Time series models and cointegration in stock portfolio selection
Time series models and cointegration in stock portfolio selection

Summary/Abstract: Cointegration has become the prevalent statistical tool in applied economics. It is a powerful technique for investigating long term dependence in multivariate time series. Our paper describes a specific portfolio selection method based on cointegration. We construct cointegration model in two stages: at first we examines the association in a long term equilibrium between the prices of a set of financial assets, and in the second stage we use a dynamic model of correlation, called an error correction model based on linear regression analysis of returns. We considered an allocation into portfolio consisting of Dow Jones Industrial Average components and thereafter we compare long term return and risk profile of portfolio focus on cointegration selection process and index DJIA. The cointegration technique enabled us to use long calibration period and provided that portfolio weights do not change too much over time and outperform the index DJIA in post-sample performance measurement.

• Details
• Contents