Download data and study materials from OSF
Sample size: 613
Field period: 05/03/2004-06/03/2004
Trust is risky: it involves a chance outcome under the control of another party. But is trusting really like taking a risky bet? This paper provides a framework to answer this question. A large body of work on trust, crossing many disciplines, assumes that the willingness to trust is closely associated with the willingness to take risk. However, in contrast to the typical risky choice task, the trust decision does not only involve the decision maker but also a second person. Not only are the latter's payoffs also affected by the decision maker's choice but in addition, the second mover and not nature determines the final outcome. Thus, attitudes towards others (or other-regarding preferences) and attitudes towards betrayal (or betrayal aversion) may affect trust decisions, in addition to (or in place
of) attitudes to risk.
Earlier research by the authors, based on laboratory experiments with student subjects, suggests that various ethnic groups' willingness to trust may be differentially affected by these vulnerability components. Members of minority groups are mostly concerned about the risk of inequity while whites mainly care about the risk of betrayal involved in trusting. Theories of status and power help account for the differences we observe. In this study, we use a representative sample of Americans to further explore this question.
Using experiments, we compare choices in a binary-choice trust game with the choices in two risky-choice tasks that offer the decision maker (DM) the same distribution of monetary payoffs: a binary-choice risky dictator game and a decision problem. The terms good and bad are used for outcomes yielding monetary payoffs to the DM that are larger or smaller than a sure payoff. The DM is confronted with a second mover in the trust game (the
Trustee) and an inert second player in the risky dictator game (the Recipient). The only difference between the trust game and the risky dictator game is that if the DM decides to trust, the Trustee determines both players' payoffs in the trust game whereas nature determines the DM's and the Recipient's outcomes if the DM chooses to gamble in the latter. The only difference between the risky dictator game and the decision problem is that the decision problem involves no second player.
To compare these three decisions, the study elicits DMs' minimum acceptable probabilities (MAPs) in each situation of getting the good outcome that just leads them to prefer the lottery to the sure payoff. If a subject's MAP is higher than the probability of receiving the good outcome, whose value was determined prior to the experiment, she receives the sure payoff. If it is equal to or lower than that probability, she will play the gamble. Our null hypothesis is that the MAPs in the three decision scenarios are perfectly correlated. Lack of correlation between a DM's MAP in the decision problem and her MAP in the risky dictator game reveals her attitudes towards others. Lack of correlation between a DM's MAPs in the trust game and in the risky dictator game measures her attitudes towards betrayal.
We find that the MAPs in the decision problem and the risky dictator game are significantly more strongly related than the MAPs in the risky dictator game and the trust game for Asians and Caucasians. In contrast, for African Americans and Hispanics, the MAPs in the risky dictator game and the trust game are significantly more strongly related than the MAPs in the decision problem and the risky dictator game. This pattern suggests that African Americans and Hispanics' decisions are most strongly affected by whether or not a second person receives payoffs while Asians and Caucasians' decisions are most strongly affected by whether or not they are playing against nature or another person. Put differently, African Americans and Hispanics care more about the potential inequity involved in trust while Asians and Caucasians care more about the potential betrayal that could result from trusting.
Null hypothesis: Trust is about taking risk. Willingness to take risk (MAPs in DP and RDG) and willingness to trust (MAP in TG) are perfectly correlated for all subjects.
Using a within-subject design, we ask participants to complete three decision tasks. Participants are asked to indicate their willingness to take risk in a binary-choice decision problem, a binary-choice risky dictator game and a binary-choice trust game. As customary in experimental economics, we use neutral language and create appropriate incentives with payoffs. Subjects are informed that they will be paid according to their decision in one of the problems predetermined by the experimenters. In addition to this behavioral data, we collect information on demographic characteristics, using the standard TESS instruments.
Willingness to trust (MAP in trust game)
Willingness to take risk (MAP in decision problem)
Willingness to take risk when risk-taking implies making another person better off (MAP in risky dictator game)
Willingness to take risk (MAP) in the decision problem is significantly more strongly related to willingness to take risk (MAP) in the risky dictator game than the MAPs in the risky dictator game and the trust game for Asians and Caucasians. In contrast, for African Americans and Hispanics, the MAPs in the risky dictator game and the trust game are significantly more strongly related than the MAPs in the decision problem and the risky dictator game. This pattern suggests that African Americans and Hispanics' decisions are most strongly affected by whether or not a second person receives payoffs while Asians and Caucasians' decisions are most strongly affected by whether or not they are playing against nature or another person. Put differently, African Americans and Hispanics care more about the potential inequity involved in trust while Asians and Caucasians care more about the potential betrayal that could result from trusting.
This pattern is in line with predictions from status and power theories. By trusting, people expose themselves to the risk of ending up worse off than the trusted party. At the same time, by trusting, they increase efficiency and make the trusted party better off than he would have been if no trust had been offered. Status and power theories suggest that minority members are more concerned about equality than majority members. It is argued that minority members' preferences for equality developed because other allocation procedures do not serve them well. Absent equality, they generally did not get what they deserved to receive, keeping everything else equal.
By trusting, people expose themselves to the risk of being betrayed by the trusted party. They cede power over their own outcome to the trusted party and accept some degree of submission to another's will. Majority members have been found to be more used to assuming powerful roles than minority members. Stereotypes can lead groups to be identified (by themselves and others) with the characteristics thought to pertain to the roles held by the group. Multiple research studies suggest than Caucasians are more likely than African Americans and Hispanics to view themselves as powerful and strong. Unbalanced power relations, such as between the trusting and the trusted parties, are unstable and are likely to lead to a variety of "cost reduction" actions or "balancing operations". Majority members confronted with the decision of whether or not to trust are likely to choose a balancing action that reestablishes their control: namely, quitting the relationship. Members of minority groups are less accustomed to holding power and thus are expected to act less to protect their power.
Trust has been defined as the "willingness to accept vulnerability." We open the black box of vulnerability and argue that people may distrust because they are averse to the risk of being worse off compared to the status quo of no trust, the risk of being worse off than someone else and/or the risk of being betrayed by the trusted party. This paper disentangles these effects and shows which motives matter for different groups of people. We focus on different ethnic groups. A group's status and power helps us understand why people do not trust each other.
We employ a method allowing participants to indicate how vulnerable they are willing to be in different situations. Their MAP, the minimal acceptable probability of success, gives us their willingness to choose a gamble over a certain option--thereby eliminating consideration of different estimates of the probability of betrayal.
We find that all groups are averse to the risk of losing money compared to the status quo. Any policy intervention or management strategy aimed at encouraging trust has to take this into account. Interventions can focus on decreasing the magnitude of the (net) losses involved or the (net) likelihood that the losses will occur. Based on our results, we expect the former strategy to work well for African Americans and Hispanics while the latter strategy seems advisable for Asians and Caucasians. Given their risk preferences, Asians and Caucasians tend to distrust because they fear betrayal. African Americans and Hispanics tend to distrust because they have a strong aversion to receiving an inferior outcome compared to their counterpart; the specter of betrayal does not loom large in their consideration.
Institutional mechanisms such as insurance and compensation encourage trust by decreasing general risk and the risk of inequality in case of betrayal. Our results suggest that such protection may successfully enhance African Americans� and Hispanics� trust. They fear losses and comparative disadvantage more than they feel any sting of pride brought on by betrayal. In contrast, Asians and Caucasians will not be as impressed by decreases in losses and payoff differences. To encourage their trust, the potential for betrayal has to be decreased, for example, by creating incentives for the trusted party to be trustworthy. Stricter law enforcement and punishment for breaches of trust, repeated interactions and reputational concerns may align the trusted party�s interests.
Recently, Alan Greenspan remarked: "American capitalism is turning back to trust and integrity as an antidote to corporate scandal." (Balls, 2004, p.1). This paper suggests that anyone wishing to increase people's willingness to trust may benefit from taking the relative importance of various motives into account and use the corresponding institutional devices to address them.
We consider a common decision format for the three situations. In each, the DM must choose between S and T. S results in a sure outcome S and T in a risky outcome that can either be G (good) or B (bad) in monetary payoffs for the DM. The DM's preference ordering is G > S > B. We first see what happens when the outcome after T is determined by the trustee, whose payoffs at B, G, and S, are respectively C, H, and S, with C > H > S. (For mnemonic purposes C pairs with B and H with G. See Figure 1.) This implies that the trustee would prefer that the DM pick T. However, if T is picked, the trustee would have the temptation to "betray" and select B, making the DM worse off than if she had chosen S. We then consider the decision when the chance outcome after T involves a lottery (the decision problem). Finally, we develop the risky dictator game, which gives a second person payoffs from the lottery decision; those payoffs are the same as those of the trustee in the trust game.
Figure 1: Monetary payoffs in the trust game, the risky dictator game and the decision problem
(First payoff to DM. Second payoff to trustee (TG) or recipient (RD). No second payoff in DP.)
(Embedded image moved to file: pic29358.jpg)
To compare the three decisions, we elicit DMs' minimum acceptable probabilities (MAPs) in each situation of getting the good outcome that just leads them to prefer the lottery to the sure payoff. We inform subjects that prior to the experiment we determined a probability, p*, of receiving G. If their MAP is higher than p*, they will earn S. However, they will play the gamble with probability p* if their MAP is lower than or equal to p*. The higher one's MAP, the higher p* must be for the person to choose T over S. Thus, the less one likes one or both outcomes in T, the higher will be one's MAP. This mechanism is incentive compatible, i.e., a rational DM should be indifferent between S and the gamble with the reported MAP, since individuals cannot affect the probability they receive in the lottery. This is equivalent to asking a consumer her willingness-to-pay for a good, with the understanding that the good will be purchased at the market price if the market price turns out to be lower than the stated price. She has no reason to misrepresent. Our procedure is related to the Becker-DeGroot-Marshak elicitation procedure, but unlike them we do not generate p* randomly from a uniform distribution.
Our experiments were run with a representative sample of Americans by TESS in June 2004. Using the experimental protocols developed by the researchers, TESS run the experiments over the internet using a standardized procedure. As in the experimental laboratory, participants included subjects who had participated in other studies before and as in the laboratory, subjects did not take part in any other study while participating in our experiments. As is customary in economic experiments, subjects were paid for performance. Specifically, at the beginning of the study, we informed subjects that they would be paid according to their decision in one of the three choice tasks predetermined by the experimenter. The order in which subjects were confronted with the three tasks was varied. We did not find any order effects. All games have identical payoffs for the decision maker. The risky dictator and the trust game include a second player. In the experiment, S=$10, B=$8, G=$15, C=$22 and H=$15. The payoffs were presented to subjects in a matrix form with neutral terminology. Payoffs were given in points and for the game chosen, converted 1:1 into US dollars. To simplify payment, all subjects were paid according to their choice in the decision problem.