References

use our credit card to charge the meal we have just eaten and will not steal the numbers for unauthorized uses. Without full trust in these situations, life would become more costly and less convenient. But why do we feel we can trust some mechanics?

Certainly there may be some motive for a mechanic to generate a reputation as being trustworthy. However, we must also acknowledge that a mechanic acting strictly as a selfish and rational actor would seek to take advantage of our lack of knowledge to some extent. Moreover, we must consider how easy it would be for the server to steal credit card numbers on a regular basis and take small amounts from customers with little chance of detection. Beyond this, the restaurant must trust that we not only have the means to pay when they first serve us the food but that we also intend to pay. Without a basic level of trust, economic transactions become costly, and some transactions will not take place.

In this chapter, we examine human behavior related to trust from a behavioral economics perspective. Trust is inherent in economic transactions. Yet in many cases, trust is a public good, and one for which private motives to be trustworthy should be minimal. Yet because both trust and trustworthiness exist and are widespread, we are all better off. This trust allows us substantial savings and increases our ability to conduct economic transactions as well as to conduct rewarding social interactions. Much of the literature on trust is closely related to the literature on fairness and altruism. We will emphasize this relationship.

EXAMPLE 16.1 Trust Relationships

Given the prominence of trust in our observed relationships, it is fair to question whether trust is a natural trait in human behavior, and if so how it came to be so. Observing trust, however, is not as easily accomplished as observing some of the other social preferences we have discussed. Fairness and altruism can be observed in a single choice or a set of choices that all occur at one point in time. However, to observe trust requires observing different players making a set of decisions that occur at different points in time. For example, we need to observe first the decision of the restaurant patron to be willing to give the server the credit card, then the server having the opportunity to either be honest or to steal the card number. For our purposes, we define trust as referring to a person’s willingness to place others in a position to make decisions that could either help or harm the person.

The canonical experiment used to measure trust was originally proposed and conducted by Joyce Berg, John Dickhaut, and Kevin McCabe. They designed an experiment in which participants were randomly assigned to one of two rooms. All the subjects were given $10. Those in Room A were told they could take any portion of their $10 and send it to a random and anonymous recipient in Room B; we refer to a person in Room A as a sender. Whatever money is sent to Room B is tripled. The people they were paired with, the receivers, could then decide to send back whatever portion of that money they wished. We refer to this as a trust game.

Let us first examine the Nash equilibrium of this game assuming selfish actors. Suppose that the sender sends $x to the receiver. The receiver must then find the split of $3x that maximizes their own utility. A selfish receiver will choose to take $3x, leaving nothing for the sender. A selfish sender will choose to send $0 no matter how much is

although senders were inclined to trust their anonymous counterparts, those counterparts were not particularly trustworthy on average.

This suggests first that people who are trusting might have incorrect beliefs about how trustworthy people are in general. This result causes some problems for the inequity aversion and fairness models. If receivers sent back less than was originally sent, this means they decided to take more than two thirds of what they received, in addition to the $10 they were given for participating, leaving the sender with less than $10. Moreover, the fact that the sender had sent a positive amount in the first place enabled the receiver to take at least some of the surplus. Thus, it seems that the sender should interpret the amount sent as a signal that the sender is being kind.

For example, suppose that the receiver employed the strategy in which she would return y = αx, where x is the amount sent by the sender. All Pareto optima in this game involve the sender sending the full $10, and in fact every allocation in this case results in a Pareto optimum. Thus, the only Pareto optimum with the receiver employing a strategy of returning αx, yields πP2 b2= πP2 b2 = 40 − 30α. The worst possible outcome for the receiver is that in which the sender sends nothing, yielding π2b2= 10. The kindness function proposed in equation 15.14 would have a value

Senders sent an average of x $5.16, with receivers returning α 0.90 of this. But if this is the case, equation 16.1 is positive. Thus, both the sender and the receiver should believe that the sender is being kind. Because this is positive, it should induce kind behavior by the receiver under the fairness-equilibrium hypothesis. But this does not occur on average. Instead, the split in the surplus return highly favors the receiver. Equation 16.1 also implies that the more the sender sends, the more the receiver should be willing to send back. In fact, there is no relationship between the amount sent and the amount returned.

Interestingly, the subgame in which the receiver decides how much money to send back to the sender is identical to the dictator game in which the receiver is the dictator. In this case, the results are fairly different from those in the dictator game. In general, the results of the dictator game did favor the dictator, but a large minority of the responses clustered very close to an even split between the dictator and the dictator’s counterpart. In this game, that would have meant a large number of receivers sending back about 1.5 times as much as was originally sent by the sender. This occurred in only a small number of cases. Thus, in response to the apparently kind behavior by the sender, the receiver appears to be less kind on average. This is a surprising result.

In an alternative treatment at a later point with different participants, senders and receivers were given a summary of the behavior from this first trust experiment. You would think that senders, given the chance to see how the game had been played by others, would send less in order to minimize their losses. This was not the case, with senders now sending an average of $5.36, more than in the previous experiments. In fact, although a larger fraction sent $0 (3/28), a larger fraction also sent the entire $10 (7/28). Seven of the 28 sent $5.

One potential explanation for this unintuitive behavior is that the information on the behavior in prior experiments created a social norm. Thus, for example, it may be that we don’t think twice about giving a waiter or waitress our credit card because we see