Multivariate Risk-return Decision Making within Dynamic Estimation

Risk management in this paper is focused on multivariate risk-return decision making assuming time-varying estimation. Empirical research in risk management showed that the static "mean-variance" methodology in portfolio optimization is very restrictive with unrealistic assumptions. The objective of this paper is estimation of time-varying portfolio stocks weights by constraints on risk measure. Hence, risk measure dynamic estimation is used in risk controlling. By risk control manager makes free supplementary capital for new investments. Univariate modeling approach is not appropriate, even when portfolio returns are treated as one variable. Portfolio weights are time-varying, and therefore it is necessary to reestimate whole model over time. Using assumption of bivariate Student's t-distribution, in multivariate GARCH(p,q) models, it becomes possible to forecast time-varying portfolio risk much more precisely. The complete procedure of analysis is established from Zagreb Stock Exchange using daily observations of Pliva and Podravka stocks.