A proposed framework for computer-driven portfolio management systems

Abstract

This research examines the effectiveness of a fully automated, price-driven portfolio management system. It examines the practical application of several seemingly unrelated theories for the purpose of creating a computerized system that makes allocation decisions given only yearly price inputs for a set of asset classes. The most relevant of these theories is the economic long wave cycle, otherwise known as the Kondratieff cycle, which was named after a Russian economist who documented the cyclical nature of price movements spanning several decades.

The proposed application is tested over a period encompassing the economic downturn during the Great Depression. Only year-end price data is used to make allocation decisions and to measure the effectiveness of the portfolio management system. As such, the solution aims to create efficient portfolios for those with relatively long-term investment horizons.

The results of this research may be of great interest to institutional investors who wish to shield themselves against what many consider rare phenomena, such as depressions and hyperinflations. For the most part, portfolio managers rule out such scenarios because they narrow their focus on relatively recent market behavior, one which they have personally experienced, and they fail to recognize the existence of cycles that extend beyond a generation or more. This thesis breaks free from the traditional linear thinking of modern portfolio theory.