Good News and Bad News (and Bad Data)

ABOUT US : Collapsed
INSIGHTS : Collapsed


Good News and Bad News (and Bad Data)

Jun. 04, 2020

In normal times seasonal adjustment factors are applied to data to smooth out the normal waxing and waning of the labor market, retail sales and the like. After all, it would make little sense to get excited about consumer spending rising in December but then to become worried when spending declined in January. However, given the speed and the magnitude of the impact of COVID-19 on the economy and on economic data, we believe it makes little sense to make adjustments for traditional patterns when, in doing so, you may be distorting the data and limiting its value as an indicator of whether the economy is improving or faltering. 

The claims data show this clearly. In February, seasonal factors deflate the number of claims filed as retailers continue to layoff workers post the holiday shopping and return season. For example, in the  first week of February, the not-seasonally adjusted claims data rose by 225,0001 but, after accounting for the normal seasonal variation during that time of the year, the seasonally adjusted claims figure (what is in all the headlines) was reported to have increased by just 201,000. In normal times, these adjustments make sense. 

Post-COVID, when the level of claims has reached unseen heights, applying a standard seasonal adjustment distorts, rather than clarifies, the picture. For the most recent week,2 the reported seasonally adjusted claims filings rose by 1,877,000 versus not seasonally-adjusted claims filings (actual, if you will) of 1,603,000. This is because this week, as can be seen on the chart below, was a big adjustment week. (To get a sense of scale, if the seasonal factor used this week was applied to data from the first week of February the impact would have been to add 34,000 filings. However, because the post COVID filings have been so large, seasonal factors added 274,000 to the past week.)

In fact, the period coinciding with COVID is a traditional period where the seasonal adjustment boosts the unadjusted data to account for seasonal variations. Indeed, if we sum up all of the seasonally adjusted claims during the COVID crisis period (mid-March to present), the number of seasonally adjusted claims filed during that time is 42,647,000, a figure that is 3.8 million larger than the unadjusted claims filed during that period. 

We believe this period of extreme overstatement is about to end. The seasonal adjustment factors (in green below) are returning towards levels where they will likely inflate claims by far less than has been the case over the last two and one-half months. In other words we believe the reported claims will fall next week, and we  expect them to fall sharply based on the lack of a seasonal factor pushing them higher.

Good News and Bad News (and Bad Data)