The news says the recovery is here. The figure on the screen is rising, the announcers use the word, the stock market has already decided, and yet the street you live on has not moved. The plant that closed has not reopened, the friends who were laid off are still looking, the shifts have not come back, and the gap between what the country is said to be doing and what your block is actually doing can feel like an insult. You are not imagining it, and you are not the last to know because the recovery forgot you. You are last because a recovery, by now, is a thing that reaches output first and jobs later, and reaches some jobs and some places later still. This chapter is about that lag: why the economy can be growing again while the hiring has not started, why the wait has tended to get longer, and what it does to the people caught in it.
The thing to see first is that output and employment do not recover together. When a recession ends, what turns up first is production, the total value of what the economy makes, and it turns up because the firms that survived can meet returning demand with the workers they kept, running them a little harder, before they hire anyone new. Employment turns up later, sometimes much later, and the space between the two lines is the recovery that skips a street. The name economists give it is the jobless recovery, and over the four decades before the pandemic the gap did not stay the same size. It grew.
The gap that opens after output returns
The comparison can be made in the American data, where two federal agencies have counted output and jobs monthly and quarterly for decades. Take each recession since the early 1980s, measure how long output took to climb back to its previous peak, and measure how long payroll employment took to climb back to its own, and the second number is always the larger. What changed over time is how much larger. Figure 4.1 lays the two recovery times side by side for each cycle.
Read across the demand-driven recessions, the lag grew each time. In the two downturns of the early 1980s and 1990s the jobs lag was measured in months; by 2001 it was measured in years; after 2007–09 it took until 2014 for payroll employment to return to where it had been in December 2007. The 2020 pandemic recession breaks the run, and it breaks it in the direction that fits what is driving the rest. That is worth being careful about, because a chart that quietly dropped 2020 to keep a tidy line would mislead by omission. The 2020 collapse was the deepest and fastest on record, payroll falling by about 14 percent in two months as the economy was ordered shut, and then the reopening pulled most of those same jobs back as fast as any recovery in the set. The reason is the mechanism behind the others, seen in the negative.
Why firms bank the reorganization
A downturn gives a firm cover to do things it wanted to do anyway. In good times, cutting staff, closing a line, replacing a process with a machine, or consolidating two departments into one is disruptive and conspicuous, and it is resisted. In a recession all of it becomes expected, even prudent, and a manager who reorganizes is seen to be responding to conditions rather than attacking the workforce. So firms use the downturn to make the changes, shedding the roles they had come to think of as excess and rebuilding the work around fewer people. When demand returns, it is met by the higher output per worker that the reorganization bought, not by rehiring to the old level. Production comes back; the payroll does not, or not yet, because the point of the exercise was to produce the same or more with less. Figure 4.2 shows the move in outline.
This is why the 2020 shutdown recovered its jobs so quickly. A pandemic closure is not a reorganization. Firms were ordered to stop rather than choosing to restructure; they kept their operations intact as far as they could, often with public money designed to keep workers attached to their employers, and most reopened into much the structure they had closed. With little banked reorganization to absorb the returning demand, the jobs came back about as fast as they had gone. The exception fits the rule: the jobless recovery is long when the downturn has been used to make the workforce smaller, and short when it mostly has not been. It also explains why the wait grew over the demand-driven cycles. As more of each downturn’s reorganizing stuck, the cut roles staying cut when demand returned, the rehiring that a recovery once required was replaced by output the surviving workers could supply, and the gap between the two lines stretched. Who ends up carrying that gap, and why a policy that treats a band of unemployment as acceptable leaves the same people waiting, is the subject of the work volume’s chapter on why governments tolerate some unemployment.
The lag on the aggregate, the scar on the person
Everything so far has been about totals, and totals have a way of healing. Eventually, in every cycle in the figure, payroll employment did return to its old peak; the country as a whole, given enough years, got its jobs back. But the total is not the experience, and the distance between them is where the damage sits. A recovery that returns the national payroll to its previous level is not the same jobs going back to the same people. Finding work again can take years, what is found often pays less, and the loss compounds over a working life in a way the monthly total never shows.
The size of that scar has been measured. Studying men who lost stable jobs in mass layoffs, using Social Security administrative earnings records, the economists Steven Davis and Till von Wachter found that displacement in a mass-layoff event costs a worker, in present-value terms, an average of 1.4 years of pre-displacement earnings when it happens with national unemployment below 6 percent, and 2.8 years when unemployment is above 8 percent. Losing a job in a bad recession is roughly twice as costly, over the following two decades, as losing one in a good year, and the difference is not a matter of a few hard months; it is a lowering of the path a life was on that persists across the twenty years the records follow. That figure is for men displaced in mass layoffs, measured in the earnings records, and the loss it names is not only money. What a long spell without work does to a person beyond the pay packet, to health, to the shape of a day, to the sense that one is needed, is taken up in the work volume’s chapter on whether people lose their purpose without work. And what happens when a whole place loses its work at once, so that the jobs do not come back to that town even after the national total has, is the subject of the countries volume’s chapter on why some regions never recover after their industry dies. The aggregate returns; the person and the place often do not.
They call it a recovery, but the jobs never came back.
Oversimplified Moderate confidence
The claim points at something real, which is why the ruling is not that its opposite holds. Recoveries do leave jobs behind, for a long time and unevenly. In every American recession since the early 1980s output returned to its previous peak before employment did, and the wait grew across the demand-driven cycles, reaching 77 months, more than six years, after 2007–09. Anyone who answers that a rising economy promptly restores the jobs is contradicted by the record, so the reverse of this claim is the less reliable reading, and that is what keeps the ruling off backwards. What the claim gets wrong is the word “never.” At the level of the national total, the jobs did come back, in every cycle in the record, including the deep one after 2008; the payroll regained its peak, and after the 2020 shutdown it did so in a little over two years. The truth in the claim is not that the total never recovers. It is that the total recovers late, and that the aggregate hides two things it never measures. A payroll back at its old level is not the same jobs returning to the same people: men displaced in mass layoffs in a bad recession lose, on average, nearly three years of earnings over the following two decades, a scar that persists and not a delay that closes. And a national recovery can leave particular towns whose industry closed with jobs that never return at all. The claim takes a real lag and a real local permanence and states them as a universal “never” that the national figure does not bear out. It is oversimplified rather than true as stated or reversed.
Sources
- Output recovers before employment, and the lag widened: U.S. Bureau of Economic Analysis, real GDP (NIPA Table 1.1.6, chained 2017 dollars, June 2026 vintage) and U.S. Bureau of Labor Statistics payroll employment (CES total nonfarm, seasonally adjusted). Months for payroll to regain its pre-recession peak: 28 (1981–82), 32 (1990–91), 48 (2001), 77 (2007–09); the jobs lag beyond output grew from about 7 to about 47 months across those four.
- The total did recover, and 2020 is the counter-case: payroll regained its February 2020 peak by June 2022, 28 months, matching 1981–82 for the fastest in the set, because a mandated shutdown and reopening banked little of the reorganization that produces a jobless recovery. Every cycle in the record eventually regained its employment peak; the claim’s “never” holds only below the aggregate.
- The scar the aggregate hides: Steven Davis and Till von Wachter, “Recessions and the Costs of Job Loss,” Brookings Papers on Economic Activity, Fall 2011 — men displaced in mass-layoff events lose, in present value, an average of 1.4 years of pre-displacement earnings when national unemployment is below 6 percent and 2.8 years when it exceeds 8 percent (Social Security administrative earnings records; the population is male workers displaced in mass layoffs). The well-being cost is treated at Work, Chapter 11; places whose jobs never return, at Countries, Chapter 11; who absorbs the tolerated slack, at Work, Chapter 3.
- Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Evidence directness is strong: output and employment are counted directly and the recovery times are read straight from the series. The binding, weaker leg is construct match. The claim is about “the jobs,” meaning the reader’s job, town, and the same jobs going back, while the figure measures aggregate payroll returning to a level, a different construct: total employment can regain its peak while specific jobs, people, and places do not. The scarring evidence closes part of that gap but is itself scoped to men in mass layoffs. The direction is not in doubt, recoveries do leave jobs behind and the lag was real and lengthening, which keeps the ruling firm and off low; the gap between the claim’s “never” and the aggregate’s eventual return is what keeps it off high.
What the lag is, and is not
The recovery that skipped your street is neither an illusion nor a betrayal aimed at you. It is the ordinary shape of a modern recovery, in which output returns first because the firms that survived can supply it with fewer people, and the jobs return later because a downturn is when the workforce is made smaller and the reorganization is banked. Across the demand-driven recessions the wait lengthened as more of that reorganizing stuck; the 2020 shutdown, which banked little, snapped back and in doing so traced the same mechanism in reverse. Two things the aggregate does not show sit inside the wait: the years of earnings a bad recession costs the people it displaces, and the towns whose work does not come back even when the country’s does. Naming the lag is not the same as calling it a plot, and the remedy is not to deny that the total recovers. It is to see who it recovers for, and when.
The jobless recovery is one face of a larger machine. The downturn that banks the reorganization is itself part of a cycle that keeps arriving, the long calm that ends in a crash and the crash that ends in a slow climb back, and the question of why an economy that has survived a dozen such cycles keeps arranging the next one is the subject of the following chapter, which turns from the recovery that skips a street to the boom and bust that make the recovery necessary in the first place.