EDITOR NOTE: The overarching theme in this smart article is that headline numbers--to which most people respond--often conceal the details (and the reality it presents) that, in many cases, can contradict the headline. When it comes to reading unemployment data, seasonal adjustment calculations often don’t give a clear picture as to what the data truly means. Hence, it’s misleading, and it takes someone who really knows how to decipher it to fully comprehend just how good or bad it really is. If you’re not one of these people, then the article below explains it in clear and easy-to-read detail.
First the data, then the data chaos.
It’s still a historically catastrophic number, but it’s the least catastrophic number since mid-May: 27.02 million people (not seasonally adjusted) continued to claim unemployment insurance under all state and federal programs, down by 1.04 million from the prior week, according to the Department of Labor this morning.
This means that 16.9% of the civilian labor force (160 million) continues to claim unemployment benefits.
Blue columns – claims under state programs:
The number of people who continued claiming unemployment insurance (UI) under state programs fell by 272k to 13.9 million (not seasonally adjusted), in line with the downtrend that had started in May.
The number of people on UI under all federal programs established by the CARES Act and some other programs fell by 769k to 13.1 million (not seasonally adjusted):
- Federal PUA claims fell by 252k to 10.97 million. The Pandemic Unemployment Assistance program, established under the CARES Act, covers contract workers, the self-employed, gig workers etc. who’d lost their work.
- Federal PEUC claims jumped by 119k to 1.41 million. The Pandemic Emergency Unemployment Compensation program, established under the CARES Act, covers workers not covered by other programs.
- State Extended Benefits more than doubled to 203k.
- State Additional Benefits programs ticked up to 2.7k. This program, available in some states, covers people who exhausted their regular state benefits.
- State STC / Workshare ticked down to 305k (employer avoids layoffs by reducing the number of regularly scheduled hours of work; employees receive some wages plus a prorata share of weekly benefits).
- Federal Employees, continued claims ticked down to 14.0k.
- Newly Discharged Veterans, continued claims ticked down to 13.6k.
The $600 a week in extra benefits – part of the CARES Act but expired at the end of July – were on top of these state and federal unemployment programs and are not included here. People who qualify for one of the state or federal programs received the extra $600 a week. This has been replaced by a program from the White House of $300 extra a week, that could become $400 a week in some states, and is now being rolled out by the states. All programs are administered through the same state unemployment office.
The newly-out-of-work: Initial UI Claims, state & federal:
Initial claims under state programs are an indication of the number of newly laid off workers. After having “unexpectedly” risen by 50k in the prior week, initial claims fell by 68k (not seasonally adjusted) in the week ended August 22, to 822k:
Initial claims under the federal PUA program for contract workers jumped to 608k, from 524k in the prior week (not seasonally adjusted).
Combined, state and federal initial claims ticked up to 1.43 million people who newly lost their work and filed for benefits during the week.
At this rate of 1.43 million initial claims per week (not seasonally adjusted) under the state programs and PUA, about 6 million people per month lose their work and file for unemployment compensation. In other words, on a monthly basis, as 6 million people still lose their work, over 6 million people must return to work for unemployment levels to improve.
Data Chaos, Part 2: Labor Department Admits “Seasonal Adjustments” Went Haywire During Pandemic.
During the early weeks of the collapse of the US labor market, it became clear to me that seasonal adjustments were exaggerating unemployment claims. So with the UI report on May 28, I switched to discussing only “not seasonally adjusted” unemployment claims. Today, in a special note, the Labor Department explained that my hunch was correct, and why it was correct, and that it would change its method starting in September.
Unemployment has seasonality. For example, there is strong hiring in October, November, and December by retailers for the Christmas Shopping Season. Then in January, these people get laid off, and many of them file for unemployment benefits, and so initial unemployment claims jump every year in late December and in January. This is predictable, and “seasonal adjustments” iron that out. There are also smaller seasonal variations over the spring and summer that are normally predictable and can be adjusted out of the data.
The way the Labor Department made those seasonal adjustments was by multiplying the not seasonally adjusted data by certain well-established factors for that season – the “multiplicative factors,” as it calls them. But during the Pandemic, the number of claims jumped by a never-before-seen magnitude (6 million in one week), and the “multiplicative factors” exaggerated the seasonally adjusted claims.
The chart below shows the difference between “seasonally adjusted” and “not seasonally adjusted” initial claims (“seasonally adjusted” claims minus “not seasonally adjusted” claims).
There is always a big difference in late December and January: not-seasonally adjusted claims jump as the retail workers get laid off, while the seasonally adjusted data irons out those jumps. You can see this on the chart by dips into the negative, showing the amount by which “seasonally adjusted” claims were smaller than “not seasonally adjusted” claims.
But starting on March 21, this method of “multiplicative factors” went totally haywire, one way, and then in early July the other way, and in late July and August it flipped again, and it continues to go haywire:
So the Labor Department is going to change the method, it announced today. And it explains:
“A multiplicative seasonal effect is assumed to be proportional to the level of the series. A sudden large increase in the level of the series will be accompanied by a proportionally large seasonal effect.”
And this caused those adjustments to go haywire during the Pandemic, or as the DOL said, it created “systematic over- or under-adjustment of the series.”
The DOL will change from its old method of seasonal adjustments (“multiplicative factors”) to its new method of “additive factors.” And this “additive seasonal effect is assumed to be unaffected by the level of the series,” and tends to “more accurately track seasonal fluctuations in the series and have smaller revisions.”
So the DOL announced today, that starting in September, “the Bureau of Labor Statistics staff, who provide the seasonal adjustment factors, specified these series as additive.”
OK, so here you have part of the reason for the data chaos in the unemployment data that has been a key theme in my coverage of this mess since March.
Originally posted on Wolf Street