A cornerstone for the evaluation of any investment strategy – besides the prediction of the expected return – is risk estimation. Both of these aspects have to be taken into consideration for any investment decision made. In order to being able to accurately assess the attractiveness of a potential investment, a risk model needs to reflect the expected risk statistics over the forecasting horizon in the best possible way. Approaching this naively as is done by some portfolio optimization techniques or some simple value-at-risk measures may lead to inaccurate risk estimations. A grossly inaccurate risk estimate, however, makes it all but impossible to accurately assess the attractiveness of any potential investment. For example, a simple Markowitz optimization requires an accurate estimation of the covariance matrix to predict the expected risk. Yet, typically only a simple empirical covariance matrix is used which can result in a substantially inaccurate market risk estimate for the portfolio.

GA Asset Management has done extensive research on quantitative risk modeling for strategy and portfolio construction. Risk models must accurately reflect the risk perception of the investor and describe market statistics as precisely as possible. GA Asset Management employs a great number of models, for example utility functions or Monte-Carlo conditional value-at-risk estimations. The latter approach facilitates the modeling of extreme risks, far beyond the standard normal distribution assumptions. For optimal portfolio constructions we utilize Bayesian methods to incorporate model risks and employ these improved risk models to optimize the portfolio. We regard the results of our research on risk models as a key element of successful strategies and portfolios.

In addition to using quantitative risk models for our portfolio construction, classic risk management has proven its place at GA Asset Management as well. We believe classic risk management is necessary for the analysis and control of market price risks, credit risks, and also operational risks. Any of these risk areas are thoroughly analyzed and controlled in our products and implementations, whenever possible by way of the best possible quantitative characterization of a given market and credit exposure. Furthermore, in the area of operational risk we strictly apply automated monitoring and controlling technologies. However, not all kinds of risk are quantifiable or can be automatically monitored, so to capture these risks in an appropriate way as well and, moreover, have a positive influence on them, we employ many traditional and approved methods of classical risk management.