Institution: Yale University Proposal Number: 9988376
In scientific disciplines, significant issues are sometimes assumed away by overmodeling, which results in lost knowledge. Such overmodeling is often caused by analytical or computational difficulty of the issues involved. By shifting the existing balance among modeling, analytical, and computational components of the current solution to an issue, one inevitably discovers new technologies. Motivated by this research philosophy, this project has three general themes. The first is to explore trade-offs between modeling and computation in finance with emphasis on advancing computational technologies for practical and theoretical finance problems. The second is to develop finance ideas to help solve computer science problems. The third is to formulate novel finance problems which break away from traditional paradigms of finance research.
This project pursues these three themes in five interrelated areas of computational finance: (1) market modeling, (2) investment and trading algorithms, (3) risk management, (4) market index design, and (5) auctions.
For the area of market modeling, the main goal is to study short-term predictability of stock markets using agent-based approaches. The models established will be used to test algorithms developed for other parts of the project.
For the area of investment and trading algorithms, this project focuses on two fundamental issues: (1) what sorts of information can be used to enhance the performance of a portfolio and (2) how to measure the performance of investment and trading algorithms. Results in this area can be useful for long-term investors such as those who save for retirement.
For the area of risk management, this project investigates two basic issues: (1) how to assess the risk of a portfolio of assets and (2) how to design optimal portfolios which meet security requirements. Results in this area may be used to help institutional investors detect and minimize unwanted risks.
For the area of market index design, novel optimization problems will be formulated to help design portfolios which are easier to manage than popular market indices but can track or outperform the indices.
For the area of auctions, the primary goal is to use auction algorithms to design solutions to computer science problems. To this end, the project will identify auction primitives which can be used to construct complicated auction mechanisms.
