The U.S. Needs Better Data. Why We Fell Behind and How We Can Catch Up.

Tracking labor and economic trends is becoming increasingly complicated. Pedestrians in New York City earlier this year. (CHARLY TRIBALLEAU/AFP via Getty Images)

Key Points

• The U.S. is experiencing a decline in the quality of federal data due to limited budgets, falling response rates, and staffing challenges.
• The Bureau of Labor Statistics lowered previously reported job gains for May and June by 258,000, and revised job creation from April 2024 through March 2025 down by 911,000.
• Adopting a blended-data approach, using private and public sources, and potentially centralizing statistical agencies could improve data quality and timeliness.

Good data matter more than ever in a complicated world, and the U.S. is falling behind.

That was hammered home during the recent 43-day government shutdown when federal statisticians stopped collecting and publishing data on an array of different topics from unemployment to inflation to retail sales.

The Federal Reserve and other agencies were forced to rely upon alternative data generated by private companies. It became clear that U.S. statistical agencies provide numbers that can’t yet be fully replicated by private data.

During the shutdown, everyone from the top White House officials to Fed policymakers to small-business owners and everyday investors lamented the dearth of reliable federal data. It was a contrast to the criticisms lobbed at agencies like the Bureau of Labor Statistics in the past year that contended the data produced wasn’t up to snuff.

As a world’s biggest economy, the U.S. needs accurate, real-time measures of activity, labor conditions, inflation, trade flows, population demographics, and revenues. The data from statistical agencies remains the “gold standard,” but that high caliber is slipping in the face of limited budgets, falling response rates and staffing challenges. That poses growing risks in the coming years, as the country faces increasingly complex changes that require leaders to make informed decisions.

The slow erosion of data quality is a “big problem” for Fed officials tasked with using economic benchmarks to guide the economy through both stable and turbulent periods, former Cleveland Fed President Loretta Mester tells Barron’s. A lot of what officials, economists and investors know about the economy is derived from looking at government statistics and using that data to calibrate economic models.

“Maybe policy today isn’t much affected, but policy tomorrow and years from now will be very much affected because the models won’t be as robust,” Mester says.

How Did the U.S. Get Here?

Producing statistics is labor-intensive, and the current methods don’t always yield timely, bulletproof information. That became clear this summer when BLS lowered previously reported job gains for May and June by a collective 258,000. That was four to five times larger than the median absolute revision since 1979.

Then the agency released revisions in early September that found the U.S. added 911,000 fewer jobs in the 12 months through March than originally reported. Instead of adding about 1.8 million positions, the U.S. economy created only 847,000 jobs from April 2024 through March 2025.

The big challenge of producing U.S. economic data is that much of the information is based on surveys of businesses and consumers. The BLS draws on the Current Employment Statistics (CES) survey of 121,000 businesses and government agencies, for example, to calculate the monthly change in payrolls and annual employment growth.

Businesses’ participation has lagged behind in recent years, making initial estimates of everything from the number of workers employed to the price of milk less reliable. The diminishing sample size typically translates into bigger revisions and a greater reliance on estimates. Not to mention the delays in obtaining an accurate view of economic conditions. That lag can have consequences when the economy shifts.

The initial response rate to the August 2025 payroll survey, for example, was approximately 57%, versus an average rate of around 70% over the past decade. Generally, by the final estimate, the collection rate tops 90%.

To be fair to the BLS, declining survey response rates are a challenge statistical agencies across the globe are grappling with. And U.S. statistical data is still considered to be among the highest quality, even at a time when BLS has lost almost 25% of its staff since February.

Making the Necessary Changes

Data-gathering methods have improved dramatically, largely due to technology. Today, you can track in real-time every purchase you make on ApplePay. Or you can get satellite snapshots of shipping containers heading into the Port of Los Angeles. Companies routinely tap these data sets to plan everything from price adjustments to targeted marketing campaigns.

Statistical agencies have been slower to adopt these dynamic data sets because they can be expensive and challenging to adopt. The BLS replaced the collection of new-vehicle prices with transaction data from J.D. Power in April 2022. It took two years to implement the change.

That’s why U.S. statistical agencies would be best-served by adopting a blended-data approach, increasing the types of data used each month alongside traditional measures to estimate the number of jobs, the unemployment rate, or even inflation levels. This is the approach that the Bureau of Economic Analysis uses to calculate GDP. The BLS also has been using various sources to help estimate the Consumer Price Index.

More recently, the Chicago Fed used roughly eight private data sources, including ADP, Indeed, and Lightcast, along with Bureau of Labor Statistics data to create estimates of the separation rate (including quits), a hiring rate for unemployed workers, and a forecast of the monthly BLS unemployment rate that’s updated twice monthly ahead of the monthly jobs report.

The labor indicators build on the blended data approach that serves as the foundation for the Chicago Fed Advance Retail Trade Summary (CARTS) series that the regional federal bank rolled out to estimate retail sales.

The process of using blended data to create these estimates has its own challenges. “One of the biggest hurdles that we face—and the statistical agencies face it too—is the data is expensive,” says Scott Brave, senior economist at the Federal Reserve Bank of Chicago. “It’s not really well-suited or well-designed to just hand off to a statistical agency without some customization, so it’s difficult to work with, it’s difficult to process, and it tends to be very expensive.”

The technological lift can be another big, upfront cost. The Chicago Fed produces the CARTS data twice a month, and that has required some significant technology investments, including cloud computing, to process the data faster while still maintaining data quality controls.

But the results of the blended data approach have been more timely, comprehensive indicators that closely track the official barometers. The Chicago Fed estimated that unemployment would rise to 4.4% on a rounded basis in September, weeks ahead of the much-delayed release from the BLS that eventually revealed the same uptick.

Beyond relying on more data sources, the BLS and other agencies should rethink how and what information they are collecting. Currently, many of the economic reports are designed to determine a particular measurement objective. The result is that consumers and companies must fill out multiple questionnaires that generate similar data sets.

Yet there are projects in play—including the U.S. Chamber of Commerce Foundation’s Jobs and Employment Data Exchange (JEDx)—that could be used to improve public-private collaboration by adopting standardized records that can be used by employers, government agencies, and individuals.

Conditions are perfect now for change; it’s a “high stakes” situation, says U.S. Chamber Foundation’s Jason Tyszko. “We’re making a lot of decisions based on not great data,” he says, but it’s “immensely feasible” to implement changes that not only lower the burden for employers, but yield more timely and robust information for statistical agencies and the public.

The chamber has piloted several tests that have made data input simpler and more standardized while maintaining the quality information needed for analysis, Tyszko says.

To increase efficiency and reduce costs, the U.S. also could move to house statistical agencies in a single national statistics agency. This would be an expensive and expansive step to improve statistical quality. But it would pay off. The U.S. is the only G-7 nation that doesn’t have a national statistics bureau.

Creating a national agency not only allows for more data-sharing and expertise, but it can also reduce costs in the long run. The most basic step would be to have the BLS, BEA, and Census all in the same agency. The USDA, Education Department, the International Trade Commission, and others have similar statistical teams that could be included.

The current setup of decentralized statistical agencies is tough, says Mike Skordeles, head of U.S. Economics at Truist. “It’s not well funded,” he adds. But a centralized agency could help alleviate some of those funding issues.

“We happen to have an administration which is less worried about doing things that are disruptive, so this is the opportunity,” says Erica Groshen, senior economics advisor at Cornell University and a former BLS commissioner.

To be sure, America’s decentralized data-agency model limits the ability to manipulate data. Moreover, quasi-redundancies can be a good thing in enabling agencies to cross-check their findings, says Tani Fukui, senior director of global economics & market strategy at MetLife Investment Management.

Yet it is hard to argue against the synergies and efficiencies of a more collaborative statistical approach.

What’s the Holdup?

Improving federal data quality will cost money. Many of the best paths forward require backtesting, technological improvements, and more staffing—at least upfront—to ensure continuity and consistency are maintained.

The other challenge is inertia. Most economists and Fed officials still consider the federal data reliable, and it is, so there’s no emergency rush to tackle these problems immediately. But when the crisis finally emerges, it may be hard to fix it in a timely fashion.

The U.S. economy is changing rapidly alongside big demographic shifts in the population. Everyone—from policymakers to business leaders to everyday Americans—will need reliable and up-to-date data to inform decisions.

That means safeguarding and enhancing the gold standard of data needs to start today.

Write to Megan Leonhardt at megan.leonhardt@barrons.com