Tuesday, 11 March 2014

US Machinery Sales - When Seasonal Adjustments Fail

Today's news that wholesale sales fell a shocking 1.9% mom sa, with key machinery sales down 0.2% mom should be treated with some caution, because it's possible that the seasonal adjustment process is doing quite as much damage to the data as the weather, or the underlying wrinkles in the US capex cycle.

One of the least conspicuous victims of the great financial crisis has been the quality of Western economic data, or, to be absolutely specific, the reliability of seasonally adjusted economic data in the US and Europe.  Seasonal adjustment mechanisms are all predicated on the usually-correct assumption that most types of economic activity are distributed unequally throughout time – more people work in the day, more people go shopping at Christmas, and more financial analyst roadshows take place in October-November than any other time of year.  So seasonally adjustments work by looking back over the past few years to work out the usual patterns, and then deflate accordingly.

Most of the time this works well enough, but sometimes events can be so unexpected and so violent that they displace the normal pattern of activity.  For example, Japan's 2011 earthquake/tsunami/meltdown disasters over-rode normality; and so too did the great financial crisis of 2008/09. These dramatic disasters throw huge rocks into the pool of economic activity, and the way seasonal adjustments are calculated mean they are pretty much guaranteed to ripple throughout the next few years.

But there's one check you can make: seasonal adjustment should merely redistribute activity around the months of the calendar year, it should never significantly add or subtract to the totals accrued for the year in question. When there's a significant difference between the annual total (or growth) of seasonally adjusted data, vs non-seasonally adjusted data, you know for certain something's wrong.

When it comes to wholesalers' machinery sales, things started to go wrong in 2011, when the seasonally adjusted dollar total came in 1.8% lower than the non-adjusted total. In 2012, things got worse, with the seasonally adjusted total coming 2.9% lower than the non-adjusted total. Although things improved considerably in 2013, the impact of the errors was still showing in the yoy comparisons, with seasonally adjusted sales claiming a rise of 12.3% in 2013, when the non-adjusted count actually came to only 9.3%.

(At a total wholesalers sales levels, there was no similarly large discrepancy – the wholesale trade as a whole evidently didn't take such an asymetric hit as the demand for capital goods.)

These errors will diminish with time, and eventually pass out of the system, and 2014 may be the year in which the seasonally adjusted data finally reconciles with the unadjusted data. But for the time being, it's worth understanding that  there's another way of reading today's wholesalers data:

  1. Wholesale sales rose 3.6% yoy, on a monthly movt which was 0.3SDs above historic seasonal trends
  2. Wholesale sales of machinery rose 11.6% yoy on a monthly movt which was 0.8SDs above historic seasonal trends. 
As the chart below shows, stripped of its seasonal adjustment, January's wholesalers' machinery sales show not significant slowdown.

No comments:

Post a Comment