Taking corporate culture seriously: group effects in the trust game.

• Author: Arce, Daniel G.
• Position: Trust Game

1. Introduction

   Kreps' (1990, 1996) Trust Game (illustrated in Figure 1) is part of the pantheon of games with proper names because it demonstrates the dichotomy between theoretical predictions of play and the way that play occurs in practice. Specifically, its unique subgame perfect Nash equilibrium (SPNE) results in gains from mutual trust going unrealized. This equilibrium is at odds with intuition that rational individuals should be able to overcome the distrust problem and the experimental evidence about how the game is actually played. Yet the logic behind the equilibrium prediction is stringent, and this is what makes it challenging. Indeed, Kreps asserts that the Trust Game (TG) is an important model of corporate culture, in that social learning to overcome this problem creates organizational efficiencies consistent with the existence of firms. Similarly, Casson (1991) opines that the formation of firms depends on the ability to overcome transaction costs, which mainly reflect the level of trust within a firm, and the level of trust depends on corporate culture.

   [FIGURE 1 OMITTED]

   In this paper, we examine the TG from the perspective of corporate culture--specifically, enduring (but finite) relations among individuals within the corporation and the periodic formation of groups therein. Groups are often used to achieve corporate goals because of the synergy created when each member is committed to the group's success (Kayser 1990). For this reason, by the 1990s, most major MBA programs emphasized teamwork in their curriculum, a practice that continues today. Moreover, Thompson (1998) asserts that whenever human beings get together into groups, the individuals in them are constantly trying to assess what sort of a group they find themselves in: Is this a group in which commitment is likely to produce the best result, or is it a group in which individual action is likely to produce the best result? We provide an answer to this question for the Trust Game.

   Our approach is based on an evolutionary game theoretic examination of the TG. Specifically, corporate culture is said to emerge as the result of pairwise matchings among cohorts in large corporate entities (Casson 1991). Indeed, Kreps (1990, 1996) suggests that because many situations within a corporation share the TG structure, it is possible that mutual trust evolves as a focal point because of its evolutionary fitness. Once the TG is set within an evolutionary context, it is found that it satisfies Wilson and Sober's (1994) criterion of a strong conflict between individual and group levels of selection. In other words, there are multiple levels of selection in which both individual strategies and groups of strategies (strategy pairs) operate as adaptive units. An important implication is that group effects can promote trust if the matching process is not independent of players' trustworthiness (assortative matching). This holds even though players who violate the trust of others do better in groups of trusting cohorts than do the trusting cohorts themselves. Group effects have been found to promote cooperation and reciprocal altruism in the Prisoner's Dilemma (Bergstrom 2003; Henrich 2004) and to select the Pareto-superior outcome in the Stag Hunt (Skyrms 2001; Bergstrom 2002). Our application to the TG is novel because the TG is an asymmetric extensive form game, whereas Prisoner's Dilemma and Stag Hunt are symmetric strategic form games.

   In the next section we briefly specify and discuss a generalized form of the TG. In section 3 we show that the TG satisfies Wilson and Sober's (1994) criterion for multilevel selection. This is followed by an evolutionary game theoretic expression of the TG in section 4. Section 5 defines the assortative matching technology and contains our main result, identifying the conditions for group effects to promote trust in the TG. We conclude with a discussion of the implications of our findings in terms of corporate culture.

2. The Trust Game

   A general form of the TG is given in Figure 2. The payoffs T (temptation), R (reward), P (punishment), and S (sucker) are often used to express and test general results for the Prisoner's Dilemma (PD). Under these payoffs, the normal form representation of Figure 2 (Box 1) makes it clear that the TG is not the PD. The TG is meant to represent the archetypal transaction with moral hazard, myriad examples of which are known to occur within and between firms, hence, the importance of the TG as a litmus test for effective corporate culture.

   [FIGURE 2 OMITTED]

   Box 1. Pairwise Matchings of Trust Game Types. Role/ Player B [??] [??] Role/Player [??] R, R S, T A [??] P, P P, P The TG also exemplifies the unraveling problem that occurs in a finitely repeated PD. Player B is in the position of honoring trust ([??]) or violating it ([??]). As T > R, rationality--in the form of subgame perfection/backward induction--predicts that player B selects [??]. As player A understands this, she selects [??] and the mutually beneficial payoff of (R, R) is forsaken for SPNE strategy pair ([??], [??]) with Pareto-inferior payoff (P, P).

   Henrich (2004) has shown that the question of whether reciprocity (in the form of a tit-for-tat strategy) can evolve in an infinitely repeated PD turns on a between-group component of natural selection that favors the evolution of reciprocity, and not individual selection, which always favors defectors. In other words, groups with a threshold level of reciprocators grow faster than other groups, even though defectors within such groups maintain their individual advantage. (1) For the finitely repeated PD, the resolution relies on informational comparative statics: either incomplete information about (possibly) irrational types or the existence of alternative commitment types that face different payoffs or have different beliefs about the structure of the game.

   Kreps (1990, 1996) notes that in an infinitely repeated TG the (R, R) payoff can result if players adopt tit-for-tat strategies, but finds this resolution unsatisfactory because, under the folk theorem, rival equilibria are abundant. Because we are taking corporate culture seriously, we add the following caveat, found in Aaron (1994, p. A14): "[W]e learned there are two ways to go: an eye for an eye (tit-for-tat), or do unto others as you would have them do unto you. In business, the latter principle is far more common, simply because it makes things work better." Finally, we assert that relying on corporate culture to resolve the TG precludes the use of informational comparative statics in the form of player (ir)rationality or...