A VECM Model of Stockmarket Returns

Observations of security prices and other financial time series usually include not only the close (C), but also an open, a high and a low (O,H,L) price for a specified interval. The multivariate vector of values (H,L,O,C) is obviously more informative than just the open or close (O, C) for modelling volatilities and volatility predictions. In this paper we capture the return generation process of security prices by using all the quoted prices (H, L, O, C) via a vector error correction (VECM) model. The results of the empirical models using daily DJI index data for a 11 year period (1990-2000) indicate some interesting stylised facts regarding the market returns. We show, via the return generation process (RGP) proposed, that the "cointegrating returns" exhibit significant explanatory power. Some insights are also provided as to why logarithmic returns tend to be non-normally distrbuted