.e. an estimate of their own undercount was 6.643 million. This includes 2.8 million of the marginally attached. My own calculations in the past did not reflect that the BLS does not count the marginally attached in its labor force number. Without the marginally attached, the BLS undercount is 3.843 million.
My own calculation of the BLS undercount is to take a 67% labor force participation rate indicative of a strong expansion and multiply it by the NIP. 0.67(238.704 million) = 159.932 million. Subtracting the labor force from this (159.932 million – 153.186) = 6.746 million. This is my calculation of the undercount. Subtracting out the 2.8 million of the marginally attached which are included in the U-6 leaves 3.946 million.
We take disemployment to be the sum of un- and under employment (U-6) plus the BLS undercount. So if we use the BLS estimate of the undercount (minus the 2.8 million counted back into the U-6), we get (25.115 million + 3.843 million) = 28.958 million disemployed.
If we take my estimate, we get (25.115 million + 3.946 million) = 29.061 million disemployed. This yields a disemployment rate of (29.061 million / 159.932 million) = 18.2%.
This number is lower than my previous estimates which hovered closer to 20%. Most of this change results from my understanding that the BLS does not count the marginally attached as part of the labor force but does count them in the U-6, a situation which since the BLS does not break down its numbers in its tables was unclear to me until now.
I can't calculate from the data how many jobs needed to be created in January to keep up with population growth. I would estimate 95,000. December-March are the lowest months of the year in terms of population growth and with more and more people leaving or not counted in the labor force, this number has declined significantly in the last few years, by about 25,000 a month. In any case, the 36,000 jobs net gained in January were well below this figure.
The take home from all of this is that the BLS needs to do a much better job, in its initial data collection, in its modeling, in the presentation and explanation of its data, and in its revising. As it is, politicians can say we've turned a corner when, in fact, nothing has changed