Communication in Games and Decision Making under Risk

Sammanfattning: This thesis is divided into two distinct parts. The first part studies communication in games and the second part investigates individual decision making under risk. Chapters 2 – 4, co-authored with Ola Andersson, investigate communication and renegotiation in dynamic games. In Chapter 2, Bertrand supergames with non-binding communication are used to study price formation and stability of collusive agreements on experimental duopoly markets. Prices are found to be significantly higher when communication is costly. A potential explanation of this finding is that, if players can communicate, they can also renegotiate, which makes the punishments needed to sustain cooperation non-credible. However, if communication is costly enough, this credibility problem may be solved; players may prefer carrying out the punishments rather than initiating a costly renegotiation process. In Chapter 3, it is shown that this reasoning is very sensitive. As soon as the action space in the Bertrand duopoly is discontinuous, renegotiation does not eliminate any relevant subgame perfect equilibrium outcome. Therefore, Chapter 4 attempts to put the renegotiation argument to a more direct test. The chapter experimentally investigates cooperation and non-binding communication in a simple two-stage game. The game has a subgame perfect equilibrium where subjects can sustain cooperation in the first stage by threatening to punish deviant behavior in the second stage. In contrast, renegotiation rules out cooperation in the first stage if intra-play communication is possible. The results provide some support for this argument. Less cooperation is observed in the first stage when intra-play communication is possible. Moreover, pre-play communication only has a significant impact on actions when intra-play communication is not allowed. Chapter 5 explores communication from another angle by applying the level-n theory of bounded rationality to a price competition game and performing an experimental test. The level-n theory predicts prices to be higher with communication than without. The experimental findings indicate that communication affects subjects in a way that seems compatible with the level-n model, indicating that people lie in order to fool other players that they believe do less thinking. The two last chapters of this thesis, co-authored with Hans-Martin von Gaudecker and Arthur van Soest, examine risk preferences. In Chapter 6, data from a preference elicitation experiment conducted on a representative sample of the Dutch population, over the Internet, is compared with corresponding laboratory data in order to investigate two types of selection effects. First, selection induced by the experimenter through the choice of sampling population is considered. In the sample of students, the frequency of errors and average risk aversion are lower than in the representative sample. Second, self-selection is showed to matter in the Internet experiment through attracting a sample where men, the more educated and persons with interest and expertise in financial matters are overrepresented. Using the same sample, Chapter 7 estimates a structural econometric model that allows for both observed and unobserved heterogeneity to study preferences toward risk, losses and the timing of uncertainty resolution. The results indicate that people on average are risk and loss averse, but indifferent towards the timing of uncertainty resolution. Females are found to be more risk averse than males. Risk aversion is lower among participants with higher education. Moreover, individuals of low age, high education and high level of wealth make more consistent choices. Finally, it is concluded that, even after conditioning on an extensive set of demographic characteristics, the remaining variation in choice behaviour is very large.