B1623
Title: Simultaneous estimates of the beta of the market line with generalized autoregressive conditional heteroscedastic errors
Authors: Hisseine Mahamat - University of French Guyana (France) [presenting]
Abstract: The systematic risk of a stock is estimated by the equation of the market line and its beta coefficient. According to the assumptions of the OLS application, the estimators are robust, and the residuals follow a white noise process. However, various papers show that there are many statistical anomalies (stylized facts) in the residuals (heteroscedasticity, autocorrelation and non-normality) that reject the properties of the estimators. The value of the beta can thus be different if an alternative estimation method is chosen. To take these anomalies into account, a class of models on the randomness of regression that have proven their effectiveness in market finance are referred to. This is the class of ARCH processes completed by a more recent model, the realized-GARCH, specific to intraday data. For example, simultaneous estimation of the market line parameters is applied with randomness on the risk premium of Societe Generale and the CAC40 (French stock market index) for the daily period from 2005 to 2015. We verify that the beta from the OLS and the other estimated models are all greater than one. However, they are significantly different, which can modify the behaviour of portfolio managers, for the example chosen, the GJR- model.
