Evidence on the equivalence of the strategic and extensive form representation of games

Our experiment tests whether strategically equivalent representations of games produce equivalent behavior when actually played. We examine representative members of the class of generic 2 x 2 extensive form games of perfect information and the equivalent strategic form games. Systematic differences exist between subjects' choices in the strategic and extensive form representations. These differences cannot be attributed to differences in subjects' ability to do backwards induction, in the salience of interpersonal preferences, or in optimization premiums between the two game forms. Instead, subjects in the extensive form are consistently more likely to allow the other player to make a meaningful choice. (C) 2003 Elsevier Science (USA). All rights reserved.