View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

September 1995

Volume 1 Number 6

The Impact of Individual Retirement Accounts on Savings
Jonathan McCarthy and Han N. Pham

Bills to expand individual retirement accounts have been introduced in both houses of
Congress this year. While proponents argue that these accounts can help reverse the nation’s
declining saving rate, recent economic research suggests that the effect of the accounts on
savings is in fact quite small.

In recent years, analysts have often charged Americans
with saving too little. As evidence, they cite the decline
in the net national saving rate from an average of
8.4 percent of net national product (NNP) in the 1950s
and 1960s to 2.3 percent in the 1990s (Chart 1). 1
Although the large government budget deficits of the
1980s and 1990s contributed to this decrease, lower
personal savings clearly played an important role. The
decline in the national saving rate troubles policymakers for two reasons: savings may be insufficient to
fund the investment necessary to raise living standards,
and workers may not be saving enough for their retirement.
To arrest the decline in personal savings, many policymakers and economists have advocated expanding
targeted saving incentives such as individual retirement
accounts (IRAs). These accounts encourage saving by
exempting contributors from certain taxes typically
imposed on financial investments. The proponents of
these incentives believe that they would induce an
increase in personal savings well above any tax revenue loss, thus boosting national savings. The influence of this view is evident in a number of bills recently
introduced in Congress that would expand IRAs.
Nevertheless, IRAs may not be as effective in
increasing savings as their proponents believe.
Although IRAs do encourage households to save by

increasing the after-tax rate of return for saving, this
higher return also means that households need to save
less to achieve the same retirement income goal.
Furthermore, because households receive a tax break
for their contributions to an IRA regardless of the
amount of new savings, households may shift assets
into the accounts without increasing their total savings.
Such a switch has no effect on net household savings
but reduces national savings by lowering tax revenue
and increasing the budget deficit.
The uncertainty about the impact of IRAs on savings has prompted many researchers to investigate the
topic. This article examines what we can learn from
existing studies about the effect of IRAs on savings. 2
Although economists have reached conflicting conclusions, the balance of evidence suggests that while IRAs
may increase savings in the long run, the impact would
probably be modest. Moreover, the current IRA proposals are structured in such a way that they would
reduce any favorable impact on savings. We therefore
conclude that these proposals will not substantially
raise saving rates from their low levels.
History of IRAs and 401(k) Plans
IRAs were established in 1974 for workers not covered
by pension plans. Taxes on both the contributions and


tions since have continued to slip (Chart 2).

the accumulated interest were deferred until the assets
were withdrawn at retirement. Withdrawals before the
beneficiary reached 59 1/2 years of age were penalized.
Because IRAs were not widely promoted nor readily
available in the 1970s, however, few eligible taxpayers
contributed. Of the 51 percent of workers eligible to
contribute to IRAs in 1978, fewer than 3 percent did,
resulting in a mere $3 billion in contributions.

In 1978, another saving incentive similar to the IRA,
the 401(k) plan, was established. These plans did not
attract much interest until the early 1980s, after the
Treasury issued clarified rules on their tax advantages.
Like IRAs, 401(k) plans feature tax-deferral of contributions and accrued interest, annual contribution limits, and restrictions on early withdrawals. Some differences exist, however. Only workers of sponsoring
employers are eligible to contribute to 401(k) plans.
Contributions are made through payroll deductions and
can be matched by the employer, whereas IRA contributions can be made at any time and have no provision

To encourage savings, the Economic Recovery Tax
Act of 1981 expanded IRA eligibility to almost every
working taxpayer. Contribution limits were raised from
$1,500 to $2,000 for single taxpayers, from $1,750 to
$2,250 for one-income joint filers, and to $4,000 for
two-income joint filers. Financial institutions
responded to the expanded market by promoting IRAs
more heavily. Correspondingly, contributions jumped
dramatically between 1981 and 1982 (Chart 2). But
although contributions remained high through 1986,
they were never more than 1 percent of GDP. Moreover,
the small fraction of eligible taxpayers who opened
IRAs were concentrated in upper income groups. For
example, only 17.8 percent of eligible tax returns in
1985 claimed an IRA deduction, but about threefourths of taxpayers with income above $75,000 did.

Chart 2

IRA and 401(k) Contributions
Billions of 1993 dollars

The Tax Reform Act of 1986 curtailed the tax
deduction of IRA contributions for many workers.
Those covered by pension plans could deduct IRA contributions fully only if their income was below $25,000
for single filers and $40,000 for joint filers; the deduction was fully phased out for single and joint filers with
incomes above $35,000 and $50,000, respectively.
Higher income earners could continue to make nondeductible contributions and earn tax-deferred interest on
their accounts. These restrictions caused tax-deductible
contributions to fall 62 percent in 1987, and contribu-


1980 81













Sources: Internal Revenue Service, Statistics of Income; IRS Form 5500.
Notes: IRA contributions for 1993 are preliminary tabulations reported in
Weber (1994). Total contributions to 401(k) plans for 1991-93 were estimated
by using data on employee contributions from the 1993 Current Population
Survey multiplied by the 1988 ratio of total contributions to employee

for employer matching. The annual contribution limit
for 401(k) plans, $9,240 in 1994, is much higher than
that for IRAs, although employers can impose additional limits on eligible contributions.

Net National Savings as a Percentage
of Net National Product
Average for 1950s and 1960s = 8.5

As Chart 2 indicates, 401(k) plans gained popularity
in the late 1980s and early 1990s, in part as a supplement to (or substitute for) employer-sponsored pension
plans. Although this increase has to some extent counteracted the decline of IRAs, the total contribution to
targeted saving incentives as a percentage of GDP has
remained less than in 1986.

Average for 1970s = 7.9
Average for
1980s = 4.3




Chart 1



Average for
1990s = 2.3


Current Proposals
Despite waning taxpayer interest in IRAs, many policymakers and economists still view the accounts as a
reliable vehicle for encouraging savings. As saving
rates have continued to fall in the 1990s, a number of











Source: U.S. National Income and Product Accounts.



bills to expand IRAs have been proposed in Congress.

Washington is to balance the budget by about 2002,
and back-loaded IRAs appear to lose less revenue over
this time frame. However, such reasoning underestimates the ultimate impact of back-loaded IRAs on the
government budget. In fact, if marginal tax rates and
interest rates are constant, the incentives that a taxpayer receives from front- and back-loaded IRAs are
the same (see the box for a more detailed explanation).
Thus the ultimate impact of front- and back-loaded
IRAs on the government debt is identical under these

Separate bills have recently been introduced in the
House and the Senate, and the Clinton Administration
has put forth its own proposal. Although the proposals
differ in their details, they do share some common features. First, to encourage more contributions, these proposals would relax the penalties on early withdrawals
so that IRA holders could remove funds to pay for
higher education, a first home, or medical expenses.
Second, each of the proposals would establish a new
type of IRA. The contributions to these IRAs would not
be deductible, but the accumulated interest and
penalty-free withdrawals would be tax-free. This type
of IRA is described as “back loaded” because most of
the tax advantages are realized at the time of withdrawal; the current IRAs are termed “front loaded”
because most of the tax advantages are realized at the
time of contribution.

Moreover, even though the incentives created by the
two types of IRAs are equivalent under the stated conditions, back-loaded IRAs may not generate the contributions expected by their proponents. Because income
is likely to decline after retirement, most taxpayers’
marginal tax rates are expected to be lower at the time
of withdrawal than at the time of contribution. Thus,
many people will choose front-loaded over backloaded IRAs, reasoning that they will ultimately pay
less in taxes. In addition, taxpayers appear to value the

The back-loaded IRAs originated as a response to
near-term deficit concerns. The current desire in

The Saving Incentives Provided by Front- and Back-loaded IRAs
We compare the incentives provided by front- and
a front- or back-loaded IRA comes from the tax-free
back-loaded IRAs by means of an example (see
accrual of interest in the accounts, rather than from
table). Suppose a taxpayer has $2,000 of before-tax
the tax treatment of the contribution or withdrawal.
income in 1995 that is to be saved until retirement in
* Note that because contributions to back-loaded IRAs are
2010, and faces a constant 5 percent before-tax rate
made with after-tax dollars, a $2,000 limit is more generous for
of return and a constant 28 percent marginal tax rate.
a back-loaded IRA than for a front-loaded IRA.
If the taxpayer places the $2,000 in a taxable savings
account (column 1), he or she pays taxes on the
$2,000 (leaving $1,440 as the amount saved) and on
A Comparison of Front- and Back-loaded IRAs
the yearly interest. The account then will grow to
All Values in Dollars
$2,448 ($1,440×[1+(.05×[1-.28])]15) by 2010.
Front-loaded Back-loaded
Savings Account

If this taxpayer contributes to a front-loaded IRA,
he or she will not pay any taxes on these funds until
they are withdrawn in 2010. After paying these taxes,
the contributor will have $2,994 to spend, $546 more
than if the funds were in a taxable account (column
2). This difference is the result of the preferential
treatment of the IRA. (It also represents the increase
in the government debt caused by the program.)
Because the gain is not realized until 2010, its present
value is $546/(1.05)15=$263 in 1995 dollars (last row).

income, 1995
Less: taxes
paid, 1995
Equals: amount
contributed to
account, 1995
balance in 2010,
5% interest rate
Less: taxes
paid, 2010
Equals: after-tax
proceeds, 2010
Increase in
government debt
from IRA, 2010
Present value
of cost, 1995

If the taxpayer contributes to a back-loaded IRA
instead, he or she will have $2,994 ($1440×(1.05)15)
to spend in 2010, the same as under the front-loaded
IRA (column 3). Even though the taxpayer pays
taxes on the contribution in 1995, the amount paid is
the same in present value as the taxes paid at
withdrawal for a front-loaded IRA ($560 =
$1,164/(1.05)15).* The gain the taxpayer receives from










(28% tax)












Source: Hubbard and Skinner (1995, p. 12).



immediate tax write-off from a front-loaded IRA, probably for the psychological satisfaction it affords.3

These conflicting findings can be traced in part to
differences in the design of the studies. Besides a wide
variation in data sources, a key difference between
these studies is the choice of households whose saving
behavior is to be compared to identify the effect of
IRAs on savings. For example, Venti and Wise (1986,
1987, 1990, 1991) compare the savings of non-IRA
contributors and IRA contributors. In contrast, Gale and
Scholz (1994) compare households that contribute the
statutory IRA limit (limit contributors) with those that
contribute less than the limit (nonlimit contributors).

The Effect of IRAs on Savings
Some Ambiguities. The effect of IRAs on savings is
more ambiguous than many IRA proponents believe. It
depends on the interest rate elasticity of savings—that
is, the degree of response shown by savings to a given
change in the interest rate. This elasticity is influenced
by two counteracting forces. A higher after-tax interest
rate causes households to want to save more to become
wealthier in the future (the so-called substitution
effect), but the higher capital income in the present
encourages households to consume more and save less
(the income effect). Thus the magnitude of this elasticity remains controversial and can only be determined
by studying actual data on the response of savings to
interest rate changes.4

Both of these choices are problematic. Venti and
Wise’s observation that IRA contributors save more
than noncontributors does not necessarily mean that
IRAs have a strong effect on savings, because IRA con-

The balance of the evidence in the studies
surveyed points to a relatively modest
short-run effect of IRAs on savings.

A further complication arises in determining the
short-run, transitional effect of IRAs on savings.
Households will receive the tax break for their contributions whether or not the contributions are new savings. During the initial years of an IRA program,
households may shift existing assets into the accounts
without increasing their saving. So even if the program
ultimately has a positive effect on savings, the shortrun effect could be reduced until households no longer
desire to shift assets into the accounts.

tributors are more likely to be high savers, even in the
absence of an IRA program. Gale and Scholz attempt to
circumvent this problem by identifying noncontributors
as low savers and effectively leaving them out of the
comparison. However, by comparing limit and nonlimit
contributors, they may underestimate the impact of
IRAs on saving. For example, in their model, past IRA
contributions help to identify a household as a high
saver. But evidence indicates that households that contributed in the past usually continue to contribute. If
such a household made its original IRA contribution
because of the program’s incentives, its current IRA

Short-Run Effect. Because IRAs have existed for a
relatively brief time, only their short-run effect on savings has been examined empirically. The table below
presents the results from several major studies. As the
table indicates, these studies have arrived at very different conclusions: some find a substantial effect, while
others find almost no effect.

Studies on the Short-Run Effect of IRAs on Savings
Author(s) and Date

Data Source

Key Results

Hubbard, 1984

1979 President’s Commission on Pension Policy Survey

“Much of the contributed funds represent marginal saving”


1983 Survey of Consumer Finances

About 50% of contributions add to net national savings


1983 Survey of Consumer Finances

45-55% of contributions add to net national savings


1980-85 Consumer Expenditure Survey

65% of contributions add to net national savings


1984-85 Survey of Income and Program Participation

66% of contributions add to net national savings

Feenberg and Skinner, 1989

1980-84 IRS/University of Michigan Taxpayer Panel

IRAs cause only marginal reshuffling of assets

Attanasio and DeLeire, 1994

1982-91 Consumer Expenditure Survey

Less than 20% of contributions add to net national savings

Gale and Scholz, 1994

1983 and 1986 Survey of Consumer Finances

2% of contributions add to net national savings

Engen, Gale, and Scholz, 1994

1979-88 IRS/University of Michigan Taxpayer Panel

About 4% of contributions add to net national savings

Joines and Manegold, 1995

1979-86 IRS/University of Michigan Taxpayer Panel

19-26% of contributions add to net national savings

Venti and Wise



lation probably exaggerates the effect of these plans.
The national assets-to-income ratio in the model initially falls because the larger government budget
deficit overwhelms higher private savings. It then takes
thirty-five to fifty years before this ratio returns to its
original level, and the ultimate effect is not reached
until after seventy years. Over this entire period, many
variables in the model, such as interest rates, are
assumed to remain constant. If the model is modified to
allow these variables to change over time, the long-run
effect on the capital stock is considerably smaller.7

savings may be said to result from these incentives
rather than from its taste for saving.
Still, we believe that the balance of the evidence in
the studies surveyed points to a relatively modest
short-run effect of IRAs on savings. The most recent
studies listed in the table, which use data sources somewhat better suited for examining the issue than those
used in earlier work, indicate that one-quarter or less of
IRA contributions add to net national savings.
Moreover, even if the higher estimates of Venti and
Wise were correct, the effect of IRAs on aggregate savings would be modest because IRA contributions have
been a small fraction of NNP. During the peak years of
IRAs (1982-86), contributions were slightly more than
1 percent of NNP. The Venti and Wise estimates then
imply that IRAs during this period would have raised
the net national saving rate 0.7 percentage points at
most, a small increase compared with the 3.6 percentage point decline in the average net national saving rate
between the 1970s and 1980s (see Chart 1).

What does the research surveyed in this article imply
about the current proposals to expand IRAs? We
believe that the proposals will probably be even less
effective in generating savings than the current programs—front-loaded IRAs and 401(k)s—for several
reasons. First, the back-loaded IRAs in these proposals
are unlikely to increase contributions substantially
because taxpayers appear to prefer an immediate tax
write-off. Second, the proposals lack the employer
matching provisions that help make 401(k) plans more
effective in generating savings. Third, even though the
relaxed withdrawal provisions may encourage more
contributions at the outset, this effect would probably
be offset by greater asset-shifting in the short run and
by more withdrawals in the long run.

Long-Run Effect. Because data on the ultimate effect
of IRAs are not available, economists must turn to artificial simulation models to study the long-run effect.
Engen, Gale, and Scholz (1994) develop a life-cycle
model that incorporates many important features of
current IRAs and 401(k) plans: tax-deductible contributions, annual limits on contributions, and early withdrawal penalties.

As for IRAs in general, the short-run effect of these
accounts on savings is difficult to establish, but any
increase in savings is likely to be extremely modest.
An IRA program maintained over the long term is also
likely to have a small impact, although it might
increase savings at the margin. In sum, such programs
by themselves are probably insufficient to reverse the
recent decline in the saving rate. Thus, a major turnaround in the national saving rate will require much
more ambitious initiatives: a total overhaul of the tax
system to favor all savings, a large-scale effort to
reduce the budget deficit, or a significant change in
household attitudes toward saving.

Their simulations indicate that the ultimate effect of
these plans is likely to be fairly small even though the
interest rate elasticity in the model is sizable. 5 IRAs
eventually would increase net national savings about
3 percent if the contribution limit is $2,000, and about
5 percent if the limit is $4,000. The 401(k) plans,
which provide greater incentives, would increase net
national savings about 8 percent if a 100 percent
penalty is placed on early withdrawals, and about 17
percent if a 10 percent penalty is placed on early withdrawals.6 Consequently, if the initial national saving
rate is 6 percent (which is close to the initial saving rate
in the model and consistent with the post-World War II
average), an IRA program would eventually raise the
saving rate to 6.3 percent at most, and a 401(k) plan
would raise it to 7 percent at most.

1. The net national saving rate is calculated using data from the
National Income and Product Accounts (NIPA). It is gross saving
minus depreciation, all divided by net national product.

Even though the tax incentive programs show only a
modest effect on the saving rate, the model appears to
suggest a substantial cumulative impact on the capital
stock—the amount of funds available to be invested by
firms. Hubbard and Skinner (1995) calculate that the
most effective IRA and 401(k) plans studied by Engen,
Gale, and Scholz would eventually increase the capital
stock $4 and $16, respectively, for every dollar
increase in the government debt. However, this calcu-

2. Other recent surveys include Gravelle (1991) and Hubbard and
Skinner (1995).
3. Feenberg and Skinner (1989) provide some evidence on such
behavior. They find that a predictor of IRA contributions is whether
a taxpayer would owe money on his or her return. They interpret
this finding to mean that taxpayers prefer opening an IRA to paying
the Internal Revenue Service.




4. For example, Boskin (1978) estimates that savings would
increase by 0.4 percent for each 1 percent increase in interest rates.
In contrast to Boskin’s figure, which is at the upper bound of empirical estimates, Hall (1988) calculates an elasticity of close to zero.
Steindel (1981) explains that such calculations can differ greatly
because the elasticity depends on both the consumption function
and the relative responses of property income and wealth to a
change in the interest rate.
5. The implied (uncompensated) interest rate elasticity of savings in
the model is between 0.15 and 0.35, compared with the largest estimates of about 0.4 (see preceding note).
6. Under current law, allowable early withdrawals are penalized at
a 10 percent rate. However, early withdrawals are allowed only in
certain circumstances, such as employee separation from a firm and
“financial hardship,” so that the effective penalty rate is between
10 and 100 percent.
7. For example, as the capital stock increased in later periods, interest rates would probably fall. This effect would reduce the increase
in the capital stock. In the extreme, a simple general equilibrium
model such as the Solow growth model would imply a constant
steady state capital-to-income ratio. An IRA plan then would have
no effect on the capital stock in the long run.

Gale, William, and John Karl Scholz. 1994. “IRAs and Household
Saving.” American Economic Review 84 (December): 1233-60.
Gravelle, Jane G. 1991. “Do Individual Retirement Accounts Increase
Savings?” Journal of Economic Perspectives 5 (Spring): 133-48.
Hall, Robert. 1988. “Intertemporal Substitution in Consumption.”
Journal of Political Economy 96 (April): 339-57.
Hubbard, R. Glenn. 1984. “Do IRAs and Keoghs Increase Saving?”
National Tax Journal 37 (March): 43-54.
Hubbard, R. Glenn, and Jonathan Skinner. 1995. “The
Effectiveness of Saving Incentives: A Review of the Evidence.”
Unpublished paper.
Joines, Douglas H., and James G. Manegold. 1995. “IRAs and
Saving: Evidence from a Panel of Taxpayers.” University of
Southern California, unpublished paper.
Steindel, Charles. 1981. “The Determinants of Private Savings.” In
Public Policy and Capital Formation: A Study by the Federal
Reserve System, pp. 101-14.
Venti, Steven F., and David A. Wise. 1986. “Tax-Deferred
Accounts, Constrained Choice and Estimation of Individual
Saving.” Review of Economic Studies 53: 579-601.


———. 1987. “IRAs and Savings.” In Martin Feldstein, ed., The
Effects of Taxation on Capital Accumulation. Chicago:
University of Chicago Press.

Attanasio, Orazio, and Thomas DeLeire. 1994. “IRAs and
Household Saving Revisited: Some New Evidence.” National
Bureau of Economic Research Working Paper no. 4900.

———. 1990. “Have IRAs Increased U.S. Saving? Evidence from
Consumer Expenditure Surveys.” Quarterly Journal of
Economics 105: 661-98.

Boskin, Michael. 1978. “Taxation, Saving, and the Rate of
Interest.” Journal of Political Economy 86: S3-S27.

———. 1991. “The Saving Effect of Tax-Deferred Retirement
Accounts: Evidence from SIPP.” In B. Douglas Bernheim and
John B. Shoven, eds., National Saving and Economic
Performance. Chicago: University of Chicago Press.

Engen, Eric M., William G. Gale, and John Karl Scholz. 1994. “Do
Saving Incentives Work?” Brookings Papers on Economic
Activity 1: 85-151.
Feenberg, Daniel R., and Jonathan Skinner. 1989. “Sources of IRA
Saving.” In Lawrence H. Summers, ed., Tax Policy and the
Economy, vol. 3. Cambridge: MIT Press.

Weber, Michael E. 1994. “Individual Income Tax Returns, 1993:
Early Tax Estimates.” Internal Revenue Service Statistics of
Income Bulletin, Fall: 11-33.

About the Authors
Jonathan McCarthy is an economist and Han N. Pham an assistant economist in the Domestic Research
Function of the Research and Market Analysis Group.
The views expressed in this article are those of the authors and do not necessarily reflect the position of
the Federal Reserve Bank of New York or the Federal Reserve System.
Current Issues in Economics and Finance is published by the Research and Market Analysis Group of the Federal
Reserve Bank of New York. Dorothy Meadow Sobol is the editor of the publication.
Subscriptions to Current Issues are free. Write to the Public Information Department, Federal Reserve Bank of
New York, 33 Liberty Street, New York, N.Y. 10045-0001, or call 212-720-6134. Back issues are also available from
Public Information.