When the Bureau of Labor Statistics released its monthly jobs report last July, the news looked good. U.S. employers had added 147,000 jobs in June, a steady pace that analysts said showed the economy’s resilience in the face of tariffs and uncertainty.
A month later, the story changed. The agency said employers had in fact added just 14,000 jobs in June, a sharp downward revision that, along with other evidence, led to renewed fears of an economic slowdown.
Last month, the story changed again. The government said employment actually declined by 20,000 in June, and that job growth over the past two years was overstated by nearly one million.
Data revisions are a fact of life for the forecasters, investors and policymakers who follow the monthly twists and turns of the U.S. economy. But the scale of recent adjustments have given some of them pause. What, exactly, is the point of a monthly jobs report that can’t reliably distinguish a solid gain from an outright loss?
It isn’t just the jobs numbers. Measures of inflation, consumer spending, home building and other aspects of the economy have seen large revisions, wild swings or other distortions and quirks. Officials at the Federal Reserve have pointed to data issues as an added source of uncertainty in their policy decisions. And a few economists have come to question the longstanding consensus that U.S. government data is the “gold standard” globally.
“I think that referring to it as the ‘gold standard’ is giving it way too much credibility,” said Steven Englander, an economist at Standard Chartered.
Most economists say data from the Bureau of Labor Statistics and other federal agencies remains among the best in the world. But they say it is important to understand its limitations.
The monthly jobs numbers, which are based on a survey of employers, are subject to significant margins of error and are regularly updated as more complete information becomes available. Sophisticated data users know to read the numbers in context — taking into account not just the latest estimate but also the recent trend, as well as evidence from other sources.
“Any one number can shift your understanding a bit, but always in the context of lots of underlying detail and lots of neighboring economic data that are all part of the overall picture,” said Jed Kolko, a senior fellow at the Peterson Institute for International Economics who helped oversee economic data at the Commerce Department during the Biden administration.
Mr. Kolko likened the monthly jobs numbers to a blood test. Few doctors would recommend radical changes based on one high cholesterol reading. Rather, they would interpret it alongside a patient’s history, fitness and other metrics, and perhaps order follow-up tests.
“That doesn’t mean you ignore the data point,” Mr. Kolko said. “It doesn’t mean that doctors shouldn’t tell you what your cholesterol number is. But it does mean that the interpretation requires other data and context.”
The monthly data last year, Mr. Kolko and others noted, mostly depicted the labor market as stuck in a “low-hire, low-fire” stasis, where employers were adding few jobs but also not laying off many workers. That basic narrative remained unchanged after the revisions.
No Signs of Political Bias
The most prominent critic of the statistical agencies has been President Trump himself. After the big downward revisions last summer, he fired the head of the Bureau of Labor Statistics, Erika McEntarfer, accusing her, without evidence, of rigging the numbers against him for political reasons.
Economists almost universally rejected those accusations, noting that there was no political pattern to the agency’s numbers and that there had been similar negative revisions under President Joseph R. Biden Jr.
Ms. McEntarfer’s firing added to fears among Democrats that Mr. Trump would try to pressure the agency into producing more favorable estimates. But there is no evidence of that happening, either. Current and former staffers say that the agency is using the same procedures as under past administrations, and that it would be impossible for the White House to interfere in its operations without detection.
Mr. Trump tried to replace Ms. McEntarfer with E.J. Antoni, a conservative economist with a history of social media posts that often appeared to distort economic statistics to support partisan positions. But the president withdrew the nomination after bipartisan backlash, and named a more traditional candidate, Brett Matsumoto, to lead the agency. Mr. Matsumoto, who must still be confirmed by the Senate, has been widely praised by economists, including former B.L.S. commissioners under presidents of both political parties.
In the interim, the agency has been led by its deputy commissioner, William J. Wiatrowski, a respected civil servant who has worked there for more than four decades.
Ms. McEntarfer herself has defended the agency in recent weeks, saying she was confident the remaining staff would speak up if they felt pressure to skew the numbers.
“You should still trust B.L.S. data,” Ms. McEntarfer wrote in a social media post last month.
Signs of Erosion
Still, just because the data isn’t politically biased doesn’t mean it is reliable. The Bureau of Labor Statistics last month said employers added just 181,000 jobs in 2025, 69 percent fewer than its initial estimate of 584,000. That followed a downward revision a year earlier that was nearly as large.
The policymakers and forecasters who follow the data most closely had plenty of warning that the revisions were coming. The B.L.S. released a preliminary version of the adjustment in September, and private-sector economists produced their own estimates even earlier. Jerome H. Powell, the Fed chair, said last fall that policymakers believed the monthly estimates were exaggerating the pace of job growth.
Economists are optimistic that the big revisions are at least partly the result of temporary factors and that the data will become more reliable going forward. The Covid-19 pandemic led to waves of business openings and closures, which are difficult for the government to track in real time, and upended the seasonal patterns that statisticians try to account for in their estimates. The surge in immigration in the early years of the Biden administration, and the sharp decline later in his term and under Mr. Trump, have broken models that were built for much more gradual demographic shifts.
“We’re getting the aftershocks of the Covid experience and how that restructured the economy,” said Michael Feroli, chief U.S. economist at J.P. Morgan. “It’s possible that normalcy will return and uncertainty will dissipate.”
But economists do have concerns about the long-term health of the statistical system, which has been strained by years of shrinking budgets, staff turnover and declining response rates to the surveys that still form the backbone of much of its data collection.
Those problems predate the Trump administration but have grown worse during it. The agencies lost hundreds of veteran employees to voluntary buyout and early retirement programs early in Mr. Trump’s term, and attrition continued after Ms. McEntarfer’s firing. Senior roles at the Bureau of Labor Statistics have been vacant for months, and staffing shortages lower in the ranks have forced the agency to limit some data collection.
The six-week government shutdown last fall further disrupted data collection and led agencies to delay or cancel dozens of data releases.
Statistical experts, including many inside the agencies, have argued for years that the system must rely less on surveys and more on data from the private sector and other parts of the government, such as the Internal Revenue Service. Such a “blended data” approach could provide statistics that are more accurate, timely and granular than survey-based estimates, and that are more comprehensive than what private companies can produce.
But developing such approaches would require time and resources that the cash-strapped statistical agencies have not had. In recent years, a number of private efforts, many funded by the Alfred P. Sloan Foundation, have been researching new approaches to data collection to try to accelerate the process.
“There’s good reason to be concerned that the quality of our statistics is going to deteriorate,” said Karen Dynan, a Harvard economist who has been involved in several of those efforts. “Even before this administration, there was reason to be concerned. The agencies have been fighting an uphill battle for years.”
Ben Casselman is the chief economics correspondent for The Times. He has reported on the economy for nearly 20 years.
The post Big Revisions Are a Reason to Question Jobs Numbers, Not Dismiss Them appeared first on New York Times.
