Chicken Little initial unemployment claims

You may have noticed on July 21 when the national employment statistics were released that many headlines throughout national media outlets shouted initial unemployment claims were up up up!

Here’s an example from Bloomberg: “US Jobless Claims Hit Eight-Month High as Labor Market Cools! Initial unemployment claims increased last week to 251,000!”

CBS reported: “Jobless claims rise to 8-month high!”

U.S. News had the headline: “Jobless Claims Reach 251,000 as Labor Market Shows Signs of Weakening.”

In terms of the art of writing headlines, these are eye-catchers. With headlines and content like this, you might think the sky was falling. Granted, an increase in initial claims isn’t great, because every new unemployment claim means someone lost their job; however, it’s useful to remember that there’s more to these numbers than just the numbers.

What was missing from most this coverage were a meaningful discussion of the size of the dataset and a sober evaluation of how messy or neat the data is. When those two things are added back into the analysis of the numbers, a less sensational perspective on the national job market emerges.

First, the size of the dataset is monstrous. It’s the entire workforce of the nation. The United States is not small. Its population is over 331.9 million people. The civilian workforce is 164 million using the seasonally-adjusted and 165 million using the unadjusted numbers. Either way, the current workforce is approximately half the population.

The size of the workforce matters in a big way when talking about things like unemployment claims. When you compare a workforce of 164 million with 251,000 initial unemployment claims and a weekly increase of 7,000 compared with the week before, those numbers fade to insignificance. Those 251,000 initial claims are 0.15% of the workforce. The 7,000 increase is a 2.87% increase in initial claims but it’s only 0.0043% of the workforce and is below the signal-to-noise ratio of the raw data. By comparison, if the initial claims increased by 7,000 every week for a year, that would be an actual trend worth looking at.

The U.S. Department of Labor usually uses a four-week average to smooth the unemployment claims data. When we smooth the data this way, the four-week moving average for the week ending on July 16 was 240,500, an increase of 4,500 from the previous week’s moving average.

We can get an appreciation for how much wobble there is in the data several different ways. Being able to measure that wobble can tell us if the 7,000 increase in initial claims is a signal that rises above the noise in the dataset. The first way to evaluate this is very simple: we can compare the difference between the weekly data and that four-week moving average. The residual is the absolute value of that difference. We can estimate the weekly wobble, or noise, in the data by taking the average of the residuals. Choosing Jan. 1, 2021 as a starting point for initial unemployment claims, the average of the residuals is 24,845 (using the not-seasonally-adjusted dataset).

Another approach to get a handle on the level of noise in the data is to calculate the standard error, which for Jan. 1, 2021 through July 16, 2022 is 25,188. For this small and simple dataset, we can stop here. While the average of the residuals is not that rigorous, standard error has some real teeth as a metric. Our 7,000 weekly increase is below the standard error.

The eight-month high reported by CBS is noteworthy but only for the seasonally-adjusted data. Initial claims had a local minimum of 166,000 for the week ending on March 19. This is a real trend; however, looking at a year-over-year comparison, the number of initial claims on March 13, 2021 was 699,000. Those numbers steadily declined for a year. The July 16 initial claims are 36% of the March 13, 2021 initial claims.

For July 16, 2022, the year-over-year comparison is 251,000 versus 424,000, so the most recent initial claims are 59% of last year’s number. The labor market is still much tighter than it was last year. What do the numbers look like for Idaho? The initial unemployment claims data for 2022 appear indistinguishable so far from the three years before the arrival of the COVID-19 pandemic.