Tuesday, February 09, 2016

Yes, the cyclicalists really won the bet


So, Bryan Caplan won a bet with Tyler Cowen regarding the unemployment rate. Cowen bet that unemployment wouldn't go below 5% at any time between 2013 and 2033. It recently did, so Caplan won the bet 17.5 years before the deadline. Paul Krugman says that this is vindication for cyclical theories of unemployment, as applied to the Great Recession, over structural theories.

Tyler says: Not so fast. Unemployment may be down, but the employment-to-population ratio has barely budged:


He then cites some labor search theory papers, such as this one by Restrepo and this one by Elsby and Shapiro, that describe various situations in which we might see substantial structural unemployment. All in all, Cowen declares that "I feel I'm the one who won the bet." 

So who is right? Caplan and Krugman are mostly right, and possibly completely right. Cowen is at least mostly wrong here.

The reason? Aging. The Great Recession came along just as Baby Boomers were starting to retire, which is why the employment-to-population ratio hasn't recovered. Instead, let's look at the best measure of employment: the employment-to-working-age-population ratio:


As we can see, there has in fact been a strong and steady recovery in employment when we just look at working-age people. About 75% of the employment drop of the Great Recession has been reversed. So if the bet has a "true" winner, or a winner in spirit, it's definitely Caplan.

What about that last 25%? That represents a little over 2 million people. Does that represent increased structural unemployment, possibly due to effects of the Restrepo type, the Elsby-Shapiro type, or others (such as de-skilling)? It might. But it will be a while before we know. So far, correctly measured employment has been on a strong, smooth, steady upward trend. Only if that trend suddenly breaks, and employment flatlines at a level below the previous one, will it be time to reach for structural explanations for the Great Recession.

Interestingly, if we're using employment recovery as our measure of structural-ness, the 2000 recession looks a lot more structural than the Great Recession. The drop in employment-to-working-age-population after 2000 was almost as steep as that of the Great Recession, and there was far less bounceback. The 2000s really were a "lost decade" for U.S. employment. Of course I suspect China trade as the culprit, based on the timing and on the recent Autor-Dorn-Hanson paper. But that is a story for another day.

Regarding the Great Recession, the conclusion is simple: Caplan has earned his $10.

Which, unfortunately, is not even enough to buy 1/2 of a small Zachary's pizza. 


Updates

On Twitter, Ernie Tedeschi shows an alternative measure, the demographically adjusted employment-to-population ratio:


This measure is designed to take into account the changing age composition of the population. My preferred measure, the employment-to-working-age-population ratio, makes no distinction between a 15-year-old and a 64-year-old. This measure does. Aging doesn't just reduce the percentage of total population in the 15-64 bucket, it also shifts the distribution within that bucket.

The "adjustment" in this measure is meant to take that into account. But it will also involve making some assumptions. It probably assumes that the relative likelihood of a a 64-year-old to be employed and a 15-year-old to be employed stay the same over time. In fact, that is not true. In the recovery from the Great Recession, employment rates rose much more strongly among the old than among the young. That means that the "demographically adjusted" employment-to-population ratio will tend to understate the degree of labor market recovery following the Great Recession. For that reason, Ernie's graph shows a slightly less robust recovery than my own, though the basic stories are the same.

But this graph does make me think I was much too quick to label the post-2000 employment stagnation "structural". Maybe a lot of that was just due to aging, as the Boomers approached (but didn't reach) the official retirement age.

18 comments:

  1. Anonymous6:29 AM

    The funny thing is that a few years ago Krugman would point to the employment to population ratio as the real measure of slack and berate everyone who said it was mostly structural.

    ReplyDelete
    Replies
    1. To be fair, those happen to be the 3 years when boomers first started retiring in large numbers.

      Delete
    2. Prof. Krugman has generally referred to the employment to working age ratio. He's been consistent warning about wrapping Baby Boomer retirement into the stat.

      Delete
    3. Phil Koop11:16 AM

      Actually Krugman's preferred measure for berating people is prime age (25-54) employment to population ratio.

      It was Caplan and Cowen who chose instead to bet on unemployment (and who so foolishly chose $10 instead of pizza as the stake.) How is that Krugman's fault?

      Delete
  2. Another take on the cyclical-structural debate:
    https://thefaintofheart.wordpress.com/2016/02/05/the-fed-the-unemployment-rate/

    ReplyDelete
  3. The trouble with Econ debates is that they are so often between one idea that is 3/4 correct and another that is 1/4 correct. Sure, it seems only logical to prefer the 3/4 correct idea: "three times the correctness!" But it leaves people unsatisfied and the folks who favor the 1/4 correct idea just continue to feel that maybe their idea is at least the beginning of a better version that would be, maybe, 7/8 correct.

    I'm with Caplan and Krugman on this one, but confidently predict that "opinions will continue to differ."

    ReplyDelete
  4. @Noahpinion: Thanks for this! Needed.

    ReplyDelete
  5. What exactly did Tyler post, anyway? Employment to population ages 15-65 rather than 15-54?

    ReplyDelete
    Replies
    1. His graph includes 3-year-olds and 93-year-olds.

      Delete
    2. Anonymous6:04 PM

      Tyler's post is employment to population for 15+ year olds. It includes 93-year-old, but not 3-year-olds.

      https://research.stlouisfed.org/fred2/series/LREMTTTTUSQ156N

      Delete
  6. I thought it was clear this was cyclical given the failure to hit the inflation target and the long stay at the zero lower bound. Aren't these clear signs of demand problems? What am I missing?

    ReplyDelete
  7. Don't trust people who come up with stats to support disproved positions. If they can't move on, their ego trumps their science.

    ReplyDelete
  8. Doc at the Radar Station11:50 PM

    I like the "ZPOP" ratio
    http://macroblog.typepad.com/macroblog/2015/09/the-zpop-ratio-a-simple-take-on-a-complicated-labor-market.html

    ReplyDelete
  9. Anonymous9:55 PM

    Why do famous economists who make bets like this always bet such tiny amounts of money? I want to see some real stakes! $100,000 to the winner! If they are willing to argue that we should steer the national economy based on their opinions, then they should be willing to put some serious personal skin in the game too.

    ReplyDelete
    Replies
    1. Why bother to bet at all, when they're not going to admit to being wrong no matter how clear the criteria?

      Delete
  10. I thought "what about retiring bombers" but then assumed that Tyler must of considered that. Silly me.

    ReplyDelete
  11. One very important distinction here is the difference between voluntary and involuntary labor force non-participation. If involuntary non-participation hadn't recovered, that would be a strong indicator that people leaving the labor force were forced out by structural conditions. Given that that 25% is almost entirely voluntarily not searching for work, an interesting puzzle is posed. That is mostly people staying in school longer, retiring earlier, and exiting the workforce for childcare. Do the people who stay in school longer stay because they want to, or are they forced out of the workforce and into school by a skills mismatch? I fall on the side that structural mismatches are not present because, as Paul Krugman loves to point out, there are not the rising wages that should accompany that.

    ReplyDelete