TAX REFORM AND

TARGET SAVING
ANDREW A. SAMWICK*

Abstract - If the United States switched debated during the primaries. Challeng-
to a broad-based consumption tax , theners have discovered a large constituency
all forms of saving would enjoy the tax-for a switch from the present income-
preferred status reserved primarily for based tax regime to a system based
retirement saving vehicles under the more on consumption taxes. Even
current income tax system. Because passing references to tax reform and tax
pensions have other unique characteris- simplification are met with considerable
tics besides their tax advantage , currentinterest from the financial community.
results on the effect of pensions on Much of the political attention is based
saving may provide an unreliable guide on the perception that a consumption
to the saving response to fundamental tax would increase the fairness and
tax reform. The net effect of reform on reduce the compliance costs of the tax
saving depends critically on household system.
motives for saving. This paper docu-
ments the considerable variation in the
There is also a belief that fundamental
reasons why households save and
tax reform would generate higher
presents a buffer stock model of saving
saving, thereby increasing the capital
that allows for both life-cycle and target
stock, worker productivity and living
saving. To the extent that specific
standards. The proponents of this belief
targets that are not currently tax sometimes cite as evidence the twin
favored motivate the saving of house-
observations that pension contributions
holds in their preretirement years,
account for a large share of private
fundamental tax reform that results in
saving and that pensions enjoy several
the elimination of current pension plans
tax advantages relative to other meth-
will reduce saving.
ods of saving.1 However, the empirical
literature on whether pension funds
actually represent "new" saving or net
additions to the capital stock has failed
to achieve a consensus.2 More impor-
tantly, the fact that pensions have
In the past two presidential election
characteristics other than their tax
years, proposals for fundamental tax
advantage that distinguish them from
reform were widely circulated and
other forms of private saving in the
'Dartmouth College, Hanover, NH 03755 and NBER, current tax regime makes it impossible
Cambridge, MA, 02138. to simply generalize that extending

621
consumption tax treatment to a greater Samwick (1995) is then used to show
share of savings vehicles will lead to that such target saving behavior is
higher saving. consistent with reasonable
parameterizations of the life-cycle model
In particular, pensions are an illiquid under uncertainty. The implications of
and, in some cases, involuntary method this model and the self-reported data
of saving over horizons that do not are then used to describe the likely
extend to retirement. Once the underly- responses of both target and life-cycle
ing model of household saving is savers to a consumption tax reform that
expanded from the standard life-cycle causes their pensions to be eliminated.
model to allow for nonretirement The last section discusses the implica-
reasons for households to save and for tions of this finding for the effect of
other capital market imperfections, fundamental tax reform on saving and
these characteristics - not the tax economic welfare.
advantage - may determine whether
pension savings are simply offset by PREFERENTIAL TAX TREATMENT
lower private savings. Chief among OF SAVING
these alternative motivations is the
desire to save as a precaution against The current tax regime is designed to
uncertainty in future budget constraints tax resources when they are earned as
or consumption needs. The main income. However, there are many
contribution of this paper is to show examples of income flows that receive
that the "target" saving described by preferential treatment in the form of a
Carroll (1997a) that may result in such alower effective tax rate. Such prefer-
model drastically changes the degree toences are often referred to as "con-
which households will offset lower sumption tax treatment," and funda-
pension wealth with higher wealth in mental tax reform that switched to a

other forms. Specifically, if pension consumption-based tax would be

coverage is not entirely voluntary and analogous to extending this treatment
people with pensions are not saving to a broader array of savings accounts.
primarily for retirement, then if employ- This section briefly illustrates the source
ers dismantle their pensions in the wakeof potential tax advantages in the
of tax reform, private saving will not context of employer-provided pensions
offset the lost pension saving and total before and after a tax reform. More
savings will fall. detailed discussions can be found in
Salisbury (1995) and Engen and Gale
The remainder of the paper is organized(1996a, b).
as follows. The next section briefly
describes the source of the relative tax Engen and Gale (1996b) use the
advantages to pensions under the following expression to denote the
current income tax system and how amount of wealth that a dollar of
those advantages would be extended to income could generate in T years if
other forms of saving under fundamen- saved by an employee in a taxable
tal tax reform. Self-reported data on account:
households' motivations for saving are
then analyzed, and several motives for
saving that result in target rather than n
retirement saving behavior are identi-
fied. The theoretical framework from B1 =(1 -t0-Sff)(1 +r( 1 -tJY
622
J TAX REFORM AND TARGET SAVING

where: income tax rate in year T and, possibly, a

penalty for early withdrawal.
t0 = the income tax rate when the
income is received, The tax advantage therefore has three
SEE = the share of the payroll tax paid components: (1) the employer's contri-
by the employee, bution is exempt from the payroll tax (if
r = the rate of return before indi- the account is employer sponsored); (2)
vidual (but after corporate) taxes the "inside buildup" on the contribu-
have been paid, and tions is not taxed; and (3) the income
f1 = the income tax rate on asset tax rate paid on the withdrawal may be
income between the current year lower than when the contribution was
and year T. made, especially if the withdrawal is
made during retirement. These advan-
Note that income and payroll taxes are tages may be sufficient to outweigh any
paid immediately on the full amount of potential penalties that are applied to
income when it is received by the early withdrawals, such as the 10
employee. Income taxes are also due percent excise tax and the 20 percent
each year on the taxable interest or withholding tax on distributions that are
dividends generated by the asset over not rolled over into qualified accounts.
the next T years. In contrast, if the These penalties are the price for
employer were to contribute that dollar converting an illiquid retirement account
to a pension account instead, it would into a liquid savings account.
generate the following amount after T
years: Under fundamental tax reform, some or
all of these relative advantages will be
eliminated from the perspective of a
a household that is actively saving for
retirement. In particular, the removal of
B2 = (1 +SfB)( 1 +r)V -tT-eT) the tax on interest (t,) will make saving
outside of pension accounts much less
disadvantaged relative to pension
where: contributions. This is the source of most
of the tax advantage to pensions and is
SER = the share of the payroll tax paid at the heart of all consumption tax
by the employer, proposals. If, in addition, the tax reform
tT = the income tax rate in year T eliminates some of the progressivity of
when the account is liquidated, the current tax system by imposing a
and, "flat tax" on consumption, then the
6t = the (possibly zero) tax penalty on advantages associated with a lower
liquidating the account in year T. marginal tax rate in retirement will
vanish. A reform that offered tax credits
Under this strategy, the initial amount for payroll taxes would eliminate the
that can be contributed includes the advantage that pensions enjoy through
employer's share of the payroll tax payroll tax deductibility. As discussed
because nonemployee pension contribu- below, if the tax advantage were made
tions are exempt from the payroll tax. more widely available, then the em-
That amount grows at the pretax rate of ployer costs of administering the
return while in the fund. Upon with- pension and complying with regulations
drawal, the funds are subject to the would also be eliminated.

623
HOUSEHOLD MOTIVES FOR SAVING motives with a "buffer stock" model of
consumption under uncertainty.
The importance of heterogeneity in
household saving behavior has recently
The SCF 1992 was the fourth in a series
become the topic of research on saving
and wealth holding.3 Economic intuition of triennial surveys conducted by the
Federal Reserve Board to obtain detailed
suggests that the magnitude of the
information on wealth and other
behavioral response to any price change
depends critically on the amount of the household characteristics. The sample
household's budget that was used to size in the SCF 1992 is 3906, comprised
purchase the item whose price changed. of an area-probability sample and an
oversampling of high-income taxpayers.
In the case of tax reform, it is the after-
tax rate of return on saving that Using the sample weights, the SCF is a
changes, and the item that is being nationally representative sample of the
purchased is, for example, retiremententire U.S. population of households.4
consumption. It is therefore essential Theto SCF contains several questions on
understand how important a motive household expectations and attitudes,
retirement is for saving. including a question that solicits up to
five of the household's most important
This section uses the household re- reasons for saving (though only one-
sponses to questions about financial third gave more than one and less than
decision making in the Survey of one-tenth gave more than two).
Consumer Finances (SCF) 1992 to
demonstrate that a majority of the Table 1 tabulates the responses to these
households are not in fact focusing on questions for all households in the
their retirement needs when making sample, stratified by the age of the head
their consumption plans. This result is of household. The head of household
quite general and is robust to control- for a married couple is the spouse with
ling for expected differences in pension the higher labor income. If the spouses
coverage, expected retirement income have the same income (usually zero),
adequacy, and the horizon for financial then the older spouse is selected as the
decision making. The next section head. The first row reports the percent-
ages of households that gave "retire-
reconciles the empirical results on saving
TABLE 1
HOUSEHOLD REASONS FOR SAVING, BY AGE

Age Category Any

Reason for Saving Under 25 25-34 35-44 45-54 55-64 Over 65 All Mention
Retirement 3.62 8.57 15.44 25.38 37.56 17.29 18.28 26.33
Uncertainty 26.57 37.75 32.83 30.78 31.69 37.60 34.07 43.73
Home purchase 7.02 9.86 4.13 2.43 1.64 0.31 4.03 5.87
Other housing 3.67 2.14 2.72 2.85 2.52 2.38 2.57 4.76
Transfers to family 9.36 13.40 15.06 9.92 2.93 3.37 9.48 16.41
Special purchases 10.04 5.81 5.29 6.12 3.19 6.95 5.87 12.99
Investment/business 0.69 1.13 1.26 0.04 0.16 0.88 0.78 1.50
Cover expenses 4.32 3.26 3.53 2.78 4.43 7.86 4.45 6.54
Just a good idea 20.35 9.21 8.49 7.97 3.52 8.35 8.51 11.44
Don't/can't save 14.35 8.89 11.26 11.73 12.36 15.02 11.97 12.52

Population share 5.22 20.63 23.17 16.48 12.77 21.74 100.00 -

Source: Authoris calculations from the Survey of Consumer Finances, 1992.
Notes:
(1) Question is: "What are your (family's) most important reasons for saving?"
(2) Total population is 95.92 million households.

624
| TAX REFORM AND TARGET SAVING

ment or old age" as their most impor- reflects other saving done to acquire
tant reason for saving. The values grow more valuable real estate, including
steadily with age before age 65, from a second homes, home improvements,
low of 3.62 percent for households and durable goods (many of which are
under age 25 to a high of 37.56 percent home furnishings). Roughly three
for households between 55 and 64. The percent of households across all age
column marked "AH" shows that, for categories save primarily for this reason.
the whole sample, 18.25 percent listed The fifth row shows that the most
retirement as their most important important reason for the saving of ten
reason. The last column in the table percent of the households is to make
("Any Mention") reports the percentage transfers to other family members,
of households who listed retirement largely through payments for their
among their most important reasons for education.7 The sixth row reflects saving
saving, even if it was not the first most done for special purchases, including
important. Approximately one-fourth of cars, vacations, own education, burial
the sample listed retirement as an expenses and charitable gifts. The last
important reason to save. two rows show that almost ten percent
of the households save primarily because
By itself, the finding that only one- they believe that saving is simply a good
fourth of the households list retirement thing to do, and a slightly larger number
as an important reason to save should do not save, primarily because they don't
cast suspicion on the most basic version believe they have enough resources.8
of the life-cycle hypothesis in which
retirement is the primary reason that One reason why Table 1 might give a
households save, regardless of their age. misleading account of saving motives is
The next row presents the analogous if survey respondents interpreted the
percentages of households reporting "saving" to exclude the saving that is
"uncertainty" as their most important implicitly done for them in the form of
reason for saving. The percentages pensions and Social Security. In order to
average 34.02 percent and show no address this possibility, Table 2 presents
systematic trend with age, although the the percentages of the working popula-
risks confronting households surely tion reporting "retirement" and "uncer-
change over the life cycle.5 The last tainty" as the most important reasons,
column shows that 43.73 percent of the controlling for expected retirement
households list uncertainty as an adequacy. The five categories are the
important reason to save. Uncertainty is households' answers to the question,
the most important reason for saving for "How would you rate the retirement
more households of every age except income you (expect to) receive from
55-64, when most households begin to Social Security and job pensions?" The
make the transition into retirement. three sets of numbers in the table are
for the full sample of households in
The remaining rows of the table present which the head is working, households
the analogous percentages for other in which at least one member is covered

savings motivations. The third row by a pension (55.14 percent) and

pertains to saving done to purchase the households in which no one is covered
household's primary residence. The by a pension (44.86 percent).
percentage starts at about nine percent
for households under 35 and then These tabulations confirm that retire-
declines with age.6 The fourth row ment is not the primary reason for

625
 TABLE 2
REASONS FOR SAVING, BY EXPECTED RETIREMENT INCOME ADEQUACY
Expected Retirement Income Adequacy

Sample Total|y Enough to Very

Saving reason Inadequate Inadequate Maintain
Workinq households
Retirement 1590 21 03 2448 17/13 2690 2045
Uncertainty 3401 37 31 3361 2971 31 06 3405
Population share 38 70 16 65 33 79 4 03 6 83

Bremen, ra9e<5514%) ^0.00 24.90 28.85 23.11 32

Uncertainty 3813 3496 3405 32 16 29 51 3498
Population share 3°^ ^.52 38.96 5.35 8.37

N°Re?irement<44 86%> 12 69 16 35 16 84 0
Uncertainty 3079 40/16 32 85 23 04 3427 32 92
Population share 484° ^.81 27.43 2.42 4.94 -
Source: Author's calculations from the Survey of Consumer Finances, 1992.
Question is: "How would you rate the retirement income you (expect to) receive from
pensions?"

saving. For the full sample, the percent-saving: there is a strong positive
age reporting retirement as the most relationship between the length of the
important reason is lower than that for planning horizon and the probability
uncertainty for each of the categories. that retirement is the most important
For the results by pension coverage, thisreason for saving, regardless of pension
result obtains in nine out of ten com- coverage. However, only 35 percent of
parisons. The percentage citing uncer- those without pensions and 45 percent
tainty fluctuates primarily between 30 of those with pensions have a financial
and 40 percent across retirement planning horizon of five years or greater.
income categories and is not particularlyIt is interesting to note that retirement
sensitive to pension coverage. The would be just as important a reason for
increasing percentages citing retirementsaving as uncertainty if the population
for higher expected retirement income were entirely like the "Over 10 Year"
are also not consistent with the objec- subsample.
tion, since households that expect low
returns from Social Security and
PRECAUTIONARY SAVING IN A BUFFER
pensions do not appear to make up for
STOCK MODEL
it by placing more of an emphasis on
saving for retirement. The results in Table 1 suggest three
different types of saving motives. The
Table 3 provides some evidence for whyfirst is the life-cycle motive, which is
retirement is not more prominently reflected by the households that cite
featured as a reason to save. The retirement as their most important
columns in the table are the households' reason for saving. The saving of these
responses to a question asking which households should be well described by
time period is most important to the the standard life-cycle model. The
household in planning its saving and second is the somewhat smaller
spending. The responses to that percentage of households that cite
question are consistent with the explicit targets such as home purchases
households' most important reasons forand family transfers. Their decision to

626
| TAX REFORM AND TARGET SAVING

TABLE 3
REASONS FOR SAVING, BY FINANCIAL PLANNING HORIZON

Most Important Time Period for Financial Planning

Sample Next Few Next Next Few Next 5-10 Over 10

Saving reason Months Year Years Years Years Total
Workinq households
Retirement 5.63 13.68 19.49 29.20 31.74 20.49
Uncertaintv 3586 3965 36 91 2923 30 69 34 12

"EZr"55 """ »5 .7.07 2264 34.53 34.30 25 30

Uncertainty 3946 4723 3930 2705 3002 3501
Population share 15 51 1134 25 64 26 96 20 56

^Retirement'4' 86%) ^ 10.36 15.41 20.63 26.47

Uncertainty 3320 3224 3382 32 76 32 56 3303
Population share
Source: Author's calculations from the Survey of Consumer Finances,
Question is: "In planning your (family's) saving and spending, which of
important to you (and your [husband/wife/partner])?"

save in one form or another should be B

motivated by how well the saving
vehicle enables the target to be
achieved. The third is the sizable
l(r-S>+!<r2n<g
proportion of the population that cites
uncertainty as the most important
reason for saving. Recent advances in where p is the coefficient of relative risk
the theory of precautionary saving by aversion, r is the interest rate, ô is the
Carroll (1992, 1997a) can be used to household's rate of time preference, o*
show that, under a broad range of is the variance of shocks to permanent
circumstances, these households will income and g is the rate of growth of
household income.
behave in a manner similar to target
savers. The model of consumption
Carroll (1 997a) shows that, if this
under uncertainty in Samwick (1995)
will be used below to illustrate this inequality is satisfied, then households
point. The model is general enough to will save in order to maintain a target
wealth-to-income ratio. He denotes this
encompass both precautionary and life-
behavior as Buffer Stock saving. The first
cycle motives for saving (though, for
term on the left-hand side of the above
simplicity, it abstracts from other explicit
targets for saving). Different para-
equation is the expected growth rate of
meterizations of the model can lead to consumption in an infinite-horizon
model with no uncertainty. Consump-
profiles of wealth accumulation over the
tion will grow more slowly if interest
life cycle that reflect varying degrees of
the two motives. rates are low, risk aversion is high,9 or
rates of time preference are high. Carroll
denotes the situation in which the
The essential result from this model can inequality holds in the absence of
be summarized by the following uncertainty (i.e., o^2 = 0) as "impa-
condition, which is based on the work tience." The second term on the left-
of Deaton (1991): hand side is the effect of income

627
uncertainty on the growth rate of
consumption. Kimball (1990) denotes
the precautionary saving motive as
"prudence." In this model, p reflects the
household's prudence as well as its
degree of risk aversion. The more
prudent the household or the greater
the uncertainty faced, the higher is
expected consumption growth. In other
words, the effect of precautionary
saving is to depress current consump-
tion and thereby raise the expected
growth rate of consumption. It is by
consuming less today that the house-
hold can provide some insurance against
the possibility of low consumption in
the future.
Thus, buffer stock behavior emerges
whenever the expected growth rate of
consumption is lower than the expected
growth rate of income, with the
magnitude of the buffer stock deter-
mined by the relative strength of
impatience and prudence. Carroll
(1997a) shows that buffer stock
behavior is more likely to obtain during
the early years of the life cycle, when
income growth is relatively high, and
shows that the buffer stock model is
consistent with a variety of empirical
regularities in the distribution of wealth. particular choice of parameters: p = 3, r
Carroll and Samwick (1997) present
evidence that the buffer stock model is
consumer will eventually behave as if
retirement is an important reason to
save as the date of retirement ap-
proaches. The key point is that, in a
finite horizon model, the appropriate
value of g on the right-hand side of the
inequality above is similar to a present
value of all future income growth rates.
Since retirement is represented in this
model by a large negative value for
income growth in the year of retire-
ment, the inequality will inevitably fail to
hold as the date of retirement ap-
proaches.
This result is presented graphically in
Figure 1 . The four different curves
correspond to progressively higher
retirement replacement rates of Social
Security and pensions, ranging from 25
to 100 percent of preretirement
income.10 A striking feature of this
graph is that all four households hold
almost identical buffer stocks for the 20
years between ages 25-45. During this
time period, the anticipated drop at
retirement is so far away that it has no
effect on wealth holdings. It is worth
emphasizing that this lack of sensitivity
to the replacement rate is not based on
any irrationality. Instead, it is due to the
= 0.02, and Ô = 0.08 in all periods; g =
0.01 5 and a = a = 0.10 before
rj e

an appropriate characterization of the retirement;

- ' r¡ and
e g - '
behavior of most households over most 0.00 after retirement. The condition
guaranteeing buffer stock behavior is
of their working lives. The buffer stock
easily satisfied at young ages.11
model is further supported by the results
for consumption profiles in Gourinchas
and Parker (1996). Using the buffer Each of the wealth profiles in which the
stock model to derive an empirical replacement rate is less than one do
specification, Carroll and Samwick eventually begin to accumulate more
(1998) find that approximately one-third
than just a buffer stock of wealth. This
of aggregate wealth can be attributedoccurs when the finite-horizon analog
to precautionary motives. to the inequality fails to hold as the date
of retirement inevitably approaches. The
A primary contribution of Samwick crossover point is earlier - and the
(1995) is to show through simulationssubsequent
of accumulation of wealth is
this model that even a buffer stock larger - for households with larger

628
| TAX REFORM AND TARGET SAVING

FIGURE 1. Wealth Profiles by Retirement Replacement Rate

anticipated drops in income at retire- rates. In the context of fundamental tax

ment. Note that this age-wealth profilereform, changing the replacement rate
is consistent with the results in Table 1
is akin to eliminating the retirement
that show that retirement becomes a income in the form of an employer-
more important reason for saving for anprovided pension. If the condition for
ever larger share of the households as buffer stock behavior is met, then the
saving response of households who save
they age. If the inequality is not satisfied
even at young ages, then the household primarily as a precaution against
never engages in a period of buffer uncertainty will be similar to that of the
stock saving and begins saving activelyhouseholds who were target savers for
for retirement immediately. Table 1 reasons such as a home purchase or
clearly shows the presence of young transfers for education. Households
households that already cite retirementwho save primarily to achieve
as their most important reason for preretirement targets will be unlikely to
saving. offset the reduction in expected
retirement income with higher values of
The salient feature of Figure 1 shows personal saving.
that over much of the working life,
households engaged in buffer stock
PENSION SAVING AFTER TAX REFORM
saving will not alter their wealth
holdings in response to fairly sizable Under the present income tax regime,
changes in their retirement replacement the tax advantages of pension plans are

629
associated with two primary costs. The employees from obtaining their exactly
first is the administrative cost to the preferred amount of tax-advantaged
employer of setting up and operating saving under current pension plans. A
the plan. If the tax advantages of complicated set of nondiscrimination
pensions are made available to employ- rules requires a rough parity of contribu-
ees outside of the pension, then a pure tions made across different categories of
efficiency gain results. The resources employees (highly versus nonhighly
that the employer previously devoted to compensated workers) if the contribu-
administering the plan will be returned tions are to be tax advantaged.13
to the shareholders as higher profits or Regulations such as nondiscrimination
the employees as higher wages. The rules are common when an exception
amount of this gain that is saved will (such as the consumption tax treatment
reflect the marginal propensities to save of pension contributions) is made to the
of those who receive it. income tax code to ensure that the
benefits of that exception are distrib-
The second is the cost associated with uted in a way that is deemed to be
the uniformity of the pension plan equitable. Garrett (1995) analyzes the
extent to which the nondiscrimination
across all workers at a given firm. Table
3 clearly demonstrates that there is rules are binding and demonstrates
substantial heterogeneity in both the large effects of the rules on the relative
motives and horizons for saving among contributions on behalf of highly and
workers with pensions. This heterogene- nonhighly compensated workers.
ity, little of which can be ascertained
systematically by the employer, meansGiven the presence of nondiscrimination
that it is not possible for the employer rules, employees who wish to access the
to design a pension plan that explicitlytax advantage of pensions must cross-
matches the desired amount of tax subsidize the pension accounts of other
advantaged saving on a worker-by- employees who would prefer to receive
worker basis. At best, employers can their compensation in wages. Enacting
establish a plan that provides the fundamental tax reform would allow the
optimal benefits for a large subset of employees who wish to save at the pre-
employees, perhaps those that the firm tax interest rate the opportunity to do
is most interested in retaining. However, so without having to comply with
given the inability of the employer to nondiscrimination rules. Because
ascertain which employees want the pensions will no longer be as valuable to
pension and which do not, the contribu- the workers who currently desire them
tions can be funded only by reducing most, it is reasonable to expect that a
the wages of all workers covered by thesizable percentage of firms that cur-
plan. The imperfect match between rently sponsor pensions will terminate
workers' saving preferences and the them as a result.14
degree to which their pensions allow
them access to tax-advantaged saving To simplify the analysis only slightly, the
imposes a welfare cost relative to a saving response by different workers to
more straightforward consumption tax fundamental tax reform can be viewed
treatment of all saving.12 as the combination of two factors. The
first is whether the worker was saving
Federal regulations on the distribution primarily to support consumption during
of pension contributions across employ-retirement as in the standard life-cycle
ees at a given firm further prevent model or to achieve more specific, near-

630
J TAX REFORM AND TARGET SAVING

term targets such as a buffer stock additionally outside of the pension to

against uncertainty, a home purchase, achieve their goals, possibly at the
or transfers to family members. The current after-tax rate of return. Given
second is whether the pension plan that the widespread result that workers
covers the worker offers sufficient covered by pensions have no less wealth
access to the tax advantage that fully outside of their pensions than do
achieves the worker's desire for saving, workers not covered by pensions, this
whether life cycle or target.15 scenario is probably the norm for
pension-covered workers.17 Maintaining
Consider first the workers who report the assumption that retirement savers
retirement as their primary reason for experience an increase in the after-tax
saving and have pensions that allow rate of return (due to the elimination of
them to achieve most of their retirement the cross-subsidy), the income effect will
saving objectives. If due to nondiscrimi- be comparatively large for them and, on
nation rules they were cross-subsidizing balance, they can be expected to
the saving of other workers covered by increase their consumption.
their pension plans, then fundamental
tax reform presents them with an The response of target savers will
increase in their after-tax rate of return analogously depend on whether their
on saving (equal to the size of the cross- pensions were previously a convenient
subsidy). This change in the return to means of achieving their target, despite
saving will induce the familiar income their illiquidity. Suppose that a worker
and substitution effects. Because was saving primarily for a down
retirement consumption is now cheaper,payment on a home purchase or a
households will save more out of child's education through a 401 (k) plan,
current income to purchase more in intending to borrow against the plan or
retirement, other things equal. The change jobs to receive the lump sum
income effect arises precisely because distribution (net of tax penalties) when
other things are not equal. Lowering the the funds were needed. Since the
price of retirement consumption meanspension met the goal at the lowest
that the level of retirement consumption possible cost, this worker would not
the household was previously expecting have large balances outside of the
can be achieved by saving less today. 401 (k) plan. Fundamental tax reform
The substitution effect leads to more would cause this saving to be done in
saving, but the income effect leads to a different tax-advantaged account,
less. Thus, even in the case of house- but the overall level of saving would
holds that are actively saving for remain about the same. Since these
retirement, a change in the after-tax targets are fairly near term, there would
interest rate when consumption tax not be large substitution or income
treatment is extended to another type effects, even if these workers were also
of income will have an ambiguous effect cross-subsidizing others covered by their
on saving.16 plan.

Next, consider workers who report Suppose instead that a target saver
retirement as their reason for saving but found that it was not possible to use a
who have pensions that do not offer pension plan to achieve the target. This
enough of a tax advantage to fully could happen if the pension had no
achieve their retirement saving objec- loan provisions or if the target was a
tives. These workers will be saving buffer stock and the prospect of

631
changing jobs just to receive the balance IMPLICATIONS FOR TAX REFORM
was not a desired outcome. Such a
An ongoing debate in the literature on
worker would be voluntarily contribut-
saving is whether amounts in pension
ing as little as possible to the pension
funds actually represent new saving. The
and saving for the target outside of the
analysis in this paper shows that, even if
plan. In the wake of fundamental tax
this claim is true, it provides an unreli-
reform, this worker would consume any
able guide to the likely effects of
pension balances accumulated on his
fundamental tax reform on saving. The
behalf. Because this worker would not
critical factor is the heterogeneity in
be cross-subsidizing any other worker
household motives for saving. Some
through the plan, he would be in the
pension balances are clearly the result of
same position as a worker not covered
extra saving to take advantage of the
by a pension in other respects. The shift
higher after-tax rate of return by
to consumption tax treatment of saving households who cite retirement as their
in liquid accounts introduces substitu-
most important reason for saving. But
tion and income effects on the entire
other pension balances are the result of
difference between the pretax and after-
households who don't actively save for
tax rates of return. Since the buffer
retirement being unable to dissave their
stock accumulated in Figure 1 is very retirement balances because what little
insensitive to the rate of return, the
wealth they do hold is earmarked for a
income effect is likely to dominate the
down payment on a home, a child's
substitution effect and the worker will
education, or a buffer stock against
consume more immediately and achieveuncertain income.
his target with lower saving at the
higher interest rate.18
When tax reform extends the higher
after-tax rate of return to forms of
Some evidence on the issue of what
saving other than pensions, the workers
would happen to pension account
balances after fundamental tax reform at each firm who valued the opportunity
to save for retirement at a higher
can be inferred from Chang (1996). She
examines the effect of the introduction interest rate will no longer need to rely
of the ten percent penalty on lump sum
on a pension plan to do it. Because the
distributions under the Tax Reform Act pension plan imposes costs of both
administration and regulation, there will
of 1986 on the probability that house-
holds who received distributions "rolled be a strong incentive for these savers to
them over" into another retirement take their compensation in the form of
account. She finds that rollover rates higher wages and save the extra pay
outside of a pension. While their total
are fairly insensitive to the tax penalty
and attributes the result to the presence
saving may not change too much in
either direction, the saving that was
of liquidity constraints among low-
previously done on behalf of workers
income households. Although the
who were otherwise target savers will
receipt of a distribution is endogenous
simply not be made up by higher
in that context,19 her results do suggest
personal saving.
that, if workers were given access to
their pension funds in liquid form in the
aftermath of a consumption tax re- This somewhat pessimistic conclusion
form, they would be inclined to spend demonstrates that there is a clear
them rather than continue to save distinction to be made between policies
them. that raise the level of saving and policies

632
| TAX REFORM AND TARGET SAVING
that raise the level of economic
welfare. While it is true that, if the
nation's capital stock were increased
exogenously or possibly due to lower
taxation of capital income (with the
revenue recouped in a more efficient
manner), the extra capital could be put
to use in production and raise the
standard of living. That is not exactly
what happens when the government
offers a tax advantage to pension
saving in an income tax system and
requires cross-subsidies to target savers.
The target savers have other economic
goals in mind, and forcing retirement
balances on them (possibly at the
direct cost of lower wages) is not the
way to make them better off. This is
especially true if they are liquidity
constrained in their consumption
decisions.
This distinction does not mean that
fundamental tax reform should not be
pursued with the goal of raising the
level of national saving. It simply
cautions that the distribution of pension 5 Households could give four different answers to
saving across the population may
currently be at odds with individual
household objectives. Under the current
tax system, both the tax advantage and
the illiquidity are important aspects of
how much wealth is held in the form of
pensions. Tax reform will open up
opportunities for households to make
themselves better off, possibly by
changing their compensation from
pension contributions to wages if
pensions lose all of their relative tax
advantages. The theory and evidence
presented in this paper strongly suggest
that this will result in a reduction in
private saving. This possibility must be
weighed against the more traditional
predictions from standard life-cycle
models, which suggest that a higher
after-tax rate of return will raise saving
in the long term.
633
ENDNOTES
An earlier version of this paper circulated with the
title, "The Effects of Tax Reform on Pension Saving
and Nonpension Saving," and was originally
presented at the American Enterprise Institute,
October 25, 1996. I am grateful to Tom Barthold,
Alan Gustman, Hank Gutman, Kevin Hassett,
Glenn Hubbard, Jon Skinner, Joel Slemrod, Steve
Venti, and two anonymous referees for helpful
suggestions. Any errors are my own.
1 See Sabelhaus (1997) for a decomposition of
personal saving into pension and nonpension
components.
2 The literature is comprised of two sets of studies.
The first pertains to Social Security and defined
benefit pension plans and is well summarized in
Gale (1995). The second set of studies analyzes
individual retirement accounts and 401 (k) plans,
which have more of a role for individual
contributions. Hubbard and Skinner (1996) provide
a partial reconciliation of the mixed results in this
literature.
3 Dynan, Skinner, and Zeldes (1997) and Carroll
(1997b) have addressed the issues of whether and
why rich households save more than other
households do. Samwick (1997) examines the
extent to which heterogeneity in tastes,
parameterized by the discount rate, is a viable
explanation for the distribution of wealth.
4 See Kennickell (1996) for documentation of the
SCF 1 992. The SCFs from 1 989 and 1 995 yield
similar results.
be
coded into this category: (1) reserves in case of
unemployment, (2) in case of illness or medical/
dental expense, (3) emergencies, "rainy days," and
other unexpected needs or for "security"/
independence, and (4) liquidity or to have cash
available/on hand. Of these, (3) is by far the most
important.
6 See Engelhardt (1996) for an analysis of saving for
downpayments.
7 The specific responses are (1) children's or
grandchildren's education and (2) "for the children/
family" or "to help the kids out." See Feldstein
(1995) for an analysis of saving for college
educations.
8 Samwick (1998) shows that results are similar
when the responses are weighted by the
household's amount of financial assets. A dollar-
weighted calculation is more appropriate when
answering the question of how much wealth will
be offset, rather than what fraction of households
will be affected. The overall dollar-weighted
fraction reporting retirement is equal to that
reporting uncertainty, with retirement prevailing by
a substantial margin between ages 45-64. Since
pension wealth is far less concentrated than
financial assets, the most appropriate tabulations
 lie intermediate between those in Table 1 and the
dollar-weighted calculations. Even so, these
differences are not large enough to be inconsistent
with the application of the buffer stock model
discussed in the text.
9 When Constant Relative Risk Aversion utility is
assumed, then the inverse of the coefficient of
relative risk aversion is the intertemporal elasticity
of substitution. A high degree of risk aversion is
therefore analogous to a low willingness to
substitute consumption from the current period to
future periods. This parameter restriction does not
affect the basic interpretation of the condition.
10 This model assumes that all pensions are defined
benefit and can be characterized by their promised
retirement income flows. See Engen (1994) for a
model in which the household jointly chooses the
level of saving and pension contributions.
11 The high value for the time preference rate is
chosen to emphasize the point that even very
impatient consumers will eventually save for
retirement as long as income drops at retirement.
The same qualitative results hold (with a larger
buffer stock) when a lower value is chosen as long
as the inequality is still satisfied. However, see
Carroll and Samwick (1997), in addition to the
results in Table 3, for a justification of a time
preference rate of this magnitude.
12 There are circumstances under which this welfare
cost can be at least partially recouped. For
example, if there are sufficiently many different
combinations of wages and pension plans across
employers, then workers can sort themselves
according to their preferences for saving without
compromising their productivity. Given the
heterogeneity in Table 3, this seems unlikely.
Another possibility is for the pension plan itself to
accommodate many different tastes for saving.
Ippolito (1996) presents a model in which this is
accomplished via the employer match rate in
401 (k) plans.
13 See McGill et al. (1996) for a description of
nondiscrimination rules.
14 Pensions will not disappear entirely because there
are other reasons to set up pensions beyond the
tax advantages. The most prominent reasons
include the ability to purchase group annuities, to
control turnover and retirement through deferred
compensation (Gustman and Steinmeier, 1995;
Kotlikoff and Wise, 1987), and to sort workers
based on unobservable aspects of productivity
(Ippolito, 1996).
15 It is worth noting that the same factors are relevant
for the more familiar debate of whether the
amounts in pension funds represent new saving
that would not have otherwise been done if
pensions were not available. That debate concerns
proposals to extend the availability of tax-
advantaged saving through the current system,
rather than through fundamental tax reform.
634
16 Note that this set of effects is similar to that
experienced by a worker saving for retirement who
did not have access to a pension. The latter worker
would experience the effect for the entire change
between the pretax and after-tax interest rates,
however, rather than just the pretax return net of
the cross-subsidy to the pretax return itself.
17 See Gustman and Steinmeier (1998) for a recent
example of such a study and a review of the
literature.
18 It is also possible that the size of the non-
precautionary targets expands (i.e., a bigger house
or a better college) or that the higher interest rate
increases consumption growth to the point where
the worker no longer engages in buffer stock
behavior.
19 Lump sum distributions are typically received only
when an employment relationship ends. The set of
workers who change employers is likely to be less
committed to retirement saving than those who
continue with an employer.
REFERENCES
Carroll, Christopher D. "The Buffer Stock
Theory of Saving: Some Macroeconomic
Evidence." Brookings Papers on Economic
Activity No. 2 (1992): 61-135.
Carroll, Christopher D. "Buffer Stock Saving
and the Permanent Income Hypothesis."
Quarterly Journal of Economics 112 No. 1
(February, 1997a): 1-55.
Carroll, Christopher D. "Why Do the Rich Save
So Much?" OTPR Working Paper No. 98-12. Ann
Arbor: Office of Tax Policy Research, University of
Michigan, 1997b.
Carroll, Christopher D., and Andrew A.
Samwick. "The Nature of Precautionary
Wealth. " Journal of Monetary Economics 40
(August, 1997): 41-72.
Carroll, Christopher D., and Andrew A.
Samwick. "How Important Is Precautionary
Saving?" Review of Economics and Statistics
(forthcoming, 1998).
Chang, Angela E. "Tax Policy, Lump-Sum Pension
Distributions, and Household Saving." National
Tax Journal 49 No. 2 (June, 1996): 235-52.
Deaton, Angus. (1991). "Saving and Liquidity
Constraints." Econometrica 59 No. 5 (Septem-
ber, 1991): 1121-42.
Dynan, Karen, Jonathan S. Skinner, and
Stephen P. Zeldes. "Do the Rich Save More?"
Federal Reserve Board. Manuscript, January, 1997.
Engelhardt, Gary V. "Consumption, Down
Payments, and Liquidity Constraints." Journal of
Money, Credit, and Banking 28 No. 2 (May,
1996): 255-71.
J TAX REFORM AND TARGET SAVING

Engen, Eric M. "Precautionary Saving and the Ippolito, Richard. "Pensions, Sorting, and
Structure of Taxation." Federal Reserve Board. Internal Discount Rates." Pension Benefit Guar-
Manuscript, 1994. anty Corporation. Manuscript, January, 1996.
Engen, Eric M., and William G. Gale. "The Kennickell, Arthur. Codebook for 1992 SCF.
Effects of Fundamental Tax Reform on Saving." Washington, D.C.: Federal Reserve Board, April,
In Economic Effects of Fundamental Tax Reform, 1996.

edited by Henry Aaron and William Gale. Wash- Kimball, Miles S. "Precautionary Saving in the
ington, D.C.: Brookings Institution Press, 1996a. Small and in the Large." Econometrica 58 No. 1
Engen, Eric M., and William G. Gale. (January, 1990): 53-73.
"Comprehensive Tax Reform and the Private Kotlikoff, Laurence J., and David A. Wise.
Pension System." Tax Notes 72 No. 3 (July 1 5, "The Incentive Effects of Private Pension Plans."
1996): 345-52.
In Issues in Pension Economics , edited by Zvi
Feldstein, Martin S. "College Scholarship Rules Bodie, John B. Shoven, and David A. Wise.
and Private Saving." American Economic Review Chicago: University of Chicago Press, 1987.
85 (1995): 552-66.
McGill, Dan M., Kyle N. Brown, John J. Haley,
Gale, William G. "The Effects of Pensions on and Sylvester J. Schieber. Fundamentals of
Wealth: A Re-Evaluation of Theory and Private Pensions. 7th ed. Philadelphia: University
Evidence." The Brookings Institution. Manuscript, of Pennsylvania Press, 1996.
June, 1995.
Sabelhaus, John. "Public Policy and Saving in
Garrett, Daniel M. "The Effects of Nondiscrimi- the United States and Canada." Canadian
nation Rules on 401 (k) Contributions." Stanford Journal of Economics 30 No. 2 (May, 1997):
University. Manuscript, December, 1995. 253-75.

Gourinchas, Pierre-Olivier, and Jonathan A. Salisbury, Dallas L. "Employee Benefits in a F

Parker. "Consumption Over the Lifecycle." Tax or Consumption Tax World." EBRI Notes 1 6
Massachusetts Institute of Technology. Manu- No. 9 (September, 1995): 1-11.
script, March, 1996. Samwick, Andrew A. "The Limited Offset
Gustman, Alan L., and Thomas L. Steinmeier. Between Pension Wealth and Other Private
Pension Incentives and Job Mobility. Kalamazoo:
Wealth: Implications of Buffer-Stock Saving."
W. E. Upjohn Institute, 1995. Dartmouth College. Manuscript, December, 1995.
Gustman, Alan L., and Thomas L. Steinmeier. Samwick, Andrew A. "Discount Rate
"Effects of Pensions on Savings: Analysis with Heterogeneity and Social Security Reform." NBER
Data from the Health and Retirement Study." Working Paper No. 6129. Cambridge, MA:
Carnegie-Rochester Series on Public Policy National Bureau of Economic Research, October,
(forthcoming, 1998). 1997.

Hubbard, R. Glenn, and Jonathan S. Skinner. Samwick, Andrew A. "The Effects of Tax
"Assessing the Effectiveness of Saving Incen- Reform on Pension Saving and Nonpension
tives. " Journal of Economic Perspectives 1 0 Saving." American Enterprise Institute.
(1996): 73-90. Manuscript, 1998.

635