A1518
Title: Measuring excess volatility \& asset mispricings through the chiarella model
Authors: Jutta Kurth - Ecole Polytechnique (France) [presenting]
Jean-Philippe Bouchaud - Capital Fund Management (France)
Adam Aleksander Majewski - QUANT Lab Pisa (Italy)
Abstract: The two most well-accepted anomalies pervading all financial markets are trend/momentum and value. Their coexistence and interaction may be studied analytically and numerically via a stochastic dynamical system known as Chiarella's model and its extensions. This framework is generalized to include arbitrary drifts in the fundamental value and a trend signal that depends on the mispricing of the asset rather than its bare log-returns, while keeping the same linear stability. The fundamental value is a direct output of the calibration and does not rely on economic pricing models. Numerically, results on synthetic data as well as on several contracts from five different asset classes show that the model provides better fits. Instead of calibrating on asset classes, it is possible to calibrate on individual price series, showcasing a powerful improvement in the fitting procedure. Stylized facts are confirmed, e.g., bimodalities in the mispricing distribution (from a phenomenological P-bifurcation in the model), suggesting that prices are more often mispriced than correctly priced, as well as the non-monotonic relation between past trends and future returns for all asset classes. The robustness of results against perturbations in each parameter and their hierarchy is studied via sloppiness analysis. The excess volatility puzzle is quantified for all five asset classes by comparing the calibrated noise strength of the noise trader with that of the fundamental value.
