Long Memory in Volatility or Parameter Inconstancy? The Case of Prague Stock Exchange

The paper summarizes some stylized facts in financial returns, in particular so called long-range dependence and IGARCH effect. The long range dependence demonstrates itself by a very slow decay of the sample autocorrelation function of absolute and squared returns, especially at larger lags. However, it has been argued elsewhere that these effects can be explained by nonstationarity due to shifts in unconditional variance. One possibility how changes of unconditional variance in GARCH model can occur is to allow for time-varying parameters. In this paper, we verify whether the above mentioned stylized facts can be observed in Prague stock index PX 50 and we perform the test of parameter constancy of the estimated GARCH model.