Preliminary and Partial Draft
Please Do Not Quote
The Decline of Defined Benefit Retirement Plans and
Asset Flows
James Poterba
MIT and NBER
Steven Venti
Dartmouth College and NBER
David A. Wise
Harvard University and NBER
Prepared for the 8th Annual Joint Conference of the Retirement Research Consortium
“Pathways to a Secure Retirement”
August 10-11, 2006
Washington, D.C.
This research was supported by the U.S. Social Security Administration through grant
#10-P-98363-1 to the National Bureau of Economic Research as part of the SSA
Retirement Research Consortium. The findings and conclusions expressed are solely
those of the author(s) and do not represent the views of SSA, any agency of the Federal
Government, or the NBER.
Just two or three decades ago, employer-provided defined benefit (DB) pension
plans were the primary means of saving for retirement in the United States. But since
that time, 401(k) and other personal retirement accounts have become the principal form
of retirement saving in the private sector. Defined benefit plans have remained an
important form of retirement saving for federal employees and for state and local
employees, although even for these employees personal retirement accounts are
becoming increasingly important. In 2000, about 87 percent of private retirement plan
contributions were to 401(k) and other personal accounts, although the contributions to
personal accounts dropped to 82 percent of the total in 2001 when there was a large
increase in contributions to DB plans. Contributions to personal retirement plans
accounted for only 12 percent of total contributions to Federal pension plans in 2000, but
had increased to 17 percent by 2004. In this paper we describe the decline in defined
benefit plans and assess the implications of the decline for the flow of assets into and out
of DB plans over the next four decades. The paper is intended to be a companion to our
paper on “New Estimates of the Future Path of 401(k) Assets”-- Poterba, Venti, and Wise
(2005). Both papers will provide input to an analysis of “Demographic Change,
Retirement Saving, and Financial Market Returns”—described in Poterba, Venti, and
Wise (2005).
In section 1, we present a cohort description of the decline in the participation rate
of employed persons in DB plans over the past two decades. Then we describe a series of
analyses that provide the basis for projections of future assets in DB plans. We begin in
section 2, with a cohort description of the dollar amount of pension benefits received by
persons over age 55, and we explain our projections of DB benefits in the future. In
section 3, we present a parallel cohort description of the probability of receipt of DB
benefits. To project the probability of receipt of DB benefits after retirement by younger
cohorts, we rely in part on the DB plan participation rate prior to retirement for these
same cohorts. Because the benefit receipt data (data on benefits received) pertain to the
population of older persons, we need to develop DB plan participation data for the
population as a whole rather than employed persons, discussed in section 1. Thus in
section 4 we consider the likelihood that a person in the population participates in a DB
plan. In section 5, we combine the information described in sections 2, 3, and 4 to
`
1
develop projections of the total value of DB benefits in future years. In section 6, we
show projections of the DB pension wealth of cohorts retiring between 1982 and 2040
and, for each cohort we compare DB wealth to projected 401(k) assets (based on our
earlier paper). In section 7, we consider projections of the total value of assets in DB
trust funds. It is the change in the assets in DB trust funds combined with the change in
the assets in personal retirement accounts that may affect the rate of return on the
investments of future generations of retirees.
1. Participation of Employees in DB Plans
DB participation data were obtained from the Survey of Income and Program
Participation (SIPP) for the years 1984, 1987, 1991, 1993, 1995, 1998, and 2003. Our
analysis here and in subsequent sections of the paper is based on organization of the data
by cohort. We sometimes define cohorts by the age of the cohort in 1984 and sometimes
by the year in which the cohort attains age 65. When referring to the cohort age in 1984,
the age is proceeded by “C.” When referring to the year the cohort attains age 65, the
year is proceeded by “A.” Thus, for example, C65 and A1984 both identify the same
cohort.
Figure 1-1 shows the data for selected cohorts. It is clear that the DB
participation rate of employed persons declined consistently with successively younger
cohorts. For example, at age 45, the participation rate of the C45 cohort—that attained
age 45 in 1984—was about 43 percent. But the participation rate of the C27 cohort—that
attained age 45 in 2002—was about 29 percent. Comparisons at ages 31, 37, and 55
show similar differences.
Not only are there differences between cohorts, with younger cohorts having
successively lower participation rate at all ages, but there is also a within-cohort decline
in the participation rate with age. The within-cohort decline with age for older cohorts is
likely explained in part by retirement. DB plans typically provide incentives to retire
early and DB participants on average retire earlier than persons without these plans; some
participants may retire as early as age 55. But even for younger cohorts there is typically
a within-cohort decline in DB participation rates with age.
For, comparison, similar cohort data for 401(k) plans (including other personal
retirement plans as well, are shown in Figure 1-2 At age 45, the 401(k) participation rate
`
2
of the C45 cohort—that attained age 45 in 1984—was only 8 percent. But the
participation rate of the C25 cohort—that attained age 45 in 2002—was about 47 percent.
Figure 1-1. DB participation rate of employed
persons, selected cohorts
50
C55
45
1984
C45
1987
Participation Rate %
40
1995
1991
1993
35
1998
30
25
C27
C13
20
2003
15
10
5
63
61
59
57
55
53
51
49
47
45
43
41
39
37
35
33
31
29
27
25
0
Age
Figure 1-2. 401(k) participation rate of employed
persons, selected cohorts
50
C27
45
C13
Participation Rate %
40
1998
2003
35
30
1993
1995
1991
25
20
1987
15
1984
10
C45
5
C55
Age
`
3
63
61
59
57
55
53
51
49
47
45
43
41
39
37
35
33
31
29
27
25
0
The cohort data illustrated in Figures 1-1 allow comparison of the participation
rate of some cohorts at a given age—say 45—but these data alone do not allow
comparison of very young and very old cohorts. To do this we need to project forward
the future participation rates of younger cohorts at older ages and project backwards the
participation rates of older cohorts at younger ages. As explained in the Working Paper
version of this paper, we make these projections by fitting the participation data with a
probit model and then using the estimated parameters to predict outside range of the
observed data. The projected data for younger cohorts suggest, for example, that for the
cohort that is born in 1984 (the C0 cohort), the DB participation rate will never exceed 19
percent, with the maximum at about age 35.
As discussed below, to project future assets in DB plans, we need to consider the
proportion of all persons that participates in a DB plan. In section 4 we return to this
issue.
2. DB Pension Benefits of Recipients and Projections
We now describe a series of analyses that provide the basis for projections of
future assets in DB plans. A more detailed description is presented in the Working Paper
version of the paper. We begin in this section with a cohort description of the dollar
amount of DB benefits received by all persons over 55 and then describe how we project
benefit amounts in the future. In section 3, we present a parallel description of the
percent of older persons that receives DB benefits.
Data on DB benefits received, like the participation data shown above, are
obtained from SIPP waves for the years 1984, 1987, 1991, 1993, 1995, 1998, and 2003.
And, as with the participation data, we first present a cohort description of the observed
data. We then fit the observed data and use the fitted model to project benefits outside
the range of the observed data.
Figure 2-1 shows the actual data for the level of benefits for selected cohorts. The
data are for all persons age 55 to 85 and thus include persons who are receiving benefits
from federal or state and local retirement programs, as well as from private sector
pensions plans. Two features of the data stand out. One is that benefits of younger
cohorts are much greater than benefits of older cohorts. For example, the benefit at age
`
4
60 for the cohort that attained that age in 1984 was about $600 per month; the benefit for
the cohort that attained age 60 in 2003 (this comparison is circled in the figure) was about
$1,600 per month. The other feature of the data is that within-cohort benefits increase
with age. In part this results from the indexing of benefits from some DB plans,
especially federal and state and local plans. In addition, private employer plans, which
are typically not indexed, sometimes grant cost-of-living increases on an ad hoc basis
after retirement.
Figure 2-1. DB benefit reciept for selected cohorts
1800
1600
Dollars per Month
1400
1200
1000
800
600
400
200
0
55
60
65
70
75
80
85
Age
We fit the data using a piecewise linear specification for age and allowing for
separate intercepts for each cohort (cohort effects). We then use the estimated parameters
to project benefits forward for the younger cohorts and to project benefits backward for
the older cohorts. There are not enough SIPP observations to reliably estimate cohort
effects for cohorts younger than A2012 or older than A1970.
The most important
projections are for cohorts younger that the A2012 cohort. We make these projections by
using the estimated age effects, together with extrapolation of the estimated year-to-year
change in cohort effects. Over the 1982 to 2003 period there is a close association
between the four-year moving average of the change in the estimated cohort effects and
the Social Security Administration average wage index. After 2003, the SSA assumes a
`
5
wage increase of 3.9 percent annually. On average the change in the cohort effects was
only about 2.9 percent over these years. Based on these and other data we have assumed
that for younger cohorts benefits will increase at 3.9 percent for each successively
younger cohort, the same as the SSA intermediate assumption for the average wage
index.
3. The Receipt of DB Pension Benefits
In section 2 above, we considered the dollar amount of the DB benefits of persons
who received DB benefits at each age. To obtain an estimate of the total dollar amount of
DB benefits for all persons, we need also to determine the proportion of persons who
receive benefits at each age. We consider the probability of benefit receipt in this
section.
We again begin with cohort data (not shown) on the percentage of persons
receiving DB benefits at each age. Two features of the data are evident. First, the cohort
effects are rather small, with the exception of the older cohorts. In other words, the
profile of benefit receipt by age is about the same for all the cohorts represented in the
SIPP data (with the exception of the oldest cohorts). For example, among cohorts
observed at age 70, the probability of receipt of benefits ranges from about 45 percent for
the youngest cohort to 49 percent for the oldest cohort. Second, the age at which the
maximum percent of persons receive benefits is about 70.
To project benefit receipt for future (younger) cohorts, that are not observed in the
SIPP data, we assume the age-benefit-receipt profile observed in the SIPP data continues
to apply. However, we allow this profile to shift (downward) for younger cohorts. To do
this we first estimate age and cohort effects for cohorts observed in the SIPP. Then we
predict benefits for younger cohorts, assuming that--except for the cohort effects--the
age-benefit-receipt profile is the same for younger cohorts as it has been for older cohorts
observed in the SIPP.
Because we cannot reliably estimate cohort effects for cohorts younger than
A2003, we assume that the benefit receipt rates of successively younger cohorts follow
the same pattern as the DB participation rates of these same cohorts when they were in
the labor force. The assumption is that if fewer persons in a particular cohort participated
when young, then fewer persons will receive benefits after retirement. We do this by
`
6
first calculating the maximum DB participation rate over ages 25-64 for each cohort. We
use this maximum rate to predict benefit receipt when the cohort is retired. We use yearto-year changes in the maximum DB participation rate, and the 2003 fitted age-benefit
receipt profile, to project benefit receipt for cohorts A2004 through A2060. Before
presenting these projections, we describe the population-based DB participation rates that
are required to make the calculations.
4. Population-Based DB Participation Rates
The calculations described in the section above require an estimate of the
maximum DB participation rate for each cohort. To predict future participation rates for
this purpose, however, we must consider not the participation rate of employed persons
(estimated in section 1), but rather the likelihood that a person in the population
participates in a DB plan. This is necessary to yield participation rates that correspond to
the benefit receipt rates that are. That is, the SIPP data we use pertains to the probability
that a person in the population receives DB pension benefit.
The person-based participation rates and the population-based rated can differ
substantially. To see this, note that the percent of population that participates in a DB
plan at age a is given by
( DB / P )a = ( E / P )a Pr[ DBa | Ea ]
The last term, Pr [ DBa | Ea ] , declined between 1984 and 2003, but the fraction of the
population employed ( E / P )a increased between 1984 and 1998 but fell in 2003.1 The
percentage of employed persons participating in a DB plan declined from 39.4 percent to
29.7 percent. The overall effect of these trends is that cohort effects are smaller when all
persons are used as the base than when employed persons are used as the base. For
example, at age 45 the difference in the participation rates of cohorts 27 and 45 is about
13 percentage points based on those employed but only about 10 percentage points based
on all persons.
1
The employment to population ratio for persons 16 and over reported by the BLS shows an increase from
59.5% in 1984 to 64.1% in 1998 and then a drop to 62.3% in 2003. The SIPP data we use are for ages 25
to 64.
`
7
Figure 4-2. Actual versus fitted profiles for selected
cohorts (persons)
50
45
Participation Rate %
40
35
30
25
20
15
10
5
63
61
59
57
55
53
51
49
47
45
43
41
39
37
35
33
31
29
27
25
0
Age
As with the participation based on employed persons, we fit the population-based
participation data using a probit model. Selected fitted cohort profiles together with the
corresponding cohort data are shown in Figure 4-2. We judge that the fitted profiles
represent the data quite well. We then use the model estimates to predict populationbased participation rates at all ages (25 to 64) for each of the cohorts for which we are
able to estimate cohort effects. We predict at all ages for each cohort because we want to
determine the age at which the participation rate is at a maximum for each of the cohorts.
Predicting at older ages for the younger cohorts suggests that by the time these cohorts
reach retirement age, their participation in DB plans will be very low. For example,
based on these predictions the DB participation rate of persons in the C11 cohort will be
only 8.8 percent when that cohort attains age 55 (cohort C11 attains age 65 in 2038).
5. Projected Benefits Paid
We want to project the total value of benefits in future years. To do this we
combine estimates of the level of benefits in section 2 with estimates of the probability of
benefit receipt in section 3. For cohorts that attain age 65 after 2003 we use populationbased estimates of DB participation to predict pension receipt after retirement. We first
use the estimates of benefit receipt for cohorts A1982 to A2003 to predict the percent of
`
8
each of these cohorts that receives benefits at age 70. As discussed in section 3, for these
cohorts there are a sufficient number of SIPP observations to obtain reasonable estimates
reliable estimates of benefit receipt. These probabilities are graphed in Figure 5-1 (the
blue bars).
To project benefits at age 70 for younger cohorts (A2004 to A2040), we use
additional data on population-based DB participation rates, discussed in section 4. From
these data, we have calculated the maximum participation rates, over ages 25 to 64, for
each cohort. We use the maximum participation rate to predict the probability of benefit
receipt at age 70 for cohorts A2004 and younger. We assume that the year-to-year
percent change in the probability of receipt of benefits at age 70 is the same as the yearto-year percent change in the maximum DB participation rate. For cohorts younger than
A2040 a two percent decline is assumed, an extrapolation of the decline for cohorts
A2031 to A2040 (shown in red in Figure 5-1). From the percent that receives benefits at
age 70, we predict the percent that that receives benefits at each age, as described above.
These estimates are shown for selected cohorts in Figure 5-2.
Figure 5-1. Percent receiving benefits at age 70, by cohort
(blue is fitted SIPP estimate and red is projected)
60
50
Percent
40
30
20
10
Age attains age 65
`
9
20
50
20
46
20
42
20
38
20
34
20
30
20
26
20
22
20
18
20
14
20
10
20
06
20
02
19
98
19
94
19
90
19
86
19
82
0
Figure 5-2. Projected percent receiving benefits for selected
cohorts attaining age 65 in years 1982 through 2040
(blue is fitted SIPP estimate and red is projected)
60
A2000
50
Percent
40
30
20
A2040
10
David: Note that 1982 is not the highest
0
55
60
65
70
75
80
85
90
95
100
Age
6. Present Value of DB benefits at 65
From the projected data on benefits, we have calculated the average present value
at age 65 of future DB benefits of all persons, including those that do not receive DB
benefits. Results are presented for each cohort and have been converted to constant year
2000 dollars. A real discount rate of 3 percent and average SSA age-specific survival
probabilities are assumed. We have also calculated the average present value of DB
benefits at age 65 of persons who receive DB benefits. Both series are graphed in Figure
6-1. We assume that the benefits of successive cohorts will increase 3.9 percent annually
`
10
Figure 6-1. Present value DB benefits at age 65: all
persons and persons with a DB
200,000
180,000
year 2000 dollars
160,000
140,000
120,000
100,000
80,000
60,000
40,000
20,000
0
1982 1986 1990 1994 1998 2002 2006 2010 2014 2018 2022 2026 2030 2034 2038
Year cohort attains age 65
All persons
Persons with DB
in the future, while the assumed discount rate is only 3 percent. Thus the top profile
shows that the average present value of DB benefits for recipients will be greater for
future than for current cohorts. However, the lower profile shows that the PV of DB
wealth, averaged across all persons, will decline in the future as fewer persons
participate. The projections show that DB wealth peaks in 2003 at about $73,000 and
falls to about $50,000 by 2020.2
For comparison, the value of 401(k) assets (reported in Poterba, Venti and Wise
2005) are shown for all persons and for persons with 401(k) assets in Figures 6-2 and 6-3
respectively. Average 401(k) assets of all persons reach the average PV of DB benefits
of all persons in 2010 when both are about $64,000. Thereafter assets in 401(k) accounts
continue to grow, reaching about $125,000 in 2020, $214,000 in 2030, and $408,000 in
2040. The assets of persons with 401(k) accounts reach the level of the PV of DB
benefits for persons who receive benefits in 2013. Thereafter the 401(k) assets continue
2
These estimates indicate that in 2000 the average of DB benefits over all persons was $67,795. Based on HRS data,
Johnson, Burman and Kobes (2004), estimate that the mean present value of employer sponsored pension income for
persons 65 to 69 in 2000 was $50,203. Our estimate should be larger than theirs because we include persons of all
ages. In particular, average benefits increase with age because death rates selectively leave in the sample persons with
higher benefits, and because of ad hoc cost of living increases and indexed benefits in many government plans.
`
11
to grow, reaching about $217, 000 in 2020, $349,000 in 2030, and $549,000 in 2040. For
persons with DB plans the average present value of benefits also continues to grow,
reaching $187,000 by 2040.
Figure 6-2. Present value DB benefits at age 65
and 401(k) assets at age 65, all persons
600,000
year 2000 dollars
500,000
400,000
300,000
200,000
100,000
19
82
19
85
19
88
19
91
19
94
19
97
20
00
20
03
20
06
20
09
20
12
20
15
20
18
20
21
20
24
20
27
20
30
20
33
20
36
20
39
0
Year cohort attains age 65
PV of DB benefits
`
401(k)assets
12
Figure 6-3. Present value of DB benefits at age 65 for
persons with a DB and 401(k) assets at age 65 for
persons with a 401(k)
600,000
year 2000 dollars
500,000
400,000
300,000
200,000
100,000
39
36
20
33
20
30
20
20
24
27
20
21
20
18
20
15
20
12
20
09
20
20
03
06
20
00
20
97
20
94
19
91
19
88
19
85
19
19
19
82
0
Year cohort attains age 65
PV of DB benefits
401(k)assets
7. Assets in DB Trust Funds and Total Benefits Paid
This work will feed into an analysis of future changes in the demand for financial
assets and, in turn, on the potential effect of changes in asset demand on future rates of
return. We made similar calculations with respect to 401(k) and other personal
retirement account assets in Poterba, Venti, and Wise (2005). This DB component,
together with the “401(k)” component will, we believe, provide a reasonable estimate of
the change in asset demand due to changing demographic trends.
One way to predict future assets in DB plans is to predict total benefits paid in
future years and then to suppose that assets held by DB plans in each year are sufficient
to pay the present value of these future obligations. The results shown here are based on
this approach. We believe, however, that this approach should yield “fully-funded”
current assets that are greater than actual assets. A second way is to predict future assets
based on an extrapolation of current assets compared to “fully-funded” liabilities.
As described above, we have projected the DB benefits of recipients by cohort
and we have projected the probability of benefit receipt by cohort. From these data,
together with demographic forecasts of the number of persons at each age in each year in
the future, we can calculate the estimated total amount of DB benefits paid each year.
`
13
Figure 7-1, shows these totals for the years 1982 to 2004, together with a constructed
series that sums together private sector DB benefits, DB benefits paid to federal
employees, and all benefits paid to state and local employees.3
Figure 7-1. Projected total benefits paid compared
to actual benefits paid, 1982 to 2004
400,000
350,000
300,000
250,000
200,000
150,000
100,000
50,000
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
0
Projected total
Actual total
The projected totals are close to the actual totals in the early years but fall short in
the later years. The discrepancy is due in part at least the growing importance of 401(k)like accounts in the state and local government plans. Thus we believe that the
comparison lends credence to our estimates, based on SIPP data, for these years. Our
forward projections, of course depend on the assumptions we have made to project
benefit levels and benefit receipt for future cohorts not represented in the SIPP data.
Our projected estimates of total DB benefits in future years are shown in Figure 72 below, in year 2000 dollars. Total benefits paid from DB plans continue to increase
until 2027. The profile turns down eventually because the probability that benefits are
received (at age 70) reaches a maximum with the cohort that attains age 65 in 2007,
3
The private sector data are from the form 5500 data and exclude benefits paid directly by insurance carriers. Federal
DB benefits include payments made by the Civil Service Retirement System, the Federal Employees Retirement
System, and the Military Service Retirement System. DB and DC benefits are not reported separately for state and
local plans. The data used here include DC as well as DB benefits. Thus our projected DB benefits should be
somewhat greater than sum of these reported government and private sector benefits.
`
14
although this cohort continues to receive benefits after 2007. From a peak of $435,000 in
2027, the real value of total benefits paid declines modestly to $392,000 by 2040. As
shown in Figures 6-2 and 6-3, the real level of benefits increases through the end of the
projection period, so the decline in total benefits in Figure 7-2 is driven by the decline in
the probability of benefit receipt.
Figure 7-2. Projected benefits paid: by year
500,000
millions of year 2000 dollars
450,000
400,000
350,000
300,000
250,000
200,000
150,000
100,000
50,000
0
1982 1987 1992 1997 2002 2007 2012 2017 2022 2027 2032 2037
Year
For comparison, projected benefits paid from DB plans are graphed together with
the projected amounts withdrawn from 401(k) plans in Figure 7-3.4 Benefits from DB
plans exceed withdrawals of 401(k) assets until 2029. After 2029 the value of DB
benefits falls each year and 401(k) withdrawals increase rapidly thereafter.
4
The 401(k) withdrawals shown in this figure are projected amounts disbursed when account owners are
alive and do not include balances that remain in accounts when account owners die. We are in the process
of refining our withdrawal assumptions.
`
15
Figure 7-3. DB benefits paid vs 401(k) withdrawals
millions of year 2000 dollars
1,200,000
1,000,000
800,000
600,000
400,000
200,000
0
1982 1986 1990 1994 1998 2002 2006 2010 2014 2018 2022 2026 2030 2034 2038
Year
DB payouts
401(k) payouts (alive)
Suppose that in each year assets held by DB plan sponsors were equal to the
present value of these future obligations. If future liabilities are discounted at 3 percent
and firms have a twenty year planning horizon, the present value of liabilities each year is
shown in Figure 7-4. The figure also shows two series representing total assets in DB
plans. The first is from the Flow of Funds accounts that include private sector DB plans
and the value of assets in DB and DC plans for government sponsors. The second series
is composed of private sector DB assets from Form 5500 reports, federal DB assets from
various federal agency annual reports, and the sum of DB and DC assets for state and
local governments from Census reports. Actual assets, for the years they are available, are
substantially below this calculation of the present value of liabilities in all years, with the
possible exception of 1999.
It is perhaps not surprising that actual assets are less than our estimates of the
assets employers would have to hold to fully fund projected liabilities (based on our
assumptions). Private sector plan sponsors have substantial latitude in the assumption of
interest rates, investment returns, when benefits will be paid, and other features that
`
16
determine funding levels.5 There are even fewer restrictions on the funding of federal,
state, and local plans and many are thought to be substantially under-funded.
Figure 7-4. Present value of DB liabilities vs DB
assets
millions of year 2000 dollars
7,000,000
6,000,000
5,000,000
4,000,000
3,000,000
2,000,000
1,000,000
19
82
19
85
19
88
19
91
19
94
19
97
20
00
20
03
20
06
20
09
20
12
20
15
20
18
20
21
20
24
20
27
20
30
20
33
20
36
20
39
0
Year
Flow of Funds assets
F5500+EBRI+Census assets
PV of future liabilities
8. DB and “401(k) Assets Combined
How will assets in DB and “401(k)” plans combined change in future years? In
our prior work we projected assets held in 401(k) plans. These 401(k) assets, DB plan
liabilities (“fully-funded” DB assets) and the sum of both series are shown in Figure 8-1.
We also consider total “contributions” to DB and 401(k) plans combined and total
withdrawals from DB and 401(k) plans combined. Figure 8-2 below shows contributions
to 401(k) plans, the annual change in DB liabilities (a rough measure of contributions to a
fully funded DB plan) and the sum of these “contributions.” Projected total
“contributions” are relatively flat between 2000 and 2040. Withdrawals from 401(k)
plans, DB benefits paid, and the sum of the two are shown in Figure 8-3. The sum of
benefits paid grows rapidly through 2040.
5
Bergstresser, Desai, and Rauh (2006) discuss these issues in the context of earnings manipulation.
Bradley Belt (2005) estimated that private DB plans were underfunded by $450 billion in 2004.
`
17
Figure 8-1. Projected assets: 401(k), DB, and
combined
18,000,000
millions of year 2000 dollars
16,000,000
14,000,000
12,000,000
10,000,000
8,000,000
6,000,000
4,000,000
2,000,000
36
33
30
27
24
21
18
15
39
20
20
20
20
20
20
20
20
12
20
20
09
06
20
00
97
94
03
20
20
20
19
88
85
91
19
19
19
19
19
82
0
Year
401(k)
DB
Combined
Figure 8-2. Projected contributions: 401(k), DB
(change in liabilities), and combined
millions of year 2000 dollars
600,000
500,000
400,000
300,000
200,000
100,000
19
83
19
86
19
89
19
92
19
95
19
98
20
01
20
04
20
07
20
10
20
13
20
16
20
19
20
22
20
25
20
28
20
31
20
34
20
37
20
40
0
Year
Change in DB liabilities
`
18
401(k) contributions
combined
Figure 8-3. Projected withdrawals: 401(k), DB (benefits
paid), and combined
millions of year 2000 dollars
1,600,000
1,400,000
1,200,000
1,000,000
800,000
600,000
400,000
200,000
19
82
19
85
19
88
19
91
19
94
19
97
20
00
20
03
20
06
20
09
20
12
20
15
20
18
20
21
20
24
20
27
20
30
20
33
20
36
20
39
0
year
DB benefits paid
401(k) withdrawals
combined
Figure 8-4 shows the total projected contributions to and withdrawals from DB
and 401(k) plans combined. Figure 8-5 shows withdrawals minus contributions for DB
and 401(k) plans combined. Withdrawals exceed contributions after 2019 and the excess
reaches about $0.56 trillion by 2040 (in year 2000 dollars).6
Whether demographically induced changes in DB and DC asset levels will have
an appreciable effect on financial rates of return will depend upon the size of swings in
DB and DC assets relative to the size of the total asset market. The greatest concern of
some analysts seems to be the possible effect of demographic trends on equity markets.
We will consider the possible effect in subsequent analysis. For now, an example may
provide some intuition.
6
Schieber and Shoven ( 1997) consider the implications of population aging for private pension fund
saving and project saving as a percent of payrolls. Although their projection method is very different from
ours, their conclusions seem rather similar. Their projections suggest that total private pension
withdrawals will exceed contributions beginning in 2024 and would thus cease being a net source
of saving.
`
19
The effect depends on the size of the change in net withdrawals of pension
equities relative to the value of the total equity market. Between 1965 and 2005 the real
value of the equity market increased by 4 percent on average. If this rate were to
continue (on average) until 2040, the value of the equity market would increase from
$16.1 trillion in 2005 to 63.7 trillion in 2040 (in 2000 dollars). Based on our projections,
equities in pension plans would increase from $4.6 trillion to $9.1 trillion over this same
period (in 2000 dollars), but the pension plan share of the total equity market would
decline from about 28 percent to about 14 percent.7 In 2005, net pension plan
withdrawals (withdrawals minus contributions) from the equity market are about -$0.09
trillion. By 2040, net withdrawals would be about $0.5 trillion, or about 0.8 percent of
the total equity market. The magnitudes seem rather small to us.
Figure 8-4. Projected contributions and
withdrawals: DB and 401(k) plans combined
millions of year 2000 dollars
1,600,000
1,400,000
1,200,000
1,000,000
800,000
600,000
400,000
200,000
19
83
19
86
19
89
19
92
19
95
19
98
20
01
20
04
20
07
20
10
20
13
20
16
20
19
20
22
20
25
20
28
20
31
20
34
20
37
20
40
0
Year
Contributions
7
Withdrawals
Assuming that 60 percent of pension assets are in equities and that DB plans are 75 percent
funded.
`
20
Figure 8-5. Projected withdrawals minus
contributions: DB and 401(k) plans combined
millions of year 2000 dollars
1,000,000
800,000
600,000
400,000
200,000
0
19
82
19
85
19
88
19
91
19
94
19
97
20
00
20
03
20
06
20
09
20
12
20
15
20
18
20
21
20
24
20
27
20
30
20
33
20
36
20
39
-200,000
Year
9. Summary
The initial motivation for undertaking this analysis was to consider the potential
effect of demographic trends on the demand for financial assets and, in turn, on rates of
return in financial markets. This paper on DB pension plan assets is a companion paper
to our earlier paper on 401(k) assets. We have described the rather dramatic decline in
DB participation in the past three decades and we have assessed the implications of this
decline for the flow of funds into and out of DB trust funds over the next four decades.
We have compared the results for DB assets with our prior results on 401(k) assets. We
find that the projected withdrawals minus contributions to DB and 401(k)-type plans
combined rises from approximately -$0.09 trillion in 2005 to about $0.5 trillion by 2040
(in year 2000 dollars). This amount, however, seems to us to be small relative to the
likely size of the total equity market by 2040. Thus we suspect that the effect on rates of
return is unlikely to be large. Subsequent formal analysis will help to clarify the
magnitude.
We also projected the present value of DB benefits at age 65 for cohorts who
reach age 65 in each year from 1982 to 2040. These estimates are of considerable
interest independently of the effect of the changing level of retirement assets on financial
`
21
market rates of return. Our projections suggest that the average, over all persons, of the
present value of DB benefits (in year 2000 dollars) at age 65 will reach a maximum (at
$72,637) in 2003. The present value of benefits at 65 for persons with a DB plan also
reached a maximum in 2003 (at $153,561). Comparable data for 401(k) plans show that
the value of 401(k) assets surpasses the present value of DB benefits for all persons in
2011 and for persons with DB or 401(k) plans in 2012. Thereafter the value of 401(k)
assets grows rapidly, reaching levels much greater than the present value of DB benefits.
Over all persons, the value of 401(k) assets in 2030, for example, will be over three times
as great as the maximum value of DB assets attained in 2003. Based only on persons
with a DB or a 401(k) plan, the average of 401(k) assets at 65 will be over twice as large
as the maximum of DB assets attained in 2003.
Although our findings show that retirement saving through 401(k)-type plans in
future years will be much greater than retirement saving in DB plans has been, 401(k)
plans as well as DB plans have at least one important limitation. Like DB plans, 401(k)
plans are often not provided in small firms with predominantly low-income employees.
Perhaps the greatest improvement to the personal retirement plan system would come
from policies to hasten the penetration of 401(k) plans to lower-income employees.
References
Belt, Bradley D, Testimony Before the Committee on Education and the Workforce,
United States House of Representatives, March 2, 2005.
Bergstresser Daniel, Mihir A. Desai and Joshua Rauh “Earnings Manipulation and
Managerial Investment Decisions: Evidence from Sponsored Pension Plans”;
Quarterly Journal of Economics, vol 121, no. 1, 2006
Poterba, James, Steven Venti, and David A. Wise, 2004, “The Transition to Personal
Accounts and Increasing Retirement Wealth: Macro and Micro Evidence,” In D.A.
Wise, (ed.), Perspectives on the Economics of Aging, University of Chicago Press.
Poterba, James M., Steven F. Venti, and David A. Wise, "Demographic Change,
Retirement Saving, and Financial Market Returns." Working Paper Prepared for the
7th Annual Joint Conference of the Retirement Research Consortium, "Creating a
Secure Retirement System," August 11-12, 2005, Washington, D.C.
Schieber, Sylvester and John Shoven. “The Consequences of population Aging for Private
pension Fund Saving and Asset Markets,” in M. Hurd and N. Yashiro (ed.) The
Economic Effects of Aging in the United states and Japan. The University of
Chicago Press, 1997.
William Wiatrowski, “Documenting Benefits Coverage for all Workers”, USDOL/BLS,
December 2005.
`
22