engleski

Theoretical Distributions in Risk Measuring on Stock Market

For any investor on stock market it is very important to predict possible loss, depending on if he holds "long" or "short" position. By forecasting stock risk investor can be ensured "a priori" from estimated market risk, using financial derivatives, i.e. options, forwards, futures and other instruments. In that sense we find financial econometrics as the most useful tool for modeling conditional mean and conditional variance of nonstationary financial time series. Besides the assumption of normal distributed returns does not represent asymmetry of information influence, normal distribution also is not the most appropriate approximation of the real data on the stock market. Using assumption of heavy tailed distribution, such as Student's t-distribution in GARCH(p, q) model, it becomes possible to forecast market risk much more precisely. Even more, using Student's distribution with non-integer degrees of freedom leads approximation to minimal differences between theoretical and real values. Such modeling enables time-varying risk forecasting, because the assumption of constant risk measures between stocks is unrealistic. The complete procedure of analysis has been established using real observed data at Zagreb Stock Exchange. For this purpose daily returns of the most frequently traded stocks from CROBEX index is used.

theoretical distribution comparison; non-integer degrees of freedom; heavy-tails; scale and shape parameters; risk measuring; conditional variance; risk forecasting of stock returns

nije evidentirano