Ian Shepherdson knew he was sticking his neck out. But last Thursday, he went public with a startling forecast: The next government labor market report would show that the U.S. economy had created 850,000 jobs in December.
For Shepherdson — and many others on Wall Street — the jobs forecast was both an unmitigated flop and an illustration of how difficult the pandemic makes it for even the most astute observers to assess the $23 trillion economy.
“Everybody wants to be pursuing precision,” he said Monday in an interview. “But even before covid, it was like hitting a moving target from a moving vehicle. Now, we’ve got a blindfold on as well.”
Indeed, economic prognostication long has provided proof for a remark attributed to baseball great Yogi Berra: “It’s tough to make predictions, especially about the future.”
But the pandemic is making a tough job even tougher. As the virus waxed and waned, it delivered some of the wildest ups and downs in U.S. labor market history. The covid recession and recovery also erased long-standing economic relationships, emptying downtown office districts, sending workers to toil in their living rooms and leaving economists fumbling for insight.
That’s resulted in an economy perched uncomfortably between the world of early 2020 and whatever reality will emerge once the pandemic is a memory. Meanwhile, familiar relationships among key economic variables have gone haywire.
Though wages are growing faster than at any point in the decade before the pandemic, for example, the number of Americans lured back into the labor market has been disappointing. Two years after covid-19 first hit the United States, the labor force participation rate remains near its lowest mark since the 1970s, when women began entering the workforce in significant numbers.
Yet a different measure, which tracks the number of employed Americans 25 to 54 years of age, relative to the total population, last year showed the fastest one-year gain since the government began keeping track in 1949.
“This is a historically unique U.S. economy,” tweeted economist Skanda Amarnath, executive director of Employ America, a nonprofit that promotes tight labor markets.
Monthly jobs gains also have been exceptionally volatile. In the last half of 2019, each month’s hiring ranged between 161,000 and 234,000 individuals. The last six months of 2021 saw swings between 199,000 and 1.1 million, a much wider band.
All this tumult has left government and private-sector economists struggling to fathom what businesses, workers and consumers will do next.
“It is extremely difficult to forecast in the current environment,” said Gregory Daco, chief U.S. economist for Oxford Economics. “It requires a heavy dose of humility. We’re more likely to be wrong, given how rapidly things are shifting.”
Even if economists are well schooled to calculate future levels of hiring, investment or trade, they are no better qualified than anyone else to foretell the next move by a shape-shifting virus — or the likely policy response in a hyper-polarized capital.
Still, after more than three decades tracking major economies, Shepherdson is no novice. He launched the research firm Pantheon, which is based in Newcastle upon Tyne in the United Kingdom, in 2012 after stints at two other firms. And he is a two-time winner, in 2014 and 2003, of the Wall Street Journal’s award for best economic forecaster.
He also wasn’t alone in botching the December jobs call. The consensus of professional economists called for more than twice as many jobs as were actually added. Moody’s Mark Zandi expected 750,000 while analysts at Goldman Sachs predicted 500,000.
Economic forecasters rely on computer models to predict the future. In layman’s language, they analyze how companies and workers have behaved in the past under various economic conditions to predict how they will behave in the future.
Ideally, economists make predictions by comparing current conditions to a previous period when the underlying structure of the economy was similar, said economist Michael Strain of the American Enterprise Institute.
But there is no precedent for determining what happens when a globalized economy operates amid a lethal respiratory virus that has killed more than 5 million people worldwide.
“The problem is what possible period do we have when the economy is close to what it is today?” said Strain, a former Federal Reserve system economist. “Some people are just using the same models they’ve been using and that’s not working.”
Shepherdson isn’t one of them. He says he realized that the pandemic had rendered traditional models — based on macroeconomic indicators such as industrial production and oil prices — “more or less useless.”
So he completely overhauled his proprietary formula to take account of the economy’s ongoing makeover, placing greater reliance upon high-frequency data from HomeBase, a provider of scheduling and payroll software for small businesses.
The British-born economist blends the HomeBase data with additional real-time input from ADP, a payroll processing company that releases its own employee count, as well as other economic inputs, to get a monthly payrolls estimate.
HomeBase data has been praised by Federal Reserve officials for its usefulness in anticipating labor market moves. But the firm has been producing its monthly reports only since the start of the pandemic, while traditional government labor market data extends to the late 1940s.
Relying on a shorter data series inevitably means a fatter margin of error, Shepherdson said.
“Until we’ve got five years of this HomeBase data, we won’t really know how useful it is,” he said, adding that he may begin emphasizing the wide range of possible outcomes when he issues future payroll estimates.
Forecasters today face a double-barreled challenge: uncertainty over where the United States is in the business cycle and what the post-pandemic economy will look like, according to Erica Groshen, senior economic adviser to Cornell University’s School of Industrial and Labor Relations.
The adjustments that government economists use to compensate for routine seasonal fluctuations, including retailers’ big holiday season hiring and firing cycle, also are misfiring amid the pandemic. Lower response rates to the government’s monthly surveys are further complicating assessments.
“Models are predicting what’s normal in a world that isn’t normal,” said Groshen, a former head of the Bureau of Labor Statistics.
This is not the first time models have failed. In 2008, many Wall Street firms were stunned when the housing implosion triggered huge losses on mortgage-backed securities. The big banks’ “value-at-risk” models, based on years of housing market data, had made no provision for housing prices to decline on a national basis.
Since previous downturns had been limited to specific regions, such as Southern California, Wall Street’s best-and-brightest assumed future declines would be similarly limited. When they weren’t, a wave of foreclosures led to massive losses on securities made up of repayment streams from hundreds of individual mortgages.
Venerable firms such as Bear Stearns, Lehman Brothers and Merrill Lynch failed or were swallowed up by rivals. American households’ net worth fell by more than $11 trillion.
“This entire market depended on finely honed computer models — which turned out to be divorced from reality,” concluded the 2011 report of the Financial Crisis Inquiry Commission.
Wall Street’s reliance on the past to predict the future has limits. Most investment solicitations carry some version of the warning “past performance is no guarantee of future results,” Wall Street’s way of saying your mileage may vary.
But if the pandemic has so unhinged traditional economic relationships and undermined the quality of key inputs, some wonder what’s the point of issuing forecasts.
“I personally think all these forecasts are just marketing for the firm,” said Barry Ritholtz, whose New York-based investment firm manages more than $1.5 billion in assets. “The least valuable part of the forecast is the actual number. It’s the process that brings something to the table.”
The prediction game won’t get any easier, at least any time soon. Fallout from the virus’s omicron variant may distort the Labor Department estimate for January hiring, due early next month, economists said.
Meanwhile, Shepherdson took Friday’s setback in good humor. That evening, he posted a self-deprecating Twitter thread, which included a photo of an Old-Fashioned cocktail labeled “there is only one answer.”
That was just “the start of the evening,” he said on Monday. “It had been a very upsetting day.”