Imagine living in a country where you can’t trust the government’s numbers.
I don’t have to imagine it. I spent time in China and the Soviet Union in the 1980s. Numbers that reflected poorly on the government or the ruling party sometimes didn’t appear at all or were massaged carefully before they were published.
Over drinks in Beijing more than 40 years ago, a senior Chinese statistician explained the earnest efforts that he and his colleagues were making to measure the economy. “It can be awkward,” he said. “The officials don’t always want to see what we have to report.”
No informed person believed all the official data, but there were workarounds. Foreign governments used satellite photos of rice paddies, steel plants during shift changes and the lights in major cities at night, along with other measures, to estimate the size of the harvest, steel production and electricity consumption. Even rough estimates were valuable.
U.S. official data isn’t tainted like this, but it may be heading there.
President Trump didn’t like the numbers he was seeing in the latest government jobs report last Friday. So he fired the head of the Bureau of Labor Statistics, the agency in charge of producing them.
The president said the agency had manipulated the numbers for political reasons, but, as my colleague Ben Casselman has reported, most economists dismiss that claim.
The furor over statistics comes at a critical moment for the economy. So much is changing — with soaring tariffs, rising deportations and the artificial intelligence boom — that we need to know where the economy is heading. Mr. Trump has cast doubt on the official data precisely when we need it most.
As a thought experiment, take this a few steps further. Imagine that there are more firings, more presidential accusations of bias, more intervention by political leaders in technical and scientific processes. Eventually, when enough technocrats are replaced and the dire consequences of truth-telling are clear, it could become impossible to rely on the official numbers.
“Sometimes you don’t know what you’ve got till it’s gone,” Erica Groshen, commissioner of the Bureau of Labor Statistics from 2013 to 2017, said in a phone conversation. “A very important part of our information infrastructure is at risk.”
Ms. Groshen emphasized that as of now, the official data was solid. An able acting commissioner, William J. Wiatrowski, has been running the operation since the ouster of Erika McEntarfer, the widely respected and experienced commissioner. Civil Service employees are in place.
If anyone tampered with the data, Ms. Groshen said, “we’d hear from whistle-blowers, and employees would resign.” We’d know about it.
Even if those numbers couldn’t be counted on, there would be plenty of workarounds. In a big data and A.I. world, private companies already extract and produce vast streams of information on what we do, how much we do it and what it means for the economy. You can’t buy an ice cream cone without some data outfit knowing it and making inferences about your personal preferences, the state of the ice cream industry and consumers at large.
Compiling all of this information into reliable, open, public statistics, and making it available to everyone at the same time, is another matter. This information is crucial for decision-making at every level: from personal choices about what to buy and how to invest to executive decisions about the directions that companies need to take.
“Official statistics are a public good,” Ms. Groshen said. “Over the long run, there’s no replacement for the government. No private-sector company has the incentive to produce trustworthy official statistics that are available to everybody at no cost.”
The Work Goes On
At the Bureau of Labor Statistics, long-established procedures for data review and revision resulted last Friday in the downward adjustments of the jobs numbers that infuriated Mr. Trump.
Rather than accept that his own policies may have been responsible for the negative economic picture, the president attacked the bureau, saying the agency’s “numbers “were RIGGED in order to make the Republicans, and ME, look bad.”
In fact, a review by Bespoke Investment Group of the bureau’s jobs reports over seven presidential election cycles since 1998 shows that it has been impartial. Periodic revisions of jobs numbers sometimes worked in favor of Republicans, sometimes Democrats.
The Bespoke report noted that before the 2020 and 2024 elections, the bureau’s initial jobs numbers contained a “pretty significant overestimation of the job market,” benefiting the incumbents in the White House. Both lost (though Mr. Trump, of course, won his second bid for re-election in 2024). The bureau has been following transparent procedures without concern for politics, Bespoke noted.
What to Look For
There are reasons to be worried about data integrity beyond Mr. Trump’s assault on the Bureau of Labor Statistics.
Science statistics have been vanishing from government websites, with private volunteer efforts underway to preserve them. Commerce Secretary Howard Lutnick has disbanded expert advisory committees devoted to improving the quality of data on inflation, employment and gross domestic product. Funding of statistical agencies has been cut.
Because the commissioner of the Bureau of Labor Statistics doesn’t see the monthly jobs report until it has been compiled, Ms. Groshen said, the appointment of a highly political commissioner wouldn’t necessarily alter the bureau’s data.
“There are many processes and people in place who would protect it,” added Ms. Groshen, now a senior adviser at Cornell’s School of Industrial and Labor Relations. Serious trouble might be ahead, however, if a new commissioner arrived with an entourage of political appointees.
There is a precedent for that.
Census Bureau statisticians came under political pressure in 2020 during the first Trump administration. The president ordered the bureau to rapidly count all undocumented immigrants, with the aim of subtracting them from the population tally then underway. That was expected to benefit Republicans in the allocation of House seats because regions where immigrants were concentrated tended to lean Democratic. According to an account by my colleagues Emily Bazelon and Michael Wines, the Trump administration added four political jobs in the bureau’s top ranks in June 2020 to make the change happen.
In a telephone interview this past week, John Abowd, who was the Census Bureau’s director of research and chief scientist, said he and other high-level career statisticians had struggled with “the politicals” to protect the data.
“Basically," he said, “we made it impossible for them to bypass the senior executive staff and ask questions directly of the people below” — the people doing most of the actual work. The only data that the political appointees “were allowed to see were data that were, or at least in principle could have been, in final publication form.”
Adherence to well-established procedures ensured that the 2020 census was “a clean product,” Mr. Abowd said. It did not exclude undocumented immigrants. On Thursday, Mr. Trump ordered that for 2030, “people who are in our Country illegally WILL NOT BE COUNTED IN THE CENSUS.”
Early Warnings
Should government data be manipulated, academics and private sector employees would soon know it. Erik Brynjolfsson, a Stanford economist who works with the firm ADP to produce a monthly report on the labor market, said in a video chat that the latest ADP numbers and the Bureau of Labor Statistics revisions “track very closely.”
If there were politically induced distortions, he said, “many private data sources are now available” and alarm bells would go off. For example, inflation statistics are available from PriceStats, a company that evolved from a project started by scholars at M.I.T. and Harvard.
But all the private data is intended as a supplement to government statistics, not a replacement for it. “We need much more investment in this, not less,” Professor Brynjolfsson said.
Without reliable official numbers, people with inside access to the best information would be at an even greater advantage than they are now.
Losing trust in government statistics wouldn’t be the end of the world, but it would be the end of the world as we know it.