گزینه های سرمایه گذاری حساب شخصی و انتخاب نمونه کارها: درس های رفتاری از برنامه های 401 (k)

This paper examines how the menu of investment options made available to workers in defined contribution plans influences portfolio choice. Using unique panel data of 401(k) plans in the U.S., we present three principle findings. First, we show that the share of investment options in a particular asset class (i.e., company stock, equities, fixed income, and balanced funds) has a significant effect on aggregate participant portfolio allocations across these asset classes. Second, we document that the vast majority of the new funds added to 401(k) plans are high-cost actively-managed equity funds, as opposed to lower-cost equity index funds. Third, because the average share of assets invested in low-cost equity index funds declines with an increase in the number of options, average portfolio expenses increase and average portfolio performance is thus depressed. All of these findings are obtained from a panel data set, enabling us to control for heterogeneity in the investment preferences of workers across firms and across time.

Over two dozen countries around the world now include individual investment accounts as part of their public pension systems. Other countries, notably including the U.S., are also considering reforms that would allow individuals to directly invest part of their Social Security contributions in individual accounts. A key issue in designing such a program is determining what investment choices to offer individual participants. The simplest portfolio theory suggests that it is sufficient to provide a choice consisting of one portfolio of risky assets – the market portfolio – and one risk-free asset, and then to allow individuals to mix these two portfolios in accordance with their individual risk preferences. Most public and private plans in the U.S. and abroad, however, provide a large number and broad range of choices. For example, in the U.S., the vast majority of private sector defined contribution pension plans offer multiple investment options, often allowing individuals to choose from among several equity, bond, market and balanced fund options. Individuals also have thousands of mutual funds to choose from when allocating their non-pension portfolios. In Sweden, the Social Security system provides participants a menu of investment options that includes over 650 funds from which to choose. The central question of this paper is whether the mix of investment options available to participants in an individual accounts program matters for portfolio allocation. In particular, we are interested in the “behavioral” response to the selection of fund options, over and above any “mechanical” link by which we mean changes that flow directly from adding or relaxing a binding constraint. For example, suppose an individual is prohibited from owning a particular asset class. It is clearly the case that this constraint will alter their portfolio choice if, in the absence of the constraint, the individual would have invested in this asset class. Instead, our focus is on the “behavioral” response, which might occur when a change in the menu of investment options leads to a large change in asset allocation, even though the investment opportunity set has not significantly changed. For example, imagine that an investor, faced with a choice between a diversified stock fund and a diversified bond fund, chose to allocate 50% of her portfolio to each fund. If this individual were provided a second diversified stock fund as a third investment alternative, then the overall investment opportunity set of this individual has not substantially changed because the additional stock fund is largely redundant of the first. In this case, standard portfolio theory suggests that this individual's optimal allocation would still be close to 50% bonds and 50% stocks. A growing body of evidence suggests, however, that many individual portfolio decisions may be influenced by plan design. For example, Benartzi and Thaler (2001) suggest that many people follow a “naïve” diversification strategy, such as evenly dividing contributions across all available assets (e.g., a “1 / n” strategy). If this is the case, then simply changing the relative number of stock and bond funds may alter the allocation of an investor's portfolio between stocks and bonds. If people behave in this way, then firm managers or policymakers who are charged with determining the set of investment options to make available to participants in a corporate or government individual accounts program should consider how the choice of fund options will influence individual portfolio allocations. The private pension system in the U.S. – in particular, 401(k) plans, which are now the single most common private pension in the U.S. – provides a useful research laboratory with which to learn about these issues. Because plan sponsors have significant leeway in choosing which investment options to make available, there is considerable time series and cross-sectional variation in the set of investment opportunities facing 401(k) participants. Using a rich panel data set on fund options and fund contributions to 401(k) plans during the 1990s, we examine several issues in this paper related to how the structure of investment options within a plan affects participant diversification. Unlike past work in this area that focuses on cross-sectional analyses, we are able to exploit the panel nature of our data to control for differences in the investment preferences of workers across firms and over time to better identify the effect of a change in 401(k) plan characteristics on participant behavior. We focus primarily on how the set of investment alternatives affects portfolio diversification. While some prior research has suggested that overall portfolio diversification is affected by the menu of options (Benartzi and Thaler, 2001), Huberman and Jiang (2006) highlight some of the difficulties in making inferences about individual behavior based on plan level data. For example, even if there is evidence at the plan level that contributions are divided roughly evenly across plan options, this could be because individuals are each following a 1 / n strategy, 1 / n of the individuals are each concentrating their portfolio in one option, or other intermediate combinations. Nonetheless, in considering the set of investment options to make available to individuals through a large national pension system, the effect of the plan choice set on aggregate portfolio allocations is of substantial interest. Our data allow us to do a more comprehensive examination of plan level portfolio responses both within and across multiple asset classes over many years. We find evidence that the allocation of contributions is significantly affected by the number and mix of investment options available, even after accounting for plan-level fixed effects. For example, our estimates suggest that by increasing the share of equity funds from 1 / 3 to 1 / 2 (such as by adding an additional equity fund option to a plan that already offers company stock, one equity fund, and one bond fund), overall participant allocations to equity funds increase by 7.5 percentage points. We find similar strong effects for allocations to bond funds and balanced funds. Given that the fraction of equity options in the plan tends to increase substantially and the fraction of bond options tends to decrease substantially as plans offer more options over time, this leads to 401(k) participants taking on greater risk in their retirement portfolios with their higher equity allocations. We also address how the composition of actively versus passively managed fund options has affected allocation decisions within 401(k) plans. From 1990 to 2003, the number of mutual funds available to investors in the U.S. rose from 3100 to 8100, with the dramatic growth in actively-managed equity funds (which quadrupled over the period) accounting for nearly 70% of the total increase (Investment Company Institute, 2005). Consistent with this, we find the vast majority of the new funds added to 401(k) plans are also high-cost actively-managed equity funds as opposed to lower-cost equity index funds. For example, actively-managed equity funds represent nearly 2 / 3 of the new funds added to 401(k) plans in our sample over the period 1998–2002, while index funds represent only 8 percent of the new funds added. This result has strong implications for the relation between the investment performance of these retirement plans and the composition and number of fund options provided. For example, we estimate that a move from a plan with only index funds to a plan with only actively-managed funds is associated with a 35 basis point increase in the annual plan-wide expense ratio paid by participants (as participants are investing more in higher-cost funds which are much more likely to be added on the margin). Given historically that actively-managed funds do not outperform index funds (even before expenses), these investment patterns (coupled with actively-managed funds being much more likely to be the marginal fund added to the plan) could result in a nontrivial reduction in resources available for retirement. While not the main focus of our paper, we also address the importance of firm-level heterogeneity and investment restrictions upon the investment behavior of 401(k) participants. A concern with any of the past cross-sectional analyses that have documented a strong relation between plan parameters and participant decisions is that an omitted factor such as differences in either the firms or the workers across firms could explain both the design of the 401(k) plan as well as the workers' investment decisions (e.g., differences in the risk aversion or investment philosophy of workers across firms). Ruling out such an alternative explanation is essential for identifying what role the characteristics of the plan (as opposed to unobserved characteristics of the participants of the plan) play in influencing investment decisions and is thus essential for public policy purposes. We present evidence that both plan parameters and firm-level heterogeneity are important determinants of how 401(k) assets are allocated across asset classes. For example, the number of equity and bond options in the 401(k) plan is an important predictor of contributions to equity and bond funds, respectively, even after accounting for firm-level fixed effects. On the other hand, while both the relation between the number of options in the plan and allocations to company stock and the “endorsement” effect found by Benartzi (2001)1 are very strong in the cross section, they are substantially diminished once firm-specific effects are accounted for in the regression. This suggests that investments in company stock in pension plans appear to be driven in no small part by firm-level heterogeneity in the desirability of company stock investment, which manifests itself both in employee contributions to company stock, how many non-company stock options are present in the 401(k) plan, and employer match policy. This paper proceeds as follows. Section 2 provides background information on the range of investment options offered in selected pension systems around the world, as well as in some U.S. reform proposals. We also review the relevant literature relating to the influence of investment options on portfolio allocation. In Section 3, we describe our data on 401(k) plans and provide some initial summary statistics about the range of investment options available. In Section 4, we provide evidence on how the number and mix of investment options influences the allocation of contributions in 401(k) plans. Section 5 demonstrates how trends in the mutual fund industry over the past decade such as the explosion of high-cost, actively-managed equity funds have affected 401(k) plans and participants. Section 6 concludes with a discussion of policy implications.

A key policy issue in the design and implementation of any defined contribution pension system, whether publicly or privately provided, is how to structure the investment options that will be made available to participants. While it is well understood that the structure of investment options can have an important effect on the level of administrative expenses of an individual accounts program, we show that the number and mix of investment options can also have a first order effect on the risk and return profile of average portfolios. Even after controlling for individual firm and year fixed effects, we find a strong positive relationship between the share of investment options provided to employees in a particular asset class and the share of their portfolios invested in that asset class. This strongly suggests that average participants are not optimally allocating their portfolios according to standard finance theory predictions, but instead are following naïve strategies that subject them to “manipulation” by non-binding changes in the number and mix of investment options. A key policy implication is that the number and mix of investment options will have an important effect on overall asset allocation in the individual accounts. In short, it appears to be possible to influence the portfolio allocation of individual participants by altering the mix of equity and bond funds, even if the overall investment opportunity set remains unchanged. By exploiting panel data on 401(k) plans, we find that both plan parameters and firm-level heterogeneity are important in explaining investment decisions. Investments in company stock, in particular, seem to be driven more by differences in workers across firms (or unobserved differences in the firms themselves) than by cross-sectional differences in pension plan characteristics (such as whether the match is in company stock or the number of non-company stock options in the retirement plan). This highlights the importance of exploiting panel data in identifying behavioral responses to pension plan characteristics. Finally, the proliferation of actively-managed equity funds in general, which has trickled down to offerings in 401(k) plans as well, has resulted in the vast majority of money in retirement plans flowing into higher-cost actively-managed equity funds as opposed to lower-cost equity index funds (such as S&P 500 funds). This results in plan participants paying higher fees, which will depress overall portfolio performance, and thus, in the long-term, depress retirement wealth.