Distinguishing Risk Models
In this chapter we discuss distinguishing some popular risk models. We begin by tracing the background of the company and model that historically has been the most popular, MSCI-Barra, then offer a general review and analysis of reporting capabilities on FactSet for any risk model. We demonstrate an analysis through an example portfolio risk attribution report. Additionally, we’ll spend some time illustrating that though level calculations can vary somewhat, risk model to risk model, they all tend to capture the trends of risk well, and, importantly, the differences between level estimates of risk models can be used to derive rules of thumb for estimation error estimates.
The history of Barra is pretty interesting and can be found detailed by John Guerard Jr.1 Barra was the first organization to bring quantitative techniques to bear for risk management; hence a review of their beginnings is a great story. In 1973 Barr Rosenberg and Walt McKibben of the Universities of Berkeley and Western Ontario, respectively, published a landmark paper on the prediction of systematic and specific risk in common stocks.2 They went on to show that these predictions can be based on simple firm accounting data. These data were: