Comment on Article by Windle and Carvalho

Abstract

Abstract. The article by Windle and Carvalho introduces a fast update procedurefor covariance matrices through the introduction of higher frequency sources ofinformation for the underlying process, demonstrated with a nancial application.This discussion focuses on outlining the assumptions and constraints around theiruse in nancial applications, as well as an elicitation of some key choices made forcomparison with traditional benchmarks, that may ultimately a ect the results.Keywords: Stochastic Volatility, Financial application, EWMA, Covariance up-date 1 Introduction We congratulate Windle and Carvalho for a very exible and empirically sound contribu-tion, where the authors build a bridge between traditionally used practical approaches,and the more theoretically sound models. Their approach, termed (UE) in the paperand this discussion, is a practical solution for updating the covariance matrix throughan additional source of information, namely higher frequency data (Kyj2008), usuallyavailable for the most liquid assets, while maintaining a tractable exponentially-weightedapproach for the update, and keeping a coherent, scalable statistical model underneath.As such, it builds a link between traditional purely statistical models, relatively popularwithin the sell side arena in nance, and more ad hoc, \quick-and-dirty approaches,namely the Exponentially Weighted Moving Averages (EWMA) updates of covariances,that are still relatively widespread (Shephard2005), mainly within the buy side arena in nance (Fabozzi2008). Since the authors propose this solution from the point of viewof a nancial application, we will focus our discussion on the pros and cons of its usefor nancial applications in general, as well as their particular application. There willbe many other cases where their approach is a natural t, but some aspects in the usefor nancial assets should be stressed, either as potential additions to future research,or as topics to consider prior to using their approach.