The Employment Numbers and the Unreported 4 Million Jobs

I described the August employment news as "good, not great" on Friday.

Larry Kudlow was more upbeat, and has a point:

The August jobs report should put to rest any fears that the economy is burning out. Following upwardly revised increases for June (134,000) and July (121,000), companies added 128,000 nonfarm payrolls last month. Meanwhile, the all-important but rarely mentioned household survey of people working gained by 250,000, sending the unemployment rate back to 4.7 percent from the July reading of 4.8 percent.

The cult of the bear, fussing about a housing-related recession, has once more been proven wrong.

The August employment showed once again that, with rare exception, there has been a fundamental disconnect between the number of new jobs reported in the Establishment Survey (phone calls to employers) compared to the Household Survey (calls to households). It's a difference that has been building since the economy began righting itself after enduring and adjusting to the trauma of the September 11, 2001 attacks.

Specifically, as shown below, the the Establishment Survey shows 4,892,000 net new jobs created since January of 2002, while the Household Survey shows 8,881,000 net new jobs:


According to the more comprehensive Household Survey (the survey that happens to be the basis for the reported unemployment rate), almost 4,000,000 more people have been added to the employment rolls than the Establish Survey shows (The Household Survey is more comprehensive because the Establishment Survey excludes farm employment). Another observation is that the Establishment Survey is only picking up 55% of the job growth being detected in the Household Survey.

Yet the employment changes that get reported each month by the business press come from the less comprephensive and much more volatile Establishment Survey (more volatile because in the 1990s, the Establishment Survey consistently showed more net new jobs than the Household Survey).

Why is this? Why wouldn't the primary basis for jobs reporting from the government and the press be the same as the one used to report the unemployment rate? At a minimum, why aren't BOTH numbers reported instead of ignoring the Household Survey's new jobs number? Does the press think we can't handle the idea that two different measurement methods would yield two different results?

Cross-posted, with minor changes, at

Economy Unemployment