Conservatisme, représentativité et ancrage dans un contexte dynamique: Une approche expérimentale Avril 2006

Abstract

Several heuristics have been developed by economists and psychologists in order to explain economic behaviour on financial markets. They stress the cognitive bias that affect individual judgments and that partially could explain anomalies observed on financial markets. The aim of our experiment is to test the pertinence of one or the other of conservatism, representativeness and anchorage-adjustment heuristics in a financial context. Its specificity relies on its dynamical context. Fifteen periods along, subjects are givenfinancial information on firm profitability. They are asked to formulate beliefs and to update them accordingly to new information. Econometric treatment of our experimental panel data refutes Bayesian updating: subjects underestimate high probabilities and overestimate low ones. Representativeness heuristic seems to be invalidated in the same way: subjects underweight the most intensive signals and thus, never exploit the whole information. On the contrary, conservatism and anchorage-adjustment are jointly accepted: subjects underweight new information when updating, but this behaviour becomes actually obvious when distinguishing situations in which subjects move away from the anchoring value, from those in which they move closer this value.