E-Paper
Real Strains Inside the BLS Made It Vulnerable to Trump’s Accusations
The agency has faced tighter budgets, falling response rates to its surveys and a staff shortage.
President Trump’s decision to fire the head of the Bureau of Labor Statistics on Friday ignited a crisis at the U.S.’s top economic-stats agency. But it wasn’t the start of the BLS’s challenges.
In the past few years, the BLS has faced tighter budgets and falling response rates to its surveys. In recent months, it has faced staff shortages brought on by a hiring freeze Trump declared in January.
Those internal pressures made the agency more vulnerable to attack from the president, who seized on routine statistical revisions to claim that the bureau is stacking the numbers against him for political reasons.
Trump and other White House officials have offered no proof of wrongdoing, and instead cite the size of recent revisions to the jobs numbers. Economists rushed to defend the agency’s former chief, Erika McEntarfer, and said regular revisions are the hallmark of data integrity. Yet the latest revisions, when they were released last week, caused a stir in the usually staid and steady world of economic data. They point to a key challenge that has dogged the agency for years: a requirement to quickly produce accurate economic data when many households and businesses respond to surveys late, or not at all.
The update that infuriated Trump was a revision that showed the economy added 258,000 fewer net new jobs than previously thought in May and June. Monthly revisions to the job numbers are a normal process dating back decades, and spring from a difficult trade-off between speed and accuracy.
Each month the BLS surveys about 120,000 employers by phone or online to track the number of jobs in the economy, one of the world’s largest sophisticated monthly polls. Many businesses—about 30% to 40% in recent months—don’t reply on time, up from under 20% a decade ago.
At the end of each month, the BLS uses the data it has on hand to make a first estimate. Then it updates that estimate in each of the following two months, based on late replies and a revised seasonal adjustment.
If the employers replying late happen to give different information from the first batch, as happened this summer, big revisions are likely. For example, Friday’s data showed that in May and June, there were about 109,000 fewer jobs for educators at state and local governments than previously thought. Late-responding schools were responsible for the bulk of that revision, said Omair Sharif, founder of analytics firm Inflation Insights.
“The decline in response rates for the initial release, that’s just going to make the range of revisions larger,” said Jonathan Pingle, an economist at UBS. “They’re getting a significantly larger amount of incremental information after the initial deadline.”
A year ago, when he was running for his second term, Trump was enraged by a different BLS revisions process—a twice-a-year benchmarking that reconciles the monthly figures with a comprehensive set of state-level data that, in theory, includes just about every employer in the country. Using this bigger data set gives the BLS a direct look at the number of companies that have been founded and have gone out of business, a key blind spot in the bureau’s month-to-month data.
When this revision was first estimated in August 2024, it suggested that the economy had created 818,000 fewer jobs between March 2023 and March 2024 than the BLS had initially estimated. Trump said at the time that the figures showed that earlier jobs numbers had been artificially inflated to make the economy look stronger. A final revision, in March 2025, showed that the original figures had overstated the number of jobs by 598,000, not 818,000.
Like those for the survey of businesses, response rates for the BLS’s huge surveys of individual households have been eroding slowly too. A decade ago, the BLS got responses from nearly 90% of families in the monthly household survey, a phone survey that determines the unemployment rate. In recent months, the figure has hovered below 70%.
The falling response rates so far haven’t put a dramatic dent in the reports’ accuracy, economists said. But only with more funding could the BLS pursue long-term solutions, such as creating an online version of the household survey to improve response rates, said the Friends of the Bureau of Labor Statistics, a group that advocates for the bureau to get more resources.
Revisions can be a confusing headache. Yet they are an important sign that the BLS can be trusted to respond to the latest data regardless of the consequences, economists said.
Tabulating economic statistics is like “measuring a person for a suit of clothes while the person is running at full speed forward,” said David Wilcox, who previously chaired an external statistics advisory committee to another government agency that the Trump administration disbanded.
Over many decades, revisions appear to be trending lower, according to an analysis by Ernie Tedeschi, director of economics at the Yale Budget Lab and a former Biden White House economist.
In the 1970s, for example, typical months often saw the initial figure for total employment get revised by roughly 1% by the BLS. More recently, revisions have averaged closer to 0.2%, he found, looking at data through March 2024.
“To be clear, the two-month revision we just got was really big,” Tedeschi said.
The BLS has also been strained by tighter budgets and an acute staffing shortage that has forced it to rely more on guesswork to calculate the rate of inflation.
“The idea that the BLS can’t do better, I don’t buy that,” said Guy Berger, a labor economist. “But the idea that it’s easy to do better, that’s just not supported either.”
Former BLS officials say that tight funding has prevented investment that could help it modernize. The bureau’s allocation from Congress is down by 22% in inflation-adjusted terms since 2010, per a calculation by Friends of the BLS.
This year, a federal hiring freeze has prevented the BLS from replacing staff members who leave, a particular burden on data collection for the inflation report, which typically relies on hundreds of workers who visit businesses across the country to check prices.
Staffing shortfalls have forced the BLS to stop collecting prices in Lincoln, Neb.; Provo, Utah; and Buffalo, N.Y. In the rest of the country, the bureau is also missing an average of about 15% of the price data it usually collects, it said last week.
The BLS releases its next inflation report on August 12.
Write to Matt Grossman at matt.grossman@wsj.com
