The Delta Jobs Slowdown?

Four quick thoughts on what looks like unpleasant news

The delta wave is leaving its mark in the jobs data. The good news is that the surge in new cases has already crested, although it may take another month or two for the improvement to show up in the employment numbers. Moreover, some of the recent weakness in the most recent job figures may also be somewhat overstated thanks to seasonal adjustment quirks applied to public and private schools.

American employers ostensibly added just 194,000 jobs in September, after adding only 366,000 jobs in August. That would be a solid set of months if the U.S. economy were fully repaired from the damage of the pandemic, but total payroll employment is still down about 5 million compared to February 2020. Moreover, the 280,000 jobs added in August and September, on average, is far less than the 776,000 jobs added on average each month between mid-March and mid-July.1

The obvious explanation for this disappointment is the fourth wave of the coronavirus. The transmissibility of the delta variant and the refusal of a sizable minority of Americans to get vaccinated have added up to a large number of additional infections and deaths. That’s led to a predictable pullback in economic activity. The good news is that the number of new cases is turning lower so we should expect some improvement in the months ahead.


There are two ways to see this.

First, the sectors that have consistently been among the most sensitive to the virus saw employment growth slow sharply. Leisure and hospitality employers went from adding an average of 512,000 jobs a month from mid-March to mid-July to just 56,000 net new jobs a month from mid-July to mid-September. The “other services” category, which mostly consists of personal services and social activities, went from 98,000 net new jobs a month to just 9,000.

If we exclude those sectors, there has been no jobs slowdown at all. If anything, employment in the broader economy has been rising at a slightly faster pace in recent months as challenges afflicting the construction and manufacturing sectors appear to have eased. That said, the average gain of around 215,000 jobs a month since mid-May is not particularly fast given the magnitude of the outstanding shortfall.

Second, recall that the Bureau of Labor Statistics runs two surveys each month. In addition to the survey of employers that asks about who is getting paid, the BLS also asks individuals each month whether they are working or not. Crucially, the BLS also asks categorizes those with jobs into subcategories such as “employed with an unpaid absense” or the broader “with a job, but not at work”. These measures aren’t seasonally-adjusted, so they usually tell us more about when people like to take vacations than anything else.

Since the pandemic, however, there have been noticeable patterns in these data. Compared to the pre-pandemic average for any given month, far more Americans tend to be counted as “employed with an unpaid absence” or “employed, not at work” when the virus is spreading rapidly.


Some of this may reflect people staying home because they are sick with the virus, but the bigger issue is that workers on unpaid furlough may be counted as employed when they should be counted as being on temporary layoff. The BLS pointed this out early on, warning that its surveyors were “misclassifying” workers. Nobody at the BLS has tried to fix the raw numbers, instead simply warning readers of the issue in a technical note.

In July 2021, there were 86,000 fewer workers reported as “employed with an unpaid absence” than there were in all Julys on average in 2000-2019. By September, however, the number was there were 572,000 more workers counted as “employed with an unpaid absence” compared to the 2000-2019 average for Septembers. That implies a total increase in the number of workers on pandemic-related furlough of around 658,000. If that hadn’t happened, payroll job gains in August and September would have averaged something like 550,000 a month.

There are also several bits of good news to highlight.

The apparent 180,000 drop in payroll employment at public and private schools in September is probably bogus. As the BLS put it:

Recent [education] employment changes are challenging to interpret, as pandemic-related staffing fluctuations in public and private education have distorted the normal seasonal hiring and layoff patterns.

The typical pattern is that employment falls in the summer, peaks in the fall, drops a bit in the winter, and then rises again in the spring. Since the first wave of layoffs took place in the spring of 2020, the normal summer layoffs were mostly missed even though the subsequent rehiring waves weren’t really displaced from their normal spot in the calendar.

Since July, there have been 1.7 million net jobs added in public and private schools on an unadjusted basis, for a 14% gain in employment. That’s less than the typical 15% gain between July and September, but it’s hard to see it as a definitive sign of trouble.

The number of American workers who consider themselves to be “permanently” unemployed—that is, they lost a longer-term job and don’t expect to be rehired by their former employer—has plunged in recent months, from 3.2 million in May and June to 2.3 million in September. In February 2020, there were 1.3 million “permanently” unemployed Americans, which means we’ve closed about half of the gap in three months.

Finally, the drop in “permanent” unemployment has also coincided with a plunge in the number of Americans working part-time who would rather be working full-time. After spiking by 6.5 million people during the worst of the pandemic compared to February 2020, this measure of underemployment has essentially returned to normal.

There is clearly a long way to go before we fully recover from the economic impact of the pandemic, but there are still good reasons to be optimistic about the months ahead.


Throughout this piece, unless otherwise stated, I’m comparing total nonfarm payroll employment against the revised benchmark estimates for March 2021. As I explained a couple of months ago, the revisions had a big impact on the levels of employment by large sector. We’ll have to wait until February 2022 until we get detailed monthly data for all job categories.