A Descriptive Model

Abstract

In this chapter, we present a descriptive model which characterizes experimental subjects’ behavior when they play the announcement mechanism. The observation of experimental data and, particularly, some regularities in the deviation of data on average from theoretical predictions suggest that there is a possible explanation of aggregated data. We have observed that players are divided into two groups: players who cheat and players who do not cheat (see Subsection 2.5.2). Moreover, the average values from which players start to cheat are very close for the three different positions (81.7, 81.2 and 80.5). We will say that a value is low when it is strictly lower than 81, and high when it is equal to or higher than 81. As we have seen in Subsection 2.5.2, players who cheat do not do so according to the theoretical prediction. There, we have proposed a simple model that describes how players who cheat announce their values. Our aim, in this section, is to present a more general model, which assigns the conditional probability that a player cheats when he announces in the different positions, given his value and the others’ announcements. To estimate these probabilities, we use a logit model. After having estimated the probability that a player cheats when he announces in one of the three positions (Section 3.1), we show how the descriptive model works (Section 3.2). We simulate players’ behavior in our descriptive model, and we compare the simulated players’ behavior with experimental subjects behavior.