🔒CT experts defend state and federal jobs reports, citing falling survey response rates for large revisions

Since January 2024, the state Department of Labor has issued 18 monthly labor market reports tracking nonfarm job gains or losses, total employment and the unemployment rate for the prior month. Of those, all 18 were later updated with additional data. In 15 cases, the reported jobs change was revised — sometimes significantly. A Hartford Business Journal analysis found that 10 of those revisions were negative, meaning fewer monthly jobs were gained or more were lost than first reported. Five revisions were positive, and three months saw no change. For the 18 reports, the average revision was 1,156 jobs. The largest came in March 2024, when a preliminary gain of 4,900 jobs was revised downward to a gain of 500 jobs — a swing of 4,400 jobs. Such large revisions might make you wonder about the accuracy of the reports, or whether someone might be gaming the numbers. On a national level, that certainly is the opinion of President Donald Trump, who on Aug. 1, fired Erika McEntarfer, the commissioner of the U.S. Bureau of Labor Statistics, following the release of a report showing that the nation had gained just 73,000 nonfarm jobs in July. That report also significantly revised jobs figures for the two previous months, slashing them by a combined 258,000 jobs. It was the largest two-month revision since April 2020, at the start of the COVID-19 pandemic. Trump said McEntarfer, a Biden appointee, had “rigged” the report to hurt him, without citing any evidence to support that claim. To succeed McEntarfer, Trump nominated economist E.J. Antoni, a conservative critic of the BLS who has suggested replacing monthly jobs reports with quarterly ones. An economist in Connecticut says that would not solve the issue of significant revisions, and in fact, could make things worse, while also joining a state Labor Department official in defending both the monthly jobs reports and the revisions.

Survey says

To understand why revisions are so common, you need to know how the data for the report is collected, and what factors have changed since the onset of the pandemic.

Patrick Flaherty, director of the state Department of Labor’s office of research and information, said the monthly jobs report has been based on two national surveys — an employer survey for job counts and a household survey for the unemployment rate. Each monthly national survey involves participants from all 50 states and three U.S. territories — the District of Columbia, Puerto Rico and Virgin Islands. About 121,000 businesses participate nationwide, including 1,000 in Connecticut, whose numbers the state uses for its report. According to Flaherty, the biggest change since the pandemic is that the business survey response rate has fallen “quite significantly.” Previously, the response rate was about 60%; now it’s about 40%, Flaherty said, meaning “we’re only getting about two-thirds of the amount of responses we were getting before the pandemic.” With fewer responses, Labor Department statisticians must model the missing data, Flaherty said. “It is a rigorous process; it’s well vetted,” he said. “It is not at all biased, but it is a model, so it’s not based on actual responses.” Some employers delay their survey submissions by up to a month, he added. When that late data does show up, it allows the department to replace modeled estimates with actual survey responses. That leads to the revisions. “So, when we shift from a model-based estimate to a sample-based estimate, you can get a pretty big swing in the numbers,” Flaherty said. He noted that the revised figures are ultimately more accurate than the preliminary ones because they incorporate more real data. Steven Lanza, an economist and associate professor at the University of Connecticut, adds that the data collected is just a snapshot of the jobs market, both for the state and nation. “It’s survey data, so it’s a sampling,” he said. “It’s not a complete count of the entire universe of jobs that are out there in the economy.”

‘Well-respected models’

A more complete and accurate accounting of the jobs numbers does actually happen. At the end of each year, the numbers for both the state and national jobs reports are benchmarked to business tax filings with the Internal Revenue Service, which collects employee payroll data, Flaherty said. The data also tracks businesses that have opened and closed. The updated, benchmarked jobs numbers are released in March, but Flaherty noted even those figures get revised, often because businesses amend their tax filings. “And that amends our data,” he said. “And so, even the data we benchmarked last March, suddenly there’s a slight change.” Still, the benchmarking is “very reliable data and is very accurate,” he said. “So, that is one advantage, particularly to the jobs numbers, that we have a way of setting it to reality on an annual basis.” Lanza agrees, saying anyone who suggests there might be a conspiracy to alter labor numbers doesn’t understand the difficulty involved. “When you’ve got thousands of trained professional researchers working for the Labor Department and crunching numbers, using well-established, well-respected models of data collection and analysis, it is really hard to cook those numbers,” he said.

Patriotic duty

The question, then, is whether there is a way to improve the reports and the process that produces them. Antoni’s suggestion to end the monthly reports and issue quarterly ones instead won’t solve the problem, Lanza said. “I think there would always be some effort to fill that void,” he said. “Even if the Labor Department were to delay its reporting of this information, private groups and individuals are going to take that limited information, and they’re going to make their own projections. … So, I’m not sure it would make a really huge difference.” Lanza added there is another downside to altering the labor report, because the lack of updated information could make other countries and outside investors “more suspicious and anxious” about investing in the U.S. economy. “That would be to our long-term detriment,” he said. Both he and Flaherty say that, at least for the moment, there isn’t an obvious better way to collect the data. “I think you’re always going to end up with basically some kind of poll, some kind of survey, all limited information you’re always going to have, and that’s the problem,” Lanza said. “It’s just the nature of the beast.” Flaherty said it could take years to change the process because all 50 states are involved. “The beauty of the program is it is a state-federal partnership,” he said. “But getting all of those entities … to move is difficult.” Still, changes do happen, he said. “There are processes we’re always trying to improve,” Flaherty said. “We have committees and work groups and all sorts of communication between our counterparts and our federal partners all the time to try to make improvements in data collection, data quality, and the modeling procedures.” Ultimately, though, in Flaherty’s opinion the simple answer is to do whatever is necessary to get businesses to provide more timely information. He understands that businesses may see filling out the survey as just “one more government form,” but it’s more than that, he added. “I’m hoping that we can get the message out that it’s not that you’re doing paperwork for the Labor Department,” Flaherty said. “You’re doing paperwork that’s going to help the entire public understand what’s happening in our economy, … and it really is our goal, our only goal, our only purpose, to try to provide accurate information about what’s happening in the economy.”