Scott Sumner  

Prediction markets and loss aversion

PRINT
The new Hypermind NGDP futures... The Behavioral Econ of Paperwo...

Bob Murphy left a comment after my MoneyIllusion post on the NGDP prediction market (and David Henderson had a similar question):

In a traditional financial market, people have skin in the game and that helps to yield the "wisdom of crowds" results that work so well in markets, but work poorly in (say) presidential elections.

Are you just saying this is better than what we have now? I.e. it would be better still if true experts could put hundreds of thousands of dollars of their own money if they perceived a large mispricing of the NGDP futures contract, but letting people use this forum is still better than nothing?

I'm sure you get my point but for others: Suppose that instead of having Google stock trade on the open market, instead we let everybody who wanted to play guess on Google's 2017 profitability. And then at the end of the year we gave $10,000 in prize money to the person whose guess was closest. Surely that procedure wouldn't be nearly as good as estimate of Google's profitability as our current system.


I partly agree, but would like to separate out a couple of distinct issues. Consider two betting markets, in each case where people predict whether outcome A or B will occur (such as a two person election contest.):

Market A: People bet $5000, and get back $10,000 if they guess right.
Market B: People bet nothing, and win $10,000 if they guess right.

In market B, people have no "skin in the game", but in my view they would have just as much incentive to bet wisely as in market A. After all, in market A it's about whether you will win $5000 or lose $5000. In market B it's about winning $10,000 or winning zero. In both cases, you are $10,000 better off by making the right prediction.

Research in behavioral economics suggests that many people have "loss aversion", which cannot be explained by purely "rational" models of behavior. I put 'rational' in quotes because economists define the term differently from psychologists (or the average person for that matter.) If there is loss aversion, then market A might motivate more effort into searching out the truth.

I think Bob's more important point is that in a normal market a trader can invest a lot of money and earn large profits if their predictions are more accurate than the consensus. He's right that the sums involved here (while they will end up being much more than $10,000--I'll announce a big gift very soon) are too small to interest big Wall Street traders.

Of course it's much harder to set up a true futures market. For instance, I'd have to go through the difficult process of getting SEC approval, with no guarantee of success.

And don't write off the usefulness of simple prediction markets. Research by Robin Hanson, Justin Wolfers and others shows that these markets can be surprisingly efficient.

Emile Servan-Schreiber left a helpful comment in response to Bob over at my MoneyIllusion post. Here it is:

There is actually very little data to back-up the idea that real-money markets make better forecasts than play-money market, but there is a lot of data to the contrary. One significant advantage of play-money markets is that they correlate much better wealth and past prediction accuracy. In forecasting the highest-impact recent political events in the U.K., U.S., and France, Hypermind systematically outperformed real-money markets such as IEM, PredictIt and Betfair. Even among real-money markets, those that are most regulated and constrained - IEM and PredictIt - tended to make better predictions than the largest, deepest, least constrained Betfair... Hypermind has been carefully designed to attract and retain only the most dedicated "superforecasters", and it has an enviable track record of accuracy.
PS. Speaking of predicting the future, how can I resist this great picture:

Screen Shot 2017-05-23 at 1.55.33 PM.png


Comments and Sharing


CATEGORIES: Finance




COMMENTS (12 to date)
Andrew_FL writes:

Servan-Schreiber's comment is distinctly anti-libertarian.

Why not set up play-money markets to centrally plan the entire economy, if they're so efficient? It seems you've "inadvertently" vindicated Market Socialism.

My comment is not ideological in any way. It is just an empirical observation.

jj writes:

Requiring market participants to have skin in the game increases your signal-to-noise ratio.

My trading contribution to the NGDP market would be mostly noise, but if it's a free lottery ticket I might as well grab one, right?

(On the other hand I don't want to spend ANY time filling out forms so I probably won't bother).

Dikran Karagueuzian writes:

I take Bob's comment to mean that large markets that already exist contain implicit predictions of NGDP. For example, if we suddenly came to expect that next year would see a 10% drop in NGDP, we would immediately see a big drop in the S&P 500.

The question is then whether these implied predictions are better than the explicit predictions of a (much smaller) NGDP futures market.

The problem with the implicit predictions is that there is no single clear method for translating them into explicit predictions. Setting up the NGDP futures market would help a lot with that translation.

Scott Sumner writes:

Andrew, I don't see your point. Lots of firms use prediction markets to try to get optimal forecasts. Why is this anti-libertarian?

JJ. Yes, I suppose that makes sense. Hypermind tries to screen out uninformed traders.

Dikran, I agree that NGDP and stocks are correlated, but it's a very loose correlation. There were huge stock gains under Obama, for instance, but very slow NGDP growth. On the other hand, stocks crashed in 1987 and NGDP was quite strong in 1988.

Andrew_FL writes:
Lots of firms use prediction markets to try to get optimal forecasts. Why is this anti-libertarian?

How much do you know about the Socialist Calculation Debate? Because sometimes it's like you don't even realize you're on the Market Socialist side of the argument.

Charlie writes:

I find the quote especially ironic, because the first invocation of "The Wisdom of Crowds" is most often cited to Galton's analysis of a contest to guess a hogs weight that was written up in Nature as "Vox Populi." [My latin is not so good, but google tells me the literal translation is "voice of the people"]. http://www.all-about-psychology.com/support-files/the-wisdom-of-crowds.pdf

About 800 people were issued tickets and the closest to guessing the hog's weight would win a prize. Galton noted the median guess of 1207 was quite close to the actual hog's weight of 1198.

That seems a lot closer to this competition than the way a stock market works, though Galton did note, "The sixpenny fee deterred practical joking, and the hope of a prize and the joy of competition prompted each competitor to do his best."

brokilodeluxe writes:

Scott,

I'd think you'd need a minimum $1 entry fee just to keep bots and trolls out, right?

ant1900 writes:

[Comment removed for supplying false email address. Email the webmaster@econlib.org to request restoring your comment privileges. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

HL writes:

Would you see a meaningful difference between this type of market (NGDP futures) and Shiller / Farmer type NGDP-linked bond markets?

I believe it is somewhat easier to build the economic (interest bearing asset) and political (tail risk in public debt ratio trajectory as in Blanchard, Mauro and Acalin 2016) case than it would be for NGDP futures market. After all, you have a vibrant inflation-linked bonds market, but not Shiller housing index futures market (which Shiller tried to build but hasn't gotten much traction)...Just curious about your view.

And markets already have experience with these...like GDP-linked bonds in Latin America (Brady Bond cases like Bulgaria, Bosnia / Herzegovina, and Costa Rica), Argentina (30 year one!), and even Singapore (well, not quite, but GDP-linked in the payment structure, like New Singapore Shares in 2001).

BC writes:

One difference between Market A and Market B is that, in A, there is actually a third option to not bet anything and make zero with certainty. In Market B, there is no option to make $5000 with certainty. So, a purely rational person that knows he doesn't have an information advantage over others might decline to participate in Market A while making a random guess in Market B. That self-selection may make a difference.

foxhuntingman writes:

Why would you say that participants in Market B have no skin in the game? They do!

Comments for this entry have been closed
Return to top