Casaubon's Book

20% of the US is Un- or Under- Employed

In January, the same month that unemployment “fell” to 9.7% (by which we mean we only lost a few hundred thousand jobs, and we hadn’t yet done the inevitable upward revision), we learn that about 20% of Americans were underemployed or unemployed and finding it hard to make ends meet. This compared to the BLS numbers at U6 at 18%, which have been widely criticized, but are still being used. This means that the Labor Departments reliance on the birth-death model in recession. is yet again called into question.

BTW, several commenters correctly pointed out that I made an error in my original post – make sure you read their comments and corrections for clarification.



  1. #1 Mike Cagle
    February 23, 2010

    Wait, I’m confused. How can it be that “20% of Americans who had jobs were … unemployed”? Underemployed I get, but how can someone who has a job be unemployed?

  2. #2 Adrienne
    February 23, 2010

    The actual article says 20% are underemployed… 9.7% are unemployed (according to statistics) so the way I read it that makes a total of nearly 30% who are out of work to one degree or another.

  3. #3 bsci
    February 23, 2010

    If you read the linked article, it notes the BLS underemployment is 16.5% compared to Gallup’s 20% and it says BLS does seasonal adjustments. BLS collects and publishes multiple employment numbers. If you want to learn about these numbers instead of ranting, you can read:

    The not seasonally adjusted U-6 value is currently 18%.

    The BLS sample is 66,0000 households (around twice as many people) each month. The Gallup survey is 1000 different people per day.

    If you have a actual reason to believe the 18% compared to 20% underemployment is based on flaws in the birth-death model rather than the very different sampling procedures, I’d love you hear your evidence.

  4. #4 bsci
    February 23, 2010

    I know there’s a link limit in a simple comment, but here’s the details of the BLS surveying technique:
    It bring households in and out of the survey over time to get some longitudinal information. (Note I mistyped 66K instead of 60K above)

    Here’s all the info I could find on how the Gallup poll was created:

  5. #5 Sharon Astyk
    February 23, 2010

    Oops, Mike, sorry, posting fast and not paying enough attention to grammar. Will fix now.

    BSCI, even the BLS admits that there are major issues with the utility of the birth-death model during a recession:

    We can see the weakness in the models – the BLS missed 1.2 million job losses last year that were replaced in later revisions.


  6. #6 Sharon Astyk
    February 23, 2010

    Adrienne, you’d think, but no, they included unemployed as well: “Gallup estimated that about 30 million Americans are underemployed, meaning either jobless or able to find only part-time work.”

  7. #7 bsci
    February 23, 2010

    I think you are mixing up two surveys. The Bloomberg article seems to be referring to the Current Employment Statistics: (i.e. the payroll survey) This is a survey of business payrolls that has a birth-death model.

    The Current Population Survey is a survey directly contacting a sample of households across the country. When the government is releasing the U-3 unemployment rates or the U-6 underemployment rates, those are data from the CPS. The 9.7% you reference is the seasonally adjusted U-3 number from the CPS.

    Even if I’m wrong, in the above post you’re writing that the government claims the rate is 9.7% while Gallup is measuring 20%. It would be more truthful to include the governments not seasonally adjusted U-6 of 18% as a comparison to the Gallup poll.

  8. #8 Tim Bartik
    February 23, 2010


    You really should correct this post. bsci is completely right.Whatever deficiencies there are in employment data based on a sample of establishments (the BLS’s birth death model that you mention, for example) have nothing whatsoever to do with any errors in measuring employment/unemployment/underemployment from the household survey, on which BLS’s measures of unemployment and underemployment are based.

    Why do BLS’s numbers for underemployment differ somewhat from Gallup’s? It’s hard to say without really getting into the Gallup data, but here are some possibilities: (1) BLS has a larger sample size; (2) the data are not collected in exactly the same time frame; (3) it is not clear that Gallup uses exactly the same questions and procedures as BLS does to measure underemployment; (4) both Gallup and BLS use separate samples, and each sample will have some sampling error; (5) Gallup and BLS may have different approaches to dealing with non-response bias, and (6) finally, as noted, Gallup does not do seasonal adjustment, whereas BLS does.

    I suspect that BLS does a better job than Gallup of measuring the various concepts of unemployment that it tries to measure. It is a legitimate argument that the official measure of unemployment may understate the unemployment problem during serious recessions, and that some of BLS’s alternative measure of labor market slack may be better. Perhaps we should be emphasizing these alternative unemployment measures more in our discussions of the economy. But BLS does do a very professional job of measuring a wide variety of measures of labor market conditions.

  9. #9 Sharon Astyk
    February 24, 2010

    Tim, BSCI, you are right – I should have used the U6 and will change it.

New comments have been disabled.