Sensitivity of seasonality effects on mean and conditional volatility to error distributional assumptions: evidence from French stock market

• Autores: João Monteiro
• Localización: Ecorfan Journal, ISSN-e 2007-1582, Vol. 3, Nº. 8, 2012, págs. 879-898
• Idioma: inglés
• Enlaces
  • Texto completo (pdf)
• Resumen

  • This paper examines seasonality effects on both the return and volatility of a GARCH family model for the French CAC-40 daily index returns. Four calendar effects � day-of-the-week, turn-of-the-month, month-of-the-year and holiday effect � are simultaneously examined. We examine the changes in inferences that might occur when the error terms of descriptive models modeling volatility are specified under different error distributions: normal, Student�s-t, generalized error distribution and double exponential distribution. The usefulness of the in-sample significant estimated seasonality patterns for out-of-sample forecasts in return and volatility is also examined. We find that the few significant seasonality patterns in descriptive models, in the mean and conditional volatility equations, are sensitive to the underlying distributions of the error term. Additionally, the significant estimated effects are not useful in explanatory models and do not introduce predictive ability against the random walk model.