Asset MNGT PDF

ESSAYS ON 401(K) ASSET MANAGEMENT
Takeshi Yamaguchi
A DISSERTATION
in
Insurance and Risk Management
For the Graduate Group in Managerial Science and Applied Economics

Presented to the Faculties o f the University o f Pennsylvania in Partial Fulfillment
of the Requirement for the Degree o f Doctor o f Philosophy
2007
Supervisor o f Dissertation
Prff-cWGraduate Group Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI N um ber: 3261011
Copyright 2007 by
Yamaguchi, Takeshi
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
UMI
UMI Microform 3261011
Copyright 2007 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company

300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
COPYRIGHT
Takeshi Yamaguchi
2007
To Aya, Eric and Justin
111
ACKNOWLEDGMENTS
I am grateful to many people for their support and contribution to this dissertation.
First, I wish to thank my committee, Olivia S. Mitchell (Chair), Stephen Shore,
Alexander Muermann, and Brigitte Madrian for their inputs, suggestions, criticism and
directions to this research. Especially, I would like to thanks my advisor, Olivia Mitchell.
Without her guidance, encouragement, support and patience, it is impossible to complete
my dissertation. It is she who taught me how to conduct academic research, how to well
organize works, and even how to struggle with frustrations and difficulties. I am also
grateful to Steve Utkus, Gary Mottola from Vanguard, for helpful comments throughout
the whole process and recordkeeping data under restricted access conditions. I am also
indebted to Masao Tamura, Minoru Nakamura and Hiroko Tominaga for their thoughts,
suggestions and comments from the U.S-Japan comparison perspective.
I would like to acknowledge my fellow doctoral students with whom I spent this
compact and dense four years. They offered me invaluable inspiration, discussion and
friendship. We shared insights o f classes, seminars and research, supported and
encouraged each other in the whole doctoral training process.
I would like to thank Pension Research Council at the W harton School o f the
University o f Pennsylvania, Michigan Retirement Research Center, Bradley Foundation,
Steven H. Sandell Foundation, Nomura Research Institute and Nomura Securities for
their financial support.
iv
Finally, I am grateful to my wife Aya and two sons, Eric and Justin. It is Ayas
support, patience and love that encourage me to devote m yself to my doctoral research;
the Angels smile o f Eric and Justin provides me unwavering courage and power to
overcome any difficulty toward this destination o f my Ph.D. voyage.
Takeshi Yamaguchi
Philadelphia, Pennsylvania
April 2007
ABSTRAT
ESSAYS ON 401(K) ASSET MANAGEMENT
Takeshi Yamaguchi
Olivia S. Mitchell (Advisor)
In the first chapter o f this dissertation, we investigate the determinants o f trading

activities in defined contribution (DC) pension plans. Using a rich new dataset on 1.2
million workers in over 1,500 plans, we find that most DC plan participants are
characterized by profound inertia. Almost all participants (80%) initiate no trades, and an
additional 11% makes only a single trade, in a two-year period. Even among traders,
portfolio turnover rates are one-third the rate o f professional money managers. Those
who trade in their DC plans are more affluent older men, with higher incomes and longer
job tenure. They tend to use the internet for DC plan account access, hold a larger number
o f investment options, and are more likely to hold active equity funds rather than index or
lifecycle funds. Some plan features, including offering own-employer stock, also raise
trading levels.
Next, we explore the determinants o f investment performance in retirement
accounts. The conventional view is that periodic rebalancing is required for optimal
accumulation patterns. Our evidence confirms that rebalancers in 401(k) accounts
perform better than those fail to rebalance. Passive rebalancers, i.e., who only hold
balanced funds, earn 85 basis points per year more than other nontraders. Active
rebalancers gain about 26 basis points annually compared to other traders. In other words,
rebalancing improves participants investment performance. Evidence from brokerage
vi
accounts suggests that many investors trade too much, a pattern associated with
overconfidence. In the 401(k) environment, we confirm that those with higher turnover
ratios earn 72 basis points less than those with the lowest turnover ratio, consistent with
the overconfidence view.
Finally, we examine how employee portfolio selection behavior changes when
DC plan sponsors introduce life-cycle (LC) funds to the investment menu. Using
difference-in-difference estimation methodology, we find that introducing LC funds does
change asset allocation patterns, particularly for younger and female employees, as well
as for those with shorter tenure, lower household income, and less financial wealth. Also,
introducing LC funds boosts participants equity exposure and improves their expected
risk-adjusted returns. In addition, adding LC funds also reduces idiosyncratic risk for
both participants and plans. Consequently, it would appear that plan sponsors can assist
participants manage their retirement assets better by offering LC funds.
vii
TA B LE O F CO NTEN TS
C h ap ter 1: T he Inattentive P articip an t: Portfolio T rad in g B ehavior in 401(k) P lans
1.1 Related S tudies...............................................................................................................2

1.2 Descriptive Statistics..................................................................................................... 7
1.3 Multivariate Analysis o f Trading Patterns................................................................ 12
1.4 Empirical Findings........................................................................................................14
1.5 Discussion and Conclusions.......................................................................................20
C h ap ter 2: W inners an d Losers: 401(k) T rad in g an d Portfolio P e rfo rm a n c e .................34
2.1 Related S tudies.............................................................................................................36
2.2 Empirical Strategy....................................................................................................... 39
2.3 Hypotheses.................................................................................................................... 43
2.4 Data and Descriptive Statistics...................................................................................45
2.5 Performance R esults.................................................................................................... 48
2.6 Discussion and Im plications....................................................................................... 53
Appendix.............................................................................................................................. 56
C h ap ter 3: How Life Cycle Funds Affect 401(k) Portfolio B ehavior...................................64
3.1 Related Studies and H ypotheses................................................................................ 67
3.2 Empirical Hypotheses Flowing from the Literature................................................ 71
3.3 How Introducing LC Funds Changes Asset Allocation: An Overview................73
3.4 Multivariate R esults..................................................................................................... 76
3.5 Discussion and Conclusions....................................................................................... 89
B ibliography.................................................................................................................................... 103
LIST OF TABLES
Table 1-1. Plan-Level Statistics for Analysis o f 401(k) Plan Trading Behavior.....23
Table 1-2. Participant Statistics for Analysis o f 401(k) Plan Trading Behavior
24
Table 1-3. Summary o f Two-Year Trading Statistics (1/2003-12/2004).................25

Table 1-4. Distribution o f Key Trading Measures (1/2003-12/2004)...................... 26
Table 1-5. Summary o f Predicted Marginal Effects (1/03-12/04).............................27
Appendix Table A l-1 . Multivariate Analysis o f the Probability o f Trading:
Regression Coefficients............................................................................................ 30
Appendix Table A 1-2. Multivariate Analysis o f the Probability o f Being an Active
Trader: Regression Coefficients.............................................................................. 31
Appendix Table A l-3. Multivariate Analysis o f the Number o f Trades: Regression
Coefficients................................................................................................................. 32
Appendix Table A 1-4. Multivariate Analysis o f Portfolio Turnover: Regression
Coefficients................................................................................................................. 33
Table 2-1. Socio-demographic Characteristics o f 401(k) Plan Participants............ 58
Table 2-2. Distribution o f Accounts by Trader T y p e................................................. 59
Table 2-3. Realized 401(k) Account R eturns...............................................................60
Table 2-4. Determinants o f 401(k) Risk-Adjusted Returns........................................61
Table 2-5. Determinants o f 401(k) Risk-Adjusted Returns for Traders O nly.........62
Table 2-6. Characteristics o f Active Rebalancers........................................................ 63
Table 3-1. Hypothesized Impacts o f Offering Life Cycle Funds on Employee
Outcomes..................................................................................................................... 91
Table 3-2. Adoption Rate o f Life Cycle F u n d s.............................................................. 92
ix
Table 3-3. Marginal Effects o f Determinants o f Choosing LC Funds.........................93

Table 3-4. Marginal Effects o f Choosing LC Only........................................................ 95
Table 3-5. Difference-in-Difference Analysis o f Equity Allocation by Age Group. 96
Table 3-6. Impact o f LC on Investment Performance: CAPM Case........................... 97
Table 3-7. Risk Source Contributing to Change in Investment Performance: CAPM
Case, Participant L evel............................................................................................. 98
Table 3-8. Change in Investment Performance From Introducing LC: CAPM Case,
Plan L evel................................................................................................................... 99
LIST OF FIGURES
Figure 1-1. Distribution o f Number o f Trades (1/03-12/04).......................................29
Figure 1-2. Distribution o f Portfolio Turnover (1/03-12/04).......................................29
Figure 3-1. Target Asset Allocation o f Life Cycle Funds.......................................... 100
Figure 3-2. Life Cycle Fund Adoption Patterns by Plan Participants...................... 101
Figure 3-3. Change in Participants Equity Allocation Resulting from Moving to LC
Funds, By A ge...........................................................................................................102
xi
Chapter 1: The Inattentive Participant: Portfolio Trading Behavior in

401(k) Plans
While many studies have explored the determinants o f employee saving behavior
in US defined contribution (DC) saving plans, 1 far less attention has been devoted to
understanding how workers manage their assets in these plans. Nevertheless, plan
sponsors, recordkeepers, money managers, and policymakers would benefit from a
deeper understanding o f how some 60 million employees manage the $2.5 trillion in their
DC pension accounts.2 This paper draws on a rich new dataset o f more than 1,500
retirement plans to analyze trading patterns o f some 1.2 million active participants in
401(k) plans over the 2003-04 period.
We relate a number o f portfolio outcomes to
employee characteristics as well as plan design features and participant investment

choices. Our results suggest that most 401(k) plan participants exhibit high levels o f
inertia: over our two-year period, most people never execute any trades in their pension
portfolios, and even among traders, portfolio turnover rates are one-third the rate o f
professional money managers. In this period o f rapidly rising stock prices, there is no
evidence o f portfolio rebalancing among the vast majority o f participants. The few who
do trade in their 401(k) plans are more affluent, older men, with higher incomes and
longer job tenure. They tend to use the internet for 401 (k) account access, hold a larger
number o f investment options, and are more likely to hold active equity funds rather than
1Mitchell, Utkus, and Yang (2005) provide a review.

2 In the US there are three times as many workers participating in DC plans today compared to defined
benefit (DB) plans and DB assets (at $2.2 trillion) are now less than DC assets (Vanguard, 2004).
index or lifecycle funds. Certain plan design features, notably the presence o f company
stock, raise trading levels, even after controlling for differences in employee
demographic characteristics. In general, most workers tend to buy and hold their pension
portfolios and are inattentive to ongoing portfolio management.
Our findings suggest several questions worthy o f future research. First, while it is
clear that most participants are inattentive to their portfolios, it remains to be seen
whether this inactivity is motivated by lack o f awareness regarding recommended
practices such as rebalancing, or whether it signals interia, implying that despite being
aware o f best practices participants require additional assistance to manage their
portfolios. Second, our analysis underscores the importance o f distinguishing between
alternative measures in evaluating the impacts o f portfolio trading. Factors such as
employee demographics, plan design, or participant holdings, can all have quite different
effects depending on whether the measure is the propensity to trade, the propensity to be
an active trader, the number o f portfolio trades, or portfolio turnover. Third, because
only a tiny minority o f participants trades actively, any efforts to address the costs due to
excessive trading would need to be targeted on the small group o f active participants.
In what follows, we first offer an overview o f related literature, and then we turn
to a high-level description o f trading activity in our dataset. Next we describe our
empirical approach, discuss findings, and outline implications.
1.1 Related Studies

Why might retirement investors alter their portfolios or trade their pension
accounts? Theoretical finance explanations have built on the capital asset pricing
framework, where portfolio choice is believed to reflect investor risk preferences, given
an efficient set o f available investment opportunities. From this perspective, portfolio
shifts are predicted if preferences change or when investors alter their forecasts o f
expected returns and risk (taking into account the transaction costs o f trading). In other
words, the rational investor trades when the marginal benefit o f trading equals or exceeds
his marginal cost (Grossman and Stiglitz, 1980).
Whether these costs and benefits change systematically as investors age is a point
o f some debate. For instance, Samuelsons (1969) theoretical approach argued that
rational investors would hold a fixed equity allocation over their lifetime, regardless o f
age or wealth (and given identically and independently distributed returns over time,
among other assumptions). From this viewpoint, portfolio reallocations or trading would
then be attributable to changing return expectations, rebalancing due to fluctuating asset
prices, or perhaps the reassessment o f manager skill (if the investor employs active
portfolio managers). More recent research suggests that an inverse correlation between
age and human capital risk should lead investors to hold less risky portfolios as they age,
generating age-related trading away from equities.3 This latter view is consistent with
advice offered by many retirement calculators and advisers, who recommend that
investors design their portfolio allocations as a function o f their goals, risk tolerance, and
other factors such as the job security or presence o f some other pension plan. From this
3 Ameriks and Zeldes (2004) provide an excellent summary o f the debate. They also conclude that there is
no evidence o f age-based trading away from equities.
perspective, portfolio trading other than periodic rebalancing would be expected to occur
relatively rarely, and it might be particularly associated with an investors age.4
Empirical studies on investor trading behavior are quite recent, and some o f the
most widely-cited research relies on data from investors with self-directed brokerage
accounts. In one influential paper, Barber and Odean (2000) concluded that active
traders realize substantially lower returns than do nontraders.5 Their sample also turned
over more than 75% o f its common stock portfolio annually. This finding lends support
to overconfidence theory, whereby overly-optimistic investors trade too frequently and
to their detriment, as a result o f a too-rosy estimation o f their own investment skills
(Odean 1999; Gervais and Odean, 2001). Follow-up research, again on discount
brokerage accountholders, reported a raw male/female gap o f 45% in portfolio turnover
(Barber and Odean 2001); the gap diminishes somewhat, to 23%, when controlling for
demographic factors.6 The overall average turnover rate for this sample o f brokerage
investors was quite high, about 6% on a monthly basis (or 72% annually).
On the face o f it, it seems quite unlikely that 401(k) plan participants would
exhibit turnover rates as high as a group o f self-directed brokerage account holders, yet
4 For example, most o f the participants in our dataset received an initial investor questionnaire that, if
completed, recommended a target asset allocation based on the investors time horizon, risk tolerance and
other factors such as job security. Rebalancing was also recommended as an annual strategy. More
generally, Bodie, Kane and Marcus (2002) discuss how investors risk tolerance and ability to recover from
losses declines with age, implying a shift toward conservative assets over time. The Certified Financial
Planner (CFP; Tacchino and Littell, 1999) and the Certified Financial Analyst curricula (CFA; Bronson,
Scanlan and Squires, forthcoming; and Maginn et al. forthcoming) emphasize the importance o f life stage
and time horizon in investors ability to take risk. Writers who link age to equity exposure include Brennan
(2002) and Evensky (1997).
5 Active traders in their study posted an average annual return o f 11.4% versus the average annual return of
16.4% for all households and a average annual return o f 17.9% for the market.
6 This figure is our calculation from reported regression results.
7 Barber, Lee, Liu and Odean (2004) also find that in a study o f market trading from Taiwan, trading is a
zero-sum gain, with profits gained by institutional investors exactly equal to losses incurred by individuals.
relatively few studies have evaluated this issue. In an analysis o f participants in highereducation 403(b) plans, Ameriks and Zeldes (2004) report that almost three-quarters o f
those participants never changed their investment holdings over a 10-year period.
Whether their findings may be generalized to the broader 401(k) universe is not yet
known. In a study o f a single large corporate 401(k) pension, Madrian and Shea (2001)
report that participants who were automatically enrolled did not change their investment
allocations much over time, instead remaining in the conservative cash fund selected by
their employer as the default account. Another study o f a single plan by Agnew,
Balduzzi, and Sunden (2003) again finds strong evidence o f 401(k) participant inertia,
with almost 90% o f plan participants making no trades in a given year. Those authors
also report an average o f 0.26 trades per year (or about one trade every four years), with a
mean annualized portfolio turnover rate o f 16%. Interestingly, men trade 56% more
often than women, and portfolio turnover o f male traders is 53% higher than female
traders. Whether these results can be generalized is again unclear, since that company
had previously permitted participants to invest only in a stable-principal investment
contract fund, so participants held only one-quarter o f their plan money in equities. This
is a low fraction compared to 401(k) plans generally.8
Two studies have explored the role o f internet access on account trading behavior.
Barber and Odean (2002) focus again on discount brokerage investors, and they conclude
that investors who switch to internet trading are also those who trade more frequently,
8 According to the joint Employee Benefit Research Institute/Investment Company Institute (EBRI/ICI)
data base o f defined contribution plan participants (Holden and VanDerhei, 2004), participants held 67% of
their assets in equities as o f December 2003, the month prior to the beginning o f our sample period.
hold more speculative investments, and see their investment performance deteriorate.
Choi, Laibson, and Metrick (2002b) compare trading patterns prior to and post internet
trading access, and they suggest that trading frequency doubles, while turnover rises only
by half when online trading is introduced, particularly among young, male, and wealthier
participants. As we note below, o f course, it is unclear whether investors planning to
trade also choose to adopt internet account access, or whether internet access in and o f
itself induces more trading.
Another factor o f particular interest in 401(k) plans is whether there are the links
between employee trading patterns and plan design decisions made by the employer.
Investment decisions in 401(k) plans are the joint outcome o f employers selection o f
investment offerings, and participants elections among the available options. While past
research has not addressed this issue specifically in the trading context, a handful of
studies have linked plan menu design and participant behaviors o f other sorts. For
instance, Benartzi and Thaler (2001) infer from experimental evidence that menu design
can lead participants to naively diversify their portfolios; Elton, Gruber and Blake (2004),
argue that some 401(k) investment menus prevent participants from constructing efficient
portfolios; and Iyengar, Huberman and Jiang (2004) suggest that participants may suffer
from choice overload, where complex investment menus discourage participation in the
plan. In what follows, we focus not only on trading patterns by investor characteristics,
but also how menu design may shape trading outcomes.9
9 Mitchell, Utkus, and Yang (2005) describe how plan features influence savings behavior.
1.2 Descriptive Statistics

Our dataset consists o f a two-year extract o f 401(k) plans and participants drawn
from the recordkeeping systems o f Vanguard from 2003 to 2004. The same encompasses
1,530 defined contribution retirement plans and includes asset allocation and trading
patterns for nearly 1.2 million active participants in those plans.10
Plan-level statistics. Table 1-1 provides key plan-level descriptive statistics: for instance,
we see that the average plan has 776 active participant accounts with assets o f $38.4
million. These plans offer an average o f 17 investment choices. Almost all plans include
one or more equity index funds; half o f plans offer lifecycle funds;11 15% offer employer
stock (also known as company stock); 93% offer one or more international options; and
only 3% offer a brokerage option.
Nearly three-quarters o f the sample plans permit
participants to take a loan from their own plan assets (up to a legal maximum). The vast
majority o f the plans (90%) permit employee contributions; only a few are completely
employer-financed.
11
Table 1-1 here

10 This file is taken from a larger dataset of more than 2,000 plans and 2.5 million participant accounts. Our
selection criteria identified those individuals who were active participants in their 401(k) plans over the 24month window, and who were in plans in continuous existence over the same period.
11 Lifecycle funds are investment options designed to provide one stop portfolio diversification in a
single fond. They are typically offered as a series within a plan. Static allocation funds offer a range of
funds based on risk characteristics (e.g., conservative, moderate, aggressive), while target maturity funds
are based on an expected retirement date (the 2005 fond, the 2015 fond, the 2025 fond, etc.), with equity
allocations higher for longer-dated funds, and automatic reductions in equity exposure over time.
12 In a 401(k) brokerage option, participants may transfer all or a portion o f their account assets (depending
on plan rules) to a brokerage account within the plan. They may invest in mutual funds, exchange traded
funds, individual stocks and bonds or other securities (again depending on the restrictions imposed by the
plan). They incur retail-level brokerage commissions for their investment transactions.
3 We use the term 401(k) plan interchangeable with defined contribution plan, recognizing that there
are other types o f employee-contributory plans, such as 403(b) plans for non-profits, as well as employeronly contributory plans, including standalone profit-sharing and money purchase plans. Omitted are
standalone Employee Stock Ownership Plans (ESOPs) since by law they are mainly invested in employer
stock.
Participant-level statistics. While the plan-level measures o f our dataset are heavily
skewed toward small firms, the 1.2 million participants in the sample are mainly found in
the larger firms.14 Some 68% o f the participant accounts in our sample are found in the
largest 10% o f plans; 97% o f participant accounts are in the top half o f plans. Since our
trading analysis is conducted at the participant, rather than the plan level, our universe is
more characteristic o f the participant behavior found at medium- and large-sized firms.15
Panel A o f Table 1-2 provides summary statistics at the participant account level.
It shows that the average plan participant has a 401(k) account balance o f more than
$86,000,16 is 44 years old, has been on the job for eight years, and has an average
household income o f just over $88,000. About half the sample is identified as male and
about a quarter female (another quarter o f accounts lack an identifier for sex); the ratio o f
men to women is approximately 2:1 assuming no bias in missing data. Table 1-2 also
indicates the distribution o f imputed non-retirement household financial assets: 32% o f
the participants are classified low wealth, 45% as medium wealth, and 23% as high
wealth.
17
Some 37% o f the participants are registered to access their account via the
internet (as o f January 2003).

Table 1-2 here
14 This is true for the 401(k) world more generallycf. Mitchell et al. (2005).
15 We use the terms participant and participant account interchangeably here, though in practice 4% of
the participants in our sample have accounts with different plans.
16 This average balance is much higher than balances reported by other sources (for instance, Vanguard
reported an average 2002 year-end balance o f just over $45,000 for all its plans). The reason is that our
sample includes only active contributors continuously participating in their employers plans over the twoyear period of interest; hence it excludes small accounts o f job changers and inactive participants.
17 Data from the IXI company are used to impute non-retirement household financial wealth at the ZIP+4
level. The data, which are categorical in nature, are collapsed into three groupings as follows: poor (wealth
< $7,280), middle class (wealth between $7,280 and $61,289), and rich (wealth >$ 61,289).
Panel B o f Table 1-2 reports on 401(k) plan features offered and actually taken up
by sample participants. It is interesting that the average participant has access to nearly
18 options in his plan, but he utilizes only 3.5 investment funds in his investment
portfolio. Earlier, we saw that 15% o f plans offers employer stock, but since these are
the larger firms, over h alf (52%) o f the participants have access to employer stock and
one-third (32%) holds an employer stock investment. Almost all o f participants have
access to equity index funds (99%) and international funds (98%), but only one-half
(53%) and a fifth (20%), respectively, actually invest in these choices. Another notable
finding is that some 85% o f participants have access to a 401(k) loan feature, but only
11% have a loan outstanding.
Trading Patterns. O f particular interest, o f course, is an overview o f 401(k) trading
activity in our dataset. It is worth noting that a substantial portion o f observed asset
movement turns out to be sponsor-initiated rather than participant-initiated. Employers
have responsibility for designing the fund menu offered to participants, and they may
periodically add or delete fund choices in response to changing investment manager
process or performance or other concerns.18 When a new investment option is added, the
sponsor will typically provide information on the new fund and will allow participants to
decide whether or not to invest in it. But when a fund is to be deleted, the sponsor will
typically notify participants and after a certain period will transfer any remaining
holdings in the deleted fund to another fund in the plan. In the latter case, this will
18 Sponsors may terminate money managers not only due to concerns about performance but also due to
changes in the managers investment objectives, style, investment process, staffing or organization. On the
fiduciary front, some sponsors in our sample period terminated certain third-party money managers who
had failed to adequately control market timing within their funds.
generate observed trading which is employer-initiated rather than employee-driven. To

estimate the extent o f such activity we examined asset balances by fund and plan over
time; if the assets in a given fund declined by 98% or more in a given month, we
excluded all trading in the fund in that one month as sponsor-initiated. 19
Evidently, this sort o f plan sponsor behavior accounts for a meaningful volume o f
gross 401(k) trading activity over the 2003-04 period. Table 1-3 defines a trade as a
fund purchase and sale occurring on a given day in a single account (this is because 401(k)
purchases and sales are only priced once per day).20 Here we see that 30% o f accounts
had at least one trade over the period, with one-third o f the trades classified as sponsorinitiated. Similarly, the mean number o f trades over the two-year period is 0.76, but
adjusting for sponsor-initiated trading, the average falls to 0.60. Accordingly, about 20%
o f all trades are therefore sponsor-related (=0.16/0.76). In the remainder o f this paper, we
focus attention only on participant-initiated trading activity.
Table 1-3 here
After these adjustments for sponsor-related trading, four-fifths o f the accounts
experience no participant-directed trades at all over a two-year period. This lack o f
activity is reflected in a variety o f trading statistics. For all accounts, including traders
and non-traders, the mean number o f trades is 0.6 per account over two years. Portfolio
19 This is likely a lower-bound estimate o f sponsor-initiated trading, since participants may have several
months notice of plans for fund deletions, and some may trade after the actual mapping o f funds.
20 Participants can also alter their contribution allocations or way in which future contributions are to be
invested, but this is not our focus in the current paper.
10
turnover, defined as the average amount traded divided by the average balance,21 is 18%
over the two-year period. For traders these measures are higher: the mean number o f
trades for accounts with trading is 3 over two years; the mean two-year turnover is 90%)
Medians are dramatically lower and underscore the skewed distribution o f trading
activity. For the entire sample, the median number o f trades and turnover rate over two
years is zero; for those trading, the median number o f trades is 1 and median two-year
turnover is 48%. We also define active traders as the subset o f participants having 6+
trades over two years (representing the top 2% o f accounts). For this group portfolio
activity is far greater, as the mean number o f trades is 13 (9 median) and the mean
portfolio turnover rate is 347% (median 162%) over the period.
A summary distribution o f the number o f trades across all accounts and for active
traders appears in Table 1-4, and histograms are provided in Figures 1-1 and 1-2. What is
most striking is the very low level o f trading. Not only do very few people trade at all,
but even those who do trade are fairly inactive. A second striking feature is the fat
tailed phenomenon o f trading, with the percentage o f participants falling dramatically as
the level o f trading increases.
Table 1-4 and Figures 1-1 and 1-2 here
It is interesting to compare 401(k) trading patterns with trading outcomes reported
in other studies. For instance, Reid and Millar (2004) estimate a mean portfolio turnover
rate o f 117% (with a median o f 65%) for US professional equity mutual fund managers
21 In practice, we first calculate the dollar amount for each trade as the average o f (positive) purchase and
sale amounts; next we sum up this dollar amount o f all trades as the total trade amount; then we divide the
total trade amount the average o f the beginning and ending balance for the two-year period.
11
as a whole. This, o f course, is far higher than our sample, since 80% o f our sample
executes zero trades, and o f the 20% who do trade, the median annualized turnover rate
during the two-year period is 24% or one-third the turnover rate o f professional equity
fund managers.22 Even our most active 401(k) traders (the top 2%) have an annualized
turnover rate o f only 162%, equivalent to about the top half o f equity fund managers in
terms o f turnover. Ultimately, for our sample, there is very little evidence o f rebalancing
or other trading activity: that is, it appears that most 401(k) plan participants are
characterized by profound inertia, tending to buy and hold.
1.3 Multivariate Analysis of Trading Patterns

The purpose o f the multivariate empirical analysis is to explore the role o f three
sets o f factors as influences on 401(k) plan trading activity. First, we seek to test whether
and how employee characteristics influence pension trading. For instance, we wish to
determine whether, as noted in previous studies, men tend to exhibit disproportionate
portfolio churning as compared to women. Second, we hypothesize that plan features
will influence trading patterns. Since the dataset includes an exceptional variety o f plan
designs, we should be able to disentangle the effects o f plan design versus employee
demographics on trading. Finally, we evaluate how plan trading patterns are related to
actual investment holdings. For example, investors who hold passive index equity funds
may also be those less likely to trade subsequently, as compared to people who invest in
22 We recognize that comparing plan participants to fund managers is to a large extent an apples-to-oranges
comparison. Participants are not full-time money managers. Moreover, the average 401(k) investor holds a
balanced portfolio o f both equity and fixed income securities, and hence he or she manages a less risky
portfolio than the average US equity fund manager.
12
actively managed equity funds perhaps because they tend to be believers in capital
market efficiency. Conversely, people who invest in their own employers stock may
trade more if they believe they can outperform market indexes or if they have inside
information (Mitchell and Utkus, 2004).
For the ith participant account in the jth plan, we relate four participant trading
measures (summarized here as the TRADINGtj vector) to a set o f employee demographic
characteristics (DEMOij), plan design features offered to plan participants (JPDj), and (in
a subset o f cases) measures o f participants own account holdings prior to the beginning
o f the trading analysis (ACCTjj):
TRADING,j =/ ? 0 + A - DEMOtJ + fi7PDJ +P,-ACCTiJ + etJ .
So as to compare our results with prior studies, model A is estimated with demographics
only, model B adds plan design factors, and model C adds account holdings.
The four
dependent variables include: TRADER: a dummy (1/0) variable indicating whether the
participant account included a trade or not over the 2003-04 period; ACTIVE TRADER: a
dummy (1/0) variable indicating whether the participant account included six or more
trades over that same period; NTRADES: the total number o f trades the participant had
over the period; and TURNOVER: the participants two-year turnover rate (analyzed for
both non-traders and traders alike). Since the first two dependent variables are (0/1)
23 All regression models also include industry controls. The three largest sectors include manufacturing
with 31% o f the sample; business, professional and non-profit services account with 22%; and finance, real
estate and insurance at 10%.
13
indicators, they are estimated as Probit models.24 NTRADES is estimated using a negative
The fourth dependent variable, TURNOVER, is estimated as
binomial count regression.
a censored Tobit regression.26 The TRADER, NTRADES and TURNOVER regressions

are estimated for the entire sample o f traders and non-traders; the ACTIVE TRADER
regression, for the active population executing six or more trades over the two-year
period.
1.4 Empirical Findings

Table 1-5 provides estimated marginal effects for the key and statistically
significant variables o f interest. 27 Coefficient estimates for all regression models appear
in Appendix Tables A l.1-4.28
24For instance the equation for model C is E{l'RADlNGj y. = 11DEMO, PD, AC C T) =

= Pr(TRADING, j = 11DEMO,PD,ACCT)= (]3lDEMOi J + (32PDk + pÂ C C T^) where d> is
the cumulative distribution function o f standard normal distribution.
25 The negative binomial model is a Poisson model whose parameter is drawn from a Gamma distribution;
the closed-form expression o f the distribution is expressed as:
Pr [NTRADES'tJ = NTRADEStJ \ D EM O^^PD ÂCCT^)
T(9 +NTRADES ik)
1 YY
T(NTRADES,j + 1)T(6)11+ a
parameters 0 and u, a
NTRADES,
where T is a Gamma function with
[ l + cr
, E{NTRADESi j \ DEMOt J, PDj , ACCTt j )= 9 a
A , , and
Var(NTRADESi J \ DEMO, , , PD}, ACCTU ) = 9a(I + a) = Xu fl + - Xu .
v e ,J)
26 The first three models use error correction to adjust for plan-level heteroskedasticity. For the
TURNOVER model, a two-stage Heckman selection model does not produce a statistically significant
coefficient on the inverse Mills ratio for the second-stage regression. A censored Tobit model with error
correction fails to converge, probably due to the enormous size o f the dataset.
27 Marginal effects for continuous variables use either one standard deviation change in the dependent
variable (e.g. 7 more years o f tenure, $60,000 more household income) or an intuitively appealing change
in the unit o f analysis (e.g. 10 more years o f age). To evaluate changes in the account holdings, we set the
corresponding plan design variables to 1 (e.g. to assess the impact o f more company stock holdings, we set
the company stock offered variable to 1). We exclude from Table 1-5 two statistically significant plan
14
Across the board, the most robust finding across all model specifications is the
coefficient on the participants sex: all else constant, men are much more likely to be
traders, to be active traders, to execute more trades, and to have higher portfolio turnover
rates than women. For instance, the probability that a male will trade over a two-year
period is 24% versus 17% for an otherwise similar female participant, a relative
difference o f 40%. Men are also predicted to execute 91% more trades than similar
female participants, and they chum their portfolios at a rate 41-55% higher than women.
It is worth noting that the interpretation o f this effect also depends on how the results are
framed. For example, it is also correct to conclude that 76% o f men are non-traders
versus 83% o f women, a non-trading differential o f only 7%. Similarly, while men are
more likely to be active traders, this is a small group in practical terms. The quantitative
differences in turnover rates by sex are also small. Thus while we confirm the Barber and
Odean (2001) view that boys will be boys, perhaps the more salient observation is that
most men and women do not trade in their 401(k) plans in the first place.
Table 1-5 here
Another important result has to do with the influence o f other financial wealth on
401(k) trading (controlling on other factors, including income). Table 1-5 shows that
higher-wealth participants are more likely to trade, to be active traders, and to execute
design variables, the offering o f index funds and international funds, since virtually all participants are
offered such options, and few plan sponsors are likely to eliminate them. Hence the variation in our data is
likely due to some idiosyncratic behavior associated with handful o f plans lacking these choices.
28 As shown in the Appendix Tables, it is worth noting at the outset that goodness-of-fit measures improve
dramatically as the specifications are made more elaborate. For example, in the TRADER equation, the
pseudo-R2 o f the demographics-only model (A) is 3.4% which rises to 11.6% for model C. Similarly the
pseudo-R2 for NTRADES increases from 9.1% (model A) to 33% (for model C). The TURNOVER model
has the lowest R2, ranging from 1.7% to 5.6%.
15
more trades, than low-wealth participants, a conclusion that could indicate that better-off
households have more experience, knowledge, or level o f engagement with financial
matters generally, which then spills over into the 401(k) plan arena.29 Turnover rates are
substantially higher for high net worth households.
The other employee-side factors we control on include participant age, income,
and plan tenure, and these also have some relationship to trading incidence but have a
smaller impact than sex and wealth. For example, at the margin, being 10 years older
(versus a mean age o f 44) is associated with only a 5-12% increase in the probability o f
trading, and a 6-14% increase in portfolio turnover. Aging, thus, is associated with higher
levels o f portfolio attentiveness although again this is in the context o f the vast majority
o f younger and older participants not trading in the first place. In addition, whether
higher trading later in life is due to systematic age-based selling o f equities or to other
types o f trading, remains to be seen. Changes in household income o f one standard
deviation ($60,000 in household income) and job tenure (7 years) are associated with
similar single to double-digit relative differences on various trading measures, including
turnover rates. O f three important demographic characteristics age, tenure and
income tenure appears to have the stronger relative impact on all measures o f trading.
It has been said, in the context o f 401(k) savings, that stayers are savers (Even and
Macpherson, 2004). A t the margin, our results suggest that stayers are also traders
again, in the context that the overwhelming majority o f long-tenured participants do not
trade in the first place.
29 Bemheim (1998) noted a spillover effect in the other direction: workplace education programs promoted
not only 401(k) saving but also non-plan saving in the household.
16
Turning now to the impact o f plan design on trading patterns, we see that perhaps
the most significant factor influencing trading is the presence o f company stock in the
plan investment menu. While this is associated with statistically higher trading
probabilities, more trades and higher turnover, the empirical magnitudes are relatively
modest: for instance, in a plan lacking employer stock, the probability that a participant
will trade is 19% over two years, versus 21% if company stock is offered. O f course this
is a relative difference o f 13%, but the absolute magnitudes are close in practical terms.
Turnover is higher due to company stock as well by about 12% in one specification. In
other words, participants with access to company stock are, at the margin, more likely to
chum their portfolios at a somewhat higher rate. This finding is interesting in light o f our
conversations with plan sponsors and recordkeepers who note that employer stock is
often associated with active 401(k) trading. Our findings suggest two motivations for this
higher trading: plans offering company stock may have workforce characteristics that
contribute to higher trading levels generally (older, higher income, longer tenure, more
male), and also employer stock appears to have its own distinct influence on trading
activity independent o f these characteristics.
Turning to other plan design features, we find that increasing the number o f funds
offered by the plan does boost the probability o f having active traders, but it has
contradictory effects on turnover depending on the specification. Our tentative
conclusion is that the number o f funds does not appear to influence aggregate trading
levels. Offering o f a brokerage option within the 401(k) plan has a large impact on
trading activity and turnover rates, though the impact in practical terms is still small since
only 3% o f participants are currently offered such an option. Also, in the case o f the
17
brokerage option, we measure trading only in the non-brokerage component o f the 401(k)
account. Part o f this higher trading may be due to greater movement among the regular
fund options in the account; another part is likely due to the movement o f money from
these regular fund options to the brokerage feature.
OA
As noted above, Model C findings also include controls for participants account
holdings, and it stands to reason that participants initial account status (e.g. their equity
mix, whether they are registered for internet trading) will be correlated with subsequent
trading outcomes. In fact, several interesting patterns emerge. For instance, individuals
who registered for internet access to their accounts are three times more likely to be
traders and nine times more likely to be active traders; they also execute five times as
many trades. Turnover rates also differ markedly: non-web registered participants are
predicted to have a 13% turnover rate versus a 48% turnover rate for web-registered
participants, a 251% relative difference. In other words, 401(k) participants who are
internet users have higher turnover rates and they use the web to engage in smaller, more
frequent trades. That said, there remains the unsolved question regarding causality, as to
whether participants who trade more gravitate to internet trading, or whether making
internet trading available itself provoke more trading.
The other results indicate that participants who initially have their money
allocated across larger numbers o f funds are more likely to trade. Participants who own
company stock appear to have higher turnover, but holdings o f company stock per se do
30 The brokerage account can be thought o f as a sidecar. Participant and employer contributions are first
made to the regular investment options offered by the plan; participants wanting to make a brokerage trade
must then transfer these assets from the regular fund options to the brokerage, which counts for a portion of
the trading volume.
18
not contribute to higher overall trading rates or larger numbers o f trade. Overall, the
employees decision to own company stock (once it is offered) appears to have a smaller
effect on trading, than does the sponsors initial choice o f company stock for the plan
investment menu. Those with brokerage accounts are more frequent traders and have
higher turnover levels. (They are also less than 1% o f all participants and so this effect is
not broadly economically meaningful.) Meanwhile, those who initially hold index or
lifecycle funds are subsequently less likely to be traders and have lower turnover rates.
We also note that, in this sample, participants who hold international funds actually were
less likely to trade, suggesting that at least most o f these participants were not attempting
to engage in international arbitrage trading. Finally, those who had taken out a loan from
their accounts are less likely to trade, and turnover rates in aggregate are lower too.
Comparison with prior studies. Compared to the single 401(k) plan examined by Agnew
et al. (2003) during the mid-1990s, we report a surprisingly similar level o f the incidence
o f trading despite differences in sample size and time period: 21% o f our accounts had at
least one trade over two years (10% annualized), versus approximately 12% o f their
accounts with at least one trade per year. Our mean number o f trades (0.6 over two years,
0.3 annualized) is very similar to the earlier papers mean number o f trades o f 0.26. Yet
our portfolio turnover rates (18% over two years, 9% annualized) are half the 16%
annualized for their single 401(k) plan. Not surprisingly, all 401(k) results, both ours and
those o f Agnew et al. (2003), pale in comparison to the 72% annualized turnover rate for
the discount brokerage account holders o f Barber and Odean (2001).
Like the two prior studies, we also find a pronounced effect o f the participants
sex on trading, but there is still considerable cross-study variation in magnitudes. In our
19
demographics-only models, designed to mirror the empirical models used in these prior
studies, men are 40% more likely to be traders, they execute 91% more trades, and their
portfolio turnover is 55% higher than for women. By comparison, Agnew et al. (2003)
find a lower sex-related incidence o f number o f trades, with single males executing 30%
more trades than single females (versus our 91%), yet a comparable rate o f turnover, with
male turnover 50% higher than female turnover (versus our 55%). Barber and Odean
(2001) report that men trade 45% more than women, when measured in terms o f
portfolio turnover, although as noted earlier, this refers to sample means differences
while the marginal effects are much smaller. It would appear, strikingly, that sex matters
more for 401(k) plans than self-directed brokerage accounts, perhaps because o f the self
selection inherent in brokerage account investing. Unlike the other studies, our estimated
effects due to sex are meaningful despite controls for non-financial wealth. Like Agnew
et al. (2003), we find that age, job tenure, and income influence trading, with the tenure
effect particularly pronounced. These effects are generally smaller than for sex and other
wealth.
1.5 Discussion and Conclusions

Based on our analysis o f who trades in 401(k) plans, we conclude that, at least
over our two-year period o f analysis, participants are generally inattentive in their
oversight o f retirement plan assets. Four o f five accounts execute no participant-driven
trades, even though the stock market (as measured by the Standard & Poors 500 Index)
rose by a cumulative 43% over the study period. In other words, for the overwhelming
majority o f retirement savers, there is no evidence o f portfolio rebalancing, shifts in risk
20
tolerance with age, or tactical portfolio changes. It remains to be seen whether such
portfolio inertia serves participants well given the investment choices they have available
to them.
The few people who are traders are likely to be older affluent men, with higher
incomes and longer job tenure; they use the internet to access their 401(k) plan accounts;
hold more funds in their portfolios, and at the margin they invest in active equity funds
while steering clear o f lifecycle and equity index funds. While some o f the measured
differences in behavior confirm those in prior studies the male/female difference in
propensity to trade is positive, statistically significant and, in our case, 40% in relative
terms it is crucial to emphasize the low base: fewer than one-quarter o f the men trade in
a two-year period versus 17% o f the women. Employer plan design features such as
offering company stock or a brokerage option do influence some trading outcomes
including aggregate portfolio turnover rates.
Our analysis will be extended in future work in at least two directions. First, the
panel we use here covers only the 2003-04 period. This was an exceptionally salubrious
period for stock market investing, with US stock prices gaining more than 40%. In the
future, we will expand the panel by including new periods and by reconstituting
information for the 2000-02 bear market. Second, this analysis offers a cross-sectional
view o f the panel, modeling trading over this period from a variety o f perspectives. We
will take on a more detailed time-series approach as the panel expands over time.
In conclusion, we offer thoughts on the implications o f our results for sponsors,
fund managers, and policymakers. One interpretation is that the portfolio inertia
identified here suggests that participants may require additional help managing their
21
portfolios. Automatic rebalancing services, lifecycle funds, and managed accounts can
be useful in ensuring that sensible portfolio management takes place on a disciplined
schedule whether in 401(k) plans, public sector DC pensions, or even in a reformed
Social Security system with private accounts. But any these programs necessarily will
raise aggregate turnover rates, given that for most participants, current turnover rates are
already zero.
Another consideration is that any assessment o f trading at the plan level must
account for the critical influence o f plan design as well as workforce demographics.
Certain employee populations (older, more male participants, longer-tenured, etc.) will
be likely to trade more simply because o f their demographic characteristics. Trading
levels are also higher under specific plan design circumstances, notably when the
employers stock is offered, or when a 401(k) brokerage option is provided (currently
offered to only a small number o f participants). Finally, employee preferences for certain
assets (e.g., index funds versus brokerage accounts) and account features (web
registration) will also influence trading outcomes. Our analysis also shows that it is
typically the demographic and account holdings, rather than plan design p e r se, that
appear to have the strongest effects on trading. A final point to emphasize is that only a
small group o f participants is ever involved in active trading. This set o f active traders
raises transaction costs for all participants, and their activities may be disruptive to
portfolio managers. Accordingly, those seeking to reduce active trading in 401(k) plans
may seek to target remedial policies on this specific sub-set o f investors.
22
Table 1-1. Plan-Level Statistics for Analysis of 401(k) Plan Trading Behavior
Variable name
M ean V a lu e
Plan size
Number of active participant accounts

Plan assets ($ millions)
Plan assets (In)
num _ p r t
pla n _a s s e t s
ln pla n _ a s s e t s
776
$38.4
17.5
Plan design fea tu re s offered
Number of funds
Index equity funds
Lifecycle funds
Company stock
International funds
Brokerage option
Loan
Employee contributions
NFUNDS
EQUITY_IND_FLG
LC_FND_FLG
CS_FLG
INTER_FLG
VBO_FLG
LOAN_FLG
EECONTRIB
16.6
98.8%
48.8%
15.0%
93.1%
3.1%
74.3%
91.8%
Note:
Number of 401 (k) plans in sample= 1,530
23
Table 1-2. Participant Statistics for Analysis of 401(k) Plan Trading Behavior
Panel A. Participant Characteristics
Variable name
M ean
D em ographics
Age
Household income
Household income (In)
Plan tenure (years)
Sex
Male
Female
Missing
Non-retirement financial wealth
Low wealth
Middle wealth
High wealth
43.5
88,003
11.4
age
h h jn c
l n h h jn c
8.0
tenure
48.0%
26.2%
25.8%
male
fem a le
m alem s
32%
45%
23%
poor
middle
ric h
401 (k) A cc o u n t Features
Account balance
Web registered
Equity allocation
86,363
37%
BLN_PRT
WEB
66 . 2 %
Panel B. Plan Features Offered versus Features Held (Account-Level)

V ariable
nam e
Plan features offered
Number of funds
Index equity funds
Lifecycle funds
Company stock
International funds
Brokerage option
Loan
V ariable
nam e
Mean
Mean
Account holdings (12/02)

NFUNDS
INDEX
LC
CS
INTL
BROK
LOAN
EECONTRIB
17.7
99%
47%
52%
98%
5%
85%
94%
Number of funds
Index equity funds
Lifecycle funds
Company stock
International funds
Brokerage option
Loan
NFUNDS_HELD
INDEX_HELD
LC_HELD
CS_HELD
INTLJHELD
BROK_HELD
LOAN_HELD
EECONTRIB
Note:
Number of plan participants = 1,186,554

See text for definition of financial wealth.
24
3.5
53%
12%
32%
20%
0.1%
11%
88%
Table 1-3. Summary of Two-Year Trading Statistics (1/2003-12/2004)

Mean (median) trading patterns
Accounts
with any
trading
All trading activity
Less: sponsor-initiated
Participant trading
367,283
(123.8851
243,398
For all accounts*

% of
Accounts
with
Trading
30.1%
-9.6%
20.5%
NTRADES
TURNOVER
For traders only*
NTRADES
TURNOVER
For active traders only*
NTRADES
TURNOVER
0.76(0.00) 23.7% (0.00%)
2.44(1.00) 76.5% (42.0%)
12.68(9.00) 340.0% (160.4%)
0.60(0.00)
2.92(1.00) 89.7% (47.7%)
12.86(9.00) 346.9% (162.4%)
18.4% (0.00%)
* Note: Traders are those having 1 or more trades over the two-year period; active traders are those with six or more trades over two years. NTRADES
are the number of trades in a two-year period for traders and non-traders. TURNOVER is the percent portfolio turnover rate for both traders and non-traders.
Note:
25
Table 1-4. Distribution of Key Trading Measures (1/2003-12/2004)

Panel A. Distribution of Number of Trades
NTRADES
0
1
2-5
6-50
Over 50
Percent
79.5%
10.9%
7.4%
2.2%
0.04%
100.00%
Number of
accounts
943,156
129,504
87,864
25,585
445
1,186,554
Panel B. Distribution of Participant Turnover

TURNOVER
0%
0-50%
50-100%
100-200%
200%-500%
Over 500%
Percent
79.5%
10.6%
5.3%
3.1%
1.1%
0.4%
99.6%
Number of
accounts
943,192
125,775
63,006
36,783
12,933
4.865
1,186,554
Note:
Reproduced with permission of the copyright owner. Further reproduction prohibited without perm ission.
Table 1-5. Summary of Predicted Marginal Effects (1/03-12/04)

DEMOGRAPHICS
Sex
Male
Female
% change
TRADER
B
A
C
23.6% 23.4% 19.4%
16.9% 16.8% 14.7%
I 40%
40%
32% |
ACTIVE TRADER
NTRADES
TURNOVER
A
B
C
A
B
C
A
C
B
2.6%
2.7%
1.3%
0.694 0.685
0.448
33.3%
33.3% 26.1%
1.1%
1.1% 0.5%
18.5%
21.5%
21.6%
0.363 0.362
0.275
149% 147% 146% | I 91 %
89%
63% | I 55%
54%
41% |
W ealth
High wealth
Low wealth
% change
25.0% 24.6% 20.1%

16.7% 16.7% 15.2%
I 50%
47%
32% I
2.5%
1.5%
71%
2.5%
1.5%
62%
1.1%
0.678 0.660
0.8%
0.427 0.427
33% | I 59%
55%
0.437
34.1%
0.326
22.4%
34% | I 52%
33.7%
22.6%
49%
26.2%
19.9%
32% |
Age
10 years older
Predicted Mean
% change
21.3% 21.4%
20.4% 20.2%
I
5%
6%
19.5%
17.4%
12% |
2.1%
1.9%
13%
2.2%
2.0%
15%
1.2%
0.587 0.590
1.0%
0.536 0.530
30% | I 10%
11%
0.437
29.4%
0.375
27.6%
17% | I
6%
29.7%
27.6%
8%
25.9%
22.8%
14% |
Income
$60K increase
Predicted Mean
% change
22.2% 22.0%
20.4% 20.2%
I
9%
9%
18.2%
17.4%
5% |
2.1%
1.9%
12%
2.2%
2.0%
11%
1.0%
0.590 0.583
1.0%
0.536 0.530
4% | I 10%
10%
0.393
30.3%
0.375
27.6%
5% | I 10%
30.2%
27.6%
10%
24.0%
22.8%
5% |
Tenure
7 year increase
Predicted Mean
% change
23.3% 23.0% 19.1%

20.4% 20.2% 17.4%
I 14%
14%
10% |
2.5%
1.9%
32%
2.5%
2.0%
27%
1.2%
0.663 0.645
1.0%
0.536 0.530
22% | I 24%
22%
0.433
32.2%
0.375
27.6%
15% | I 17%
32.0%
27.6%
16%
25.4%
22.8%
11 %|
29.1%
26.0%
12%
[
23.1%
22.5%
3% I
29.6%
27.6%
7%
21.5%
22.8%
-6% |
35.3%
27.3%
29%
[
27.0%
22.6%
19% |
PLAN DESIGN
Co stock Offered
Not offered
% change
21.4%
19.0%
13% |
No. funds 10 add! funds

offered
Predicted Mean
% change
Brokerage Offered
Not offered
% change
2.2%
1.7%
I 29% I
0.588
0.473
24% I
2.5%
2.0%
I 28% |
25.9%
20.0%
29% |
2.9%
1.9%
[ 52% |
0.759
0.521
46% |
EE cont.
Allowed
ER only
% change
28.2%
19.4%
46%
[
23.3%
16.5%
41% |
Loan
Offered
Not offered
% change
27.0%
30.9%
I -12%
22.3%
25.7%
-13% I
27
Table 1-5 (contd). Summary of Predicted Marginal Effects

ACTIVE TRADER
B
TRADER
B
TURNOVER
B
NTRADES
B
ACCOUNT HOLDINGS
Web
Registered
Not registered
% change
No. funds 5 add'l funds

used
Predicted Mean
% change
32.1%
11.1%
189%|
27.4%
17.4%
58% |
3.7%
0.4%
825% I
1.001
0.209
379% |
48.1%
13.7%
251%|
2.5%
0.666
1 .0 %
0.375
78% |
31.0%
22 .8 %
36% I
155%|
Company Held
stock
Not held
% change
24.7%
22 . 0 %
12%
Brokerage Held
Not held
% change
46.7%
17.3%
169%|
Index
Held
Not held
% change
16.4%
18.2%
Held
Not held
% change
15.7%
17.8%
- 12 % |
0.7%
-36% I
Held
Not held
% change
15.7%
17.7%
- 11 %|
0.7%
0.310
0.387
I -20% |
Held
Not held
% change
15.2%
17.6%
-14%[
Inti
Lifecycle
Loan
6.9%
1 .0 %
616%|
1.176
0.375
I 214%|
52.4%
22 . 8 %
130%|
0.336
0.421
20.9%
24.7%
-15%1
0 .8 %
1.2 %
- 10%|
-20% |
0.317
0.392
1 .1 %
1. 0 %
-35% I
-19% I
20.5%
23.4%
- 12%|
20 . 1%
23.3%
20.9%
22.9%
-9% |
Note:

Column A = demographics only; B = demographics and plan design; C = demographics, plan design and 12/02 account holdings.
28
Figure 1-1. Distribution of Number of Trades (1/03-12/04)
100%
oo
a
*->
c
na
'o
n
a
79.5%
80%
60%
40%
ea> 20%
aA 0%
-10.9%
7.4%
a.
2 .2 %
0.04%
6-50
Over 50
2-5
Number of trad e s over a 2-year period
Note:
Number o f plan participants = 1,186,554
Figure 1-2. Distribution of Portfolio Turnover (1/03-12/04)

= 100%
o
o
o
A
79.5%
80%
60%
Q.
40%
20 %
A
eA
0.
10:6 % -
5.3%
3.1%
1 . 1%
0.4%
200%500%
Over
500%
0%
0%
0-50%
50-100% 100-200%
Turnover rates over a 2-year period
Note:
29
Appendix Table Al-1. Multivariate Analysis of the Probability of Trading:

Regression Coefficients
Probit model specification
(A) - Demographics only

(B) - Demographics and plan design features offered
(C) -- Demographics, plan design and account holdings on Dec. 2002
TRADER (mean: 20.5%)
Mean
Demographics
AGE
TENURE
MALE
HH INC (In)
MIDDLE
RICH
(A)
0.003
0.014
0.242
0.118
0.163
0.291
43.5
8.0
47.5%
88.0
45%
23%
***
***
***
***
***
***
(B)
0.004
0.014
0.239
0.116
0.157
0.279
***
***
***
***
***
***
(C)
0.008
0.009
0.187
0.065
0.099
0.188
***
***
***
***
***
***
Plan design features offered
NFUNDS
NFUNDS2
INDEX
INTL
CS
BROK
LC
LOAN
EECONTRIB
PLANASSETS
17.7
495.1
99%
98%
52%
5%
47%
85%
94%
414,818
0.009
0.000
-1.079
0.186
0.086
0.194
-0.027
-0.091
0.247
0.004
*
**
**
***
*
-0.004
0.000
-1.15 *
0.182 **
0.018
0.135 *
0.023
-0.089
0.201
-0.011
Account holdings (Dec 2002)
37%
WEB
NFUNDS HELD
3.5
INDEX_HELD
53%
INTL HELD
20%
CS_HELD
32%
BROKJHELD
0%
LC HELD
12%
LOAN HELD
11%
Observations
1,186,554
1,186,554
-log(L)
581,429
577,103
Pseudo R-squared
3.4%
4.2%
8,652 ***
Chi-square
All models are Probits.
***: significant at the 1 % confidence level
**: significant at the 5% confidence level
*: significant at the 10% confidence level
Sector variables and missing dummy variables included but not reported.
30
0.756
0.068
-0.063
-0.091
0.060
0.730
-0.091
-0.083
1,186,554
532,166
11.6%
89,874
***
***
***
***
*
***
***
***
***
Appendix Table Al-2. Multivariate Analysis of the Probability of Being an Active

Trader: Regression Coefficients
Probit model specification
(A) -- Demographics only

(C) - Demographics, plan design and account holdings on Dec. 2002
ACTIVE TRADER (mean: 2.2%)
(A)
(B)
(C)
Mean
Demographics
AGE
TENURE
MALE
HH INC (In)
MIDDLE
RICH
43.5
8.0
47.5%
88.0
45%
23%
0.005
0.017
0.368
0.089
0.122
0.219
***
***
***
***
***
***
0.006
0.015
0.367
0.085
0.113
0.198
***
***
***
***
***
***
0.010
0.011
0.331
0.027
0.053
0.105
***
***
***
***
***
***
NFUNDS
NFUNDS2
INDEX
INTL
CS
BROK
LC
LOAN
EECONTRIB
PLANASSETS
17.7
495.1
99%
98%
52%
5%
47%
85%
94%
414,818
0.013 **
0.000
-0.13
-0.008
0.105 ***
0.178 **
-0.028
-0.069
0.180
0.021
-0.001
0.000
-0.15
0.002
0.051
0.138 *
0.042
-0.101
0.147
0.005
37%
WEB
NFUNDS HELD
3.5
INDEXJHELD
53%
INTL_HELD
20%
CS HELD
32%
BROK HELD
0%
LC HELD
12%
LOAN HELD
11%
1,186,554
Observations
1,186,554
118,962
117,942
-log(L)
Pseudo R-squared
5.0%
5.8%
Chi-square
2,040 ***
All models are Probits.
***: significant at the 1% confidence level
31
0.890
0.075
-0.179
-0.165
0.026
0.725
-0.180
0.009
1,186,554
105,206
15.9%
25,473
***
***
***
***
***
***
***
Appendix Table Al-3. Multivariate Analysis of the Number of Trades: Regression

Coefficients
Negative binomial model specification
(A) - Demographics only

Mean
Demographics
AGE
TENURE
MALE
HH INC (In)
MIDDLE
RICH
(A)
0.009
0.031
0.648
0.185
0.273
0.462
43.5
8.0
47.5%
88.0
45%
23%
NTRADES (m ean: 0.6)

(B)
***
0.011 ***
0.028 ***
***
***
0.637 ***
0.182 ***
***
0.262 ***
***
***
0.436 ***
(C)
0.015
0.020
0.488
0.091
0.159
0.293
***
***
***
***
***
***
NFUNDS
NFUNDS2
INDEX
INTL
CS
BROK
LC
LOAN
EECONTRIB
PLANASSETS
17.7
495.1
99%
98%
52%
5%
47%
85%
94%
414,818
0.019
0.000
-0.88
0.264
0.217
0.376
-0.014
-0.165
0.489
0.021
**
*
***
***
-0.006
0.000
-1.19
0.243
0.098
0.237
0.090
-0.187
0.410
-0.001
***
**
*
**
37%
WEB
3.5
NFUNDS HELD
INDEX HELD
53%
INTLJHELD
20%
CS HELD
32%
BROK HELD
0%
LC HELD
12%
11%
LOAN HELD
1,186,554
1,186,554
Observations
194,814
-log(L)
198,002
10.5%
Pseudo R-squared
9.1%
6,376 ***
Chi-square
Ail models are negative binomial regressions.
*: significant a t the 10% confidence level
32
1.567
0.115
-0.218
-0.215
0.084
0.918
-0.269
-0.018
1,186,554
145,927
33.0%
97,773
***
***
***
***
***
***
***
Appendix Table Al-4. Multivariate Analysis of Portfolio Turnover: Regression

Coefficients
Censored Tobit specification
(A) -- Demographics only

_______ TURNOVER (m ean: 18.4%)
(A)
(B)
(C)
0.012 ***
0.010 ***
0.020 ***
***
0.034
0.035 ***
0.023 k k k
0.687 ***
0.681 * * *
0.514 * k k
0.279 k k k
0.151 * * *
0.283 ***
kkk
0.393 ***
0.374
0.233 k k k
***
0.668 ***
0.634
0.409 * * *
Mean
Demographics
AGE
TENURE
MALE
HH INC (In)
MIDDLE
RICH
43.5
8.0
47.5%
88.0
45%
23%
NFUNDS
NFUNDS2
INDEX
INTL
CS
BROK
LC
LOAN
EECONTRIB
PLANASSETS
0.017 k k k
0.000 * * *
-2.512 k k k
0.480 ***
0.182 ***
0.417 ***
-0.012
-0.210 * * *
0.574 k k k
0.025 * * *
17.7
495.1
99%
98%
52%
5%
47%
85%
94%
414,818
-0.011
0.000
-2.651
0.471
0.040
0.269
0.113
-0.214
0.502
-0.013
kkk
kkk
kkk
***
***
kick
kkk
kkk
*k*
kkk
37%
WEB
NFUNDS HELD
3.5
53%
INDEXJHELD
20%
INTL HELD
32%
CS HELD
BROKJHELD
0%
12%
LC HELD
LOAN HELD
11%
1,186,554
1,186,554
Observations
916,762
913,795
-log(L)
2.1%
1.7%
Pseudo R2
5,933
Likelihood ratio te st1
All models are censored Tobit regressions.
1: Likelihood test is to com pare the models: B with A, C with B.
kkk
33
1.931
0.095
-0.233
-0.205
0.155
1.106
-0.281
-0.103
1,186,554
880,451
5.6%
66,689
kkk
kkk
kkk
kkk
kkk
kkk
kkk
kkk
kkk
Chapter 2: Winners and Losers: 401 (k) Trading and Portfolio

Performance
In the US and in other countries around the world, participant-directed defined

contribution (DC) accounts are increasingly replacing professionally-managed defined
benefit (DB) plans. As a result, employees must take an increasingly active role in
managing their retirement assets. Yet little research has examined how active workers
manage their 401(k) plan assets, and even less is known about how a critical aspect o f
investment decision-making, trading activity, affects DC pension performance.31 Relying
on a unique new data set o f about one million active 401(k) participants in some 1,500
DC plans, this study is the first to evaluate in detail the impact o f workers trading
decisions on the performance o f their DC portfolios.
As observed in our related study on the propensity to trade in 401(k) plans, the
dominant trading behavior in 401(k) plans is not active or even somewhat inactive
trading, but rather non-trading (see Mitchell, Mottola, Utkus and Yamaguchi, 2006).
Most DC plan participants do not trade at all, a small group trades infrequently, and a
minority engages in quite active trading. In this study, we show that, as a group, traders
outperform nontraders when returns are not risk-adjusted. But because traders assume
higher portfolio risk, the difference in returns between the two broad groups disappears
after adjusting for risk. Further, we also find significant differences in risk-adjusted
3'Throughout this paper we use the terms 401(k) and DC plans interchangeably, though we recognize there
are other types o f DC plan designs including money purchase plans, standalone or 401(k)-paired profitsharing or Employee Stock Ownership Plans (ESOP) plans, 403(b) plans for the non-profit sector, and 457
plans for government entities.
34
returns among specific groups o f DC plan participants based on trading activity. In

particular, some types o f trading are beneficial, while other types are not. Rebalancing
seems to be a particular beneficial strategy on a risk-adjusted basis. Passive rebalancers,
or investors who hold only balanced or lifecycle funds and leave trading to the funds
portfolio manager, realize excess annual returns o f 84 basis points on a risk-adjusted
basis. Active rebalancers, who move their 401(k) portfolios equity allocation back to a
given target on their own, earn 26 basis points in excess risk-adjusted returns. Yet by our
estimates only about 10 percent o f participants rebalance their 401(k) account on an
active or passive basis thus leaving some 90 percent o f participants forfeiting the
potential advantages o f rebalancing. We also find that while some degree o f trading is a
return-enhancing trading strategy, very high portfolio turnover is not. Among those who
trade, investors who most actively chum their accounts lose 72 basis points per year
compared to traders with the lowest turnover ratios.
Our findings should be o f great interest to corporate plan sponsors and
policymakers charged with managing DC retirement systems. In particular, our research
underscores the value o f rebalancing as an essential investment approach. Currently,
most DC plans do not impose automatic rebalancing as the default; rather, each
participant must actively rebalance his portfolio from time to time, unless he selects a
single balanced or lifecycle fund that is professionally rebalanced on his behalf. In view
o f the clear rewards from rebalancing as an investment strategy, plan sponsors should ask
35
whether using an automatically rebalanced account should become the default.32 Finally,
since we find that high turnover rates in 401(k) plans harm investment performance, it
would appear that discouraging active trading would produce superior risk-adjusted
returns and ultimately higher retirement savings.
In what follows, we first discuss related research and then turn to a discussion o f
our methodology. N ext we describe the data and then turn to an analysis o f participant
raw and risk-adjusted returns. Subsequently we report our multivariate analysis o f the
effect o f trading patterns on portfolio behavior. A final section offers a short discussion
o f implications.
2.1 Related Studies

Relatively little is known about what motivates plan participants to trade in their
DC pension accounts. Nevertheless, conventional neoclassical models o f investor
behavior imply that a rational agent should continuously rebalance his portfolio to his
target allocation determined by his risk tolerance, as long as there are no transaction costs
(Sharpe 1964, Merton 1973). In the real world, o f course, even rational investors may
rebalance their portfolios less than instantaneously when they have drifted from their
target allocations, due to transaction costs. Nevertheless, such an investor still should
return to his target allocation when the marginal cost o f trading is less than the benefit,
and he should earn relatively more than an investor who does not follow this rebalancing
32 In a related development, the US Department of Labor recently issued new default fund regulations for
401 (k) plans under the 2006 Pension Protection Act, and these regulations encourage adoption of
automatically rebalanced balanced/lifecycle funds or managed account options as plan defaults.
36
approach (Grossman and Stiglitz, 1980). That said, in markets which are efficient and
where all agents act rationally, there should be no gains from trading at all. Returns o f
traders and nontraders should be identical.
By contrast, behavioral finance theorists suggest that trading may be both
irrational and detrimental to performance (Barber and Odean, 2000 and 2001; Gervais
and Odean, 2001). At one extreme, overconfidence may lead some investors to
misestimate the gains to be realized from trading, contributing to excessive and costly
portfolio turnover. Those more subject to overconfidence as a bias would thus realize
lower returns than those who are less biased. At the other extreme, procrastination or
inertia may lead investors to fail to trade, even though they might benefit by trading.
Participants in 401(k) plans do not usually directly pay for trading costs, so if they
are rational investors they would be expected to rebalance their retirement portfolios
frequently.
Yet research to date has found that pension participants trade rather
infrequently in their DC plans: for instance, Mitchell et al. (2006) report that only 20
percent o f participants trade in a broad set o f 1,500 plans over a two-year period. The
few who do trade tend to be affluent older men with long job tenure, who hold more
funds in their portfolios, and who are less likely to invest in index or lifecycle funds. In
their study o f a single employer plan, Agnew et al. (2003) find that only 10 percent o f
33 When a participant trades, the transaction is consolidated with all other buy and sell transactions for the
401 (k) fund option. The funds portfolio manager must buy and sell securities only on the net transactions,
and thus in many instances any given trade by a single participant may be offset in whole in part by
opposite trades by others, resulting in zero transaction costs for all or part o f the trade. Finally, any
transaction costs from a net sales or purchase position are shared across all holders o f the fund, and not just
the participants or other investors undertaking a transaction. Thus, even in the worst case where every
participant trade results in a purchase or sale of securities by the portfolio manager, virtually all o f the costs
are borne by other holders. Recently, there has been a move to impose redemption fees due to excessive
trading, and these fees would be borne directly by participants. During our study period, they applied only
to a limited number o f not-widely-held fund options.
37
their participants traded in a given year. In their study o f teachers, Ameriks and Zeldes
(2001) report that almost three-quarters o f participants never changed their investment
holdings over an entire decade.34 Thus far, no research has examined the impact o f
trading on 401(k) portfolio performance in the US context, which is what we undertake
below.
A few studies evaluate household investment and trading patterns in their
portfolios but do not focus on pensions (c.f. Calvet et al. 2006, Grinblatt and Keloharju
2000,2001; Odean et al. 2006; Guiso and Jappelli 2006). In general, these authors
confirm that household investors tend to underperform institutional investors. Odean
(1999) and Barber and Odean (2000) also investigate trading patterns and performance
but they are limited to a very special subset o f investors, namely investors holding retail
brokerage accounts. Their analysis shows that active stock traders realized substantially
lower risk-adjusted returns compared to nontraders for two reasons: traders must pay
transaction costs which lowers returns, and traders also trade a great deal due to
overconfidence. By contrast, 401(k) participants are usually offered pooled investment
vehicles such as mutual funds, where transactions fees from trading are borne by all
investors. In other words, DC plan traders are in effect subsidized by non-traders, which
may make 401(k) trading relatively more profitable than trading in a discount brokerage
account.
34 Madrian and Shea (2001) also find evidence of inertia in their study on automatic enrollment.
38
2.2 Empirical Strategy

Our research aim is to examine whether and how trading activity alters the
investment performance o f workers DC pension accounts. To evaluate this question, we
calculate raw (or non-risk-adjusted) returns, as well as risk-adjusted returns based on two
models: a variant o f the traditional Capital Asset Pricing Model (CAPM, Sharpe 1964),
as well as a Fama-French (1982,1993) multi-factor model.
Realized returns and risk. For each 401(k) participant account, the raw realized monthly
return is the asset-weighted average o f the returns realized by each o f the participants
beginning-of-month positions.35 The realized monthly return Ru o f a particular 401(k)
j
participants portfolio is defined as Rit = ^
>where R irt is the return o f participant i
7=1
in month t , j represents each fund in investor i s portfolio, ajtt is the dollar weight o f fund
j in beginning-of-month portfolio for investor /, RJ>t is the total return o f fund j in month t,
and J is the total number o f held funds. We also calculate a relative return measure for
each participant portfolio and a risk measure. The Own Relative Benchmark return is
computed as the difference between what the participant actually realized and what he
would have earned had he always rebalanced back to the allocation reflected in his
contribution allocation.36 This is expressed as R fwn = Ri t - RJ" where RAowniitis the
35 Following Barber and Odean (2000), we assume that all transactions occur at the end o f a month and
therefore ignore the impact o f intra-month transactions.
36 Each participants own benchmark return is computed using the contribution allocation that was on
record when the first contribution was recorded. Our analysis confirms that contribution allocations tend to
be extremely stable over time. In only a handful o f cases, a fund to which a participant had directed his
initial contribution was later dropped. If such an instance, if the fund was merged into another one, we
assumed that he would have selected the acquiring fund; if the fund was simply dropped, we use a fund
39
own relative benchmark return o f participant i in month t, Ri>t is the total return o f
participant i in month t, and
is the monthly return on his/her own benchmark.
Finally, we calculate portfolio risk as the average monthly excess portfolio standard
deviation.
Risk-adjustment methods. Following Fama-MacBeth (1972), we begin by calculating
factor loadings for each o f the underlying assets held by the 401 (k) participants in our
universe. We use the returns o f the fund investments from the five-year period prior to
our study - January 1998 to December 2002 - as the period for estimating these factors.
There are two sets o f factors calculated: one for the CAPM variant, and a second for the
Fama-French model.
The investment options included in our dataset include a wide variety o f domestic
and international stock funds, bond funds, balanced or lifecycle funds, investment
contract funds, and money market funds. For the modified CAPM model, we take this
breadth o f asset choices into account by regressing the excess return37 for each o f the
401(k) funds in our universe on three market indices: the value-weighted CRSP portfolio,
the Lehman Brothers Aggregate Bond Index (LBA), and the Morgan Stanley Capital
International (MSCI) Europe, Australia and Far East (EAFE) Index. These, respectively,
represent the US equity market, the US bond market, and the international equity market.
The residuals o f the LBA and MSCI EAFE are regressed on the CRSP US market index,
from the same asset class, if available at the plan, or a benchmark from that asset class to compute
benchmark returns.
37 Throughout the paper, all returns for participant portfolios and market indicesare excess returns,
calculated using the return o f a US Treasury Bill as the risk-free investment.
40
in practice, to create orthogonal factors because LBA and MSCI EAFE are highly
correlated with CRSP. This regression function can be written as
R jJ ~ R f ,t ~
(2 '1 )
( R c RSP j ~ R f , t ) + P l ^ - L B A X , t + P i ^ M S C I X , l + S j , t
where Rjit is the total return o f fund j in month t, and R /x is the risk free return as defined
above, R crsp.i, R lbaxj, and R mscixj are returns on the CRSP value-weighted market
portfolio and residuals o f the LBA and MSCI EAFE indexes regressed against CRSP,
respectively. /?/, P 2 and
are the regression coefficients or factor loadings; % is the
error term.
The Fama-French multi-factor approach follows the same methodology but adds
factors for firm size and book-to-market ratio:
R j,t ~ R f ,t
=A
(R a tS P ,! ~ R f , t
) + f i l ^ L B A X ,f + ^ 3 ^ M S C I X J +
+ S jjt
(2-2)
where Rsm.t is the Fama-French small-minus-big (SMB) index and R hmlxj is the residual
o f Fama-French high-minus-low (HML) index regressed on SMB. We again employ the
residuals o f the HML factor because HML and SMB factors are highly correlated during
the period.
As a result, each investment option in our universe has two sets o f factor
weightings based on the preceding five-year period/?/, /^ a n d
for the CAPM model,
and another /?y through Psfox the Fama-French model. We next impute each participants
exposure to these factors for the subsequent 24-month period, January 2003 to December
2004, the period over which we observe participant trading activity. The participant
41
exposure to these factors is a weighted average of the factors for each fund held by the
participant over the 24 months.
10
In the end, for the CAPM model, each participant has a weighted exposure to
three distinct factors: B itcRSP is the participants average exposure to the CRSP valueweighted US stock portfolio factor (controlling for the US equity market); B^lbax is the
participants average exposure to the residual Lehman Brothers Aggregate Bond Index
factor (capturing residual effects o f the US bond market not already captured in the CRSP
index); and B^mscix is his average exposure to the residual MSCI EAFE International
factor (capturing residual effects o f non-US stocks not already captured in the CRPS
index). For the Fama-French set o f factors, each participant has an unique exposure to
the three preceding factors, plus B ^ mb, capturing the effects o f small versus large US
stocks, and B^hmlx capturing the residual effects o f growth versus value stocks.
Finally, we regress the average realized excess return o f each participants
portfolio on his risk exposure to each factor, and a set o f behavior variables to evaluate
how trading effects 401(k) investment performance. The CAPM and Fama-French
regressions are as follows:
R i ,e x =
Yo
Y 2 ^ i , l b ax + Yiî,Msax + STRADEt +
+ Y \ B i ,c r s p
Rl,ex = Yo + Y 1B lCRSI> + Y 2^1,LBAX
YiîJHSCIX + Y A
(2-3)
+Y î,HMLX + STR A D E j +
(2-4)
q ft
Specifically g
04
psa
Z j
jL u
t=Jan ,03 7=1
_ _L_ v*
S' a
jJ r
j,k
where Bhk is participant i s risk loading on factor k, a is the
allocation o f fundj in beginning-of-month portfolio in month t, and
k is the regression coefficient of
fundj on factor k.
42
where Rlex indicates participant Vs realized average excess return.39 Of most interest is
the vector 8 on the TRADE, variable which varies across specific models, as described
below.
2.3 Hypotheses
As part o f our empirical approach we developed several hypotheses based on the
previous literature. First, according to conventional neoclassical investment theory, in
the near-frictionless world o f 401(k) trading, individuals would not be hindered by
transaction costs as they are in other types o f investment accounts, and so might trade
more frequently. But in an informationally efficient market, such trading would be for
rebalancing purposes only and there would be no sustainable gains to be realized by
trading for other reasons. As a result, from this point o f view, traders and nontraders
would realize similar risk-adjusted returns:
H I : Traders will earn no more than nontraders on a risk-adjusted basis.
To test this hypothesis, we define TRADE as a dummy variable where TRADE = 1 if the
participant trades in his or her account. If H I holds, we would anticipate that coefficient
o f TRADER, 8trade, should be zero.
From a behavioral perspective, prior research has also identified a tendency
among investors, including 401(k) participants, to be subject to procrastination and
inertiato be inattentive in managing their portfolios. As a result, workers might
go
Specifically
= _
94
Dec , 04
Z_i
where R,t, is the participant account return (defined above) for a given
r
hi
l = Jan ,03
month..
43
overlook the potential advantages o f periodically rebalancing their portfolio based to their
preferred level o f risk. Under this behavioral approach, the attentive investors - the
rebalancers - should do better:
H2: Rebalancers outperform nonrebalancers on a risk-adjusted basis.
Here we define a type o f rebalancers known as active rebalancers. They are investors
who return to their target equity allocation when they trade. (As described later in more
detail, we define active rebalancers as those who always trade so their asset allocation
falls within +/-10 percent o f their target allocation.) Accordingly, if TRADE=1 when the
investor is an active rebalancer and zero otherwise, this hypothesis predicts that dtrade > 0.
A further behavioral model is that while certain individuals may strive to
rebalance their portfolios, they may still be subject to procrastination or inertia to some
degree. As a result, they may fail to rebalance on a consistent, disciplined basis. Under
this hypothesis, even those who are observed rebalancing may be inattentive from time to
time. As a consequence, those who have their portfolios rebalanced by a more attentive
agent, such as a professional money manager paid to rebalance a portfolio, should earn
superior risk-adjusted returns:
H3: Passive rebalancers peform better than active rebalancers on a risk-adjusted

basis.
By passive rebalancers, we mean investors who hold only balanced or lifecycle funds in
their account during the entire 24-month period - in effect, the situation where the funds
portfolio manager rebalances on the investors behalf. If even some active rebalancers
44
are inattentive, we would expect passive rebalancers to do better. In other

words, 8 pr > 8 ar, where 8 pr, 8 ar refer to passive and active rebalancers, respectively.
A final behavioral test focuses on overconfidence. We know from our prior work
that active traders are more likely to be affluent males, and other research suggests that
such individuals may be overconfident generally and also overconfident with respect to
portfolio trading in particular. In keeping with the theory o f overconfidence, we would
anticipate that higher levels o f portfolio turnover lead to lower risk-adjusted returns, at
least among those trading.
H4: Traders with high turnover rates will earn less than those with low turnover.
Accordingly, the final variation o f the model defines the TRADE variable as the
investors turnover ratio conditional on having traded. Then we test 8turnover < 0 . If
traders are overconfident, returns would be lower for the highest turnover quintile.
2.4 Data and Descriptive Statistics

To test these hypotheses we use a dataset o f administrative records on 401(k) plan
participants provided by Vanguard, covering 1,483 retirement plans offered by a wide
range o f employers. No previous study has had access to the diversity o f plans and
richness o f data on over a million active participants, including participant investment
45
holdings, trading patterns, contributions, and demographic characteristics.40 The trading

data are available for the 24 month period, January 2003 through December 2004.41
Descriptive statistics for the entire sample appear in Table 2-1, along with
summary information comparing traders and nontraders. Panel A shows that the median
participants account balance is about $42,000, but traders have significantly higher
balances than nontraders. Overall, the average equity exposure is 64 percent, and here
traders do not differ from nontraders. Traders reallocate their balances about three times
over the two-year period; the mean portfolio turnover rate over the same period is 92
percent. Panel B indicates that the participants average age is 44, about half are men,
mean plan tenure is 8 years, and average household income is about $87,000. Our prior
study on 401(k) trading pattern showed that traders tenet to be longer tenured, more
affluent, males, as compared to nontraders.42
Table 2-1 here
Trading patterns are summarized in Table 2-2, where we see that just over onefifth o f participants traded during the period. Overall, some 17 percent are active traders
and only three percent are active rebalancers. (As a percentage o f all traders, 15 percent
are active rebalancers, and 85 percent are active traders.)
Table 2-2 here
40 See Mitchell et al. (2006) for more discussion o f the data. Active participants as those who made a
contribution to their plan during the sample period. Also participants who invest in privately-held company
stock funds are excluded.
41 Our database also included historic fund return information, as described in more detail in Appendix I.
42 We were able to obtain data generated by the IXI Corporation on non-retirement wealth and
assign these to participants by zip code. These are categorical variables collapsed here into three
groups: Poor (wealth<$7,280), Medium ($7,280 to $61,289), and Rich (>$61,289).
46
We define an active rebalancer as a participant whose sole trading during the

observation period always returns him to within + /-1 0 percentage points o f his target
equity/fixed income ratio. Empirically, the target is set at each employees own equity
percentage associated with his (first observed) plan contribution. To illustrate, suppose
that A s first observed contribution to his 401(k) account was a 50/50 equity/fixed
income allocation. He would be classified as an active rebalancer if he was a trader and
all o f his trading maintained the equity/fixed income ratio within this + /-1 0 percent band.
All other types o f traders are defined as active traders.
Nontraders can be further subdivided as well. The Table shows that close to six
percent o f all participants are passive rebalancers participants who never traded on
their own and invested their entire account balance in one or more balanced or lifecycle
funds over the entire 24-month period.43 In effect, the holdings o f passive rebalancers are
rebalanced regularly by the funds (or funds) portfolio manager, with no trading activity
by the participant. A notable fact is that total rebalancing, by our definition, accounts for
less than 10% o f participants: 3.1% are active rebalancers, while 5.5% are passive
rebalancers. Other nontraders, a group comprising 74 percent o f participants, are those
participants who never made a trade and who did not invest exclusively in balanced or
lifecycle funds.
43 Nearly 90% o f passive rebalancers hold only one fund, but a few hold more than one balanced and/or
lifecycle fund (possibly because they switched their contributions to a new fund but maintained balances in
an old fund); accordingly the mean number o f funds held by passive rebalancers is 1.1.
47
2.5 P erform ance R esults

Table 2-3 presents realized return and risk measures, along with excess or relative
returns compared to the participants own benchmark. Columns 1 and 2 show that
trading seems to have a positive impact on investment performance in the 401(k)
environment, using raw or non-risk-adjusted data. Thus Panel A compares raw returns
for traders versus nontraders, and the difference indicates that traders significantly
outperform nontraders by 4 basis points per month or an annualized 55 basis points
(difference 1-2). But as column 3 shows, traders hold more volatile portfolios and these
differences are statistically significant.44 Columns 4 and 5 compute how well the two
groups actually did, compared to what they would have earned had they always held their
Own Benchmark. The findings prove that actual returns compared to those from the Own
Benchmark do not differ significantly between traders and nontraders (consistent with
Grossman and Stiglitz, 1980). Table 2-3 also shows that relative returns for traders
exceed those o f nontraders (columns 6-7) when measured against either their ownbenchmark by about five basis points per month or an annualized 60 basis points per year
(difference 1-2).
Table 2-3 here
To press the comparisons further, we next disaggregate traders by type, to see
whether participants who rebalance their portfolios do better than others. Column 1 o f
Panel B reports unadjusted returns, where active rebalancers outperform nontraders by 15
basis points per month (line 1), while passive rebalancers underperform nontraders by 11
44 Some 15% o f the sample holds only fixed income funds in their 401 (k) portfolios. Volatility
and excess returns are low for this asset subgroup.
48
basis points per month (line 2). This is a substantial advantage on an annualized basis
(Column 2). Yet there are important differences within trading subgroups. Thus active
rebalancers portfolios are riskier and passive rebalancers less risky (Column 3, D iff 1-4).
In all, Columns 4-5 show that active rebalancers do better than nontraders, given their
own benchmarks, but passive rebalancers underperform nontraders (these results do not
yet correct for risk differences). Relative realized returns over the benchmark appear in
Columns 6-7, where we see that both rebalancer groups (the passives and the actives) do
not achieve significantly higher excess relative returns, but other traders do. This is not
surprising because the rebalancer groups always maintain its then-current risk profile and
keeps a relatively constant asset allocation.45
Panel C o f Table 2-3 groups the sample into quintiles according to portfolio
turnover, where Quintile 1 contains traders with the lowest turnover rates, and Quintile 5
includes traders with the highest turnover rates.46 On a realized-retum basis, lowtumover traders (Quintile 1) outperform higher-turnover traders (Columns 1-2). Lowtumover traders in Quintile 1 have somewhat higher standard deviations (Column 3).
Columns 4-5 indicate that performance falls with turnover. Finally, excess returns using
the Own Benchmark do not differ across quintiles.
These results, as they are unadjusted for risk, seem to show that traders
outperform nontraders, and active rebalancers do better than passive rebalancers. But
45 This statement is consistent with investors having a constant relatively risk averse utility function, as
illustrated in the classic financial economics literature (c.f. Sharpe 1964).
46 Turnover is computed for each participant by dividing the dollar value o f assets traded from 1/03 to
12/04 by the participants average assets balance at the beginning and end o f the sample period.
49
risk-adjusted returns in Table 2-4 lead to quite different conclusions.47 For instance, Panel
A includes the dichotomous TRADE variable which takes a value o f 1 if the participant is
a trader (0 else): after controlling for risk, the trade advantage is greatly attenuated.
Specifically, in the CAPM formulation (Column 1), traders have only an excess riskadjusted return o f 2 basis points per month, while the Fama-French approach (Column 3)
finds no significant impact. In other words, based on the more robust risk-adjustment
approach o f Fama-French, trading has no impact on risk-adjusted returns. Thus, H I,
based on the neoclassical view that traders and nontraders should realized similar riskadjusted returns, is not rejected.
Table 2-4 here
A more granular trader definition is offered in Panels B and C, where participants
are classified according to whether they are active rebalancers, active traders (but not
rebalancers), passive rebalancers, or nontraders. Here the reference category is nontraders.
In Column 1, we see that active rebalancers outperform other nontraders by 6 basis points
per month or 72 basis points on an annualized basis. Meanwhile, passive rebalancers
outperform other nontraders by 8 basis points per month or 108 basis points on an
annualized basis in the CAPM specification. Similar though slightly smaller results are
evident in Column 3 using the Fama-French model. Active rebalancers outperform other
nontraders by 2 basis points per month or 26 basis points per year, while passive
rebalancers outperform by 7 basis points per month or 85 basis points per year. In both
47 For both the CAPM and Fama-French multi-factor regressions, all standard errors are robust across plans
and all functions are estimated by Ordinary Least Squares. We use Rogers (1993) approach for clustered
samples to compute robust standard errors; see Appendix 2.
50
formulations, active traders do no better than nontraders, and rebalancers outperform all
other types o f trading. We cannot reject H2 the notion that some investors are
inattentive and fail to rebalance at all, leading rebalancers to realize superior risk-adjusted
returns. We also find that passive rebalancers do relatively better, consistent with H3. In
other words, even active rebalances are perhaps less attentive than they should be, and as
a result, those whose portfolios are rebalanced by a third party realize superior riskadjusted returns.
Earlier we noted that those who engage in higher turnover in their 401(k)
accounts have lower returns unadjusted for risk. Table 2-5 compares risk-adjusted
performance for two turnover measures. Panel A breaks the sample into five quintiles
from lowest to highest turnover while Panel B presents the results o f a polynomial model
that tests for a curvilinear relationship between turnover and risk-adjusted performance.
In both models, those with higher turnover earn less. For instance, in Column 3, traders
in the highest turnover group (Quintile 5) underperform the lowest turnover group by 6
basis points per month or 72 basis points on an annualized basis.
Table 2-5 here
To test whether the turnover relationship is linear, Panel B uses the turnover rate
as a continuous variable along with its square and cube. Again, both models yield similar
results: the overall turnover effect is negative and higher 401 (k) portfolio turnover means
lower risk-adjusted returns. Yet the positive coefficient for turnover squared suggests
that some amount o f turnover helps improve performance. In other words, those engaged
in modest turnover earn higher risk-adjusted returns than those who fail to trade at all.
But the negative coefficient on turnover cubed shows that at high levels o f trading
51
performance declines. In other words, some turnover enhances returns, but high turnover
is costly. Our hypothesis H4 regarding overconfidence is largely confirmed
overconfidence as measured by excessive trading is costlybut some trading remains
beneficial.
Evidently, workers who rebalance their accounts do better from a risk-adjusted
performance standpoint. In particular, active rebalancers do best, yet they are a rare
breed, accounting for only a handful o f participants. To better understand who they are,
we present a Probit model in Table 2-6 where active rebalancers are coded as 1 and all
other traders are the reference group. Explanatory factors include demographic
characteristics, plan design variables, and investment holdings. The results show that
active rebalancers are slightly younger, somewhat less affluent, and more likely to be
women, compared to other traders. Some plan design variables are also related to the
prevalence o f active rebalancers: for instance, when more funds are offered in the 401(k)
menu, the likelihood o f being an active rebalancer decreases. This suggests that larger
plan menus tend to encourage trading beyond traditional rebalancing. We also see that
web-registered participants with online access to their accounts are less likely to be active
rebalancers, and more likely to be a more-active trader. Traders who invest in index
funds are more likely to be active rebalancers, perhaps because they are attracted to the
buy-and-hold approach o f index funds. In addition, trading restrictions imposed by index
funds may discourage other types o f traders from using index funds.
Table2- 6 here
52
2.6 Discussion and Implications

Over the last half-century, US company pensions have shifted from a system
dominated by defined benefit (DB) plans to one dominated by defined contribution plans,
usually o f the 401(k) variety; it is estimated that 43 million private-sector employees
currently manage $3.7 trillion in their DC pension accounts. A similar trend is evident
around the world. Accordingly, it is o f keen interest to investigate how trading in 401(k)
plans affects DC pension portfolio performance. We conclude that those who trade in
their accounts seem to earn higher returns before adjusting for risk, but traders fail to
outperform after risk adjustment. Also, we find that passive rebalancers perform best (on
a risk-adjusted basis): these are investors who hold only balanced or lifecycle funds
where their portfolio manager rebalances on their behalf. A passive rebalancer can earn
substantially more - over 80 basis points per year - compared to traders and other
nontraders.
Further, we conclude that some types o f trading are productive while others are
not. That is, active rebalancers who always steer their equity allocation back to a target
allocation earn 26 basis points more per year than non-rebalancing traders. Since only
about 10 percent o f participants take advantage o f rebalancing strategies in their 401 (k)
accounts, either passive or active, this means that some 90 percent leave money on the
table. Last, active churning o f 401(k) accounts is detrimental: participants in the highest
turnover quintile lose, on average, more than 70 basis points per year on an annualized
basis, compared to those in the lowest turnover quintile. A little turnover may be
beneficial as the case o f rebalancing suggests. But a great deal o f turnover is costly.
53
Our findings should be o f great interest in the current environment where plan
sponsors and policymakers seek to improve the performance o f DC pension plans.
Currently, most DC plans do not induce automatic rebalancing; rather, participants must
actively decide whether to rebalance their own portfolios periodically, or opt to invest in
professionally-rebalanced funds. In view o f the rewards from passive rebalancing as an
investment strategy, our research underscores the value o f offering a rebalancing fund or
service as an investment default, such as a balanced or life cycle fund, or a managed
account. Employers and recordkeepers overseeing 401(k) plans may also want to
consider whether automatic rebalancing o f 401(k) accounts should be the default design.
Furthermore, policies designed to discourage active trading in 401(k) plans would
likely produce superior risk-adjusted returns, and ultimately higher retirement saving
since high turnover rates harm investment performance. Round-trip restrictions and early
redemption fees are two examples o f policies that have been recently introduced in the
US to deter excessive market-timing trading by investors. These or similar policies
would appear to improve returns and reduce transaction costs for all participants since, in
the commingled investment offerings o f most DC plans, transactions costs are borne by
all holders, not just the traders.
Future research could address several unanswered questions. First, this paper
classifies participants as rebalancers based on investment patterns and trading behavior.
In future work, it would be valuable to survey participants to enhance our understanding
o f investor motivation. Second, this paper confirms the widespread reality o f inertia
among investors: only 20 percent o f plan participants traded over our period, only 3
percent actively rebalanced their accounts, and only 6 percent passively rebalanced by
54
investing in balanced funds. Survey research might also assess whether participant
inertia is driven by financial illiteracy or a conscious decision not to act. Third, we
examine participant portfolios over a two-year period favorable to equity investing. In
future work, we will include additional years to see whether our results generalize to
different market conditions. Fourth, ongoing research is evaluating whether changes in
investment menus offered to 401(k) participants might influence trading patterns in the
long run.48 A last intriguing question is whether there are other potential explanations,
besides rebalancing, for why passive rebalancers do better than active rebalancers and
active traders. We have already empirically ruled out the explanation that passive
rebalancers generate superior returns by investing disproportionately in low-cost index
funds.49 However, evaluating alternative explanations for this better risk-adjusted
performance remains an important area for future research.
48 Elton et al. (2006) find that plan participants alter their allocations in response to investment menu
changes.
49 Many o f the balanced funds in our study are index-based, so they tend to have lower expenses than
actively managed funds. By definition, passive rebalancers are more likely than other participants to invest
in these lower-priced funds, so it is possible that the superior performance o f passive rebalancers is due to
the fact that they simply choose funds with low expense ratios. To test for this possibility, we reran the
regressions and added a new variable that we defined as the percent of a participants portfolio invested in
index funds. In all instances, this new variable was non-significant and had no impact on the coefficients in
the equations -- suggesting that the superior performance o f passive rebalancers is not due to their tendency
to invest in lower-priced funds.
55
Appendix
2-1: Fund Return Data
The Vanguard return database is constructed for the 737 distinct investment
options held by our participants in their 401(k) plans, as well as a variety o f market
benchmarks. The 401(k) assets included publicly available mutual funds, company
stocks, commingled funds, as well as privately managed separate accounts. In total, our
returns database encompassed seven years - five years prior to our study and the two
years for our trading analysis. Monthly total returns from January 2003 to December
2004 were used to calculate the actual portfolio returns realized by our over one million
active participants, as well as a variety o f risk measures. These returns are discussed in
the text. Monthly asset returns for the prior five years, January 1998 to December 2002,
are used in estimating risk-adjusted returns using both CAPM and Fama-French riskadjustment models.
2-2: Robust Estimation of Standard Errors in the Clustering Case

As sketched in Agnew et al. (2003), Rogers (1993) general formula for the
covariance matrix with heteroskedasticity and clustering in a panel data formulation takes
the following form:
=1
\ i
where i is the cluster, V is the negative inverse o f the Hessian o f the log-likelihood, and u,
is the vector o f contributions o f cluster / with Tt observations to the scores o f the
likelihood expressed as:
A dLnLi ,
< = 2 / dS
(= i
56
where Li t is the likelihood o f the t-th observation for individual i and <?is the parameter
vector.
As our sample has a multi-level structure but is not panel in design, we rewrite the
estimated robust variance-covariance matrix as:
(
=1 (=1
V 1 P
P =1
( N
1=1
i= i
P ~
/=1
where p refers to the plan or cluster, i refers to the participant, ZitPis the vector o f
observations on the independent variables, and s i
is the residual o f the regression.
In practice, SAS provides this estimator in Proc Surveyreg for linear regression
with least-squares estimation, and Proc Surveylogistic for nonlinear maximum-likelihood.
57
Table 2-1. Socio-demographic Characteristics of 401(k) Plan Participants

Whole Sample
Mean (1)
Nontraders
Traders
Median (2)
Mean (3)
Median (4)
Mean (5)
Median (6)
Panel A: A ccount C haracteristics

83,248
42,255
133,635
74,739
70,460
36,471
Equity Allocation of Balance
0.643
0.753
0.630
0.737
0.647
0.758
Trading Propensity
0.202
Number ofTrades
0.582
0.000
2.876
1.000
Turnover Ratio
0.186
0.000
0.917
0.497
43.536
44.000
44.982
46.000
43.169
43.000
Balance ($)
NA
NA
Panel B: Participant C haracteristics

Age (years)
Male (male=1)
0.465
Plan Tenure (years)
7.940
6.081
9.353
7.669
7.582
5.665
Household Income
87,184
86,319
97,084
86,319
84,671
86,319
0.550
0.444
IXI_Rich (yes=1)
0.226
0.291
0.209
IXI_Medium (yes=1)
0.448
0.456
0.446
IXI_Poor (yes=1)
0.326
0.253
0.345
Note: Balance a s of 01/03. A trade is defined as any fund allocation change on a given day. A trader is a
participant who ever executed at leat one trade from 01/03 to 12/04. Trading propensity refers to the percent of
participants who traded over the period. Data on non-pension wealth are provided by the IXI corporation and
matched by ZIP code. IXI_Rich refers to non-pension wealth>$61,289, IXI_Medium refers to $7,208-61,289, and
IXI_Poor refers to <$7,280.
58
Table 2-2. Distribution of Accounts by Trader Type

P ercen t (%)
B alan ce Equity Allocation
Total
100.0
0.643
Traders
20.2
0.630
Active R eb alan cers
3.1
0.721
Active Traders
17.2
0.613
79.8
0.647
5.5
0.565
74.2
0.653
A ccount Type
Non-traders
P a ssiv e R eb alan cers
O ther N ontraders
Note: S e e Table 2-1. Active reb alan cers a re participants who reb alan ce their fixedincom e/equity ratio to their targ eted fraction, which defined a s the fixed-incom e/equity ratio
of their first contribution. Active traders a re trad ers who are not active reb alan cers. P a ssiv e
reb alan cers a re p articipants w ho hold only balanced funds throughout th e sa m p le period and
did not trade. O ther n ontraders a re th e remaining of nontraders.
59
Table 2-3. Realized 401(k) Account Returns

Raw Return (Non-risk adjusted)
Monthly
Own B enchm ark Return
Monthly
A nnualized
(3)
(2)
(1)
Standard
D elation
(4)
A nnualized
(5)
E x c e ss R eturns over Own

B enchm ark
Monthly
(6)
A nnualized
(7)
P anel A: Traders vs. Non-traders

(1)
0.0133
[0.1718]
0.0241
0.0127
[0.1635]
0.0006
Non-traders (2)
0.0129
[0.1663]
0.0222
0.0128
[0.1649]
0.0001
[0.0012]
Difference
0.0004
[0.0055]
0.0019
-0 0010
1-0 0012J
0.0005
[0.0060]
Traders
(l)-(2)
[0.0072]
P anel B: D isaggregated Trading C ategories

[-0 0012]
Active rebalancers
(1)
0.0145
[0.1886]
0.0251
0.0146
[0.1900]
Activer trad ers
(2)
0.0131
[0.1690]
0.0239
0.0124
[0.1594]
0.0007
[0.0084]
P a ssiv e reb alancers
(3)
0.0119
[0.1525]
0.0193
0.0119
[0.1525]
OOOOO
[OOOOO]
O th er non-traders
(4)
0.0130
[0.1677]
0.0224
0.0128
[0.1649]
0.0002
[0.0024]
(1)-(4)
0.0015
[0.0209]
0.0027
0.0018
[0.0218]
-0.0003
[-0.0012]
(1)-(2)
0.0014
[0.0195]
0.0012
0.0022
[0.0267]
-0.0008
[-0.0024]
Q1
(1)
0.0146
[0.1900]
0.0249
0.014
[0.1816]
0.0006
[0.0072]
Q2
(2)
0.0140
[0.1816]
0.0242
0.0135
[0.1746]
0.0006
[0.0072]
Q3
(3)
0.0132
[0.1704]
0.0234
0.0126
[0.1621]
0.0006
[0.0072]
Q4
(4)
0.0120
[0.1539]
0.0229
0.0113
[0.1444]
0.0007
[0.0084]
(5)
0.0127
[0.1635]
0.0251
0.0123
[0.1580]
0.0004
[0.0048]
[0.0195]
OOOOO
[OOOOO]
Traders
Non-traders
Difference
P anel C :Tum overR atio Quintiles
Q5
Difference
(1)-(3)
0.0014
[0.0196]
0.0015
0.0014
(1)-(4)
0.0026
[0.0361]
0.0020
0.0027
[0.0329]
0 0001
[-0 0012}
(1)-(5)
0.0019
[0.0265]
-0.0001
0.0017
[0.0206]
0 0002
[0 0024]
Note: All returns a re nominal returns. Overall average monthly raw return is 0.01298. S hadow a re a indicates s ta tistic s a re not significant a t th e 5% level. Raw return is the
average of monthly dollar-weighted returns (01/03 -12/04). O w n-benchm ark return is th e return th a t an investor would have earned had h e alw ays rebalanced to his desired
contribution allocation. N um bers in sq u are b ra c k ets are annualized returns. Q 1, Q2, Q3, Q4, Q 5 refer to turnover ratio of (0-16.9%],(16.9% -36.8% ],(36.8% -65.5% ],(65.5% 109.6%] an d 109.6% + , respectively.
60
Table 2-4. Determinants of 401(k) Risk-Adjusted Returns

(OLS models with robust standard errors)___________________
C A P M + T rading
F a m a -F r e n c h M ultiF a c to r+ T ra d in g
C oefficient
t-value
C oefficient
t-value
(1)
(2)
(3)
(4)
P a n e l A : T ra d ers vs. N o n tra d e rs

In tercep t
0 .0 0 3 7 ***
8 .1 2
0 .0 0 1 9 ***
7 .0 6
CRSP_V W
0 .0 1 3 3 ***
2 4 .1 9
0 .0 1 4 3 ***
3 9 .6 6
LBAX
-0 .0 0 1 3
0 .0 0 0 7
1 .1 7
0 .0 0 5 3 ***
3 .0 9
HMLX
0 .0 1 0 8 ***
1 1 .6 2
SM B
0 .0 0 5 9 ***
MSCIX
-1 .2 7
0 .0101 ***
4 .0 4
0 .0 0 0 2 *
T rad ers
1.71
-0.0001
4 .3 7
-1.11
N o n -trad e rs (re fe re n c e g ro u p )
N
1 ,0 1 5 ,5 5 7
1 ,0 1 5 ,5 5 7
7 0 .9 %
R2
8 6 .1 %
P a n e l B : A c tiv e T ra d ers v s. N o n tra d e rs

In tercep t
0 .0 0 3 7 ***
8 .1 2
0 .0 0 1 9 ***
7 .0 7
C RSP_V W
0 .0 1 3 3 ***
2 4 .1 8
0 .0 1 4 3 ***
3 9 .6 5
-1 .2 7
0 .0 0 0 7
1 .1 7
0 .0 0 5 3 ***
3 .0 9
HMLX
0 .0 1 0 8 ***
1 1 .6 2
SM B
0 .0 0 5 9 ***
4 .3 8
LBAX
-0 .0 0 1 3
MSCIX
T rad e rs
0 .0101 ***
4 .0 4
A ctive R e b a la n c e r
0 .0 0 0 5 ***
5.01
A c tiv e T ra d e r
0.0001
0 .9 4
0 .0 0 0 2 **
-0.0001
2.41
-1 .5 5
N o n tra d e rs (re fe re n c e g ro u p )
N
1 ,0 1 5 ,5 5 7
1 ,0 1 5 ,5 5 7
R2
71 .0 %
8 6 .1 %
P a n e l C: A c tiv e T ra d e rs v s. O th e r N o n tra d e rs
In te rc e p t
CRSP_V W
0 .0 0 3 6 ***
8.11
0 .0 1 3 3 ***
2 4 .3 2
-0 .0 0 1 4
LBAX
MSCIX
-1 .3 2
0 .0 1 0 0 ***
3 .9 6
HMLX
SM B
T rad e rs
N o n -trad e rs
A ctive R e b a la n c e r
0 .0 0 0 6 ***
5 .4 2
A ctiv e T ra d e r
0.0001
1.57
P a s s iv e R e b a la n c e
0 .0 0 0 8 **
2 .1 9
0 .0 0 1 8 ***
7 .2 9
0 .0 1 4 3 ***
4 0 .0 5
0 .0 0 0 6
0 .9 5
0 .0 0 5 2 ***
2 .9 9
0 .0 1 0 9 ***
11 .7 2
0 .0 0 5 9 ***
4 .3 2
0 .0 0 0 2 ***
3 .0 2
-0.0001
-1 .0 5
0 .0 0 0 7 **
2 .3 6
O th e r N o n tra d e r (re fe re n c e group)

N
1 ,0 1 5 ,5 5 7
1 ,0 1 5 ,5 5 7
R2
71 .0 %
8 6 .2 %
N ote: S e e T a b le 2-2. **,**,* in d ic a te s ta tis tic a lly sig n ific a n t a t 1 % ,5 % ,1 0 % level, re s p e c tiv e ly . C R S P _ V W is th e
e x c e s s re tu rn o f th e v a lu e -w eig h ted C R S P m a rk e t portfolio; L B A X is th e re s id u a l o f th e L e h m a n B ro th e rs
A g g re g a te B o n d Index e x c e s s re tu rn r e g r e s s e d o n C R S P _ V W : M SC I is th e re s id u a l o f M SC I E A F E in te rn a tio n a l
eq u ity in d ex e x c e s s re tu rn r e g r e s s e d o n C R S P _ V W ; H M LX is th e re s id u a l o f F a m a -F r e n c h HML fa c to r
r e g r e s s e d o n S M B fa c to r; S M B is th e F a m a -F r e n c h S m a ll m in u s Big factor.
61
Table 2-5. Determinants of 401(k) Risk-Adjusted Returns for Traders Only

(OLS models with robust standard errors)
CAPM+Trading
Fam a-French MultiFactor+Trading
Coefficient
t-value
Coefficient
t-value
(1)
(2)
(3)
(4)
Panel A : Traders b y Turnover Ratio 1
Intercept
0.0045 ***
7.32
0.0019 ***
7.09
CRSP_VW
0.0130 ***
21.88
0.0144 ***
41.58
-1.40
0.0005
0.90
0.0060 ***
3.41
HMLX
0.0117 ***
12.91
SMB
0.0048 ***
3.55
LBAX
MSCIX
-0.0015
0.0117 ***
3.73
Q1 (Reference group)
Q2
-0.0003 ***
-2.99
-0.0002 ***
4.67
Q3
-0.0006 ***
-3.62
-0.0003 ***
-5.49
Q4
-0.0011 ***
-4.53
-0.0006 ***
-6.12
Q5
-0.0010 ***
-3.71
-0.0006 ***
-4.38
205,557
205,557
R2
66.1%
81.6%
Panel B: Traders b y Turnover Ratio 2
Intercept
0.0039 ***
7.36
0.0016 ***
5.70
CRSP_VW
0.0131 ***
22.15
0.0145 ***
41.52
-1.41
0.0005
LBAX
MSCIX
-0.0015
0.0059 ***
3.38
HMLX
0.0118 ***
12.84
SMB
0.0047 ***
3.45
-2.39
-0.0927 ***
-4.02
0.3967 ***
2.87
0.2700 ***
4.47
-0.0021 ***
-2.74
-0.0014 ***
-4.50
Tumover/1000
Tumover_sq/100000
Tumover_cube/100000
0.0117 ***
-0.1000 **
3.69
0.91
205,557
205,557
R2
65.7%
81.4%
Note: S e e Table 2-4 for detail
62
Table 2-6. Characteristics of Active Rebalancers

(Probit models with robust standard errors)
D ependent Variable: Active R ebalancer Dummy (M ean=15% )
M ean
Unit
Model A
Model B
Model C
D em ographics Only
+P lan Design
+ Prior Holding
R egression
Marginal
R egression
Marginal
R egression
Marginal
Coefficient
Effect
Coefficient
Effect
Coefficient
Effect
C onstant a n d D em ographics
Intercept
AGE
PLAN TENURE
1.00
-0.5478 ***
44.98
Y ear
9.35
Y ear
-0.0079 ***
0.0019
-1.0874 ***
-1.3171 ***
-0.0177
-0.0087 ***
-0.0194
0.0027
0.0021 *
0.0030
-0.0173
-0.0004
-0.0005
-0.0181
Male=1
-0.0829 ***
-0.0199
-0.0852 ***
-0.0204
-0.0774 ***
97.08
$K
-0.0062
-0.0007
-0.0076
-0.0009
-0.0168 ***
-0.0018
MEDIUM
0.46
-0.0051
-0.0012
0.0029
0.0007
-0.0165 *
-0.0038
RICH
0.29
-0.0188
-0.0044
-0.0068
-0.0016
-0.0367 ***
-0.0084
-0.0064 *
-0.0102
-0.0092 ***
-0.0102
MALE
0.55
HHJN C
II
E
3
-0.0079 ***
Rich=1
Plan Design
NFUNDS
NFUNDSQ
17.73
494.93
0.0001
0.0001 **
INDEX_OFFER
0.98
Y es=1
0.7831 ***
0.1183
0.7111 ***
0.1086
INTER_OFFER
0.98
Y es=1
0.0672
0.0152
0.0262
0.0059
C S JD F F E R
0.51
Y es=1
0.0172
0.0040
-0.0320
-0.0073
V BO_OFFER
0.08
Y es=1
-0.0029
-0.0007
-0.0252
-0.0057
LCJDFFER
0.45
Y es=1
0.0020
-0.0268
-0.0061
LOAN_OFFER
0.84
Y es=1
-0.0748 **
-0.0175
EE
0.95
Y es=1
0.1961 ***
0.0418
$M
-0.0162 ***
-0.0004
PLANBLN
466.50
0.0085
-0.0798 *
-0.0192
0.1475 ***
0.0313
-0.0074
-0.0002
-0.1161 ***
-0.0269
0.0439 ***
0.0559
0.0391
Prior Holding
W EB
0.64
NFUNDSJHELD
3.95
INDEX_HELD
0.59
Y es=1
0.1599 ***
INTERJHELD
0.24
Y es=1
0.1034 ***
0.0244
Y es=1
0.0495 *
0.0077
-0.0857
Y es=1
CSJHELD
0.33
VBO_HELD
0.01
Y es=1
-0.4762 ***
LC_HELD
0.11
Y es=1
0.1806 ***
LOANJHELD
0.10
Y es=1
0.0056
BLN
133.64
$K
0.0403
-0.0014
0.0117 **
0.0010
205,557
O bs
-log(L)
87,340
86,948
85,263
Pseudo-R 2
0.48%
0.93%
2.85%
Note: S e e Table 2-2. Plan d e sig n variables a re a s of 01/2003. The marginal effects of AGE, PLAN TENURE, H H JN C , NFUNDS,
PLANBLAN, NFUNDSJHELD a n d BLN respectively refer to th e ch an g e of probability of being Active R eb alan cer w hen a n investor's a g e
in c re a se s 10 y e ars, plan ten u re in c re a se s 6 y e ars, household incom e in c re a se s 6K, num ber of funds in c re a se s 10, plan balance
in c re a se s 50M dollars, num ber o ffen d s held in c re a se s 5 and balan ce in c re a se s 6K dollars, respectively. The marginal effects of MALE,
MEDIUM, RICH refer to th e probability of being Active R ebalancers a s m ale, in m edium or rich group of non-pension w ealth com pared
with reference group of female, an d poor group. Marginal effects of INDEX_OFFER, INTER_OFFER, C S_O FFE R , V B O J3F F E R ,
LC_OFFER, LOAN_OFFER, E E refer th e probability of being active trader in a plan th a t offers indexed equity fends, international fends,
com pany sto ck , brokerage option, life-cycle fends, loan and is em ployee contributory com pared with in a plan th a t d o e s not offer th e s e
investm ent options an d is not em p lo y ee contributory. The marginal effects of W EB, INDEXJTELD,
INTER_HELD,CS_HELD,VBO_HELD,LC_HELD,LOAN_HELD refers th e ch an g e of probability of being Active R ebalancer w hen a
participant h a s a registered internet acco u n t, holds indexed equity fends, international fends, com pany sto ck , brokerage option, life
c y cle fends, and loan co m p ared with w hen h e d o e s not hold th e s e investm ent options.
63
Chapter 3: How Life Cycle Funds Affect 401(k) Portfolio Behavior
As the responsibility for asset management in defined contribution (DC) pension

plans has shifted from professional fund managers to individual employees, plan
participants have looked for help in allocating their 401(k) plan assets. In response,
employers have in recent years added Life Cycle (LC) funds to their investment menus.
Under a LC framework, the participant is offered a choice o f funds o f funds in which
the mix o f underlying mutual funds is automatically rebalanced from time to time. There
are two major types o f Life Cycle funds, with the more traditional being the Static
Allocation (S A) approach, where the employee selects a hybrid fund that promises to
maintain a pre-set risk level, for instance, o f the Conservative, Growth, Aggressive
Growth, or Income varieties. A second, and increasingly popular, way o f constructing
Life Cycle funds is termed the Target Maturity (TM) approach, where the participant sets
his target retirement date and then chooses the investment manager to rebalance his
portfolio according to a predetermined glide path. That path generally requires the
participant to be more exposed to equities early in life, and then, as the target retirement
date nears, the participant is automatically rebalanced into an increasingly conservative
portfolio. Figure 3-1 illustrates some typical asset allocation patterns for each.
Figure 3-1 here
64
Despite the rising popularity o f Life Cycle funds in the DC environment,50 little is
known about how or whether this financial innovation had changed investors portfolio
allocation patterns. For instance, prior analysts have not evaluated how participants react
to the introduction o f LC funds and how these premixed funds affect pension investment
performance. Here we contribute to knowledge by using a unique new dataset covering
450,000 active 401(k) participants across 400 plans which added LC funds between 1/0312/05, to address several questions. First, we assess who elects LC funds and whether
workers select only these funds, or whether they include them in a broader investment
portfolio. Second, we evaluate how LC funds alter investors equity allocations in 401(k)
portfolios. Third, we examine how the introduction o f LC funds changes participants
and plans risk and return patterns. These questions are particularly salient in view o f the
2006 Pension Protection Act, wherein the US Congress identified Life Cycle funds as
suitable defaults for pension participants who failed to chose an asset mix on their own.
Our results show that workers find Life Cycle funds quite appealing, with onesixth o f existing workers electing them when they are introduced, and adoption rates o f
one-third for new hires. The LC funds prove most attractive to younger and lower income
workers, as well as women and the less affluent. Employees are more likely to adopt
Target Maturity date funds, compared to Static Allocation funds. Further, introducing the
LC funds boosts the equity fraction in workers portfolios across the board, with the
strongest effect among younger employees (under age 35) by some 20% points, and with
50 Poterba et al. (2006) identified some 250 Life Cycle funds in the mutual fund market in 2005; the
Investment Company Institute (ICI2006) estimated that almost half of all DC plans offered Life Cycle
funds in 2005 amounting to $120B in these instruments.
65
less o f an impact for older investors. Offering LC funds also changes participants riskadjusted returns. Using a CAPM framework, an employee in his 20s who adopted a Life
Cycle fund only would expect to earn 51 basis points more than without it; this positive
effect is smaller, at 27 basis points, for adopters over age 65. Additionally, selecting Life
Cycle funds reduces idiosyncratic risk both at the participant level and at the plan level,
since LC funds invest in indexed funds.
Our findings will be o f interest to researchers, as it is the first empirical study to
evaluate the quantitatively important impact o f introducing Life Cycle funds into the
401(k) menu. Specifically, Life Cycle funds do change investment patterns; further, they
help those who elect them earn more, and allow them to avoid incurring the effort o f
actively rebalancing their retirement assets periodically. This study will also interest
stakeholders charged with designing and managing 401(k) plans. Currently, most DC
plans offer money market funds as the default investment if employees fail to select
specific asset allocations. But earlier research has demonstrated that workers tend to stick
with the fund they are defaulted into (Madrian and Shea, 2001), so it would appear that
offering a Life Cycle fund as the default could provide workers with potentially large
risk-adjusted rewards.51 We also argue that our findings are relevant to research on the
self-control problems noted by behavioral theorists (c.f. O Donoghue and Rabin 2001).
Thus even present-oriented workers stand to reap potentially large future benefits when
offered a LC alternative.
51 In Australia and Chile, target maturity-type funds based on age have been the default allocation in DC
plans for some time.
66
In what follows, we briefly review research regarding LC models and funds.

Next, we describe our empirical strategy and hypotheses, followed by a discussion o f our
database and descriptive statistics. Subsequently, we present our empirical finding
showing how Life Cycle funds influence employees decision making outcomes and
evaluate how introducing LC funds changes behavior. A final section offers conclusions
and discusses future research avenues.
3.1 R elated Studies an d Hypotheses

In this section, we review a range o f theoretical and empirical models that speak
to how workers make their retirement portfolio allocation choices, focusing first on
perspectives from the finance literature, and next on hypotheses from the behavioral
economics literature.
Finance Perspectives. The finance literature is in some disagreement regarding how
investors should alter their equity fraction with age. For instance, Samuelson (1969) and
Merton (1969) argued that the rational risk-averse individual investing for the long term
should hold a fixed equity allocation over his lifetime (given i.i.d risky asset returns over
time). More recently, analysts have contended that equity shares should optimally
decline with age, due to the inverse correlation between human capital risk and
borrowing constraints (c.f. Bodie, Merton and Samuelson, 1988; Evensky, 1997; Gollier,
2001; Campbell, Cocco, Gomes and Maenhout, 2001; Brennan, 2002; and Campbell,
Gomes and Maenhout, 2005). This latter viewpoint therefore corresponds to an inverse
relationship between equities and age, consistent with the Target Maturity date approach
to investment. In practice the negative age/equity relationship is supported empirically
67
by Agnew, Balduzzi and Zunden (2003, p215) who find a negative effect o f age on the
equity share, concluding that each extra year translates into a lower allocation to stocks
by 93 basis points.. .remarkably close to the practitioners rule o f thumb o f decreasing
ones equity exposure by 1 percent for each additional year o f age. However, Ameriks
and Zeldes (2004) demonstrate using TIAA-CREF data that participants equity fraction
declines very little when investors age, suggesting that Static Allocation funds would
seem to be more appealing to the defined contribution plan investor. Accordingly, it
would be o f interest to establish more clearly the relationship between workers age and
equity allocation patterns in their 401(k) portfolios, and how these change when LC funds
are provided.
O f course from the financial perspective, adding one or several LC funds would
not be expected to change participants asset allocations, unless the added options
provide new assets that had not previously been available. In a CAPM world, workers
would need only to determine what fraction they would like to hold o f the market
portfolio and risk-free assets, both o f which are generally accessible in 401(k) plans.
Consequently, adding Life Cycle funds would not be expected to change portfolio
allocations, as they are simply linear combinations o f the preexisting assets already on
offer.
Only one cross-sectional study tests this hypothesis (Huberman and Jiang 2006),
and it argues that 401(k) participants do not follow a naive diversification strategy but
rather allocate their retirement assets equally across the asset classes they prefer (what
they call a conditional 1/N rule).
52 Mitchell, Utkus, and Yang (2005) offer a comprehensive analysis of the relationship between 401(k)
design and participant saving behavior.
68
Another point made by financial economics is that workers should rebalance their
accounts periodically, if they incur no transaction costs when they do so (Sharpe 1964,
Merton 1973). Even with positive transaction costs, investors would still be expected to
return to their target allocations, when the marginal benefit o f trading exceeds its
marginal cost (Grossman and Stiglitz, 1980). O f course higher transaction costs are
likely to discourage investors from rebalancing their portfolios (Vayanos 1998). In the
401(k) context, since participants do not generally incur any explicit commission or fee
for reallocating their portfolios, it might be thought that frequent trading would be both
easy and natural to do. Yet there may still be an advantage to Life Cycle funds, insofar as
automatic rebalancing will require less effort than remembering to reallocate ones
retirement portfolio periodically. Empirical analysis on cross-sectional 401(k) portfolios
shows that people do not trade much in their 401(k) plans, and passive rebalancers who
hold only LC/balanced funds earn higher risk-adjusted returns, compared to other
investors (Yamaguchi, Mitchell, Mottola, and Utkus 2006).53 W hether adding LC funds
to portfolio menus enhances participant returns is a question that we examine below.
Behavioral Economics Perspectives. The behavioral economics literature suggests

several additional reasons that workers might react positively to the provision o f LC
funds in the 401(k) menu. One reason is that these funds may help retirement investors
overcome a number o f problems associated with inability to make risky decisions. That
53 A few studies evaluate household investment and trading patterns in their portfolios but do not focus on
retirement accounts and rebalancing motivated trading (c.f. Odean 1999, Barber and Odean 2000, Odean
2001, Campbell et al. 2006, Grinblatt and Keloharju 2000, 2001; Odean et al. 2006; Guiso and Jappelli
2006). In general, these authors confirm that household investors tend to underperform professional
investors, and active trading generates worse investment performance, because individual investors trade
their asset due to their overconfidence.
69
is, some workers may procrastinate, forgoing making their investment allocation today
because they plan to do more research and learn more about their fund choices tomorrow
- but never get around to it (Samuelson and Zeckhauser, 1998; O Donoghue and Rabin,
1998; Choi, Laibson, Madrian and Metrick, 2001, 2003). From this perspective, firms,
policymakers, and others can create incentives to help present-biased people act in their
own long-run best interests, for example by benign paternalism. LC funds could
therefore play an important role.
There is ample evidence that workers display substantial inertia in their retirement
accounts, and that their asset allocation decisions in these accounts are apparently not
very sophisticated. For instance, Ameriks and Zeldes (2004) report that almost 75% o f
participants in higher education retirement plans never change their asset allocations over
a 10-year period, despite large changes in underlying asset market values. Agnew et al.
(2003) find that many participants in the large firm they studied allocated all o f their
assets either to cash or equity, and only 10% o f employees traded their assets each year.
Strong inertia also characterized 401(k) plan participants in Vanguards file o f 1.2 million
participants in 1,530 plans (Mitchell et al. 2006). Other researchers including Benartzi
and Thaler (2001) have argued that workers follow a naive diversification approach,
allocating their retirement funds equally across offered funds (1/N rule); subsequent
research confirms that even in plans with large menus, participants interpret a large
number o f equity funds as the employers endorsement o f equity investment (Liang,
Nellie and Weisbenner, 2006; Karlsson, Massa and Simonov, 2006).
A handful o f recent studies has sought to assess how employee saving behavior is
influenced by plan sponsor behavior, though most do not focus directly on how
70
investment performance might change after introducing new investment choices. The
study by Madrian and Shea (2001) detected large increases in enrollment rates after a
large firm introduced automatic enrollment; they termed this dramatic change the power
o f suggestion. Applying psychological principles, Benartzi and Thaler (2004) designed
a program o f automatic saving escalation called Save More Tomorrow, in which
employers invited their workers to pre-commit to save more when they received a salary
raise. They found that saving rates jum ped over four times for those who adopted this
mechanism. Analysis o f complex 401(k) menus by Iyengar, Huberman and Jiang (2004)
shows that offering large investment menus subject workers to information overload,
discouraging them from making investment decisions. Other studies find that workers
are often overconfident in their own knowledge about their firms, and/or believe that
employers endorse their own company stock when they put this stock in the 401(k)
investment menu (Mitchell and Utkus, 2004). Conversely, employees believing they are
more competent than average - due to having higher education or larger asset balances may be less interested in premixed asset allocations.
3.2 Empirical Hypotheses Flowing from the Literature.

No prior study has analyzed changes in investment patterns due to the
introduction o f LC funds, but from the previous literature just cited we draw several
testable hypotheses:
H I: Spanning Hypothesis: Rational savers will not alter their 401(k) investment
choices as long as introducing LC funds offers no new assets in the investment
menu and transaction costs are unimportant;
71
H2: Transaction Cost Hypothesis: Rational savers will find LC funds appealing in
their 401(k) plans, if these save on time and effort associated with rebalancing.
Additionally, younger employees with a longer saving horizon would elect LC
funds more than older ones, as they have longer to reap the benefit o f the costsavings.
H3: Investor Competence Hypothesis: Investors who believe they are more
skillful or knowledgeable than average would not use Life Cycle funds with pre
mixed allocations;
H4: Familiarity/Inside information hypothesis: Plan participants who have access
to their own employer stock in the 401(k) portfolio will be less likely to elect Life
Cycle funds if offered.
H 5: Procrastination/inertia hypothesis: Employees who anticipate suffering from
procrastination or inertia will tend to elect LC funds for their 401(k) plans so as to
avoid these future costs.
Table 3-1 summarizes how these hypotheses can be tested in our dataset, about
which we say more below. In particular, the Spanning Hypotheses (H I) would predict
that adding LC funds would have no impact as long as transactions costs are small (or are
controlled with a set o f socioeconomic variables to be discussed below). A less stringent
test (H2) would recognize that younger participants would be likely to adopt LC funds,
because they will benefit more from automatic rebalancing due to their longer investment
horizons, and perhaps TM over SA funds, if they take into account their human capital
profiles. To the extent that lower-income, less wealth, and female employees are less
financially literate, they would be predicted to be more likely to adopt LC funds (H3;
72
Heath and Tversky 1991; Lusardi and Mitchell, 2007). As plan tenure may be a proxy
for familiarity with the plan, longer tenure and the offering o f employer stock would be
expected to disincline employees from electing LC funds (H4). N ew hires might,
however, be more likely to elect LC funds, if existing workers suffer from inertia (H5).
Lastly, as these factors surely might take time to work, we predict that Life Cycle fund
adoption rates would start low and rise over time since inception.
Table 3-1 here
3.3 How Introducing LC Funds Changes Asset Allocation: An Overview

In this section, we provide an overview o f how 401(k) plan participants altered
their asset allocation choices when Life Cycle funds were newly introduced, followed in
the next section by multivariate analysis. Our dataset consists o f a three-year panel o f
401(k) plans and participants drawn from the Vanguard recordkeeping system over the
period 1/03-12/05. It encompasses 1,548 defined contribution plans covering more than
1.7M active 401(k) participants, and provides rich information on participants
demographic and economic characteristics (sex, age household income, and household
nonpension wealth); contribution amounts by each participant to each fund by month; and
plan design features including the number o f funds offered, when and which Life Cycle
funds where adopted, and whether the employer offered company stock each month.54
We focus on plans that added a Life Cycle fund during the sample period. We have
confirmed that the LC addition did not alter the range o f assets provided in the
54 For a more detailed description of the Vanguard dataset see Mitchell, Mottola, Uktus, and Yamaguchi
(2006a and 2006b).
73
investment menus, as all plans previously included at least one U.S stock index fund and
one U.S bond index fund, and 98% also had an international equity index fund.
Accordingly similar investment frontiers were available both before and after the
introduction o f the LC choices.
Further information on what specific changes occurred in the plan investment
offerings is as follows:
1) 40% o f the plans added TM but not SA funds (None/AddTM);
2) 17% o f the plans added SA but not TM funds (None/AddSA);
3) 26% o f the plans added TM and kept existing SA funds (KeepSA/AddTM); and
4) 17% o f the plans added TM but dropped existing SA funds (DropSA/AddTM).
Therefore cases 1), 3) and 4) all added Target Maturity funds, while case 2) added only
Static Allocation funds. Interestingly, although more than 100 plans added both SA and
TM funds, most o f the new Life Cycle fund holders elected the Target Maturity date
funds. In terms o f timing, in 2003 most o f the plans added SA funds, but in 2004 and
2005 most added TM. In our dataset, we observe the plan on average for ten months
after the LC fund was introduced.55
Table 2-2 presents the adoption rate for LC funds among participants. We see that
18% o f participants selected Life Cycle funds when they were offered; o f these, just
under half (8%) chose only Life Cycle funds, while the other group (10%) also elected
other funds. N ew enrollees are more likely to use LC funds compared to existing
55 When a 401(k) plan drops a fund from its investment menu, it will usually announce the menu change in
advance and ask participants to move their contributions and account balances. If an employee makes no
move, the employer will switch the money to a default fund, usually the identical asset type as the fund
replacing the one that was dropped.
74
employees (35% vs. 15%). There is a clear pattern cross different plan types. The
adoption rate for a plan adding TM (column 1) is much higher than in plans introducing
SA (column 4). This difference becomes much clearer if we focus on existing
participants or new enrollees only.56 Both the overall adoption rate and the fraction o f
participants contributing to only Life Cycle funds is twice that when TM funds are
adopted, versus in firms SA, for existing workers.
Table 3-2 here
To illustrate the impact o f offering LC funds, Figure 3-2 plots the adoption rate
over time for existing workers. Panel (a) refers to plans that introduced TM or SA from
afresh - that is they had no LC options previous; panel (b) refers to plans that added TM
on the top o f preexisting SA funds. The evidence clearly shows that participants display
some inertia: as time passes, they are more likely to adopt LC funds in both cases. Further,
as time passes, the adoption rate rises more quickly for the Target Maturity versus the
Static Allocation Funds. Panel (b) reveals much higher overall adoption rates, because
when TM was added, substantial numbers o f participants already held the SA funds.
Figure 3-2 here
O f most interest is an examination o f how workers portfolio allocations to equity
change as a function o f the LC fund introduction date. Here we focus on how workers
allocate their contributions, or new money to the funds, because contributions are usually
considered as desired allocations57 reflecting workers preferences and/or expectations
56 Here we define new enrollees as those whose first contribution is observed after the plan introduced Life
Cycle funds.
57 For example, Agnew et al. (2003), Mitchell et al (2006a, 2006b).
75
regarding risky asset returns. To pinpoint the before date, we use the month before the
LC funds were adopted, and the after period is immediately after the LC funds were
CO
adopted. Figure 3-3 illustrates age patterns o f adoption by Life Cycle Only.
Prior to
the adoption o f LC, the age pattern o f equity allocation displays a hump shape by age.
After the inception o f LC funds, the fraction in equity rises dramatically for younger
workers, particularly for those who elected only LC funds. In fact, the fraction o f the
portfolio in equity is about 20% for those under age 35. (Smaller changes in overall
equity allocations are detected among those who held LC as well as other funds).
Thereafter, equity allocation fractions fall nearly monotonically with age.
Figure 3-3 here
3.4 Multivariate Results

Next we present multivariate analysis on three key questions: (1) who selects Life
Cycle funds; (2) what impact do Life Cycle funds have on participants equity allocation;
and (3) how is investment performance affected by offering LC funds. Two sources o f
identifying variation are available in our sample. The first arises from the cross-sectional
difference between participant elections and their characteristics as well as plan design
features; the second arises from the before and after experiment for plans introducing
Life Cycle funds. Accordingly, we use the full sample to analyze LC adoption rates,
because we seek to investigate how they differ between existing employees and new hires.
58 We also check the equity allocation changes of those not electing LC funds, as well as workers in plans
lacking a LC option and in plans that had LC funds from the beginning o f the period. In all these cases,
only 10% o f workers changed their equity allocations in their contributions.
76
We use existing employees to analyze the impact on equity allocation using a

difference-in-difference tactic - that is, we compare not only cross-plan differences,
but also exploit before and after changes. Finally, to evaluate the impact on
investment performance, we focus separately on LC Only Choosers and Other LC
Choosers.
The Impact of Offering LC Funds on Fund Election. When offered a new LC set o f
funds, the employee has a two-fold decision problem. First, he m ust decide whether he
will invest in the new funds; and conditional on opting for LC, he must decide whether he
should only use the Life Cycle funds or combine them with some other fund choices. To
explore who and how employees are using Life Cycle funds, we evaluate whether and
how employee characteristics influence the probability o f adopting the Life Cycle fund.
We also hypothesize that plan design features will influence portfolio outcomes, and we
also investigate whether offering Target Maturity versus Static Allocation funds influence
workers allocation decisions differently.
For a participant i in plan j at month t after his plan introduces Life Cycle funds,
we relate the binary response (LCChoseniJt) to a set o f demographic and economic
explanatory variables (X,), a plan design vector (Zj) and a Life Cycle fu n d s treatment
vector (LCj) with the following multi-level probit model:
Pr (LCChosenjjit= l \XhZj.LCj) = a X j+ flZj + y ,L C j+ u,+r}j+ Tt+iJit
(3-1)
The dependent variable LCChosenjjj is set to 1 if a participant chooses any Life Cycle
funds, and else it is 0. The vector X, includes participant-specific variables such as age,
77
plan tenure, household income, plan account balance, non-retirement wealth,59 and a
dummy variable to indicate whether he is a new enrollee or not. The vector Z,- represents
each plans characteristics including the number o f offered funds, the number o f
participants, and binary indicators for whether the plan offers company stock and permits
the worker to take out a loan against his fund balance. Time is a counter set to 1 in the
month that the LC fund is introduced, 2 for the following month, and so forth. LCj is a
treatment effect indicating the LC fund was offered, and the definitions o f
None/AddTM, KeepSA/AddTM, and DropSA/AddTM are as in Table 3-2.
Three versions o f this empirical model are estimated in practice, and the marginal
effects o f changes in the key variables appear in Table 3-3. Model A includes only
employee demographic and economic characteristics; Model B adds plan design features;
and Model C adds the Life Cycle treatment and time effects.60 Overall, the coefficient
estimates are quite statistically significant, suggesting that offering LC funds does alter
participants fund selection. Accordingly we reject H I, the spanning hypothesis. There is
support for H2, the transaction cost hypothesis, in that the marginal effect o f age is
significantly negative:61 specifically, the LC adoption rate rises by 0.5% as age falls by
10 years The negative age impact may be partially explained by transaction cost:
59 Data o f non-retirement financial wealth groups was based on the categorical variables collected by IXI, a
financial information provider. We imputed the original wealth classes into dollar term, and then collapsed
into three groups: Poor (wealth<$7,280), Medium ($7,280 to $61,289), and Rich (>$61,289).
Because o f the multi-level data structure, we adjust for serial correlation and heteroskedasticity o f error
disturbances clustered at the participant level (u<), control both plan design characteristics and the plan
average o f participants demographic and economic characteristics in order to capture plan-level error term
( r jj) , and include time dummies to control the fixed effect occurred in the same month ( t J , finally we
assume the random error term e,7i, as normally distributed. We also control on eight industrial sector
dummies and missing dummies for all demographic and economic factors.
61 The regression also includes a squared term on age; the higher order term is included when computing
the marginal effect o f age.
78
holding other things constant, younger workers may benefit more from automatic
rebalancing due to their longer investment horizon. In addition, the adoption rate in plans
adding TM is higher than in plans adding SA, suggesting that the way TM funds
rebalance investors portfolios by age is more appealing that SA funds which do not.
Table 3-3 here
The results also provide support for several other hypotheses. Focusing on H3, the
investor confidence hypothesis, we see that LC adopters are likely to be female, less
wealthy, have shorter plan tenure, and smaller account balances. For example, one
standard deviation increase in plan tenure (7 years) reduces the probability o f adopting a
LC fund by about 30%. Conversely, men, longer tenure employees, and the better off
may believe they are more skilled and knowledgeable investors, and so they are less
likely to select Life Cycle funds. H4 spoke to the thesis that some workers invest based
on perceived familiarity, and we find support for that view as well: specifically, plans
which offer company stock have lower LC adoption rates. This is consistent with Brown,
Liang and W eisbenner (2006) who suggest that 401(k) participants believe they can
predict the returns o f their company stock due to having inside information. Next, the
procrastination/inertia hypothesis is also confirmed (H5), in that female, less wealthy,
lower income, and younger workers are more likely to procrastinate according to
Madrian and Shea (2001). Our evidence shows that Life Cycle funds with pre-mixed
asset allocations are more attractive for these individuals. Across all specifications, new
enrollees are much more likely to use LC funds; that is, other things constant, 23% o f
new enrollees would elect LC funds compared to 10% for existing workers. Another
interesting result is that the LC adoption rate rises over time since introduction: that is,
79
three months after inception, the adoption rate rises by 0.5% points. This is consistent the
view o f O Donoghue and Ratin (2001) who posit self-control may take a while to
implement.
Next we turn to an assessment o f whether the workers investing only in LC funds
differ from those who hold LC as well as other funds. The LC Only chooser can be seen
as a pure strategy investor, while other investors might have most o f their portfolios
rebalance automatically but also seek risk premiums that Life Cycle funds do not capture.
The following probit analysis is employed to explore this difference:
P r (LCOnlyjjj=l \XitZj,LCj LCChosenijX= l)=a Xi+PZj + f L C j + vt+ r/j+ Tt+siJit
(3-2)
where LCOnlyijj is set to 1 if a Life Cycle funds user contributes only to LC funds (0
else). A set o f controls for demographic, economic, plan design, and treatment effects are
included as in Table 3-3, and the error structure is as in equation (1). If the coefficients
and marginal effects are not significant individually or jointly then the two groups do not
differ from each other.
Results appear in Table 3-4. Here we provide the marginal effects o f specific
variables on the probability o f being a LC Only chooser, versus not, conditional on
selecting a Life Cycle fund. It is clear that Life Cycle Only holders differ from those who
also use other funds, as many coefficients are statistically significant from zero.
Specifically, women, those with less tenure, less financial wealth, smaller account
balances, and who have a plan without employer stock or loan options, are more likely to
devote their 401 (k) contributions to LC funds only. Further, probably due to inertia,
existing employees tend to simply add LC funds to their other holdings. Also, a worker
80
offered only a Target Maturity date plan is 20%-50% more likely to hold Life Cycle
funds alone, compared to those offered Static Allocation funds. Finally, as time passes,
participants are increasingly likely to switch all their contributions entirely to Life Cycle
funds. In particular, the probability o f being a Life Cycle only chooser three months after
Life Cycle funds were offered increases by about 10%, compared to the LC inception
date.
Table 3-4 here
In summary, the evidence suggests that adding Life Cycle funds to the 401(k)
plan menu significantly alters participants fund elections. Lifecycle funds are more
likely to be adopted by new hires, females, lower wage and lower wealth workers, and
those with shorter plan tenure, probably because they are less confident and more likely
to value the investment advice inherent in the premixed funds. W hen company stock is
available, workers are less likely to invest in Life Cycle funds, maybe because they feel
more confident about investing in that stock. Finally, adoption rates o f LC funds rise with
time since inception.
The Impact of Offering LC Funds on Portfolio Allocation. Next we evaluate how

offering LC funds alters employee equity allocation patterns. As mentioned above, there
is a debate in the finance literature about whether people change their percent in equity as
they age; similarly, there is a debate in the behavioral finance literature regarding how
plan design alters workers investment choices. For instance, researchers have
alternatively argued that 401(k) participants follow a 1/N rule, or a revised 1/N rule, or a
81
1/N rule limited only to equities, or a conditional 1/N rule.62 Yet no study has examined
both the age effect and the menu effect in a single analysis,63 and no study has explored
how investment vehicles such as Life Cycle funds might change portfolio allocations.
Accordingly, we are interested in asking whether and how the fraction o f investors
portfolios changes due to the introduction o f LC funds in 401(k) plans, and how these
change with age.
To evaluate this question we exploit cross-sectional and time-series variation,
before and after LC funds are introduced with a difference-in-difference estimator
(Todd, 2006). This can be stated as two-way fixed-effect model as follows:
EQ%ij, t,a=ccXi+PZj + 8 PostLCjtt+yiNone/AddTM)+y2KeepSA/AddTMj
+/3DropSA/A ddTMj
+y4None/AddTM*PostLCji(+y5KeepSA/AddTMj*PostLCjit
+ ytDropSA/AddTMj
(3-3)
+Vi+ T lj+ T t+ S iji t .
Here EQ%iJ:t}a represents the equity allocation o f participant i in plan j in month t\ we

also evaluate each age group a separately. Because a Difference-in-Difference analysis
requires observations both before and after the treatment, we focus on existing
participants and limit our attention to the [-6, +6] month time window concentrated
62 Benartzai and Thaler (2001) argue that participants allocate their contributions equally to all offered
funds (1/N); Brown, Liang and Weisbenner (2006) argue that the fraction contributed to each class is
significantly positively related to the offered fraction o f that asset class in the investment menu (revised 1/N,
i.e., framing effect); Geotzmann, William and Kumar, Alok (2001) find that the 1/N rule holds only when
the set is constrained to offered funds o f a given type; Huberman and Jiang (2006) claim people allocate
their 401 (k) assets equally to the funds they chose (conditional 1/N rule).
63 Agnew et al (2003) discuss equity allocation by age, but because they only focus on one plan, they
cannot identify the framing effect from variation in menus.
82
around the date o f the adoption o f the Life Cycle fund.64 Controls include demographic,
economic, and plan design vectors, as before.; we also include PostLCj,t , a dummy
variable that is set to 1 if month t comes after the LC introduction month. The variables
None/AddTMjt KeepSA/AddTMj, DropSA/AddTMj refer to the treatment effects that
capture cross-sectional variation, while the interaction terms capture the difference-indifference effects indicating whether Life Cycle funds still have measurable impacts on
equity allocation patterns after controlling on time and cross sectional patterns.65
Table 3-5 reports the empirical findings from equation (3-3), with the columns
indicating the effects differentiated by age group. It will be recalled that, according to
hypothesis H I, adding Life Cycle funds should have no impact on workers equity
holdings if they are simply seen as a linear combination o f the same funds already
available. Clearly this hypothesis is rejected, as the LC treatment effect (S) is positive
and statistically different from zero, and is relatively similar across age groups. For
younger people the increase in equity allocations is about 2%, and for older people about
4-5%. We also find that all TM treatment variables (//,
72, 73)
are positive and significant,
implying that plans adding Target Maturity funds have more o f an equity increase than
plans which add SA funds. We also find that most o f the regression coefficients on
difference-in-difference factors ( 74,
75,
76) are not statistically different from zero,
implying that portfolio equity allocations do not differ between participants adopting TM
64 We checked the contribution allocation change after a participant chose Life Cycle funds. Over 90% of
the Life Cycle fund choosers continuously use Life Cycle funds; some o f Life Cycle funds choosers
switched their contribution allocated to other funds to Life Cycle funds, i.e., switched from other Life Cycle
funds users to Life Cycle funds only user but did not change the equity allocation.
65 Again, we apply clustering technique to capture participants heterogeneity, control the plan design and
the plan average o f demographic and economic features to absorb the error terms at the plan level, and add
the time dummies to capture time fixed effects. All regressions are estimated by OLS.
83
versus SA funds, holding other things constant. Evidently equity allocation fractions in
plans that adopted TM were higher to begin with, compared to participant equity
allocations in plans that added SA funds. In summary, introducing Life Cycle funds does
boost equity allocation o f 401(k) participants, and there is little difference across TM and
SA funds.
Table 3-5
Economic Impact of Life Cycle Funds. Next we ask whether and how Life Cycle
funds alter expected investment returns for participants and plans. We do not observe
investors realized returns over long time periods, so instead we study the properties o f
employees 401(k) portfolio allocations by estimating the moments o f asset returns and
then we impute the first and second moments for all individuals portfolios. Following
Campbell et al. (2006), we calculate factor loadings for each o f the underlying assets held
by the 401(k) participants in our sample using fund returns from 1/98-12/05, which is the
longest period available to estimate fund betas. The funds in our dataset include a wide
variety o f domestic and international stock funds, bond funds, balanced, and Life Cycle
funds. For the modified CAPM model, we take this breadth o f asset choices into account
by regressing the excess return66 for each o f the 401 (k) funds in our universe on three
market indices: the value-weighted CRSP portfolio, the Lehman Brothers Aggregate
Bond Index (LBA), and the Morgan Stanley Capital International (MSCI) Europe,
Australia and Far East (EAFE) Index. These, respectively, represent the US equity
66 Throughout the paper, all returns for participant portfolios and market indices are given as excess
returns calculated with US Treasury Bill returns as the risk-free investment.
84
market, the US bond market, and the international equity market. This regression
function can be written as:
R n ,t - R f ,t = PlRcRSPRF, t> + P 2RLBARFJ+P 3 R m SC IR F j+n,t,
(3 -4 )
where R yt is the total nominal net return o f fund n in month t, and R f ity is the risk free rate,
R c r sp r f \t, R lb a r f j ,
and R m s c ir f j are excess returns on the CRSP value-weighted market
portfolio, LB A and M SC IEA FE indexes, respectively.
P i,
/Tând /7?are the regression
coefficients or factor loadings; siyt is the error term.

A
From equation (3-4), we obtain the risk loading matrix B = (bx,...,bn)', and bn is
A
the estimated loading vector o f fund n, which can be written as bn - (/?", /?2 , P " )'.
Taking the average o f factor vector Rf = (RCRsprf t R-lbarfj >Rmscirfj ) an<3 the varianceA
covariance matrix I,f o f factors, we compute the risk-adjusted excess return R e o f fund n
and the variance-covariance matrix o f all assets E as follows:
K = b ' R f
A
A , A
(3-5)
A
E = B 2 yB + D
where D is the diagonal matrix with diagonal elements computed as the square o f
A
s n estimated from equation (4). Here, we can decompose the variance E into systematic
risk E'yv = B E; B and idiosyncratic risk b do.
Finally, using the first and second moments o f all assets estimated from equation
(3-4)-(3-5) and the weights in the participants portfolio, we impute the risk-adjusted
85
returns and variance o f the asset allocation before and after he chooses Life Cycle
funds. We can then assess the impact o f the change using the following:
:=2
>a '
n
(3-6)
=1
Z, = a>w a>t/t, = tolM$ r
+ b id)G>^t =
where o)int is the weight o f fund n in participant V s contribution portfolio, and t takes
two values which indicate before and after LC adoption, respectively. We can also
decompose the variance o f every individual portfolio E, into the systematic risk Zf and
the idiosyncratic risk E ' 0 .
To explore how LC fund adoption affects participants, we focus on the subset o f
workers who elected the LC funds (65,561 in total), o f which a third are Life Cycle Only
choosers; the remainder also elected some other funds. Table 3-6 summarizes results,
where the first column presents the change in risk-adjusted returns for Life Cycle Only
choosers by age group. The results show that Life Cycle Only choosers improve their
risk-adjusted returns by selecting LC funds, and the improvement is significant and large
for younger workers (Column l).67 For example, if an investor in his 20s switched all o f
his contributions to LC funds, he would be expected to earn an additional 51 basis points
per year above his baseline; this reward falls to 27 bps if an investor switched over at the
age of 65. This makes sense because younger switchers - particularly those who move to
a Target Maturity date fund - hold more in equities than older participants. Moreover,
Columns 5 and 8 show that Life Cycle fund choosers reduce their idiosyncratic risk
67 The positive effect is not significant for those who mix LC funds with other assets (Column 2).
86
significantly. For an instance, those electing Life Cycle Only portfolios experienced an
increase in the portfolios annual variance from 12% to 17%; however, their nonsystematic risk share fell by 10 percentage points, and the effect is most marked for
younger investors.
Table 3-6 here.
To explain the why expected investment performance changes in Table 3-6, we
decompose the sources o f risk into two portions: one has to do with a change in the share
o f risky assets, and the other pertains to different risk loadings. We implement this by
classifying all the funds available to investors in our sample according to their holdings
o f equity, bonds, and cash,68 and then we define the equity and bond holdings as risky
assets. Table 3-7 shows how asset allocations changed on the introduction o f LC funds
by age, for Life Cycle Only choosers versus the other Life Cycle choosers. Not
surprisingly, investors under age 55 increased their equity exposure but reduced their
share o f cash and bonds; by contrast, investors over age 55 shifted from stocks to more
cash and bonds. Interestingly, all investors (even the older ones) reduced their cash
holdings. For example, before the LC funds introduction, the over-65 investor allocated
11% o f his contributions to cash; after selecting the LC funds, he increased his bond
holdings by 13% (his equity fraction did not change much). This portfolio change is a
key reason why even the older investors improved their risk-adjusted returns. Columns
(7-9) and (10-12) illustrate the changes in risk loadings estimated from equations (4-6)
for Life Cycle Only choosers and other Life Cycle fund users. For both groups, most o f
68 Money Market funds as are treated as cash and balanced funds are allocated into equity and bond
holdings using each funds reported targeted allocation from its prospectus.
87
the under-age 55 investors boosted their domestic stock /?/ (Columns 7 and 9) and
domestic bond loading
(Columns 8 and 10), although the magnitude depends on the
investors age. Another fact worth noting is that LC funds raised all investors
international equity exposure. For example, the international stock market loading (f3$ o f
Life Cycle Only choosers under age 55 climbed to twice what they held before LC funds
were adopted (Column 9).
Table 3-7 here
Having shown how introducing LC funds influences behavior o f those electing
these funds, next we assess how offering Life Cycle funds influence plan-level outcomes.
Using the same methodology as described above, we evaluate some 600,000 individual
portfolios and decompose the change in investment performance into two parts: the
portion due to participants adopting Life Cycle fluids, and the portion attributable to the
fact that workers electing Life Cycle funds change their risk-adjusted returns and
variance. Table 8 presents estimation results. Over all, adding LC funds does not appear
to yield significant plan-wide improvements in total risk-adjusted returns, but it does
reduce idiosyncratic risk by 1.5% points. Plans adding TM have a higher Life Cycle
adoption ratio, but plans adding SA benefits appear to have higher risk-adjusted returns,
mainly workers in plans that add SA had less equity to begin with. Across all plans, the
major changes in risk are attributable to a notable increase in international equity
loadings (column 7). On average, 401 (k) plans increased their exposure to foreign stock
markets by about 13% when adding LC funds, compared to beforehand (from 8.3% to
9.4%). By offering LC funds, plan sponsors can therefore help investors diversify their
88
retirement portfolios internationally better than they would have done on their own. As a
result, at the plan level, offering LC funds will reduce non-systematic volatility.
Table 3-8 here
3.5 Discussion and Conclusions

In recent years, many employers have added Life Cycle (LC) funds to the set o f
investment options offered to the employees covered by defined contribution pension
plans at their firms. Traditional financial economics suggests that offering a new fund
should not alter participant asset allocation as long as the existing fund menu is welldiversified. Yet previous studies have found that pension plan design can influence
participant portfolio selection. Here we use a unique new dataset on almost half a million
401(k) plan participants to assess whether offering these premixed investment choices
influences workers portfolio selection behavior.
Our results show that workers find Life Cycle funds quite appealing, with onefifth o f existing workers electing them when they are introduced, and adoption rates at
two-thirds for new hires. The LC funds prove most attractive to younger and lower
income workers, as well as women and the less affluent. Employees are more likely to
adopt Target Maturity date funds as contrasted to Static Allocation funds. Further,
introducing the LC funds appears to boost the fraction o f the workers portfolios held in
equity across the board, with the strongest effect among younger employees (under age
35) by some 20% points but less o f an impact for older investors. Offering LC funds also
changes participants risk-adjusted returns. Using a CAPM framework, an employee in
his 20s who adopted a Life Cycle fund would earn 51 basis points more than prior to its
89
adoption; this positive effect amounts to 27 basis points for adopters over age 65.
Additionally, selecting Life Cycle funds reduces idiosyncratic risk both at the participant
level and at the plan level, since LC funds invest in indexed funds.
Our findings should be o f great interest to plan sponsors, policymakers, and fund
managers. We conclude that the support given to Life Cycle funds by the recently passed
2006 Pension Protect Act is likely to be beneficial to workers. Having the default be a LC
fund does not put older workers at greater risk, and it can improve younger workers riskadjusted returns while reducing non-systematic risk for participants and plans.
Future research avenues may be suggested. We need to investigate whether plans
that offer LC differ from those that do not;69 it would be interesting to measure the planwide impact o f LC introduction; moreover, we also need to track the LC impact over
longer time periods. It will be useful to examine whether offering LC funds reduce
participants trading activity, since professional managers would rebalance these
portfolios regularly. We will also assess whether the impacts we have found on
investment performance are robust over a longer time horizon. Last, we seek to evaluate
in more detail why apparently similar individuals offered Target Maturity date funds are
more likely to switch than Static Allocation funds, and relatively few offered either type
actually opt for these menu choices. Future research is needed to determine whether this
is due to procrastination, financial illiteracy, or because the limited pre-mixed asset
allocation patterns do not match participants preferred risk-retum profiles.
69 When we compare the characteristics of plans in our sample that introduced LC vs. those that did not, the
plans appear similar on most counts. At the plan level, the average size is about 1,100 thousand active
participants with a total balance o f $50 million. Characteristics o f their work forces, average age,
household income and non-plan wealth are also very similar.
90
Table 3-1. Hypothesized Impacts of Offering Life Cycle Funds on Employee

Outcomes.
Notes: LC, TM, and SA respectively refer to Life Cycle funds, Target Maturity funds, and Static Allocation
funds. Age and Py stand for a participants age and his job tenure; Male refers to the participants sex;
H h jn c is the households annual income imputed from ZIP code data provided by Claritas; Poor and
Medium indicate non-retirement financial wealth groups from IXI data grouped as follows: Poor
(wealth<$7,280), Medium ($7,280 to $61,289), and Rich (>$61,289). Balance refers to the participants
account balance; New is a dummy variable set to 1 if a participants first observed contribution month was
after his plan introduced the LC fund. CS_offer and Loan_offer are dummies set to 1 respectively if the plan
offered company stock or loans (else 0). Nfunds refers to the number o f funds offered, and Num_prt the
number o f participants. None/AddTM refers to the plan which had no LC and then introduced TM;
None/AddSA refers to a plan that had no LC and then introduced SA; KeepSA/AddTM is a plan that kept its
existing SA and added TM; DropSA/AddTM refers to a plan that dropped its SA and introduced a TM fund.
Time is a time index where the LC-introduction month equals 1, the following month as 2, etc.
Probability of
Being LC Chooser
Change of Equity
Allocation
Change of Riskadjusted Return
Dem ographic and E conom ic Factors
Age
Py
Male
Hh inc
Poor
Medium
Balance
New
Netural
Plan Design
CS_offer
Loan offer
Neutural
Nfunds
Num_prt
Lifecycle F und Treatm ent
None/AddTM
KeepSA/AddTM
DropSA/AddTM
Time
Neutural
Neutural
91
Table 3-2. Adoption Rate of Life Cycle Funds.

New hires are those whose first observed contribution is after the introduction of LC in their plan. LC
Only Choosers are those who contribute only to LC. Other LC Choosers are those who contribute to
several funds including LC. Data are as o f the last observed contribution month over the sample period.
Notes and definitions o f other variables: See Table 1.
Total
Number of Participants ('000)
None/Add
SA
(1)
(2)
Keep
Drop
SA/Add TM SA/Add TM
(3)
(4)
Total
(5)
203.3
77.8
112.0
63.8
456.9
14.2
11.0
20.6
34.0
18.0
LC Only Choosers
4.7
3.8
12.3
17.8
8.2
Other LC Choosers
9.6
7.1
8.3
16.3
9.8
181.2
60.7
86.9
57.6
386.3
11.5
5.4
17.1
32.4
14.9
LC Only Choosers
2.2
0.9
8.9
16.1
5.6
Other LC Choosers
9.3
4.4
8.1
16.3
9.3
Number of Participants (000)
22.1
17.0
25.1
6.2
70.5
Adoption Rate (%)
36.7
30.8
32.7
49.5
35.0
LC Only Choosers
25.0
14.2
23.9
33.5
22.7
Other LC Choosers
11.8
16.6
8.8
16.0
12.3
Adoption Rate (%)
Pre-existing Number of Participants ('000)

Participants
Adoption Rate (%)
New Hires
None/Add
TM
92
Table 3-3. Marginal Effects of Determinants of Choosing LC Funds.

This table reports marginal effects estimated from a multi-level Probit model: Pr (LCChoseniji,)=a,Xi+p,Zj
-HyLCj+Oi+rij+Tt+Sij t The dependent variable, LCChosen, is a binary response term set to 1 if participant i
in plan j in month t after his plan introduces a life cycle fund selects it (else =0). Xt is a vector of
participants demographic and economic characteristics defined as in Table 1, with additional variables
Index_offer, lntl_offer, CS_offer Loan_offer which are dummy variables set to 1 if a plan offers indexed
domestic equity, international equity, company stock and loans (else 0). Serial correlation and
heteroskedasticity o f error disturbances are adjusted at the participant level (u), and both plan design
characteristics and the plan average o f participants demographic and economic characteristics are
controlled to capture plan-level errors (rjj). Time dummies control for month fixed effects (r,.); s^, is
assumed normally distributed. The marginal effects o f Age, Plan tenure, H hinc, Balance, Nfunds,
Num_prt, and Time refer to the change in probability o f LC being selected if, respectively, the investor is
10 years older; plan tenure rises 6 years; household income is $6,000 higher; the participants balance
increased by $3,000; the plan added 5 funds to the investment menu; the number o f participants increased
by 500; and time since LC is introduced rises 3 months. Marginal effects o f Male, Poor and Medium refer
to the change in probability o f choosing a life cycle fund for men, or for people in poor or medium non
pension wealth groups (vs Female and Rich). Marginal effects o f Index_offer, Inter_offer, CS_offer, and
Loan_offer refer the probability o f selecting a LC fund if the plan offers indexed equity funds, international
funds, company stock, and loans, versus a plan that lacks these options. Marginal effects o f None/AddTM,
KeepSA/AddTM, and DropSA/AddTMrefer the change in the probability o f selecting a LC fund in a plan
that added TM without SA, or added TM and kept SA, or added TM but dropped SA compared to a plan
that added SA. Marginal effects o f Age, Plan Tenure, and Time also include squared terms (which are
statistically significant). The values of Hh inc, Balance and Num_prt are taken in natural log form.
Missing dummy variables for all independent controls are also included. ***(**) indicates coefficient
differs from 0 at the 1% (5%) level.
93
Dependent Variable: 1= LC Chooser, 0=Others Mean: 14.1%

Mean
Model A
Model B
Model C
(1)
(2)
(3)
(4)
Demographic and Econom ic Factors
Age
43.10
-0.0054 ***
-0.0034 ***
-0.0042 ***
Py
7.55
-0.0357 **
-0.0358 ***
-0.0334 ***
Male
0.44
-0.0237 ***
-0.0038 ***
-0.0043 ***
-0.0010 **
-0.0009 **
Hhjnc
74.84
0.0002
Poor
0.38
0.0298 ***
0.0137 ***
0.0141 ***
Medium
0.37
0.0148 ***
0.0109 ***
0.0114 ***
Balance
18.96
-0.0027 ***
-0.0024 ***
-0.0008 ***
0.11
0.0971 ***
0.0928 ***
0.1310 ***
New
Plan Design
CS_offer
0.30
-0.0075 ***
-0.0015
Loan_offer
0.66
-0.0533 ***
-0.0186 ***
Ntunds
32.8
-0.0097 ***
-0.0165 ***
3,019
-0.0022 ***
Num_prt
0.0001
Lifecycle Fund Treatment
None/AddTM
0.32
0.0189***
KeepSA/AddTM
0.32
0.2513 ***
DropSA/AddTM
0.10
0.2117 ***
Months Since LC Added
0.0035 ***
Time Fixed Effect
No
Yes
Yes
Industry Fixed Effect
No
Yes
Yes
Plan Average of Demographics
No
Yes
Yes
Yes
Yes
Yes
4,973,086
4,973,086
4,973,086
456,869
456,859
456,869
4.6%
7.7%
11.5%
Clustering at Participants
Observations
Number of Participants
R-Squared
94
Table 3-4. Marginal Effects of Choosing LC Only.

This table reports the marginal effects estimated from a multi-level Probit model o f the form: Pr
(LCOnlyij t= l| LCChosenij t= l)= a ,Xi+PZj -HyLCj +Ui+r)j+Tt+8yitwhere the dependent variable, LCOnly, is
a binary response, which is set to 1 if a participant i in plan j in month t after his plan introduced LC selects
only the LC fund, conditional on selecting any LC fUnd (else =0). Independent variables and errors are
handled as in Table 3). Marginal effects refer to the change in probability being a LC Only Chooser and
independent variables are given in Table 3.
Dependent Variable: 1= LC Only Chooser, 0=O ther LC Chooser Mean: 46.4%
Mean
Model A
Model B
Model C
(1)
(2)
(3)
D emographic an d E conom ic F actors
Age
Py
Male
40.44
0.0235 ***
0.0169 ***
0.0173 ***
5.46
-0.0183 **
-0.0326 ***
-0.0328 ***
0.32
-0.0349 ***
-0.0303 **
-0.0296 ***
73.52
-0.0154 ***
-0.0062 **
-0.0060 **
Poor
0.44
0.0291 ***
-0.0001
-0.0046
Medium
0.36
0.0081
-0.0001
-0.0048
Balance
7.79
-0.0796 ***
-0.0551 ***
-0.0646 ***
New
0.26
0.0258 ***
0.0632 ***
0.1115 ***
H h jn c
Plan D esign
lndex_offer
1.00
-0.3308
-0.2720
lntl_offer
1.00
0.0551
-0.0339
CS_offer
0.27
-0.1696 ***
-0.1995 ***
Loan_offer
0.60
-0.1624 ***
-0.1748 ***
Nfunds
33.1
-0.0190 ***
-0.0223 ***
2,874
-0.0032 ***
-0.0020 ***
Num_prt
Plan T ypes b y Life C ycle Funds F eatures
None/AddTM
0.23
0.2436 ***
KeepSA/AddTM
0.42
0.0998 ***
DropSA/AddTM
0.20
0.2110 ***
Months Since LC Added
0.0455 ***
Time Fixed Effect
No
Y es
Yes
No
Y es
Y es
No
Y es
Yes
Clustering at Participant
Yes
Y es
Y es
Observations
703,194
703,194
703,194
89,272
89,272
89,272
8.3%
25.9%
25.9%
R-Squared
95
Table 3-5. Difference-in-Difference Analysis of Equity Allocation by Age Group.

This table reports the regression results from the following Difference-in-Difference equation:
E Q % = a X i+ P Zj +SPostLCt+YiNone/AddTMj+y 2KeepSA/AddTMj
+ yiDropSA/AddTMj+y4None/AddTM*PostLCt+YiKeepSA/AddTMj *PostLCt
+Y6DropSA/AddTM ]+vi+ t]j+ rt+Sij,t
The dependent variable, EQ%, refers to the equity allocation o f a participant i in plan j in month t before
and after his plan introduced the LC Fund. PostLCj is set to 1 in months after LC introduction (else =0).
Other variables and error structure are as in Table 3-3. Controls include demographic, economic, and plan
design variables defined in Table 3-3.
Dependent Variable: Equity Allocation

-25
26-35
36-45
46-55
56-65
65+
0.028 ***
0.018 ***
0.022 ***
0.022 ***
0.045 ***
0.074 **
Time-series Variation
PostLC
0.023 **
Cross-sectional Variation
None/AddTM
0.293 ***
0.135 ***
0.056 ***
0.063 ***
0.064 ***
KeepSA/AddTM
0.198 ***
0.039 ***
0.016 **
0.047 ***
0.044 ***
0.090 **
DropSA/AddTM
0.232 ***
0.083 ***
0.031 ***
0.067 ***
0.078 ***
0.073 **
0.000
-0.001
None/AddSA (Reference Group)

Difference-in-Difference E ffect
PostLC*None/Add7M
-0.009
-0.024 ***
-0.004
0.001
PostLC'KeepSA/AddTM
0.038 ***
0.006 **
-0.001
-0.003
0.000
-0.007
P ostLC*DropSA/AddTM
-0.003
-0.007 ***
-0.001
0.000
-0.004
-0.015
PostLC*None/AddSA (Reference Group)

Controls
Yes
Yes
Yes
Yes
Yes
Yes
Time Fixed Effect
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Clustering at Participants
Observations
R-Squared
Yes
Yes
Yes
Yes
Yes
Yes
76,200
595,200
878,880
941,220
456,264
40,584
6,350
49,600
73,240
78,435
38,022
3,382
36.5%
28.7%
20.0%
11.0%
7.7%
11.7%
96
Table 3-6. Impact of LC on Investment Performance: CAPM Case.

This table reports the risk-adjusted returns and the return variance structure for participants who elected
Life Cycle funds. Participants risk-adjusted returns and variance are imputed from the first and second
moments o f the assets held, and the assets return and variance are evaluated using a CAPM asset pricing
model as below:
Rn.l - R f . r P lR cR SPR F ,b+ P2R1.BAIU',
/+@ 3 RmSCI,RF\ 1+ s n.I,
where R n t is the total return o f fund n in month t, and R fiX is the risk free return, R c r s p r f.u R lb a r f.I
R m scirf.i are excess returns on the CRSP value-weighted market portfolio, Lehman Brothers Aggregate
Bond index and MSCIEAFE indexes, respectively. p h p2 and p 3 are the regression coefficients or factor
loadings; su is the error term. Systematic risk is the variance that can be explained by market factors, and
the rest o f the variance is idiosyncratic risk. The share o f idiosyncratic risk is the ratio o f idiosyncratic risk
over the total variance. Shadow areas indicate not statistically significant at the 5% level.
Annual Riskadjusted Return
<%)
LC Only Other LC
Chooser Chooser
s
s
(1)
6.45
(2)
6.43
After
6.82
Difference
Before
Overall Before
-25
26-35
36-45
46-55
56-65
65+
Variance (%)
LC Only Choosers
Total
Idiosyncra Share of
Variance tic Risk Idiosyncrati
c Risk
(3)
(4)
(5)
Other LC Choosers
Total
Idiosyncra Share of
Variance tic Risk Idiosyncrati
c Risk
(6)
(7)
(8)
24.40
34.42
8.40
11.95
1.24
10.38
6.50
16.51
0.10
0.63
24.69
6.37
25.80
0.37
OOd
4.57
-1.14
-9.75
0.28
-2.03
-8.62
6.21
6.19
8.51
0.69
8.12
20.14
6.92
34.34
After
6.73
6.36
15.72
0.11
0.71
24.59
5.53
22.49
Difference
0.51
0 17
7.21
-0.58
-7.41
4.44
-1.39
-11.86
Before
6.49
6.42
11.63
1.00
8.62
23.51
7.60
32.30
After
6.88
6.45
18.95
0.11
0.58
25.40
5.75
22.66
Difference
0.39
7.33
-0.89
-8.05
1.89
-1.84
-9.65
Before
6.50
6.47
12.36
1.20
9.72
24.70
8.13
32.93
After
6.88
6.54
18.13
0.10
0.57
25.28
6.23
24.65
Difference
0.38
0 .07
5.77
-1.10
-9.15
058
-1.90
-8.29
Before
6.45
6.44
12.70
1.56
12.30
25.73
9.50
36.94
After
6.79
6.55
14.61
0.10
0.68
24.63
7.20
29.23
Difference
0.34
0 11
1.91
-1.46
-11.62
-1.10
-2.30
-7.71
Before
6.31
6.38
11.74
1.54
13.14
24.49
9.19
37.52
After
6.63
6.48
10.76
0.09
0.87
21.89
6.69
30.56
Difference
0.32
0 10
-0 97
-1.45
-12.27
-2.60
-2.50
-6.96
Before
6.15
6.24
9.58
0.76
7.98
19.38
5.67
29.23
After
6.42
6.21
8.46
0.11
1.28
17.31
4.74
27.37
Difference
0.27
-0.03
-1 12
-0.66
-6.70
-2.08
-0.93
-1.86
97
Table 3-7. Risk Source Contributing to Change in Investment Performance: CAPM Case, Participant Level
This table reports the risk source contributing to the change o f participant investment performance. LC Only Choosers and Other LC Choosers are defined as in
Table 3-3. Cash includes money market funds and investment contracts. Risk loadings are computed as in Table 3-6.
C hange in A s s e t Allocation(%)
LC Only C hoosers
All
4 6 -5 5
65-
C ash
Equity
Bond
(1)
(2)
(3)
(4)
(5)
(6)
S tock
Bond intem ation
M arket M arket al Equity
(Pi)
(32)
(Pj)
(7)
(8)
(9)
O ther LC C hoosers
S to c k
M arket
Bond
M arket
Intemation
al Equity
(Pi)
(10)
(pz)
(p3)
(1 1 )
(12)
5.9
52.0
42.2
13.7
68.1
18.2
0.459
0.357
0.063
0.586
0.212
0.113
61.5
38.5
9.8
70.5
19.7
0.564
0.240
0.119
0.627
0.177
0.133
-5.9
9.6
-3.7
-3.9
2.4
1.5
0.105
-0.117
0.057
0.041
-0.035
0.020
9.4
44.0
46.6
21.9
62.1
16.0
0.388
0.381
0.055
0.510
0.209
0.120
0.154
Before
0.0
64.3
35.7
12.9
73.0
14.1
0.540
0.254
0.110
0.628
0.147
-9.4
20.4
-11.0
-9.0
10.8
-1.9
0.153
-0.127
0.055
0.117
-0.062
0.033
5.1
52.3
42.5
14.3
68.7
17.0
0.461
0.360
0.064
0.583
0.195
0.121
After
0.0
70.3
29.7
10.1
74.0
15.9
0.605
0.193
0.128
0.645
0.135
0.143
Difference
-5.1
17.9
-12.8
-4.2
5.3
-1.1
0.144
-0.168
0.064
0.063
-0.060
0.022
Before
Before
5.1
53.2
41.7
12.1
70.0
17.9
0.472
0.354
0.063
0.602
0.199
0.113
After
0.0
63.5
36.5
8.9
72.2
18.9
0.594
0.207
0.127
0.642
0.156
0.133
Difference
0.021
-5.1
10.3
-5.3
-3.2
2.2
1.0
0.122
-0.147
0.064
0.040
-0.043
Before
5.8
52.9
41.4
13.0
67.8
19.2
0.468
0.350
0.064
0.593
0.225
0.107
After
0.0
55.0
45.0
9.4
68.0
22.6
0.533
0.279
0.114
0.617
0.208
0.126
0.019
-5.8
2.1
3.6
-3.6
0.2
3.4
0.066
-0.071
0.050
0.025
-0.017
Before
8.3
49.7
42.0
14.9
64.5
20.6
0.438
0.358
0.060
0.568
0.257
0.103
After
0.0
47.3
52.7
11.0
63.0
26.0
0.459
0.353
0.096
0.573
0.266
0.118
10.8
-3.9
-1.5
5.4
0.020
-0.005
0.036
0.005
0.009
0.014
23.6
0.376
0.063
0.526
0.250
0.099
Difference
5 6 -6 5
Bond
0.0
Difference
3 6 -4 5
LC Only C hoosers
Equity
Before
After
2 6 -3 5
O ther LC C hoosers
C ash
After
Difference
-2 5
C hange in Loadings of Risky A s s e ts
Difference
-8.3
-2.4
Before
11.4
43.4
45.2
15.0
61.4
0.0
42.0
58.0
11.9
57.9
30.2
0.396
0.392
0.080
0.507
0.273
0.103
-11.4
-1.4
12.8
-3.1
-3.5
6.6
0.020
0.016
0.017
-0.019
0.023
0.004
After
Difference
98
0.376
Table 3-8. Change in Investment Performance From Introducing LC: CAPM Case, Plan Level
This table reports the plan-level change in investment performance from introducing LC funds, measured by comparing investment performance o f participants
prior to the LC introduction month and six months afterward. The risk-adjusted return and variance decomposition are computed as in Table 6. Statistical tests
are applied to the Annual Risk-adjusted Return (Column 1) and the Share o f Idiosyncratic Risk (Column 2); ***.** indicates significant at the 1% and 5% level,
respectively.
Change in Investment
Perform ance (% pts)
Annual RiskS h are of
adjusted Return Idiosycratic Risk
(1)
All
None/AddTM
DropS A/AddTM
None/AddSA
Change in Risk S ource

Equity
Stock Market Bond Market International Equity
Allocation (%) Loading (pi) Loading (p2)
Loading (p3)
(3)
(4)
(5)
(6)
(7)
Before
6.14
21.9
NA
66.6
0.583
0.143
0.083
After
6.20
20.4
16.4
67.0
0.575
0.140
0.094
Difference
0.06
-1.50 **
0.5
-0.008
-0.003
0.011
Before
5.89
26.34
NA
66.8
0.571
0.111
0.091
After
5.87
23.51
15.1
67.1
0.561
0.102
0.101
0.3
-0.010
-0.009
0.010
Difference
KeepSA/AddTM
(2)
LC
Adoption
R ate (%)
-0.01
-2.83 ***
Before
6.30
18.74
NA
71.3
0.608
0.169
0.078
After
6.31
18.61
19.8
71.0
0.596
0.171
0.088
Difference
0.01
-0.13
-0.3
-0.011
0.002
0.010
Before
6.24
22.82
NA
65.7
0.628
0.314
0.104
After
6.35
21.85
20.1
66.0
0.619
0.287
0.120
Difference
0.10
-0.97 **
0.3
-0.009
-0.027
0.016
Before
5.57
16.38
NA
59.7
0.548
0.072
0.059
After
5.79
16.02
8.4
61.7
0.551
0.089
0.073
Difference
0.22 ***
-0.36
2.0
0.003
0.016
0.014
99
Figure 3-1. Target Asset Allocation of Life Cycle Funds.

Panel (a) indicates the asset allocation patterns of the target maturity date funds offered by Vanguard, with
age 65 assumed as the target retirement age. Panel (b) presents the asset allocation patterns o f the
Vanguard static allocation funds. Source: Authors tabulations from Vanguard Fund Prospectuses
(a). Target Maturity Type
2005 2010 2015 2020 2025 2030 2035 2040 2045 2050
65
60
U.S S to ck
55
50
45
i Int'l S tock
40
35
I U .S Bond
30
25
MMF
(b). Static Allocation Type
In co m e
U .S S to ck
C o n serv ativ e
Inf I S tock
M o derate
U .S Bond
Grow th
D C ash
100
20
Figure 3-2. Life Cycle Fund Adoption Patterns by Plan Participants.

This figure indicates the adoption rate by workers newly offered Life Cycle funds. TM refers to Target
Maturity funds and SA refers to Static Allocation funds. The sample includes only plans with at least 12
months o f data after the LC introduction over the period 1/03-12/05.
%
(a) Adoption Rate in Plans Adding TM/SA

without LC Options
7
6
5
4
3
2
1
0
1
10
11
12
Month Since LC Introduction

None/AddSA
isNone/AddTM
(b) Adoption Rate in Plans Adding TM on Top of SA
20
Period with TM
16
Month since TM introduced

-12
KeepSA/AddTM
10
is DropSA/AddTM
101
Figure 3-3. Change in Participants Equity Allocation Resulting from Moving to LC

Funds, By Age.
Panel (a) illustrates age specific changes in portfolio equity allocation for LC Only Choosers, while Panel
(b) focuses on Other LC Choosers. Before refers to the equity allocation in the month prior to the LC
selection; after refers to the first month in which the participant had a LC fund. Other terms are defined
in Table 3. The sample consists of 22,170 LC Only Choosers and 43,391 Other LC Choosers.
(a) LC Only C hooser
-2 5
2 6 -3 5
3 6 -4 5
5 6 -6 5
4 6 -5 5
65+
Before
(b) O th er LC C ho o ser
75
70
65
60
55
50
45
40
35
30
in
-2 5
2 6 -3 5
3 6 -4 5
Before
4 6 -5 5
5 6 -6 5
65+
After
102
Bibliography
Reference o f C h a p te r 1
Agnew, Julie, Pierluigi Balduzzi, and Annika Sunden. 2003. Portfolio Choice and
Trading in a Large 401(k) Plan. The American Economic Review. 93(1): 193-215.
Ameriks, John and Zeldes, Stephen P. 2004. How Do Household Portfolio Shares Vary
With Age? Working Paper.
http://www2.gsb.columbia.edu/facultv/szeldes/Research/.
Barber, Brad M. and Terrance Odean. 2000. "Trading is Hazardous to Your Wealth: The
Common Stock Investment Performance o f Individual Investors." Journal o f
Finance. LV(2). 773-806.
Barber, Brad M. and Terrance Odean. 2001. "Boys will be Boys: Gender,
Overconfidence, and Common Stock Investment." Quarterly Journal o f
Economics. 116 (1). 261-292.
Barber, Brad M. and Terrance Odean. 2002. "Online Investors: Do the Slow Die First?
Review o f Financial Studies. 15(2). 455-487.
Barber, Brad M., Yi-Tsung Lee, Yu-Jane Liu, and Terrance Odean. 2004. Who Gains
from Trade? Evidence from Taiwan. Working paper.
faculty.haas.berkeley.edu/odean/Current%20Research.htm.
Benartzi, Shlomo and Richard H. Thaler. 2001. Naive Diversification in Defined
Contribution Savings Plans. American Economic Review. 91(1). 79-98.
Bemheim, B. Douglas. 1998. Financial Illiteracy, Education and Retirement Saving.
In Olivia S. Mitchell and Sylvester J. Schieber, eds., Living with Defined
Contribution Plans. University o f Pennsylvania Press, Philadelphia, PA. 38-68.
Bodie, Zvi, Alex Kane and Alan J. Marcus. 2002. Investments: Fifth Edition. McGrawHill Higher Education. New York.
Brennan, John J. 2002. Straight Talk on Investing. John Wiley and Sons, Hoboken, NJ.
Bronson, James W., Matthew H. Scanlan and Jan R. Squires. Forthcoming. Managing
Individual Investor Portfolios. In Managing Investment Portfolios: A Dynamic
Process, 3rd Edition. CFA Institute, Charlottesville, VA.
Choi, James J., David Laibson, Brigitte C. Madrian and Andrew Metrick. 2002a.
Defined Contribution Pensions: Plan Rules, Participant Choices and the Path o f
103
Least Resistance. In Janies Poterba, ed., Tax Policy and the Economy. MIT
Press, Cambridge, Massachusetts. 16. 67-113.
Choi, James J., David Laibson, and Andrew Metrick. 2002b. How does the Internet
affect trading? Evidence from investor behavior in 401(k) plans. Journal o f
Financial Economics. 64 (3): 397-421.
Choi, James J., David Laibson, and Brigitte C. Madrian. 2004. Plan Design and 401 (k)
Saving Outcomes. National Tax Journal. LVII(2). 275-298.
Elton, Edwin J., Martin J. Gruber and Christopher R. Blake. 2004. The Adequacy o f
Investment Choices Offered by 401(k) Plans. Stem School Working Paper,
New York University.
Even, William E. and David A. Macpherson. 2004. Determinants and Effects o f
Employer Matching Contributions in 401(k) Plans. Working Paper.
www.sba.muohio.edu/evenwe/papers/matching_in_401 k.pdf.
Evensky, Harold R. 1997. Wealth Management: The Financial A dvisors Guide to
Investing and M anaging Client Assets. McGraw-Hill Publishing.
Gervais, Simon and Terrance Odean. 2001. "Learning to be Overconfident." Review o f
Financial Studies. 14(1). 1-27.
Grossman, Sanford. J and Joseph E. Stiglitz. 1980. On the Impossibility o f
Informational Efficient Capital Markets. American Economic Review, 70(3).
393-408.
Holden, Sarah and Jack VanDerhei. 2004. 401(k) Plan Asset Allocation, Account
Balances and Loan Activity in 2003. ICIPerspective. Investment Company
Institute, W ashington DC. 10(2).
Iyengar, Sheena, Gur Huberman, and Wei Jiang. 2004. How Much Choice is Too
Much? Contributions to 401(k) Retirement Plans. In Pension Design and
Structure: New Lessons from Behavioral Finance. Olivia S. Mitchell and Stephen
P. Utkus, eds. Oxford, United Kingdom: Oxford University Press: 83-95.
Madrian, Brigitte, and D.F. Shea. 2001. The Power o f Suggestion: Inertia in 401(k)
Participation and Savings Behavior. Quarterly Journal o f Economics 116:11491525.
Maginn, John L., Donald L. Tuttle, Dennis W. McLeavey and Jerald E. Pinto.
Forthcoming. The Portfolio Management Process and the Investment Policy
Statement. In Managing Investment Portfolios: A Dynamic Process, 3rd Edition.
CFA Institute, Charlottesville, VA.
104
Mitchell, Olivia S. and Stephen P. Utkus. 2004. The Role o f Company Stock in Defined
Contribution Plans. In Olivia S. Mitchell and Kent Smetters, eds., The Pension
Challenge: Risk Transfers and Retirement Income Security. Oxford University
Press. 33-70.
Mitchell, Olivia S., Stephen P. Utkus, and Tongxuan (Stella) Yang. 2005. Turning
Workers into Savers? Incentives, Liquidity and Choice in 401(k) Plan Design.
NBER Working Paper 11726.
Odean, Terrance. 1999. "Do Investors Trade Too Much?" American Economic Review.
89. 1279-1298.
Reid, Brian and Kimberlee Millar. 2004. Mutual Funds and Portfolio Turnover. IC I
Research Commentary. Investment Company Institute, W ashington DC.
November 17.http://www.ici.org/pdf/rc_vl n2.pdf.
Samuelson, Paul A. 1969. Lifetime Portfolio Selection by Dynamic Stochastic
Programming. Review o f Economics and Statistics. 51(3). 239-246.
Tacchino, Kenn Beam and David A. Littel. Planning fo r Retirement Needs. Third
edition. The American College, Bryn Mawr, PA. 1999.
Vanguard. 2004. How America Saves: A Report on Vanguard D efined Contribution
Plans 2004. Vanguard Center for Retirement Research, Malvern, PA.
www.vanguardretirementresearch.com.
Reference o f C h a p te r 2
Trading in a Large 401 (k) Plan, American Economic Review, 93(1): 193-215.
with Age? TIAA-CREF Institute Working Paper 6-120101 .Working Paper.
Finance, LV (2): 773-806.
Barber, Brad M. and Terrance Odean. 2001. "Boys Will be Boys: Gender,
Overconfidence, and Common Stock Investment", Quarterly Journal o f
Economics, 116(1): 261-292.
105
Calvet, Laurent, John Campbell, and Paolo Sodini. 2006. Down or Out: Assessing the
Welfare Costs o f Household Investment Mistakes," NBER Working Paper No.
12030, February.
Choi, James J., David Laibson, and Andrew Metrick. 2002. How Does the Internet
Affect Trading? Evidence from Investor Behavior in 401(k) Plans, Journal o f
Financial economics 64(3): 397-421.
Elton, Edwin J., Martin J. Gruber, and Christopher R. Blake. 2006. Participant Reaction
and The Performance o f Funds Offered by 401(k) Plans. Working Paper.
Fama, Eugene F. and James D. MacBeth. 1972. Risk, Return, and Equilibrium:
Empirical Tests, Journal o f Political Economy, 2: 607-636
Fama, Eugene F. and Kenneth R. French. 1982. The Cross-section o f Expected Stock
Returns, Journal o f Finance, June:427- 465
Fama, Eugene F. and Kenneth R. French. 1993. Common risk factors in the returns on
stocks and bonds, Journal o f Financial Economics 33: 3-56.
Gervais, Simon and Terrance Odean. 2001. "Learning to be Overconfident", Review o f
Financial Studies, Spring, 14(1): 1-27.
Goetzmann, William N., Massimo Massa, and K. Geert Rouwenhorst, Behavioral
Factors in Mutual Fund Flows International Center for Finance at the Yale
School o f Management Working paper.
Grinblatt, Mark and Matti Keloharju. 2001. What Makes Investors Trade? Journal o f
Finance, April. 56 (2): 589-616.
Grinblatt, Mark, and Matti Keloharju. 2000. The Investment Behavior and Performance
o f Various Investor Types: A Study o f Finlands Unique Data Set. Journal o f
Financial Economics 55: 43-67
Guiso, Luigi and Tullio Jappelli. 2006. Information Acquisition and Portfolio
Performance. CeRP Working Paper 52/06. Center for Research on Pensions and
Welfare Policies, Moncalieri, Italy.
http://cerp.unito.it/publications/information acquisit.
Investment Company Institute. (ICI). 2006. Research Fundamentals, Vol. 15, No. 5,
Washington, D.C.: Investment Company Institute.
Participation and Savings Behavior. Quarterly Journal o f Economics 116: 11491525.
106
Merton, Robert. 1973. An Intertemporal Capital Asset Pricing Model. Econometrica 41:
867-887
Mitchell, Olivia S., Gary R. Mottola, Stephen P. Utkus, and Takeshi Yamaguchi. 2006.
The Inattentive Participant: Portfolio Trading Behavior in 401(k) Plans.
Wharton Pension Research Council Working Paper 2006-05. Pension Research
Council, University o f Pennsylvania, Philadelphia, PA.
http://www.pensionresearchcouncil.org/publications/document.php?file=::10.
Munnell, Alicia H., and Pamela Perun. 2006. An Update on Private Pensions. Center
for Retirement Research at Boston College. Chestnut Hill, MA.
http ://www.bc.edu/centers/crr/ib 50. shtml.
Odean, Terrance. 1999. "Do Investors Trade Too Much?" American Economic Review,
Vol. 89, December, 1279-1298.
Odean, Terrance, Brad Barber, Yi-Tsung Lee and Yu-Jane Liu. 2006. Who Loses from
Trade? Evidence from Taiwan, Working Paper, The Hass School, University o f
California at Berkeley.
Poterba, James, Steve Venti and David Wise. 2004. The Transition to Personal
Accounts and Increasing Retirement Wealth: Macro and Micro Evidence. In D.
Wise, ed., Perspectives on the Economics o f Aging, Chicago, IL: Chicago
University Press: 17-71.
United States Department o f Labor. 2005. National Compensation Survey: Employee
Benefits in Private Industry in the United States, March 2005. U.S. Department o f
Labor Statistics, Summary 05-01.
Reference of Chapter 3
Trading in a Large 401(k) Plan, American Economic Review, 93(1): 193-215.
with Age? TIAA-CREF Institute Working Paper 6-120101.Working Paper.
Finance, LV (2): 773-806.
Barber, Brad M. and Terrance Odean. 2001. "Boys Will be Boys: Gender,
Overconfidence, and Common Stock Investment", Quarterly Journal o f
Economics, 116(1): 261-292.
107
Benartzi, S. and Thaler, R.H, 2001, Naive diversification strategies in defined

contribution saving plans, The American Economic Review 91(l):79-98.
Bodie, Zvi, Robert C. Merton, and William Samuelson, 1991, Labor Supply Flexibility
and Portfolio Choice in a Life Cycle Model, Journal o f Economic Dynamics and
Control 16, 427-449.
Brown, Jeffrey R. Nellie Liang and Weisbenner Scott, 2006, Individual Account
Investment Options and Portfolio Choice: Behavioral Lessons from 401(k)
Plans, Working Paper, University o f Illinois.
Campbell, John Y., Joao Cocco, Francisco Gomes, and Pascal Maenhout, 2001,
Investing Retirement Wealth: A Life-Cycle Model, in John Y. Campbell and
Martin Feldsten (ed), Risk Aspects o f Investment-Based Social Security Reform,
University o f Chicago Press
Campbell, John, Laurent Calvet, and Paolo Sodini. 2006. Down or Out: Assessing the
Welfare Costs o f Household Investment Mistakes," NBER W orking Paper No.
12030, February.
Choi, James J., David Laibson, and Andrew Metrick. 2002. How Does the Internet
Affect Trading? Evidence from Investor Behavior in 401(k) Plans, Journal o f
Financial economics 64(3): 397-421.
Gervais, Simon and Terrance Odean. 2001. "Learning to be Overconfident", Review o f
Financial Studies, Spring, 14(1): 1-27.
Grinblatt, Mark and Matti Keloharju. 2001. What Makes Investors Trade? Journal o f
Finance, April. 56 (2): 589-616.
Grinblatt, Mark, and Matti Keloharju. 2000. The Investment Behavior and Performance
o f Various Investor Types: A Study o f Finlands Unique Data Set. Journal o f
Financial Economics 55: 43-67
Guiso, Luigi and Tullio Jappelli. 2006. Information Acquisition and Portfolio
Performance. CeRP Working Paper 52/06. Center for Research on Pensions and
Welfare Policies, Moncalieri, Italy.
Huberman, Gur and Wei Jiang, 2006. "Offering versus Choice in 401 (k) Plans: Equity
Exposure and Number o f Funds," Journal o f Finance 61(2), p763-801
Investment Company Institute. (ICI). 2006. Research Fundamentals, Vol. 15, No. 5,
Washington, D.C.: Investment Company Institute.
Iyengar, Sheena, Gur Huberman and Wei Jiang. 2005. How Much Choice is Too Much?
Contributions to 401 (k) Retirement Plans, In Pension Design and Structure:
New Lessons from Behavioral Finance. Edited by Olivia S. Mitchell and Stephen
P. Utkus. Oxford University Press: 83-95.
Karlsson, Anders, Massa, Massimo and Simonov, Andrei, 2006, "Portfolio Choice and
Menu Exposure" W harton Pension Research Council Working Paper 2006-23.
Pension Research Council, University o f Pennsylvania, Philadelphia, PA.
108
Lusardi, Annamaria and Olivia S. Mitchell. 2007. Baby Boomer Retirement Security:
The Roles o f Planning, Financial Literacy, and Housing Wealth. Journal o f
Monetary Economics. 54(1) January forthcoming.
Participation and Savings Behavior. Quarterly Journal o f Economics 116: 11491525.
Marquez, Jessica, 2005, Life Cycle Funds Can Help Companies Mitigate Risk and
Boost Employee Savings, Workforce Management, April 65-67
Merton, Robert. 1973. An Intertemporal Capital Asset Pricing Model. Econometrica 41:
867-887
Merton, Robert C., 1969, Lifetime Portfolio Selection under Uncertainty: The
Continuous Time Case, Review o f Economics and Statistics 51, 247-257
Mitchell, Olivia S., Gary R. Mottola, Stephen P. Utkus, and Takeshi Yamaguchi. 2006.
The Inattentive Participant: Portfolio Trading Behavior in 401 (k) Plans.
Wharton Pension Research Council Working Paper 2006-05. Pension Research
Council, University o f Pennsylvania, Philadelphia, PA.
Mitchell, Olivia S., Stephen P. Utkus, and Stella Yang. 2005. Dimensions o f 401(k) Plan
Design. In Restructuring Retirement Risks. Edited by David Blitzstein, Olivia S.
Mitchell and Stephen P. Utkus. Oxford University Press.
Munnell, Alicia H., and Pamela Perun. 2006. An Update on Private Pensions. Center
for Retirement Research at Boston College. Chestnut Hill, MA.
Odean, Terrance. 1999. "Do Investors Trade Too Much?" American Economic Review,
Vol. 89, December, 1279-1298.
Odean, Terrance, Brad Barber, Yi-Tsung Lee and Yu-Jane Liu. 2006. Who Loses from
Trade? Evidence from Taiwan, Working Paper, The Hass School, University o f
California at Berkeley.
O Donoghue, Ted, and Matthew Rabin, 2001, Choice and Procrastination, Quarterly
Journal o f Economics, Feb 2001, pl21-160.
O Donoghue, Ted, and Matthew Rabin, 2006, Incentive and Control, Working Paper.
O Donoghue, Ted, and Matthew Rabin, 1999, Procrastination in Preparing for
Retirement, in Henry Aaron, ed., Behavioral Dimensions o f Retirement
Economics, Brookings Institution Press and Russell Sage Foundation.
Poterba, James, Joshua Rauh, Steven Venti, and David Wise, 2006, Life Cycle Asset
Allocation Strategies and The Distribution o f 401(k) Retirement Wealth, NBER
Working Paper No. 11974.
Poterba, James, Steve Venti and David Wise. 2004. The Transition to Personal
Accounts and Increasing Retirement Wealth: Macro and Micro Evidence. In D.
Wise, ed., Perspectives on the Economics o f Aging, Chicago, IL: Chicago
University Press: 17-71.
109
Samuelson, Paul, 1969, Lifetime Portfolio Selection by Dynamic stochastic

Programming, Review o f Economics and Statistics 51, 239-246Samuelson,
William, and Richard J. Zeckhauser, 1988, Status Quo Bias in Decision
Making, Journal o f Risk and Uncertainty 1, p7-59.
Thaler, Richard H. and Shlomo Benartzi, 2004, Save More Tomorrow: Using
Behavioral Economics to Increase Employee Saving, Journal o f Political
Economy, 112(1), pp.S164-S187.
Todd, Petra E., 2006 Evaluating Social Programs with Endogenous Program Placement
and Selection o f Treated, in H and o f Agricultural Economics, Volume 4, edited
by Robert E. Evenson and T.Paul Schultz.
United States Department o f Labor. 2005. National Compensation Survey: Employee
Benefits in Private Industry in the United States, March 2005. U.S. Department o f
Labor Statistics, Summary 05-01.
Vayanos, Dimitri, 1998, Transaction costs and asset prices: a dynamic equilibrium
model, Review o f Financial Studies, 11 (1). pp. 1-58.
Yamaguchi, Takeshi, Olivia S. Mitchell, Gary R. Mottola, Stephen P. Utkus, 2006.
Winners and Losers: 401(k) Trading and Portfolio Performance. Pension
Research Council Working Paper 2006-26. The Wharton School, University o f
Pennsylvania, Philadelphia, PA
110

Asset MNGT PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Asset MNGT PDF

Uploaded by

Copyright:

Available Formats

ESSAYS ON 401(K) ASSET MANAGEMENT

For the Graduate Group in Managerial Science and Applied Economics

Prff-cWGraduate Group Chairperson

UMI N um ber: 3261011

All rights reserved.

ProQuest Information and Learning Company

To Aya, Eric and Justin

In the first chapter o f this dissertation, we investigate the determinants o f trading

1.1 Related S tudies...............................................................................................................2

Table 1-3. Summary o f Two-Year Trading Statistics (1/2003-12/2004).................25

Table 3-3. Marginal Effects o f Determinants o f Choosing LC Funds.........................93

Chapter 1: The Inattentive Participant: Portfolio Trading Behavior in

We relate a number o f portfolio outcomes to

employee characteristics as well as plan design features and participant investment

1Mitchell, Utkus, and Yang (2005) provide a review.

1.1 Related Studies

1.2 Descriptive Statistics

Nearly three-quarters o f the sample plans permit

Table 1-1 here

internet (as o f January 2003).

generate observed trading which is employer-initiated rather than employee-driven. To

1.3 Multivariate Analysis of Trading Patterns

binomial count regression.

a censored Tobit regression.26 The TRADER, NTRADES and TURNOVER regressions

1.4 Empirical Findings

24For instance the equation for model C is E{l'RADlNGj y. = 11DEMO, PD, AC C T) =

Pr [NTRADES'tJ = NTRADEStJ \ D EM O^^PD ^ACCT^)

T(9 +NTRADES ik)

where T is a Gamma function with

, E{NTRADESi j \ DEMOt J, PDj , ACCTt j )= 9 a

Var(NTRADESi J \ DEMO, , , PD}, ACCTU ) = 9a(I + a) = Xu fl + - Xu .

1.5 Discussion and Conclusions

Number of active participant accounts

Plan design fea tu re s offered

401 (k) A cc o u n t Features

Panel B. Plan Features Offered versus Features Held (Account-Level)

Plan features offered

Account holdings (12/02)

Number of plan participants = 1,186,554

Table 1-3. Summary of Two-Year Trading Statistics (1/2003-12/2004)

For all accounts*

For traders only*

For active traders only*

0.76(0.00) 23.7% (0.00%)

2.44(1.00) 76.5% (42.0%)

12.68(9.00) 340.0% (160.4%)

2.92(1.00) 89.7% (47.7%)

12.86(9.00) 346.9% (162.4%)

Number of plan participants = 1,186,554

Table 1-4. Distribution of Key Trading Measures (1/2003-12/2004)

Panel B. Distribution of Participant Turnover

Number of plan participants = 1,186,554

Table 1-5. Summary of Predicted Marginal Effects (1/03-12/04)

25.0% 24.6% 20.1%

23.3% 23.0% 19.1%

No. funds 10 add! funds

Table 1-5 (contd). Summary of Predicted Marginal Effects

No. funds 5 add'l funds

Number of plan participants = 1,186,554

Figure 1-1. Distribution of Number of Trades (1/03-12/04)

Number of trad e s over a 2-year period

Figure 1-2. Distribution of Portfolio Turnover (1/03-12/04)