Strategic Asset Allocation with Switching Dependence

This paper revisits the problem of the strategic asset allocation between stocks and bonds. The novelty of our approach is to model the influence of economic cycles on the marginal distributions of asset returns and their dependence structure by a single hidden Markov chain. After a brief review of selected statistical distributions (Student't and Weibull) and copulas (elliptic and Archimedian), we describe how the switching regime model is calibrated using two indices: the CAC 40 for stocks and the SGI Bond 10 years, for bonds. We then propose a dynamic investment policy based on the estimated probabilities of sojourn in each state of the Markov chain. Even though the Markov chain ruling the assets dynamics is hidden, a Bayesian procedure can be used to infer the probabilities of being in a certain state of the economy. The asset allocation can then adapted to provide the highest yield given the most likely state. Having calibrated and estimated the parameters of the model, the performance of static and dynamic strategies are compared by conducting Monte Carlo simulations. Our results show that dynamic strategies, which exploit the additional information relating the probable regime state, perform better than static policies with a limited risk and an acceptable number of reallocations