WHY U.S. JOBS GAINS ARE SO HARD TO COUNT DURING COVID-19 / THE WALL STREET JOURNAL

Why U.S. Job Gains Are So Hard to Count During Covid-19

Government gets less data from employers, and economists have a hard time understanding how the pandemic affects consumer, business behavior

By Josh Mitchell, Anthony DeBarros and Andrew Barnett  

During the Covid-19 pandemic, economists’ estimates for job growth have often been off by hundreds of thousands of jobs./ PHOTO: JUSTIN LANE/EUROPEAN PRESSPHOTO AGENCY/SHUTTERSTOCK

The monthly U.S. jobs report moves trillions of dollars in market trades and influences key policy decisions such as the Federal Reserve’s interest-rate plans.

But during the Covid-19 pandemic, economists have had an especially hard time predicting the report’s headline number of jobs added. 

Meanwhile, the government itself has routinely made big revisions to its initial estimates.

Why? 

Two big reasons. Economists have struggled to guess the behavior of consumers and companies during unprecedented government stimulus, labor-market shifts and virus fears. 

Second, the government has seen a sharp decline in the payroll data it collects from employers. During one of the most volatile periods in recent memory, private and public-sector economists have a less firm grasp of what the labor market is doing.

During the pandemic, economists’ estimates for job growth have often been off by hundreds of thousands of jobs. 

So far this year, for instance, economists’ estimates have cumulatively surpassed the government’s initial reports by about 1.3 million jobs.

In the days leading to the report on November payrolls, economists surveyed by The Wall Street Journal believed employers added 573,000 jobs that month. 

The Bureau of Labor Statistics said the actual number was 210,000, and it added 82,000 more jobs to its initial October payrolls estimate for a new total of 546,000.

Even in normal times, predicting job growth is hard. For one, the jobs report, typically released on the first Friday of each month, is one of the first glimpses of economic activity the government offers.

A second reason: the sheer size of the labor market, currently 149 million jobs. 

A forecast off the mark by 200,000 qualifies as a big miss. 

But as a percent of overall jobs, it is a rounding error. 

For November, the median economist forecast was off by 363,000 jobs added, which represented 0.2% of total U.S. payrolls.

The pandemic has added uncertainties. 

Congress sent households trillions of dollars in stimulus payments and enhanced unemployment benefits. 

Forecasters were caught off guard by how quickly consumers spent that money, much of it on goods. 

Economists also struggled to ascertain how quickly businesses would reopen and consumers would return to restaurants and stores. 

This summer’s Delta variant added uncertainty about whether employers would cut jobs.

The government has far more data than private-sector economists do. 

But even its estimates have been far off.

BLS surveys about 145,000 employers in the middle of every month and produces an initial estimate that makes up the headline payrolls number of the jobs report. 

The agency subsequently provides two revisions to that estimate, as it collects more survey responses.

In any given month, many employers respond to the survey late or not at all. 

The BLS says its big revisions reflect a sharp drop in survey respondents during the pandemic, as often happens during times of economic turmoil. 

The response rate fell from 59% in February 2020 to 45% last month, BLS data show.

“You’re now asking companies to pony up economic data when they themselves are struggling to keep in business,” said Georgetown University professor Keith Hall, former head of the BLS. 

In many cases employees who fill out the questionnaire are preoccupied with other tasks or have been laid off.

When businesses don’t respond, the agency must guess their payroll size, said Cornell University’s Erica Groshen, who also previously headed the BLS. 

Often late survey responders are “biased” in one direction—they hired or fired en masse, Ms. Groshen said. 

When the agency finally collects their responses, it incorporates them into new estimates, leading to big revisions.

The BLS also routinely reaches out to new companies to join the survey sample, to keep up with changes in the types and size of companies that constitute the labor force. 

There, too, the agency has run into trouble. 

The share of companies agreeing to be surveyed—the so-called “initiation rate”—has fallen by half over the pandemic to 32% in October. 

This has led to a smaller sample size, which has likely led to more “noise,” or bigger swings in estimates and re-estimates, Ms. Groshen said. 

“As your sample decreases, even if it’s representative, you’re going to get more random variation because you don’t have everybody in it,” she said.

Then there is seasonal adjustment. 

The BLS tweaks the raw survey data to account for seasonal patterns of hiring, such as when retailers boost hiring for the holidays.

The pandemic disrupted those seasonal patterns. 

It led to the mass closure of schools.

Households shifted spending to goods from services, and to online from physical stores. 

Retailers likely reduced staffing at in-person stores, said Stephen Stanley, chief economist at analytics firm Amherst Pierpont. 

The shift may have affected retail employment in November, typically a big month for hiring.

The BLS reported that retailers cut 20,400 jobs in November, after accounting for seasonal factors. 

Without adjustments, the industry added 331,600 jobs.

Seasonal effects also caused November’s overall employment figure to look weak.

Without adjustments, the economy added 778,000 jobs, the largest non-seasonally adjusted gain in any November on records dating from 1939.

“The evolution of the structure of the economy has accelerated because of the pandemic,” Mr. Stanley said. 

He believes part of the problem is the labor shortage, which has prevented many businesses from hiring in months that typically include big job gains.