Diversifying Trends

Abstract

This paper provides a new method to disentangle the systematic component from the idiosyncratic part of the risk associated with trend following strategies. A simple statistical approach, combined with standard dimension reduction techniques, enables us to extract the common trending part in any asset price. We apply this methodology on a large set of futures, covering all the major asset classes, and extract a common risk factor, called CoTrend. We show that common trends are higher for some cross-asset class pairs than from intra-asset class ones, such as JPY/USD and Gold. This result helps to create sectors in a portfolio diversification context, especially for trend following strategies. In addition, the CoTrend factor helps to understand arbitrage-based hedge fund strategies, which by essence are decorrelated with the standard risk factors.

Keywords: Time series momentum, Portfolio construction, Factor analysis

JEL Classification: G11, G12, G15, F37