Factor analysis of security returns aims to decompose a return covariance matrix into systematic and specific risk components. While successful in many respects,traditional approaches like PCA and maximum likelihood suffer from several drawbacks. These include a lack of robustness and strict assumptions on the underlying model of returns.
We apply convex optimization methods to decompose a security return covariance matrix into a low rank, sparse and diagonal components. The low rank component includes the market and other broad factors that affect most securities. The sparse component contains narrow factors such as industry and country, which affect only small subsets of securities. The diagonal component describes the specific risk of each security. We measure the accuracy of a low rank plus sparse decomposition on simulated data, and we illustrate the decomposition on an empirical data set consisting of 400 global equities drawn from 6 countries.
We apply convex optimization methods to decompose a security return covariance matrix into a low rank, sparse and diagonal components. The low rank component includes the market and other broad factors that affect most securities. The sparse component contains narrow factors such as industry and country, which affect only small subsets of securities. The diagonal component describes the specific risk of each security. We measure the accuracy of a low rank plus sparse decomposition on simulated data, and we illustrate the decomposition on an empirical data set consisting of 400 global equities drawn from 6 countries.
- Start date: 2016-06-14 11:00:00
- End date: 2016-06-14 13:00:00
- Venue: 639 Evans Hall at UC Berkeley
- Address: 639 Evans Hall, Berkeley, CA, 94720