Tag Archives: seasonal adjustments

Retail Sales: Inflation Plus Extrapolation

The footnotes are the most interesting section of every financial and economic reports.  They also happens to be least studied section of these reports.  Those who prepare these reports rely on this fact.

The monthly headline retail sales is based to a large extent on estimates, guesswork, invalid assumptions and statistical magic.  Examine the line-item details in this retail sales report link. Note the numerous lines for the May “estimate” that contain “(*).” Then scroll down to the footnotes.

“(*)” indicates, per the footnotes, that “advance estimates are not available for this kind of business.” Footnote 3 further explains that “Advance estimates are based on early reports obtained from a small number of firms…”.   In other words, a significant percentage of the retail sales are based on guesswork and inference.

Scroll further down the retail sales report and Table 2 shows the summary table (Table 2) which presents the month to month percentage change comparison for the latest month’s report.  The data in first four lines in this table is the data used for the headline reports.

Everyone uses these numbers, most without any knowledge whatsoever about the degree to which the data “behind” the numbers is comprised of highly questionable guesswork and unsubstantiated, if not entirely problematic, statistical inference and  adjustment calculus.

Additionally,  there’s a section in the report that explains methodology for the guesswork.  “Advance estimates are computed using a link relative estimator.”  A “link relative estimator” is a polite descriptor that basically means, “we assume that the historical growth rates implied by our historical reports can be applied to growth rate we assume in  this month from the previous month.”  On top of all of that, the Census Bureau then applies its nefarious “seasonal adjustment” factors to the data.  Keep in mind that a significant portion of the data is pulled out their ass.

All of this methodology is explained in further detail in the tabs on the main Monthly Retail Trade page of the Census Bureau. The information spread out in this section substantiates every assertion I have put forth above. It requires sifting through the “how data are collected,” “definitions” and “FAQs.”  I’m probably one of the few analysts curious enough to subject myself to this brain damage.

By the Census Bureau’s own trumped up numbers, most of the “gain” in retail sales from April to May, if indeed a bona fide gain occurred, was from gasoline and clothing inflation.   The numbers in the report are expressed in nominal terms.  They are not adjusted for the effects of price inflation.  Removing the effect of price inflation would yield the change in “unit” volume of retail sales.  This would be the number of true interest.

Finally, the estimated change in retail sales is not consistent with the patterns in consumer credit.  Based on the Fed’s consumer credit report, the use of revolving credit (credit cards, checking overdraft accounts, etc) has been contracting.  With the savings rate at an all-time low, the only way that retail sales unit volume could possibly increase is through the use of credit.  Thus, while guesswork and inflation is driving today’s headline report, in all likelihood unit volume of sales declined.  This latter assertion is indeed supported by recent manufacturing, factor order, durable goods and wage growth data.