Date

2014

Document Type

Dissertation

Degree

Doctor of Philosophy

Department

Information and Systems Engineering

First Adviser

Thiele, Aure ́lie C.

Other advisers/committee members

Storer, Robert H.; Yale, Wilson; Zuluaga, Luis

Abstract

This dissertation investigates robust optimization models for performance attribution analysis in investment management. Specifically, an investment manager seeks to evaluate the performance of fund managers who manage funds he might invest his clients' money in. A key difficulty for the investment manager is to quantify the fund manager's skill when he may not know the fund manager's allocation precisely. This introduces two main sources of uncertainty for the investment manager: the stock returns and the fund allocations. This dissertation proposes and analyzes robust, quantitative models to address this challenge. We study a robust counterpart to the mean-variance framework when the fund managers' precise allocations are uncertain but belong to known intervals and must sum to one for each manager, present an algorithm to solve the problem efficiently and analyze the investment manager's allocation in the various funds as a function of the benchmark return. Further, we consider the case where the stock returns are also represented as uncertain parameters belonging to a polyhedral set, the size of which is defined by a parameter called the budget of uncertainty, and the investment manager seeks to maximize his worst-case return. We describe how to solve this problem efficiently and analyze how the investment manager's degree of diversification and his specific allocations in the funds vary with the budget of uncertainty.

Included in

Engineering Commons

Share

COinS