Level-n bounded rationality and dominated strategies in normal-form games

Dominated strategies play a crucial role in game theory and its solution concepts. While empirical studies confirm that humans generally avoid dominated strategies, they also suggest that humans seldom believe others will avoid such strategies. Hence, the iterated dominance solution is not likely to be a good predictor of one-shot behavior. We investigate how the salience of a dominated strategy affects the extent to which players believe that others will recognize and avoid it. Level-n theory serves as a useful tool in this empirical investigation, as it is able to classify behavior into levels of bounded rationality and provide clear statistical tests for model comparisons. We find that even the most obviously dominated strategies do not induce consistently significant behavioral differences in a variety of one-shot games. Nevertheless, the fit of the Level-n model can be improved by hypothesizing that the Level-0 choices and Level-1 beliefs are tilted slightly away from the uniform distribution to the extent that the average payoff of a strategy falls below a threshold.