JOLTS of reality

Stephen Stanley | January 15, 2021

This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.

The December monthly employment report, the monthly JOLTS report for November and a 200,000 surge in weekly unemployment claims have offered three different perspectives within the last week on the state of the labor market. While the employment report, as always, is the best source of information, the JOLTS series includes data on the pace of layoffs, which suggest that the unemployment claims figures remain grossly inflated.

JOLTS data

The Bureau of Labor Statistics introduced the JOLTS report in 2002 to offer more detail than the payroll figures could offer. JOLTS offered a picture of the immense flows in the labor market. In a normal month, payroll employment might rise by 200,000 or so, but that figure masks the enormous churn in the labor market. In normal times, something like four to five million workers start a new job in any given month, while a similar number leave a job—often, many of the same people.

JOLTS provided tallies for job openings, hires and separations, with separations divided into quits, layoffs, and “other,” to offer new detail on that churn. The data come from a monthly survey of establishments, about one-tenth the size of the monthly establishment survey used for the higher-profile payrolls readings.

Market participants for the most part do not pay much attention to the JOLTS report, perhaps in part because it lags the more consequential employment report by more than a month. For example, the report released January 12 was for November, while the employment report released on January 8 was for December. To the extent that the JOLTS report garnered attention pre-pandemic, it was for the count of job openings. In recent years, the elevated level of job openings at a time when the unemployment rate was setting new 50-year lows offered confirmation of the tightness of the labor market and the shortage of workers facing many employers.

Tracking layoffs during the pandemic

During 2020, the JOLTS report came to offer a new insight. The changes to the unemployment benefits program in the wake of the pandemic, the broadening of eligibility and massive bonus weekly payments, created unprecedented conditions that have rendered the data difficult to interpret, as the pace of new filings in particular has run multiples of pre-pandemic levels since March.

There have been numerous reports suggesting that the initial claims figures were inflated by a variety of factors. State labor offices apparently booked duplicate claims as well as erroneously counting ongoing beneficiaries as initial claims every week. In addition, as time has passed, an increasing number of states have revealed that their programs were plagued by floods of fraudulent claims, as criminals filed claims using stolen identities, and overwhelmed labor offices were unable to properly vet the applications in due time.

Since late spring, the initial claims figures seemed hopelessly distorted. The JOLTS data offer a reality check on the pace of layoffs. One slight complication that is easily resolved is that the claims figures are weekly, whereas the JOLTS data are monthly. Since market participants are more familiar with the initial claims data, it is reasonable to convert the JOLTS series of layoffs to a weekly proxy by dividing by four.

Several things stand out in comparing initial claims to the JOLTS data (Exhibit 1). First, pre-pandemic, the JOLTS layoffs series was running roughly double the pace of initial unemployment claims. This makes sense, because, prior to the pandemic, only a subset of the workforce was eligible to file for unemployment insurance. Second, the two series moved roughly in tandem during the first stages of the surge in unemployment. Note that the March readings are quite similar. Then, after Congress passed the CARES Act at the end of March and expanded the unemployment insurance program to include those previously ineligible and offered extra $600 a week in payments, filings continued to explode higher even as the JOLTS series offers evidence that, in reality, layoffs, while still disastrously high, were slowing in April.

The unemployment claims figures continued to run at unprecedented levels well after lockdowns ended and businesses began to reopen in May and June. By the summer, the JOLTS reading implied a pace of layoffs consistent with a garden-variety weak labor market. As perspective, initial claims ran in the neighborhood of 400,000 a week for much of 2001 until surging toward 500,000 after the 9/11 terrorist attacks. Initial claims also ran in the 400,000’s in 2010 and most of 2011. By comparison, the JOLTS readings implied a weekly pace of layoffs in the 400,000’s in June and July and in the 300,000’s in August and September.

Exhibit 1: Initial unemployment claims and JOLTS layoffs (000s)

Source: Labor Department, BLS.

To be clear, the level of employment remained calamitously depressed throughout the summer. But surveys and anecdotes suggested that the pace of new layoffs came off sharply, as far more businesses were reopening or expanding operations than closing or curtailing their operations. The fact that initial unemployment claims remained close to 1 million per week throughout that period, higher than at any other time in history, is convincing evidence of problems with that set of data. The other trouble with reading the initial claims data is that as fraud and reporting errors were rooted out over time, it has been difficult to interpret the declining trend seen through November. How much of it was a function of a slowing in underlying payoffs and how much simply reflected a reduction in the error rate?

The JOLTS data also seems to more accurately track the subsequent softening in the labor market late last year. While the initial claims figures remained on a downtrend through November, the JOLTS layoff series bottomed out in September and began to rise again in October, when the pandemic began to get really bad in the Midwest and when payroll gains began to slow substantially. More recently, both series signal that the pace of layoffs has accelerated, likely an accurate read, as restaurants and other businesses in many parts of the country were forced to reduce their operations or close in November and December.

Heading into 2021, the initial claims data will once again be subject to new complications, as the re-introduction of supplemental payments with the passage of the December fiscal package—this time, $300 a week—likely induced many to file again, perhaps contributing to the massive rise in new filers reported for the latest week. While the JOLTS data are somewhat dated, they are probably a better proxy for the actual underlying pace of layoffs than the unemployment insurance tallies.