Performance analysis of GARCH family models in three time-frames

Abstract

The proposed alternative p-value method can be used in finding the best performing models. The rank of the p-values namely t-test and z-test statistics can overcome the constraint imposed when using the Mean Absolute Percentage Error as the measurement error. It is crucial to select the right model in the right period so that the model can interpret volatility correctly. This study aimed to provide empirical analyses on the volatility of the Dhaka Stock Exchange market during the market crash in 2011. Three sub-samples were considered to represent pre-crisis, crisis, and post-crisis between November 16, 2009 to July 31, 2013 representing 889 observations. Various GARCH family models were fitted in order to capture the volatility and their performances were compared.