Drivers of investors’ portfolio choices: Separating risk preferences from frictions

Why don’t more households participate in the stock market? The two reasons often cited in the literature are risk aversion and frictions that prevent participation, such as the cost of maintaining a brokerage account or of paying attention to the markets. This column exploits firms switching their default asset allocation from a money market fund with no stock market exposure to an equity fund to disentangle these two drivers. The findings suggest that frictions are the main reason behind limited stock market participation. Absent such frictions, the authors estimate that 95% of retirement savers would participate in the stock market.

Many households, including those with significant financial wealth, do not participate in the stock market (Mankiw and Zeldes 1991, Campbell 2006, Gomes et al. 2021). This limited stock market participation is difficult to reconcile with standard economic theory, which predicts that all investors should hold at least a small amount of stocks when the equity premium is positive.

Why don’t more households participate in the stock market? The literature has proposed many explanations that fall into two broad categories. The first line of explanations argues that households perceive stocks to be too risky and prefer to hold safer assets. This is, for instance, the case when investors are loss-averse (Gomes 2005), have pessimistic beliefs about stock market returns (Briggs et al. 2021), or do not trust the financial sector (Guiso et al. 2008, Christelis et al. 2024). The second line of explanation argues that investors may prefer stocks over safer assets but still not participate due to frictions (Vissing-Jørgensen 2002). These frictions could include the real costs of creating and maintaining a brokerage account or the cognitive cost of paying attention or making a financial plan.

Although these two categories of explanations predict similar participation behaviour, it is important to distinguish between them. If non-participation is primarily driven by frictions (e.g. hassle costs), nudging savers toward stocks can reduce choice frictions and improve welfare. Conversely, if non-participation primarily reflects investors’ preferences for holding safe assets (e.g. due to loss aversion), nudging them to hold stocks may not be such a good idea.

In new research (Choukhmane and de Silva 2023), we make progress on disentangling risk preferences and frictions as drivers of household stock market participation by leveraging changes in the default asset allocations inside defined-contribution 401(k) plans. Our evidence suggests frictions (rather than risk preferences) are the main reason behind limited stock market participation. Absent such frictions, we estimate that 95% of retirement savers would participate in the stock market.

Separating risk preferences and frictions: The ideal experiment

Imagine that we randomly assign investors who do not participate in the stock market accounts with either stocks or safe assets. If non-participation is due to risk preferences (e.g. loss aversion), those assigned an account with stocks should sell their stocks and rebalance their portfolio toward safe assets. In contrast, if non-participation is driven by one-time or adjustment frictions, those assigned to the stock account should remain invested in stocks, while those given safe assets should progressively rebalance toward equity. Finally, the speed of rebalancing and the extent to which some investors remain passive and always stick with their assigned allocation is informative about the size of adjustment frictions. Without frictions, there should be no passive investors, and with infinitely large frictions, all investors should be passive and stick with their assigned allocation.

Investors defaulted into stocks keep their stocks, while those given bonds rebalance into stocks

To approximate this ideal experiment, we use administrative account-level data from a large US 401(k) provider and study firms that switch their default asset allocation from a money market fund, which has no stock market exposure, to a target date fund. In both cases, workers can opt out of the default option and choose their own allocation. We compare the choices of two groups of investors: a ‘control’ group of investors hired before the change and auto-enrolled into a money market fund; and a ‘treatment’ group of investors hired after the change and auto-enrolled into a target date fund.

Our first empirical finding (see Figure 1) is that more than 95% of workers automatically enrolled into an equity fund (the treatment group) maintain participation in the stock market throughout their tenure. In contrast, those automatically enrolled into a money market fund (the control group) slowly rebalance away from the money market fund and towards stocks, thus revealing their preference for stock market participation. The slow convergences of equity share across the two groups highlight the importance of adjustment frictions: it takes years for the allocations of the control and treatment groups to converge.

Figure 1

Investors’ preferences differ markedly from their choices

Our second result is that investors’ preferences, identified from active decisions, differ substantially from their observed choices. We build on and extend the theoretical framework of Goldin and Reck (2020) to translate our empirical findings into estimates of investors’ preferences. In the left panel of Figure 2, we plot how stock market participation rates vary with age in the raw data. Consistent with prior literature, participation is well below 100% and increases over the life cycle (Cocco et al. 2005, Catherine 2022). In the right panel, we plot how our estimates of investors’ preferences (based on active decisions) vary with age. In contrast to their observed choices, investors’ participation rates absent frictions would be high and flat over the life cycle. When we do the same analysis for equity shares, we find that preferred stock shares are much higher than observed in the raw data and decline with age. These preferences are more in line with the predictions of textbook portfolio choice models (Merton 1969, Gomes 2020), which complements Robertson and Choi (2018) and highlights how frictions can obscure the mapping between choices and frictions.

Figure 2

Figure 3

Investors’ behaviour is consistent with a portfolio choice model with adjustment costs

Our empirical evidence suggests that the overwhelming majority of retirement investors prefer to hold stocks but, unless equity is the default allocation, it can take many years for them to overcome inertia and start participating. How large are the adjustment frictions required to match this observed behaviour? To answer this question, we build a lifecycle portfolio choice model to quantify the preferences and frictions needed to replicate the observed behaviour. A key ingredient in the model is that investors must pay a cost every time they want to adjust their portfolios. Our estimates match the empirical evidence in Figure 1 with a coefficient of relative risk aversion of around 2 and a portfolio adjustment cost of $200. In contrast, a model without adjustment costs cannot simultaneously match the behaviour of investors in the control and treatment groups: by carefully modelling frictions, we can match the behaviour of both with the same preferences.

Conclusion and policy implications

Our research shows that, in the context of retirement accounts, investors’ lack of stock market exposure is primarily explained by frictions associated with investing in the stock market. These frictions can drive a large wedge between investors’ choices and their risk preferences. Once we account for these frictions, investors’ asset allocations are consistent with moderate risk aversion and with the tenets of standard lifecycle portfolio theory. These results have implications for the ongoing policy debate surrounding the design of retirement savings plans (e.g. Leganza and Garcia-Miralles 2024), such as the welfare implications of the rise of target date funds as the most common default asset allocation in 401(k) plans following the 2006 Pension Protection Act (Parker et al. 2022), as well as the broader discussion of ways to increase financial inclusion (Yogo et al. 2023, Derenoncourt et al. 2024).

source : voxeu