Firing the BLS Chief Makes the Next Good Jobs Report Suspect

At 8:30 a.m. on Friday, August 1, the Bureau of Labor Statistics^[1] reported that payrolls grew by 73,000 in July and that May and June had been revised down by a combined 258,000. That afternoon, President Trump^[2] posted on Truth Social that the commissioner, Erika McEntarfer, had “faked the Jobs Numbers before the Election,” and that he had directed his team to fire her “IMMEDIATELY.” He provided no evidence for the charge, the Associated Press noted.^[3] Labor Secretary Lori Chavez-DeRemer backed the decision, saying it would “ensure the American People can trust the important and influential data coming from BLS.” Later that day the president wrote^[4] that, in his opinion, “today’s Jobs Numbers were RIGGED in order to make the Republicans, and ME, look bad.”

Revisions that large deserve hard questions. Firing the person who published them answers none. It is shooting the messenger, and it takes the country onto dangerous ground. Elections should change policy. They should not change measurement. The Federal Reserve, Congress, employers and investors act on the same public number at the same moment, and markets need truth, not spin. Removing the commissioner does not make that number less accurate. It makes an accurate number harder to trust. Fewer employers and households answer the bureau’s surveys than a decade ago, and that calls for money, research and independent review. A firing supplies none of those, and it adds doubt about the next good report.

Start with what a commissioner controls. The commissioner is the bureau’s only political appointee, at an agency of hundreds of career civil servants, according to the AP.^[3] The Senate confirmed McEntarfer 86 to 8 in January 2024, and JD Vance, then a senator from Ohio, voted yes.^[5] Each month the payroll survey gathers reports from about 121,000 businesses and government agencies^[1] covering some 631,000 worksites, and the bureau puts the 90% confidence interval for a monthly change at roughly 136,000 jobs either way. By that yardstick, which counts sampling error alone, July’s 73,000 is consistent with anything from a loss of 63,000 jobs to a gain of 209,000. The first estimate uses the reports in hand by the deadline. Revisions come from reports that arrive later and from recalculated seasonal factors, and the bureau publishes every one. Its stated creed^[6] is almost aggressively literal: asked whether the glass is half empty or half full, “At BLS, we see an 8-ounce glass containing 4 ounces.”

The trend that deserves scrutiny sits upstream of the commissioner’s desk. Fewer of those sampled answer. The share of sampled employers responding to the payroll survey fell from 61% in April 2015 to 42.6% in March 2025,^[7] and the household survey behind the unemployment rate fell from 88.3% to 68.1%^[8] over the same decade. Resources are thinning too. The bureau stopped collecting consumer prices entirely^[9] in Lincoln, Nebraska, and Provo, Utah, in April and in Buffalo, New York, in June, and it says it makes such cuts when “current resources can no longer support the collection effort.” It has also suspended roughly 15% of the price sample,^[10] on average, in its 72 other areas. By its own simulation, losing the three cities barely moves the national inflation rate, though it warns that local and item-level indexes may get noisier. Federal Reserve Chair Jerome Powell^[11] said in June that the Fed could still do its job, but that shrinking surveys are “going to lead to more volatility.” A response rate measures participation, not error, and the bureau notes^[12] that response rates “don’t relate well to nonresponse bias.” The decline does not establish whether accuracy has suffered or, by itself, explain this year’s revisions. It does argue for paying for collection and for studying who stops answering. Replacing the commissioner does neither.

Response rates to the two surveys behind the jobs report, 2015 to 2025

Trust is the product, and it is not the same thing as accuracy. In the same answer, Powell called good economic data “a huge public good.” An accurate count keeps informing the Fed and employers even when many voters doubt it. What doubt erodes is the count’s role as common ground, the figure both parties accept before they argue about what to do. Argentina shows the full cost of losing that. Its government intervened in INDEC,^[13] the national statistics institute, in January 2007. From late 2007 to early 2011, economist Alberto Cavallo found, online supermarket prices rose at an average annual rate of about 20%, against about 8% in the official figures. The International Monetary Fund censured Argentina^[14] in 2013 over the inaccuracy of its inflation and GDP data. China stopped publishing its youth unemployment rate^[15] in August 2023 after it hit a record 21.3%, and Fu Linghui, a spokesman for the National Bureau of Statistics, said the statistics “need to be improved.” The series returned in January 2024^[16] under a new method that excludes students, at 14.9%. A statistics office that looks steerable loses the benefit of the doubt, even on the days it tells the truth.

The law permits the firing: the commissioner serves four years “unless sooner removed,” the statute says.^[17] Voters elect a president, not a bureau, and a president who answers for the economy can reasonably want people he trusts running its instruments. Vance’s communications director, William Martin, said^[18] the confirmation vote showed only that Vance at times let nominations advance despite disagreeing, and that the president “has the right to hire and fire the people he wants.” Kevin Hassett, director of the National Economic Council, said on NBC News^[19] that “the revisions are hard evidence,” called them a “historically important outlier” and rejected the charge that the president was shooting the messenger. The data give him material. By my count from the bureau’s revision table,^[20] the July report’s combined downgrade of 258,000 was the second largest in raw jobs of any report since 1979, behind only one from the pandemic spring of 2020. Every month from January through June has been revised down, by a combined 461,000. And the benchmark for March 2024, previewed at 818,000 fewer jobs,^[21] came in at 598,000 fewer.^[22] If first estimates keep flattering the economy, voters are entitled to ask why.

But the arithmetic does not follow the party in power. Under President Biden, with McEntarfer in office from January 2024,^[3] the bureau revised the first half of that year down^[20] by a combined 340,000 jobs, and its March 8, 2024 release^[23] cut 167,000 from December and January in one stroke. Trump’s^[2] August 1 post cited two of the bureau’s own revisions as grounds for his accusation: an overstatement of about 818,000 jobs for March 2024 and one of 112,000 for August and September. The bureau published that 112,000 correction itself^[24] on November 1, four days before the election.^[25] The 818,000 was a preliminary estimate the bureau posted on August 21, 2024,^[21] well before the vote; his second post^[4] placed it on November 15, after the election. Scale matters too. As a share of payroll employment,^[26] by my count, the July report’s downgrade ranks seventh since 1979, and all six larger ones corrected months during or just before the recessions of 1980, 1981 to 1982 and 2020.^[27] A downgrade that size is a reason to watch the economy more closely. A revision is a statistical agency admitting error in public, on schedule, with the arithmetic attached. That is the behavior worth protecting.

Revisions to first payroll estimates, January 2024 to June 2025

The firing changes how the next report will be read. William Beach, whom Trump appointed commissioner in his first term,^[28] called the firing “totally groundless”^[19] and predicted that if even the best possible successor reports a bad number, people will assume reality is worse because they will “suspect political influence.” Senator Rand Paul, a Kentucky Republican, said that “we have to look somewhere for objective statistics.” The bureau’s first commissioner, Carroll D. Wright, described its mandate^[6] as “the fearless publication of the facts.” Fearlessness is a lot to ask of career statisticians who have just watched a boss lose her job on the day her agency published a bad number. The next president of the other party inherits the precedent, and the risk is a cycle in which each side fires the statistician whose numbers it dislikes and the other side discounts whatever the replacement publishes. A government that punishes its messengers blinds itself.

If the White House suspects the payroll numbers are biased, the remedy is oversight in daylight, and the first step needs no new law. The Evidence Act,^[29] approved in January 2019 during Trump’s first term, requires statistical agencies to conduct “objective” statistical activities, which the law defines^[30] as “accurate, clear, complete, and unbiased.” A rule the Office of Management and Budget^[31] finalized last October requires statistical agencies to produce work “free from undue influence and the appearance of undue influence.” It has each parent department’s inspector general review compliance, including whether the agency has the resources to do its job, and report to Congress. The first scheduled reviews cannot start before December 10, 2026, but the Interagency Council on Statistical Policy may request one at any time if it sees a substantial change in circumstances. A commissioner fired on the day of a bad report is such a change, and the council should request a review now. After that, the Labor Department should restore the bureau’s two outside advisory committees, which it ended in February,^[32] saying they had “fulfilled their intended purpose.” Congress should give a bipartisan panel, three members from each party, the job of auditing the methods, so that a finding of bias, or of none, carries both parties’ names. It should also fund the surveys and research into why response has fallen. All of that is slower than a firing, and the delay is a real cost. Unlike a firing, it examines the numbers.

Take the bureau’s creed literally. On August 1 the glass turned out to hold less than the country had been told. Firing the person who read the level adds no water. It adds a question for the next time the glass reads full: who topped it up? Let the next commissioner report 3 ounces in an 8-ounce glass and keep the job.