Risk Model Construction
This chapter focuses on the recipes for constructing your typical risk model; in particular, reference will be made to those models on FactSet. As usual, we’ll offer a little history and develop a theme from there. We conclude with an introduction to historical methods, including Arch/Garch methodologies.
MULTIFACTOR PRESPECIFIED RISK MODELS
Barra, Inc. was started as a project between Barr Rosenberg and Andrew Rudd in 1975 while at the University of California, Berkeley. The two were specialists in operations research. They began what became an industry standard in risk modeling. Barra used cross-sectional models predicated on factors that were defined from fundamental financial statement data, but deferred to cross-sectional regressions, most likely because of the founders’ backgrounds in operations research (decision science), which relies on this method heavily. Later, with the introduction of the Fama-French equations, support for the usage of financial statements as factor exposures grew, and other risk model vendors entered the market, producing variations of risk models that were still mostly constructed the same way, through regressions in cross-section for a given time period, done sequentially to create a time series of factor returns (betas).1 These models became ...