Visar MALAJ & Vincenzo GENTILE

Abstract In this paper, we investigate some innovative experimental extensions
of the multiplicative error model, introduced by Engle (2002) for positive valued processes and specified as the product of a conditionally autoregressive scale factor and an innovation process with positive support. We suggest a flexible MEM specification in which the process for realized volatility can be seen as a mixture of two Gamma MEMs according to De Luca and Gallo (2007). We also implement an original Gamma model for a MEM with time-varying parameters which depend on the value of the past. These specifications take into account the long memory pattern of the volatility process and the typical slow decay of the respective autocorrelation function. We apply these models to the volatility of ‘STOXX Balkan 50 Equal Weight’ Index, which represents the largest and most liquid companies across eight Balkan countries.
Key words Stock Market Data, Volatility, MEM models, Balkan stock index
JEL classification G15, C58
PDF format below