COMOVIMIENTO EN SERIES TEMPORALES APLICADO A LAS RENTABILIDADES DE LAS ADMINISTRADORAS DE FONDOS DE PENSIONES (AFP) CHILENAS

Abstract

In this work different methods to measure the comovement between time series are presented. Comovement is a concept suffciently studied in the economic context, for example in the international stock market. The comovement is focused on measuring the degree of correlated movement between different time series. Over time there have been various methods for measuring the comovement, so in this work we summarize 8 techniques found in the literature: nonparametric approach, time-frequency space, Box's M Test and principal component analysis, dispersion coefficient, GARCH models, Granger causality test, copula and linear regression. Finally, to learn about the agreement between these methods an actual dataset related to the returns of Chilean pension fund administrators (AFP) is considered. The magnitude of comovement, measured with different methods, is almost perfect among all di erent pension funds indicating that the AFP are investing in the same places and getting the same returns, i.e., there is little competition between them.