WASHINGTON — Friday morning, the Labor Department released its grim December jobs report: 524,000 jobs were lost that month, bringing the 2008 total to 2.6 million and pushing the nation's unemployment rate to 7.2 percent, the highest since 1993.
But how exactly does the government figure unemployment?
Like many big reports, the Labor Department's monthly jobs report is based on data extrapolated from a randomly selected sample of all U.S. households, a much smaller group than the total number it seeks to determine. The department contacts 60,000 households to determine the unemployment picture for the entire work force, which consists of about 154 million Americans.
This is how almost all big surveys are done. Statistically speaking, if you're trying to get data out of a large population — like a national presidential poll — once you survey about 1,000 randomly selected people, you can be 95 percent confident that the answers you get are within 3 percentage points of the actual number. If you count every person, it's not a survey; it's a census. (And that is very expensive and time-consuming.)
Each month, 15,000 of the sample's households are switched out so the 1,500 U.S. Census workers who take the labor data aren't talking to the same people each month.
Interestingly, people who are interviewed for the monthly survey are never asked: Are you employed or unemployed? If that sounds silly, it's because the agency has criteria to classify various kinds of employment: full, part-time, unemployed but looking, unemployed but not looking and so forth.
Instead, the poll-takers ask, for example: Last week, were you on a layoff from a job? Last week, did you do any work for pay or profit?
Once the data are taken, they are weighted to take into account age, race, sex and so on, to make certain the sample numbers more closely represent the general population numbers. So, for instance, if a certain 60,000-household sample contains more residents of Hispanic origin than exist in the general population, the Hispanic sample numbers are weighted downward.
Given all this, how accurate are the unemployment reports?
"A sample is not a total count, and the survey may not produce the same results that would be obtained from interviewing the entire population," reads the Labor Department's Web site. "But the chances are, 90 out of 100, that the monthly estimate of unemployment from the sample is within about 230,000 of the figure obtainable from a total census.
"Since monthly unemployment totals have ranged from 5 million to 8 million in recent years, the possible sampling is not large enough to distort the total unemployment picture," the site reads.
At December's unemployment rate of 7.2 percent, about 11.1 million Americans are out of work.