What Drives Investors' Portfolio Choices? Separating Risk Preferences from Frictions

Abstract

We separately identify the role of risk preferences and frictions in portfolio choice. Individuals may choose not to participate in the stock market because of non-standard preferences (e.g. loss aversion) or frictions impacting their choices (e.g. participation costs). We overcome this identification problem by using variation in the default asset allocation of 401(k) plans and estimate that, absent frictions, 94% of investors would prefer holding stocks in their retirement account with an equity share of retirement wealth that declines over the life cycle, which differs markedly from their observed behavior. We use this variation to estimate a structural life cycle portfolio choice model with Epstein-Zin preferences and find the evidence consistent with a relative risk aversion of around 2.1. This estimate is significantly lower than most estimates in the life cycle portfolio choice literature and highlights how choice frictions can hamper the identification of risk preferences. *MIT Sloan. Corresponding emails: tahac@mit.edu; tdesilva@mit.edu. First draft: November 18th, 2021. We thank Hunt Allcott, Joesph Briggs (discussant), Sylvain Catherine (discussant), Nuno Clara, Cary Frydman (discussant), Francisco Gomes (discussant), Debbie Lucas, Christopher Palmer, Jonathan Parker, Antoinette Schoar, Frank Schilbach, Larry Schmidt, Eric So, David Sraer, David Thesmar, Adrien Verdelhan, Toni Whited, and audience members at MIT Economics, MIT Sloan, Duke Fuqua, University of Illinois at Urbana-Champaign, Chicago Fed, Princeton, 2022 NBER Behavioral Finance Spring Meeting, 2022 CEPR Conference on Household Finance, 2022 Western Finance Association Annual Meeting, 2022 Society for Economic Dynamics Annual Meeting, and the 2022 Texas Finance Festival for their insightful comments. We also thank the data provider, a large U.S. financial institution, for making available the data used in this paper, helpful discussions, and technical support. Many households do not participate in the stock market, including households with significant financial wealth (Mankiw and Zeldes 1991; Guiso, Haliassos, and Jappelli 2002; Campbell 2006). This limited participation in the stock market is difficult to reconcile with standard economic theory, which predicts that all investors should hold at least a small amount of stocks in the presence of a positive equity premium (e.g. Merton 1969; Campbell and Viceira 2001).1 In principle, an investor may choose to not allocate their financial wealth to the stock market for two reasons. First, this investor may prefer holding safe assets over stocks (e.g. due to loss aversion, ambiguity aversion, or pessimistic beliefs about returns). Alternatively, this investor may prefer holding stocks over safe assets, but not participate due to frictions. These frictions could include the real costs of setting up and maintaining a brokerage account or the cognitive cost of making a financial plan and paying attention. Although these two explanations have similar predictions for this investor’s participation in the stock market, distinguishing between them has important normative implications. For example, interventions designed to encourage more stock market participation may be more desirable if non-participation is due to high participation costs rather than a preference for safe assets. In this paper, we propose and implement a new empirical approach to recover investor preferences in the presence of frictions. We begin by showing that the life cycle profile of participation in the Survey of Consumer Finances is consistent with very different calibrations of a standard life cycle portfolio choice model: (i) a risk aversion below 2 and an extremely high adjustment or participation cost or (ii) a risk-aversion above 30 and a lower adjustment or participation cost. This illustrates the challenge in separately identifying investors’ risk preferences, as well as the size and specification of choice frictions. Next, we overcome this identification problem using quasi-experimental variation in the default asset allocation of 401(k) plans. An ideal experiment for separating between preferenceand friction-based explanations for non-participation would be to randomly give investors who are not participating in the stock market an investment account with stocks, which would effectively remove any one-time adjustment costs associated with participation. If these investors dislike holding risky assets (for instance, due to loss aversion), or if they face large per-period participation costs, they should sell the stocks and move their holdings toward safer assets. Alternatively, if onetime frictions were responsible for these investors not participating beforehand, we would expect them to keep the stocks, durably switching from stock-market non-participation to participation as a consequence of the treatment. To approximate this ideal experiment, we rely on account-level data from a large 401(k) plan 1With strictly increasing and differentiable utility, agents should be risk-neutral over small risks (Rabin 2000).