Bayesian Evaluation and Selection Strategies in Portfolio Decision Analysis

Practically all firms pursue goals by selecting a portfolio of courses of action that consume resources. Yet, due to uncertainties such as unforeseen market developments, it may not be possible to know in advance how much value these actions yield. Thus, the selection must be based on uncertain ex ante estimates about the future values of the actions. In this paper we show that the explicit Bayesian modeling of these uncertainties serves to increase both the expected value of the selected portfolio and the expected number of those selected actions that belong to the ex post optimal portfolio. Bayesian modeling is also shown to eliminate the expected negative gap between the true and estimated portfolio value, which is caused not by biased value estimates but by the statistical properties of optimizing under uncertainty. Furthermore, we offer analytic results for the optimal targeting of additional evaluations and show that it often suffices to re-evaluate only a subset of the actions. Finally, we propose a novel project performance measure, which is the probability that the project belongs to the optimal portfolio.