Working Paper: NBER ID: w4756
Authors: A. Craig Mackinlay
Abstract: A number of studies have presented evidence rejecting the validity of the Capital Asset Pricing Model (CAPM). This evidence has spawned research into possible explanations. These explanations can be divided into two main categories - the risk based alternatives and the nonrisk based alternatives. The risk based category includes multifactor asset pricing models developed under the assumptions of investor rationality and perfect capital markets. The nonrisk based category includes biases introduced in the empirical methodology, the existence of market frictions, or explanations arising from the presence of irrational investors. The distinction between the two categories is important for asset pricing applications such as estimation of the cost of capital. This paper proposes to distinguish between the two categories using ex ante analysis. A framework is developed showing that ex ante one should expect that CAPM deviations due to missing risk factors will be very difficult to statistically detect. In contrast, deviations resulting from nonrisk based sources will be easy to detect. Examination of empirical results leads to the conclusion that the risk based alternatives is not the whole story for the CAPM deviations. The implication of this conclusion is that the adoption of empirically developed multifactor asset pricing models may be premature.
Keywords: CAPM; multifactor models; asset pricing; risk-return tradeoff
JEL Codes: G12; G14
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| deviations attributed to missing risk factors (C29) | detection difficulty (C52) |
| expected return deviations due to missing risk factors (G17) | increased variance (C46) |
| non-risk-based explanations (D91) | observable deviations (C90) |
| irrational investor behavior (G41) | observable deviations (C90) |
| data snooping biases (C52) | observable deviations (C90) |
| market inefficiencies (G14) | non-zero intercepts in CAPM tests (C29) |
| confounding factors (D91) | complicate identification of deviations (L15) |