Economic data face imminent threats

The federal statistical system is at risk. Declining data quality would compound the harm of an economic downturn or financial crisis.

Today’s post is from economist Jed Kolko. Jed served for two years as Under Secretary for Economic Affairs in the Department of Commerce, where he oversaw the Census Bureau and the Bureau of Economic Analysis. He was previously chief economist at Indeed and Trulia.

---

The state of federal statistics is obviously not this week’s emergency, but as US tariff announcements roil financial markets, the economic uncertainty makes threats to federal statistics more urgent.

In economic crises, data serves as the indicator lights — but crises also create an incentive for governments to fudge or suppress data. When we most need transparent, comprehensive, trustworthy data, it’s at greatest risk.

So far, recent threats to economic data have been more collateral damage than intentional harm. But damage to federal statistics would be hard to reverse, and there’s no private-sector substitute for the public economic dashboard that policymakers and businesspeople base critical decisions on. Today, the call to action is to watch and document changes that could harm economic data, before real damage is done and the data is beyond repair.

The five risks to federal statistics

The US statistical system is the global gold standard. Official US statistics are widely trusted, transparent, and detailed. Economic data is both timely and accurate, and the government maintains this delicate balance with regular revisions. When the statistical agencies make errors, they acknowledge them quickly and document them fully. And the agencies themselves (and their supporters) have been upfront about ways that economic statistics need improvement and modernization.

Since the current Administration came into office, some existing risks to federal statistics have worsened and new ones have emerged. To date, the heightened risks to economic statistics have been primarily collateral damage, with many of them more anticipatory than realized.

Here are five risks currently facing our federal statistical system, roughly in order of when they emerged:

1. Disappearing data: Numerous datasets were removed from statistical websites, suddenly and without explanation. Some have been fully restored; others have been altered; and others remain missing. The still-missing or altered datasets cover topics directly related to Administration priorities, such as gender identity, environmental justice, and law enforcement. In addition, a much wider swath of data — including economic data — temporarily went missing. Large sections of the Census Bureau’s website were inaccessible for days, an apparently unintended consequence of the bureau’s need to review and confirm that no documents about gender identity were publicly accessible. As a result, in the week leading up the jobs report for January 2025, key benchmarking data on Census population estimates were unavailable.

2. Budget cuts: Statistical agencies are expecting severe cuts in the still-under-debate Fiscal Year 2025 federal budget. Those cuts would be on top of recent high attrition and Executive Orders instituting a hiring freeze and workforce reductions. Many staff who conduct surveys in-person or gather on-site data like retail prices are probationary and therefore at elevated risk of being laid off. Such cuts could force the agencies to cease production and publication of some economic indicators, or to substitute modeled estimates for hard data.

3. Lost expertise: Expertise that guides the statistical system is being lost. The seven outside-expert committees that advise the Census Bureau, Bureau of Labor Statistics, and the Bureau of Economic Analysis were disbanded. Research staff at the Social Security Administration, the Federal Housing Finance Agency, and the National Center for Education Statistics were cut or put on leave. These committees and staff provide advice and do research that improves the quality of government statistics, which is especially important when a fast-evolving economy demands new methodologies.

4. Privacy risks: Increased data sharing across the federal government could improve economic data in some ways while harming it irrevocably in others. A recent Executive Order to reduce waste, fraud, and abuse through increased data sharing could fix some inconsistencies in federal statistics that arise because statistical agencies have unequal access to the same data. But data sharing introduces clear risks to privacy and can therefore breed distrust. Just as the IRS sharing taxpayer data to support immigration enforcement could reduce future tax compliance, increased data-sharing among federal agencies could reduce future survey response rates if people think that their responses to surveys could be used against them. Federal statistics might not survive further declines in survey response rates, which have already been declining for years.

5. Reduced trust in government data: Redefining or recalculating principal economic indicators, like gross domestic product (GDP), could reduce trust in economic data. Markets interpret changes to economic indicators as attempts to hide bad news — unless the changes are transparent and follow accepted statistical practices, and enough underlying detail still gets published to calculate the indicators as traditionally defined. Recently the Commerce Secretary raised hackles when he proposed excluding government spending from GDP, which would break historical and international standards and is conceptually tricky.¹ But so far that’s resulted only in some extra charts in the GDP report with informative detail on the government's contribution to economic output. Some spin is normal, but suppressing or altering key indicators would be a trust-wrecking five-alarm fire. Governments hide or manipulate the numbers only when they’re bad, as Argentina did with inflation, Greece with public finances, and China with its youth unemployment rate. Paul Krugman warns: “America has never cooked its economic books. But there’s a first time for everything.”

A damaged federal statistical system would be very hard to replace or repair

Many of the risks to economic statistics are, at the moment, hypothetical or anticipatory.

Still, they are worth preparing for because the harm to economic data would be severe. It would be nearly impossible to replace or repair the federal statistical system, even if some functions of the statistical system are weakened only temporarily, for four reasons.

First and most important, private-sector data cannot replace official statistics. The private sector does or could do many things better than the government, but producing economic data is not one of them. Private-sector datasets — such as Indeed’s job postings, Zillow’s home value estimates, and bank credit-card transactions — have improved in recent years and were instrumental in tracking the economy during and after the pandemic. Private-sector data can be faster, more granular, and more nimble than official statistics. But those sources lack the authority, representativeness, historical comparability, and quality of official statistics. Furthermore, official statistics are used to benchmark and validate private-sector data. Private data is a complement to, not a substitute for, official statistics. If official statistics are damaged, private-sector data will fill some of the gap temporarily but will also become less reliable over time.

Second, federal statistics are a complicated, interdependent machine. Statistical agencies produce inputs into other agencies’ indicators. The monthly jobs report from the Bureau of Labor Statistics (BLS) is benchmarked to unemployment insurance data from states’ workforce agencies and to population estimates from the Census Bureau. GDP is constructed by the Bureau of Economic Analysis almost entirely from components produced by Census and BLS, as well as from Treasury, the Fed, the Energy Information Administration, and other agencies. Cuts to seemingly obscure or unrelated statistical products could leave holes in critical indicators like the jobs report or GDP.

Breaks in federal statistics can’t be repaired. Surveys ask households and businesses about recent behaviors, so data collections can’t be paused and later conducted retroactively. Interpreting data correctly depends on long, continuous time series: For instance, seasonal adjustment methods need several unbroken years of data. Finally, consistency in federal statistics is essential to building credibility and trust. A temporary break in a key series would permanently reduce trust in federal statistics.

Finally, modernizing federal statistics requires the resources and expertise that are being lost. The federal statistical system faces challenges that long pre-date this Administration, including declining survey response rates, threats to privacy and confidentiality, and societal shifts toward economic activities that are harder to measure. Innovation is urgently needed: the statistical agencies should rely more on administrative data and get expanded access to other federal agencies’ data. But new data sources or measurement approaches need to be carefully scrutinized and tested, which requires staff resources and expert outside advice. Better data sharing works only with privacy guardrails in place, which requires staff and technical tools. If the statistical system loses staff, resources, and access to outside expert advice, successful modernization and innovation can’t happen.

If economic statistics degrade, the private sector and the Administration will suffer

Of course, the earliest and loudest voices pointing out the risks to federal statistics were data users and researchers — especially those who rely on datasets that disappeared from websites starting in late January. The professional associations of statisticians, demographers, and scientists, as well as data librarians and archivists, have been quick to call out and catalog threats to public data.

But the Administration and the private sector also depend on high-quality, trustworthy federal statistics. It’s not just that “DOGE comes for the data wonks,” as the Economist newspaper put it. The data wonks are the front-line users, but the damage goes much further. Academic economists overwhelmingly agree that cuts to statistical-agency budgets, staff, and advisory committees will degrade federal statistics and impair policy-making and business planning.

An Administration that wants a less wasteful, more efficient government needs accurate and comprehensive federal statistics. Trillions of federal dollars are allocated to places and people according to statistics-driven formulas that determine who is deserving of funds. With less accurate data on population, income, or poverty, funds could be allocated less efficiently. Statistics and research inform which programs work and which are underperforming. . Also, the Administration needs accurate statistics to get credit for strong economic growth, low inflation, or job creation. Data measuring these successes needs to be widely trusted, including by the financial markets, if policy wins are to lead to private-sector investments and favorable economic expectations.

In the private sector, businesses use federal statistics for investment and marketing decisions. Official statistics on population growth, housing conditions, local demographics, and local spending patterns drive decisions about where to build factories, open stores, locate jobs, and construct housing. Companies use official statistics for benchmarking firm performance against their sector or the economy and for setting compensation and pricing strategies. Even when businesses aren’t hands-on users of federal data, they rely on research from consultants or industry associations that use official statistics. Financial markets trade on macroeconomic releases, and investors rely on clear, confident signals from the Federal Reserve, which itself depends on trustworthy economic data.

Here’s what you can do

Understandably, the risks to economic statistics aren’t the private sector’s number one issue right now. No one is paying their lobbyists to appeal to the Administration for statistical integrity. Still, the data wonks are busy doing what they do best — gathering and analyzing evidence — and you might be able to help.

If you’re looking for missing federal data, check with the Data Rescue Project. If you know of specific threats to datasets, research capabilities, or statistical agency resources, share them on the Data Foundation’s SAFE-Track and the American Statistical Association’s FedStatMonitoring sites. Finally, if you want to tell the world how important a federal dataset is for your work or for the American people, share your story on the America’s Essential Data site.

1. There can be good reasons to exclude government from statistics and focus just on private-sector economic activity. One closely watched subcomponent of GDP, final sales to private domestic purchasers, predicts future GDP better than current GDP does. And in tracking regional income inequality, I prefer looking at market income only, excluding government transfers.