Finite sample properties of causality in variance tests : a Monte Carlo study

Abstract

The principal aim of this study is to investigate the finite sample statistical properties for a group of methodologies which have been proposed in the econometric literature for the detection of causal relations in variance as well as in the mean. More specifically through extensive Monte Carlo simulations it attempts a thorough examination of the performance of the various methodologies under different states for the series studied as well as under the presence of long memory. For this purpose it uses alternative Data Generating Processes such as the GARCH and FIGARCH models that will be described in detail in the following chapters. The methodologies belong in two categories. The first one is based on the estimation of the cross correlation function for the (squared) standardized residuals that are obtained from the estimation of univariate GARCH models, while the second one uses the residuals of the same models but takes the form of a Lagrange Multiplier test. The innovative part of this study lies on the fact that for the first time to our knowledge such an extensive in depth analysis of the competing causality tests is conducted. This feature makes this study particularly important for both theoretical and applied econometricians as well as for the professionals in the field of financial economics.