Intraday-patterns in the Colombian Exchange Market Index and VaR: Evaluation of Different Approaches

Abstract

This paper evaluates the performance of 16 different parametric, non-parametric and one semi-parametric specifications to calculate the Value at Risk (VaR) for the Colombian Exchange Market Index (IGBC). Using high frequency data (10-minute returns), we model the variance of the returns using GARCH and TGARCH models, that take in account the leverage effect, the day-of-the-week effect, and the hour-of-the-day effect. We estimate those models under two assumptions regarding returns behavior: Normal distribution and t distribution. This exercise is performed using two different ten-minute intraday samples: 2006-2007 and 2008-2009. For the first sample, we found that the best model is a TGARCH(1,1) without day-of the week or hour-of-the-day effects. For the 2008-2009 sample, we found that the model with the correct conditional VaR coverage would be the GARCH(1,1) with the day-of-the-week effect, and the hour-of-the-day effect. Both methods perform better under the t distribution assumption.