By Adam Davidson
A few of Donald Trump’s claims about the labor force might generously be considered gross exaggerations, but the unemployment numbers he cites appear to be wholesale inventions. Illustration by Oliver Munday
This essay is part of a series The New Yorker will be running through the election titled “Trump and the Truth.”
The unemployment rate is part of the foundational machinery of American society. You hear it or see it every month on the news. In terrible times, it gets a lot of attention. In good times, most people barely note it. The unemployment rate as we now know it is the product of more than a century of evolution, but, for as long as most of us can recall, the process behind the number has been settled. Today, Democrats and Republicans, socialists and libertarians, union officials and corporate C.E.O.s all share the same view on how to count unemployment—a hard-won consensus between social science, economics, journalism, business, and labor. But Donald Trump doesn’t buy it.
“Don’t believe those phony numbers,” Trump declared during his victory speech after the New Hampshire primary, in February. “When you hear 4.9 and five-per-cent unemployment, the number’s probably twenty-eight, twenty-nine, as high as thirty-five—in fact, I even heard recently forty-two per cent.” By June, when he did an interview on Fox News’s “Justice with Judge Jeanine,” he’d received new data. “The number five per cent is a phony number—it’s really twenty per cent, close to twenty-per-cent unemployment. That’s just a phony number to make the politicians look good.” In an August speech on the economy, he said, in reference to the official government unemployment rate, “The five-per-cent figure is one of the biggest hoaxes in American modern politics.”
That, within a six-month span, Trump’s estimates ranged from twenty, or “close to twenty,” per cent all the way up to forty-two per cent suggest he’s not using an overly rigorous model. His are statistics verified by gut feeling and bold, bald assertion. The unemployment rate is not a hoax—it is subject to a range of checks and balances. There are tens of thousands of people who are involved in constructing and reviewing it each month, and seven decades of data from which to compare each new figure with the ones that came before it.
The first President to call for an official government study of unemployment was Chester A. Arthur, who did so in 1884. Before the Civil War, there weren’t many who thought it was relevant for farmers in South Carolina to know anything about the job market among textile workers in Boston, or vice versa. But, with railroads and the telegraph and the second industrial revolution, there was the dawning realization that, in addition to whatever was going on locally, there was also this other thing, a vague force, the national economy, that seemed to make life better or worse for everyone, everywhere, in irregular cycles. One way to get a handle on this mysterious force, it was thought, was to measure the change in people who were out of work.
Arthur could not have known that he was spurring what would become one of the greatest, and most difficult, research projects in social-science history. Employment—the number of people who do have jobs—is a tricky enough stat to measure. You have to define what a job is, figure out how to count people who have one, and sort out some way of dealing with the confusing cases, like teen-agers who help out on the family farm during harvest season for a bit of allowance, volunteer workers, drug dealers and other off-the-books workers, and the list goes on. But measuring unemployment is harder still. It’s not simply the opposite of employment, a measure of all the people in the country who aren’t working. There are lots of people who are neither employed nor unemployed: children, retirees, students, stay-at-home parents, and plenty of others.
Many of the nation’s leading economists, sociologists, and statisticians spent the Great Depression years working to develop a reliable way to figure out how many people were unemployed. But they kept getting stymied: How do you properly estimate unemployment? Some compared one year’s average number of workers with a prior year’s, but this left out the chronically unemployed and ignored demographic changes. Others asked imprecise survey questions, which made the answers from survey to survey unreliable. What the social scientists and government statisticians lacked was a seemingly simple idea: the labor force. It wasn’t until 1940 that statisticians, designing that year’s census questions, began to use the term and concept of a “labor force,” which was defined as the group of people who are capable of work, want to work, and are actively either working or searching for work. It took some time to refine the precise questions, but statisticians and economists today accept unemployment data from 1948 onward as valid.
Here’s the modern-day process that Trump has called into question. Each month, the Bureau of Labor Statistics and the U.S. Census Bureau jointly conduct the Current Population Survey. Government employees conduct phone and in-person interviews with members of sixty thousand households. The selection of these sixty thousand is itself a wonder of modern sampling science, with complex computer models randomly selecting regions of the country, then buildings within those regions, then households within those buildings. Every aspect of the sampling method—save the actual names of the people being sampled—is widely disseminated and pored over by academics, advocates, journalists, and others. The questions put to the interviewees have been debated over the years—the last major change came in 1994, when, among other things, it was decided that people can only be classified as unemployed if, in their search for work, they have taken some sort of action in the previous four weeks. Simply staying at home and reading the help-wanted ads doesn’t count. (In Canada, by contrast, this form of passive searching makes you officially “unemployed.”)
The results of the Current Population Survey are held in a secured office in the Bureau of Labor Statistics. (Precautions were put in place after an incident in 1960, when a pro-Kennedy B.L.S. staffer leaked October’s unfavorable unemployment numbers to a columnist just before Election Day.) Typically, the numbers are released to the public at precisely 8:30 A.M. on the first Friday of the month. They are then pounced on by news outlets, banks, hedge funds, advocacy groups, industry lobbyists, and countless academics and economic forecasters. Many of these people don’t just look at the headline number. They dig into the thousands of detailed data points that underpin that much-reported statistic.
It’s in those deeper stats that we learn about discouraged workers, those who want work but are so despairing of finding it that they no longer search for a job. Sometimes, when Trump claims to have discovered a previously secret unemployment rate, he seems to be referring to this result, one that is publicly released by the government every month and widely discussed. There are six different official unemployment rates in the public release, allowing us to see, with precision, various trends in how people are finding and not finding work. The highest number, known as U6, includes not only the officially unemployed and the discouraged but people who have part-time jobs but wish they were working full time. This number, currently, is 9.7 per cent, which is high, but nowhere near as high as Trump’s “close to twenty per cent” or even the “eighteen to twenty per cent” he cited in his Presidential-campaign announcement speech, last year.
Another crucial statistic, also released monthly, is the labor-force participation rate, the percentage of all working-age adults who are either working or are seeking work. Trump has correctly pointed out that this number has fallen to the lowest point in nearly four decades. It peaked in 2000, when 67.3 per cent of Americans were in the labor force, and has fallen since then, speeding up during the Great Recession. It is now 62.8 per cent. But this number is no secret, either—it may be the single most widely discussed labor statistic in America today, a constant obsession in newspaper columns, on finance TV shows, and among academic economists and politicians. Much of the decrease in working people can be attributed to positive factors: more Americans live longer and have longer retirements, more young adults stay in school longer. But it is also the result of more troubling forces, including an anemic recovery from the recession.
While a few of Trump’s claims about the labor force might generously be considered merely hyperbole or gross exaggeration, the unemployment numbers he cites appear to be wholesale inventions. I have found it impossible to track down sources for most of the numbers he has “heard” as representing the “real” unemployment rate. Last year, the Washington Post’s Glenn Kessler helpfully suggested a source for Trump’s most outrageous number: forty-two-per-cent unemployment. Kessler found the same number in a 2015 blog post by David Stockman, who served as the director of the Office of Management and Budget under President Ronald Reagan. On his Web site, Stockman regularly posts writings that he readily calls “rants” and are single-mindedly focussed on his longtime obsession with the evils of Keynesianism, central banking, and the need for a return to the gold standard.
In the post in question, Stockman was writing about “the massive and systematic falsification of asset prices,” which he considers a consequence of central banking. One key part of monetary policy is the unemployment rate—the Federal Reserve’s job is to keep it low—so Stockman was attempting to show that officials have “not the foggiest idea” what the rate actually is. To find his number, Stockman ignored all that fancy stuff about labor markets and discouraged workers, and turned instead to a far simpler calculation. He started his argument by saying that, in 2014, America had two hundred and ten million people between the ages of sixteen and sixty-eight. They all, Stockman said, could have been working forty hours a week. Therefore, there should have been four hundred and twenty billion hours worked in America in 2014. Instead, a mere two hundred and forty billion hours had been worked. “Technically, therefore, there were 180 billion unemployed labor hours, meaning that the real unemployment rate was 42.9%, not 5.5%!” he concluded.
Stockman studied theology in graduate school and has no formal training in economics. In economic circles, he is a famous crank—he often makes jokes himself about the fact that his ideas are far removed from any sort of mainstream thinking. His calculations are absurd, built on a notion that everyone—including “non-working wives, students, the disabled, early retirees and coupon clippers”—could be working forty hours each week. His post does allow for the fact that America has “drifters, grifters, welfare cheats, bums and people between jobs, enrolled in training programs, on sabbaticals and much else.” But he doesn’t think these people should matter when calculating unemployment. While researching Trump’s claims, I sketched Stockman’s ideas out for Katharine Abraham, a former head of the B.L.S., who is now a professor of survey methodology at the University of Maryland. When I finished, she was silent for a moment, and then she laughed. “If you had a population of slaves, that would be a meaningful metric,” she told me.