Is U.S. Economic Data Being 'Massaged'?

Many market-watchers claim that U.S. economic statistics are increasingly being revised downward in subsequent periods, suggesting that the figures initially being reported by Washington are "puffed up," so to speak, most likely for political purposes.

Well, I went back and had a look at the differences between the reported and revised data for various series, including monthly retail sales, nonfarm payrolls, industrial production, and durable goods orders, to try and figure out if the cynics are right.

Using data from Bloomberg, I calculated whether the revised data for each month was lower than the first-cut estimate. Then I tabulated 12-month running totals for each series to see if there has been some sort of systematic bias (in other words, whether the pattern of monthly downward revisions was trending higher instead of undulating up and down).

To make the comparisons easier, I subtracted the 12-month tally as of May 2002 (an arbitarily chosen date) from the monthly totals for all four economic series so that the starting point for each would be the same -- zero.

Based on a quick read of a graph of the data (see below), it does seem as though the pattern of negative revisions has been trending higher lately, especially during the past year or so, suggesting that the cynics may be on to something.

That said, I am not a statistician, and the results may be nothing more than "noise." There is also the possibility that my methodology is lacking (because, for example, the margins-of-error for each month's data are relatively large, or because of certain quirks that crop up when an economy is in transition).

Still, you gotta wonder...

Comments

[Matt]

Economic data is typically revised down during recessionary periods. If this is "the worst recession since the depression" one would expect to see the most significant downward revisions.

When these revisions start to fade or are revised upward, that will be a meaningful sign that we are approaching the end of this recession.

[DavosSherman]

“Well, I went back and had a look at the differences between the reported and revised data for various series, including monthly retail sales, nonfarm payrolls, industrial production, and durable goods orders, to try and figure out if the cynics are right.”

Cynics? & Try to FIGURE OUT????

For Christ’s sake - Just look at how they calculate GDP. The flipping thing is off by 40%!!!!!!!!!

You own your house. The government calculates what you’d pay in rent if you rented, which you don’t rent, and then they add that rent, which you don't pay to GDP.

I could go on for days and days about any and all the statistics.

Forget the messaging. A corpse is a corpse, with or without the masseuse.

F&*king Enron accountants would be proud of sh&t like this.

Cynics? ” “You gotta wonder?”

No, I'm not a cynic and, golly gee, I don't wonder.

Have you been watching Cramer??????????????

[no eyeballs]

the way to tell is to do a runs test, a nonparametric test (doesn't assume normality, or any distribution) of the number of sequential up or down revisions. Tells you the chance that the initial errors are just noise.

[no eyeballs]

e.g.

http://home.ubalt.edu/ntsbarsh/Business-stat/otherapplets/Randomness.htm

[DavosSherman]

http://www.chrismartenson.com/crashcourse/chapter-16-fuzzy-numbers

Scre* nonparamtric, its all about the aggregate

[DavosSherman]

PS If I'm passionate about one thing it is bad data. We, as bloggers, need to really make this point the key point.

Everything should be based on accurate data. Debt, recovery, deficits, unemployment - everything.

This is Enron accounting they are doing.

The message after the cooked books is a mere puppet show.

Take our debt. Your clock advertises 11 trillion. Add to that the liabilities of Social Security, Medicare part A and D and we are looking at another 55 trillion.

Add to that the stimulus, 2 wars, tarp, talp and the rest. Puplava pins it at 80 trillion. And we make (GDP) what 6 trillion after you back out the cooked imputations.

[carol]

One aspect is whether numbers are revised downward (GDP, factory utilization) or upward (unemployment, foreclosures). Another is how the revision compares to the initial estimates of analysts outside government.

A very clear case of manipulation was the GDP figures published a week before last year´s presidential election. Beforehand all economists predicted a GDP decline of 0,5%. No variation. Then the GDP growth was published: only 0,3% decline, almost half as bad as expected. It started a huge rally.
Later, the GDP number was silently revised to a decline of ........ 0,5%, exactly as previously estimated.

[CLR]

Eyeballs is right - non parametric test such as an odds ratio. However, you would have to check the growth side as well. This would tell you whether the errors were due to:
Lags,
Ramdom, or
Manipulation.