Sunday, September 06, 2009
US August Employment, More Comprehensive
So here goes:
1) The establishment survey is quite unreliable at this point. You have to use the household surveys. Mind you, the establishment survey will still be unreliable when things finally start to improve. There are two reasons why the establishment figures can be (temporarily) so far off during times of rapid economic change. The first is that the basic methodology of the Birth/Death assessment is always about 3/4s of a year behind. To assess the relationship of business expansions and new businesses as a ratio of business closures and contractions, BLS uses the Business Employment Dynamics survey. Here is the most recent, and it covers the last quarter of 2008. Note the sudden pop in job losses. BLS revises the establishment survey for a very long time. Even when business conditions haven't shifted that much, the first two months of the establishment survey are based on partial returns (see technical note), and the rest is imputed. The establishment data gets much better the third month, and then after about a year it gets considerably better. The second reason that the establishment survey can be off is that if a person is working two part-time jobs, the person will get counted twice. At this stage, many persons are doing just that, so there should be an upward bias in the establishment survey.
2) The household survey is a very good survey, but it too becomes far more reliable over two months. Only gross changes in monthly employment (over 430,000) pass the 90% confidence level. Here is an explanation from the technical note:
Suppose the estimate of total employment increases by 100,000 from one month to the next. The 90-percent confidence interval on the monthly change would range from -330,000 to 530,000 (100,000 +/- 430,000).These figures do not mean that the sample results are off by these magnitudes, but rather that there is about a 90-percent chance that the "true" over-the-month change lies within this interval. Since this range includes values of less than zero, we could not say with confidence that employment had, in fact, increased.If you take rolling two month summaries for any period during which the employment numbers change by less than 350,000, you will have a much better feel for what the household survey is conveying. Smaller samples (such as unemployed totals) have an even wider confidence range. So it is better to assess short-term trends by what is happening with overall employment rather than looking at the unemployment rates.
3) It is well to read the monthly report in combination with the weekly insured unemployment report, which gives us week by week estimates of initial claims. Through August the rolling four week average of initial claims showed a troubling upward trajectory, ending over the 570,000 level.
4) When reading the household survey, Table A-5 breaks down total employment by ag and nonfarm, as well as by wage and salary vs self-employed. These convey a hint of trends that may be meaningful, and in August, the numbers were both in the 90% confidence range and quite intimidating. The numbers following are SA (Seasonally Adjusted) except as noted:
- Total ag employment (not included in the establishment survey) dropped by 45,000.
- Total non-farm employment dropped by 137,000 SA, 861,000 NSA.
- Here's where the belly starts knotting up; wage and salary employment dropped by 654,000 SA, 1,129,000 (1.1 million) NSA. This passes the 90% confidence level rather impressively.
- Self-employed workers expanded by 353,000 SA, 262,000 NSA.
- Government employment declined 313,000 from June to August.
Instead, what I think we are seeing is the first group of long-term unemployed who are losing unemployment benefits and scratching a living. You only have to be employed one hour of one day of the reference week in order to be classified as employed.
A further comment about the government employment - because of the steadiness of government entities, these numbers are generally pretty reliable in both surveys. It is, however, a rather large drop in view of expanded federal employment, especially the census workers, and likely indicates the depth of the state and local combined tax/retirement problem. The combined impact will only increase for years to come.
Considering that the ADP employment survey has shown a pronounced shift in employment losses from large to small firms, I think we are seeing the collapse of small businesses, especially small retail, as well as hefty drops in hospitality employment.
However my own private data indicates that many small service businesses are considering shutting down because they are nearing the end of their rope. It is one thing to stop taking a salary for a few months. It is another to go into personal debt to keep the business open when you are losing hope, have no access to business credit and revenue continues to drop. This is the duration effect to which I have been referring. Most owners of small businesses are adjusted to surviving six months to a year of slack business, but not to two years.
On a YoY basis (non-seasonally adjusted, i.e. the Year Over Year change from August 08 to August 09):
- Non-ag wage and salary employment dropped 5,774,000.(-4.3%)
- Involuntary part-time rose 3,099,000. (+54.0%)
- Self-employed dropped 221,000. (-2.3%)
- Ag employment dropped 61,000 (-2.6%)
You might also want to see Table B-1, which gives the establishment survey breakdown by employment type. There are particularly hefty cuts in state and local government education. Also see temporary help services, which have stopped bleeding out but continued to slowly drift down by 6 or 7 thousand jobs a month this summer. Right now the absolute numbers for most of these categories are probably unreliable, but the direction is probably pretty reliable.
I will end this post with one final comment about the "almost recovery" theory: This is the real retail sales series (through July) from St. Louis Fed:
I want you all to open your little, non-polemical-because-non-government-employed eyes wide and stare at the trajectory before both of these recessions. See the flattening of real retail sales? In the US, we cannot go sideways. Real retail sales must increase more than the rate of population increase, or we wind up in a recession. This is why, from my perspective (as I wrote years ago), the 2001 recession began at the end of 1999, and the current recession began in 2006.
What happens when real retail sales consistently go sideways for a time is that profit margins drop because per-capita retail sales are dropping, which means that consumers and/or small businesses are squeezed. Because profit margins are squeezed, there are marginal effects on retail and business profits which add up to a slow erosion of jobs/incomes, the recent practical effect of which you can clearly see in the business inventories/sales ratios report:
See that slow drop beginning in 2006? Once that really sets in, gross private domestic investment starts to be constrained:
If that continues for long enough, you get a bad recession. My dad taught me the fundamentals of this method of business cycle forecasting when I was a kid, and it really works. His macro from micro method still forms the fundamentals of mine. He was a very high-end electrical engineer, and he had a manufacturing business. Btw, he died in 1991, but before he did, he predicted that Ford was the auto company which would be left standing. PDG performance!!! But I'm not surprised.
When my brother was getting his PhD in physics, he came home discussing a very interesting aspect in physics research. Not only did my dad know about it, but he knew more than my brother and the current research did. (Dad spent his evenings reading physics and math texts. He got bored easily.) My brother was astounded, and wondered out loud just when he would ever surpass my dad in knowledge. Well, TechnoBuddha, you're not the only one. When I really am confused about some aspect of the economy, I sometimes resort to modelling it with circuits.
But I digress. The best way to track real inflation is to use state and local government expenditures on stuff. The official CPI is pretty bogus. An alternate way aside from real retail sales is to use sales taxes, which will give you a similar trajectory. Sales tax receipts (adjusted for taxation changes) must increase at least 1% above the population increase or you are getting into recessionish territory, and if they stick below 1.5% over population increase for lengthy period of time you'll fall into a recession regardless. These slow small consistent changes are what generate the big changes over time. In growth recessions, the economy is limber enough to adjust on its own.
Of course, all those sources of data are always behind the trend, so the best real-time indicators by far are trends in grocery stores. Changes in grocery store stocking and pricing often will give you an indication 2 years before an official recession or a growth recession begins, and always more than a year before it does, which is very helpful indeed if you are a bank or a manufacturer.
What I am trying to explain is that by all the robust methods of economic forecasting I understand, this current adjustment (a recession is an adjustment) is not over. That's the sting in the tail of the scorpion. The underlying forces that control the economy have not reached a stable point from which consistent growth can rebound.
Next, the grocery/retail trends.
Incidentally, Ford isn't much better in my book. They're subject to the same fundamental forces as the other two--their investment trend went negative much later than GM's, but it's negative nonetheless.
My dad taught me the fundamentals of this method of business cycle forecasting when I was a kid....my dad...knew more than my brother and the current research...I sometimes resort to modelling it with circuits....
Now I understand you a bit better, MOM. I knew folks like you when I was getting my EE degrees. I always said it takes at least 2 generations to make a really good research engineer.
As for me--circuits, plumbing, the price of tea in China, they're all just terms in the transfer function.
Talk about liquidity...
Neil - I am not totally sure why, but a lot of it had to do with the way they shaped up their factories.
The key point is that the machine models the economy so closely that it becomes the economy, and vice versa. Changing the water flow on the machine changes the money flow in the actual economy.
A warning, though: If you like that sort of thing, reading one Discworld novel will cause you to have to read all of them, and there's at least 20.