Private Sector Data Reveals COVID Hit Low Income Workers Hardest

The Pandemic Was Two Different Recessions

In March 2020, as the country locked down, something strange happened. High income people did not just stop going to restaurants and barbershops. They stopped spending money almost entirely. The economists Raj Chetty, John Friedman, and Michael Stepner, along with the Opportunity Insights Team, watched this unfold in real time using anonymized data from private companies like credit card processors, payroll firms, and job listing sites. What they saw was not a uniform economic collapse. It was a cascade that started at the top and crushed the bottom.

The authors built a new public database that tracked weekly consumer spending, business revenues, job postings, and employment rates down to the county and income group level (Chetty et al., 2023). This was not the usual government survey data that arrives months late. This was live. And it told a story that official statistics had missed: the COVID recession was actually two separate recessions happening simultaneously to different groups of people.

High wage workers experienced what the authors call a V shaped recession. It lasted a few weeks. They stayed home, kept their salaries, and watched their stock portfolios recover. Low wage workers experienced something else entirely. Their job losses were much larger, and they did not come back (Chetty et al., 2023). Even by December 2021, when consumer spending and job postings had fully recovered nationally, employment rates for low wage workers remained depressed in the areas that were initially hit hardest.

This is the finding that should unsettle anyone who thinks the economy bounced back evenly. It did not. The pandemic simply revealed which workers the economy could afford to lose.

Why Did High Income Spending Matter So Much?

The mechanism the authors uncovered is almost mechanical in its cruelty. High income individuals, who account for a disproportionate share of consumer spending, slashed their expenditures sharply in March 2020. They stopped going out to eat, stopped traveling, stopped getting haircuts, stopped buying services that required in person interaction (Chetty et al., 2023). This was a rational response to health risk, not a collapse in their own finances.

But that reduction in spending did not just hurt luxury businesses. It directly reduced the revenues of small businesses in affluent, dense areas. Those businesses, which rely on high foot traffic and high margin customers, suddenly had no revenue. So they laid off their employees.

Who were those employees? Mostly low wage workers. The authors found that the layoffs were concentrated among workers in the bottom quartile of the income distribution. These were the people who served the food, cleaned the offices, staffed the retail stores, and drove the rideshares that high income people had suddenly stopped purchasing (Chetty et al., 2023).

It is a perverse kind of economic gravity. The spending decisions of the wealthy, which are driven by fear and caution, become the unemployment reality of the poor. The pandemic did not create a new economic fault line. It just exposed the one that was already there.

The Data That Made This Visible

To understand how the authors pulled this off, you have to appreciate what they had access to. Most economic data in the United States comes from surveys. The government asks a few thousand households how much they spent last month, or calls a few thousand businesses to ask how many people they employ. This takes weeks to collect and months to publish. By the time you see the numbers, the recession might already be over.

Chetty and his team took a different approach. They partnered with private companies that already had granular, real time data. Credit card processors like Affinity Solutions provided anonymized transaction data from millions of consumers. Payroll firms like Earnin and Intuit provided employment data at the individual level. Job listing sites like Burning Glass Technologies showed exactly which positions employers were trying to fill and when (Chetty et al., 2023).

The result was a database that updated weekly, not monthly. It covered every county in the United States. It broke down spending, employment, and job postings by income quartile. For the first time, economists could watch a recession unfold in something close to real time, and they could see exactly who was getting hurt and when.

This methodology matters because it changes what we can know. The authors showed that low wage workers in the bottom income quartile experienced job losses that were roughly four times larger than those of high wage workers in the top quartile during the initial shock (Chetty et al., 2023). That gap was invisible in the monthly unemployment rate, which averages everyone together. The private sector data revealed the heterogeneity that the official statistics smoothed over.

The Persistent Damage That Should Not Have Happened

Here is the part that economists are still arguing about. By the end of 2021, consumer spending had fully recovered. Job postings were higher than before the pandemic. The overall economy looked healthy. But employment rates for low wage workers in the hardest hit areas were still depressed (Chetty et al., 2023).

Why? If demand for workers had recovered, why had supply not returned?

The authors offer a hypothesis that is both intuitive and unsettling. The temporary fall in labor demand during the initial lockdowns led to a persistent reduction in labor supply (Chetty et al., 2023). In plain English: when low wage workers lost their jobs, many of them did not come back. Some left the workforce entirely. Some moved to different cities or industries. Some retired early. Some started caring for family members. Some simply gave up looking.

This is not what standard economic models predict. In a normal recession, when demand picks up, workers eventually return. But the pandemic recession was not normal. It was a health crisis that became an economic crisis, and the health effects lingered even after the economy reopened. Low wage workers were more likely to get sick. They were more likely to have jobs that could not be done remotely. They were more likely to lack child care when schools closed. They were more likely to face eviction or food insecurity.

The authors are careful not to overclaim. They show that the persistent reduction in labor supply was concentrated in areas that were initially hard hit, which suggests a causal link between the initial shock and the long term damage (Chetty et al., 2023). But they cannot fully disentangle whether this was driven by health concerns, wealth effects from stimulus payments, or structural changes in the labor market. What they can say is that the recovery was not automatic. The economy did not simply snap back to its pre pandemic shape.

Did Stimulus Payments Actually Work?

The authors also used their real time data to evaluate the fiscal stimulus policies that the government deployed. This is where the paper gets practical.

Cash stimulus payments, like the $1,200 checks sent out in April 2020, led to sharp increases in consumer spending early in the pandemic (Chetty et al., 2023). People who received the money spent it quickly, especially low and middle income households. The marginal propensity to consume, the fraction of each dollar that gets spent rather than saved, was high.

But later rounds of stimulus, like the $600 and $1,400 payments sent in late 2020 and early 2021, produced much smaller spending responses, particularly among high income households (Chetty et al., 2023). Why? Because by that point, many households had already accumulated savings from earlier payments and reduced spending during lockdowns. They did not need the money to cover immediate expenses. They saved it or used it to pay down debt.

The authors found that real time estimates of marginal propensities to consume, calculated from the private sector data, provided better forecasts of the impacts of subsequent stimulus rounds than historical estimates did (Chetty et al., 2023). This is a practical insight for policy. If you want to know whether another round of checks will boost spending, look at what people are actually doing with their money right now, not what they did during the 2008 recession.

But here is the catch. The authors also found that fiscal stimulus could not restore full employment when the initial shock to consumer spending arose from health concerns (Chetty et al., 2023). No amount of cash payments could make people feel safe enough to go back to restaurants and barbershops in April 2020. The stimulus could stem the secondary decline in spending and job losses, the cascading effect where laid off workers stop spending and cause more layoffs. But it could not fix the primary problem, which was that people were afraid to leave their homes.

This is a sobering conclusion for anyone who thinks that government spending alone can solve a health driven economic crisis. It can help. It can prevent a recession from turning into a depression. But it cannot make people feel safe. Only public health measures can do that.

What This Research Does Not Prove

The authors are transparent about the limits of their data. The private sector data is not a perfect random sample of the US population. It skews toward people who use credit cards and payroll apps, which means it may undercount the poorest Americans, who are more likely to use cash and work off the books. The authors adjust for this, but the adjustments are imperfect.

The data also cannot tell us everything about why low wage workers did not return to work. The authors show a correlation between initial job losses and persistent reductions in labor supply, but correlation is not causation. It is possible that other factors, like changes in unemployment insurance generosity or early retirement among older workers, played a larger role than the initial demand shock.

There is also an open question about whether the patterns the authors observed are specific to the COVID recession or whether they reveal something general about how the economy works. Do high income spending cuts always lead to low wage job losses in a recession? Or was this unique to a pandemic that shut down in person services? The data cannot answer that yet.

What This Actually Means

• ▸The pandemic recession was not a single event. It was two parallel recessions. High wage workers experienced a brief shock and recovered quickly. Low wage workers experienced a much deeper and more persistent downturn. Any policy that treats all workers the same will miss this entirely.

• ▸The spending decisions of high income households have outsized effects on low wage employment. When the wealthy stop buying services, the workers who provide those services lose their jobs. This is not a moral judgment. It is a mechanical fact about how consumer spending flows through the economy.

• ▸Real time data from private companies can reveal economic patterns that government surveys miss. The authors showed that low wage job losses were four times larger than high wage job losses, but this gap was invisible in the monthly unemployment rate. Policymakers should invest in building these data systems permanently.

• ▸Stimulus payments work best when targeted to the people who will actually spend them. Early in the pandemic, almost everyone needed the money. Later, only low income households did. The authors showed that real time spending data can predict who will spend future payments, which means policy can be more precise.

• ▸Fiscal policy cannot solve a health crisis. Stimulus payments prevented a secondary collapse in consumer spending and employment, but they could not bring back jobs in sectors that required in person interaction until people felt safe enough to go out again. The economy cannot recover from a pandemic until the pandemic itself is under control.

The paper by Chetty, Friedman, Stepner, and the Opportunity Insights Team is not just a study of the pandemic. It is a demonstration of what economics could look like if we had better data. The old way of measuring the economy, with monthly surveys and quarterly reports, was too slow and too coarse to capture what actually happened. The private sector data showed the truth in real time: the pandemic was not one recession. It was two. And one of them never really ended.