Are Behaviours in the “11-20” Game Well Explained by the level-k Model?


Speaker:Todd Kaplan & Lawrence Choo, University of Exeter
Date: Tuesday 8 July 2014
Time: 12.00 - 13.00
Location: Syndicate Room C, Building One

Further details

We investigate whether behaviours in Arad and Rubinstein (American Economic Review, 102(7), 2012, 3561-3573) “11-20” game are well explained by the level-k model. We replicate their game in our Baseline treatment and provided two other variations that retain the same mixed-strategy equilibrium but result in different predicted level-k behaviours. Our hypothesis test is motivated by the logic that if the Baseline and variation games capture level-k reasoning behaviours, we should find consistent proportion of level-k types in all games. We considered two types of level-k models where players were assumed to best respond stochastically and found that the level-k models were able to explain the data significantly better than the equilibrium driven alternatives. In addition, the level-k models were also able to demonstrate consistent proportions of level-k types between the dif- ferentiated games. Our findings provide support for Arad and Rubinstein (2012) assertion that behaviours in the “11-20” game can be attributed to the level-k models.