Employment growth has been overestimated by the Bureau of Labor Statistics (BLS) over the past few years, garnering the attention of a chorus of prominent individuals, from former President Donald Trump to Federal Reserve Chair Jerome Powell.
As sizable revisions to previous months’ job data become more frequent, economic observers have presented various theories, such as poor data collection methods, to ascertain the reason behind this trend.
One of the causes might be the decline in response rates.
The federal statistics agency conducts a comprehensive approach to generating employment data every month.
The first is known as the Current Population Survey (CPS). Researchers contact approximately 60,000 households—by phone, video calls, emails, mail, and personal visits—to determine the employment status of individuals 15 and older.
They present a series of standard questions relating to work, employment activity, and job search. These findings are used to produce the unemployment rate.
The second is the Current Employment Statistics Survey (CES), which consists of interviews with 119,000 nonagricultural businesses and government agencies, representing approximately 619,000 individual worksites.
These surveys offer insights into vital industry data, such as employment, earnings, and hours for nonfarm payrolls.
These findings have been integral for decades for public policymakers, financial markets, and economists.
However, according to the BLS, fewer people respond to the survey.
For the CPS, the overall response rate was 90.3 percent in January 2013 and 70.5 percent in August 2024. For the CES, the overall response rate was 63 percent in July 2014 and 43 percent in June 2024, according to the BLS.
Officials say response rates substantially decreased at the onset of the COVID-19 pandemic and that the drop “accelerated during the pandemic.”
To ensure high-quality data, the federal agency pays close attention to the nonresponse “because it has a direct effect on data quality.”
“If the rate of nonresponse is high, it increases the chance that the final survey estimates may reflect bias,” the BLS stated.
“CPS estimates may reflect bias if the characteristics and labor force statuses of nonresponding households differ from those of responding households.”
There could be various reasons behind the nonresponse, such as privacy issues, unavailable respondents, and challenges contacting respondents.
“In recent years, it has become more difficult for Census interviewers to reach respondents, either in-person or by telephone,” the BLS stated.
In June, BLS Commissioner Erika McEntarfer suggested that budgetary constraints might force the agency to reduce the CPS sample size by 5,000.
“Budgets have not been keeping up with rising survey costs, and over the course of time, many actions have been taken to try and reduce survey costs in the field,” McEntarfer said at a Federal Economic Statistics Advisory Committee meeting.
She confirmed at the Council of Professional Associations on Federal Statistics (COPAFS) quarterly meeting in late June that the sample size would be reduced and that the change would go into effect in 2025.
In a May letter to lawmakers, COPAFS recommended giving the BLS “no less than $812 million in FY 2025.”
While McEntarfer says the leadership is exploring different options, Peterson Institute for International Economics senior fellow David Wilcox said that there “is no fully articulated strategy” for substituting the CPS should the situation worsen.
“The CPS gathers information about aspects of the labor market that are difficult or impossible to measure using ‘big data’ or alternative sources,” Wilcox said.
“If the survey response rate were to implode—and the current plan to add an internet self-response mode were to encounter similar challenges—it’s not clear how the statistical agencies would proceed.”
Hit and Miss
So far this year, the BLS initially reported 1.84 million jobs. Following the first and second revisions, employment growth was adjusted lower by about 370,000 jobs.
In August, the Department of Labor’s annual benchmark revisions updated the payroll data, showing that the economy created 818,000—or 30 percent—fewer jobs from April 2023 to March 2024.
The update—Quarterly Census of Employment and Wages (QCEW)—represented the most extensive downward revision to employment gains since the global financial crisis.
In the 12 months through March, the average monthly job gains were 174,000, down from the initial report of 242,000.
While it might seem like a new phenomenon, research by Education Advisors reveals that the BLS has overestimated job growth by, on average, 5.6 percent per year, “or about 9 million jobs.”
The federal government overestimated or underestimated growth by 20 percent or more for nearly half of all job categories, the study found.
“Due to unpredictability and uncontrollable factors that exist in economic forecasting, it could be said that landing a perfect job growth prediction is harder than landing a rocket on the moon—we should expect some wide margins of error,” the study authors wrote.
“However, job growth predictions that are grossly under or overestimated provide shocking evidence as to just how wildly wrong one of the highest U.S. economic authorities can sometimes be.”
Powell told reporters at a press conference after last week’s Federal Open Market Committee meeting that monetary policymakers considered, as part of their decision to cut interest rates by 50 basis points, that the latest batch of payroll numbers might be “artificially high” and could show further downward changes.
“We had the QCEW report, which suggests that the payroll report numbers that we’re getting may be artificially high and will be revised down,” Powell said.
Economists are debating whether the revisions have altered the portrait of the economic landscape.
Bill Adams, Comerica Bank’s chief economist, said last week that the “big downward revisions” to the job numbers “show the labor market is not doing as well as previously thought.”
“Considering the downward revisions in the August jobs report as well as the preliminary benchmark revision announced in mid-August, job growth in the last 12 months was just 157,000 per month, much slower than the 218,000 12-month average through June in the data available when the Fed met last on July 30 and 31,” Adams said in a note.
However, Dean Baker, a senior economist at the Center for Economic and Policy Research, has said that the downward adjustments do not imply a weaker labor market.
“What they tell us is that the economy is creating fewer jobs than we had previously believed,” Baker said. “But if we have high levels of employment with fewer jobs, what exactly is the problem?”
The September jobs report will be released on Oct. 4.
Early forecasts suggest that the economy created 130,000 new jobs and that unemployment ticked higher to 4.3 percent.
