Misspecification-Robust Inference in Linear Asset Pricing Models with Irrelevant Risk Factors

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in www.econstor.euThe authors thank the editor (Andrew Karolyi), two anonymous referees, Anton Braun, Chu Zhang, Guofu Zhou, and seminar participants at the Federal Reserve Bank of Atlanta, Imperial College London, Queen's University, Singapore Management University, and the University of Georgia. They also thank participants at the third Annual CIRPÉE Applied Abstract: We show that in misspecified models with useless factors (for example, factors that are independent of the returns on the test assets), the standard inference procedures tend to erroneously conclude, with high probability, that these irrelevant factors are priced and the restrictions of the model hold. Our proposed model selection procedure, which is robust to useless factors and potential model misspecification, restores the standard inference and proves to be effective in eliminating factors that do not improve the model's pricing ability. The practical relevance of our analysis is illustrated using simulations and empirical applications.JEL classification: G12, C12, C52

On the properties of the constrained Hansen–Jagannathan distance

“…To understand why IGM and ICAPM are not rejected by the test based on the unconstrained HJ-distance, it is important to realize that these models are not properly identified since they contain unspanned factors that exhibit very low correlations with the returns on the test assets (for example, using the reduced rank test of Kleibergen and Paap (2006), we cannot reject the null hypothesis of lack of identification for these models at any conventional level). For such models, Gospodinov, Kan, and Robotti (2014) show that the unconstrained HJ-distance test has low power in detecting misspecification, and that this test is inconsistent under the alternative of model misspecification in the extreme case in which one or more factors are useless. In contrast, the test based on the constrained HJ-distance seems to have higher power in rejecting misspecified and unidentified asset-pricing models.…”

“…This feature, however, only depends on the properties of the test asset returns and not on the particular choice of a model. Furthermore, the population constrained HJ-distance and its parameters appear to be well defined in the presence of factors that are uncorrelated with the returns on the test assets, which is not the case for the unconstrained HJ-distance (see, for example, Gospodinov, Kan, and Robotti, 2014). On the downside, the constrained HJ-distance lacks a clear maximum pricing error interpretation and using it to compare and rank competing asset-pricing models can be problematic.…”

Asymptotic variance approximations for invariant estimators in uncertain asset-pricing models

“…Peñaranda and Sentana (2015) show that CU-GMM delivers numerically identical estimates in the beta-pricing and SDF setups. 8 By augmenting e( ) in the SDF representation with additional (just-identi…ed) moment conditions for f , V f ; and , the CU-GMM estimate of the augmented parameter vector = …”

Identification-Robust Factor Pricing: Canadian Evidence

“…In multivariate-regression financial models, finitesample testing is important because tests which are only approximate and/or do not account for non-normality can lead to unreliable empirical interpretations of standard financial models; see Shanken (1996), Campbell et al (1997), Beaulieu (2003, 2010), and Beaulieu, Dufour and Khalaf (2007, 2010a. In parallel, an emerging literature, which builds on Zhang (1999a, 1999b) recognizes the adverse effects of large numbers of factors; see Kleibergen (2009), Kan, Robotti and Shanken (2013), Kleibergen and Zhan (2013), Gospodinov, Kan and Robotti (2014), and Harvey, Liu and Zhu (2015). In this paper, we develop inference methods immune to both dimensionality and identification difficulties.…”

“…In contrast, Kan et al (2013), Kleibergen and Zhan (2013), and Gospodinov et al (2014) focus on model misspecification. assets adding momentum to the list of factors considered by L' Her et al (2004).…”