It depends on how you define trust. One way to conceptually distinguish the concept of risk aversion from trust is to define trust as a formal belief - the belief that a generic other will not cheat you, for instance. This is the path I and co-authors follow in a working paper "Trust and Cheating" (http://www.nber.org/papers/w18509). Among other things, we find that both trust beliefs and risk aversion affect trusting behavior. This is not in the paper, but our raw data suggest that trust beliefs and risk aversion are essentially uncorrelated.

Hi Jeffrey

Thanks for the response. I see trust as a form of belief about other persons behavior, that they are trustworthy. Such a belief may be naïve or optimistic/pessimistic and not be based on experience when it is related to an anonymous person but can be more realistic if it relates to a known person or group of persons with some cultural or other identity. A naïve person taking a taxi in a city he does not know may not even notice that the taxi driver takes him on a long tour to get extra pay. You don't think that such a person also could be less risk averse or loss averse?

I feel trust and risk are not comparable. While trust is form of human behavior, risk is the inherent unknown outcome of this behavior, if we are considering humans, which I understand what you meant here.Dealing with a person who is trustworthy bears relatively lower risk than the person whose character is unknown to us. However, if we move beyond the human behavior, trust can not be on the same level as risk because both are not applicable simultaneously.

The problem here may be to distinguish perceived risk and risk aversion or risk preferences more generally as behavioral characteristics. Recent studies of risk preferences have revealed differences in behavior related to losses and to gains and different behavior related to known risks and uncertainty. Shocks have also been found to affect the degree of risk aversion in risk preference experiments. It is this "preference instability" that triggered my thought that risk preference experiments with given objective probabilities are not waterproof against respondents manipulating these objective probabilities if they are "optimists" or "pessimists", are "naïve" or very have had some recent bad experience..... But how to study this?

Risk is the possible damage, loss, or danger from an unforeseen event. Trust is a behaviour or feeling that the person whom you trust will not harm you. Since risk is the possible outcomes from the uncertain events, trust may not be directly positively correlated with risk averse.

Both risk and trust relate to unforseen future events. Risk preferences and trustfulness embed subjective beliefs about the future and how nature and other persons will behave whether they "play a game with nature" or a game with other persons.... In the classical risk preference games that Binswanger played with peasants in India he found that luck in earlier rounds made players less risk averse in later games. These were games with coin tossess where the objective probability should be 0.5. Perhaps players who won believed that this is my luck day and adjusted their subjective probabilities in the following games?

Hi Stein (if I may),

In response to your reply to my reply above concerning the naive taxi cab passenger, I would agree that it seems a perfectly reasonable conjecture that the naive, trusting, passenger is also likely to be more risk tolerant or less loss averse. This conjecture is, in theory, testable, provided one can measure propensity to trust, risk tolerance and loss aversion separately and convincingly. I am not aware of very many studies that do this, precisely because it is difficult to separate tolerance for social risk (trust) from tolerance for financial risk (risk tolerance and loss aversion). The only data I am aware of is my own, which I mention above. I don't claim to have the perfect measures of any of these phenomena, but as far as I can tell the correlation between willingness to tolerate social risk and willingness to tolerate financial risk is essentially zero. I am as puzzled as you by this, but that is what the data are telling me.

I think the short answer is no. Of course the notion of trust is not universally defined, but from my own work, I find risk preferences have little predictive power as to a participant's decision to trust or not, albeit this was not my research question (Sheremeta & Shields 2013). This particular question was also studied in Eckel and Rick K Wilson (2004).

Do Liars Believe? Beliefs and Other-Regarding Preferences in Sender-Receiver Games. Roman M Sheremeta and Timothy Shields, Journal of Economic Behavior & Organizations, 2013 94

Is trust a risky decision?. Catherine C Eckel and Rick K Wilson, Journal of Economic Behavior & Organizations, 2004 vol. 55 (4) pp. 447-465

Thanks Timothy for useful references. These findings are quite surprising, as stated also by Eckel and Wilson. One of the challenges is that different approaches to measure risk preferences also show low correlation. I have found the same myself using the Holt & Laury approach with hypothetical real life framing versus experiments with money. Another difference between the Holt and Laury approach and the trust games played by Eckel and Wilson is that the H&L approach is about risk (known probabilities) while trust games with unknown stragers is really about uncertainty. This could also be one of the reasons for low correlation. Trust games between people that know each other may be different, or if people you play with are from a closed social group with clear social norms to reciprocate for example. I have played trust games with siblings and found them to trust each other more than their parents while in dictator games they tended to be more generous towards their parents than towards their sibling. And trust and generosity was closely correlated.

Hi Stein and Timothy,

If the question is about trusting behavior (e.g., amount sent in the trust game), then I think the answer is that trusting behavior is related to risk preferences simply through the monetary consequences of trusting behavior. You can see this in the Eckel and Wilson paper mentioned by Timothy: expected monetary returns are a substantial determinant of the decision to send some money to the co-player. Most empirical papers on trust control for risk preferences in their analyses precisely because of this.

My comment was in response to measuring a true underlying propensity to trust and its relation to risk preferences. As far as I know, there are no papers which tackle this issue. Ideally one would have to use something like the Bohnet and Hauser betrayal aversion (2004) methodology because this compares apples to apples when constructing the residual of trusting behavior that is not explained by preferences towards risk, altruistic (social) preferences or preferences toward uncertainty. However, all of the betrayal aversion studies I am aware of use a between-subjects design, so no individual-level comparison is possible.

Although, come to think of it, I have data from an experiment combining the betrayal aversion methodology with a within-subjects design. I just looked at the raw data: risk tolerance has a strong positive impact on the likelihood of betrayal aversion. The impact is large in magnitude: increasing the measure of risk tolerance from its lowest possible value to its highest possible value decreases the probability of being betrayal averse by 33 percentage points, which is huge, representing 70% of the sample mean of betrayal aversion.

References

Bohnet, Iris and Richard Zeckhauser (2004), “Trust, Risk and Betrayal.” Journal of Economic Behavior and Organization, 55(4), pp. 467-484.

Hi Jeffrey

Thanks for valuable ideas and references. I think we are closer to a separation of the risk preferences and trust. Bohnet et al. (2008) has an extended assessment and find that people are less willing to take risk (more risk averse or "betrayal averse") when playing with other people than when they "play with nature". It may feel worse to be fooled by people than by nature. However, still "nature risk aversion" and "people risk aversion" (betrayal aversion) may be correlated as your finding seems to indicate. Fehr (2009) is also an interesting paper on the issue how risk preferences and other social preferences are related to trust. But how much of trust can be explained by low betrayal aversion? What is left after we peel of risk aversion, betrayal aversion and generosity? Beliefs may be important for trust but they are not affected by oxytocin while trusting behavior is. Is what is remaining related to inequality aversion and expectations about peoples trustworthiness? Bolton and Ockenfels (2010) seem to find that inequality aversion may be another important issue.

References

Bohnet, I., Greig, F., Herrmann, B., & Zeckhauser, R. (2008). Betrayal aversion: Evidence from brazil, china, oman, switzerland, turkey, and the united states. The American Economic Review, 294-310.

Bolton, G. E., & Ockenfels, A. (2010). Betrayal Aversion: Evidence from Brazil, China, Oman, Switzerland, Turkey, and the United States: Comment. The American Economic Review, 100(1), 628-633.

Fehr, E. (2009). On the economics and biology of trust. Journal of the European Economic Association, 7(2‐3), 235-266.

Personally I think risk preferences do shape behavior in trust situations, however what I think you'll find is the variance in beliefs, distributional preferences (other regarding), etc... will simply swamp out the variance in risk preferences. I've ran several variations of sender-receiver games, and despite using several different elicitation methods, risk preferences are the least predictive. Again, this does not mean they are not important, but are swamped out by the variances in the aforementioned reasons (at least with N I can afford!). What I can say is the participants behavior is generally too trusting to be explained solely by risk preferences and comparable lottery choices.

Hi Timothy,

The effect of risk preferences on trusting behavior is not always negligible. If you look at Table 6 of a working paper of mine "Trust and Cheating" (http://www.eief.it/files/2012/10/wp-13-trust-and-cheating.pdf), I estimate the impact of risk aversion on the intensive margin of trusting behavior (how much to send in a trust game) to be negative and highly significant. The magnitude of the effect of a one unit increase in (Holt and Laury) risk aversion is roughly equivalent to about a five percentage point increase in the probability of being cheated---so, not huge but non-trivial. In most studies, the intensive margin and the extensive margin (whether or not to trust) are confounded or not separately considered, reducing the apparent impact of risk preferences: I too find no significant role for risk preferences on the extensive margin. This is possibly what is going on in your paper. Don't get me wrong, it is a great paper that I really enjoyed reading and would not mind discussing with you further at some point. But, in the end, trusting in your paper is a binary decision so that you only consider the extensive margin.

I completely agree with you; I think the other factors are simply swamping risk preferences out. In that particular paper the variance of beliefs overwhelms the variance in risk preferences, so it's no wonder why statistically they do not show up. I think one might be able to find something significant, but you would want (i) continuous action space, (ii) control over the value of the payoff, that is, that errors in estimating the payoff do not swamp risk preferences, and perhaps (iii) some control for beliefs.

Also, I would love to see your paper on using the betrayal aversion methodology. Let me know when it's in working paper form.