Monday, October 08, 2012
I Figured The Employment Report Out
The results of the employment report on Friday were quite schizophrenic, and I do understand why so many assumed they were the result of a BLS conspiracy. However it was not the BLS, but a confluence of factors.
Before I get into the main point of the post, I want to address Neil's quarrel with the idea that the unemployment rate has to drop. Neil is wrong, because:
This is a graph of the primary ratios produced in the Household survey. They are civilian participation in the workforce and the employment/population ratio.
The employment/population ratio has refused to move in this recovery, and that is historically aberrant. It is also the source of Snarky Mark's dire employment trend graphs. So we don't have jobs.But we do have a falling unemployment rate, because of the collapse of the civilian participation rate, due to slack employment, high travel costs, and low wages (it doesn't pay some people to work), plus retirements/disability (which are solely due to demographic effects, but exaggerated by the poor economy).
Fiscally this is a disaster, that is true. But nonetheless, unemployment rates will keep falling. Unfortunately, a falling unemployment rate no longer means that the economy is improving. One of the big factors controlling retirements is Medicare eligibility, and if you were born in 1946 you turned 65 last year, and if you were born in 1947 you turned 65 this year. So there had to be a flood of retirements at this time, and there is.
Now on to the main issue. This employment reports showed a huge gain in jobs from the Household Survey, and a very lackluster gain in the Establishment Survey. It is quite hard to figure out how you can report a real gain of 873K jobs in one survey (historically utterly massive) and a 114K gain of non-farm jobs in another survey. Further, there was no trend change shown for private jobs in the second survey, and many indicators are decisively bad. This could only occur if suddenly there was a huge explosion in new company startups in the private sector. I find this implausible - it is not supported by any other data, including gas consumption, which is pretty strong proof to the contrary. A four-week -2.5% YoY says that it didn't happen. Also temporary employment stalled in the prior month and dropped a little in this month. That also is strong proof that the economy just didn't start generating a lot of jobs.
However it does not seem to have been a BLS conspiracy. Let us at least concede to the BLS workers the intelligence to realize that they would have to FUBAR both surveys if they were to successfully conspire to get the president reelected. Nor is it that unprecedented - during times of slackness or rapid change in the economy the Household survey does show these swings:
In thousands of persons, the red line is the month-to-month change reported on the Establishment survey; the blue line is the month-to-month change reported on the Household survey. Remember that the red line is extensively revised after the fact, so that initial results wouldn't look quite like this graph. Over time, the general curves do correspond. In any given month they can be quite disparate.
You tend to get a whole lot of spikiness in the blue line either coming into a downturn or coming out of it. Over time this effect seems to have increased.
Instead of slandering BLS statisticians, it is logical to look at the two different surveys for the source of the difference. The Household survey is generated by using a Census sample which is supposed to be representative of the general population. Therefore, you can get flyers on this survey if there is a very strong effect on some segment of the population. But it cannot be that much of a "real" effect, because we know from the Establishment survey that not that much employment action was going on.
After considerable concentration on the Household survey stats, I came up with the Hispanic effect combined with the SA education effect. Here's how this goes: earlier this year, the Obama administration made the executive decision to grant work permits to a huge batch of young immigrants. Obviously this takes time to make its way through the system, but it did. Economists did think it would have an effect on unemployment rates. It did.
Here's some suggestive info. First, there was a very large and very huge pop in SA young adult employment in September:
The little blue pop in rates turns into 400K by the magic of SA. There's nothing wrong with the SA, but this is historically odd to say the least. Now some of this ought to have been young teachers/school emps replacing older teachers/school emps who are retiring in droves in some places. Despite all the yakyak over government jobs, in September government jobs racked up a 10K increase, concentrated in state education but also with a minor component in local ed. Due to oddities in school calendars, it does seem like some of the normal spread in school gains shifted from August to September this year.
But there's another piece of the pie which does seem to be related to a change in status. That's harder to pin down, but one clue is that you have this huge pop in 20-24 employment. That is the age level where one would expect newly-legalized Hispanics to show up.
This graph shows some apparently unrelated stats that tend to suggest a high correlation. The thin red line is the 20-24 employment level. The thin blue line is the African-American men level. Note that decline. The heavy yellow/umber line is self-employment, which shows a very interesting pattern. It rose very strongly earlier, and then stalled out. The green line is total Hispanic employment adjusted by dividing it by 1.8 to get it inline with the other demographic segments.
There is and has been a replacement effect going on between a group of black males and Hispanics. Further, what seems to be a mysterious trend in self-employed workers is probably not. There has been an odd gap in reported construction employment over the summer, and it is beginning to unbend. We are probably seeing the effect of more legal workers shifting onto the books - before they were probably being paid as contractors.
Lastly, the Household survey never asks about legal status and is supposed to be immune to that variable. That is not true. I know this for a fact, because I use a lot of Census data at the county level, and I've seen very big disparities between reported employment and actual employment among Hispanics. Very clearly a lot of people are working illegally and lying about it. The gray economy is pretty large, and as numbers of persons who can't work legally rise, you have to expect it to get larger.
What I believe happened is that first a number of newly-legal persons did pick up part-time jobs in retail, restaurants and seasonally-associated jobs such as warehousing. Before they weren't eligible to go on the books and now they are. Then there was a reporting effect, with more people reporting that they were working because they were now doing so legally. Finally, the seasonal adjustment happened to pick this up and magnify it.
I can't quantify the effect. Because jobs didn't show up in the August survey, I expected a big pick-up in the September survey. The official Household data is that we lost over 100K SA jobs from July to August. That didn't really happen. So I was expecting around 300K jobs to pop in there.
How are we really doing? Well, if you look at Table A-8, which splits Household survey data out in another way, you see something really interesting. Private industry SA wage and salary workers are below the numbers reported in May and June! Government workers are up over 500K since May. We have somehow mislaid around 400K private workers since the spring.
So you ardent critics who commented that these numbers weren't true based on what you could see are not wrong. The private employment picture is not improving - it is getting worse. Government employment is growing, and it is growing in relation to private employment. This brings me back to my appalled disgust at our president's spring gaffe claiming that the private economy was doing fine. It isn't.
Your skepticism is justified. Blaming it on a BLS conspiracy is not.
On the other hand the loud shouts of joy coming from the press are total BS.
Ironies do abound, but they are so painful that one cannot savor them.
The employment/population ratio has refused to move in this recovery, and that is historically aberrant. It is also the source of Snarky Mark's dire employment trend graphs. So we don't have jobs.But we do have a falling unemployment rate, because of the collapse of the civilian participation rate, due to slack employment, high travel costs, and low wages (it doesn't pay some people to work), plus retirements/disability (which are solely due to demographic effects, but exaggerated by the poor economy.
My girlfriend has a part time job 13 miles from our house. Using the IRS cost per mile rate of 55.5 cents, she spends roughly 1.5 hours of her pay getting to and from work.
Last week she was scheduled to work 25 hours. They only wanted her to work 3. She netted 1.5 hours of actual work for the week. Unfortunately, this is a trend that is becoming easier and easier to predict. Sigh.
Technically speaking, she's not unemployed though. Go figure.
As for BLS conspiracy theories, there have been 64 straight revisions to the upside in the initial claims.
I do not and have not believed that it was a grand conspiracy though (and have pointed out my beliefs in the comments). Instead, I suspect that there is a structural problem with the assumptions used in their estimates that leads them to be consistently off in the same direction. It pains me when I see my data used as an implied proof that there is a conspiracy (and it has). I think it is very reasonable to be suspicious, but it is another thing entirely to say it is proof. It most certainly is not.
I think the following sums up where our country is headed. I don't claim it is proof though.
NFL: Kansas City fans cheer as Chiefs quarterback Matt Cassel lay injured
I sure hope it isn't. Sigh.
So the answer is: "OK Mr. Smarty-Pants Nitpicker, they don't necessarily have to retire but in fact, they are."
Pesky facts, always getting in the way of a good theory.
I still suspect that if hiring really improves coming out of the next trough, the unemployment rate will be sticky due to Boomers taking semi-retirement jobs.
Thanks very much for the info.
Every month I compare the establishment non-farm private jobs non-seasonally adjusted to the same month last year. The best picture was in February and the number of jobs have declined by 351,000 since then.
To see the seasonal pattern and how I simply compare year over year:
My blog is not updated except to show examples like this.
Links to this post: