Do fat tails matter in GARCH estimation: testing market efficiency in two transition economies

The use of the GARCH-class of models is commonplace when examining stock market returns. In this paper we use data on stock markets in two transition economies, the Czech Republic and Romania, to demonstrate the importance of using the correct GARCH specification. When residuals are characterised by fat tails or kurtosis, the use of a GARCH-t specification is appropriate. Diagnostic tests suggest that the GARCH-t specification is appropriate for modelling stock market returns in Romania, whilst the standard GARCH specification is adequate for the Czech Republic. Using a standard GARCH specification leads to rejection of the null hypothesis of market efficiency in Romania, whereas this null hypothesis cannot be rejected using the GARCH-t specification. The null hypothesis of efficiency cannot be rejected in the Czech Republic using either specification. Thus, we find that the presence of fat tails can have important implications for inference in the analysis of stock market returns.