Jobs: Can You Trust Unemployment Data?

The report from the Bureau of Labor Statistics was just as dismal as anticipated. Unemployment in January jumped to a 16-year high of 7.6 percent, as 598,000 jobs were slashed from U.S. payrolls in the worst single-month decline since December, 1974. With 1.8 million jobs lost in the last three months, there is urgent desire in Washington to boost the economy as quickly as possible: hence, the nearly trillion-dollar stimulus package wending its way through Congress. But Washington would do well to take a deep breath before reacting to the grim numbers.

Collectively, we rely on the unemployment figures and other statistics to frame our sense of reality. They are a vital part of a matrix of data that we use to assess if we're doing well or doing badly, and that in turn shapes government policies and corporate budgets and personal spending decisions. The problem is that the statistics aren't an objective measure of reality; they are simply a best approximation created by smart people working for government agencies. Directionally, they capture the trends, but the idea that we know precisely how many are unemployed is a myth. That makes finding a solution all the more difficult.

First, there is the way the data is assembled. The official unemployment rate is the product of a telephone survey of about 60,000 homes. There is another survey, sometimes referred to as the "payroll survey," that assesses 400,000 businesses based on their reported payrolls. Both surveys have problems. The payroll survey can easily double-count someone: if you are one person with two jobs, you show up as two workers. While that can overstate the number of people employed, it could also overstate the consequences of payrolls contracting (if you have two jobs and lose one, that's bad, but not nearly as bad as losing the only job you have). The payroll survey also doesn't capture the number of self-employed (estimated at 9.5 million people), and so says little about how many people are generating an independent income.

The household survey has a larger problem. When asked point-blank, people tend to lie or shade the truth when the subject is sex, money or employment. If you get a call and are asked if you're employed, and you say yes, you're employed. If you say no, however, it may surprise you to learn that you are only unemployed if you've been actively looking for work in the past four weeks; otherwise, you are "marginally attached to the labor force" and not actually unemployed.

Critics have been saying for years that current statistics underplay how dire the employment situation is because the number of workers no longer looking for work has gone up; because many of those officially employed are "involuntarily" working fewer hours; and because saying you're "self-employed" is a face-saving way of saying you're unemployed. On the flip side, you can say you're self-employed in the household survey but also be on the payroll of a company.

Not only are the headline numbers based on a statistical fiction called the "U.S. workforce," they are also an average that masks huge variations. If you are 25 years old and have no high-school diploma, your chance of being unemployed isn't 7.6%; it's 12%. If you are African-American, it's 12.6% If you have a bachelor's degree, it's 3.8%. The sharpest increase in the past year has been young men without a college degree losing their jobs, not the white-collar workers that have been the subject of such attention in the media and in Congress. In short, race, class and education are more serious issues for overall employment than the headline numbers convey.

The urge to quantify is embedded in our society. But the idea that statisticians can then capture an objective reality isn't just a will-o-the-wisp. It also leads to serious misjudgments. If the present is seeing an upsurge of entrepreneurial activity (as often happens in economic downturns) in the context of payrolls being trimmed, then massive stimulus to restore those lost jobs isn't necessary (stimulus to arrest further job loss is another story, as is insurance for those needing unemployment benefits). If the ranks of the unemployed are being swelled by the poor and those with less formal education, then that should augur for programs targeted to help those earn degrees. Democrats and Republicans can and will take sides on those issues, but a more crucial concern is that both are basing major policy decisions on guesstimates rather than looking at the vast wealth of raw data with a critical eye and an open mind.