Description:
<jats:p>The presence of long memory in realized volatility (<jats:italic>RV</jats:italic>) is a widespread stylized fact. The origins of long memory in <jats:italic>RV</jats:italic> have been attributed to jumps, structural breaks, contemporaneous aggregation, nonlinearities, or pure long memory. An important development has been the heterogeneous autoregressive (<jats:italic>HAR</jats:italic>) model and its extensions. This article assesses the separate roles of fractionally integrated long memory models, extended <jats:italic>HAR</jats:italic> models and time varying parameter <jats:italic>HAR</jats:italic> models. We find that the presence of the long memory parameter is often important in addition to the <jats:italic>HAR</jats:italic> models.</jats:p>