Parsimony - A Model Risk Paper
Models are representations of certain aspects of reality and as such are not reality itself. This discrepancy is model risk; the risk that behavior predicted by a model is divergent from actual behavior.
A parsimonious model choice embodies an economy of conceptualization. We choose a parsimonious model in terms of only that which is needed to understand our problem and robustly extrapolate. By doing so, we better understand and mitigate model risk.
Selecting a germane model which informs us in the context of decision-making under uncertainty calls for a model that faithfully describes historic behavior and is robustly predictive. Moreover, the model needs to be neither too simplistic nor overly elaborate but be just right!
In this paper, we systematically explore how to understand and mitigate model risk. Guidance is given on finding parsimonious models within a spectrum that neither under fit nor overfit data in order to rationally and consistently make informed business decisions.