Almost immediately after the Bureau of Labor Statistics (BLS) released Friday’s employment report, conservatives claimed a conspiracy to cook the books to help get President Obama reelected. How does the unemployment rate drop 0.3 percent, they asked, when GDP growth is miniscule and there’s barely enough job growth to absorb a growing population?
There’s no conspiracy here. Sometimes you get unbelievable numbers when you combine an imprecise survey with tricky data adjustments. It’s wrong for the Republicans to claim that the hard working analysts at the BLS are behaving dishonestly. But it’s also wrong for Democrats to claim that the 0.3 percent drop in the unemployment rate indicates a healthy job market.
Let’s start with the problematic numbers. The BLS’s survey of households showed an 873,000 person increase in employment in September, the largest increase since 1983 not resulting from a statistical adjustment. By contrast, the establishment survey showed just 114,000 jobs being created, resulting in a 759,000 job gap between the two surveys, the biggest since 2003.
The BLS household survey also showed that the number of people officially unemployed declined by 456,000 last month. Because 114,000 jobs is only slightly more than is needed to keep up with population growth, this number seems wrong.
While these numbers are not believable, measurement error is the most plausible explanation. The BLS’s survey of households has a huge margin of error. The statistical agency is 90 percent sure that it’s household employment measure is within ±436,000 jobs of the actual number. That means, of course, that the actual number from September’s household survey could be as low as 437,000 or as high as 1.3 million.
The two series also define employment differently. The household survey includes people who work in agriculture, are self-employed, are on unpaid leave, and household and family workers not receiving a paycheck; but it does not count the multiple jobs that some people hold. To make the household survey more comparable to the establishment survey, the BLS also reports an adjusted household measure of employment, which showed that only 294,000 jobs were created in September.
The BLS adjusts its data for seasonality and sometimes its seasonal adjustment factor needs to be changed. The September jobs number might be evidence of that. The big increase in employment lies in the 582,000 people who started to work part time for economic reasons last month. But in 2011 the BLS estimated that the number of people working part time increased by 483,000 in September and declined 480,000 in October. Similarly in 2010, part time employment rose 579,000 in September and fell 419,000 in October. These offsetting movements suggest that something might not be right in the BLS’s adjustments for seasonality.
The BLS might need to fix its population estimates. As Harvard economist Greg Mankiw writes on his blog (http://gregmankiw.blogspot.com/), “If the BLS uses incorrect estimates of the size of the population, these errors will be reflected in its estimates of household employment.”
Finally, the household survey is prone to mistakes that occur when surveyors call people on the phone and ask for information. If those being queried refuse to answer or give inaccurate information then the survey results may be biased.
All of these measurement issues suggest that the drop in the unemployment rate to 7.8 percent isn’t an indicator of a robust job market. If it were, then the BLS’s measure of the percentage of the labor force that is unemployed plus those who are marginally attached and those who are working part time for economic reasons (called U6) should also have shrunk. But it remained at 14.7 percent in September.
While it is more interesting to claim political conspiracy than measurement error, the truth is that inaccurate economic data is a better explanation than ill intent for what happened to the unemployment rate in September.