Correspondence: Daniel Satchkov, E-mail: DSatchkov@factset.com

¹PhD is a Chief Scientific Adviser at FactSet Research Systems, Inc. Dr Novosyolov's role at FactSet is to verify mathematical foundations of quantitative products and to help create solutions to critical questions in the field of risk measurement. He also teaches risk theory at Siberian Federal University. Founded in 1978, FactSet Research Systems, Inc. combines integrated financial information, analytical applications, and client service to enhance the workflow and productivity of the global investment community.

²CFA is Vice President at FactSet Research Systems, Inc. At FactSet he leads research and development of multi-asset class portfolio risk measurement products. Founded in 1978, FactSet Research Systems, Inc. combines integrated financial information, analytical applications, and client service to enhance the workflow and productivity of the global investment community.

Received 28 January 2008; Revised 28 January 2008.

Abstract

Principal component analysis (PCA) is a technique commonly applied to the interest rate markets to describe yield curve dynamics in a parsimonious manner. Despite an increase in global investing and the growing interconnectedness of the international markets, PCA has not been widely applied to decomposing joint structure of global yield curves. Our objective is to describe the joint structure with a model that can potentially be used for scenario analysis and for estimating the risk of interest rate-sensitive portfolios. In this study, we examine three variations of the PCA technique to decompose global yield curve and interest rate implied volatility structure. We conclude that global yield curve structure can be described with 15–20 factors, whereas implied volatility structure requires at least 20 global factors. The procedure that we identify as preferable is a two-step PCA, with local curves decomposed in the first step and combined local PCs decomposed into a joint structure (PCA of PCs) in the second step. This procedure has a key advantage in that it makes any scenario analysis more meaningful by keeping local PCA factors, which have important economic interpretations as shift, twist and butterfly moves of the yield curve.