Markov Regime Switching Model Implementation to the Stockholm Stock Market & Comparison with Equal Weight Portfolio

The unpredictable behaviour of financial time series has long been a concern for econometricians, making it difficult to find appropriate models with a satisfactory fit. The Markov regime switching model is a popular approach, much in behalf of the way it takes the shifts in the time series behaviour into account. The model in this thesis is based on a mixture of normal distributions, extended to include a Markov switching behaviour. As the behaviour of the time series changes, regime switches are assigned to it, making the time series alternate between a predetermined number of states. After the implementation, on two portfolios à seven stocks selected from the Stockholm stock market, the examinations indicated that the fit of the model could be improved by changing the number of states assumed in the estimation. It was found that a Markov regime switching model with three states had the most satisfactory fit to the time series. Subsequently, one of the modelled portfolios was allocated to maximize the Sharpe ratio. This led to some unfavourable extreme allocations, and upon comparison with a portfolio of equal weights containing the same assets the results were poor. Despite a higher yearly return, the modelled portfolio displayed significantly larger volatilities, leaving the results of this evaluation inconclusive. Nevertheless, the implementation lead to a significant improvement in the autocorrelation of the absolute residuals, along with giving the residuals a substantially more homogenic appearance. These results indicate that most of the significant dependence structure has been captured, in particular by the three-state model