ABSTRACT
Time-varying volatility models have been popular since the early 1990s in empirical research in finance, following an influential paper ‘Generalized Autoregressive Conditional Heteroskedasticity’ by Bollerslev (1986). Models of this type are well known as generalised autoregressive conditional heteroscedasticity (GARCH) in the time series econometrics literature. Time-varying volatility has been observed and documented in as early as 1982 (Engle 1982) and was initially concerned with an economic phenomenon – time-varying and autoregressive variance of inflation. Nevertheless, it was data availability and strong empirical research interest in finance, motivated by exploring any kind of market inefficiency, that encouraged the application and facilitated the development of these models and their variations. For instance, the GARCH in mean model is related to asset pricing with time-varying risk instead of constant risk in traditional models such as the CAPM. An exponential GARCH (EGARCH) model addresses asymmetry in volatility patterns which are well observed in corporate finance and financial markets and can sometimes be attributed to leverage effects. GARCH with t-distribution reflects fat tails found in many types of financial time series data where the assumption of conditional normality is violated. Finally, multivariate GARCH models are helpful tools for investigating volatility transmissions and patterns between two or more financial markets.
