Multiperiod portfolio selection with transaction costs

This paper provides an innovative multiple stages portfolio selection approach with random portfolio weight vector for each period. Quadratic transaction costs are joined into considerations. The underlying assets time series data are supposed to follow a discrete time triangular cointegrated vector model with an autoregressive error model. Under the specified cointegrated model and some mean-variance formulations, the optimal portfolio allocations via the proposed -stage or a standard one-stage approach, are derived in closed forms. Moreover, relationships among the discussed optimal solutions are analytically investigated.