Anxiety in Returns

AbstractWe provide empirical evidence that risk-averse investors become anxious about investments in stocks whose realized losses reveal the downside of risk. Contrary to short-term reversal and in support of convex risk aversion, the latter stocks yield significantly lower returns in the subsequent period. Our findings are based on a novel measure of time-varying risk aversion, but can also be observed when using a well-established measure of risk aversion. Moreover, anxiety predicts cross-sectional returns in out-of-sample tests, suggesting that risk-averse investors’ preferences drive empirical risk premia.Keywords: AnomalyAsset Pricing FactorsCross SectionRisk AversionJEL CLASSIFICATION: G12G41G17C31 Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 We do not consider the accounting-implied value of the well-known free cash flow model, as we neither measure expected firm values nor expected cash flows empirically, as, e.g., in Francis et al. (Citation2000). The exact definition of a payback does not change the following considerations and is interchangeable with, for instance, dividends2 We refer to our finding as an anomaly, as the observed investor’s behavior is likely to be irrational, i.e., there is no apparent reason not to invest in assets with a preceding negative return despite having a high level of risk aversion, and as it yields high out-of-sample return predictability that results in remarkably strong long-short returns over multiple decades, which is a commonly accepted characteristic of an anomaly; see, for instance Hou et al. (Citation2015).3 Note that for brevity and based on our assumption of homogeneous investors we omit a subscript for the investor.4 The data can be downloaded from their website: www.openassetpricing.com.5 The results are available from the authors upon request.6 The clustered standard errors and reported t-statistics are obtained by regressing rt+1,i on appropriately chosen specifications of dummy variables indicating low (high) risk aversion and positive (negative) preceding returns as well as interactions thereof.7 Data on MPU and FS is publicly available on https://www.policyuncertainty.com.8 The data and details on the definitions of the respective factors can be obtained from https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html