What’s the opposite of a jobless recovery? A jobful recession

The phrase “jobless recovery” dates back to at least 1938, when the end to a recession that had begun the previous year wasn’t immediately followed by strong employment growth. For the next half-century, jobs bounced back after recessions as gross domestic product and other economic indicators did, so there was little need for the term.

But slow employment growth in the early 1990s brought it back, and the economic situation of the early 2000s seemed to pretty much define jobless recovery, with nonfarm payroll employment still falling nearly two years after the recession arbiters at the National Bureau of Economic Research had determined that the economy began growing again.

In the above chart, I’ve used the average of GDP and the lesser-known gross domestic income — another measure churned out by the US Bureau of Economic Analysis that should in theory be identical to GDP but never quite is — because the combination is probably a more accurate reflection of economic activity than GDP alone. It also has some current relevance.


Last summer, after two consecutive quarterly declines in real GDP prompted many observers to note that this met the informal definition of a recession, I and others pointed out that GDI and the GDP/GDI average had actually grown in those quarters. Subsequent revisions changed this, and during the past two quarters rising GDP has been more than offset by falling GDI. As the measures stand now (there are many revisions to come), the US economy appears to have contracted in four of the last five quarters. Meanwhile, payroll employment has grown and grown.

Last summer, the disconnect between GDP and the jobs numbers brought some talk of a “jobful recession.” For those of us who look at the GDI numbers, the contrast seems even more glaring now. But was it truly a recession then? Is it one now?

The more or less formal definition of a recession in the US is whatever the Business Cycle Dating Committee of the NBER, a nonprofit founded in 1920 and based in Cambridge, Massachusetts, says it is. So far the eight economics professors on the committee have not made that call. As two of its members, Christina Romer and David Romer of University of California at Berkeley, wrote in a 2020 paper, the chief metrics the committee has considered in making recession determinations in recent decades have been:

real GDP, nonfarm payroll employment, industrial production, real manufacturing and trade sales, and real personal income less transfers. Among monthly indicators, employment and real personal income are considered the broadest and most reliable.

Here’s how those have changed since the end of 2021 (I’ve listed them in descending order of their increase since then):

Last spring, three of the five indicators were headed down, but employment and industrial production were quite strong. Late last year, industrial production fell sharply, but other indicators were mixed. Payroll employment has kept rising throughout, joined in recent months by the other “most reliable” indicator, real personal income.

Contrast this with the picture the NBER committee saw during the recession and beginning of the jobless recovery of the early 2000s (it’s not exactly the picture they saw, because the numbers have been revised since then, but close enough):

It wasn’t an especially deep or long recession, but there were times in the spring and summer of 2001 when every monthly indicator was in decline. Coming after the boom times of 1997 to 2000, when growth in real GDP (and the GDP/GDI average) topped 4% every year, that was a big shift.

So it does seem reasonable that the NBER committee declared a recession in 2001, which it got around to just as that recession was ending in November, and hasn’t (yet) this time around. The economy has been sending mixed signals in 2022 and so far in 2023. The most comprehensive measure available, the real GDP/GDI average, has declined in four of the past five quarters, but those declines have been so modest that some or all may disappear in future revisions. Also, as long-run economic growth slows along with population growth, quarters when GDP or GDI or both decline while the economy shows few other signs of distress are going to become more common as they have in shrinking Japan.

NBER staffers started publishing recession dates in 1929, and the Business Cycle Dating Committee was formed in 1978. In their 2020 paper, the Romers conclude that the organization has done a pretty good job through the decades of identifying when the US economy shifts from its usual “regime” of growth to a different regime of “relatively rapid and substantial declines in economic activity.” But they also note that NBER hasn’t been especially clear or consistent about this, suggesting that it redefine recessions as periods with “significant and rapid increases in the shortfall of economic activity from normal rather than significant declines in economic activity” and use a statistical technique called “Markov switching” to help identify when the regime has changed.

They also propose paying more heed to a metric that used to play a big role in NBER recession calls but was dropped in the 1970s: the unemployment rate. Unemployment numbers are available just a few days after the month they measure, and unlike the nonfarm payroll data released at the same time, they aren’t revised in subsequent months. That’s because they come from a monthly sample survey of 60,000 households conducted by the Census Bureau, and there is no additional data that trickles in later. This also means the numbers can be a bit noisy from month to month, but if you use the rolling three-month average, the picture becomes clear. If the unemployment rate jumps, there’s a recession.

In 2019, economist Claudia Sahm — then at the Federal Reserve, now, among other things, a Bloomberg Opinion contributor — proposed federal stimulus payments that would kick in automatically “when the three-month average national unemployment rate rises by at least 0.50 percentage points relative to its low in the previous 12 months.” This “Sahm rule” has since taken on a life of its own as a recession signal, somewhat to the dismay of its creator. Since 1948, when the unemployment data series begins, every such increase in the unemployment rate has been followed by further increases and a recession. But the Covid-19 recession and subsequent recovery have been unlike anything the US economy has ever experienced. As Sahm wrote in December, “if the Sahm rule is ever gonna break, it’s next year.”

At present, the rule is nowhere near kicking in, with the three-month average unemployment rate just 0.03 percentage points above the 12-month low. The jobful recession-that-isn’t-quite-a-recession continues. This can’t go on forever. At some point either the job market will crack or other economic measures such as GDP and GDI will begin to catch up with it. The indicators the NBER recession committee follows seem to have been pointing lately toward the latter outcome, but most have changed direction before over the last year and a half.