In this work portfolio management strategies are applied in a state dependent setting. The two states which were considered are crisis times and normal times. To identify the different states different indicators are applied. The results show, that an indicator based on a Markov switching model for three states offered the best results. In this context macro-economic data is used to predict the switching probabilities between the the different states. The other indicators applied are a heuristic indicator as well as a US recession indicator. To assess the effect of state dependent asset-allocation on performance a variety of well-established portfolios were examined in several out-of-sample tests. The main findings show that certain simple switching strategies based on crisis indicators achieve better results compared to the equally-weighted portfolio. These results are statistically significant with respect to the Sharpe ratio and Omega. In the backtests, the 1/N strategy (or ERC) is the best performing allocation method during normal times. On the other hand, during turbulent times, risk considerations seem play a major role leading to minimum-variance as the key strategy. In a further step an additional constraint on the diversification of the asset weights was implemented in the asset-allocation algorithms. The implied weight diversification of a strategy is forecasted using a linear regression model. Also the first and second eigenvalue of the covariance matrix are forecasted using a linear regression model based on macro-economic data.      «

In this work portfolio management strategies are applied in a state dependent setting. The two states which were considered are crisis times and normal times. To identify the different states different indicators are applied. The results show, that an indicator based on a Markov switching model for three states offered the best results. In this context macro-economic data is used to predict the switching probabilities between the the different states. The other indicators applied are a heuristic...     »