Forecasting and estimating the number of jobs lost in a given month.

Forecasting and estimating the number of jobs lost in a given month.

Forecasting and estimating the number of jobs lost in a given month.

Answers to your questions about the news.
Sept. 3 2009 6:45 PM

How Does the Government Count Lost Jobs?

And why are there so many different estimates?

The U.S. economy shed 298,000 jobs in August, according to a report released by Automated Data Processing on Wednesday. The estimated job losses were heavier than the consensus forecast of 250,000. The official job numbers from the Bureau of Labor Statistics will be released Friday morning. Where do all these figures come from?

Monthly surveys of employers, proprietary payroll information, and lots of market research. Each month, the government begins its calculation of total U.S. jobs by sending surveys (PDF) to about 400,000 employers. All businesses with more than 1,000 workers are automatically included, and the remainder of the surveys are distributed in such a way as to accurately reflect the national economy. Every business that receives the questionnaire must report how many employees it had, how many hours those people worked, and the total amount of wages paid out during a particular pay period. Once the data are in, the government statisticians make a series of corrections; for example, they assume that some of the nonresponders went out of business.


Because hundreds of billions of dollars can be added to or taken out of the stock market based on the jobs report, financial firms strive to get an idea of what the government numbers will be ahead of time. Automated Data Processing releases its estimate two days earlier and makes use of the fact that it prints the paychecks of about one in six private-sector employees. The company bases its estimate on 400,000 employers (the same number used for the government estimate) selected from its client list. Since this sample may not accurately reflect the entire economy, ADP has to massage the data a bit more than the government does. (In fact, it uses past reports from the Labor Department to build its models.) Still, the monthly estimates tend to fall within about 5 percent of the official numbers.

To get a jump on both the government report and the ADP estimate, more than 70 different forecasters put out their own predictions earlier in the month. These typically break the economy into sectors and then use data from each one to predict how its employment numbers might change. For the manufacturing sector, for example, a forecaster might consult new orders and production data from the Institute for Supply Management and the regional Federal Reserve banks to determine which way the industry is trending. Every forecaster has its own secret method for crunching the numbers.

Just because the government report is "official" doesn't mean it's accurate. Every month, the Bureau of Labor Statistics updates the previous three monthly estimates to reflect data acquired after the initial report. At the end of the year, the agency corrects the past 12 monthly reports using information from state unemployment insurance programs, which includes a much bigger sample of the 7.7 million employers in the United States and thus gives a more complete picture.

Got a question about today's news? Ask the Explainer.

Explainer thanks Joseph Brusuelas of Moody's

Brian Palmer covers science and medicine for Slate.