This paper introduces FOLIO, an expert system designed to assist portfolio managers by automating the complex process of asset allocation. FOLIO addresses the inherent uncertainty and incompleteness in client data by employing a heuristic reasoning approach and a goal programming algorithm to create optimal portfolios.
TL;DR
FOLIO is an expert system that helps portfolio managers by interviewing clients to understand their investment goals and then recommending an optimal asset allocation. It tackles the challenge of uncertain client data using heuristic rules and a goal programming algorithm to find portfolios that best fit client objectives.
What Problem Does This Paper Address?
Modern portfolio theory, while powerful, struggles with the practical reality of client data. Key client characteristics like risk tolerance, tax status, and investment goals are often uncertain, incomplete, or even contradictory. This makes it difficult for even sophisticated quadratic programming algorithms to produce truly optimal and trustworthy asset allocations. The paper aims to bridge this gap by developing an asset allocation program that is sensitive to this inherent uncertainty in client information.
Methods
FOLIO employs a three-pronged approach:
- Interview Functions: A set of simple interview functions to gather client information.
- Heuristic Inference Rules: A forward-chaining production system that uses expert knowledge to infer the client's goals from the interview data. These rules are designed to handle uncertainty and contradictions.
- Goal Programming Algorithm: A relaxation method based on goal programming (specifically, a modified two-phase simplex algorithm) to resolve the client's multiple, potentially conflicting goals into an optimally fitting portfolio. This algorithm minimizes deviations from target values for each goal, with penalties for exceeding or falling short.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Handling Uncertainty is Crucial: The paper highlights that the accuracy of asset allocation is fundamentally limited by the quality and completeness of client data. FOLIO's core innovation is its explicit attempt to model and reason about this uncertainty.
- Heuristics for Goal Inference: Instead of directly asking clients for precise, quantifiable goals (which they may not be able to provide), FOLIO uses heuristic rules to infer these goals from their stated preferences, financial situation, and risk tolerance.
- Goal Programming for Optimization: Goal programming provides a flexible framework for optimizing portfolios against multiple, potentially conflicting objectives. It allows for "good enough" solutions rather than demanding perfect adherence to every single goal, which is often unrealistic.
- Separation of Concerns: FOLIO focuses on asset allocation to fund classes, leaving the selection of specific securities within those classes to the human investment advisor. This simplifies the problem for the expert system and leverages the advisor's expertise.
Key Results
The paper describes the design and operation of FOLIO, including examples of its recommendations. It notes that the algorithm takes approximately five minutes to produce a portfolio for a client with three or four accounts on a lightly loaded DEC 11/730. While quantitative performance metrics are not detailed, the system is presented as a functional expert assistant.
Limitations and Future Work
- Conflict Resolution: Currently, conflicts in inferred goal parameters are resolved by simple heuristics (e.g., taking the maximum penalty). The authors suggest that a more robust theory of reasoning about uncertainty, as discussed in Cohen and Grinberg (1983), is needed for more sophisticated conflict resolution.
- Untested System: FOLIO is described as "as yet untested," though it is being run on extant accounts. Empirical validation and comparison with human advisors are future steps.
- Potential for Advisor Efficiency: The authors anticipate that FOLIO could significantly increase advisor efficiency and serve as a natural front-end for electronic filing systems, potentially alerting advisors to accounts needing adjustments.
- Extending to Individual Securities: While FOLIO currently allocates to fund classes, future work could potentially extend its capabilities to individual securities.
Summary updated