Spot The Adjustment To The Seasonal Adjustment

Tyler Durden's picture

Recall how the prevailing excuse for all economic data missing in the past two months have been attributed to inclement weather which mysteriously has not been captured in seasonal adjustments? Well, something curious happened in today Durable Goods report.

As we noted previously, the reason why stocks posted a bounce when the report came out was the stronger than expected bounce in Durable Goods, and especially the Durable Goods ex transports print which rose 1.1% on expectations of a modest drop. It is here that a somewhat puzzling observation can be made. First, the Non-seasonally adjusted number in the series posted a $5 billion decline from December into January, in other words a 3.2% drop compared to the 1.1% seasonally adjusted change. But it is not the sequential data that is notable but the annual change which is where the seasonal adjustments are most obvious.

What happened in that set of data is that while the NSA Durables ex transport posted a tiny 0.4% increase, amounting to $533 million, the SA series showed a far more sizable $1.9 billion increase. This can be seen on the chart below.

Why is this curious? Because in all recent prior years, the year-over-year change was a lower increase in the SA data compared to the NSA.

 

And more to the point, here is the same chart showing just the difference between the two year over year data sets. Spot the outlier:

 

What does this mean? Simply, in January, the Department of Commerce decided to apply a far greater than historically applicable seasonal adjustment, boosting and skewing the data far more than history suggests it should, and giving the seasonally adjusted number the benefit of about $2.3 billion in extra "oomph" solely due to the discretion of the person applying a far stronger than deserve seasonal adjustment!

Sure enough, had an inline adjustment been applied, the Durable Goods ex Transports seasonally adjusted number would have been a decline sequentially, and in fact would have printed at -0.3%. Precisely in line with expectations. But that would hardly be enough to send the headline kneejerk algos into a momentum buying spree, which in a market that was threatening to drop on the overnight newsflow was not acceptable...

So much for seasonal adjustment not adjusting for the impact of the "harsh" January weather.