A multi-criteria model to explain some empirical violations of basic assumptions of expected utility theory

In economics, the Expected Utility (EU) model is the normative benchmark for decision making under risk. However, empirical studies have shown that many behavioural phenomena are inconsistent with the predictions of this model. Observing empirical evidence, behavioural economics has demonstrated that people do not always make rational, or optimal, decisions, even if they have the information and tools with which to do so. People’s choices are influenced by their emotions and impulsiveness, and depend on their environment and circumstances (context): overconfidence, loss aversion, and self-control are the main statements in behavioural economics. Starting from past studies concerning the use of multi-criteria method based on prospect theory for describing people behavior (Gomes and Lima 1991), this paper aims at proposing a multi-criteria method as a tool to model some behavioural anomalies. It begins by analysing some decision problems proposed in the literature (Kahneman and Tversky 1979), showing that people’s preferences may violate some of the axioms of the EU model. Then, it proposes a multi-criteria approach to model human behaviour in these problems, that is, when emotional factors and cognitive biases influence people’s choices: e.g., the certainty effect and the reflection effect. Among several multi-critria methods, the ELimination Et Choix Traduisant la REalité (ELECTRE) method is used due to its main features: it allows dealing with positive and negative reasons in order to model the preferences, including with heterogeneous scales; its preference and indifference thresholds allow taking into account an imperfect knowledge of the data; no systematic compensation exists between “gains” and “losses”. The results show that the ELECTRE III method provides the same preferences expressed by the majority of respondents in the experiments considered. In other words, a multi-criteria tool accounts for some behavioural anomalies, in particular, those relating to certainty and reflection effects.