A real drop in unemployment
Feb 4th 2011, 20:17 by G.I. | WASHINGTON
DESPITE the good efforts of my colleague, there’s still a lot of misinterpretation of the drop in the unemployment rate from 9.4% in December to 9.0% in January. Some on Wall Street say it’s a bad sign, attributing it to a decline in the labour force as people gave up looking for work. But that decline in the labour force is a statistical illusion. When you remove that illusion, the entire drop in the unemployment rate can be attributed to the unemployed finding jobs.
I’ll dig a bit more deeply into the disparity. The low payroll employment number seems to be due to bad weather. Normally you can look to the separate tally of employment from the survey of households for a different perspective but that figure was distorted by the estimation procedure that always happens between December and January.
Each month the Census Bureau on behalf of the Bureau of Labour Statistics surveys 60,000 households, then extrapolates the results to the entire American population. The resulting estimates of the labour force and employment thus depend heavily on what Census thinks the entire population is. Each January, it revises that estimate and it now thinks the population is 347,000 smaller than it previously estimated, primarily due to the presence of fewer Hispanics than expected. Perhaps the tougher economy has cut down on immigration while increased enforcement has reduced the inflow of illegal immigrants.
The BLS helpfully provides “smoothed” data that adjusts for the change in the estimation procedure. By this tally, employment jumped 589,000 between December and January, not by the meagre 117,000 using the unadjusted numbers. The labour force grew by 2,000 instead of shrinking by 504,000. So the unemployment rate fell because more of the unemployed became employed.
Here’s a summary of the data:
(check link)
There still remains, however, the puzzling disparity between the household employment number and the payroll number. Household employment is defined slightly differently from payroll employment; it includes farm workers and the self-employed, for example. On an apples-to-apples basis, household employment, defined to be similar to payroll, rose a robust 540,000, compared to the mere 32,000 recorded in the payroll survey.
I have two possible explanations: some people who answered the household survey thought they were still working even though they stayed home because of weather. The second is that this is the usual statistical noise you get from trying to measure a huge number like the number of employed Americans through two different methods. The payroll number is usually the more reliable, so go with that. But the drop in the unemployment rate looks to be for real—although why the labour force has grown so slowly given that a recovering economy would normally suck more people in to work remains a mystery.