That's one way to leave the labor force.

Photographer: William West/AFP/Getty Images

What's Really Wrong With the Unemployment Rate

Justin Fox is a Bloomberg View columnist. He was the editorial director of Harvard Business Review and wrote for Time, Fortune and American Banker. He is the author of “The Myth of the Rational Market.”
Read More.
a | A

The U.S. unemployment rate was 4.9 percent in July, according to the Bureau of Labor Statistics. That’s the same as in July 1997. Does this mean the labor market is as healthy now as it was then, in the early days of the late-1990s boom? No, it doesn’t.

The broader U-6 unemployment measure, for example, which includes discouraged workers and involuntary part-timers, was still at 9.7 percent in July, compared with 8.6 percent in July 1997. The job growth reported by employers was, at 1.7 percent over the past 12 months, well short of the 2.6 percent pace of July 1997. And most discouragingly, the share of Americans 16 and older who were in the labor force (that is, either working or looking for a job) was just 62.8 percent, compared with 67.2 percent in 1997.

These are all good reasons to take the headline unemployment number with a grain of salt, and to question claims that happy days are here again. The unemployment rate is a problematic measure, especially when used for historical comparison.

Does this mean that the unemployment rate is some sort of “big lie” or “hoax,” charges that seem to be coming up ever more frequently? Well, if the unemployment rate is a hoax,  it’s quite the long-running one. Yes, there have been modest shifts through the decades in how unemployment is defined, the last ones in 1994. But if the rate seems less useful now than it once was, that probably has far more to do with changes in the economy and society than with anything the economic statistics-gatherers have done.

One of central questions in measuring unemployment has been how to divide those who would work if work were available from those who shouldn’t be considered part of the labor force. One can just ignore the question, and look instead to the employment-to-population ratio -- a statistic that seems to be getting more attention lately (in my columns, at least). But policy-makers and number-crunchers have for various reasons always wanted a metric that excluded those who couldn’t or wouldn’t work.

As David Leonhardt explained in a great New York Times column in 2008, this all started in the U.S. with Carroll D. Wright, who as head of the Massachusetts Bureau of the Statistics of Labor during the economic hard times of the 1870s set out to measure joblessness while excluding people he considered malingerers:

The survey asked town assessors to estimate the number of local people out of work. Wright, however, added a crucial qualification. He wanted the assessors to count only adult men who “really want employment,” according to the historian Alexander Keyssar. By doing this, Wright said he understood that he was excluding a large number of men who would have liked to work if they could have found a job that paid as much as they had been earning before.

Wright went on to become the first commissioner of what is now the BLS. And when the U.S. government finally started measuring unemployment on a monthly basis in 1940 it was with a similar understanding that you didn’t count as unemployed unless you really wanted to work. The employment survey was initially conducted by the New Deal-era Works Progress Administration, then handed over to the Census Bureau in 1943. In 1959 the BLS took over responsibility for analyzing and publishing the data, although the Census Bureau continued (and continues) to collect it in a monthly in-person and telephone survey of 60,000 households.

In the surveys of the early 1940s, those who said they would have looked for work but didn’t think there were any jobs available (what are now called discouraged workers) were counted as unemployed. There were concerns that these answers weren’t reliable -- anybody can say they want to work, but if you aren’t actually looking for work you may not mean it -- so in 1945 the survey was changed to count only those who were actively looking for work as unemployed.

As Josh Zumbrun reported in the Wall Street Journal last week, that didn’t stop Reader’s Digest from alleging in 1961 that the government was including discouraged workers to boost the unemployment rate and provide “fodder for the communist line.” Amid the resulting uproar, President John Kennedy appointed University of California at Berkeley economist Robert A. Gordon (yes, the father of noted productivity-growth skeptic Robert J. Gordon ) to head a Committee to Appraise Employment and Unemployment Statistics.

Gordon dismissed the Reader’s Digest charges as nonsense, but in an attempt to make the measurement process even less subjective his committee recommended that the employment survey-takers ask more detailed questions and only count as unemployed those who had looked for a job in the previous four weeks and could describe at least one search method that they had used. The recommendations went into effect in 1967.

Tests of the new survey in 1966 found that it reduced the measured unemployment rate by only about one-tenth of a percentage point. And the new, more detailed questionnaire gave the BLS its first actual count of discouraged workers, which Commissioner Julius Shiskin used along with other metrics in 1976 to compile a range of unemployment measures labeled U-1 through U-7. “No single way of measuring unemployment can satisfy all analytical or ideological interests,” he said at the time.

Two years later, President Jimmy Carter set up yet another panel to examine the unemployment statistics, with George Washington University economist Sar Levitan at the helm. I’m not clear on what led Carter to do this (people at the BLS seemed to think it was a waste of time), but one of the Levitan Commission’s main findings was that the classification of discouraged workers was still too arbitrary, and that people should be counted as discouraged only if they had looked for a job sometime in the past six months. (Remember, only people who have looked for work in the past four weeks are counted as unemployed.)

It took a while, but the BLS and Census Bureau finally adopted a weakened version of this recommendation in 1994, from then on counting only people who had looked for work in the past year as discouraged (with those who last looked more than a year ago counted simply as out of the labor force). In test surveys conducted in 1993, the change was found to reduce the reported number of discouraged workers from 1.1 million to just 424,000. This had no impact on the headline unemployment rate, because discouraged workers had been excluded from that calculation since 1945, but it did affect some of the alternative unemployment rates that were rejiggered in 1994 as U-1 through U-6.

Other changes made in 1994, which the BLS and Census Bureau said were occasioned partly by the growth of services and the entry of women into the workforce and partly by the availability of laptop computers for survey takers, had the effect of raising the headline unemployment rate by half a percentage point during the test year of 1993. The main reason was that changes in the wording of questions -- especially the switch from “Did you do any work at all last week, not counting work around the house?” to “Last week, did you do any work for pay?” -- led to more women being classified as both employed and unemployed.

Since 1994, there haven’t been any meaningful changes to the unemployment-survey questions. There have been lots of minor changes in methods, mainly to the population controls used in weighting the surveys, and there’s been a decline in survey response rates that could well be pushing the unemployment rate down slightly.

I don’t want to dismiss concerns that these changes, or the bigger shifts in 1994 and 1967, can make unemployment comparisons misleading -- although, as I wrote Tuesday, I would dismiss the charge that the Obama administration or its predecessors have been actively manipulating the unemployment rate.

Still, all these survey issues appear to pale in significance beside the real changes in the labor market over the past couple of decades that have made the unemployment rate less informative. Consider, for example, the epic rise in the percentage of prime-age men (ages 25 to 54) who neither have jobs nor have looked for one anytime recently.

In the mid-1950s this was about 700,000 men. In July 1997 it was about 4.5 million. Now it’s around 7 million. Some of these guys are out of the labor force for positive reasons: They’re taking care of kids, they’ve gone back to school, they’ve retired early to take up windsurfing full time. But they’re a minority. As Leonhardt put it in his column back in 2008, just as the labor-force participation rate was beginning to take another big dip:

Instead, these nonemployed workers tend to be those who have been left behind by the economic changes of the last generation. Their jobs have been replaced by technology or have gone overseas, and they can no longer find work that pays as well.

That these people aren’t numbered among the unemployed doesn’t constitute a lie or a hoax -- there are good reasons to try to keep the measurement of unemployment consistent over time, and counting labor-force dropouts as unemployed would be a dramatic break from the practice that has prevailed since 1945. But to really understand the long-term health of the job market, there are better places to look than the unemployment rate.

  1. It’s not.

  2. The other big source of monthly employment data, the establishment survey of 146,000 businesses, nonprofits and government agencies from which the nonfarm payroll numbers are derived, is conducted by the BLS.

  3. It’s worth noting that during World War II the U.S. was plagued by labor shortages, meaning that policy-makers wanted a measure of how many people really, truly were available and willing to work.

  4. The senior Gordon was also the husband of noted health, employment and education economist Margaret S. Gordon, who died in 1994, and the father of noted labor economist David M. Gordon, who died in 1996. In the father’s obituary in 1978, the New York Times said the family “has been called the Flying Wallendas of economics.”

  5. U-3 is the headline unemployment rate.

  6. One fun example of the impact that even minor differences in survey design can have comes from Canada, where the definition of looking for work is less demanding than here. As University of Ottawa economist Miles Corak put it:

    The crux of the matter is that flipping through the want-ads in a newspaper gets you classified as unemployed in Canada, but not in the United States.

    Corak estimated that, as of early 2012, Canada’s unemployment rate would have been 1.1 percentage points lower if its statisticians used the U.S. definition of looking for work.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Justin Fox at

To contact the editor responsible for this story:
Stacey Shick at