Working Paper: NBER ID: w26418
Authors: Hui Chen; Winston Wei Dou; Leonid Kogan
Abstract: We formalize the concept of “dark matter” in asset pricing models by quantifying the additional informativeness of cross-equation restrictions about fundamental dynamics. The dark matter measure captures the degree of fragility for models that are potentially misspecified and unstable: a large dark matter measure signifies that the model lacks internal refutability (weak power of optimal specification tests) and external validity (high overfitting tendency and poor out-of-sample fit). The measure can be computed at low cost even for complex dynamic structural models. To illustrate its applications, we provide quantitative examples applying the measure to (time-varying) rare-disaster risk and long-run risk models.
Keywords: dark matter; asset pricing; GMM; model fragility
JEL Codes: C52; D81; E32; G12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| larger dark matter measures (C55) | higher expected overfitting (C52) |
| larger dark matter measures (C55) | lower statistical test power (C12) |
| choice of estimation technique (C51) | degree of overfitting (C52) |