Why the Bureau of Labor Statistics regularly revises jobs data

Boston Globe

a day ago

But they also say the bureau, like statistical agencies around the world, is struggling to measure a rapidly changing economy and needs to adopt new techniques to produce reliable data.
Advertisement
Those concerns, normally the stuff of statistics conferences and economics podcasts, burst into public view Friday — first with the shocking size of the revisions, which shaved more than a quarter-million jobs off the recent total, and then with Trump's decision to fire the official in charge of the agency, Erika McEntarfer.
Trump claimed without evidence that McEntarfer had 'rigged' the numbers to hurt him politically. His aides, in TV interviews over the weekend, argued that she had deserved to lose her job because the agency had failed to produce accurate statistics on her watch.
Economists across the political spectrum decried McEntarfer's firing, and dismissed the notion that she would, or even could, meddle with the monthly jobs figures. They said that while it was understandable that revisions — especially large ones — could foster skepticism, the data produced by the agency remained reliable.
Advertisement
The challenges facing the agency, they added, long predate McEntarfer's leadership and will be difficult for her successor to address without increased funding. Instead, the Trump administration has proposed further cuts to the agency's budget.
'The solution to that is not to fire the head of BLS but to make the immediate investments in the federal statistical system,' said Nancy Potok, who served as chief statistician of the United States during the first Trump administration.
Measurement challenges
There is a fundamental tension inherent in all economic data: accuracy vs. timeliness.
Policymakers, investors and businesses want information as quickly as possible so it can inform their decisions. But the most complete data is often based on tax returns, Social Security filings or other records that aren't available until months or years later.
Revisions are the imperfect solution to this problem. Statistical agencies release preliminary estimates of job growth, inflation, gross domestic product and other measures, then revise them as more complete data becomes available.
'If we waited for perfect data, our statistics wouldn't be timely,' said Jed Kolko, who oversaw economic data at the Commerce Department during the Biden administration. 'If we only looked at timely statistics, they wouldn't be accurate. Revisions are how we get timely statistics and accurate statistics.'
In the case of job growth, the statistics bureau conducts a monthly survey of more than 100,000 employers to collect data on how many people they employ, hours worked and wages paid. It revises the numbers twice in subsequent months, as data comes in from employers that responded late.
Advertisement
Once a year, the agency releases a larger 'benchmark revision' that aligns the survey data with more definitive — but much less timely — records from state unemployment insurance offices.
Regular users of the jobs data know to expect revisions and adjust accordingly, averaging the monthly data over time and combining multiple sources to get a more complete picture.
The downward revisions Friday helped bring the monthly payroll numbers more closely in line with other measures of the labor market. But to people who don't follow economic news closely, such big changes can be shocking.
'These numbers are intended to speak to lots of different audiences,' said Guy Berger, a labor economist who has followed the monthly jobs reports for years. 'It's hard to communicate to normies.'
Shifting patterns
The revisions announced Friday were the biggest in decades, outside the pandemic. But on the whole, revisions have grown smaller over time, according to an analysis by Ernie Tedeschi, director of economics for the Yale Budget Lab. The average revision during Trump's first six months in office is almost identical to the average during the four years of the Biden administration.
What has been unusual during the early months of Trump's term is that the revisions have been consistently in the same direction: down. That was true for much of President Joe Biden's term as well, other than a period of large positive revisions in the immediate aftermath of the COVID-19 pandemic.
A pattern of negative revisions is often a bad sign for the economy. The initial monthly payroll estimates repeatedly understated job losses early in the Great Recession, for example, and did the same during the pandemic. Some forecasters have suggested that the recent revisions could be a sign that the labor market is weakening again.
Advertisement
There could be other factors at play, however. To produce timely estimates, statistical agencies have to make assumptions to fill in the gaps left by incomplete data. But the pandemic and its chaotic aftermath disrupted hiring patterns. Big shifts in immigration did the same. It is possible that the models used by the statistics bureau haven't kept up with rapid changes in the labor market in recent years.
'Given the atypical features of this cycle, it is not surprising to me that we've had some large revisions,' Berger said. 'There are fundamental issues that are worth talking about finding ways to improve.'
Trying to adapt
The statistics bureau has acknowledged a need to modify its methods, and has already begun making changes.
Response rates to government surveys have been falling for years, a decline that accelerated during the pandemic. Lower response rates don't yet seem to have had a major effect on the reliability of the statistics, but experts say it is only a matter of time before they do. And existing measures don't always reflect the reality of the modern labor market, such as the rise of gig work.
'It would be better if we were doing a better job capturing some of the changes in the economy,' said Katharine G. Abraham, who led the agency during the Clinton and George W. Bush administrations.
Federal statistical agencies have been trying to improve response rates and rely less on surveys. Major companies can now report their payroll numbers directly to the statistics bureau, for example, rather than needing to fill out a survey form. And the bureau is using private-sector data on prices for cars, gas and medical services rather than relying on traditional survey methods.
Advertisement
Such approaches may prove to be more accurate and less expensive. But not right away — changing methods is painstaking, requiring experimentation and testing to make sure that the new data sources are comparable to the old ones. That requires resources and expertise.
The bureau's budget has declined about 18 percent in inflation-adjusted terms since 2009, according to a 2024 report from the American Statistical Association, a professional group. That report warned that budget cuts were making it hard for the bureau and other statistical agencies to modernize their operations.
The problem has grown worse since Trump took office. A federal hiring freeze, combined with the buyout and early retirement programs offered to many government workers, has led more professionals to leave federal agencies. Trump also disbanded advisory committees through which private-sector experts provided feedback and guidance to the statistical agencies. And his budget proposed steep cuts to the bureau's funding.
A Labor Department spokesperson, who declined to be named, said the administration had offered options for addressing declining response rates and other challenges, and had provided funding to prevent the cancellation of an annual survey on displaced workers, at the bureau's request.
In June, the bureau said it had stopped collecting price data in three cities, and cut back data collection across the country because of funding constraints. That forced the agency to rely more on statistical techniques to fill in the gaps, potentially reducing the reliability of its data. (The Labor Department spokesperson said the department had learned about the cuts through media reports.)
Advertisement
'If you were serious about modernizing and improving official statistics, you would be spending more money, you would invest in this, you would make sure that the experts who work at these agencies are given incentives to stay and you would take every advantage of opportunities to get feedback and input from the private sector,' Kolko said.
This article originally appeared in
.