Normal versus student in measuring value at risk: an empirical bayesian overview

• Autores: Gonzalo García-Donato Layrón, Pedro Gento Marhuenda, Juan Francisco Ortega Dato
• Localización: Documentos de Trabajo ( Universidad de Castilla La Mancha. Facultad de Ciencias Económicas y Empresariales ), Serie 1, Nº. 4, 2001, 15 págs.
• Idioma: español
• Texto completo no disponible (Saber más ...)
• Resumen

  • Gaussian distribution is usually assumed in financial returns (the Variance- Covariance method) to calculate Value At Risk (VAR). It is well known that financial returns exhibit high positive kurtosis, therefore the Student model has been proposed as more adequate for explaining the distribution of financial returns. Besides, the models are managed from a frequentist statistical inference view. We apply Empirical- Bayesian (EB) techniques to the Normal and Student models in order to obtain VAR. Both VAR are easy to calculate and that resulting from the Student model has an appealing interpretation in terms of kurtosis. Both methods are applied to daily return observations in six international indexes during the last decade. The results show that the Student model is more accurate than the Normal one. They also demonstrate that some statistical characteristics, other than positive kurtosis, exist in the financial returns, which a_ect the goodness of the results.