A New Test of Risk Factor Relevance

Abstract

Textbook models assume that investors try to insure against bad states of the world associated with specific risk factors when investing. This is a testable assumption and we develop a survey framework for doing so. Our framework can be applied to any risk factor. We demonstrate the approach using consumption growth, which makes our results applicable to most modern asset-pricing models. Participants respond to changes in the mean and volatility of stock returns consistent with textbook models, but we find no evidence that they view an asset's correlation with consumption growth as relevant to investment decisions.THE TEXTBOOK EXPLANATION FOR WHY some assets have higher expected returns than others involves "extending the principles behind fire and casualty insurance to investment portfolios (Cochrane (1999))." While investors in

Millionaires speak: What drives their personal investment decisions?

“…Fourth, non-standard preferences may be relatively less important for explaining portfolio risky shares and the equity premium than other factors. Fifth, our results may be consistent with the view, forcefully argued by Chinco, Hartzmark, and Sussman (2021), that positive asset pricing and portfolio choice theories centered on period-by-period covariance of returns with consumption growth may be a paradigm we should abandon.…”

“…The latter explanation becomes less likely if most people believe the asset to have low risk or a low expected return. Chinco, Hartzmark, and Sussman (2021) argue that adjusting a return for a particular risk exposure does not make sense if investors do not report trying to avoid that risk.…”

“…Respondents can in principle infer causality without observing exogenous variation in the candidate mechanism because they have observed the process by which outcomes arose. Related approaches are taken by Graham and Harvey (2000), who ask CFOs how often they use particular techniques and the importance of various considerations when making financial decisions, and Chinco, Hartzmark, and Sussman (2021), who ask respondents what factors, information, and goals they considered in making an experimental investment decision. If one's equity portfolio share !…”

“…Despite its distinguished pedigree in finance research (Lintner, 1956), this approach had largely fallen out of favor in the field. Recent examples of papers returning to survey methods include Greenwood and Shleifer (2014), Kuhnen and Miu (2017), Kuchler and Zafar (2019), Choi and Robertson (2020), Das, Kuhnen, and Nagel (2020), Giglio et al (2020Giglio et al ( , 2021, Chinco, Hartzmark, and Sussman (2021), and Liu et al (2021). 2 Although surveys are a useful tool for gaining insights about individual investors, they are often less informative about the determinants of prices and aggregate quantities because very wealthy investors, who possess a disproportionate share of the economy's assets, are usually a tiny fraction of survey samples.…”

“…sample a total of 2,484 U.S. respondents, each of whom has at least $1 million of investable assets, 18% of whom have at least $5 million, and 4% of whom have at least $10 million. Other recent papers that have surveyed wealthy investors include Giglio et al (2020Giglio et al ( , 2021, who measure stock return and GDP growth beliefs in a sample of Vanguard clients with an average account balance of about $500,000 and relate these beliefs to their trading behavior; and Chinco, Hartzmark, and Sussman (2021), who measure how investors react to return covariance with economic growth in a sample that includes finance professionals and top clients of a large asset manager.…”