ABSTRACT
Econometric modelling of time series, discussed in the previous chapters, was mainly centred around modelling the conditional first moment (or conditional mean). Temporal dependencies in the variances and covariances were considered as model misspecifications. To this effect, researchers had developed ways of correcting such misspecifications and time series were studied in the context of homoscedasticity. However, developments in financial econometrics noted that most of the financial time series showed time-varying variances and covariances that captured the risk or the uncertainty element in financial assets. This meant that heteroscedasticity was not to be considered as model misspecification but as an important feature present in financial time series that should be modelled and not corrected.
