Schedule - Parallel Session 3 - Social ChoicesIMC Room 246 - 15:40 - 17:10
Intuition and Deliberation in Giving and Punishment
I study why individuals differ in their tendencies to give altruistically and punish norm violators. An online experiment investigates the role of dual systems of processing information, using a 40-item self-report questionnaire to measure (1) individual reliance on unconscious, intuitive feelings and (2) personal tendency to engage in effortful, deliberate thinking. I find that people who think more deliberately tend to resist the impulse to punish. Moreover, a higher cost of punishing reduces both punishment and giving. Lastly, higher reliance on intuitive feelings is associated with greater sensitivity of punishment to a cost increase than to a cost decrease; this asymmetry of sensitivity might be explained by loss aversion.
Rebalancing the Scale of Fairness: Dictator Game with Workloads
Ning Liu; Valentina Bosetti
Fairness presents commonly in daily arguments but often absent in neo-classic economic theories that assumes full selfishness. Experiments on dictator games generally find dictators sharing non-zero proportions of gains with the receivers (Engel, 2011). Such results challenge the descriptive validity of the selfishness assumptions. This challenge would be less relevant for behaviors in the field if people treat experimental gains differently from their own money. One common drawback in dictator game experiments is that the money to allocate is often endowed, and therefore it’s worth suspicion if subjects feel entitled to such money. Started by Konow (2000) and Cherry, Frykblom, and Shogren (2002), studies have investigated the effects of earned money in dictator games. In these studies, the allocation phase is preceded by an earning phase in which the subjects earn the money to allocate in the next phase. Under such settings, i.e., when the dictator earns the money, the allocation to the receivers often plummets to almost zero. These results are usually interpreted as evidence of low concerns for fairness with earned money. However, this interpretation is only about the fairness concern in the dimension of money. Another perspective of looking at the results is that subjects are rebalancing the scale of fairness by allocating zero money to the receiver, an allocation merely matching the allocating of effort between the two. In the current paper, we propose an experiment in which the dictators allocate effort (money) given an exogenous allocation of money (effort). By variating the given allocation, we can test if the effort (money) allocation of the dictator matches the given money (effort) allocation in a symmetric manner in cases where the exogenous allocation of effort (money) favors the dictator or the receiver. In out setting, a symmetric manner would indicate a concern for fairness in allocation, and this indication is not biased by the predetermined allocation in one dimension, as in dictator games with earned money. In addition, our design also allows us to explore how the initial allocation of effort/money interact with the sense of fairness of the dictator.
Experiments on Aggregate Informational Cascades
Elisa Cavatorta; Antonio Guarino; Steffen Huck
A feature in many everyday-life decisions is that decision makers can gather aggregate information about the decisions of other people, but rarely observe all the individual decisions. E.g., a customer deciding whether to dine in a restaurant has some private information on the restaurant’s quality and is able to look inside and see how many people are already dining. He can only speculate about how many people passed by and decided not to enter. The work by Guarino et al. (GEB,2011) studies the properties of social learning in these situations. They show that, in equilibrium, an “aggregate informational cascade” arises only on the observable action – e.g., if a restaurant goes sees many people in a restaurant, he joins the crowd disregarding his own signal. A cascade on the unobservable action never arises – e.g., an empty restaurant with good reviews will not remain empty forever. In related work, Herrera&Hörner (GEB,2013) characterise the process of social learning when agents can only observe one of two types of actions and can observe time passing. Motivated by these theories, we studied how subjects learn from the actions of others in the lab. In a first set of treatments, based on Guarino etal.’s model, subjects make a binary decision in a random sequence and are not aware of their own position. When deciding, subjects are only informed about the total number of people who chose the observable action before them. We put Guarino et al.’s model to test in a small group treatment of size 3, whereby incentives to follow the equilibrium outcome are strong, and a large group treatment of size 10, whereby incentives to follow the equilibrium outcome are weaker – hence a stronger test of the theory. Our results show that on average subject decisions are close to Guarino etal.’s equilibrium predictions. In the small group, theory suggests that an informational cascade occurs when 2 agents are observed choosing the (observable) action. In the lab, 86% of subjects follow the others despite their private bad signal in line with theory. In the large group, theory predicts that the cascade forms after 4 agents are observed choosing the observable action. In the lab, 70% of subjects follow the predecessors’ decision (disregarding their own signal) after observing 5 predecessors making that action, as if they are in a cascade. Informational cascades on the unobservable actions rarely occur. Albeit not every subject behave according to the theoretical predictions, the theory explains the data rather well. In recent treatments, we allow subjects to observe predecessors who made the observable actions with some randomness, and never predecessors who made the unobservable action.Subjects know their order (which plays a role similar to the passing of time in Herrera&Hörner).In this case,we should observe cascades on both actions. We also examine the heterogeneity of behaviour in the various treatments in terms of cognitive skills and personality traits.
Liability Judgment Under Uncertainty and Minimization of Social Losses
The updating of information is an important step in many decisions involving uncertainty and is non-trivial for understanding behavior in dynamic contexts such as entrepreneurial activities or the politics of the precautionary principle. It is sometimes not obvious or even debatable that there is information update involved, and intuitions would diverge. However, except for the discovery of base-rate neglect in the 80s, not much recent attention seems to have been given to either description or application of alternative models of behavior in the updating process. This paper presents a curious result of information update process in liability judgements and hopes to inspire more discussions of this interesting aspect of decision-making. We think of liability as a perceived cause, such as an individual act, of an undesirable event, such as a social loss. It is intuitive to expect the working of a liability attribution system, formal or informal, to enhance minimization of social losses that are subject to externality. Under certainty, liability attributions enable social losses to be directly translated to individual losses, and remove any negative externalities in an individual’s decision process. Even when there is uncertainty with respect to the perceived cause of a social loss, one would still expect liability attributions to work at least in the direction of reducing externality in social losses. Shavell (1983) for example compared two liability systems and concluded that one is more effective than the other in incentivizing socially optimal behavior. In comparison to having no liability system though, both systems would clearly trump in reducing social loss. This paper contends that this is not necessarily the case. Under uncertainty, the judgement of liability can be considered as judgement of posterior probabilities of a potential causing event conditional on the occurrence of a social loss. According to Bayes’ rule, this posterior probability would depend on both the base-rate of the potential causing event and the conditional propensity that this causing event would generate a social loss. An individual trying to minimize liability would then try to reduce both the base-rate and the conditional propensity. However, when there is complementarity (e.g. constant sum) between the loss-generation processes, minimization of total probability of a social loss would mandate opposing directions of movement of the base-rates and the conditional propensities. That is, when base-rate of one particular causing event is lower than the base-rate of the other causing events, it would be optimal for the one causing event to take over more conditional propensities from the other causing events in the loss-generation processes, rather than the other way around. Furthermore, with the presence of base-rate neglect, this dilemma would be resolved and individual would be trapped in keeping lowest conditional propensities and highest base-rates.