Federal Reserve officials use government data to determine when to raise or lower interest rates. Congress and the White House use it to determine when to extend unemployment benefits or send out stimulus payments. Investors place billions of dollars in bets tied to monthly reports on job growth, inflation and retail sales.
But one new study says that the integrity of that data is increasingly at risk.
The report, released Tuesday by the American Statistical Association, concludes that government statistics are currently reliable. But that could change quickly, the study warns, citing factors such as shrinking budgets, declining survey response rates and the potential for political interference.
The authors – statisticians from George Mason University, the Urban Institute and other institutions – compared the statistical system to physical infrastructure such as highways and bridges: essential, but often ignored until something goes wrong.
“We identify this kind of downward spiral as a threat, and that’s what we’re trying to address,” said Nancy Potok, who served as the U.S. chief statistician from 2017 to 2019 and was one of the report’s authors. “We’re not there yet, but if we don’t do something, that threat could become a reality, and it could happen in the not-too-distant future.”
The report, “The Nation’s Data at Risk,” highlights the threats facing federal government statistics, including data on education, health, crime and demographic trends.
But the risks to economic data are particularly striking because of the attention that policymakers and investors pay to them. Most of that data is based on surveys of households or businesses. And response rates to government surveys have fallen precipitously in recent years, as have private polls. The response rate to the Current Population Survey — the monthly survey of about 60,000 households that forms the basis for the unemployment rate and other labor force statistics — is dropped to about 70 percent in recent months, compared to almost 90 percent ten years ago.
“This is a slow-moving train wreck,” said Erica Groshen, who led the Bureau of Labor Statistics, part of the Labor Department, during the Obama administration.
The problem could get worse. Faced with rising costs to conduct its surveys and a budget that can’t keep up, the Bureau of Labor Statistics recently said it expected to reduce the size of the Current Population Survey by about 5,000 households in the next fiscal year, which begins in October.
Response rates for other government surveys — including those used to produce data on wages, vacancies and consumer spending — also fell sharply during the pandemic and have only partially recovered, if at all. Statistical offices in other countries are facing similar problems: Britain’s Office of National Statistics last year publication temporarily suspended of data from the Labour Force Survey due to concerns about its reliability.
There is no evidence that U.S. economic data has suffered a similar erosion in quality. Statistical agencies say they routinely monitor their data for evidence that declining response rates are leading to biased results and that they have confidence in their data — a conclusion supported by the American Statistical Association report.
But William Beach, who led the Bureau of Labor Statistics from 2019 to 2023, said the agency was approaching the point where it could no longer publish reliable monthly employment and unemployment data for smaller demographic groups, such as Asian Americans and teenagers, or for more sparsely populated states.
“The first thing you’ll see going forward is less data being reported,” he said.
Representatives from the Bureau of Labor Statistics and the Census Bureau said they were confident in the reliability of their data. But they acknowledged that declining response rates were a challenge.
“Issues such as privacy concerns, challenges in contacting respondents in households with only a cell phone, and the availability of respondents when contact is made have all contributed to the decline,” Ron Jarmin, deputy director of the Census Bureau, said in a written statement. “We have been conducting research and testing ways to stabilize or reverse this trend, because higher response rates mean higher quality data.”
Statistical agencies and outside experts agree that federal statistics will eventually need to incorporate more private-sector and administrative data alongside traditional surveys. That process has already begun: The Census Bureau, for example, uses data from private-sector aggregators Circana and Nielsen to supplement survey data for its monthly retail sales report.
But such an approach is resource-intensive, requiring government statisticians to collect and verify external data, figure out how to merge different sources, and test the resulting statistics to ensure their reliability—all while still producing reports using traditional methods.
Resources for that kind of innovation are scarce. Funding for the Bureau of Labor Statistics has fallen 18 percent in inflationary terms since 2009, the American Statistical Association report found. Other agencies have also seen their budgets shrink, while the costs of simply maintaining existing operations have risen.
“Agencies are tasked not just with producing this month’s numbers or this quarter’s numbers, but also with continually modernizing so that they don’t lose trust and quality,” Ms. Groshen said. “But the history of underfunding over the last 20 years certainly means that they’ve had less opportunity to do the research and develop the modernization plans that they know are needed.”
The new report calls on Congress to provide more funding to the statistical agencies, to ensure the continued reliability of their existing data and so they can expand coverage of parts of the economy that aren’t well captured by traditional statistics. For example, existing employment data has struggled to capture the rise of gig work.
But funding isn’t the only challenge, the report’s authors stressed. They said Congress should also make it easier for agencies to share data with each other, so they can operate more efficiently. And they called for more explicit protections to keep statistical agencies free from political interference.
The agencies’ independence came under increased scrutiny during the Trump administration, when the Commerce Department sought to add a citizenship question to the 2020 census — a move critics said was aimed at suppressing responses in Democratic-leaning areas. (The Trump administration denied a political motivation.)
The Supreme Court eventually blocked the attempt, but the event showed that there are few legal provisions ensuring that statistical agencies can operate independently.
“The fact is that we have relied on norms for the autonomy of statistical agencies, and sometimes norms are overturned,” said Ms. Potok, one of the report’s authors. “That’s why we say Congress should pass legislation to make this issue of professional autonomy law.”
