Friday, August 03, 2007
Interesting Employment Release
The headline numbers are 4.6% unemployment (an uptick of .1%), with the establishment survey showing 92,000 new jobs created. That differs from ADP's much more pessimistic numbers released a few days ago. Government employment showed a surprising loss of 28,000 jobs, so the job-generation number is better than it seems in the establishment survey. Retail only lost 1,000 jobs in this survey, which is a definite improvement on recent numbers and probably real due to the increasing concentration of retail in large establishments.
But looking at the household survey ought to raise a few eyebrows. According to the household survey, the total number of employed persons dropped by 30,000 from June to July. Also notable is that first quarter employment average is given as 146,044 while the second quarter employment average came in as 145,956, for a total quarter to quarter drop of 88,000 jobs. Given the household survey's somewhat lax criteria for unemployment (a person who worked at all, even if unpaid, during the survey week is counted as employed), that's a rather serious number. It does, however, conform to the weekly unemployment report, which has been running consistently at about 90,000 more continuing claims in July than last year.
It's important for everyone, including government economists, to understand that the big losses in RE-related employment mostly do not show in the insured employment statistics. The theory is that you can pick them up in the household survey, but of course large losses in illegal alien employment aren't picked up there either. Most of the recent immigrant population doesn't have land lines, and the residential pattern is crowded to say the least. Nor are you likely to find anyone willing or able to answer a bunch of questions about who is working where if you do call a land line.
You would expect to have at least losses of 400,000 in effective construction employment before it started to show inofficial statistics, and I believe we are well past that number. (The way that this works is that recent illegal immigrants are likely to lose contract employment and be replaced by citizens, residents and other well-established groups who do show in the employment stats. Call it the buddy system.) The converse is that these workers don't necessarily show on the radar when they are originally employed, either. It is quite likely that employment last year was stronger than reported, and that employment this year is weaker, but no one can truly know whether that is true or not. I assume that it is, because of physical sampling and some payroll card stats I got from friends.
Table A-5 shows that YoY, persons employed in non-farm part-time jobs for economic reasons increased from 4,349 to 4,456. These are the NSA numbers, which are best fro YoY comparisons. The SA numbers showed that this number increased from 4,194 in June to 4,240 in July.
Table A-9 gives a breakdown of duration of unemployment. Using the SA numbers, a clear pattern of increased duration of employment is emerging. It's an early trend, but real. Workers unemployed for 15 weeks and over increased from 33.2% in March to 34.0% in July, and median duration of unemployment has risen from 8.5 in March to 8.9 in July.
Table A-13 gives multiple jobholders, which shows a rise YoY. This could easily be a function of economic stress, however, because the largest increase is in the number of persons working full-time and a second part-time job.
Table A-11 gives a YoY comparison of unemployment rates by industry. Most notable here are the increase in leisure and hospitality (6.8 > 7.3) and transportation & utilities (4.2 > 5.1). Because of the smaller category samples, you need a larger change to be able to comfortably assume that the change is real. Both of these numbers make sense given the economic patterns we are seeing, however, and both should be statistically significant.
In about six months we will get a revision of the models used to generate many of these figures, and based on early sampling (the BED survey), we will see some changes. They may be highly significant, or they may not. Still, the overall trends have slowly shifted to show a moderate trend in increased unemployment and early indicators of employment stress.
What is not picked up in the employment surveys are some rather dramatic changes relating to the new economy. It is far more common for persons to work as contractors now than a few decades ago. Because of the way sampling is done, measures of full employment or underemployment for these workers continually runs below the radar for six or more months. The best way to sample for this are from self-employment tax receipts, and there is a large underground economy that doesn't pay these taxes, especially in the handyman/home repair segment. So it is likely that overall, revenues for many of these workers are dropping fast, but we cannot know that with any degree of certainty.
It's very hard to know what to make of this report. I am guessing that the establishment survey, which is generally biased in its sampling to some larger organizations, may now be a bit of an oversample. However, one should not ignore the relative loss of government jobs compared to private employment. We may have to wait seven months to get a better base read on what is happening. Another big question are the effects of demographics. Many older displaced workers may just choose retirement, even early retirement, if they can afford to do so. Even if they can't truly afford to do it, they may have no option if they cannot get new jobs. The most likely outcome would be taking early SS and a part-time job to supplement income, which does seem to be happening in some areas.