On Apr 3, 2011, at 2:50 PM, 123hop at comcast.net wrote:
> "A couple of things about last Friday’s job report. The seasonally adjusted increase was 216,000. The unadjusted number was 25,000. What this comes down to is how much you trust the BLS’ modeling of what is happening in the economy. My own impression is that it is assuming we are in a fairly standard recession-recovery cycle. I don’t think this is the case. The recession wasn’t like anything we have seen since the 1930s. And a common complaint I make is that no economists are looking at the economy as a kleptocracy or in kleptonomic terms. So the BLS modeling assumptions are almost certainly wrong as regards job creation.
>
> Also as I have said, we need to look for corroboration for the BLS stand on job creation. We should be able to find this in two places. The first is in our everyday experience. If the job situation is improving, then we should begin to see more local stories about more hiring and new businesses. The second is in other statistics that the government keeps such as hours, wages, capacity and capacity utilization, and tax receipts. I don’t see in either of these evidence that there is a lot of new hiring going on.
>
> Put simply in the absence of corroborating evidence I would question the reliability and accuracy of this BLS jobs number."
>
> Doug, you say the BLS folks are reliable. Comment?
Much about that is wrong. The BLS adjusts its so-called birth/death modeling, which is what the first paragraph is sorta talking about without actually understanding it apparently, after every annual benchmarking exercise. (The benchmark revisions, applied every February to the preceding March's data, adjust the employment estimates using the near-complete coverage of the employment universe provided by the unemployment insurance system.) So they're not basing the monthly estimates (which come from a survey of about 300,000 employers, aka the establishment survey) on their pre-existing idea of a standard cycle. There are also quarterly reports, also based on the UI system, which aren't used to benchmark the monthly estimates, but which can be compared with the monthly numbers; the most recent estimates, for December, are quite close to the monthly numbers. I'm not sure what the klepto part has to do with it, since I'm not sure how one's ethical or political critique can be incorporated into a statistical model.
The first paragraph seems to confuse seasonal adjustment with the birth/death model. Seasonal adjustment is basically a high-tech version of a moving average. Seasonal adjustment contains no assumptions about the business cycle - it's all based on historical experience, with the recent past more heavily weighted than the distant past. The cliched example is Christmas: retailers hire a lot of people in November and lay them off in January. It would make no sense to look at the raw numbers - otherwise you'd think there was a boom in November and a bust in January. So you use SA to smooth out those predictable things. There's also an issue around Easter - sometimes it's early, sometimes it's late. That also affects hiring patterns. The BLS tries to adjust for that as well. Or school calendars - those too must be adjusted for.
The second is wrong in this: there's plenty of confirmation of a modest recovery in the anecdotal data - and in tax receipts, both at the federal and state level. The hour and wage data come from the same source as the employment estimates, the survey of employers, so if you doubt the payroll numbers, you'd have to doubt the rest. And the BLS's monthly survey of households, which is separate from the employer survey, is telling a story very similar to the establishment survey.
Doug