Tuesday, September 14, 2010

August Retail Sales: Debunking Mainstream Media Coverage

The 'Helicopter Economics Investing Guide' is meant to help educate people on how to make profitable investing choices in the current economic environment. We have coined this term to describe the current monetary and fiscal policies of the U.S. government, which involve unprecedented money printing. This is the official blog of the New York Investing meetup.

The August retail sales numbers were out this morning and media reports stated they were up 0.4%. Stock futures immediately rose because this was ahead of a 0.3% expected gain - or was it?  Buried in the middle of the government's press release was this statement, "The Census Bureau [the issuer of the report] does not have sufficient statistical evidence to conclude that the actual change is different than zero."

September 14th is primary election day in a number of U.S. states. It is amazing that political polls are reported more accurately by the press than government economic numbers. Like all political polls, most of the economic numbers are derived from surveys. All surveys have sampling bias, as well as other biases, that introduce errors. It is standard procedure for the U.S. media to report the sampling bias (such as plus or minus 3.4%) when reporting on any political poll. The mainstream media almost never does this with government economic reports even though their sampling error rates can be far more significant. While the mainstream media will report a political race as too close to call because of the margin of error in the polls, it will almost never say that that the numbers from a government report are meaningless gibberish because of statistical error.

The more accurate reporting of August 2010 retail sales is +0.4% (plus or minus 0.5%), so the number could actually have been minus 0.1% and the headline could have stated, 'No Conclusive Evidence Retail Sales Grew in August'. This is just the error from statistical sampling however. Unlike political polls, government economic reports have other serious problems. Incomplete data is one of them. As I mentioned in a previous article, the weekly unemployment claims for the first week of September was missing data from 9 out of 50 states, so the numbers for those 9 states were simply made up (and when the government makes up numbers for some reason it seems to err on the side that makes things look better).  Is there any missing data in today's August retail sales report?  The following items were not available:

Appl., TV and cameras
Computer and software stores
Building mat. and supply dealers
Beer, wine and liquor stores
Pharmacies and drug stores
Men's clothing stores
Women's clothing stores
Shoe stores
Department stores (incl. L.D.)
Other general merchandise stores
Warehouse clubs and supercenters
Electronic shopping and mail order houses
GAFO (firms that specialize in department store type merchandise)

The lack of data should be a reason to lack confidence in the final number in this report. Lack of consistency with other reported data is an even bigger concern. According to today's retail sales release, auto sales and parts were only 1.5% lower in August 2010 than they were in August 2009. Industry source Autodata found that the drop in auto sales between these dates was 21%. This problem could be reconciled if the government reported the change in auto sales as -1.5% (plus or minus 100%). It would also be a good idea if the government used the plus or minus 100% for all of its numbers. That would give the public a clear indication of how accurate they are, although it is highly unlikely the mainstream media would report it.
The government's press release on August retail sales can be found at: http://www.census.gov/retail/marts/www/marts_current.html

Disclosure: No positions.

Daryl Montgomery
Organizer, New York Investing meetup

This posting is editorial opinion. Like all other postings for this blog, there is no intention to endorse the purchase or sale of any security.

No comments: