Economic statistics were never designed to measure the sudden shutdown and restart of large segments of the U.S. economy. Still, if there is one question that the government seemingly should be able to answer, it is this: How many Americans are receiving unemployment benefits?
Since the start of the pandemic, however, federal data on the unemployment insurance system has been plagued by errors, double counting and other issues. And even after the initial flood of layoffs slowed, the problems have only grown in recent weeks, in part because of an apparent spike in fraudulent claims for benefits.
The biggest problems appear to involve Pandemic Unemployment Assistance, a program created by Congress in March to cover freelancers, self-employed workers and others who are left out of the regular unemployment system. Federal data implies that nearly 15 million Americans are now receiving benefits under the program, but some economists believe that overstates the true number by millions.
The scale of the overcounting issue varies by state. In Texas, figures for Pandemic Unemployment Assistance claims closely match the federal government’s. But in Montana, the state says just 9,000 people are receiving benefits under the program, versus the more than 60,000 reported by the federal government.
The biggest problems, at least in absolute numbers, are in California. The federal data suggests that nearly seven million Californians are receiving pandemic benefits. The state’s data shows that number is under two million.
The counting issues don’t change the broad contours of the crisis: By any measure, millions of Americans are relying on unemployment benefits to buy groceries and pay rent. But they do make it harder to answer basic questions about how quickly the economy is improving and how successful government programs have been at mitigating the damage.
“This does really underscore just how important it is that we make key investments in our data infrastructure, because now we know what it feels like when we don’t have good data,” said Heidi Shierholz, director of policy for the Economic Policy Institute.
The complex unemployment system makes it hard to get a national picture.
The United States doesn’t have an unemployment insurance system. It has 53 systems, one for each state plus the District of Columbia, Puerto Rico and the Virgin Islands. Each operates independently, with its own rules and procedures, subject to policies set at the federal level.
State unemployment offices report data to the U.S. Labor Department, which compiles the numbers into a weekly report. One number in that report, known as “continuing claims,” counts filings of people who have previously filed for benefits and have remained unemployed since the previous week.
That figure is often treated by economists as an estimate of the number of people receiving unemployment benefits. But that isn’t actually what it measures, at least not directly. It counts applications, not all of which are approved. And rather than counting the number of individuals applying for benefits, it counts the total number of weeks of benefits they apply for.
That distinction doesn’t matter much in normal times, when most people apply for benefits on a weekly basis and are quickly approved. But because benefits are paid retroactively, if there are delays processing applications, people can end up applying for multiple weeks of benefits at once, skewing the continuing-claims number.
That seems to be a particular issue in California, according to a new analysis of state unemployment data by researchers at the California Policy Lab. Some of the recent flood of applications for Pandemic Unemployment Assistance there are from people saying they lost jobs in the early weeks of the pandemic, meaning they could be owed months’ worth of benefits, said Till von Wachter, an economist at the University of California, Los Angeles, who was an author of the Policy Lab analysis.
State officials say many backdated claims in that new flood may be fraudulent. But others may not be, Mr. von Wachter said. Someone in the film industry, for example, might not have applied for benefits right away last spring, on the assumption that business would bounce back relatively quickly. But now, with no reopening in sight, the worker might decide to file — and to claim, legitimately, to have been out of work since April.
Weekly unemployment reports are a timely but flawed measure of the economy.
Weekly unemployment filings were not intended to be an economic indicator. They aren’t collected by the Bureau of Labor Statistics, the Labor Department agency that produces the unemployment rate and related measures, and they aren’t subject to the quality controls applied to official statistics.
Instead, the data is collected by the states and reported to the Employment and Training Administration, a Labor Department agency charged with overseeing the states’ unemployment systems. Asked about the data discrepancies, the department said the numbers were intended primarily for administrative purposes, like allocating federal funding for state employment agencies.
Economists pay attention to unemployment filings because they’re often an early-warning system for trouble in the labor market. But once the alarm has been sounded, economists usually turn to more reliable monthly and quarterly data to get a more complete picture of what is going on.
The speed of the present crisis has put a premium on timely data. At the same time, state unemployment systems, many of which run on decades-old software, were overwhelmed by the flood of applications for traditional unemployment benefits, while carrying out a new program that covered a separate category of workers. That made it hard for them to report accurate data.
“It’s a fast number, but that doesn’t make it a good number,” said Eliza Forsythe, a University of Illinois economist who studies unemployment.
Data on the pandemic assistance program has been especially confusing.
The standard unemployment system leaves out a lot of people: freelancers, self-employed workers and people with too little work history to qualify. (That can include some low-paid part-time and low-wage workers.) The Pandemic Unemployment Assistance program is meant to fill that gap.
By many measures, the program has been a success, helping millions of workers who would otherwise have had no source of income. But data on the program has been troubled from the start. Many states took weeks to get the program up and running, and when they did, many did not begin reporting data right away.
Once the data started coming in, it was often hard to interpret — some states would report thousands of recipients one week, then zero the next. Processing backlogs made it hard to separate recent job losses from layoffs that happened early in the pandemic.
“The claims that are coming in are borderline nonsensical sometimes,” said Kathryn Anne Edwards, an economist at RAND who studies the unemployment system.
The actual number of people receiving pandemic benefits is probably below 10 million.
The Labor Department says about 13 million people are receiving benefits under regular state unemployment programs. An additional 1.5 million or so are covered by various programs for people whose regular benefits have run out. Economists consider those figures generally reliable.
But few economists believe the federal government’s figures for the pandemic assistance program, which on paper is now larger than the regular state programs.
California’s data alone indicates the count of continuing claims could be overstated by about five million. Other states report their own discrepancies that, taken together, suggest the federal count could be inflated by a further two million or more, although too few states are reporting individual-level data to allow for a precise estimate.
Other sources, including surveys and federal spending data, likewise suggest that the number of people receiving benefits under the pandemic program is below 10 million, and perhaps as low as five million. That would mean the number of people receiving unemployment benefits of any kind right now is 20 million to 25 million, rather than the 30 million suggested by federal continuing-claims data.
Some people are still left out of the programs.
The official unemployment insurance figures almost certainly overstate the number of people receiving benefits. But they might still underestimate the number of people whose livelihoods have been affected by the pandemic.
Some groups, like undocumented immigrants, are excluded from the unemployment system. Others have been improperly denied benefits, or have been unable to apply. Surveys and other evidence suggest a sharp increase in food insecurity during the pandemic, a sign that even the expanded benefit programs aren’t reaching everyone in need.
“It’s both an overcount and an undercount at the same time,” Ms. Forsythe said.
The good news is that there is little evidence that the recent increase in unemployment claims, particularly in the pandemic program, reflects a real-world increase in the rate of job losses. While layoffs are continuing, most public and private data sources show a gradual improvement in the labor market. But those same sources suggest that progress has slowed in recent weeks, and that the absolute level of joblessness remains high.