Essais sur la modélisation de la dynamique du taux de change à travers les enseignements de la finance comportementale

The thesis draws lessons from behavioural finance and nonlinear econometrics to build models that provide better explanatory and predictive powers of exchange rate dynamics than traditional models. The first article justifies the failure of traditional models by the failure of their underlying hypotheses and claims that the unique consideration of macroeconomic fundamentals is not sufficient to explain exchange rate dynamics. The second article analyses the success of order flows models and shows the necessity of considering agents’ behaviours to find a robust exchange rate model. The third article considers heterogeneous behaviours to explain exchange rate dynamics. It shows that a rule based on stylised facts about agents’ behaviours provides better exchange rate forecasts than the random walk. The fourth article proposes an exchange rate model based on conventions that prevail in the market. The convention model significantly beats the forecasting power of traditional models. The main message of the thesis is that macroeconomic fundamentals and agents’ behaviours are both essential components to explain and forecast exchange rate dynamics.