Factor Forestry: Pruning using an Economic Chainsaw

Abstract

Hundreds of anomalies, factors, or characteristic portfolios have been discovered whose risk-return spread is not explained by benchmark empirical factor models. Each of these is a potential candidate as a factor in such models. Efforts to narrow down this plethora of candidates to a parsimonious set has mainly relied on econometric and statistical tools. Standard textbooks, however, emphasize that factors in models proxy for the marginal utility of aggregate consumption growth. Imposing the economic restrictions that this link implies, we find that only 10-20 per cent of this forest of factors impound news about future consumption growth, the low-frequency risk of consumption growth or its extreme tails i.e. bad times. Interestingly, new factors used in models that perform better than the benchmark models do better in passing economically motivated hurdles. Our results imply that imposing theoretical restrictions, on this forest of potential factors can impose discipline, in the search for an ex-post MV efficient portfolios. In addition, applying an economic chainsaw to yet to be discovered anomalies could curb the growth of this forest.