Cross-predictability of stock returns: A study of the limited-information model

No Thumbnail Available
Journal Title
Journal ISSN
Volume Title
School of Economics | Master's thesis
Degree programme
PURPOSE OF THE STUDY The purpose of this study is to examine limited-information models which posit that returns on economically linked assets cross-predict each other, and to determine whether they provide a compelling explanation for stock return predictability in a time series. Furthermore, this paper investigates the model predictions concerning the effect of informed investors and investor geographic specialization on return cross-predictability. This paper also investigates self-financing trading strategies that capitalize on return cross-predictability effects. DATA This paper analyzes two samples which include all publicly listed companies traded in Eurozone and EU27 countries over the time period ranging from January 2000 to December 2009. The accounting and stock market data used in this paper are from Worldscope and Thomson One Banker databases, respectively. In addition, consolidated Eurostat input-output tables for years 2000 and 2005 are used to identify customer and supplier industries for each sample company. RESULTS The empirical evidence in this paper shows that previous-month supplier industry returns cross-predict stock- and industry-level returns. On the other hand, previous-month returns in customer industries exhibit only weak cross-predictability effects, particularly in the Eurozone. In addition, the results show that the magnitude of return cross-predictability is negatively related to the number of informed investors. Furthermore, this paper is able to provide new empirical evidence indicating that the magnitude of return cross-predictability is positively related to the geographic dispersion of informative signals diffusing from related industries. Finally, the results show that cross-predictability effects can be economically significant: self-financing trading strategies based on return cross-predictability are able to generate mean annual abnormal returns of up to 9.7%.
limited-information model, return cross-predictability, gradual information diffusion, investor specialization, market segmentation, informed investor, input-output table