Retail Sales Data: Lies and Statistics

Feb.15.10 | About: VanEck Vectors (RTH)

On Friday the US. Census Bureau announced the advance estimates of US. retail and food services sales for January. The headline numbers quoted in the media was a 4.7% year-on-year rise. This is a seasonally adjusted number and I'll discuss seasonal adjusting later in the article.

Firstly I want to focus on the non-seasonal raw data. Zero Hedge was understandably skeptical of the upbeat number referring to it as "yet another economic data fabrication." I view things slightly differently, more along the lines of an "economic headline fabrication", i.e. issue a press release with a good number knowing that the media merely regurgitates the press release rather than analyzing the underlying data. I've shown in a previous article that the raw, i.e. non-seasonally adjusted, data from this survey correlates well with sales tax data so I am comfortable about the raw numbers. The table below shows the situation in terms of raw data (sales revenue is in US$ millions):

Click to enlarge

Note that in these tables the columns run from Jan 09 to Jan 10, Jan 08 to Jan 09 and Jan 07 to Jan 08. So we're still well down on pre-global financial crisis (GFC) metrics, however the numbers understate the decline because had a GFC not occurred nominal retail sales would have continued to grow through 2008 and 2009 as we can see in the chart below:

Click to enlarge

The data released by the Census Bureau is in nominal dollars. For a more accurate assessment of where sales are, the next table shows retails sales in real dollars. Real dollar sales were calculated using the CPI, a series which in itself is frequently criticized, and normalized to 1992 dollars.

Click to enlarge

Note that since a CPI figure for January 2010 hasn't been released I left the CPI adjustment at zero. From this table is seems that retail sales have stabilized and small growth has returned. The data doesn't indicate any strength in the recovery. The next chart shows how badly retail sales have slipped in real terms.

Click to enlarge

One of the things in the GDP numbers for Q4 2009 that had me "puzzled" was the growth number for consumption expenditure. When you look at the real sales data quarter by quarter the real growth doesn't look anywhere near as robust as headlines would suggest.

Click to enlarge

Unfortunately we cannot get raw GDP data and the stuff that makes headlines is real, annualized, and seasonally adjusted, i.e. statistical yoga.

Here is the seasonally adjusted retail sales data:

Click to enlarge

I don't have a problem with the concept of seasonal adjustments, but the implementation gives me cause to be very skeptical of the numbers in press releases.

My understanding of seasonal adjustments is that by definition the same adjustment should be made at the same point in each year, i.e. at the same point in the seasonal cycle. Therefore, year-on-year (yoy) differences should be the same regardless of whether the number is seasonally adjusted or not. So my eyebrows were raised when the headline growth number came in at 4.7% yoy. We know that January is a quiet month therefore an amount is added to the measured number in January as a seasonal adjustment. However there is a 1.5% discrepancy between the yoy seasonally adjusted and non-seasonally adjusted numbers for January 09/10. When you look at the raw data, the Census Bureau added $26.185 billion or 8.35% to the number in January 2009, but added $32.089 billion or 9.91% in January 2010. In other words they magically pulled another 1.5% from the seventh planet in the solar system, to pump up the January number.

To justify this sort of statistical yoga you would have to be assuming that sales were seasonally lower in January 2010 not because of economic conditions but because of other factors -- bad weather kept consumers away from stores, there were too many good shows on TV to leave home and shop, and so on. It doesn't seem credible.

Disclosure: No positions