Some data signal more weakness than high-profile payroll survey implies
By Gwynn Guilford of The WSJ. Excerpts:
"The payroll survey showed a gain of 339,000 jobs, while the household survey showed employment falling 310,000 and the number of unemployed leaping 440,000 to its highest level since February 2022.
The two surveys often diverge because of statistical noise or because they define employment differently. For example, the self-employed are counted by the household survey but not the payroll survey, and their numbers fell sharply in May.
Historically, economists consider the payroll survey a more reliable indicator of labor market health, except at turning points in the economy.
For example, from 2007 to 2010, a period dominated by recession and a weak recovery, the payroll survey overstated jobs by a cumulative 1.7 million, as shown by subsequent, more comprehensive tax data.
A major cause of such overestimates is related to jobs created by startups and lost by business closures. The survey has no way of capturing businesses it doesn’t yet know exist, or whether a company that doesn’t respond to the survey is ghosting it, or has closed, until many quarters later, when tax data become available.
Until then, the Bureau of Labor Statistics uses a “birth-death model” to extrapolate from recent trends how many jobs are created by new companies and lost to business closures. Its contributions are significant, said Omair Sharif, founder of Inflation Insights. “In some years, business net births make up around 40% of the payroll increase.”
But when a weakening economy is closing companies and snuffing out new business, the model might erroneously count jobs that haven’t actually been added."
"It is also possible that the birth-death model is overestimating net new business creations, as it has at previous economic turning points."
"Jonathan Pingle, chief U.S. economist at UBS, said that the level of nonfarm payroll employment at the end of 2022 was likely too high by several hundred thousand, and that the overstatement might have carried into 2023."
"The employment decline it [the household survey] showed in May could be noise given gains in the prior two months, and the series is prone to swings."
"three months of declines “would signal . . . that the labor market is at an inflection point and we are now truly shedding jobs”"
See also Comparing employment from the BLS household and payroll surveys from the BLS.
Related posts:
Why U.S. Job Gains Are So Hard to Count During Covid-19 (2021)
2 comments:
I am not sure how job numbers could ever be accurate based on the information provided in this article. Self employment is tricky and the only way to tell if their employment is halted is via unemployment benefits? As for startups, that is a another tricky metric in itself, but I'm sure they have payroll which makes no sense as to why they wouldn't be counted.
Thanks for reading and commenting.
The article says
"The survey has no way of capturing businesses it doesn’t yet know exist, or whether a company that doesn’t respond to the survey is ghosting it, or has closed, until many quarters later, when tax data become available."
So yes, they are counted. It just takes time
Post a Comment