Saturday, February 15, 2014

Bet with Kurt Mitman about unemployment



I have a new bet with Kurt Mitman, a PhD candidate at the University of Pennsylvania, about unemployment benefits. Kurt, like some other economists, believes that unemployment benefits are holding down employment in the U.S. by paying people not to work. He has written a paper with three coauthors that examines cross-state evidence and concludes that unemployment benefits are a significant disincentive to work.

It now looks like there is a very good chance that Congress will fail to extend unemployment benefits. Using the results from his paper, Kurt predicts that this will cause a lot more Americans to go get jobs. I have decided to bet against this happening. Because it's possible that the expiration of unemployment benefits will simply lead to a bunch of people no longer claiming to be looking for work (but still not looking for work), we made two separate bets on two separate numbers: A) the unemployment rate, and B) the employment-to-population ratio. Kurt bets that more people will get work, so unemployment will fall while employment rises. Being a general skeptic about the importance of policy, I bet that nothing much will happen, so that both numbers stay the same. Recent evidence from North Carolina suggests we might both be wrong - expiring benefits there seem to have led to a fall in unemployment and a corresponding rise in dropouts from the labor force.

Since Kurt is still an impoverished grad student, we kept the size of the bets small - one pizza per bet. As with my bet with Brad DeLong, Miles Kimball is officiating. Here are the official terms of the bet, as stated by Kurt:
We are interested in the December 2014 jobs report, to be released in January 2015.  The variables of interest are the unemployment rate and employment to population ratio.  Call these Ut and Et
The prediction of my model is that unemployment will be 5.2% on that date, and an increase in the employment to population ratio of 2 percentage points, call these Um and Em.
We are interested in the difference between the headline rate on the last month when extended unemployment benefits were in effect.  As of right now that is the December 2013 jobs report numbers, 6.7% for unemployment and 58.6% for employment population ratio.  If benefits get extended again, it would be the last month that they are in effect.  Call these U0 and E0.
I win the bets if U0-Ut > (U0-Um)*t/24
and
Et-E0 > (Em-E0)*t/24
where t is the months between the final month when extended benefits were in effect and December 2014.
For example, if benefits do not get re-extended, t=12, I win if:
6.7%-Ut > (6.7% - 5.2%)*12/24
or
Ut < 5.95%
Basically, the fall in U or increase in E is "pro-rated by the number of weeks.
If benefits were to be re-authorized through March 2014, I would win if
U_march - Ut > (U_march - 5.2%)*9/24
And we are betting one large pizza for each variable.
And this time, just for fun, I promise not to hedge my bet with a side bet.

18 comments:

  1. Economists are thinking about this issue all wrong because their frame of reference is completely off. When a low income worker decides not to take a jobtheir attitude is really:

    "I need $15/hr to survive, but this douchebag expects me to bust my ass for $8/hr. If I won't make enough to survive whether I take this job or not, why the hell would I decide to deal with annoying customers, a shitty job, and an abusive boss?"

    ReplyDelete
  2. Hey Noah, is there a chance you are betting that ObamaCare will offset the employment gains? :P

    What happens if you win U and your friend wins E? It seems to me from his wording that he has to win both to get free pizza.

    Its interesting that he is willing to bet on a decreasing E, because while the unemployed for more than 26 weeks have reduced their number since the peak, the civilian unemployment ratio has not seen any rise:

    http://research.stlouisfed.org/fred2/series/UEMP27OV

    http://research.stlouisfed.org/fred2/series/EMRATIO

    If UI increased unemployment, I would think that after 99 weeks have passed people would see incentives to start looking for work and E would rise. Which would mean a spike in employment around 2-3 years after the ARRA - and there no such thing last year.

    ReplyDelete
  3. Anonymous1:13 PM

    What is going to (or already has) expire is long-term unemployment benefits. In addition, the economy has been slowly improving, and may continue to do so even if dropping UI adds an additional drag & results in slower economic improvement than might have occurred otherwise. Which is to say that it seems to me you could be correct and still lose the bet on these terms.

    I wonder if it might be better / more specific to be on U6, and/or the unemployment rate for those who have been unemployed >27 weeks.

    ReplyDelete
  4. A fall in unemployment and employment ratio should be anyone's default.

    ReplyDelete
  5. Anonymous10:06 PM

    I'm looking at the picture . . . so the winner of the bet gets some kind of a casserole?

    ReplyDelete
    Replies
    1. Good point! If you guys think the stuff in the picture is a real pizza, neither of you has any judgment and you both deserve to lose. Come to New York and we'll show you what real pizza is, and we'll throw in some some discussion of the real-world impacts of policy on people's lives at no extra cost.

      Delete
  6. Try this

    1. Rise in crime
    2. Rise in abandoned children
    3. Rise in homelessness
    4. Rise in suicides

    Wanna bet?

    ReplyDelete
    Replies
    1. Don't forget:

      5. Abortion rates
      6. Substance abuse
      7. Domestic abuse (Technically a crime I guess, but this and addiction abuse will jump before other crimes do.)

      Delete
  7. We are dividing by 24 in this formula because... why? "The prediction of my model is that unemployment 5.2% on that date, and an increase in the employment to population ratio of 2 percentage points."

    There are plenty of ways to win this bet even if the model is right but the coefficients are not precisely estimated.

    ReplyDelete
    Replies
    1. Noah and I discussed, the idea is that my model makes a prediction - which is the mean from 1000 simulations - and so I would win if the data got halfway to my prediction. Scaled by 24 to take into account that it takes time for these stock variables to adjust. So that formula takes that into account.

      Delete
  8. Noah, one thing that you're both assuming is that the BLS's seasonal adjustments for Christmas are either correct or wrong in your favor. ISTR that that has been problematic recently, because the BLS is trying to figure it out in a depressed economy.

    ReplyDelete
  9. Am I not seeing something, or does your bet not account for the possibility that Congress does wind up extending UI at some point in the interim?

    ReplyDelete
    Replies
    1. If that happens, the bet is off.

      Delete
  10. Anonymous6:00 PM

    People are receiving UI because they lost their jobs. They are not being paid not to work. Or has that impoverished grad student documented cases in which employers have been unable to find workers?

    ReplyDelete
  11. Anonymous10:18 PM

    Unemployment Insurance. Wage earners pay for insurance.

    By calling it some thing else it sounds as if there were no obligation on the other side and the other side is giving a gift. Notice that unemployment insurance is a required payment by workers.

    ReplyDelete
  12. Anonymous10:22 PM

    The word insurance is not in the article.

    Very good point Anonymous 6:00 PM.

    ReplyDelete
  13. Anonymous3:39 PM

    Anyone want over/under 300k deaths in the next tsunami?

    ReplyDelete