Decomposing trust into risk preferences, altruism, and subjective beliefs: An experimental analysis

There is a sizeable economic literature dedicated to understanding trust and the extent to which it influences decision making. Although trust is difficult to measure, experimental economics has commonly used the 'amount sent’ in the Berg, Dickhaut and McCabe (1995) investment game to elicit levels of trust from players of the game. However, there is a growing body of literature suggesting that factors such as risk preferences, altruism, and subjective beliefs may confound this measure of trust, thereby questioning the validity of using the 'amount sent’ to elicit people’s levels of trust. To understand these factors, we designed and conducted an incentive-compatible economic experiment with students from the University of Cape Town. These participants completed the investment game, the dictator game (to measure levels of altruism), and a random lottery pair risk preferences task (to gauge risk preferences). We also included an information treatment where students were shown the conditional distributions of amount sent decisions made by students in the previous, baseline treatment. This was done to evaluate whether knowledge of the actions taken by other students would ground students’ beliefs and influence the decisions they made. We estimate a set of standard statistical models to gauge determinants of the amount sent and a complementary maximum likelihood estimation approach to estimate Expected Utility models and Rank-Dependent Utility models in order to further evaluate our data. Our results show that caution needs to be taken when using the amount sent as a measure of trust as the relationship between risk preferences and the amount sent is a nuanced one. Moreover, altruism has a statistically significant association with the amount sent and with risk preferences. We also found that those who were part of the information treatment sent significantly more than those who were not, and they were on average also less risk averse. This indicates that while subjective beliefs do influence behaviour in the investment game, they also affect risk preferences. Thus, our results suggest that researchers should not use the amount sent in the investment game as a pure measure of trust because its measurement is confounded empirically by altruism and subjective beliefs, and theoretically by risk preferences