Advances and Applications in Statistics
Volume 9, Issue 2, Pages 153 - 176
(August 2008)
|
|
LOCAL DEPENDENCE FUNCTIONS FOR NORTH AND SOUTH AMERICAN STOCK INDEX RETURNS
Thierry Ane (France) and Guillaume Fabre (France)
|
Abstract: Although sophisticated measures of dependence, like copula functions, have been recently introduced in finance, the correlation coefficient remains the most popular measure of association for portfolio managers. In this paper we discuss a new measure of local dependence which is a localized version of Pearson?s coefficient. A simulation study shows that it can easily and accurately be estimated by a kernel-based nonparametric method.
The local dependence functions obtained for North and South American countries allow for a typology of shapes that is consistent with the economic situation of all countries.
We explain how this local dependence function can be integrated partially to obtain several average measures of association very useful for portfolio management. In particular, measures of association in the extremes can be formed when the local dependence function is integrated beyond a threshold.
Without distributional assumptions, we then observe that extreme negative returns are, on average, 8.54 times more dependent than their positive counterparts. |
Keywords and phrases: local dependence, kernel estimation, correlations, association. |
|
Number of Downloads: 356 | Number of Views: 1053 |
|