A Satisficing Model For Project Selection

Thispaperconsiderstheproblemofprojectselection, subjecttoabudgetandotherconstraints. The return of each individual project is uncertain, and its probability dis-tribution is only partially characterized. The criterion for project selection is based ona satisficing approach that evaluates how well the uncertain total return achieves a pre-specifiedtargetreturn, consideringthediversificationpreferenceofthedecisionmaker.Thus, our model maximizes the value of a risk aversion parameter, while keeping thecertainty equivalent of the uncertain returns under exponential utility above a giventarget. We allow interactions, for example synergies, between the returns of differentprojects. We also allow correlation between the uncertain returns of different projects.Our solution procedure solves a small sequence of subproblems via binary search andapplies a cutting plane procedure to test feasibility at each subproblem. An extensivecomputational study shows that the satisficing model generates more effective projectportfolios than any other project selection approach.