Identifying Small Mean Reverting Portfolios

TL;DR

This paper develops methods to identify small, mean-reverting portfolios from multivariate time series, balancing sparsity and predictability, with applications across different markets.

Contribution

It formulates the portfolio selection as a sparse canonical correlation analysis and explores algorithms for sparse generalized eigenvalue problems.

Findings

Effective algorithms for sparse eigenvalue problems

Tradeoff analysis between sparsity and predictability

Impact of market conditions on portfolio mean reversion

Abstract

Given multivariate time series, we study the problem of forming portfolios with maximum mean reversion while constraining the number of assets in these portfolios. We show that it can be formulated as a sparse canonical correlation analysis and study various algorithms to solve the corresponding sparse generalized eigenvalue problems. After discussing penalized parameter estimation procedures, we study the sparsity versus predictability tradeoff and the impact of predictability in various markets.