Tuesday, September 14, 2010

Are Retail Sales Overstated?

Mish doesn't buy the recent retails sales figures (hat tip reader Mike Hardy):

The issue...

I don't buy it. If retail sales were back to within 4.3% of the pre-recession peak, sales tax collections would be back towards the pre-recession peak, if not exceeding the pre-recession peak.

Why might they exceed the peak? Because of numerous state sales tax hikes.
In spite of numerous sales tax hikes, tax collections are still 5.9% lower than two years ago. Moreover, June of 2008 was not the pre-recession peak. November of 2007 was the pre-recession peak.
The conclusion?
Retail sales are down 10% or more from the pre-recession peak, especially if one factors in tax hikes.
Many states look at the Census report trying to figure out why their sales are lagging the national averages. The problem is the Census Bureau national averages are a blatant distortion of reality. The key to the states' conundrum is Census Bureau sampling methodology does not take into consideration stores that have gone out of business. Sales tax collections obviously do. Closed stores make no sales and collect no taxes.
In other words, same store sales are up because the overall number of stores has declined more than the level of overall sales (a smaller pie, but larger slices). The thought is the Census is then extrapolating those higher same stores figures to a base that is no longer there.

My sanity check on his hypothesis is to compare the decline in retail sales relative to business, manufacturer, and wholesale sales (under the assumption that retail stores have closed at a faster pace / are more numerous and harder to track than businesses, manufacturers, and wholesalers).

While there could be alternative reasons (i.e. price levels, type of sale, etc...), it will be worth following this to see if he may be on to something...


It looks like my "smell test" may have holes. Mish reached out and noted retail sales peaked in November 2007. In looking at the data since that peak, a different trend can be seen:

In a nutshell, it appears retail sales slipped first, then sales of business, manufacturing, and wholesale followed (thinking these sales "react" to the end consumer). More recently, it looks like retail sales have rebounded first and again, have been followed by business, manufacturing, and wholesale.

So what may cause the discrepency between retails sales and tax collections? Back to Mish:

1. Sales Tax holidays
2. Online sales
3. Government purchases

Source: Census


  1. Only if Census Bureau hires competent economists!

  2. I'm increasingly convinced that USA economic statistics are not much more valid than the old USSR statistics.

    The key difference seems to be "A" versus "SR".

  3. Fantastic digging! I think your split of the data into different series is based on a good assumption, this is definitely an interesting result, and it will be interesting to watch.

    With regard to the US vs USSR comment (i.e., is this all just propaganda?) - I personally think there is perhaps a little bit of that, but mostly the data series that seem somewhat corrupted (like this and others such as CPI with hedonics and owner-equivalent rent) may seem propaganda-like but the bias is systemic. That's an interesting property in that it does allow for better spin but also still allows you to apply the inverse of the systemic bias once you understand it, peel back a layer and infer the more "real" data set cleanly underneath (with something like CPI-CS, to continue the example).

    So at least in my humble and maybe naive opinion, we're not quite living in a USSR or Ministry of Truth dystopia, but it certainly pays to be skeptical, break series into constituent parts and think hard about the methodologies and motivations of the series publishers. Fun stuff there quite often, and maybe useful to steal a march on the market - esp. if this retail sales weakness is true and we're just living in a recognition / impulse/response window per Hussman and you get a short position (or simply de-risk) on pre-general-recognition.

    Thanks for the wonderful posts

  4. There are a couple other factors to consider.

    1. Sales Tax holidays
    2. Online sales
    3. government purchases

    Those sales collect no tax

    However, online sales are about 6% of sales, sales tax holidays are pretty rare. I do not know about point #3. It was just brought to my attention.


  5. The peak in retail sales was November 2007

    Can you take the numbers back to then?


  6. Looks like Mish blew it.

  7. Not sure anyone has blown it yet - I think another plausible explanation is that the hypothesized methodological error may be present in all of the sectors (businesses / wholesalers etc are closing too!) and the divergence in this particular chart may just be showing the non-retail inventory cycle vs the more leading retail cycle.

    Obviously that could be incorrect, just saying that it is basically unclear why tax collections and sales are diverging, I haven't seen anyone that has data showing exactly what things are causing it and how much is attributable to each thing...