Performance Perspectives Blog

Minimum requirements for Karnosky-Singer attribution

by | Dec 7, 2012

Many software vendors and custodians claim to offer the Karnosky-Singer multi-currency attribution model, but do they really?

The name has become synonymous with a robust approach to identifying the contributions from currency movements on a portfolio’s excess return, and comes from the monograph Denis Karnosky and Brian Singer wrote for the CFA Institute. Though a relatively short piece (as monographs are intended to be), it covers a lot of ground. Carl Bacon has done an admirable job of describing the model in his books, which arguably should be considered a companion piece to the model, and present on all performance measurement professionals’ bookshelves.

But when we speak of the K-S model, what do we mean? I discussed this recently with my colleagues John Simpson and Jed Schneider, and then confirmed our conclusions with Brian.

First, we expect to see currency’s attribution effect bifurcated into two components (am I being redundant? how many components could BIfurcation create?):

  • the contribution that results from currency movement on the market; that is, as we take securities from the local to base currency, the change in value that’s attributed to currency fluctuations
  • the contribution that comes from currency management; that is, the investments in currency futures in order to hedge the portfolio, or to expose the portfolio to other currency weights that differ from the portfolio’s investments.

Second, that interest rates be broken away from the market side of attribution and included with the currency side, because, as Denis and Brian point out, interest rates have a direct bearing on currency movement.

We have observed some firms who claim to offer the K-S model, but who “lump” or combine the effects into a single value; i.e., the currency effect is shown as one, not two numbers. This fails to provide the highlights that the K-S model offers from an analytic perspective, robbing the recipient of knowing (a) how much was contributed from the currency movement across time on the underlying assets and (b) how much came from currency management (i.e., the hedging decisions that occurred). To lump them together converts the K-S model into a naive model, and should deprive the firm from claiming that they do, in fact, offer Karnosky-Singer attribution.

As always, your thoughts and insights are welcome.