The Curious Case Of The “Strong” January Retail Sales: It Was All In The Seasonal Adjustment

There was hardly a blemish in today’s retail sales report: the January numbers not only beat expectations across the board, including the all important control group which printed at 0.6% or the highest since May, but the December data was also revised notably higher. At first glance, great news for those who hope consumer spending is finally getting some traction from collapsing gasoline prices.

And yet, even a modestly deeper look below the strong retail sales headline numbers once again reveals just how this “across the board beat” was accomplished.

It was all in the seasonal adjustment, something which plagued the January non-farm payrolls report as well as numerous sellside analysts lamented.

The thing about seasonally adjusted retail sales is that while they are supposed to smooth out month-to-month changes in any given data series, they should be virtually identical to the non-seasonally adjusted retail sales on a annual, year-over-year basis. After all the same “seasonal” adjustment that was applicable this January, was applicable last January, the Januarybefore it, and so on, unless of course, something changed.

To the best of our knowledge nothing changed, even though while seasonally adjusted sales rose modestly by $800 million to $449.9 billion, on an unadjusted basis retail sales actually dropped by $112.7 billion with a “B.”

And indeed, when looking at the annual change in headline retail sales data we find that, as expected, the seasonally-adjusted (blue) and unadjusted (red)retail sales series are almost identical…

… but not quite.

If one zooms in on the most recent data, one finds something surprising: a substantial rebound in SA retail sales, which according to the Dept. of Commerce rose 3.4% – the best print since January 2015 – while unadjusted retail sales rose by just 1.4% – the worst monthly print since August, and hardly inspiring confidence that what is happening on a seasonally adjusted basis is indicative of what is really happening.

To isolate the problem we decided to look at only the annual (YoY) change in January data. The chart below shows the surprising finding: while virtually every January in the prior 5 years saw an almost identical change in the SA and NSA data, this January, there was a major disconnect: in fact on an NSA basis, January retail sales were matched for the lowest increase since the financial crisis at 1.4%, a far cry from the far more respectable and adjusted 3.4%.

To show just how much of an outlier January 2016 was compared to January in prior years, here is the seasonal “adjustment ratio” for the month of January for every year since 2010 to 2016, by which we define the ratio of “seasonally adjusted” to “unadjusted” retail sales. Spotting the outlier should be easy enough.

In other words, any “strong” rebound in January retail sales was all in the adjustment factor. We wonder if Yellen’s “dot plot” will likewise come in unadjusted and seasonally adjusted flavors from now on to reflect both the actual underlying economy and the economy the Fed would like to see when observed through the filter of an Arima X 12 seasonal adjustment model?