On Jan 9, 2012, at 1:09 AM, nathan tankus wrote:
> "That thing is absolute crap. "
>
> interesting. why?
It being Shadow Economic Stats, for those who missed the antecedent.
He doesn't disclose his techniques in any detail. But his work on prices, for example, claims giant errors by the BLS which no one else has ever discovered. The BLS addressed some of his claims (though only mentioning his site in footnote 57) here:
http://www.bls.gov/opub/mlr/2008/08/art1full.pdf
As far as I know, he's never responded. He's certainly never disclosed his techniques with this level of detail
On the employment numbers, I spend a lot of time with these things. I really know this stuff well and from what he writes I don't think he really does. He just says stuff without any supporting evidence. He makes routine revisions sound like a conspiracy. He says things like the BEA targets the Wall Street consensus on its early GDP estimates, which is just baseless nonsense.
He wrote this in February 2008:
> The Bureau of Economic Analysis (BEA) reported annualized real (inflation-adjusted) growth in fourth-quarter 2007 GDP at 0.64% +/- 3%, which was statistically indistinguishable from a meaningful contraction. The reported growth was down from nonsensical growth of 4.91% for the third quarter. The BEA tries to target consensus forecasts (which were 1.2%) for the advance estimate, since they have to guesstimate more than 90% of the underlying data. The reported result suggests the first-cut estimate actually was a contraction. Any number of bad assumptions currently in place, if altered slightly, would have given that result. Annual growth reportedly slowed from 2.8% to 2.5%. This series is so heavily politicized that it is little more than political propaganda. Underlying reality suggests both the quarterly and annual rates of change should have been negative, as will be discussed in the upcoming newsletter.
The actual, fully benchmarked growth rates for 2007Q3 and 2007Q4 were +3.0% and +1.7% (annualized). They came in under consensus for Q4 on the first release, which turned out to be too low (as was the consensus). Annual rates are now 2.5% and 2.2% - not all that far from the original estimate.
> With Consistent Seasonal Adjustments, Payrolls Fell by 100,000. The Bureau of Labor Statistics (BLS) reported a seasonally-adjusted jobs loss of 17,000 (a loss of 393,000 net of revisions and benchmark revisions) +/- 129,000 for January, following a revised 82,000 (previously 18,000) jobs gain in December. Given the BLS’s ability to have brought the reported monthly change in at any desired level,
This is just a lie. These BLS and BEA people are honest civil servants who have to resist political pressure and do a very good job. This is just know-nothingism.
The economy was starting to fall apart in early 2008. There's no way the BLS could have kept up with that. January 2008 is now reported as a *gain* of 13,000 - this is fully benchmarked with the records of the unemployment insurance system, which has 99% coverage of the employment universe. There's not much difference between -17,000 and +13,000, really. But I don't get his point. Paulson wanted a small contraction? What administration official wants negative jobs numbers? The loss really should have been -100,000? Why?
> One factor that may have affected the result was a revamping of the bias factor (birth-death model), which was a net subtraction of 175,000 jobs in January 2007. At that time, the decline had been offset adequately by a change seasonal adjustments. For January 2008, the bias factor was revised to subtraction of 378,000.
The birth-death model is an attempt to keep up with business births and deaths between annual benchmark revisions. Most of the time it works pretty well. It breaks down at turning points, like 2007-2008. But conspiracist loons like to turn it all into a big plot.
He also says:
> seasonally-adjusted January payrolls, using consistent seasonals, would have shown a 100,000 jobs loss for January
Which we now know to be +13,000. But this "consistent seasonals" bit is nonsense. The BLS has been using "concurrent" seasonal adjustment, which (unlike the CPI and retail sales) doesn't use a canned series derived from previous years, but is constantly revised every month. For example, there's usually a lot of retail hiring for Xmas. A "normal" amount of hiring should get adjusted away and only the departure from normal would survive adjustment. But if there's less than normal hiring in November and December, you'd expect less than normal layoffs in January. Concurrent adjustment is supposed to adapt to this, and "expect" lower layoffs after weaker hiring. In any case, over a year, the additions and subtractions from seasonal adjustment sum to 0, so any "lie" that would be told through SA can't last very long.
I wrote this for The Real News website a while back:
> A few words about some of the comments that have been posted here and around the web.
> First, someone pointed to a critique of the productivity stats from John Williams of Shadow Government Statistics. Williams is a real piece of work. I spend a lot of time tearing into these numbers and talking to the people who collect them. I really don't think that Williams does. His stuff often reads like he just pulls it out of the air. When he writes about the technicalities of putting together the official data, Williams frequently sounds like he doesn’t know what he’s talking about—and he rarely discloses his own methodology. Officialdom discloses its techniques at great length.
> Some specifics. Williams’ productivity critique <http://www.shadowstats.com/article/productivity-miracle> comes along with a silly essay on the Federal Reserve’s flow of funds accounts, their quarterly taxonomy of financial assets and liabilities, from 2006. He mocks the downward revision to the Fed’s earlier estimates of foreign holdings of U.S. Treasury bonds for late 2005, and its apparent shift to U.S. households.
> But almost all stats are regularly revised, sometimes up and sometimes down. But he treated this as a confidence trick to hide the government’s indebtedness. But five years later, the rest of the world’s holdings of U.S. Treasury bonds have increased by almost half, a huge increase that no one’s tried to revise away.
> Williams did similar stuff with employment stats early in the recession, when the job losses were pretty small. A few months later, the Bureau of Labor Statistics had no problem reporting huge, sustained job losses, the worst in modern history.
> On productivity, Williams revises the official number for “methodological gimmicks that were added over time.” He doesn’t tell us any more—what these gimmicks are, or how he did his magical revisions. There are some complicated issues with how to adjust price changes for quality changes. The crux of the problem: today’s computer is more than twice as fast as 2008’s, but probably costs the same or less. Since you’re getting twice as muich computer as you were, what happened to its price? But do people really take advantage of all that speed? There are a lot of products like this—and it’s not easy to measure what their value is over time. The BLS tries to adjust for it. You can disagree with the way it's done, and I have, but it's not bullshit. Williams certainly doesn’t know any better.
> The BLS is staffed by serious people who are independent of political pressure and try their best to do a good job. It's hard, conceptually and practically. No one ever said they're perfect, but they're pretty good.
>
> And here's an important political point: these stats are collected not for the broad public, which has no idea what they mean, but for elites in business and finance. They want honest accounts, not bullshit. Why would a ruling class's government lie to that ruling class?
> I've long said that the official stats are damning enough as they are. The notion that productivity is up and wages aren't, which has allowed for a massive increase in profits and upper-bracket incomes, is hardly at odds with most people's experience. Everything we want to know about poverty, inequality, overwork, bad health, you name it, is in the official stats. Why some guy with a website and an attitude is more credible than statistics collected by thousands of serious, dedicated people is beyond me. But people believe what they want to.
Doug