Schedule - Parallel Session 4 - Group and Interactive Decision Making

WMG IMC Auditorium 0.02 - 11:00 - 12:30

Group-Shift and the Consensus Effect

David Dillenberger; Collin Raymond


Group decision-making is ubiquitous in social, economic and political life. It is well-known that individuals tend to make different choices depending on whether the outcome of interest is a result of their choice alone or also depends on the choices of others in a group. These ”choice shifts” in groups have been demonstrated in a variety of contexts across fields. The available evidence supports the idea that choice shifts in groups are largely predicted by the preference of the majority of individuals. Group decisions reflect two underlying processes: aggregating information known by the members of the group and aggregating preferences of the members of the group. A robust literature has been exploring how private information is aggregated within groups. In contrast, we focus on the latter process. In doing so, we assume that there is no private information, thus simplifying the role of the group decision to that of aggregating preferences. If individuals are expected utility maximizers, their behavior in group situations should be identical to their behavior in isolation (where their decision is the only one that matters). In contrast, we show that individuals have preferences that are strictly quasi-convex in probabilities if and only if they exhibit a ”consensus effect” a decision-maker who is indifferent between two options when choosing as an individual will, in a group context, instead actually strictly prefer the option that is sufficiently likely to be chosen by the group. More precise results are obtained with respect to several well-known models of non-expected utility which can satisfy quasi-convexity, including rank-dependent utility, quadratic utility, and Koszegi and Rabin’s (2007) model of reference-dependence. We extend our analysis to a full equilibrium setting, showing that with quasi-convex preferences we can accommodate observed phenomena such as group polarization, the bandwagon effect, and preference reversals. For example, we demonstrate that, in contrast to the results under expected utility, when individuals exhibit the consensus effect, in equilibrium, group decisions can fail to aggregate preferences properly and in fact strictly Paretodominated equilibria may result.

David Dillenberger

Assistant Professor, University of Pennsylvania

The Rationale of Team Reasoning

Guilhem Lecouteux


Team reasoning theory (Sugden 1993, Bacharach 2006) has been developed to overcome the predictive failures of standard game theory in cooperation and coordination games: the core argument of this theory is that solving coordination games requires introducing a notion of collective intentionality, and that collective intentions may also offer a satisfactory explanation of cooperative behaviours. Unlike social preferences approaches in which we alter individual preferences, team reasoning implies a transfer of agency from the individual to the collective level, without requiring a modification of individual preferences. In this paper, I develop a game theoretical framework based on Bacharach’s (1999) notion of ‘unreliable team interaction’ to model collective intentions. I argue that collective intentions can be represented as the satisfaction of collective preferences that the team reasoners have strategically chosen. I determine the collective preferences that team reasoners should choose, and show that collective intentionality may lead to aggressive behaviours toward the players outside the team in submodular games. I can then show that, in a very large class of games, team reasoning is the only procedure of choice that rational individuals can adopt: individually rational players, actuated by the intention of maximising their individual payoff, are in general better off by becoming team reasoners and being actuated by the collective intention of maximising the payoff of all the members of their team, since the choice of collective preferences gives them an opportunity to make a strategic commitment.

Guilhem Lecouteux

Teaching assistant, University of Bristol

Individual and Group Behaviour in a Life-Cycle Experiment Under Ambiguity

Enrica Carbone; Konstantinos Georgalos; Gerardo Infante


Experiments on saving behaviour reveal both the substantial heterogeneity of behaviour and performance and the difficulties that subjects face in optimally solving similar problems. Previous studies have focused on contexts in which the future income stochastic process was characterised by an objective probability distribution (risk). We present the results of an intertemporal consumption/saving laboratory experiment in the presence of ambiguity (lack of objective probabilities) regarding the future income level. We study the effects of ambiguity to the consumption/savings decisions. Behaviour at an individual choice level treatment is compared to a group treatment where pairs are asked to solve the same decision task. In addition, the experimental design provides the participants the opportunity to obtain partial information and update their beliefs regarding the stochastic income process. The latter allows us to study the effects of learning to the optimal choice. Our results contribute to the understanding of the under-saving and over-consumption patterns that are being generally observed as well as to the understanding of the difference between individual and group preferences when solving a stochastic intertemporal problem. Our results show that groups deviate less compared to individuals from the optimal theoretical predictions. We also find extended evidence of dynamically inconsistent behaviour (myopic planning horizons). The latter allows us to assess the effects of these inconsistencies to the well-being of the individuals and the groups and to explore whether policy intervention is justified.

Enrica Carbone

Associate Professor, Second University of Naples

Measuring the Impact of Social Relationships: The Value of Oneness

Chris Starmer; Simon Gaechter; Fabio Tufano


Social relationships affect many variables that naturally interest economists yet they barely feature at all in conventional economic analysis. In an extensive bank of experiments, we examine the predictive power of a simple and portable tool, developed in social psychology, for the purpose of measuring social relationships. This is the so-called oneness scale. We deploy the tool in an experiment where groups of subjects who vary in the extent of pre-existing social relationships, also play weak-link coordination games. Our results are striking. Despite no possibilities for communication, groups with high oneness are very likely to coordinate on highly Pareto-ranked equilibria while groups with low oneness never do; hence, sufficiently high oneness appears a necessary condition for coordination success. While oneness co-varies, as expected, with various objective characteristics of groups, surprisingly, in the presence of oneness no other factors are ever significant. The effects of oneness are also large when benchmarked against the impact of financial incentives. We view our results as providing significant proof of concept for oneness measurement as a potentially highly productive tool for economic research.

Chris Starmer

Professor, University of Nottingham