Electronic copy of this paper is available at: http://ssrn.com/abstract=958585
Heuristics and Biases in Retirement Savings Behavior
Shlomo Benartzi and Richard Thaler
Shlomo Benartzi is Associate Professor and Co-Chair of the Behavioral Decision-Making
Group, Anderson School of Management, University of California at Los Angeles, Los Angeles,
California. Richard H. Thaler is Robert P. Gwinn Professor of Behavioral Science and
Economics, and Director of the Center for Decision Research, Graduate School of Business,
University of Chicago, Chicago, Illinois. Their e-mail addresses are <email@example.com> and
Electronic copy of this paper is available at: http://ssrn.com/abstract=958585
All around the world, in both the public and private sectors, retirement plans are shifting
away from “defined benefit” plans and toward “defined contribution” plans.
Poterba et al.
(2006), for example, followed the cohort of Americans who were 45 years old in 1984 and report
a decrease in defined benefit plan coverage from about 40 percent to 20 percent and a
corresponding increase in defined contribution plan coverage from about 5 percent to more than
30 percent. Defined contribution plans have many attractive features for participants, such as
portability and flexibility, but these attractions come with an increased responsibility to choose
wisely. The plans also provide economists with an attractive domain in which to study savings
The standard economic theories of saving (like the life-cycle or permanent income
models) contain three embedded rationality assumptions, one explicit and two implicit. The
explicit assumption is that savers accumulate and then decumulate assets to maximize some
lifetime utility function (possibly including bequests). The first implicit assumption is that
households have the cognitive ability to solve the necessary optimization problem. The second
implicit assumption is that the households also have sufficient willpower to execute this optimal
Both of the implicit assumptions are suspect. Even among economists, few spend much
time calculating a personal optimal savings rate, given the uncertainties about future rates of
return, income flows, retirement plans, health, and so forth. Instead, most people cope by
1 A defined benefit plan promises a benefit determined by a formula that typically includes a salary history and
length of employment. A defined contribution plan specifies how much goes into a worker’s retirement account, but
then transfers much of the decision-making authority about whether to participate, how much to save, and how to
invest from the employer or government to the employee.
adopting simple heuristics, or rules of thumb. However, psychology teaches that such heuristics,
though often useful and accurate, can lead to systematic biases (Griffen, Gilovich, and
Kahneman, 2002). In this paper, we investigate both the heuristics and the biases that emerge in
the area of retirement savings. We do not discuss how to determine whether people are saving
enough for retirement; that topic is covered in a companion paper by Jonathan Skinner in this
issue. Instead, we examine the decisions employees make about whether to join a savings plan,
how much to contribute, and how to invest. We then discuss the possible role of interventions, in
terms of either education or plan design.
Enrollment Decisions: To Join or Not to Join
Defined contribution retirement plans are attractive vehicles for saving. Contributions are
tax deductible, and accumulations are tax deferred. In addition, many employers offer to match
employees' contributions. For example, a common plan is to match 50 percent of employees'
contributions up to some threshold, such as 6 percent of salary. Taking advantage of this match
should be a no-brainer for all but the most impatient and/or liquidity-constrained household.
Nevertheless, enrollment rates in such plans are far from 100 percent.
One extreme example of reluctance to join an attractive retirement plan comes from the
United Kingdom, where some defined benefit plans do not require any employee contributions
and are fully paid for by the employer. They do require employees to take action to join the plan.
Data on 25 such plans reveals that only half of the eligible employees (51 percent) signed up.
2 We thank David Blake and the U.K. Department of Work and Pensions for providing us with the data.
Another extreme example involves those workers for whom joining a retirement plan
amounts to an arbitrage opportunity. Choi et al. (2004b) identify one group of workers with this
arbitrage opportunity, namely employees who are 1) older than 59
years old, so they face no
tax penalty when they withdraw funds from their retirement account; 2) have an employer match;
and 3) are allowed by their employer to withdraw funds from their retirement account while still
working. For this group of employees, joining the plan is a sure profit opportunity because they
can immediately withdraw their contributions without any penalty, yet they get to keep the
employer match. Nonetheless, Choi et al. find that 40 percent of these individuals either do not
join the plan or do not save enough to get the full match.
One method to encourage worker participation in retirement plans is to change the
default, so that instead of workers being outside the retirement plan unless they choose to opt in,
they would be enrolled in the plan unless they choose to opt out. This strategy, called automatic
enrollment (or negative election), has proven to increase enrollment in U.S. defined contribution
plans (Madrian and Shea, 2001a; Choi et al., 2004a, 2002). Under automatic enrollment,
workers are notified at the time of eligibility that they will be enrolled in the plan (at a specified
savings rate and asset allocation) unless they actively elect not to participate or they change the
default selections. In one plan Madrian and Shea studied, participation rates under the opt-in
approach were barely 20 percent after three months of employment, gradually increasing to 65
percent after 36 months of employment. When automatic enrollment was adopted, enrollment of
new employees jumped to 90 percent immediately and increased to more than 98 percent within
36 months. Automatic enrollment thus has two effects: participants join sooner, and more
participants join eventually.
3 Duflo et al. (2005) find a similar unexploited arbitrage opportunity in the context of tax filers eligible for the
savers tax credit.
Does automatic enrollment merely overcome the inertia to help workers make the choice
they would actually prefer? Or does automatic enrollment somehow seduce workers into saving
when they would prefer to be spending? Under automatic enrollment, very few employees drop
out of the plan once enrolled. For example, in the four companies Choi et al. (forthcoming)
studied that adopted automatic enrollment the fraction of 401(k) participants who dropped out of
the plan in the first year was only 0.3 to 0.6 percentage points higher than it had been before
automatic enrollment was introduced. This finding suggests that workers are not suddenly
discovering, to their dismay, that they are saving more than they had wanted.
Closely related to automatic enrollment is the idea to require that workers make an active
decision whether to join the plan (Choi et al., 2005), such as requiring employees to check a
“yes” or a “no” box for participation. With active decision making in place, employees have to
state their preferences and there is no default choice. One company switched from an opt-in
regime to active decisions and found that participation rates increased by about 25 percentage
Another related idea is to simplify the enrollment process. Choi et al. (2005) tested this
idea by analyzing a simplified enrollment form. New employees were handed enrollment cards
during orientation with a “yes” box for joining the plan at a 2-percent saving rate with a pre-
selected asset allocation. Employees did not have to spend time choosing a saving rate and asset
allocation but could just check the “yes” box for participation. Choi et al. report an increase in
participation rates during the first four months of employment from 9 percent to 34 percent.
However, both automatic enrollment programs and active decision plans are typically set
with a relatively low default saving rate of 2 or 3 percent and a very conservative investment
choice, such as a money market account. Madrian and Shea (2001a) found that many employees
continue saving at the default rate of 2 percent, a rate far too low to provide sufficient funding
for retirement, and also that many employees remain in the default investment fund. We will
later discuss policies that might keep the high participation rates of automatic enrollment plans
and also promote higher contribution rates and more broadly diversified portfolios.
While automatic enrollment or “quick” enrollment makes the process of joining a
retirement plan less daunting, expanding the choices of funds available to participants can have
the opposite effect. Iyengar, Huberman and Jiang (2004) find a negative correlation between the
number of investment options offered in the plan and participation rates. They estimate the
addition of 10 funds to the menu of investment options reduces the likelihood of employee
participation by two percentage points.
In a typical opt-in retirement savings plan, employees are first asked whether they wish to
participate, and then asked how much they want to contribute. In this paper, we do not attempt
to answer the difficult question of whether the average employee is contributing “enough.”
do want to make two general points, however. First, for workers who do not have other
significant sources of retirement income, the savings rates typically observed in 401(k) plans are
unlikely to provide anything close to complete income replacement in retirement. Second, many
employees believe that they should be saving more. Choi et al. (2002) report that 68 percent of
4 While saving rates are very low and often negative in the U.S., U.K., Canada and Australia, saving rates are much
higher in Asia. It is beyond the scope of this paper to address any cultural differences that could explain the cross
sectional variation in saving rates.
401(k) participants feel their saving rate is “too low,” 31 percent feel their saving rate is “about
right,” and only 1 percent believe their saving rate is “too high.”
How do participants choose their contribution rate? Many people spend very little time
on this important financial decision. In a survey of faculty and staff at the University of
Southern California, we found that 58 percent spent less than one hour determining their
contribution rate and investment elections (Benartzi and Thaler, 1999).
Apparently, many people are using shortcuts or “saving heuristics.” For example, in
many plans, participants are asked to state a desired saving rate as a percentage of pay. Hewitt
(2002a) finds that the distribution of contribution rates spikes at multiples of 5 percent, even
though this analysis excludes plans that offer an employer match with a threshold of either 5 or
10 percent, thus ruling out the possibility that employees simply maximize the amount
contributed by their employer on their behalf (a strategy we discuss below).
Another saving heuristic we explored in a joint research project with Hewitt Associates is
picking the maximum contribution rate allowed by the plan (Hewitt, 2002b). This strategy can
be a sensible one. However, changes in the tax code enabled us to explore whether the rule of
“saving the max” is the result of careful thinking or just a convenient rule of thumb. The
Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) retained dollar caps on
contributions to retirement accounts but eliminated the restrictions on the percentage of salary
that could be contributed. As a result, people with a low salary (for example, a part-time worker)
below the dollar cap could choose to save 100 percent of their pay. This strategy could be
5 Economists sometimes belittle such statements of intention, and partly for good reason, given that few of the
participants who say they should be saving more make any changes in their behavior in. Yet such statements are not
meaningless or random. Many people announce an intention to eat less and exercise more next year, but very few
say they hope to smoke more next year or watch more sitcom reruns. We interpret the statement “I should be saving
(exercising) more” to imply that people making such a statement would be open to strategies that would help them
achieve these goals. We discuss one such plan below.
attractive in certain cases: for example, a couple that wants to save all or most of the second
earner’s pay. Traditional economic analysis predicts that EGTRRA would likely result in
increased contributions to retirement accounts. But if some employees used the maximum
contribution percentage prior to EGTRRA as a saving heuristic, and those employees now need
to choose their own percentage (because saving the maximum 100 percent of wages is not
feasible for most people), then raising the maximum share of income that can be contributed
could result in lower contributions to retirement accounts.
To study this possibility, we compared the distribution of contribution rates for those
joining one plan at the fourth quarter of 2001, when the maximum was 16 percent, with those
joining the plan at the first quarter of 2002, just after the limit was raised to 100 percent. Figure
1 shows the results. Prior to EGTRRA, 21 percent of new hires deferred 16 percent of their
income. After EGTRRA, five percent deferred 16 percent and seven percent deferred more than
16 percent. Thus, the percentage of employees saving at least 16 percent decreased from 21 to
12. We believe that some employees who would have been attracted to the “maximum
contribution heuristic” prior to EGTRRA found the 100 percent maximum rate too high and
switched to the “multiple-of-five heuristic,” which explains the increased popularity of
contribution rates of 10 and 15 percent.
Another common rule of thumb is to contribute to a retirement account the minimum
necessary to get the full employer match. For example, if the employer matches employees'
contributions up to 6 percent of pay, then many employees contribute 6 percent. The employer
in Figure 1 matched up to 6 percent, and 28 percent of the participants contributed at that level.
If participants are behaving this way, then firms desiring to encourage employee savings might
alter their matching formula to achieve this goal. For example, we suspect that changing the
match formula from 50 percent on the first 6 percent of pay to 30 percent on the first 10 percent
of pay would result in higher contribution rates. Those who use the match threshold as a rule of
thumb would save more with a higher matching threshold. Also, picking a round number as the
threshold would also capture those who use the “round number heuristic” we discussed above.
Naïve diversification strategies
Having decided to join the plan, and having picked an amount to save, participants must
then decide how to invest their contributions. When asked about how he allocated his retirement
investments in his TIAA-CREF account, Nobel laureate Harry Markowitz, one of the founders of
modern portfolio theory, confessed: “I should have computed the historic covariances of the
asset classes and drawn an efficient frontier. Instead, ... I split my contributions fifty-fifty
between bonds and equities” (Zweig, 1998). Markowitz was not alone. During the period when
TIAA-CREF had only two options--TIAA invests in fixed income securities and CREF invests
in equities--more than half the participants had selected a 50/50 split.
Markowitz’s strategy can be viewed as naïve diversification: when faced with “n”
options, divide assets evenly across the options. We have dubbed this heuristic the “1/n rule.”
Consider the following experiment Read and Loewenstein (1995) conducted on Halloween night.
The “subjects” were trick-or-treaters. In one condition, the children approached two adjacent
houses and were offered a choice between two candy bars (Three Musketeers and Milky Way) at
each house. In the other condition, they approached a single house where they were asked to
“choose whichever two candy bars you like.” Large piles of both candies were displayed to
ensure that the children would not think it was rude to take two of the same. The results showed
a strong diversification bias in the simultaneous choice condition: every child selected one of
each candy (see also earlier work by Simonson, 1990). In contrast, only 48 percent of the
children in the sequential choice condition picked different candies.
While the consequences of picking two different candies are minimal, applying naïve
diversification heuristics to portfolio selection could have more significant consequences. In one
study, UCLA employees were asked to allocate their retirement contributions among five
investment funds. One group of employees was presented with four equity funds and one fund
investing in fixed-income securities, whereas another group of employees was presented with
four fixed-income funds and one equity fund. The menu of funds had a strong effect on portfolio
choices. Those offered one equity fund allocated 43 percent to equities, whereas those offered
multiple equity funds ended up with 68 percent in equities. This experiment was designed to
replicate the actual menu of funds then offered to UCLA employees and pilots at Trans World
Airlines (TWA), with TWA having the equity-dominated menu of funds. The study results are in
line with the actual equity exposure of the two plans, which are 34 percent for UCLA and 75
percent for TWA (Benartzi and Thaler, 2001).
To complement this experiment, we also examined cross-sectional data on 170 retirement
saving plans. We used the number of equity funds relative to the total number of funds offered
to categorize retirement saving plans into three equal-sized groups.
The relative number of
equity funds for the three groups was 0.37, 0.65, and 0.81, respectively. For a plan with ten
investment options, for example, a 0.37 figure implies that roughly four of the options are equity
6 We made some adjustments for the time each investment fund was introduced to the plan, because inertia predicts
that newer funds will be slow to attract money, everything else being equal. See Benartzi and Thaler (2001) for
more details on the exact calculations.
funds. We found that the mean allocations to equities for each group were 48 percent, 59
percent, and 64 percent. Consistent with the diversification heuristic, the relative number of
equity funds is positively and significantly correlated and the percentage invested in equities.
The heuristics people use depend on the complexity of the situation. At a buffet dinner, if
the number of choices is small, then some version of the 1/n strategy works fine (take a bit of
each item). But when the number of options gets large, people have to devise other simplifying
strategies, such as to take one item from each category. Using this logic in the world of
retirement savings plans, it follows that adding options to plans will no longer have an effect
once the number of options gets large. Along these lines, Huberman and Jiang (2006) find a
positive correlation between the fraction of equity funds offered and the resulting allocation to
equities for plans that offer up to ten investment choices, but the correlation is no longer
significant in plans with more than ten funds.
Huberman and Jiang (2006) also find additional evidence consistent with naïve
diversification. The vast majority of participants choose a small number of funds, with the
median between three and four funds, and then tend to divide assets equally among the funds
chosen, what Huberman and Jiang call the “conditional 1/n rule.” The use of the conditional 1/n
rule appears related to the ease of applying it. When 100 is divisible by n, the conditional 1/n
rule is quite popular, but when 100 is not divisible by n, the 1/n rule is rarely used. For example,
when participants choose n = 2 or n = 4, 37 to 64 percent of them adopt the 1/n rule, but when n
= 3 the rule is only used by 18 percent of the participants. Instead, when choosing three funds,
many people adopt some other arithmetically simple division, such as .50, .25, .25.
The finding that people choose a small number of funds led us to wonder whether
participants were limited in the number of funds they could choose. An informal poll of several
members of the University of Chicago finance and economics community revealed they
incorrectly thought that four funds was the maximum number allowed. A glance at the sign-up
form revealed why faculty had this false impression: The form has only four lines for investment
elections. To choose more than four funds, a second form is needed.
This finding led us to consider whether small details, such as the number of lines on the
sign-up form, might influence the number of funds selected. We conducted an experiment on the
Morningstar.com website, which combines mutual fund and other financial information for
individual investors. We asked two groups of Morningstar.com subscribers to indicate how they
would allocate their retirement funds among a hypothetical list of eight funds. The first group
was presented with a form with four lines on it, though the participants could easily select
additional funds by clicking on a highlighted link with these instructions: “Based solely on the
above, please indicate how you would allocate your retirement contributions. You may choose
up to four funds. (If you would like to elect more than four funds, please
click here.)” The
second group of participants was shown an election form with eight lines on it. Despite the ease
of simply clicking on the link, only 10 percent of the subjects with the four-line form selected
more than four funds. In contrast, 40 percent of those viewing the eight-line form picked more
than four funds. The evidence in support of the “number of lines” hypothesis is quite strong.
As the number of funds increases, and the 1/n rule becomes impractical, investors must
adopt some other strategy. Iyengar and Kamenica (2006) report that people reduce their exposure
to equities as the menu of funds expands and becomes overwhelming. They estimate that the
addition of ten funds increases the fraction allocated by participants to money market and bond
funds by 3.28 percentage points.
Many plans have attempted to help participants deal with the difficult problem of
portfolio construction by offering “lifestyle” funds that blend stocks and bonds in a way designed
to meet the needs of different levels of risk tolerance. For example, an employer might offer
three lifestyle funds: conservative, moderate, and aggressive. These funds are already
diversified, so individuals need only pick the fund that fits their risk preference. Some funds also
adjust the asset allocation with the age of the participant.
Do participants understand how to use these diversified funds? We studied one plan that
offers both lifestyle funds and core funds. The three lifestyle funds are conservative, moderate,
and aggressive, and the six core funds include an equity index fund and a growth fund, among
others. The results are displayed in Table 1. Those who invest in the conservative lifestyle fund
allocate just 31 percent to that fund, with the rest being allocated to the core funds. Because the
menu of core funds is dominated by equity funds, the resulting equity exposure of those in the
conservative fund is 77 percent. These participants end up with a fairly aggressive portfolio,
probably without being aware of it. Participants seem reluctant to stick with one fund, even
when that fund already contains several different funds. Vanguard (2004) reports similar
findings using a much larger sample of plans. They find that participants who elect a lifestyle
fund allocate only 37 percent of their account balance to the lifestyle fund.
Indeed, individuals choosing among lifestyle funds can end up picking the same level of
risk as those constructing their own portfolios from the core options. We conducted an
experiment in which UCLA employees were assigned to one of two conditions. One group was
asked to allocate their funds between a stock fund and a bond fund. The second group was asked
to choose one of five lifestyle funds whose equity allocation varied from zero to 100 percent in
25 percent increments. Economic theory predicts that the choices made under these two
conditions should be roughly the same. However, dramatic differences arose between the two
conditions. Under the mix-it-yourself condition, individuals found the fifty-fifty allocation fairly
attractive (32 percent selected it), and only 15 percent chose an all-equity allocation. In contrast,
51 percent of those who chose among the pre-mixed portfolios selected the most aggressive
portfolio of 100 percent stocks. We ran this experiment in the late '90s, and we believe that
many people chose the 100 percent stock allocation because they were attracted to the
remarkable stock performance at the time.
The diversification heuristic does not seem to apply when people pick among pre-mixed
funds, as all the funds are perceived to be equally diversified to a naïve investor who confuses
diversification with the number of funds (see also work by Fox and Langer, 2005). Our
experiment was designed to replicate the actual difference between 401(k) plans in the U.S.,
where most people construct their own portfolios, and the Chilean social security system, where
individuals pick one of several lifestyle funds. Our findings are troubling, because small
variations in the framing of the problem result in dramatically different portfolio choices. Thus,
the findings raise difficult questions for policymakers with respect to the design of social security
systems or other retirement saving programs.
One extreme example of poor diversification occurs when employees invest in their
employer’s stock. Five million Americans have over 60 percent of their retirement savings
invested in company stock (Mitchell and Utkus, 2004). This concentration is risky on two
counts. First, a single security is much riskier than the portfolios offered by mutual funds.
Second, as employees of Enron and WorldCom discovered the hard way, workers risk losing
both their jobs and the bulk of their retirement savings all at once.
Many employees still do not think these risks apply to their own employer. First,
employees do not seem to understand the risk and return profile of company stock. When the
Boston Research Group (2002) surveyed 401(k) participants, it found that despite a high level of
awareness of the Enron experience, half of the respondents said that their company stock carries
the same or less risk than a money market fund. Similarly, Benartzi et al. (forthcoming) find that
only 33 percent of the respondents who own company stock realize that it is riskier than a
“diversified fund with many different stocks.” Even after financial education initiatives by fund
providers and plan sponsors, participants in surveys conducted by John Hancock Financial
Services during the 1992 to 2004 period continued to rate company stock as safer than a
domestic stock fund.
Second, plan participants tend to extrapolate past performance into the future. Benartzi
(2001) sorted firms into quintiles based on their stock performance over the prior 10 years and
examined subsequent allocations to company stock. Employees at the worst-performing firms
allocated 10 percent of their retirement contributions to company stock, whereas those at the best
performing firms allocated 40 percent of their contributions to company stock. Benartzi also
examined the subsequent stock performance and found no evidence that employees have any
superior information regarding their firm’s future prospects. Specifically, there was no
correlation between the allocation to company stock and subsequent stock performance.
Third, employees who receive their employer matching contribution in company stock
view their employer’s decision to match in company stock as implicit advice (Benartzi, 2001).
In particular, those who are required to take the employer match in the form of company stock
allocate 29 percent of their discretionary contributions (that is, the money they have control
over)to company stock , while those who have the option but not the requirement take the
employer match in the form of company stock allocate only 18 percent of their own funds to
Meulbroek (2002) has estimated the costs of investing in a single security instead of a
diversified portfolio (see also calculations by Poterba, 2003; Ramaswamy, 2004). The relative
value to the employee of a dollar of company stock, as opposed to a diversified stock portfolio, is
inversely related to the proportion of wealth held in company stock, the number of years the
stock will be held, and the volatility of the stock. For example, with an assumed investment
horizon of ten years and 25 percent of the assets in company stock, a dollar in company stock
only provides 58 percent as much risk-adjusted value as a diversified portfolio. Lengthening the
investment horizon to 15 years, and increasing the allocation to company stock to 50 percent,
would further reduce the value to 33 cents on the dollar. These results probably underestimate
the costs of being under-diversified, because they ignore the correlation between human capital
and the performance of company stock.
Given the substantial costs of being under-diversified, why do some employers require
that employees receive the match in company stock? Roughly speaking, employers are spending
a dollar to give employees 50 cents of benefits (in risk-adjusted utility terms). How could such
an equilibrium persist? Why does Congress even permit the use of company stock in 401(k)
plans? Note that no other individual stock is allowed to be offered in a 401(k) plan. Company
stock, however, is exempt from the diversification requirements all other 401(k) investment have
to comply with.
To understand why some employers provide the match in company stock, we surveyed
Vanguard clients (Benartzi et al., forthcoming). Employers believe that the potential increase in
motivation and productivity, the advantageous tax treatment of company stock, and placing
shares in friendly hands are the most important factors.
Some of the alleged benefits employers
attribute to company stock are overstated. For example, the evidence on the increase in
productivity is at best mixed (Prendergast, 1999). Increases in productivity are uncorrelated with
the degree of employee ownership (Kruse and Blasi, 1995). This finding is unsurprising,
because in a large firm each employee owns an extremely small fraction of the firm and has an
incredibly small effect on the overall performance. The tax advantages of company stock are
also exaggerated; we estimate them at somewhat less than 10 percent of the value of the stock.
As to the friendly hands argument, if employers are requiring their employees to hold shares in
the company to avoid takeovers, their claims to protection by the law are rather flimsy.
Market Timing: Buy High, Sell Low
Throughout the 1990s, participants were increasing their equity allocation, both in terms
of the percentage of money contributed each year and the account balances held. At the time we
speculated that there could be two reasons for this behavior. One remote possibility was that
investors had spent the decade pouring over finance and economics journals, had learned that
stock returns were substantially higher than bond returns over a long period, and so decided to
invest more in stocks. The other possibility was that more investors had come to believe that
stocks only go up, or, that even if stock prices fall, that is just another buying opportunity since
7 The tax advantages of company stock from the employer’s perspective have changed a lot over time. See Benartzi
et al. (forthcoming) for more details on the specific tax benefits.
they quickly rise again. The stock market provided an opportunity to test these competing
hypotheses during 2000-2002.
Using data from Vanguard, we calculated the mean allocations to equities from 1992
through 2002. Our calculations are based on the allocations of contributions rather than account
balances, since balances are heavily influenced by the performance of the funds, and. because of
the strong inertia exhibited by existing participants, the choices of new participants provide more
insight into the current thinking of investors. The results are shown in Panel A of Figure 2. The
equity allocation of all participants did increase from 52 percent in 1992 to 65 percent in 2000
and did not change much thereafter. However, new participants were already allocating 58
percent of their assets to equities in 1992, but that percentage rose to 74 percent in 2000.
next two years, however, the allocation to equities fell back to 54 percent. The market timing of
new participants in increasing their exposure to equities was exactly wrong.
Similar behavior is observed in the asset allocations within equities, in the plans in which
investors can choose funds that specialize in particular industries or sectors. To determine how
this option affects investors, we studied data from Hewitt Associates on a plan that offers a
technology fund. Panel B of Figure 2 displays the percentage of new participants selecting the
technology fund as well the fund’s performance. The fraction of new participants selecting that
fund increased dramatically from 12 percent to 37 percent over the course of two years, and then
it decreased by half, from 37 percent to 18 percent over the course of one year. Again,
participants were buying into the technology fund most aggressively at the peak.
8 One caveat is that the data we obtained were a snapshot of the plan participants as of midyear 2002. While we
knew the enrollment date for each participant, we observed their allocations as of midyear 2002. To the extent that
participants may have made changes to their allocations over time, any bias should be fairly minimal for the 2000 to
2002 samples, however, as those data points are relatively recent.
9 See also Elton, Gruber and Blake (2005) for related findings for plan sponsors and participants.
Mental Accounting and Framing
Mental accounting refers to the implicit methods individuals use to code and evaluate
transactions, investments, gambles and other financial outcomes (Kahneman and Tversky, 1984;
Thaler, 1985). We believe that participants use separate mental accounts for “old money”
(amounts they have already accumulated in the plan), and for “new money” (amounts they have
not yet contributed). The propensity to adjust the allocation of old money is much lower than
that of new money. Perhaps investors fear the potential regret of reallocating old money and
observing the new investment choices underperforming the original choices. With regard to new
money, however, a reference point has not been set yet, so less potential exists for regretting any
changes. Ameriks and Zeldes (2000) report that over the 1987–1996 period, only 27 percent of
the TIAA-CREF participants they studied reallocated their accumulated assets, though 53
percent reallocated their future contributions.
Mental accounting also affects company stock. In particular, employees seem to view
company stock as a unique asset class that is neither stocks nor bonds. In our sample, plans that
do not have access to company stock have half in stock funds and half in bond funds, whereas
plans with access to company stock have 42 percent in company stock and the remaining 58
percent split evenly between stock funds and bond funds (Benartzi and Thaler, 2001). As a
result, those with access to company stock invest 71 percent (42 plus half of 58) in equities.
10 One caveat is that TIAA-CREF had certain limitations on the reallocation of old money, which could explain the
lower frequency of reallocating old money versus new money. However, data from another provider which does not
restrict the reallocation of old money reveals a similar pattern (Hewitt Associates, 2005). In particular, 16.7 percent
of plan participants reallocated old money in 2004, whereas 21.4 percent reallocated new money.
11 Additional evidence on the powerful role of inertia and lack of rebalancing activity is provided by Mitchell et al.
Individuals investing in company stock do not seem to realize that company stock is part of their
Framing is another important factor in participants’ behavior. Providing plan participants
with short-term rates of return on the different investment funds induces “myopic loss aversion”
(Benartzi and Thaler, 1995). Loss aversion refers to the tendency of individuals to weigh losses
about twice as much as gains (Kahneman and Tversky, 1979, 1991), whereas the myopic
component is the tendency of individuals to evaluate their portfolios too often. As a result,
individuals become hypersensitive to short-term losses. We ran an experiment in which we
showed individuals one-year returns or long-term simulated returns for a stock fund and a bond
fund. We found that those viewing the one-year returns allocated just 41 percent to stocks,
whereas those viewing the longer-term returns allocated 82 percent to stocks (Benartzi and
Thaler, 1999). These results have significant implications for how often plan sponsors and plan
providers should convey information to plan participants.
Rational but unsophisticated investors may ask a knowledgeable expert for help. But
while individuals do ask others for advice, their “advisors” tend to be their spouses and friends,
who don’t necessarily qualify as experts (Benartzi and Thaler, 1999). One interesting anecdote
comes from a chain of supermarkets operating in Texas.
The plan provider noticed that
participants’ behavior in each supermarket was remarkably homogeneous, but the behavior
across supermarkets was fairly heterogeneous. It turns out that most of the supermarket
employees considered the store butcher to be the investment maven and would turn to him for
12 We thank Ken Robertson from the 401kcompany for sharing his data and experience with us.
advice. Thus, depending on the investment philosophy of the butcher at each individual location,
employees ended up heavily invested in either stocks or bonds.
Similar strong peer effects are documented by Duflo and Saez (2000, 2002) in a study of
the retirement plan participation at 11 libraries of a large university. In this system, prospective
librarians are interviewed and hired by the central library, so there is no reason to expect a large
variation across the libraries in demographics or in the propensity to save. Indeed, the data
confirmed that there were no demographic differences across libraries, yet plan participation
varied dramatically across libraries, from a low of 14 percent to a high of 73 percent, illustrating
strong peer effects.
How Much Is Investor Autonomy Worth?
One advantage of defined contribution retirement plans is that they allow for variation in
individual tastes, both for saving and for risk bearing. The trend over time has been to allow
more flexibility, both in savings rates and investments. For example, although private 401(k)
plans are relatively new, defined contribution plans have existed at universities since 1918 when
TIAA was formed. In these original plans, there was only one option (TIAA—a fixed income
vehicle) until 1952, when CREF (which invests in equities) was launched. The number of
options remained at two until 1988. Furthermore, at many universities, a minimum savings rate
is specified by the university. The employee is required to save at least x percent and the
university will contribute y percent. These minimums are quite high (x+y is often between 10
and 15 percent) relative to the average savings rates in private plans.
Do private plans that offer more choice in savings rates have higher contribution rates
than the university plans that are more restricted? We know of no thorough analysis of this
question, but the low savings rates observed in some private plans certainly raise questions about
In studying asset allocation, we have investigated whether participants do a good job--as
judged by themselves--in picking a portfolio. Using a plan with participants defaulted into a
professionally managed account based on their age (Benartzi and Thaler, 2002), we studied the
choices of those participants who elected to opt out of the default investment and form their own
portfolio. Using software provided by Financial Engines (the financial advice firm founded by
William Sharpe), we projected for each employee the distribution of retirement income for three
portfolios: a) the employee’s self-constructed portfolio; b) the average portfolio for all
employees who had opted out of the professionally managed accounts; and c) the professionally
managed account the employee turned down. We presented the subjects with the three
(unlabeled) distributions of projected retirement income and asked them to rate the three
investment programs on a scale of one (very unattractive) to five (very attractive).
Participants’ self-constructed portfolios received the lowest average rating, 2.75, the
average portfolio received slightly higher mean ratings of 3.03, and the professionally managed
portfolios received the significantly higher mean rating of 3.50. Even among the sample of
participants who stated a preference to construct their own portfolios, 80 percent found the
managed account solution more attractive! These employees were not behaving in a directly
inconsistent manner, since when they made their initial decision to reject the default asset
allocation and form their own, they were probably not using specialized financial software. Of
course, a firm could try to improve individual investment choices by providing similar software,
but firms that have made such software available have not found a very high usage rate.
Another more indirect test of the value of active portfolio choices in retirement plans
comes from the partial privatization of the Swedish social security system launched in 2000
(Cronqvist and Thaler, 2004). Private accounts were created for each worker, and a portion of
the payroll tax was contributed to this account. Workers could choose from an array of 456
funds, one of which was designated as the default fund. (The number of funds has since grown to
over 600.) The default fund was carefully constructed, well diversified, and had very low fees
(16 basis points), but participants were urged by the Swedish government to eschew the default
fund and select their own portfolio of up to five funds. Two-thirds of participants took this
advice and formed their own portfolios. The average portfolio actively selected had higher fees
(77 basis points), more risk, and strong “home bias” (French and Poterba, 1991) with a very high
concentration of Swedish stocks (48 percent). The active portfolios also underperformed the
default fund by 9.7 percent (cumulative) over the first three years of the system.
system has since stopped encouraging active decision making, and of those workers joining the
system for the first time in 2003, only 8.4 percent made an active choice.
One might wonder whether suboptimal portfolio choices are costly. Calvet, Campbell
and Sodini (2006), for example, find that a lot of Swedish households own portfolios that are
close to the efficient frontier, so perhaps suboptimal portfolio choices are inconsequential. We
tend to disagree for several reasons. While many Swedish households own well-diversified
portfolios, Calvet, Campbell and Sodini also report that 38 percent of Swedish households do not
participate in the equity market, estimating the return loss from nonparticipation at 4.3% per
year. Brennan and Torous (1999) also point out that picking the wrong portfolio along the
efficient frontier could be very costly. Using the calculations in Brennan and Torous, Benartzi
13 Similar evidence from the U.S. is provided by Yamaguchi et al. (2006) who find that participants in balanced
funds (or lifestyle / lifecycle funds) earn the highest risk-adjusted rates of return.
and Thaler (2001) showed that investors using the 1/n heuristic could experience utility loss of
more than 25 percent by picking portfolios that are either too conservative or too aggressive for
their own preferences.
Choosing Between Defined Benefit and Defined Contribution Plans
The employees of state governments with retirement programs are sometimes given a
choice between a defined benefit and a defined contribution retirement plan, which present
another opportunity to study high-stakes decision making in the savings domain. We studied one
such large public employer, which offered all employees three options: remain in the existing
defined benefit plan, choose a new defined contribution plan, or a hybrid option to keep existing
benefits under the defined benefit plan and to accumulate future accruals under the defined
contribution plan. Those (vested) employees who switched from the defined benefit to the
defined contribution plan would receive an actuarially-fair lump sum contribution to their
defined contribution plan.
In this setting, it was usually not possible to determine the “rational” choice for a given
participant, since personal preferences were not known, with one important exception. The
defined benefit and the defined contribution plan had very different vesting schedules: one year
for the defined contribution plan and six years for the defined benefit plan. Thus, an employee
whose tenure in the defined benefit plan was less than six years received no benefits. This meant
that employees needed to estimate their expected tenure with the employer to make a good
choice. Turnover for young or new employees was particularly high, so these employees were
almost certainly better off choosing the defined contribution plan. For example, the plan actuary
estimates that a 31-year-old employee with one year of service has approximately a one-in-ten
chance of working for the same employer through the plan’s normal retirement age of 62.
In Figure 3 we illustrate the projected income replacement ratio for a 31-year-old
employee under the defined benefit and the defined contribution plans as a function of the age at
which the employee terminates employment. Whereas the defined benefit plan could provide an
income replacement ratio of 66.7 percent after 32 years of service, slightly higher than the 59.9
percent for the defined contribution plan, under most scenarios the defined benefit plan provides
a lower income. The likelihood of breaking even under the defined benefit plan--that is working
long enough for the current employer so that the defined contribution and defined benefit
replacement ratios are identical or the defined benefit plan is a better choice--is only 13 percent.
[Insert Figure 3 About Here]
Data on participants’ choices reveals that only 7 percent of those with less than two years
of service selected the defined contribution plan (as of February 28, 2003). There are several
potential explanations. First, the defined benefit plan was set as the default choice, and 63
percent of the participants ended up in the defined benefit plan by default. Interestingly, when
surveyed beforehand, only 10 percent of the participants planned on being defaulted into the
defined benefit plan, and many more predicted they would choose the defined contribution plan
than actually did. Second, employees vastly overestimate their expected tenure Working for the
state. For instance, when new employees are asked about the likelihood of remaining with their
current employer until retirement age, the gap between participants’ expectations and the plan
actuary’s predictions reaches 40 percent. Third, in spite of a serious effort to educate the
employees about their options, they had very little understanding of the plan’s features. For
example, only 19 percent of the employees realized that there was a one-year vesting
requirement under the defined contribution plan. And finally, the choice came in the second
half of 2002, in the midst of a bear market. This timing likely discouraged participants from
choosing the DC plan.
Other studies on the choice between defined benefit and defined contribution plans also
suggest that relatively few workers select the defined contribution option and that the default
choice could have a dramatic effect. Papke (2004), for example, reports that only 1.6 percent of
the corrections workers covered by the State of Michigan Employee Retirement System elected
the defined contribution plan.
Additional evidence on suboptimal choices between defined benefit and defined
contribution plans come from Brown and Weisbenner (2006). They investigated elections made
by State of Illinois employees who were offered a choice among a traditional defined benefit
plan, a portable defined benefit plan and a defined contribution plan. The authors argue that
under most circumstances the portable defined benefit plan dominates the defined contribution
plan, as it has a more generous employer match. Yet, a non trivial fraction of new employees
(15 percent) elect the defined contribution plan. Furthermore, many more employees elected the
14 A study by Yang (2005) is an exception, reporting take-up rates for a defined contribution plan of up to 50
percent, though two caveats are worth mentioning. First, the plan choice in Yang’s study took place in March 2000,
at the peak of the bull market. Second, the information provided to employees in this plan displayed the projected
defined contribution benefits with the full employer match. Those selecting either a low contribution rate or not
contributing their own money to the plan would not get the full employer match and could expect much lower
benefits from the defined contribution plan. Interestingly, Yang finds that those who did not make a choice and
were defaulted into the defined benefit plan were more similar to the defined contribution choosers than the defined
benefit choosers. For example, the average age of defined benefit defaulters was 38, closer to that of defined
contribution choosers at 40 than to that of defined benefit choosers at 53.
defined contribution plan in 1999, right before the market crash. This is consistent with our
earlier discussion of negative market timing by plan participants.
Interventions by Plan Sponsors
What can employers do so that more plan participants enroll in retirement plans,
contribute an amount that will build a reasonable retirement nest-egg, and allocate the funds
among assets in an appropriately diversified way? There are two broad classes of interventions:
education and plan design.
Many employers have tried to educate their employees to make better decisions or
supplied tools to help them improve their choices. The empirical evidence does not suggest that
these methods are, in and of themselves, adequate solutions to the problems. The same large
employer discussed above that offered its employees the chance to switch from a defined benefit
to a defined contribution plan offered its employees a financial education program free of
charge. The employer measured the effectiveness of this education by administering a before-
and-after test of financial literacy. The quiz used a True/False format, so random answers would
receive, on average, a score of 50 percent. Before the education, the average score of the
employees was 54; after the education, the average score jumped to 55. As professors know,
teaching is hard.
Using education to increase participation and contribution rates has generally led to
disappointing results. Employees often leave educational seminars excited about saving more,
but then fail to follow through on. For example, Choi et al. (2002) measured the effectiveness of
employee seminars. At the seminar everyone expressed an interest in saving more, but only 14
percent actually joined the savings plan, not much better than the 7 percent of comparable
employees who did not attend a seminar and joined the savings plan. Similarly, Duflo and Saez
(2003) find that the attendance at a “benefit fair” has only a small effect on participation in a tax-
deferred savings account.
The difficulties of explaining the “right” choices to people are nicely illustrated by an
experiment conducted by Choi, Laibson, and Madrian (2004b), who as discussed earlier tried to
investigate why participants fail to join a retirement plan with a company match even when
joining the plan is an arbitrage opportunity (because the employees are over 59.5 years old and
can immediately withdraw contributions without penalty). Choi et al. conducted an experiment in
which some employees received a survey about this free lunch and instructions explaining how
to go about eating it. Filling out this survey had a negligible and insignificant effect on behavior.
The most optimistic results for how education can improve saving are found by Bernheim
and Garrett (2003) and Bernheim et al. (2001). They use cross-sectional surveys of individuals
from the population, rather than the employees in a specific company. For example, Bernheim
and Garrett use a survey that asks people whether financial education is available in their
workplace. They find that workers who report that financial education is available where they
work are more likely to save, both retirement saving and other forms. However, this method
faces problems that could induce a spurious correlation. For example, workers who are likely to
save may also be more likely to be aware of the availability of financial education. Discovering
such a correlation does not show that education will affect the behavior of young worker, who
are not thinking about retirement and not in the retirement plan.
The main alternative to education as a method of influencing decisions about retirement
savings plans is to choose the features of the retirement savings plan in a way that will promote
the desired objectives. The simplest change is automatic enrollment. Although automatic
enrollment is very effective at getting new and young workers to enroll sooner than they would
have otherwise, participants tend to stick with the default contribution rate, which is typically
quite low. To mitigate this problem, we devised a program of automatic escalation of
contributions to such plans called Save More Tomorrow.
Save More Tomorrow was constructed with certain psychological principles in mind.
First, people find it easier to accept self-control restrictions that take place in the future. For
example, many people plan to start their diet the next day and to join a gym next month. Second,
potential losses have roughly twice the effect on people’s decision-making as gains (Kahneman
and Tversky, 1979, 1992). Third, losses are evaluated in nominal terms (Kahneman, Knetch, and
Thaler, 1986; Shafir, Diamond, and Tversky, 1997). Fourth, inertia plays a powerful role in
With the above principles in mind, Save More Tomorrow invites participants to pre-
commit to save more every time they get a pay raise. By synchronizing pay raises and savings
increases, participants never see their take-home amounts go down, and they don’t view their
increased retirement contributions as a loss. Once someone joins the program, the saving
increases are automatic, using inertia to increase savings rather than prevent savings. When
combined with automatic enrollment, this design can achieve both high participation rates and
increased saving rates.
15 Save More Tomorrow is available at no charge to vendors who are willing to share data for research purposes.
Many retirement plan administratorshave adopted the idea, including Vanguard, T. Rowe
Price, TIAA-CREF, Fidelity, and Hewitt Associates, and it is now available in thousands of
employer plans. The first implementation, at a mid-sized manufacturing firm, provides the
longest time series of results. Initially, employees were invited to chat with a financial
consultant, and about 90 percent accepted that offer. Given the very low savings rates of the
plan participants, the advisor almost always told the employees that they needed to save much
more than their current rate, but he capped his recommended saving increase at 5 percent points
of pay, fearing that people might find larger increases impossible to implement. Twenty-five
percent of the participants too this advice and immediately increased their savings rates by the
recommended 5 percent points. Those who rejected the advisor’s advice were offered the Save
More Tomorrow program. Specifically, they were told that their saving rates would go up by 3
percentage points every time they got a pay raise. Pay raises were about 3.25 to 3.50 percent.
Out of the group who could not increase their savings rate immediately, 78 percent joined the
program to increase their contribution every time they get a pay raise. The results were dramatic.
Those in the Save More Tomorrow program started with the lowest savings rate, around 3.5
percent. After three and a half years and four pay raises, their saving rate had almost quadrupled
to 13.6 percent, considerably higher than the 8.8 percent savings rate for those who accepted the
consultant’s initial recommendation to raise savings by 5 percentage points. In addition, most
people in the program remained in it through the entire period. Most of the few that did leave the
program just stopped the increases; they did not set their retirement rate back to where it had
been prior to joining the program.
The Save More Tomorrow program could be extremely effective at increasing saving
rates if joining the program is made easy, or even an automatic default choice. One plan that
automatically enrolled participants into the program reports that less than 5 percent of the
participants opted out after the first savings increase.
We find this result especially
encouraging because the savings increases and pay raises were not synchronized, so participants
did see their take-home amounts decrease.
Plan design features could also be used to improve participants’ portfolio choices. One
option is to offer a set of model portfolios that have varying degrees of risk. For example, a plan
sponsor could offer conservative, moderate and aggressive “lifestyle” portfolios. All the
participants need to do is select the lifestyle fund that best fits their risk preferences. Another
option available to plan sponsors is to offer plan participants “target maturity funds.” Target
maturity funds typically have a year in their name, like 2010, 2030, or 2040. A participant simply
selects the fund that matches his or her expected retirement date. Managers of the target
maturity funds select the degree of risk and reduce the allocation to stocks as the target date
Some vendors and plan sponsors have started to offer automated solutions for portfolio
selection. In particular, some plan sponsors automatically assign participants to the target
maturity fund based on a standard retirement age. Others are defaulting participants into
“managed accounts,” which are typically portfolios of stocks and bonds that are based on the age
of the participants and possibly other information. In either case, participants can opt out of the
default investment and choose their own portfolio. Providing these sensible default investment
is an idea with considerable merit, given that participants find the risk and return
16 We are grateful to Jodi Dicenzo, who made this field experiment and many others happen.
17 Kamenica (2006) finds that people with more typical characteristics are more likely to accept the default asset
allocation in their 401(k) plans, a helpful finding in thinking about how to structure these plans.
profiles of the automated solutions more attractive than their self-constructed portfolios.
Iyengar, Huberman and Jiang (2004) noted, simplifying the investment selection process could
encourage more employees to join the retirement plan.
Saving for retirement is a difficult problem, and most employees have little training upon
which to draw in making the relevant decisions. Perhaps as a result, investors are relatively
passive. They are slow to join advantageous plans; they make infrequent changes; and they
adopt naïve diversification strategies. In short, they need all the help they can get. Fortunately,
many effective ways to help participants are also the least costly interventions: namely, small
changes in plan design, sensible default options and opportunities to increase savings rates and
rebalance portfolios automatically. These design features help less sophisticated investors while
maintaining flexibility for more sophisticated types.
18 One way that firms could address the “company stock” problem is by implementing something called the “Sell
More Tomorrow” plan (Benartzi and Thaler, 2003), in which employees are first educated on the risk and return
profile of company stock in plain English, and then they are offered a gradual selling program that automatically
divests them of a small portion of their holdings every month. To ensure that employees don’t feel like they are
“missing the boat,” the program could be set to keep a small portion of their portfolio (say 5 percent) in company
Benartzi is grateful for financial support from Reish Luftman McDaniel & Reicher and
Vanguard. We are also grateful to Jodi Dicenzo, Wayne Gates of John Hancock, Lori Lucas of
Hewitt Associates, John Rekenthaler of Morningstar, Jason Scott of Financial Engines, Brian
Tarbox, Steve Utkus of Vanguard, and Carol Waddell of T. Rowe Price for all the data they have
provided us over the years. We received very helpful comments from Emir Kamenica and all the
editors. An earlier version of this paper was presented to the AARP Public Policy Institute.
Ameriks, John, and Stephen P. Zeldes, 2000, “How do Household Portfolio Shares Vary with
Age?” working paper, Columbia University.
Benartzi, Shlomo, 2001, "Excessive Extrapolation and the Allocation of 401(k) Accounts to
Company Stock?" Journal of Finance 56.5, pp. 1747-1764.
Benartzi, Shlomo, and Richard H. Thaler, 2003, "Using Behavioral Economics to Increase
Employee Diversification in 401(k) Plans: Solving the Company Stock Problem,” working
paper, University of California-Los Angeles.
Benartzi, Shlomo, and Richard H. Thaler, 2002, "How Much Is Investor Autonomy Worth?"
Journal of Finance 57.4, pp. 1593-1616.
Benartzi, Shlomo, and Richard H. Thaler, 2001, “Naive Diversification Strategies in Retirement
Saving Plans,” American Economic Review 91.1, pp. 79-98.
Benartzi, Shlomo, and Richard H. Thaler, 1999, “Risk Aversion or Myopia? Choices in Repeated
Gambles and Retirement Investments,” Management Science 45.3, pp. 364-381.
Benartzi, Shlomo, and Richard H. Thaler, 1995, "Myopic Loss-Aversion and the Equity
Premium Puzzle," Quarterly Journal of Economics 110.1, pp. 73-92.
Benartzi, Shlomo, Richard H. Thaler, Stephen P. Utkus, and Cass R. Sunstein, forthcoming,
“Company Stock, Market Rationality, and Legal Reform,” Journal of Law and Economics.
Bernheim, Douglas, B., and Garrett, Daniel M., 2003, “The Effects of Financial Education in the
Workplace: Evidence from a Survey of Households,” Journal of Public Economics 87, pp.
Bernheim, Douglas, B., Garrett, Daniel M., and Maki, Dean M., 2001, “Education and Saving:
the Long-term Effects of High School Financial Curriculum Mandates,” Journal of Public
Economics 80.3, pp. 435–465.
Brennan, Michael, J., and Walter N. Torous, 1999, “Individual Decision Making and Investor
Welfare,” Economic Notes 28.2, pp. 119-143.
The Boston Group, April 25, 2002, “Enron Has Little Effect on 401(k) Participants’ View of
Company Stock,” Boston: MA.
Brown, Jeffrey R., and Scott J. Weisbenner, 2006, “Who Chooses Defined Contribution Plans?”
Working Paper, University of Illinois at Urbana-Champaign.
Calvet, Laurent, E., John Y. Campbell, and Paolo Sodini, 2006, “Down or Out: Assessing the
Welfare Costs of Household Investment Mistakes,” Working Paper, Harvard University.
Choi, James, J., David Laibson, Brigitte Madrian and Andrew Metrick, forthcoming, "Saving For
Retirement on the Path of Least Resistance,” in Behavioral Public Finance, eds Slemrod and
McCaffrey, Russell Sage.
Choi, James, J., David Laibson, Brigitte Madrian and Andrew Metrick, 2005, “Optimal Defaults
and Active Decisions,” NBER Working Paper No. 11074.
Choi, James J., David Laibson, Brigitte Madrian and Andrew Metrick, 2004a, “For Better or For
Worse: Default Effects and 401(k) Savings Behavior,” in Perspectives in the Economics of
Aging, David Wise, ed., pp. 81 – 121. University of Chicago Press.
Choi, James J., David Laibson, and Brigitte Madrian, 2004b, “$100 Bills on the Sidewalk:
Violation of No-Arbitrage in 401(k) Accounts,” Working Paper, University of Pennsylvania.
Choi, James J., David Laibson, Brigitte Madrian and Andrew Metrick, 2002, “Defined
Contribution Pensions: Plan Rules, Participant Decisions, and the Path of Least Resistance,”
in Tax Policy and the Economy, James Poterba, ed., pp. 67 -113, Volume 16, MIT Press.
Cronqvist, Henrik, and Richard H. Thaler, 2004, “Design choices in Privatized Social-Security
Systems: Learning from the Swedish Experience,” American Economic Review 94.2, pp.
Duflo, Esther, William Gale, Jeffrey Liebman, Peter Orszag, and Emmanuel Saez, 2005, “Saving
Incentives for Low- and Middle-Income Families: Evidence from a Field Experiment with
H&R Block,” NBER Working Paper No. 11680.
Duflo, Esther, and Emmanuel Saez, 2002, “The Role of Information and Social Interactions in
Retirement Plan Decisions: Evidence from a Randomized Experiment,” NBER working
Duflo, Esther, and Emmanuel Saez, 2000, “Participation and Investment Decisions in a
Retirement Plan: the Influence of Colleagues’ Choices,” working paper, MIT.
Fox, Craig R., and Thomas Langer, 2005, “Biases in Allocation under Risk and Uncertainty:
Partition Dependence, Unit Dependence, and Procedure Dependence,” working paper,
University of California – Los Angeles.
French, Kenneth R., and Poterba, James M., 1991, “Investor Diversification and International
Equity Markets,” American Economic Review 81.2, pp. 222-226.
Gilovich, Thomas, Dale Griffin, and Daneil Kahneman, Ed., 2002, “Heuristics and Biases: The
Psychology of Intuitive Judgment,” Cambridge University Press: New York.
Hewitt Financial Services, 2005, “Hewitt Research Report: How Well are Employees Saving and
Investing in 401(k) Plans.”
Hewitt Financial Services, 2002a, “Investing in Round Numbers.”
Hewitt Financial Services, 2002b, “How High Should Plan Sponsors Go In Raising the Before-
Tax Plan Maximum Under EGTRRA?”
Huberman, Gur, and Wei Jiang, 2006, “Offering versus Choice in 401(K) Plans: Equity
Exposure and Number of Funds,” Journal of Finance 61.2, 763 – 801.
Iyengar, S.S., Huberman, G., and Jiang, W. (2004), “How Much Choice is Too Much:
Determinants of Individual Contributions in 401K Retirement Plans,” in O.S. Mitchell and S.
Utkus (Eds.), Pension Design and Structure: New Lessons from Behavioral Finance (pp. 83-
95). Oxford: Oxford University Press.
Iyengar, Sheena and Emir Kamenica, 2006, “Choice Overload and Simplicity Seeking,” working
paper Columbia University.
Kahneman, Daniel, and Amos Tversky, 1984, "Choices, Values and Frames," American
Psychologist 39, pp. 341-350.
Kahneman, Daniel, and Amos Tversky, 1979, “Prospect Theory: An Analysis of Decision Under
Risk,” Econometrica 47, pp. 263-291.
Kahneman, Daniel., Jack L. Knetsch, and Richard H. Thaler, 1986, “Fairness as a Constraint on
Profit Seeking,” American Economics Review 76, pp. 728-741.
Kamenica, Emir., 2006 “Contextual Inference in Markets: On the Informational Content of
Product Lines” University of Chicago working paper
Kruse, Douglas L., and Joseph Blasi, September 1995, “Employee Ownership, Employee
Attitudes, and Firm Performance,” NBER Working Paper No. 5277.
Madrian, Brigitte C., and Dennis F. Shea, 2001a, "The Power of Suggestion: Inertia in 401(k)
Participation and Savings Behavior," Quarterly Journal of Economics 116, pp. 1149-1525.
Madrian, Brigitte C., and Dennis F. Shea, 2001b, “Preaching to the Converted and Converting
Those Taught: Financial Education in the Workplace,” working paper, University of
Meulbroek, Lisa, 2002, “Company Stock in Pension Plans: How Costly Is It?” working paper
02-058, Harvard Business School.
Mitchell, Olivia S., Gary R. Mottola, Stephen P. Utkus, and Takeshi Yamaguchi, 2006, “The
Inattentive Participant: Portfolio Trading Behavior in 401(k) Plans,” Pension Research
Council Working Paper #2006-5, The Wharton School, University of Pennsylvania.
Mitchell, S. Olivia, and Stephen P. Utkus, 2004, “The Role of Company Stock in Defined
Contribution Plans,” In Olivia Mitchell and Kent Smetters, eds., The Pension Challenge:
Risk Transfers and Retirement Income Security, Oxford: Oxford University Press, pp. 33-70.
Poterba, James, Steven Venti, and David A. Wise, 2006, “The Decline of Defined Benefit
Retirement Plans and Asset Flows,” Working Paper, MIT.
Poterba, James M., 2003, “Employer Stock and 401(k) Plans,” working paper, MIT.
Prendergast, Canice, 1999, “The Provision of Incentives in Firms,” Journal of Economic
Literature 37, pp. 7-63.
Ramaswamy, Krishna, 2004, “Corporate Stock and Pension Plan Diversification,” In Olivia
Mitchell and Kent Smetters, eds., The Pension Challenge: Risk Transfers and Retirement
Income Security, Oxford: Oxford University Press, pp71-88.
Read, Daniel, and George Loewenstein, 1995, "Diversification Bias: Explaining the Discrepancy
in Variety Seeking between Combined and Separated Choices," Journal of Experimental
Psychology: Applied 1.1, pp. 34-49.
Shafir, Eldar, Peter Diamond and Amos Tversky, 1997, “Money Illusion,” Quarterly Journal of
Economics CXII (May), pp. 341-374.
Siegel, Jeremy, 1998, “Stocks for the Long Run,” McGraw-Hill Companies.
Simonson, Itamar, 1990, "The Effect of Purchase Quantity and Timing on Variety-Seeking
Behavior," Journal of Marketing Research 28, pp. 150-62.
Skinner, Jonathan, “Are You Sure You’re Saving Enough for Retirement?” Journal of Economic
Perspectives, in this issue.
Thaler, Richard H., 1999, "Mental Accounting Matters," Journal of Behavioral Decision Making
12, pp. 183-206.
Thaler, Richard H., and Shlomo Benartzi, 2004, "Save More Tomorrow: Using Behavioral
Economics to Increase Employee Savings," Journal of Political Economy 112.1, Part 2, pp.
S164 - S187.
Tversky, Amos, and Daniel Kahneman, 1992, “Advances in Prospect Theory: Cumulative
Representation of Uncertainty,” Journal of Risk and Uncertainty 5, pp. 297-323.
Vanguard, 2004, “How America Saves: A Report on Vanguard Defined Contribution Plans
Yamaguchi, Takeshi, Olivia S. Mitchell, Gary R. Mottola, and Stephen P. Utkus, 2006, “Winners
and Losers: 401(k) Trading and Portfolio Performance,” Pension Research Council Working
Paper #2006-26, The Wharton School, University of Pennsylvania.
Yang, Tongxuan, 2005, “Understanding the Defined Benefit versus Defined Contribution
Choice,” Pension Research Council Working Paper #2005-4, The Wharton School,
University of Pennsylvania.
Zweig, Jason, January 1988, “Five Investing Lessons from America's Top Pension Fund,”
Money, pp. 115-118.
The Distribution of Contribution Rates for New Participants before and after the Maximum Rate
was Increased from 16% to 100% of Pay
Percentage of Participants
The above chart displays the distribution of contribution rates at a large defined contribution plan
administered by Hewitt Associates. In 2002, the maximum contribution rate allowed under the
plan was increased from 16 percent to 100 percent of pay in accordance with EGTRRA. The
chart displays the distribution of contribution rates for participants who joined the plan in 2001
versus those who joined in 2002 after the maximum rate was increased. For more details, see
Hewitt Financial Services (200b).
Panel A: The Equity Allocation of New versus All Plan Participants
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Prct. of New Contributions in Equitess
New Participants All Participants
Panel B: Percentage of New Participants Selecting the Technology Fund
1997 1998 1999 2000 2001
% Selecting the Tech Fund
1997 1998 1999 2000 2001
Tech Fund Share Price
Panel A displays the percentage of new contributions allocated to equities by new versus all plan
participants. “New” participants are those entering the plan in a given year. The chart was
constructed from data provided by Vanguard. Panel B reports the allocations of new participants
at a large plan that offers a technology fund. The left axis displays the percentage of new
participants allocating some of their contributions to the technology, and the right axis shows the
fund’s share price. Data were provided by Hewitt Associates.
Income Replacement Ratios for a Defined Benefits Plan vs. a Defined Contribution Plan
30 35 40 45 50 55 60 65
Age Terminating Employment
Income Replacement Ratio
The above chart displays income replacement ratios for a large employer offering a choice
between a defined benefits plan versus a defined contribution plan. The illustration is based on a
30-year-old individual with one year of service credit, and it shows income replacement ratios
for the defined benefit versus the defined contribution plan as a function of the age at which the
employee terminates employment. The data were provided by the plan actuary.
Allocation of Contributions for a Plan Offering a Mix of Lifestyle Funds and Core Funds
NOT in any
Core Funds 66% 55% 54% 100%
Conservative Lifestyle Fund 31 1 0 N/A
Moderate Lifestyle Fund 3 42 4 N/A
Aggressive Lifestyle Fund 0 2 42 N/A
Total Equity Exposure 77 80 89 78
The above table displays investment elections made by employees at a large 401(k) plan offering
a choice among pre-mixed model portfolios (i.e., the Conservative, Moderate and Aggressive
Lifestyle funds) and core funds (for example, an equity index fund). The table describes the
average allocations of future contributions among the model portfolios and the core funds. The
table also provides the total equity allocations for those investing in the various model portfolios
as well as those not investing in any of the model portfolios.