SUMMARY
This article studies the serial dependence and the speed of adjustment to newinformation of weekly portfolios returns of stocks traded in the Santiago deChile stock exchange. Portfolios grouped by size and traded volume during theperiod 1991-2000 are considered. With the purpose of detecting the predictivepower of the lagged returns of certain groups of stocks on others, the studyanalyzes autocorrelations, crossed-serial correlations, Dimson regressions andvector autoregressions.The evidence indicates that weekly returns are significantly autocorrelated, witha significant crossed-serial effect as well: a 1 percent shock in the returns of themost traded and large (prime) stocks predicts a significant cumulative returnbetween 0,4 and 0,5 percent in the other stocks. There is also evidence of aseparate liquidity effect and, in a smaller magnitude, of a size effect, whichimply the existence of cross-serial correlation. Above all, however, the jointeffect prevails. This evidence supports the hypothesis of a delayed reaction toinformation of the smaller and less liquid stocks. Given the order of magnitude,the effect could be exploitable.