Job Numbers: Adjusting The Adjustments

Last week’s widely reported figures for initial unemployment claims were distorted, as usual – but at least they were less distorted than in the recent past.

The Labor Department reported Thursday that the seasonally adjusted claims figure for the week ending August 29 was 881,000, compared to 1.011 million in the prior week. That’s a nice-looking dip of nearly 13%. But because it is a seasonally adjusted figure, it doesn’t reflect the experience of actual human beings.

In the real world, the Labor Department tallied 833,352 new claims for unemployment in the most recent week, compared to 825,761 the week before that. So there was no actual decline in unemployment claims, although the flesh-and-blood reality in both cases was better than the official statistics typically trumpeted in the press. Is this good news or bad news? Take your pick.

Seasonal adjustments are useful for statisticians and policymakers. They reduce the noise that is inevitably introduced by the rhythms of ordinary life in a series of recurring measurements. Policymakers need to know, for example, that a big pickup in employment at the start of the winter holidays is the result of merchants and delivery services staffing up to meet normal demand; it is not because policies are suddenly yielding exceptional results. Similarly, when employers let those seasonal employees go after New Year’s Day, it does not mean the wheels are falling off the economic wagon.

Since early this year, however, normal seasonal rhythms have had almost nothing to do with how the labor market (or the rest of the economy) has performed. Everything has been driven by the pandemic and by the measures government has taken to address it. As a result, the seasonal adjustments in the weekly unemployment claims reports have introduced considerably more distortion in the data than they have suppressed.

This is not a novel observation. I pointed out in this space back in April that in the course of a few weeks, as the nation hunkered down, the widely reported claims for unemployment were being overstated by millions as a result of seasonal adjustments that had no basis in 2020’s reality. Conditions were still awful, and tens of millions of people really were out of work. Even so, the measurements of both the downturn and the rebound that followed were seriously distorted.

Of course, anyone who wanted to could focus on the unadjusted figures. The Labor Department publishes them every Thursday, alongside the seasonally adjusted ones. But the old adage in the news business is “If it bleeds, it leads,” and the adjusted figures were bloodier. If you wish, you can add to that explanation your own adjustment for the biases, laziness and intelligence (or shortage thereof) of the journalists and outlets doing most of the reporting.

Last week’s seasonal adjustments produced less distortion than earlier ones because of a change in the way the Labor Department computed them. This was a nod to the reality that the pandemic continues to throw the normal rhythms of economic life out of whack. I suppose the government’s experts concluded that although the seasonal patterns are overwhelmed, they are still present, and so it makes sense to retain some degree of adjustment. As always, users of the data are free to focus on the unadjusted figures instead.

As those of us fortunate enough to have jobs return to work after Labor Day, what picture does the data paint of the labor market following nearly six full months of unprecedented economic restrictions?

Broadly, that it is in bad shape – but not nearly as bad as it was in the spring. And it was not quite as bad in the spring as the distorting seasonal adjustments made it seem. As of August 22, about 13.1 million Americans were collecting regular state unemployment benefits, accounting for roughly 9% of the insured workforce. This was down from 9.5% the previous week, and down from a peak of 15.6% in mid-May, a time when the seasonal adjustment produced an insured unemployment rate of 17.1%.

But in mid-August, nearly 13.6 million people were receiving federal pandemic unemployment insurance targeted to self-employed and “gig” workers. These are the musicians, actors, ride-hailing service drivers, beauticians and others who have been put out of work or had their income severely crimped by the shutdown. In other circumstances, they are ineligible for unemployment coverage, so there is no prior baseline with which to compare. The number of these claims has fluctuated widely through the summer. The reasons are not obvious, although I suspect at least some of it is due to the way various parts of the country lifted and then reimposed restrictions as they experienced surges in infections.

It has been a tough year for a lot of people. Any honest set of statistics would reflect that reality. But when we make comparisons across time periods, whether it is comparing this year’s economy to prior contractions, or comparing how we are doing now to how we were doing in earlier parts of the year, we need measurements that are legitimately comparable. It would help, too, if the people reporting those measurements took the trouble to understand what they mean.

• Create Content With AI
• Become A Funded Futures Trader