Someone stage an intervention on this guy. On Oct 11, 2012, at 9:05 AM, Jim Devine <jdevin...@gmail.com> quoted Jack Rasmus:
> An addendum, and partial correction, is in order to my previous post > on ‘Are the September Jobs Numbers Cooked Controversy’, as follows: > > My reference to the Current Population Survey, CPS, surge in jobs by > 870,000 last month, and my associating that surge with the labor > department’s ‘Business Employment Dynamics’ model, was incorrect. No again. The BED is yet another program, derived from the Quarterly Census of Employment and Wages, which gets its raw numbers from the unemployment insurance system. BLS home page here: http://www.bls.gov/bdm/ It measures gross job gains and losses, but with a considerable delay - the most recent figures are for 2011Q4. It's useful to measure turnover in the job market, and to decompose the net gain/loss numbers into gross flows. He's confused the BED with the birth/death model http://www.bls.gov/web/empsit/cesbd.htm which is an attempt to adjust the monthly payroll survey to account for business formations and closures between annual benchmarks. What follows is completely untrue: > The BED raw data on jobs are added to the labor department’s second jobs > data source, the Current Establishment Survey, CES, and then together > seasonally adjusted. The birth/death estimates - detailed in a table on the birth/death page - are added to the raw CES (payroll) numbers before SA. The BED has nothing to do with this. > For the past two years I had been pointing out that the BED raw jobs > data were artificially boosting the CES jobs numbers in the winter > months, November-January, and in turn likely artificially > underestimating the numbers of jobs in the summer months, May-August. Except that the birth/death model subtracted 398,000 jobs, not added, between Nov 2011 and Jan 2012, and added 467,000 between May and August, so he's got this exactly backwards. > (See my blog, jackrasmus.com, for that argument in detail in my > earlier postings on jobs this past winter and in previous winters. My > recent book, Obama’s Economy: Recovery for the Few, also documents > these excessive swings up and down in jobs for the past three years). > I still think that is true and that the CES-BED is generating false > seasonal jobs growth—inflated in the winter and deflated in the > summer. Some people have argued that the worst months of the job collapse in late 2008/early 2009 distorted subsequent seasonal adjustment figures, producing this artificial seasonal pattern. But if you apply the 2005-2006 seasonal factors to the raw data, you get the same effect. Rasmus doesn't seem aware of this controversy at all. > If so, what that means is that we will likely see an artificial surge > in CES jobs numbers in the November-January months coming up, just as > we’ve seen a collapse in job creation in the summers the past three > years. That of course does not explain the 870,000 September CPS > numbers. So what accounts for the CPS 870,000 is the obvious next > question. Is the CPS now picking up numbers that the CES is > underestimating in September? Or do we have some kind of parallel > seasonality adjustment problems now going on with the CPS as well as > the CES? My guess is that the inordinate surge in CPS jobs in > September is related to the extraordinary surge of 582,000 involuntary > part time jobs in September. Could be. Could also be that the household numbers are extremely volatile and one shouldn't pay much attention to these short-term anomalies. The o survey often differ, and often by a lot. There's nothing new about this. Since January 2011, the average change on the payroll (CES) side has been 150,000; on the adjusted household, 154,000 - trivial. But while the absolute value of the payroll change has averaged 150,000 (i.e., they've all been positive), the average on the adjusted household is twice that, 311,000. The average difference (payroll - adj HH) has been -73,000; the average difference in absolute values, more than twice that, 180,000. The max monthly change in payroll is 275,000 - vs. 879,000 on the HH side (adjusted, again); the min payroll is 45,000; min HH, -495,000. Trying to parse the differences in the monthly numbers is a waste of time. > Whichever the case, the bigger picture beyond September worth > considering is that we are apparently getting increasing divergences > and magnitude swings between the two jobs surveys, the CPS and CES. > That volatility is not an indicator of a stabilizing jobs market. Sometimes at turning points, the HH survey catches moves that the CES doesn't (because new establishments are created at a pace that's underestimated by the birth/death model). So in periods of acceleration, HH employment can outpace payroll employment over the course of several months. (If it's less than several months, forget about the difference.) That's meaningful. The reverse happens during periods of slowdown/recession: the b/d model overestimates new establishment and underestimates failing ones, to HH employment lags its CES counterpart. But what Rasmus says here is meaningless. > Quite the contrary. Moreover, something is going on here with jobs > calculations in both surveys worth further investigation. It could be > that seasonality assumptions and adjustments being used for both > surveys, CPS and CES, were perhaps more appropriate in a pre-2007 > economy and not today’s very much altered employment market > environments. The SA technique for the payroll survey is not canned - it's "concurrent," meaning based on recent patterns, with the past three years most heavily weighted. In other words, the SA machinery has largely forgotten the pre-2007 economy. Details here: http://www.bls.gov/web/empsit/cessa_oview.htm > Or it could mean we can expect to see a crash in the CPS > numbers soon and a simultaneous moderate rise in the CES jobs numbers > in coming months—i.e. the opposite of what’s been happening over the > past summer and in September. In either case, the labor department’s > jobs numbers will then appear even less convincing. They're more convincing than Rasmus' analysis. Doug _______________________________________________ pen-l mailing list pen-l@lists.csuchico.edu https://lists.csuchico.edu/mailman/listinfo/pen-l
