paper on evolutionary stable play in the lab

[Christoph Kuzmics]

Dear all,

Daniel Rodenburger and I just had a paper accepted at Economic Theory entitled "A case of evolutionarily stable attainable equilibrium in the lab".

My short over the top sales pitch would be this: "We stumbled across an old political science experiment in which lab subjects recurrently play a somewhat complicated 14 player game. While there are many equilibria subjects actually play the more efficient of the two evolutionarily stable strategies of this complicated game." We felt such a feat needed to be recorded somewhere.

The complete truth is of course a bit more complex - see the paper

Here is the abstract: We reinvestigate data from the voting experiment of Forsythe, Myerson, Rietz, and Weber (1993). In every one of 24 rounds 28 players were randomly (re)allocated into two groups of 14 to play a voting stage game with or without a preceding opinion poll phase. We find that the null hypothesis that play in every round is given by a particular evolutionarily stable attainable equilibrium of the 14 player stage game cannot be rejected if we account for risk-aversion (or a heightened concern for
coordination), calibrated in another treatment.

Best wishes,

Christoph

[Jonathan Newton]

Thanks for the paper, Christoph.

I read it, except for the introduction and appendix, and enjoyed it. 

Typo page 7: "thee" -> "three".  ;)

One observation/musing:

One thing of note is the way the experiment is constructed with 4,4,6 players respectively who have A,B,C respectively as their preferred outcome. This creates a tie-breaking component to the problem that might not exist if, for example, the numbers were 5,3,6. Indeed, this seems to be a lesson from the opinion poll treatment.

The initial treatment is of a form that might not occur too often in real world situations in which behavior would have evolved. For example, it has the tie in numbers (4=4) and there is no deliberation or discussion (is this correct?). Consequently, it might be that specific evolutionary adaptation to such circumstances is not very strong. Ergo, it is not even clear that equilibrium for this context is an appropriate concept rather than some application of some behavioral maxim learned in other contexts. This could actually be quite close to your discussion at the end of Section 3 of players who see the game as a challenge to achieve coordination.

[Christoph Kuzmics]

hi jonathan, thanks for the comments!

i want to briefly publicly reply to your statement "it might be that specific evolutionary adaptation to such circumstances is not very strong":

of course i don't expect subjects having played such an artificial game ever before, and i do not think that they have literally evolved to play something specific in this game. i interpret evolutionary stabilty as follows. a strategy that subjects play constantly in recurrent and (reasonably) anonymous interaction and with sufficient feedback about their opponents' average choices will probably be an evolutionary stable strategy. if it is not people will abandon it at some point (probably very quickly).

imagine subjects playing 2x2 pure coordination games with actions A and B recurrently in randomly matched anonymous pairs with some feedback about what subjects on the whole play. i do not believe that they will play the 50-50 mixed equilibrium in the long run, because it is not evolutionarily stable. no real long run evolution of their genes has to happen for this to be true, i believe.

best

christoph