Carpe Diem

Take out one outlier month, and the ‘stunning growth’ in part-time employment falls from 59% to 19% of jobs added


A blog post by John Lott titled “Numbers continue to be Stunning. 96 percent of net jobs added this year have been part-time jobs,” has received a lot of attention and was featured by Greg Mankiw, Don Boudreaux, and Tyler Cowen and “critiqued” (I think) by Brad DeLong.

First of all, it turns out that the 96 percent figure was actually wildly inaccurate, and was later corrected by John Lott – only 59.4 percent of the jobs added from January to August were part-time jobs (497,000 of 837,000 total jobs, see chart above, BLS data here for full-time employment and here for part-time employment). Fair enough, it was a huge, but honest mistake that should probably be corrected by Greg and Don on their blogs – they are still both reporting the original and inaccurate 96% figure. Here are some other issues:

1. Much of the 59% part-time figure for jobs added during the 8-month period from January to August period was driven by some pretty wild numbers in the month of March. It’s widely known that the BLS jobs data from the Household Survey are extremely volatile in certain months, and March was a perfect example. According to the BLS, there was a loss of 240,000 full-time jobs in the month of March and an increase of 360,000 part-time jobs, and both of those monthly job numbers deviated significantly from the averages over the last 12 months for those two job categories of 143,000 new full-time jobs and 24,000 new part-time jobs per month.

If you omit the month of March as an outlier, and consider the other seven months of 2013, part-time jobs represented only 19.1 percent of the total jobs created so far this year (see chart above).

2. The chart above shows what percentage of total new jobs created were part-time jobs over various periods from the last eight months to the last 24 months. For example, if we add two more months, and consider the last 10-month period from November 2012 to August 2013, we find that only 14.96% of the total jobs created were part-time (130,000 out of 869,000). Likewise, over the last 12 months through August 2013, only 14.37 percent of the total jobs created were part-time positions (288,000 out of 2,004,000 total new jobs).  Adding just one more month illustrates the extreme volatility of the Household Survey jobs data. In September of last year, the BLS estimated that 767,000 full-time jobs were added in just one month, compared to a loss of 19,000 part-time jobs, and both of those figures deviated significantly from averages over longer periods of time, and brought the share of part-time jobs down to only 5.9 percent for the last 13 months.

3. Over longer periods of time that help smooth out the volatility of the monthly data (e.g. 12, 18, and 24 months), we see that a large majority of the jobs that have been created have been full-time (almost 86 percent over the last 12 and 24 months periods) and part-time employment has accounted for only about 14% of the new jobs added during those periods.

Bottom Line: Although it’s certainly possible that part-time jobs are increasing relative to full-time jobs, that trend can’t yet be confirmed. Therefore, the 96 percent 59 percent figure for part-time jobs created over the last 8 months really doesn’t deserve much attention, given the facts that: a) it’s being driven by one outlier month (March), b) monthly jobs data from the Household Survey are so extremely volatile, and c) the 59 percent figure for part-time jobs declines significantly over longer periods of time that include more data, e.g. adding just two months brings the part-time percentage below 15%.

Update: To further illustrate the extreme volatility of the BLS Household Survey data, consider that for the most recent month available – August 2013 – the number of full-time jobs increased by 118,000 while the number of part-time jobs fell by 234,000. So based on the most recent labor market data, part-time jobs are falling relative to full-time jobs; but the extreme volatility of these monthly data makes that finding relatively meaningless, which further illustrates the main point of this post.    

22 thoughts on “Take out one outlier month, and the ‘stunning growth’ in part-time employment falls from 59% to 19% of jobs added

  1. Wow, Perry is on the data. Congrats.

    Mankiw and Cowen got caught in the quicksand. What does that say about their economic “instincts”? As usual, Boudreaux is a johnny-come-lately.

    FRED Graph

    • Well now they’re NOT exsactly cheering over at Marginal Revolution since they apparently see it different…

      Mankiw and Cowen got caught in the quicksand“…

      Could that have been Ben Engebreth’s quicksand?

      Are all these guys finding it problematical with regards to finding a credible source of data?

      I agree that John Lott’s numbers are more than a bit pessimistic and maybe even questionable…

      • Could that have been Ben Engebreth’s quicksand?

        No, Juan Doze The Idiot. It was Donald Coffin on Lott The Turd’s blog.

        What is it with scumbags like Lott who unilaterally change the body of their blog posts to protect their sorry butts. Lott changed his post without indicating same. Updates are fine. Scamming his statistical ignorance isn’t.

  2. “According to the BLS, there was a loss of 240,000 full-time jobs in the month of March and an increase of 360,000 part-time jobs.”

    “If you omit the month of March as an outlier,…”

    Statisticians and full-time job seekers: beware the ides of March.

    • If you cherry pick the data you can come up with anything. If you are going to omit March because it is too high – you should also omit some low month as well.

      we know this economic data is perfectly valid of course since months later the data is always revised grossly downwards.

  3. i’m not sure i understand the justification for leaving out march.

    i mean, the economy from 2003-13 looks very different if we leave out the the second half of 2008 and the first part of 2009 too.

    they were certainly “outliers” as well, but leaving them out would seem to hide more than it reveals.

    unless there is some reason to suspect that the data in march was inaccurate, why would we exclude it?

    it also looks like we are seeing a general rise in the percentage of part time work in jobs creation.

    given the large number of firms that have expressed plans to shift more people to part time status to avoid obamacare rules, this does not seem surprising nor, i fear, does it imply good prospects for a move to full time in the near future.

    • Well, we can also add just two more months, and the 10-month percentage falls to less than 15% compared to the 8-month percentage of more than 59%, suggesting some extreme volatility in the monthly data.

      Here’s another example: In the most recent month available (August), part-time employment fell by 234,000, while full-time employment rose by 118,000.

      • it also looks like we are seeing a general rise in the percentage of part time work in jobs creation.

        Quit blowing smoke. The 01/2013 part-time employment for economic reasons to full time employment ratio was 5.41159%. Likewise 08/2013 it was 5.22095%. That’s a decline.

        • marmico-

          you do realize that captures only part of part time jobs and says nothing definative about the overall number created, right?

          for a guy so anxious to heap abuse on others, you sure seem to leave out some awfully pertinent stuff…

          • look at the numbers posted.

            the ratio for the last 12 months is higher than that least 24.

            that means that the recent 12 months were higher than the 12 months before them.

            the 11, 10, and 9 month figures are higher still, which demonstrates a further increase.

            this is 8th grade math marmico.

            perhaps you should take you own advice about smoke.

      • “Well, we can also add just two more months, and the 10-month percentage falls to less than 15% compared to the 8-month percentage of more than 59%, suggesting some extreme volatility in the monthly data.”

        volatility is not a reason to ignore things.

        and why use trailing averages?

        this makes it difficult to compare year to year changes.

        for example, what if we compare the first 8 months of this year with the first 8 of last year?

        if the trialing months being added in drops the % so much, that would seem to imply that the percentages are significantly higher this year, even including august.

        i also wonder if there is not an inflection point in this data around the delay of the obamacare employer mandate which seems to have had an effect on hiring patterns.

      • Including today’s new bls data how do things average out now?

        And thank you Mark for your detail, fairness and concern for accuracy here. Am very impressed and this has helped me learn more about how this volitile part time jobs space can be best understood… Tim

  4. The data is the data. I don’t understand the value of arguing how things look differently if March is left out. The 9 month average is as-is. There is no reason to toss out a month just because it is higher or lower than one likes.

    One can argue that there was little change between the 24 and the 12 month averages; therefore one needs not be too alarmed. That would be a fair point.

    • rk-

      on the other hand, we can look at the 8 months to august and see a 19.1% figure.

      we can then look at the 14.3% 12 month figure and the 14.1% 24 month figure, and compute that the 12 months before the most recent 12 had a 13.9% ration of part time jobs. thus the last 8 months even without march are considerably higher that the out 12 month number or even the trailing 12 month figure.

      this would seem to imply that there has been some sort of meaningful change in the last 8-10 months even if we exclude march, but including it makes this change look very dramatic.

      if the 12 months that ended 12 months ago were 13.9 and the last 8 have been 59%, that’s a massive increase.

      diluting it by adding more months to a moving average just hides the inflection point.

    • marque-

      i’m not sure, but we can look at the data to get a sense of recent figures.

      if the last 8 months have seen 59% of jobs be part time and the last 24 have seen 14.1, we can do some simple math and see that if:

      ((8 X 59) X (16 X n))/24 = 14.1 then the ratio for the 18 months leading up to 2013 (but not including it) was 11.5%. perhaps we can use that as a baseline, but i do ot know if that is elevated by the standards of the last 20 or 50 years.

      thus, the last 12 months show a significant uptick from that level, and the last 8-9 months show a violent uptick.

      this seems to imply that starting sometime around december, something meaningful changed.

      august seems to have been a reversal of this trend (perhaps for seasonal reasons, perhaps due to the delay in the employer mandate).

      this will make sept an interesting number to watch.

  5. Do you have some kind of agenda here?
    IF there is a problem, highlight it. You correctly highlighted an error… good.
    However, if I’d built a load of homes on an earthquake fault and seen large volatility on a seismograph I should just say don’t worry folks, its just an ‘outlier’.
    Violent swings are likely to be indicative of something worthy of further investigation.
    Statistical analysis is about seeing the wood for the trees, not just throwing the baby out with the bath water.

  6. Interesting. So lets look at the numbers. I am looking at the sites you listed.

    In November, 2008, part time jobs rose by 2500 per month, from the 25,000 to the 27,500 range. What caused that? An initial distrust of the new president, I would guess.

    Second, March 2013 is among the lowest part time job creation numbers since the prior March. Its not, as you say, 24000, but in the 27,400 range. Removing the trend line, the number for March part time jobs creation is within the standard deviation (on the negative side) for that two year period.

    Regretfully, gotta go to work, but so far the part time numbers aren’t looking especially weird. However, I did get the generalized trend by looking from 2009 to 2013. When you look at the linearized numbers for the last two years, there is a small upward trend in part time job numbers. I’ll look at the full time jobs later, but so far, I’m not seeing “volatility” in these numbers(not from a statistical point of view), and there is a small but consistent increase. Full time numbers may be different, which I guess will be the case considering that its not the March part time number that is out of line.

  7. So, now I’m confused. The full time jobs don’t look anomalous. You can see a reduction in full time job creation starting January (its very visible) and obviously connected to the tax hike. But that March anomaly doesn’t show up in the part time job data that you linked to.

    Was this a mistake in the Feds data set?

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>