Survey_HMS_2003

Abstract

In view of this relation, tests of the Laplace distribution against the GED based on fitting unconditional distributions are thus expected to produce misleading results, because the kurtosis due to non–normality of the conditional distribution is overestimated when ignoring GARCH volatility dynamics. We perform the test using daily (closing) returns on 25 German stocks (see Table 1 for details), which are included in the German DAX index, a blue chip index comprising the 30 largest firms in terms of exchange turnover and market capitalization. As the composition of the index changes over time, we use the 25 corporations that were included in the index over the whole sample period (see Theissen, 2003).1 The data range from December 1996 to October 2001, yielding 1,220 observations for each stock. Continuously compounded percentage returns are used, i.e., rt = 100(log Pt ? log Pt?1), where Pt is the price at time t. Table 1 reports statistical properties of the 25 return series, along with Lagrange multiplier test results for conditional heteroskedasticity, which reveal highly significant ARCH effects of order 5 for all return series but one (Bayer). However, the ARCH effects of order 6 are significant at the 10% level for Bayer.