The April unemployment numbers, released this morning, were pretty good. According to the U.S. Bureau of Labor Statistics, which produces the monthly estimates, the economy added 165,000 jobs last month and the unemployment rate ticked down 0.1 percentage point, to 7.5 percent.
The best news, however, may have been what the numbers told us about what happened before April. The government revised upward its estimate for the number of jobs created in March and February. The original March numbers had been disappointing, a measly 88,000 jobs created, but the new number rose sharply to 138,000. The revised February estimate reached 332,000, up from 236,000.
Why does the BLS do these revisions, and what do they mean? The monthly jobs numbers are actually based on two separate surveys, the household survey and the payroll survey. The terms are pretty self-explanatory: In the household survey, people are simply asked to describe their employment status, while the payroll survey (also called the establishment survey) aggregates payroll data from companies and public-sector employers. The household survey produces the percentage, which is the number most people remember. Economists think it’s the less reliable of the two measures, though, because its sample size, about 60,000, is much smaller than the payroll survey, which covers 557,000 workplaces. The payroll survey provides the monthly number of jobs added (or lost).
Collecting data from 557,000 job sites is a lot of work, and that’s what gives the payroll survey more reliability. The BLS starts collecting data every month as soon as the first pay period is complete, but because lots of workplaces send out paychecks every two weeks rather than every week, this doesn’t happen until nearly halfway through the month. That doesn’t leave much time before the jobs report comes out early the next month, and many businesses can’t get their payroll information submitted in time. So the BLS revises the figures once, and then a second time over the next two months as more and more employers get their data in.
Even the revised numbers are still estimates, of course, based on survey data. To correct for any potential errors that remain after revision, once a year the BLS benchmarks its estimates, replacing the survey data with numbers from a comprehensive count of the unemployment insurance rolls. Those are the numbers that economic historians of the future will rely on. For those of us trying to figure out what sort of economy we’re living in, it’s important to keep in mind that every monthly jobs number comes with an asterisk.