Making sense of the monthly jobs numbers

Everyone's puzzling this morning over the government's January jobs numbers. A household survey asking a representative sample of Americans whether they're employed showed a healthy decline in the jobless rate, to 9.0 percent from 9.4 percent. But a separate survey, which asked employers about additions to their payrolls, showed that the United States added just 36,000 jobs last month—far fewer than expected, and not nearly enough to account for the 0.4 percent drop in the unemployment rate.

So what's going on?

Some have argued that the difference is due to changes in survey methodology. This month, the government changed its estimate of the baseline size of the U.S. population—as it does each January—to a figure nearly 100,000 lower than previous estimates. These analysts say the decrease in population brought down the unemployment rate in the household survey. In addition, December's payroll numbers were revised upward by 18,000—effectively lowering the number of jobs added in the new payroll survey.

But Heidi Shierholz of the Economic Policy Institute and Gary Burtless of the Brookings Institution—both seasoned observers of the monthly numbers—separately tell The Lookout that neither of these changes explains the discrepancy.

Even when you account for the change in the baseline population, said Shierholz, the household numbers "unambiguously show big improvement from December to January." Burtless agreed. The Labor Department "devoted two pages to taking out the part that's due to a benchmark revision,"—i.e., the downward adjustment in the overall population count—"and separating that from what they think is the change," he said. "The effect of the benchmark revisions is pretty trivial."

As for the revision to December's payroll numbers, that number, too, "is trivial compared with size of job gains being registered in the household survey," Burtless said.

So what does explain the conflicting survey results? Part of the issue, both Shierholz and Burtless say, is that the two surveys framed their questions about employment differently—and during periods of bad weather like this past January, that can create a discrepancy. The household survey simply asks Americans whether they'e employed, while the payroll survey asks companies about the size of their weekly payrolls. If bad weather causes outdoor work like construction to taper off, workers will still likely say they're employed. But their employers' weekly payrolls won't reflect that. During bouts of bad weather, in short, the household survey will likely show a higher rate of employment than the payroll survey does.

But Shierholz and Burtless both stressed a larger point. If you compare the two surveys over a longer time period—since the job market began to recover a year ago—they don't diverge nearly as much. Both show very slow job growth, which never registers a big enough month-to-month spike to significantly bring down the jobless rate.

Dean Baker of the Center for Economic and Policy Research points out that "such discrepancies are not uncommon, especially in the month of January." He notes that there were similar divergences in January of 1992, 1994, and 1997. "The numbers in the surveys tend to converge over time," he writes.

What this all suggests, as Felix Salmon of Reuters argues, is that despite the media's fixation on shifts in the monthly jobs numbers, those figures aren't necessarily the most useful guide to what's happening in the labor market. To better understand the jobs picture, we need to take a longer view.

(Job seekers at a Phoenix career center, Friday, January 2011: Ross D. Franklin/AP)