The structure of multifactor equity risk models
- PDF / 131,145 Bytes
- 10 Pages / 598.664 x 767.575 pts Page_size
- 54 Downloads / 251 Views
Jason MacQueen is Chairman of Alpha Strategies LLC, an investment consulting firm operating out of Philadelphia and London. He is also the founder of QUANTEC Ltd, a leading provider of investment technology and risk models for international investors. He was educated at Oxford and London Universities, where he read mathematics and theoretical physics. Chairman, Alpha Strategies, 17a Park Lane, Bath B1 2XH, UK Tel: ⫹44 (0)1225 448571; Fax: ⫹44 (0)1225 443931; e-mail: [email protected]
Abstract A number of commercial risk models have been available to institutional fund managers for the last two decades, and while there has been considerable discussion as to their different choices of factors, their different methods of construction have rarely been questioned or compared. This paper seeks to lay out the important choices to be made in building linear, multi-factor risk models. Its key insight is simply that stock risk models are not built for stock risk analysis, but for portfolio risk analysis, so that the usefulness of any of the various alternative methods of model construction needs to be evaluated at the portfolio level, not at the single stock level. The paper includes a brief review of the more well-known risk models, showing how they fit into the framework discussed, and concludes by looking ahead to the development of customised, hybrid risk models, designed to match specific investment processes. Keywords: risk models; factor models; covariance matrix
Introduction The purpose of this paper is to review the various methods currently available for building multifactor risk models, and to discuss their respective strengths and weaknesses. The main part of the paper is theoretical, but it concludes with a summary of the more well-known equity risk models commercially available from the main vendors. This summary will review their main features, and show where they fit into the general framework of risk models. Few vendors are completely open about the details of their methodology, however, so these reviews will be based primarily on their marketing literature.
䉷 Henry Stewart Publications 1479-179X (2003)
The first important point to make is that no one builds stock risk models because they care about individual stock risk. If they did, they almost certainly would not use the same model for, say, 20,000 different stocks, but would probably use 20,000 different models, one for each stock. Security analysts, for example, sometimes build simple spreadsheet models for each of the stocks they follow, in order to capture the idiosyncrasies of particular stocks. More typically, they build generic models for companies in different industries, on the grounds that stocks within a particular industry will be subject to much the
Vol. 3, 4, 313-322
Journal of Asset Management
313
MacQueen
same common factor effects as each other, so that one generic model for each industry will be sufficient. Risk models are a different case, however. A single risk model is built for all stocks in a universe, because portfoli
Data Loading...
