Portfolio Optimization: Living Up to Expectations

Abstract

<jats:title>Abstract</jats:title> <jats:p>This paper addresses some of our experiences in working with practitioners dealing with the adoption and implementation of formal risk and portfolio decision methods. The emergence of project investment case histories based on formal Decision and Risk Assessment (D&amp;RA) methods provides a mixed bag of results. Though generally better than historical deterministic methods, the failures are just as illuminating as the successes. In our analysis of the case histories, we find that some of the failures can be attributed to several simplifications formalized in the process. In this paper we explore the impacts of three areas of simplification that, whilst wellintentioned, have potential to cause adverse initial reactions by professionals and management.</jats:p> <jats:p>The first simplification covers the handling of the quality and validity of the input data used to generate the output metrics ultimately forming the foundation of the portfolio decisions. We show that ignoring intra-asset dependencies, or worse, using a P10,P50, P90 characterization derived from a base case plus an upside and downside (or even just guestimates) significantly distorts the real NPV distributions.</jats:p> <jats:p>The second area arises from simplifying or ignoring the dependencies (or correlations) between the various assets. These inter-asset dependencies may be regional, fiscal, financial, global, technical, or life-cycle, to name but a few.</jats:p> <jats:p>The final area of simplification is in optimizing against the expected value of the portfolio. The issue here is that there really is a distribution of possible returns for any given level of risk. If there are binding constraints then there is a substantial probability that the expected return will not be achieved. (This is analogous to the arguments that insist we need to look at the full distribution of NPV for an individual asset, not just its expected value). This problem can be overcome by applying stochastic optimization techniques.</jats:p> <jats:p>Using a typical suite of assets that a company might select a portfolio from, we demonstrate that these simplifications cause some very real and, economically, very significant concerns. They result in selection of a sub-optimal set of assets and participation levels, which potentially leads to a different portfolio risk profile and lower returns than expected.</jats:p>