tax break - Brookings · president’s proposal to cut dividend and capital gains taxes and with proposals for fundamental tax reform. Section IX closes with comments on reform of

Leonard E. Burman, William G. Gale, and Peter R. Orszag, Tax Policy Center, take a look at some of President Bush’s tax-free saving proposals. Leonard E. Burman, William G. Gale, and Peter R. Orszag, Tax Policy Center, take a look at some of President Bush’s tax-free saving proposals.

The Administration’s SavingsProposals: Preliminary Analysis

I. IntroductionIn its fiscal year 2004 budget, the Bush adminis-

tration proposes to create a new set of tax-preferredaccounts that would expand opportunities and con-solidate rules for tax-advantaged saving. The initialreaction to the proposal was not particularly positive.Less than a week after the budget was released, con-gressional Republicans dismissed the plan. The WhiteHouse reportedly reacted by blaming the idea onrecently departed Treasury Secretary Paul O’Neill(Vanderhei 2003). Congressional leaders, however,later softened their initial remarks and indicated thatsome elements of the proposal might be included inlegislation aimed at expanding access to pensions(Rojas 2003). Then White House officials were said toabandon the idea (Andress 2003).

Despite its uncertain prospects, the proposal isworth considering in detail because it would dramati-cally alter the tax treatment of saving, via the creationof Lifetime Saving Accounts (LSAs), individual Retire-ment Saving Accounts (RSAs), and Employer Retire-ment Saving Accounts (ERSAs). Some elements of theproposal — in particular, some of the simplifications— might form the basis of a useful pension reformpackage. Other elements are troubling because theywould be regressive, could reduce saving among themost vulnerable populations, and would exacerbatethe already bleak long-term budget outlook. In thispaper, we provide a preliminary analysis of selectedaspects of the proposal, with the following principalconclusions:

• The proposal would substantially expand op-portunities for tax-sheltered saving, particularlyfor households with incomes above the currentlimits on IRAs. LSAs would represent a historicchange in the tax treatment of saving — allow-ing significant amounts of tax-free saving($7,500 per account per year) for any purpose,with no restrictions on age or income. RSAswould be designed similarly, but tax-free with-drawals could be made only after age 58 or thedeath or disability of the account holder. RSAswould significantly expand annual contributionlimits for retirement saving outside of work;would remove all eligibility rules related to age,pension coverage, or maximum income;eliminate minimum distribution rules while theaccount owner is alive; and allow conversionsof traditional and nondeductible IRAs into thenew backloaded saving vehicles without regardto income.

• The proposal would result in growing revenuelosses over time, with the annual revenue lossexceeding 0.3 percent of GDP within a decadeand more thereafter. The transfer into LSAs ofexisting taxable assets and of future saving thatwould have been done in taxable form other-wise could drain between $200 billion and $300billion in revenues over the next decade, withthe amounts increasing as a share of GDP overtime. By 2013, the annual revenue loss wouldexceed 0.3 percent of GDP and cost even morethan the administration’s dividend tax pro-posal. Contributions to RSAs that would haveinstead been saved in taxable accounts wouldfurther reduce revenue over time. Rollovers oftraditional IRAs into the new RSAs would raiserevenue in the short run, but reduce it even moreover the long run in present value terms. Al-though there is considerable uncertainty aboutthe exact short-term revenue effects, all of thesefactors suggest significant long-term revenuelosses. After 25 years, the loss in revenue is like-ly to exceed 0.5 percent of GDP per year (theequivalent today of more than $50 billion a year,given the current size of the economy).

• Higher-income households would receive a dis-proportionate share of the benefits from the pro-posal. The loosening of rollover rules, forexample, would affect only households with in-comes exceeding $100,000. The elimination ofincome limits on eligibility for retirement savingaccounts would mainly benefit households witheven higher income. The ability to shift existing

tax breakby Leonard E. Burman, William G. Gale,

and Peter R. OrszagTAX ANALYSTS®

The authors are codirectors of the Tax PolicyCenter (http://www.taxpolicycenter.org). Burmanis a Senior Fellow at the Urban Institute. Gale isthe Arjay and Frances Fearing Miller Chair inFederal Economic Policy at the Brookings Institu-tion. Orszag is the Joseph A. Pechman senior fellowat Bookings. The authors thank David Gunter, MattHall, and Jeffrey Rohaly for programming assis-tance, and Mark Iwry, Craig Copeland, JaneGravelle, Bob Greenstein, and Jim Horney for help-ful comments and discussions. The views ex-pressed are their own and should not be attributedto the trustees, officers, funders, or staff of theBrookings Institution, the Urban Institute, or theTax Policy Center.

TAX NOTES, March 3, 2003 1423

(C) T

ax Analysts 2003. A

ll rights reserved. Tax A

nalysts does not claim copyright in any public dom

ain or third party content.

taxable assets into LSAs will predominantlybenefit households with substantial existing as-sets, which tend overwhelmingly to be thosewith high income. As a result, the proposal willgrow increasingly regressive over time.

• The proposal would not raise private saving bymuch, if at all, over the next decade and wouldbe almost certain to reduce national saving. Thetransfer of existing taxable assets into LSAswould reduce taxes but not raise private saving.New contributions to RSAs by very-high-income households currently constrained byIRA income limits are also unlikely to representnet additions to saving.

• The proposal could result in reduced employer-based pension contributions for rank-and-fileworkers. Reduced pension contributions amongrank-and-file workers would reduce the per-centage of American households that are savingadequately for retirement.

• The proposal would undermine some of thepotential salutary effects of the president’s pro-posal to reduce taxation of dividends. That pro-posal would encourage firms to reduce taxsheltering to some degree and pay more divi-dends but only to the extent that shareholderswould otherwise face income taxes on their div-idends and capital gains. The massive expan-sion in tax-free saving that LSAs would producewould significantly increase the proportion ofshareholders that do not face individual tax andso would dull the incentives created by the div-idend plan.

• The saving proposal should not be confusedwith fundamental tax reform. A revenue-neutralconsumption tax might raise growth moderate-ly, but the current proposal is not revenue-neutral, especially in the long run. Furthermore,while a consumption tax would exempt only thereturn to new capital investments, the adminis-tration’s proposal would exempt the return to ex-isting capital. In other words, under the proposal,much of the revenue loss would simply provide awindfall gain on old investments, which doesnothing to spur economic growth. As a result, theproposal would retain the regressivity of manyconsumption tax proposals without many of thepotential growth benefits. Unlike a comprehen-sive consumption tax, the proposal would alsocontain incentives for inefficient tax shelters thatwould further undermine revenue.

• The proposal as a whole would simplify the taxsystem somewhat. LSAs would be an even moresignificant simplification if they replaced exist-ing targeted saving accounts, but they insteadsupplement those plans, which could compli-cate tax planning. RSAs would simplify retire-ment saving. The interaction between both LSAsand RSAs and the current-law saver ’s creditwould, however, lead to considerable com-plexity for lower-income taxpayers and createtraps for the unwary. ERSAs could conform pen-

sion rules that apply to several different typesof retirement plans. They would also simplifythe nondiscrimination rules, but potentially atthe cost of reducing benefits for lower- andmiddle-income workers.

• While simplification efforts are needed, this pro-posal would also create some new complexities.Moreover, simplification does not require mas-sive increases in contribution limits or substan-tial new tax-preferred savings vehicles. Themost costly features of the new proposal havenothing to do with simplification and much todo with allowing high-income households toshelter much more wealth than under currentlaw.

• A better strategy would encourage expandedpension coverage and participation among low-and middle-income households, which wouldboost national saving and build wealth for manyhouseholds who are saving too little. Reformsalong these lines could include expanding theincome eligibility range for the saver ’s creditand making the credit refundable, changingdefault choices in 401(k) plans, improving finan-cial education, encouraging diversification of401(k) accounts, and simplifying pension rules(including exempting a modest amount per in-dividual from the minimum distribution rulesand transforming the nondiscrimination rulesinto a minimum contribution regime).

Section II describes the proposed changes. SectionIII examines the potential for tax simplification. SectionIV briefly describes what we view as the major struc-tural changes in saving rules and the likely set ofrelevant behavioral effects to consider. The followingthree sections examine the effects on revenue, the dis-tribution of benefits from the tax cut, and private andnational saving. Section VIII discusses some broaderaspects of the plan, including its interaction with thepresident’s proposal to cut dividend and capital gainstaxes and with proposals for fundamental tax reform.Section IX closes with comments on reform of tax rulesfor pensions and other saving.

II. The Proposal

Under the president’s proposal, every individualcould set up a Lifetime Saving Account (LSA) and aRetirement Saving Account (RSA).1 Contributions toeach account would have to be in cash, would not betax-deductible, would be limited to $7,500 per accountper year (indexed for inflation), and could be made bypersons other than the account holder. There would beno maximum income or age limits on contributions,

1Unless noted otherwise, all details about the proposal aregleaned from U.S. Treasury (2003b).

(Text continued on p. 1427.)

COMMENTARY / TAX BREAK

1424 TAX NOTES, March 3, 2003

(C) T

ax Analysts 2003. A




Table 1Tax-Preferred Savings Accounts

YearCreated

Qualified Useof Funds

EligibilityRules

ContributionLimits

TaxTreatment

Changes UnderAdministration

Proposal

Archer Medicalsavings account

1996 Unreimbursedmedical ex-penses

Self-employed,small businessowners, and em-ployees ofsmall businessowners. All par-ticipants mustbe covered by ahigh deductiblehealth plan(HDHP).

75% of HDHPdeductible(family); 65% ofHDHP deduct-ible (self-only);no more than in-come

Tax-deductiblecontributions,earnings anddistributionstax-free for qual-ified expenses

Could be con-verted to anLSA before2004. Balancewould be taxedon conversion.Could coexistwith theproposed plans.Administrationhas proposedan increase inthe contributionlimits.

CoverdellEducationSavings Ac-counts

1997 Certain elemen-tary, secondary,and highereducation ex-penses

Beneficiarymust be under18 (excludingspecial needsbeneficiary)Modified AGImust be below$220,000 (mar-ried filing joint-ly), under$110,000(single);phaseouts beginat $190,000 and$95,000, respec-tively

$2,000 perbeneficiary

Tax-free earn-ings and with-drawals forqualified ex-penses

Could be con-verted to LSAbefore 2004.Balance wouldnot be taxed onconversion.Could coexistwith theproposed plans,including newcontributions.

Qualified statetuition program(Section 529plan)

1996 Tuition, room,board, fees,books, and sup-plies at ac-credited institu-tions of highereducation.

Anyone can es-tablish an ac-count for adesignatedbeneficiary.

Varies by state;the contributionlimit is oftenhigh; for ex-ample, inCalifornia con-tributions areallowed untilthe accountreaches $267,580

Tax-free earn-ings and dis-tributions forqualified ex-penses

Could be con-verted to LSAbefore 2004.Balance wouldnot be taxed onconversion.Could coexistwith theproposed plans,including newcontributions.

Lifetimesavings account

Proposed No limitations Unrestricted $7,500 Tax-free earn-ings and dis-tributions

Sources: Internal Revenue Service, “Medical Savings Accounts (MSAs),” Publication 969; Internal Revenue Service, “TaxBenefits for Education,” Publication 970; UPromise, “529 Questions and Answers,” available athttp://www.upromise.com/savingsplans/what529/faq.html; http://www.savingforcollege.com, “The 529 PlanEvaluator”; U.S. Department of Treasury, “General Explanations of the Administration’s Fiscal Year 2004 Revenue Pro-posals,” February 2003.



(C) T

ax Analysts 2003. A




Table 2Individual Retirement Accounts

YearCreated

EligibilityRules

ContributionLimits

TaxTreatment

InteractionsWith Other

Plans


Proposal

Traditional IRA 1974 Have earned in-come and (a)not be in an em-ployer-spon-sored retire-ment plan or(b) have AGIunder $64,000(married filingjointly) or$44,000 (single);phaseouts beginat $54,000 and$34,000, respec-tively.Must be under701⁄2

$3,000 for thoseunder 50;$3,500 for those50 and over

Contributionsare tax-deduct-ible; distribu-tions before591⁄2 are subjectto penalty withthe exception ofdeath, dis-ability, a seriesof substantiallyequal pay-ments, or:1) health in-surancepremiums forthe unemployed2) qualifiedhigher educa-tion expenses3) qualifiedfirst-time homebuyer expenses($10k max)

Contributionlimit applies tosum of contrib-utions to allIRAs.

No new contri-butions after2004, but ac-counts couldremain in exist-ence. Balanceconverted to anRSA would betaxed. Taxes onconversionbefore 2004could be spreadover four years.

NondeductibleIRA

1974 Universal $3,000 for thoseunder 50;$3,500 for those50 and over

tax-free buildup Contributionlimit applies tosum of contri-butions to allIRAs.

No new contri-butions after2004, but ac-counts couldremain in exist-ence.

Roth IRA 1997 AGI under$160,000 (mar-ried filing joint-ly) or $110,000(single);phaseouts beginat $150,000 and$95,000, respec-tively.

$3,000 for thoseunder 50;$3,500 for those50 and over

Contributionsare taxed; dis-tributions aretax-free afterthe owner hasheld the ac-count for 5years andreaches 591⁄2,dies or becomesdisabled, oruses the fundsfor qualifiedfirst-time homebuyer expenses

Contributionlimit applies tosum of contri-butions to allIRAs.

Would becomean RSA as of2004.

Retirementsavings account

Proposed Anyone withearned income

$7,500 Contributionsare taxed; dis-tributions aretax-deductibleafter a personreaches 58

Sources: CCH Editorial Staff, 2003 U.S. Master Tax Guide (Chicago: CCH Inc., 2002); U.S. Department of Treasury, “GeneralExplanations of the Administration’s Fiscal Year 2004 Revenue Proposals,” February 2003.



(C) T

ax Analysts 2003. A




nor any requirement to withdraw the funds while theaccount holder is alive.2 Trustee accounts could be es-tablished in a minor ’s name and would become theproperty of the named individual on turning 18. In-dividuals could transfer their accounts to other people,subject to gift tax rules.

The plans differ in several ways. LSAs are meant toprovide tax-free saving. Anyone could have an ac-count, and withdrawals at any time for any purposewould be tax-free. RSAs are meant to provide ac-cumulations that will be available in retirement — al-though as noted above, there is no requirement thatbalances actually be withdrawn during the lifetime ofthe account holder. For a single person, RSA contri-butions may not exceed the account holder ’s taxablecompensation. For a married couple, total RSA contri-butions could not exceed their combined compensa-tion. Withdrawals after age 58 (regardless of retirementstatus) or in the event of death or disability would betax-free. Other withdrawals would face income tax andpenalty on the difference between the withdrawalamount and the amount contributed to the account(that is, the basis).

To encourage consolidation of existing nonretire-ment saving accounts into LSAs, individuals could rollover balances from Archer Medical Savings Accounts(MSAs), Coverdell Education Saving Accounts (ESAs),and section 529 college saving plans before January 1,2004. Table 1 lists the features of these plans. Since MSAcontributions are tax-deductible, MSA rollovers wouldbe subject to income tax. Contributions to 529 plansand ESAs are made with after-tax dollars, so no taxwould be owed on rollover.

MSAs, ESAs, and 529 plans would continue to existand could accept new contributions under thepresident’s proposal. In fact, the administration’sbudget proposes an increase in MSA contributionlimits.3

To encourage consolidation of individual retirementaccounts, Roth IRAs would be renamed RSAs, andtraditional or nondeductible IRAs could be rolled overinto RSAs. Table 2 lists the features of these plans. Thecurrent-law income limits on conversions would beeliminated. The conversion amount, less any non-

deductible contributions (basis), would be subject totax. Conversions could be made at any time, but tax-payers who converted IRAs into RSAs in 2003 wouldbe permitted to spread the income (and tax) at-tributable to the conversion over four years. Taxpayerswould also be permitted to roll over distributions frompensions directly into RSAs, paying tax on the balanceless after-tax contributions (basis) in the account.

Traditional and nondeductible IRAs would continueto exist under the proposal but could not in generalaccept any new contributions. Traditional IRAs, how-ever, could accept rollover distributions from employerplans.

The proposal would also create an Employer Retire-ment Saving Account (ERSA) to replace seven differenttypes of defined contribution plans (see Table 3).ERSAs would generally follow current rules governing401(k) plans, but the nondiscrimination rules would bechanged.

Employees could contribute up to $12,000 annuallyto an ERSA ($14,000 if age 50 or over), increasing to$15,000 ($20,000) by 2006.4 ERSAs could exist as eitherpretax or after-tax plans. A pretax ERSA would excludethe employee contribution from income tax but wouldfully tax distributions. Employee contributions to anafter-tax ERSA — called a Roth ERSA — would not bedeductible and distributions would not be taxed.5 Em-ployers could simply rename their existing plansERSAs, maintaining the pretax or after-tax charac-teristic of contributions from the original plan, subjectto the new rules. The preexisting plan could also bemaintained, but could not accept new contributionsafter 2004.

Under the proposed reform, preretirement lump-sum distributions could be rolled over into a rolloverIRA or into an RSA. Rollovers from a pretax plan intoan RSA would be taxable on conversion. Preretirementdistributions not taken as an annuity would be subjectto tax and penalty once total distributions exceed basis.The 20 percent withholding tax on pre-retirement dis-tributions that are not rolled over would continue toapply.

Lower-income contributors to any of the new ac-counts could qualify for a current-law saver ’s tax creditof up to 50 percent of contributions up to $2,000.6 (Trea-sury 2003a) As under current law, a taxpayer may notclaim a saver ’s credit if he or she made a withdrawalin the current or prior two tax years.

2The distribution rules after the account passes to an heirwould apparently mimic current rules for Roth IRAs. In gen-eral, amounts in a Roth IRA must be distributed within fiveyears after the owner ’s death, unless the owner had nameda designated beneficiary. If the owner had named a desig-nated beneficiary, the Roth IRA may be distributed over thebeneficiary’s lifetime. More generous rules apply if the solebeneficiary is the owner ’s spouse (e.g., a spouse could delaydistribution of the Roth IRA until the owner would havereached 701⁄2). See reg. section 1.408A-6, Q&A 14. Amountsdistributed from the Roth IRA due to the death of the ownerare not generally subject to income tax.

3MSA contributions are currently limited to 75 percent ofthe health insurance plan deductible for a married couplefiling a joint return (up to $3,788 in 2003) and 65 percent ofthe deductible for a single person (up to $2,178). Thepresident’s budget would increase the MSA contributionlimits to 100 percent of the deductible, up to $5,050 for acouple and $2,500 for an individual.

4The maximum combined employer and employee con-tribution to all defined contribution plans of the employerwould be limited, as under current law, to the lesser of$40,000 or 100 percent of the employee’s compensation.

5As under current law, employer contributions to eithertype of ERSA would be deductible for firms, excluded fromincome taxes for individuals, and exempt from payroll tax.

6The maximum credit is available for joint returns withincomes under $30,000 and single returns with incomesunder $15,000. The credit rate phases down to 20 percent andthen 10 percent before being eliminated $50,000 (joint) and$25,000 (single). The credit is not refundable.




(C) T

ax Analysts 2003. A




Table 3Defined Contribution Plans

YearCreated

EligibilityRules

ContributionLimits

TaxTreatment

InteractionsWith Other Plans


Proposal

401(k) 1978 Private-sectoremployees

Lesser of $12,000or AGI for thoseunder 50;$14,000 for those50 and over

Contributionsare tax-deduct-ible; distribu-tions are taxed.Distributionsallowed upon ter-mination of em-ployment orplan, in case offinancialhardship, orwhen the par-ticipant reaches591⁄2.

Limits IRA de-ductibility.

Would becomean ESRA as of2004 (as wouldall plans listedbelow).

403(b) 1958 Employees of cer-tain tax-exemptand public educa-tion institutions


Contributionsare tax-deduct-ible, distribu-tions are taxed.

Contributionlimits applies toall plans.Limits IRA de-ductibility.

No new contri-butions after2004. Accountscould remain inexistence.

Govern-mental 457Plan

1978 Non-school stateand local govern-ment employees


Contributionsare tax-deduct-ible; distribu-tions are taxed.Distributionsallowed upon ter-mination of em-ployment, incase of financialhardship, orwhen the par-ticipant reaches701⁄2.

Not necessary tocoordinate maxi-mum deferralwith contri-butions to otherretirement plans.Limits IRA de-ductibility.


SIMPLE IRA 1996 Small businessemployees

$8,000 plus em-ployer contri-bution (man-datory matchingcontribution notexceeding 3% ofcompensation ornonelective con-tribution of 2%of the first$200,000 of com-pensation); 50and over, $1,000unmatched catch-up contributionallowed

Contributionsare tax-deduct-ible; distributionrules the same astraditional IRAs

Employer maynot maintain anyother retirementplan.


SIMPLE401(k)

1996 Small businessemployees

$8,000 plus em-ployer contri-bution (man-datory matchingcontribution notexceeding 3% ofcompensation ornonelective con-tribution of 2%of the first$200,000 of com-pensation); 50and over, $1,000unmatched catch-up contributionallowed

Generally sameas 401(k) withadditional limitson the employerdeduction

Employer maynot maintain anyother retirementplan.


(Table 3 continued on next page.)



(C) T

ax Analysts 2003. A




III. The Potential for Simplification

A. LSAsUnder current law, individuals may establish a

variety of tax-preferred saving accounts. Among theaccounts that focus on nonretirement saving, the rulesdiffer for qualified uses of the funds, the taxation ofcontributions and distributions, age- and income-based eligibility rules for contributions, annual con-tribution limits, and distribution requirements (Table1, p. 1425).

On paper, LSAs would represent a historic shift inthat they give preferential treatment (under the income

tax) for substantial saving without regard to incomeand for any purpose and time frame.7 In practice,though, LSAs may represent less of a liberalization inthe allowable tax-exempt uses of saving than appearsat first since penalty-free IRA withdrawals are alreadyallowed for many purposes and withdrawals of con-

YearCreated

EligibilityRules

ContributionLimits

TaxTreatment

InteractionsWith Other Plans


Proposal

Salary reduc-tionsimplifiedemployeepension(SAR-SEP)

1986 Small businessemployees.After 1996, noemployer couldestablish a SAR-SEP.

Lesser of $12,000or AGI for thoseunder 50;$14,000 for those50 and over.

Contributionsare tax-deduct-ible, distribu-tions are taxed.


Thrift plans 1954 Nonqualifiedplans thatprimarily benefithighly compen-sated employeeswho have con-tributed the max-imum to a quali-fied plan; specialrules apply torank-and-file em-ployees who par-ticipate becausethe employerdoes not offer aqualified plan

None Employerdeducts contri-butions to afunded plan; em-ployee paystaxes on em-ployer contrib-ution if substan-tially vested;investment in-come tax-free, ex-cept for incomeattributable toemployee con-tributions aboveemployer contri-butions


Employerretirementsavings ac-counts

proposed in2003

Any employercan sponsor


Contributionsare tax-deduct-ible; distribu-tions are taxed.Distributionsallowed upon ter-mination of em-ployment orplan, in case offinancialhardship, orwhen the par-ticipant reaches591⁄2. After-taxcontributionsand tax-free dis-tributions alsoallowed, subjectto the maximum$40,000 (or 100%of salary) com-bined employerand employeecontributionlimits to definedcontributionplans.

Sources: CCH Editorial Staff, 2003 U.S. Master Tax Guide (Chicago: CCH Inc., 2002); U.S. Department of Treasury, “GeneralExplanations of the Administration’s Fiscal Year 2004 Revenue Proposals,” February 2003.

7The only precedent of which we are aware relates toAll-Savers Certificates, which allowed a lifetime exclusion ofup to $1,000 of tax-free interest income from special one-yearcertificates of deposit purchased between October 1981 andDecember 1982.



(C) T

ax Analysts 2003. A




tributions (but not earnings) from a Roth IRA are tax-and penalty-free for any purpose (see Table 2, p. 1426).Nevertheless, the reduced “hassle factor” would haveto be considered significant.

As a result, if they were to replace existing tax-favorednonretirement saving accounts, LSAs would represent asignificant simplification of present law for many tax-payers. But rather than replacing existing accounts, theadministration’s proposal simply adds LSAs to theroster of tax-preferred alternatives. Individuals whomeet the eligibility requirements may contribute to allexisting education and medical saving accounts and toan LSA. Indeed, the administration’s proposed in-crease in contribution limits to MSAs undermines theincentive to consolidate accounts or simplify tax plan-ning for those who have access to MSAs. Those whoare saving specifically to help pay future medical ex-penses would likely want to retain their MSA and con-tinue making contributions, since the contributions aredeductible and the withdrawals are tax-free if used forqualified purposes. In addition, those who save sub-stantial amounts would be likely to continue to use thetargeted savings accounts, especially 529 plans, be-cause they allow the sheltering of large amounts ofadditional savings from tax. Indeed, those with a 529plan could shelter even greater amounts of income byconverting existing 529 balances, which can total asmuch as $267,000 per covered individual, into an LSAin 2003 and creating a new 529 plan in subsequentyears.

Lower-income people will also face some new com-plexity. As noted, withdrawals from any of the newaccounts would disqualify them for the saver ’s taxcredit. Thus, a qualifying low-income person who iscontributing to a pension plan at work or through anRSA might avoid setting up an LSA, because with-drawal of even a dollar from an LSA would disqualifyhim from the saver ’s credit for three years (the currentand two subsequent tax years). This could cost as muchas $3,000 in lost tax credits, assuming the taxpayerearns less than $30,000 per year, has at least $1,000 oftax liability, and contributes at least $2,000 per year toan RSA or ERSA.8

LSAs would, however, simplify rules for upper-middle-income families who do not qualify for the saver ’scredit and save less than $7,500 per year outside ofretirement plans. With minimal paperwork and few

restrictions, such families could establish and maintainan LSA for the same purposes as the medical or educa-tion plans, as well as for other needs. The flexibilityand ability to tap funds in an emergency without penal-ty should make LSAs more attractive than the alterna-tives for most taxpayers.9

B. RSAsExisting individual retirement accounts vary regard-

ing the taxation of contributions and withdrawals,qualified uses of funds, minimum distribution rules,age-, income-, and pension-based eligibility rules, andother factors (Table 2, p. 1426). Although taxpayerswould not be required to convert traditional and non-deductible IRAs to RSAs, the existing IRAs could nolonger accept contributions, other than pension rol-lovers. This would give individuals incentives to con-solidate their accounts by rolling existing accounts intoan RSA, and it would remove complexity regardinghow to allocate new contributions. In addition, RSAshave no income, age, or minimum withdrawal require-ments. Thus, taxpayers would not have to address min-imum distribution rules, nor would they have to beconcerned with income-eligibility cutoffs.

The elimination of income limits and phaseouts ofeligibility would simplify tax compliance for higher-income taxpayers, especially those near the currentlimits. Those taxpayers could make contributions toRSAs during the year without the risk that they wouldbe ineligible at tax time. Income phaseouts also createimplicit taxes, which are undesirable, although thereare good policy reasons to limit savings incentives tomoderate-income people, as discussed below. Theremoval of income limits, whatever its undesirabilityon policy grounds, could facilitate further simplifica-tion. Without an income limit, for example, therewould be no need for the provisions that allow tax-payers to undo contributions before the due date fortheir tax return without penalty. Similarly, without in-come limits, the provisions that allow contributionsafter the end of the calendar year to IRAs would be nolonger necessary, which would eliminate ambiguityabout whether contributions on April 1 should be at-tributed to the prior or the current tax year and endinga source of complex tax planning. This type of potentialsimplification, however, was not included in thebudget proposal.

One potential ly important tax planning con-sideration is that unlike current Roth IRAs, whichallow tax- and penalty-free withdrawals for a varietyof nonretirement purposes, RSAs would permit tax-free withdrawals only after age 58 or in the event ofdeath or disability. To maximize flexibility, taxpayers

8LSAs would also create simple opportunities for taxavoidance in combination with the saver ’s credit. For ex-ample, a taxpayer could borrow $2,000 to deposit into an LSAon April 15 for the preceding tax year, claim the saver’s creditof up to $1,000 on the LSA contribution for that precedingtax year, and then withdraw the entire amount the followingday to repay the loan. The taxpayer nets $1,000 in tax savingswithout altering his or her saving at all. The antichurningrules would disqualify the taxpayer for the saver’s credit forthe current tax year (which includes the April 15 in whichthe deposit was made to the LSA) and the following two taxyears, but the process could be repeated again after that. Asimilar scheme would work under current law using RothIRAs, so this is not a new tax shelter. But the total lack ofrestrictions on LSAs might make such schemes easier tomanage.

9Even for these taxpayers, though, complications may stillarise due to interactions between LSAs and college financialaid formulas; double-dipping rules regarding LSA with-drawals used for college education and other education sub-sidies — such as the HOPE credit, Lifetime Learning Credits,and the education deduction; and asset tests in means-testedprograms, such as Medicaid, Food Stamps, and Supplemen-tal Security Income. These complications also arise for exist-ing savings accounts.



(C) T

ax Analysts 2003. A




could put all saving into LSAs first, and only makecontributions to RSAs after their LSA contributionlimits were met. But some theories suggest that peopleuse restricted retirement accounts as a way of guaran-teeing that saving actually is available for retirement,rather than spent prematurely. Under that behavioralmodel, the flexibility of LSAs may actually be a draw-back.

C. ERSAsAt fi rst glance, ERSAs look l ike a dramatic

simplification, since they unify the disparate rulescovering a wide variety of defined contribution plans(see Table 3, p. 1428). But closer inspection suggests

that unifying the rules only represents simplificationalong certain dimensions. It does not represent sub-stantial simplification for any employer who has al-ready chosen a specific type of plan or any worker whoparticipates in only one plan. Unification of disparaterules across types of defined contribution plans is logi-cal and reasonable, but it does not substantiallysimplify many pension planning problems.10

Table 4Behavioral Responses

BehavioralResponse Comments Revenue

PrivateSaving

NationalSaving Distribution

(1) Transfer exist-ing taxable assetsinto LSAs

Taxpayers shouldtransfer as muchas possible toLSAs.Transfers may beslowed becausethey must be madein cash.

$200B-$300B over10 years.0.2%-0.3% of GDPin year 10.Will continue togrow over time.

No direct effect(But higher wealthdue to transferscould reduce newsaving — see row3.)

Negative Regressive.Increasingly regres-sive over time.

(2) Place futuresaving that wouldhave been doneanyway in an LSA

After transferringexisting assets, tax-payers will placetheir usual savingin LSAs.

Growing revenueloss over time.

No direct effect(but see row 3).

Negative Regressive.Increasingly regres-sive over time.

(3) Change level ofnonretirementsaving

Higher rate ofreturn may raisesaving.Higher wealthfrom rows 1 and 2would reduce newsaving.

Uncertain. Uncertain. Uncertain. Uncertain.

(4) Rollover tradi-tional and non-deductible IRAsinto RSAs

Shifts revenuefrom the long termto the short termPresent value ofrevenue falls.

No effect Higher in theshort term.Lower in the longterm.Present value falls.

Only benefitshouseholds withincome above$100,000.

(5) Place in RSA fu-ture saving thatwould have beenvia a traditionalIRA

Same as above. No effect Same as above None.

(6) Change in levelof retirementsaving contri-butions

Could rise forthose constrainedby current limits.Could fall becauseof elimination ofup-front deduction.

Uncertain Uncertain Uncertain Likely regressive(see text)

(7) Change in pen-sion coverage

May rise due tosimplification ofpensions.May fall due to ex-pansion of non-employer basedsaving options.

Uncertain Uncertain Uncertain Uncertain

(8) Change in pen-sion contributions,given coverage

Will fall, given theweakening of non-discriminationrules.

Higher in theshort term.Lower in the longterm.Present value rises.

Falls, since contrib-utions bymoderate incomeworkers tend toraise saving.

Falls in the longterm

Reduces benefitsfor low andmoderate incomehouseholds.

10It would simplify efforts to consolidate pensions forpeople who have several retirement funds from several jobs,but that can already be accomplished by rolling the fundsover into an Individual Retirement Account.



(C) T

ax Analysts 2003. A




The real potential for simplification in ERSAs comesthrough proposed changes in the nondiscriminationrules. Nondiscrimination rules are intended to ensurethat tax-preferred pensions are distributed equitablyacross income groups. The rules typically serve torestrict the amount of contributions that highly paidworkers can make relative to the amount that rank-and-file workers make (or that the employer makes ontheir behalf). The proposed rules for ERSAs wouldsimplify, and reduce the stringency of, those rules inseveral ways (see Appendix A). As discussed in sub-sequent sections, that simplification may come at thecost of reduced pension coverage.

Another issue is that an after-tax ERSA would in-volve an additional layer of complexity, just as Roth401(k) plans will. Amounts attributable to employermatching contributions, which were not taxed whenmade, would presumably be taxable to the individualupon withdrawal. After-tax contributions made by theemployee (which are not taxable upon withdrawal inretirement) will therefore have to be tracked separatelyfrom employer matching contributions (which is alsonecessary in similar situations under current law).

IV. Key Policy Changes

Because the proposal involves a substantial numberof changes, it is helpful — as a prelude to consideringthe revenue, distributional, and savings effects — tobreak down the proposal into the key changes in thestructure of saving rules and the principal channelsthrough which households and firms would be likelyto respond. Taken together, the proposals represent:

• a significant expansion in the types of savingthat could be undertaken with tax-preferredstatus;

• a significant increase in the amount of tax-preferred saving that could be undertaken out-side of retirement accounts;

• a significant expansion in annual contributionlimits for retirement saving;

• the removal of all eligibility rules related to age,pension coverage, or maximum income on con-tributions to individual retirement accounts;

• the removal of all minimum distribution ruleson those accounts while the account owner isalive;

• the removal of all income eligibility limits onconversions of traditional and nondeductibleIRAs into back-loaded saving vehicles;

• the termination of contributions to traditionalIRAs with tax-deductible contributions; and

• simplification and weakening of the nondis-crimination rules.

These policy changes would drive a series of be-havioral responses. Our analysis of the revenue, dis-tributional, and saving effects of the proposal in thefollowing sections will highlight the behavioralresponses summarized in Table 4 (see p. 1431).

V. Revenue Effects

The Treasury Department estimates that LSAs andRSAs would raise $15 billion in revenue over the first5 years, and lose $13 billion over the subsequent 5years, for a net gain of $2 billion over the 10-yearbudget period (Table 5, above). Treasury also estimatesthat ERSAs would reduce revenue by about $3 billionover the decade. All told, the official estimates implythe saving proposal would have negligible effects onthe federal budget over the next decade.

Our revenue estimates differ significantly. Our es-timates do not include all eight channels for revenueloss shown in Table 4, and are likely to understate therevenue loss attributable to the proposal. Thus, it is strik-ing that our estimates suggest large revenue losses withinthe 10-year budget window. In addition, the revenuecosts will be even larger outside of the 10-year window(and especially outside of the 5-year budget window thatthe administration uses). Our revenue estimates focus onthree key elements of the plan: the shifting of existingtaxable assets and future taxable saving that would havebeen done anyway into LSAs, the termination of on-goingcontributions to deductible IRAs, and rollovers of tradi-tional and nondeductible IRAs into RSAs.

A. Shifting Funds Into LSAsContributions to LSAs may not exceed $7,500 per

year and must be made in cash — taxpayers may nottransfer existing stocks or bonds into the accounts. Noother restrictions apply. As a result, most taxpayerswho could shift existing taxable assets to LSAs arelikely to do so.11

Table 5Administration Estimate of Revenue Effects ($ billions), 2004-13

Fiscal Year

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Total 2004-2013

RSA/LSA 10.6 4.8 1.9 -0.6 -1.8 -1.9 -2.4 -2.7 -2.9 -2.9 2.0

ERSA -0.2 -0.3 -0.3 -0.3 -0.3 -0.3 -0.3 -0.3 -0.4 -0.4 -3.0

Source: United States Treasury Department (2003b).

11Taxpayers with very small balances, especially those whodo not expect to be taxable any time soon, might rationallydecide that it is not worth the trouble to establish a new ac-count. Another factor that could slow the transition is that theincentive to shift depends on interest rates, which are currentlyquite low. It is also possible that such accounts will have min-imum balance requirements or higher fees if the accounts aresubject to more reporting requirements or are restricted in someway compared with regular savings accounts.



(C) T

ax Analysts 2003. A




To estimate the revenue costs of transfers of alreadyexisting assets and future saving that would have beendone anyway into LSAs over the next 10 years, weconsider how much a household would be able to con-tribute to LSAs over time and then estimate the cor-responding level of capital income that would be ex-empt from tax. Because the cumulative contributionlimit rises over time, so too does the implied capitalincome exclusion. For example, a single person couldcontribute $7,500 in the first year. With a 6 percentnominal return, a first-year transfer of $7,500 in taxableassets would make $450 (= .06*7,500) of capital incomeexempt from tax in year 1. Ignoring compounding, inthe second year, $900 would be exempt from capitalincome tax, and so on.

As shown in Table 6 above, an individual who trans-ferred $7,500 of existing assets into LSAs in each yearstarting in 2003 could shelter $450 in capital income in2004, $2,659 by 2008, and $6,567 by 2013 (assuming theassets earned a 6 percent nominal return). For marriedcouples, the effective exclusion would be twice aslarge. Est imates using the Tax Policy CenterMicrosimulation model imply that a capital incomeexclusion that corresponds to these amounts over timewould reduce revenue by $330 billion over the decade.The costs would rise over time, both in absolute terms

and as a share of GDP. By 2013, the annual revenueloss would be $59 billion or more than 0.3 percent ofGDP.12

Table 6Lifetime Savings Accounts: Revenue Cost ($ billions), 2004-131

Fiscal Year2

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Total2004-2013

Baseline: current law for EGTRRA

6% ROI3 10.2 14.9 19.3 23.9 28.7 33.8 39.1 46.9 53.6 59.9 330.2

6% ROI, Dividend Plan4 9.8 14.2 18.2 22.4 26.6 31.0 35.5 42.3 48.2 53.7 301.9

6% ROI, $3,750 MaximumContribution 6.0 8.9 11.7 14.7 17.9 21.3 24.9 30.1 34.7 39.1 209.0

3% ROI5 5.9 8.7 11.3 14.1 17.0 20.0 23.2 27.8 31.7 35.5 195.3

Baseline: EGTRRA extended through 2013

6% ROI 10.2 14.9 19.3 23.9 28.7 33.8 39.1 44.8 50.6 56.8 322.0

6% ROI, Dividend Plan 9.8 14.2 18.2 22.4 26.6 31.0 35.5 40.4 45.6 51.0 294.6

3% ROI 5.9 8.7 11.3 14.1 17.0 20.0 23.2 26.6 30.1 33.7 190.6

15 year accumulation in 20046 37.2 — — — — — — — — — —

20 year accumulation in 20047 45.6 — — — — — — — — — —

25 year accumulation in 20048 53.3 — — — — — — — — — —Equivalent Capital IncomeExclusion9 $450 $938 $1,467 $2,040 $2,659 $3,328 $4,049 $4,827 $5,665 $6,567 —

Source: Urban-Brookings Tax Policy Center Microsimulation Model.1Lifetime Savings Accounts are assumed to have a $7,500 per account maximum after-tax contribution.2Fiscal-year estimates assume a 75-25 split.3The annual nominal return on investments inside the LSA is assumed to be 6 percent.4The president’s proposal to exempt some dividends from personal taxation is included in the baseline in this calculation5The annual nominal return on investments inside the LSA is assumed to be 3 percent.6Assumes that LSAs have been in effect for the past 15 years with a real contribution of limit of $7,500 in 2003 dollars anda 3.5 percent real return.7Assumes that LSAs have been in effect for the past 20 years.8Assumes LSAs have been in effect for the past 25 years.9Equivalent capital income exclusions are in current dollars, and they correspond to the previous calendar year in whichthe fiscal year began.GDP for FY2004 is estimated to be $11,309 billion CBO (2003).

12These calculations contain at least five sources of am-biguity. First, it is unclear what portion of the funds wouldcome from existing assets and what portion from future tax-able saving that would have been done anyway. Second,among the share that comes from transfers of already existingassets, it is unclear how much capital gains tax revenuewould be generated from the sales of existing assets so thatthey could be placed in the LSA. Because taxpayers haveinterest-bearing assets and can choose which assets with cap-ital gains or losses to sell, we expect that the revenue gainfrom increased realizations would be small. Third, it is un-clear how much of whatever increase in capital gain revenueoccurs would have been obtained anyway from the sale ofthe asset in the future. Including the last two factors wouldreduce the revenue loss within the 10-year budget window,but increase the loss in years beyond 2013. Fourth, the effectsof the minimum distribution rules for inherited accounts arenot incorporated into the analysis. Those rules limit to somedegree of accumulation of funds in the new tax-shelteredaccounts over time, an effect not incorporated into the reve-nue analysis. Fifth, some of the contributions made to LSAsmay displace other tax-preferred saving. In that case, the netrevenue loss would be smaller than shown here.




(C) T

ax Analysts 2003. A




Box 1. Back-Loaded Saving Plans as Budget Gimmicks:Or, Why Rollovers Cost the Government Money

Traditional (or front-loaded) IRAs allow a deductionfor contributions and tax withdrawals. This reducescurrent government revenue but raises future revenue.In contrast, in back-loaded plans (like Roth IRAs undercurrent law and RSAs under the administration’s pro-posal) contributions are not deductible but with-drawals are not taxable. The creation of Roth IRAs in1997 was widely derided as a budget gimmick for atleast three reasons.a First, switching on-going contri-butions from traditional IRAs to Roth IRAs raises cur-rent tax revenues but reduces future revenues. Second,Congress gave taxpayers incentives to convert tradition-al IRAs into Roth IRAs and pay taxes on the traditionalIRA balance. These conversions also raised revenues inthe short term but only at the expense of reducedrevenues in the long term. Thus, under then-currentbudget rules, the revenue “gained” in the short termfrom both of these changes could be used to financeother spending programs or tax cuts, even though theshort-term revenue gain did not represent an improve-ment in the government’s overall revenue stream.

Third, and perhaps less obvious, conversions ofasset balances and on-going contributions not onlyshift revenues to the present by mortgaging futurerevenues, they do so at very unfavorable terms for thegovernment. In particular, the conversions reduce thepresent value of future revenues by more than theyincrease the present value of short-term revenues.This is true because a conversion increases theamount of tax sheltering that can occur. For example,a balance of $x in a traditional IRA represents t*x inpre-paid tax payments, where t is the tax rate, and(1-t)*x in after-tax wealth. In contrast, all of the fundsin a tax-deferred, or back-loaded, account like a RothIRA or RSA are tax-sheltered. Thus, converting anIRA balance to an RSA, and paying the tax on therollover with taxable assets, increases the amount ofsheltering, which in turn reduces government reve-nue in the long term by even more than it raisesrevenue in the short term.

An example helps illustrate this point. SupposeSally has $1 million in a traditional IRA and that thebalance will double over the next 10 years due tocompound interest. Suppose also that Sally faces aconstant 35 percent tax rate and has $350,000 in ataxable interest-bearing account. Sally could keepher traditional IRA or roll it over into a back-loadedaccount — i.e., an RSA.

If she keeps the traditional IRA and cashes it outafter 10 years, she will withdraw $2 million and keep $1.3million after paying $700,000 (= 35 percent of $2 million)

in taxes. She would also have a total of $552,000 inher taxable interest-bearing account (assuming forconsistency that the pre-tax rate of return for thataccount is the same as for the IRA and the taxes dueon the interest each year are paid out of the interestearne d).b Her after-tax wealth would be $1,852,000.

Suppose instead that she immediately rolled the$1 million in the traditional IRA into an RSA and paidtaxes of $350,000 (= 35 percent of the IRA balance)on the rollover. The tax payment would exhaust hertaxable interest-bearing account. But her RSAbalance would be $2 million after 10 years, whichcould be withdrawn tax-free. Thus, with the strokeof a pen, the rollover raises Sally’s after-tax wealthby $148,000 (= $2 million - $1,852,000) after 10 years.

The effect of the rollover on government revenuesis the opposite — if Sally pays $148,000 less, the gov-ernment receives $148,000 less.c In present value terms,the $148,000 is worth $74,000 currently. This represents21 percent of the actual tax paid at the time of therollover. Put differently, every dollar of revenue paidnow through this scheme costs the government 21 centsin present value. That is, it reduces the present valueof future revenues by $1.21. This cost would rise if theholding period were longer, interest rates were higher,the time-constant tax rate was higher, or if tax rateswere expected to rise in the future.

As noted above, the driving factor behind thesecalculations is that the rollover lets Sally shelter morefunds. To see this, note that the revenue conse-quences of rolling over the after-tax value of the tradi-tional IRA are the same as not rolling over the wholeIRA balance. If the after-tax value were rolled over,Sally would have to pay taxes out of the IRA balanceitself, and she could only rollover $650,000. Thiswould grow to $1.3 million in the RSA after 10 years,leaving her with a total of $1,852,000 (=$1.3 million+ $552,000) after tax. This would give her, and thegovernment, the same outcome as not rolling overthe IRA balance.

aFor example, Steuerle (1997) notes that “Almosteveryone recognizes that they [Roth IRAs] were enactedpartly because of perverse budget accounting that does nottake into account long-term effects of the deficit.” Halperin(1998) writes that “ . . . the main purpose of the Roth alter-native was to hide the budget cost.” See Gravelle (1993) forcalculations of the differences in the timing and level ofrevenues between front- and back-loaded plans.

bThe pre-tax rate of return required to double assetbalances over 10 years is 7.2 percent. Thus, the post-taxreturn on the taxable interest-bearing account would be 4.7percent = ((1-.35)*7.2). The result reported in the text oc-curs because 350,000*(1.04710)) = 552,000.

cThe net decline in government revenues has severalcomponents. First, the government receives $350,000 im-mediately upon rollover, but forgoes $700,000 in revenuesin year 10 that would have been forthcoming had Sally keptthe traditional IRA. These two items net to zero in presentvalue terms (By assumption, the rate of return is such thatasset balances will double over the decade. As a result, thepresent value of the $700,000 loss in year 10 is equal inmagnitude and opposite in sign of the present value of the$350,000 revenue gain in year zero.) Second, because Sallypays taxes on the rollover with taxable assets, the govern-ment forgoes future tax revenue from those balances. Thisreduces revenues by $148,000 (the difference between the$700,000 balance the assets would have generated at thepre-tax rate of return and the $552,000 they generate at thepost-tax return).



(C) T

ax Analysts 2003. A




In comparison, the administration’s dividend pro-posal is slated to reduce revenue by $385 billion over thedecade and $53 billion in 2013. Notably, our estimatesthroughout this article use current law as the baseline —that is, they assume that the dividend proposal is notenacted. If the dividend proposal were enacted first, theincremental costs of the saving proposal would declineto $301 billion over the decade (Table 6, p. 1435).

As noted above, LSA deposits would have to be madein cash, which might reduce the speed at which existingassets were transferred. For example, if each individualwere only able to transfer a maximum of $3,750 per year,the implied capital income exclusion for a single personwould be $225 in 2003, rising to $3,283 in 2013. Even so,transfers of existing assets would reduce revenues by$209 billion over the decade, and by 0.2 percent of GDPin 2013. Likewise, if the nominal rate of return were only3 percent, transfers of existing assets would reducerevenues by about $195 billion over the decade.

The revenue losses would grow over time. For ex-ample, using 2003 tax law and income levels, and as-suming the accounts generate real returns of 3.5 per-cent per year (which coupled with our baselineinflation assumption of 2.5 percent generates nominalreturns of 6 percent), after 15 years the annual revenueloss from transfer of existing assets would be 0.34 per-cent of GDP and after 25 years the annual loss wouldbe 0.5 percent of GDP. Note that the assumed realreturn on the assets in the accounts, 3.5 percent, islower than the real return assumed in other contexts.For example, the president ’s Commission toStrengthen Social Security assumed a 4.6 percent realreturn (after administrative costs) in evaluating its in-dividual account proposals.13 A higher assumed returnwould increase the revenue loss.

There is a great deal of uncertainty about these reve-nue estimates. It is possible, for example, that take-upof the new tax incentive might be much slower thanwe have estimated in any of our scenarios. But Table 6shows that there is great potential for erosion of therevenue base over the long term from this provisionalone.

B. Termination of Contributions to Deductible IRAsThe proposal would end contributions to deductible

IRAs. Using the TPC simulation model, we estimatethat this could raise revenues by as much as $25 billionover the next decade (assuming that traditional IRAcontributions that would have been made would nothave been withdrawn during the next decade). Thisshort-term revenue gain corresponds to revenue losseswith equal or greater present value in the future (be-cause taxable withdrawals from the traditional IRAsno longer take place). So the $25 billion in short-termrevenue gain portends a revenue loss of at least $25billion plus interest mostly occurring when ourbudgetary situation is much worse than it is now.

C. Rollovers of IRAs Into RSAsConversions from traditional IRAs into RSAs trigger

a current tax payment but eliminate all future tax pay-ments on withdrawals from the account. As a result,they raise short-run revenues, but reduce long-term reve-nue by even more in present value. Rollovers are in es-sence a very expensive way for the government to borrowmoney.14 (See Box 1, p. 1434.) Taxpayers will choose toroll over their IRA balances only if they expect their futuretax savings to be worth more than the current tax pay-ment — that is, only if they expect the transaction willreduce the present value of their tax payments. This canoccur because conversions effectively increase theamount of saving that can be tax-sheltered and/or be-cause the account holder expects her tax rate in retire-ment to exceed her current tax rate.15

Currently, only households with income below$100,000 are allowed to convert their traditional IRAsinto Roth IRAs. The president’s proposal would allowall households to do so. As an added incentive formaking the conversion soon, the income attributableto any conversion in 2003 could be spread over fouryears, thus reducing the present value of tax pay-ments.16

The revenue effects of induced rollovers are subjectto substantial uncertainty because it is difficult topredict the extent to which households would exploitthe new rollover opportunities. One useful piece ofinformation in forecasting such behavior is that in1998, when households with income below $100,000

13President’s Commission to Strengthen Social Security,Strengthening Social Security and Creating Personal Wealth forAll Americans, December 2001, page 97.

14This calculation assumes that the risk-adjusted rate ofreturn that taxpayers can obtain is the same as the return thegovernment can obtain on its investments. Even under thisassumption, rollovers cost the government money in thelong-term because as discussed in the text they allow tax-payers to increase the amount of sheltered saving. If tax-payers can earn higher risk-adjusted returns on investmentsthan the government can, a conversion from traditional IRAsto Roth IRAs will further reduce government revenues, evenaside from the higher amount of sheltering that would occur.The reason is that in traditional IRAs, the government sharesin the accrual of wealth and so benefits from the higher returnthat the private sector is able to obtain. If the traditional IRAis converted to a Roth IRA, the government loses out on thatopportunity (Dusseault and Skinner 2000).

15Even if tax rates fall somewhat in retirement, individualsmay find it advantageous to convert because of the highereffective contribution limits in back-loaded plans. Burman,Gale, and Weiner (2003) found that 71 percent of taxpayerswould benefit from choosing a Roth IRA (analogous to anRSA) based on actual income patterns over a 13-year periodand assuming no change in tax law. A majority of taxpayersin every income and age group (under 65) would havebenefited from this choice. This calculation assumed perfectforesight about future tax rates. If taxpayers are risk averse,there is an additional benefit from locking in a certain taxrate now, rather than taking a risk on future tax rates.

16With a real discount rate of 3.5 percent and an inflationrate of 2.5 percent, the ability to spread income over fouryears reduces the present value of the tax liability by about8 percent, assuming a constant tax rate. With a progressiveincome tax, spreading income reduces the odds that a largelump sum would push the taxpayer into a higher tax bracket,which would reduce the present value of tax payments byadditional amounts.



(C) T

ax Analysts 2003. A




became eligible to convert their traditional IRAs intoRoth IRAs, it appears that about 4 percent of eligiblebalances were converted.17

Conversions generated by the new proposals arelikely to come from taxpayers with incomes over$100,000, since those with lower incomes can alreadyconvert balances.18 Thus, as described in Appendix B,we estimate rollover activity in three steps. First, weposit several determinants of rollover behavior. Roll-overs will be more likely to occur the larger is thepresent value of the tax savings as a share of requiredtax payment. Rollovers will be less likely to occur thelarger is the current tax liability on the rollover as ashare of the household’s liquid financial assets. Otherthings equal, rollovers will be less likely to occur as ahousehold’s portfolio becomes more concentrated inretirement assets. Second, using data from the 1998Survey of Consumer Finances, we choose thresholdlevels of those determinants that would be consistentwith the 4 percent conversion rate of IRA balances forhouseholds with income below $100,000 observed after

1997. Third, we apply those thresholds to the IRAbalances of households with income above $100,000.

Using this approach, we adopt as our base case therule that households would convert their traditionalIRAs into RSAs only if (a) the projected future taxsaving exceeds 20 percent of the up-front tax payment,(b) the up-front tax payment is less than 20 percent ofthe household’s liquid financial assets, and (c) thehousehold has less than half of its total financial assetsin retirement funds. Under these assumptions, we es-t imate that 2 .9 percent of IRA balances amonghouseholds with income below $100,000 would be con-verted (Appendix Table B-1, p. 1442). Thus, our basecase understates rollover behavior relative to the 1997experience. Nevertheless, our estimates imply substan-tial rollover activity would occur. Specifically, we es-timate that the proposal would induce $109 billion inIRA conversions by households with income above$100,000 (Appendix Table B-2, p. 1442). This representsabout 12 percent of traditional IRA balances amongthese households. This figure is substantially higherthan the 2.9 percent conversion rate for lower-incomehouseholds, because higher-income households aremuch more likely to generate substantial tax savingsupon rollovers and to have sufficient liquid assets topay the taxes due.

These rollovers would generate short-term revenueof $37 billion (Appendix Table B-3, p. 1442), but theywould reduce the present value of long-term revenueby $49 billion, or 132 percent of the short-term gain(Appendix Table B-4, p. 1442). We emphasize, andshow in Appendix B, that these estimates are subjectto substantial uncertainty. In particular, a reasonablerange of parameter estimates generates short-term con-version totals that vary from $49 billion to $270 billion,with short-term revenues of between $17 billion and$93 billion. A key reason for the uncertainty is that thedeterminations of rollover behavior under the admin-istration’s proposal may differ from the 1998 ex-

Table 7Comparison of Benefits From the Creation of Lifetime Savings Accounts: Distribution of Income Tax

Change by AGI QuintileAGI Quintile1 2003 2010 Mature LSA2

Share ofTax Cut

Share of CumulativeCut

Share of Tax Cutin the 10th Year

Share of CumulativeTax Cut

Share of Tax Cutin the 25th Year

Lowest Quintile 1.8 1.2 0.7 0.8 *

Second Quintile 5.0 3.9 3.0 3.1 0.5

Middle Quintile 10.3 8.4 6.8 7.1 3.2

Fourth Quintile 22.5 21.8 20.5 19.7 15.6

Next 10 Percent 21.2 19.8 18.4 17.9 15.0

Next 5 Percent 15.2 15.1 14.8 15.0 15.9

Next 4 Percent 17.8 21.6 24.5 23.5 28.8Top 1 Percent 6.2 8.2 11.4 13.0 21.0

All 100.0 100.0 100.0 100.0 100.0

Source: Urban-Brookings Tax Policy Center Microsimulation Model.* Less than .05 percent.1Filers with negative AGI are not included in this category but are included in totals.2For purposes of this calculation CY2003 is used as endpoint of 25 years of accruals into LSAs with a maximum contri-bution of $7,500 in constant 2003 dollars.

17In 1998, taxpayers filing a total of 1.4 million returnsconverted $39.3 billion from traditional to Roth IRAs.(Campbell, Parisi, and Balkovic 2000) This might overes-timate slightly the ultimate level of conversion activity, be-cause some conversions were reversed after tax returns werefiled. Households with adjusted gross income (AGI) of lessthan $100,000 in 1997 owned IRAs worth at least $1 trillionin 1998, according to the Survey of Consumer Finances (SCF).The SCF includes both IRA and Keogh assets in the samecategory. We excluded all assets in this category if an in-dividual indicated the presence of only Keogh assets, andassumed a 50-50 split if the individual indicated the presenceof both IRAs and Keoghs.

18One caveat to this observation is that currently eligiblehouseholds who want to maintain only one individual retire-ment account would likely end up converting to an RSA ifthey plan to make any new contributions.



(C) T

ax Analysts 2003. A




perience. The removal of income limits on rolloversmight generate significant advertising regarding thebenefits of rollovers that, combined with the sophisti-cated tax advice available to many higher-incomehouseholds, could imply far greater rollover activitythan our calibration exercise suggests. In all cases,however, the conversions cost the government long-term revenues that exceed the short-term gains. Theratio of the long-term losses to short-term gains rangesbetween 123 percent and 137 percent. Put differently,raising money through IRA conversions costs the gov-ernment 23 to 37 percent more than raising cashthrough conventional borrowing. Allowing for the factthat rollover tax payments could be spread over fouryears would raise government costs even more.

D. Other EffectsTable 4 lists several other behavioral responses that

would produce revenue effects but that we are unablequantify currently. Perhaps most important amongthese is that the removal of income limits on individualretirement saving would generate an increase in con-tributions to such accounts by very high-incomehouseholds. In addition, for anyone over age 58, RSAwithdrawals would be unrestricted and so could bethought of as a second LSA. Incorporating these twoeffects would increase the amount of tax sheltering thatwould occur under the proposal and increase the reve-nue loss further.

VI. Distributional Effects

As with the revenue effects, we provide distribu-tional estimates of certain aspects of the proposal.Taken as a whole, though, there is no doubt that theproposal will disproportionately benefit higher-in-come households.

The conversion of traditional IRAs to RSAs, for ex-ample, will cause a one-time increase in after-tax in-come of about $12 billion in our base case, with thetotal flowing entirely to households with income inexcess of $100,000, since only these households areaffected by the income limits under current law.

Selected distributional aspects of the conversion ofexisting taxable assets and future taxable saving thatwould have been done anyway to LSAs are shown inTable 7 (see p. 1436). As in the previous section, theLSA is modeled as an increasing capital income ex-clusion over time, for purposes of considering thetransfer of existing assets and new contributions thatwould have instead been made to taxable accounts. Thetable shows that the transfer of existing assets wouldgenerate regressive tax cuts that would become in-creasingly regressive over time. In 2003, the bottom 40percent of filers would obtain about 6.8 percent of thebenefits. By 2013, they would obtain only 5.1 percentof the cumulative benefits. After 25 years, they wouldhave obtained only 3.9 percent of the total tax cut inplace at that time. Their share of the new tax cut pro-vided in each year is even smaller, just 3.7 percent in2013 and 0.5 percent in the 25th year. Their share of thebenefits would decline because they have fewer assetsto shift to LSAs. In contrast, households in the top 20percent of the income distribution would receive 60

percent of the benefits in 2003, rising to 70 percent ofthe new benefits provided in 2013, and 80 percent ofthe new benefits provided in the 25th year. All of thesetrends would continue after the 25th year.

Note that the share of the tax cut provided to thetop 1 percent is more modest than one may have ex-pected, largely because asset incomes for such filerstend to be so large that the proposal would initiallyexempt a relatively small share of their capital incomefrom taxation. It is worth emphasizing that the resultsabove account for only the transfer of existing assetsand the transfer of new contributions from taxable ac-counts into LSAs. Other aspects of the proposal aredifficult to quantify, but also suggest the tax cut wouldbe of particular benefit to higher-income households.

Some households will increase their saving inresponse to LSAs and thus benefit from the tax treat-ment of their new saving. Again, however, high-income households will likely benefit to a dispropor-tionate degree, since the LSA gives high-incomehouseholds a bigger incentive to save by raising theirafter-tax return by more than lower-incomehouseholds. For example, households facing a 30 per-cent tax rate would have their after-tax return rise byabout 43 percent (from (1- .30)*r to r), whereashouseholds in the 10 percent bracket would only seean 11 percent rise in their rate of return (from (1-.90)*rto r). Also middle-income households, and even manywith moderately high incomes, are much less likely tobe constrained by the contribution limits on existingtax-preferred savings vehicles than are those with veryhigh incomes.

Even more dramatically, RSAs are virtually ex-clusively a benefit to those with very high incomes.Even before the employer contribution limits were in-creased in 2001, less than 5 percent of households con-tributed the maximum to an IRA and less than 5 per-cent contributed the maximum to their employer ’sretirement plan.19 RSAs only provide an additional taxbenefit — and an additional incentive to save — to thetiny minority of households who are constrained byexisting tax-preferred retirement saving incentives. Be-cause the existing contribution limits are set in absoluteterms, not as a share of income, it is plausible that onlythe very-highest-income households would be con-strained by current limits.

In addition, it is likely that the changes in pensioncoverage discussed in further detail in the next sectionand in Appendix A also would reduce benefits for mid-dle- and lower-income workers, and could well raisebenefits for higher-income workers.

19Only about 4 percent of eligible taxpayers made the max-imum $2,000 contribution to a traditional IRA in 1995 (Carroll2000, Copeland 2002). About the same percentage of workerscontributed the maximum ($9,500) to their 401(k) plan in 1996(Richardson and Joulfaian 2001). Some individuals may havecontributed less than $9,500, but still have been constrainedby limits imposed by their employers or required by law;however, it is unlikely that this group is very large.



(C) T

ax Analysts 2003. A




VII. Private and National Saving

The proposal is designed to affect private saving andalmost certainly will. But the impact on private savingmay not be positive, especially in the first 5 to 10 years.And the effect on national saving — private saving pluspublic saving — is almost surely negative over the firstdecade. Although we have provided revenue effects ofselected proposals above, it is difficult to quantify theimpact on private saving without more detailedmodels. Thus, we discuss the qualitative effects of thevarious behavioral responses, following the format ofTable 4.

The transfer of existing taxable assets to LSAs is afirst-order concern, as it will have literally no impacton private saving, but will drain public revenues.Likewise, the placing of future saving that would havebeen done anyway in an LSA rather than in some tax-able asset will have literally no impact in itself onprivate saving, but will reduce public saving. Ofcourse, the level of nonretirement private saving mayalso change. This saving could rise because of thehigher after-tax rate of return available through LSAs.On the other hand, the substantial income and wealtheffects associated with the first two changes above —the transfer of existing assets and future saving thatwould have been done anyway — could easily raiseconsumption (that is, reduce other nonretirementsaving).

The conversion of traditional IRAs to RSAs will raisenational saving in the short term because of the posi-tive revenue effects of the conversions, but will reducenational saving in the long term because of the long-term drain on revenue, which exceeds the short-termgain in present value.

It is also possible that some households will raisetheir level of nonemployer retirement saving. Thiscould occur for two groups: those that are alreadysaving the maximum amount in an IRA or 401(k); andthose with very high income (above $160,000) whowould be newly eligible to make tax-advantaged con-tributions. As noted above, however, very fewhouseholds are constrained by current contributionlimits. Moreover, the households that are constrainedtend to be higher-income households, as of course arethe households that would be made newly eligible forthe contributions via the removal of income eligibilitylimits. Several studies of 401(k) plans, pensions, andIRAs find that contributions to the plans by householdswith high levels of income are less likely to representnet additions to private saving than are contributionsby lower- and middle-income households.20 Thus, thelikely impact on private saving from increased contri-

butions to nonemployer retirement accounts is likelyto be small; if it is, the effect on national saving couldwell be negative because a small increase in privatesaving would be outweighed by the reduction in taxrevenues. Moreover, some of the strongest advocatesof IRAs believe that the removal of the up-front deduc-tion for contributions could serve to reduce participa-tion significantly (Altman 2003).

The effects of the proposal on pension coverage,participation, and contributions would be a criticalcomponent of its impact on private and nationalsaving. The administration claims that the greatersimplicity of plan design and nondiscrimination ruleswill encourage some business owners to create newplans. But a recent EBRI (2002) study reports that lessthan 30 percent of small businesses that do not offerpensions cite pension rules, broadly defined, as thereason why. Concerns such as that employees preferwages or other benefits, revenue is too uncertain tocommit to a pension plan, and a highly seasonal orpart-time work force appear to weigh more heavily.About 11 percent of respondents did list the cost ofrequired contributions and about 15 percent listed ad-ministrative costs of one sort or another as the primaryreason.

A second factor affecting pension coverage wouldbe that the vast increase in the amount of tax-freesaving that taxpayers would be able to do outside ofretirement plans would reduce the incentives for smalland medium-sized businesses to offer plans. For ex-ample, under the proposal, a business owner and herspouse could shelter $30,000 per year in saving outsideof any company retirement plan ($7,500 each in an LSAand $7,500 each in an RSA). Thus, many businessowners and managers may find that they can meet allof their demands for tax-free saving without the hassleand expense of maintaining an employer-sponsoredplan.21

Holding pension coverage constant, it seems likelythat pension participation and/or average contri-butions for low- and middle-income workers wouldfall due to the relaxation of nondiscrimination rules(see Appendix A). To the extent that firms did droppension coverage, rank-and-file workers would ofcourse have the option to make up the contributionsthemselves by contributing to their own RSAs. How-ever, evidence suggests strongly that participation innonemployer-based saving plans, like IRAs, is likelyto be much lower than participation rates amongworkers eligible for an employer-based plan. For ex-

20The effects of 401(k) plans on household wealth ac-cumulation is controversial. Early studies by Poterba, Venti,and Wise (1995, 1996) found that virtually all 401(k) contri-butions represented new saving. More recent studies, usingimproved econometric techniques, have found that contri-butions by lower- and moderate-income households tend torepresent net additions to saving, but that contributions by

(Footnote 20 continued in next column.)

higher-income households tend to represent asset reshuffling(Benjamin 2000, Engelhardt 1999, Engen and Gale 2000).Engen and Gale (2000), for example, find that contributionsmade by households with income below $40,000 tend to rep-resent new saving, but that contributions by higher-incomehouseholds do not show up as increased private wealth.

21Several researchers have found evidence that theavailability of nonemployer based retirement saving options,like IRAs, reduce the amount that employees contribute toemployer-based plans (see Burman, Johnson, and Kobes2003, Engelhardt and Kumar 2002, Engen, Gale, and Scholz1994).



(C) T

ax Analysts 2003. A




ample, even in the early 1980s when IRAs were univer-sally deductible, participation rates never exceededabout one-quarter of the population in any given year,with much lower participation rates among low- andmiddle-income families. In contrast, among eligibleworkers, 401(k) participation rates are consistently inthe range of 60 to 70 percent across almost all incomelevels (Poterba, Venti, and Wise 1995). And participa-tion is highest in cases where employers offer a match-ing contribution. The weaker nondiscrimination rulesare likely to reduce the prevalence of employermatches, thus reducing worker participation evenwhere a firm continues to offer a pension.

VIII. Other IssuesA. Interactions With Dividends/Cap. Gains Proposal

Besides the saving proposal, the administrationrecently unveiled a proposal to remove the doubletaxation of dividends. The interactions between theseproposals are worth considering. Both proposals serveto exempt capital income from taxes, and it is possiblethat, in combination, they would eliminate all, or al-most all, individual-level taxes on capital income forall but the extremely wealthy. Note, however, that thetwo proposals undermine each other.

The saving proposal reduces the effectiveness of thedividend proposal in encouraging firms to report in-come and pay dividends. As Gale and Orszag (2003a)show, the dividend proposal provides no incentive forfirms to reduce sheltering to the extent that a firm isowned by tax-exempt shareholders. It only providesan incentive when the firm is held by taxable share-holders. The saving proposal will increase the share ofcapital income that is tax-exempt and hence attenuatethe effects of the dividend proposal on discouragingsheltering. The saving proposal would also reduce theincentive created by the dividend plan for firms toreduce borrowing — under the combined proposals,firms could borrow and deduct their interest payments,but individuals could avoid paying taxes by puttingcorporate bonds in their LSAs.

Likewise, the dividend proposal may reduce therelative attractiveness of both employer-provided pen-sion plans and individual tax-free savings accounts:Shares held in taxable accounts invested entirely inshares of firms with fully taxed corporate incomewould effectively enjoy tax treatment almost like RSAs.Contributions to these accounts would be made withafter-tax dollars, but dividends and capital gains thatare due to retained earnings would be untaxed at theindividual level.22

B. Fundamental Tax ReformThe president’s saving proposal is not a revenue-

neutral shift to a consumption tax. Our estimates sug-gest substantial revenue losses over the next 10 yearsthat would continue to grow in subsequent years.

Moreover, despite a superficial resemblance, theproposal is not a consumption tax. First, a consumptiontax would have consistent treatment of capital incomeand expense. It would not allow capital income to beexempt from tax — through the LSA and RSA (anddividend proposal) — while at the same time allowingpayers of interest — both corporations and individuals— to deduct their interest payments. Second, a con-sumption tax would only exempt the return to newcapital investments, not to old capital. The president’sproposal would exempt the return to old capital. Thus,it should be considered along the lines of consumptiontaxes with transition relief, or essentially more like awage tax than a consumption tax. Notably, the eco-nomic growth benefits of a consumption tax with tran-sition relief are much smaller than those of a consump-tion tax without transition relief (Altig et al. 2001).Even those smaller economic benefits may be offset bythe new opportunities for arbitrage that would becreated by allowing interest expense to remain deduct-ible while capital income is taxable, and would needto be assessed against increased regressivity, reducednational saving, and increased public debt burden.

IX. ConclusionThe administration’s tax-free saving proposals

would help simplify some aspects of tax rules forsaving. But the proposals may also create complex in-teractions with the saver ’s credit for low-incomehouseholds and with other existing tax-deferred savingplans. Moreover, the effort to simplify the system doesnot require massive increases in contribution limits orthe creation of substantial new tax-preferred savingsvehicles. That is, the most costly features of the pro-posals have nothing to do with simplification but muchto do with allowing high-income households to sheltersubstantially more assets than they can under currentlaw.

The administration’s proposals might be the rightpolicy prescription if the key problems facing the pen-sion system were that workers with income above$160,000 were unable to save adequately for retire-ment, low- and moderate-income workers were cur-rently receiving too many subsidies through the pen-sion system, and the budget outlook were so positivethat policy makers were worried about ever-expandingsurpluses.

In our view, however, by expanding tax-free savingsopportunities primarily for higher-income households,and potentially reducing saving and pension coverage

22Indeed, to the extent that the normal return to corporatestock is tax-exempt under the dividend proposal, but anyexcess return is taxable (or loss deductible), investors wouldprefer to make these investments outside an LSA. Taxation(assuming full deductibility for losses) reduces the varianceof after-tax returns without affecting the mean, whichproduces no net revenue for the government, but makes risk-averse investors better off. Capital losses are only currentlydeductible to the extent of other capital gains plus $3,000,

(Footnote 22 continued in next column.)

but Auerbach, Burman, and Siegel (2000) show that almostall losses are taken within two years. Given the sustainedstock market declines of the past couple of years, the losslimit is surely more binding now, so the benefit from riskreduction may be offset for many taxpayers because gainsand losses are taxed asymmetrically.



(C) T

ax Analysts 2003. A




among lower- and middle-income households, the ad-ministration’s proposals could reduce private savingand would almost surely reduce national saving overthe next decade, decrease the number of householdswho save adequately for retirement, make the tax sys-tem less progressive, and exacerbate an already dif-ficult fiscal situation.

We believe that one of the key problems facing thepension system is the current mismatch between wherepension benefits are provided and where those benefitswould do the most good. In our view, the private pen-sion system provides disproportionate and expensivebenefits to high-income households. Those benefitsgenerate little improvement in the adequacy of savingfor retirement, since the beneficiaries would save sub-stantial amounts even without tax subsidies. Moreover,the tax subsidies generate little increase in privatesaving, since very-high-income households tend tosubstitute existing assets or saving that would havebeen done anyway into tax-preferred vehicles, rather

than reducing their current living standards to financetheir deposits. In contrast, lower- and middle-incomehouseholds gain less from the pension system, but pen-sion benefits targeted at them can both increase savingand help households who would otherwise save inade-quately for retirement. A key objective of pensionreform should therefore be to encourage expandedpension coverage and participation among low- andmiddle-income households, a step that would boostnational saving and build wealth for households, manyof whom are currently saving too little. Reforms alongthese lines could include expansion of the existingsaver ’s credit, changes to the default choices in 401(k)plans, improved financial education, options for en-couraging diversification of 401(k) accounts, and pen-sion simplification (including exempting a modestamount per individual from the minimum distributionrules and transforming the nondiscrimination rulesinto a minimum contribution regime) (Gale and Orszag2003b).

Appendix ANondiscrimination Rules

A. Current LawCurrent law limits the benefits qualified pension

plans may provide to highly compensated employees(HCEs, basically defined in 2003 as those earning$90,000 or more) relative to rank-and-file workers,known as non-highly-compensated employees(NHCEs). One requirement hinges on coverage rates.This requirement can be met if the share of rank-and-file workers covered is at least 70 percent of the shareof HCE workers covered.23 For example, if all HCEs arecovered, 70 percent of rank-and-file workers wouldneed to be covered.

An additional set of tests applies specifically to401(k) and similar defined contribution plans. Underthese tests, the share of compensation contributed byHCEs is limited by the share of compensation contrib-uted by rank-and-file workers. For example, if theaverage contribution rate for rank-and-file workers isbetween 2 and 8 percent of salary, the average HCEcontribution rate cannot be more than 2 percentagepoints higher.24 Safe harbor rules allow firms to avoidthese tests by offering specific patterns of 401(k) em-ployer matching or nonmatching contributions; a planfollowing the safe-harbor design satisfies the testregardless of actual take-up behavior.

A variety of other rules apply as well, some makingthe allowable pattern of contributions more progres-sive and some making it less progressive. The “per-mitted disparity” rules allow higher contributions forhighly compensated employees, in a manner that wasintended to reflect the offsetting progressivity of SocialSecurity benefits. The “cross-testing” rules allowhigher contributions for older workers.25 The top-heavy rules require plans in which more than 60 per-cent of benefits accrue to “key” employees (such ascertain executives and owners) to provide a minimumcontribution equal to 3 percent of compensation torank-and-file workers.

B. The Administration’s ProposalThe administration’s proposal would change these

rules in several ways. First, it would change the non-discrimination rule applying to 401(k) and othersimilar plans so that the average contribution rate forHCEs could not exceed twice the average contributionrate for NHCEs if the NHCE contribution rate averaged6 percent or less. If the NHCE contribution rateaveraged more than 6 percent, no nondiscriminationtesting would apply. The administration’s proposalwould also loosen the safe-harbor rules allowing firmsto avoid nondiscrimination testing through specificplan designs.26 The administration’s proposal would

23This is the so-called ratio percentage test. An alternativeway of meeting the coverage requirement is through theaverage benefits test. For a description of these tests, seeOrszag and Stein (2001).

24If the average contribution rate for NHCEs is less than2 percent, the average HCE contribution rate is limited to nomore than 200 percent of the average NHCE contributionrate. If the NHCE average contribution rate is more than 8percent, the HCE average contribution rate can be no morethan 125 percent of the NHCE average rate. Two similar andparallel tests (the ADP test and the ACP test) evaluate em-ployee pretax contributions (ADP test) and the combinationof employer contributions and employee after-tax contrib-utions (ACP test).

25Orszag and Stein (2001) give examples of how the cross-testing rules can be used to subvert the intent of the nondis-crimination rules.

26To enjoy a safe harbor under the administration’s pro-posal, the plan must be such that all employees are eligibleto receive fully vested employer contributions —includingnonelective (automatic) and matching contributions — of atleast 3 percent of compensation. In addition, any matchingcontributions must follow one of two rules: (1) the employermatch would be 50 percent of employee contributions up to

(Footnote 26 continued on next page.)



(C) T

ax Analysts 2003. A




also repeal the top-heavy rules, eliminate permitteddisparity and cross-testing, and remove a secondarytest for demonstrating that a plan covers an adequateshare of NHCEs.

These proposals would have complicated effects. Togive a sense of some of the changes, the following twotables simply list the required minimum NHCEaverage contribution rate for a given HCE contributionrate (Table A1, above) and the maximum HCE averagecontribution rate given an average NHCE contributionrate (Table A2, next page), under current law and theadministration’s proposal, respectively. Note that thesetables ignore the safe-harbor rules, which would alsobe loosened somewhat under the administration’s pro-posal, and do not attempt to take into account em-ployer-matching contributions and employer-nonelec-tive contributions in general.

These tables show, for example, that if a firm wantsto allow HCE workers to contribute 10 percent ofsalary, it would currently have to ensure a contributionrate of 8 percent for NHCE workers. Under the newlaw, the firm would have to ensure only a 5 percentcontribution rate for NHCE workers. Thus, in manycircumstances, it may be possible to reduce contri-butions for lower- and middle-income workers underthe new rules. On the other hand, the firm may be more

inclined to offer the plan in the first place if the con-tributions for rank-and-file workers were lower.

A simplified example may provide further insightinto the possibility that the changes will in some cir-cumstances allow pension contributions made on be-half of rank-and-file workers to fall, while maintainingcontributions for HCE workers. For example, considera firm with two HCEs and five other employees. TheHCEs earn $110,000 and $120,000 respectively, whilethe others earn between $25,000 and $70,000. The rankand file workers contribute an average of 5 percent ofcompensation to the firm’s 401(k) plan, with anaverage dollar contribution of $2,130.27 Under currentlaw, the HCEs are allowed to contribute 7 percent ofcompensation, which is 2 percentage points higherthan the average NHCE contribution rate. The result isan average dollar contribution for the HCEs of $8,050.

This pattern of contributions results in slightly morethan 60 percent of the aggregate contributions beingmade by key employees. Assuming that the accountbalances of the key employees also account for morethan 60 percent of aggregate account balances underthe plan, the firm must provide a contribution of 3percent of pay to NHCE workers to satisfy the top-heavy rules.28 Including these employer contributions,the total contributions made to NHCE accountsaverage $3,510 and the total contributions to HCE ac-counts average $8,050 (since no employer contri-butions are required to be made on behalf of the HCEemployees under the top-heavy rules).

Under the administration’s proposal, the two HCEscould meet all of their retirement saving needs andcontribute even more to retirement just by creatingRSAs for themselves and their spouses and droppingthe firm’s pension plan altogether. If they retained thepension plan, though, the 5 percent average contri-bution rate for the rank-and-file workers would allow

Appendix Table A1Minimum Contribution Rate for NHCEs Given Contribution Rate for HCEs

HCE Contribution Required Minimum NHCE Contribution Rate

Current Law Administration’s Proposal

0 0 0

2 1 1

4 2 2

6 4 38 6 4

10 8 5

12 9.6 6

6 percent of compensation; or (2) the match rate as a percent-age of employee contributions does not increase with thecontribution and the aggregate match rate must be at leastas great as the rate that would apply under the first formula.Also, the match rate cannot be higher for HCEs than forNHCEs. These rules represent a loosening of the current safeharbor rules. For example, under current law, the least strin-gent safe harbor design requires matching contributions onbehalf of NHCEs to be 100 percent of employee contributionsup to 3 percent of compensation, plus 50 percent of employeecontributions between 3 and 5 percent of compensation.Under the administration’s proposal, by contrast, the planwould qualify for the safe harbor if matching contributionswere 50 percent of employee contributions up to 6 percent ofcompensation. The required aggregate matching contri-bution for an employee making contributions of 6 percent ofcompensation are thus reduced from 4 percent of compensa-tion (100 percent up to 3 percent, plus 50 percent between 3and 5 percent) to 3 percent of compensation (50 percent of 6percent). For workers making contributions of less than 6percent of compensation, the reduction in aggregate requiredmatching contributions is generally larger.

27For purposes of the nondiscrimination tests, the“average” contribution rate for NHCEs and HCEs is a simpleaverage of each individual’s contribution rate and is notweighted by compensation.

28We assume that the HCEs are both key employees andthat none of the NHCEs are key employees. The top-heavyrules mandate a nonelective 3 percent contribution on behalfof the NHCE employees.




(C) T

ax Analysts 2003. A




Appendix Table A2Maximum Contribution Rate for HCEs Given Contribution Rate for NHCEs

NHCE Contribution Maximum Allowed HCE Contribution Rate

Current Law Administration’s Proposal

0 0 0

2 4 44 6 8

6 8 12

8 10 No nondiscrimination testing

10 12.5 No nondiscrimination testing

12 15 No nondiscrimination testing

Appendix Table A3Hypothetical Firm Under Current Law

Worker CompensationEmployee

Contribution ($) %Employer

Contribution ($) %Total

($)

NHCE

A 25,000 1,250 0.05 750 0.03 2,000

B 35,000 2,100 0.06 1,050 0.03 3,150

C 45,000 3,150 0.07 1,350 0.03 4,500

D 55,000 2,750 0.05 1,650 0.03 4,400E 70,000 1,400 0.02 2,100 0.03 3,500

HCE

F 110,000 7,700 0.07 — 7,700

G 120,000 8,400 0.07 — 8,400

Average NHCE 46,000 2,130 0.05 1,380 3,510

Average HCE 115,000 8,050 0.07 — 8,050Average total 65,714 3,821 986 4,807

Appendix Table A4Hypothetical Firm Under Administration’s Proposal

Worker CompensationEmployee

Contribution ($) %Employer

Contribution ($) %Total

($)

NHCE

A 25,000 1,250 0.05 0 0 1,250

B 35,000 2,100 0.06 0 0 2,100C 45,000 3,150 0.07 0 0 3,150

D 55,000 2,750 0.05 0 0 2,750

E 70,000 1,400 0.02 0 0 1,400

HCE

F 110,000 11,000 0.1 0 0 11,000

G 120,000 12,000 0.1 0 0 12,000

Average NHCE 46,000 2,130 0.05 — 0 2,130Average HCE 115,000 11,500 0.10 — 0 11,500

Average total 65,714 4,807 4,807



(C) T

ax Analysts 2003. A




the HCEs to contribute 10 percent of pay (rather than7 percent under current law). In addition, the removalof the top-heavy rules would eliminate the need for anemployer contribution to NHCE accounts. The result,as shown in the table below, is that the average con-tribution made to the HCE accounts rises to $11,500,but the average contribution made to the NHCE ac-counts declines from $3,510 to $2,130. The averageacross the entire firm remains $4,807, indicating thattotal contributions to the plan remain constant (at$33,650). The distribution of those contributions, how-ever, shifts from NHCEs to HCEs. Under current law,52 percent of the total contributions are made to NHCEaccounts; under the administration’s proposal, 32 per-cent of the total contributions are made to NHCE ac-counts.

It is important to emphasize that this hypotheticalexample could not — and clearly does not — capturethe full effects of changing the nondiscrimination rules.Some analysts believe that loosening the nondis-crimination rules will reduce the costs and complexityof providing pensions, increase pension coverage, andthereby actually raise the benefits accruing to workersat all income levels (Clark, Mulvey, and Schieber 2000),although little empirical work has been conducted onthis issue. If the provisions creating LSAs and RSAswere enacted, however, the costs of nondiscriminationrules would likely weigh more heavily than under cur-rent law. If employers and highly compensated em-ployees could contribute $30,000 to these two accountsfor themselves and their spouses, there may be muchless tolerance for any costs associated with providinga employer-based pension plan.

Appendix BDeterminants and Estimates of Rollover Behavior

This appendix describes the set of calculations madeto provide revenue estimates of induced changes inrollover behavior for households with income above$100,000. We assume there would be no changes inconversions for households with income below$100,000 since they can already convert their fundsunder current law.

A. Aggregate IRA BalancesWe estimate that households with income above

$100,000 held about $900 billion in traditional IRAs in2001. This estimate is obtained by estimating tradition-al IRA values from the 1998 Survey of Consumer Finan-ces (using the procedure described in footnote 20), in-flating by the proportional increase in aggregate IRAvalues from 1998 to 2001 in the Federal Reserve’s Flowof Funds, and then deflating by 10 percent to accountfor Roth, nondeductible, and other types of nontradi-tional IRAs.

B. Determinants of the Decision to Convert to an RSAConsider an investor with $A in a traditional IRA

and $B in a taxable saving account. Let r be the pretaxinterest rate, τ be the tax rate, and n be the holdingperiod. We set n equal to 70 minus the age of the headof household (assuming that the average dollar froman IRA will be withdrawn at age 70); and set r accord-ing to the current yield curve for Treasury bonds formaturity closest to n. The tax parameter is set at theaverage marginal tax rate for households given filingstatus, income category, age (over and under 65), andnumber of children (one, two, or three or more) calcu-lated by the Tax Policy Center microsimulation modelin 2003. The tax rate is assumed to be constant overtime.

In n years, the traditional IRA balance will grow toA(1+r)n and upon withdrawal in year n would net thetaxpayer A(1+r)n(1-τ). The $B in taxable assets willgrow to equal B(1+ r(1-τ))n after taxes in n years, sothat her total after-tax wealth in n years is A(1+r)n(1-τ)+ B(1+ r(1-τ))n. If the investor converts the IRA to anRSA in 2003, the entire tax-preferred balance A growsat the tax-free rate, and is withdrawn without tax

liability in year n. But she pays a tax of τA in 2003 outof her taxable assets. Thus, she ends up with A(1+r)n

+ [B-τA](1+ r(1-τ))n.The value of her tax saving from conversion is the

difference between the two expressions, or τA(1+r)n -τA(1+ r(1-τ))n. The first term shows the benefits ofbeing able to shelter more funds. The second term isthe loss in future income due to the fact that taxes werepaid in 2003 out of taxable assets. Taking the presentvalue of that expression, and dividing by τA gives thetax savings as a share of the up-front tax payment:

PV = (1+r)n−(1+r(1−τ))n

(1+r)n = 1- (1+r(1−τ))n

(1+r)n {1}While it is reasonable to assume that the likelihood

of converting an IRA to an RSA would be an increasingfunction of the present value of the tax savings givenin {1} , it is unreasonable to believe that everyhousehold with a positive present value of convertingwould choose to convert. First, liquidity-constrainedhouseholds may be unwilling to convert even in ex-change for significant tax savings in present value. Thesmaller the required up-front tax payments relative tononretirement financial assets, the more likely it is thata conversion will appear attractive. Second, thelikelihood of converting depends on whether the tax-payer wishes to save more for retirement than he orshe has been able to do in tax-preferred form. We positthat the greater the share of retirement assets in thehouseholds’ overall portfolio of financial assets, theless likely the household would be to convert the IRAto an RSA, since the household is less likely to want toincrease retirement saving.

C. Calibrating the ParametersAppendix Table B-1 identifies the share of tradition-

al IRA balances that would be converted to RSAs forhouseholds with income below $100,000 under severaldifferent threshold values for the parameters notedabove. For example, if people converted only when (a)the present value of the tax saving exceeded 20 percent




(C) T

ax Analysts 2003. A




Appendix Table B1Percent of IRA Assets Converted in Households With AGI Less Than $100,000

PV of tax savingsgreater than: 0.1 0.2 0.3

Taxes paid on conver-sion as % of nonretire-ment financial assetsless than:

0.1 5.5 2.1 0.8

0.2 8.4 2.9 1.1

0.3 9.3 3.0 1.1

Note: Rollovers assumed to occur only when the ratio of retirement assets to total financial assets is less than 0.5.

Appendix Table B2Predicted Amount of IRA Assets Rolled Over Into RSAs ($ millions)



0.1 143,649 74,694 49,474

0.2 219,694 108,869 63,661

0.3 270,548 143,128 80,938


Appendix Table B3Predicted Short-Term Revenue Gain From IRA Rollovers ($ millions)



0.1 49,217 25,714 17,1650.2 75,475 37,018 21,956

0.3 93,381 49,481 28,242


Appendix Table B4Estimated Cost Per Dollar Raised From IRA Rollovers



0.1 1.24 1.33 1.37

0.2 1.23 1.32 1.370.3 1.24 1.31 1.37


Notes: Marginal tax rates used in these calculations do not reflect the acceleration of marginal tax rate cuts proposed in theadministration’s budget. Universe (except for Table B1): Households with 2003 adjusted gross income in $2001 over$100,000 with one and only one tax return. Married couples filing separately and nonmarried couples with two returnsare excluded. Nonretirement assets equal total financial assets minus retirement assets. Financial assets were inflatedfrom the 1998 SCF by the proportional increase in financial assets of the household and nonprofit sector from 1998 to 2001in the Flow of Funds. IRA and Keogh assets were inflated from the 1998 SCF by the proportional increase in IRA andKeogh assets from 1998 to 2001 in the Flow of Funds; the remaining retirement assets were inflated from the 1998 SCF bythe proportional increase in defined contribution assets from 1998 to 2001 in the Flow of Funds.



(C) T

ax Analysts 2003. A




of the up-front tax payment, (b) the current tax liabilityon the conversion would be less than 20 percent ofexisting financial assets, and (c) retirement assets wereless than 50 percent of total financial assets, thenhouseholds with income below $100,000 would convertabout 2.9 percent of their traditional IRA balances. Thetable also shows wide variation based on reasonablevariations in the parameters above.

We use these parameters to generate our central es-timates for rollover behavior. We compute the valueswith τ set to its current-law level. As shown in Appen-

dix Table B-2, using the parameter values noted above,we estimate that about $109 billion in IRA balanceswould be rolled over and would raise revenue in theshort term by about $37 billion. Again, there is widevariation in the results.

Finally, as noted above, rollovers represent a veryexpensive way for the government to borrow againstfuture revenues. Appendix Table B-4 reports the ratioof the present value of the long-term revenue lossdivided by the short-term gain. In the base case, thatratio is 1.32.

ReferencesAltig, David, Alan J. Auerbach, Laurence J. Kotlikoff,

Kent A. Smetters, Jan Walliser. 2001. “SimulatingFundamental Tax Reform in the United States.”American Economic Review. June; 91(3): 574-95.

Altman, Daniel. 2003. “Accounts Chock-Full or PlanHalf Empty,” The New York Times, February 1.

Andrews, Edmund. 2003. “Bush’s Plan for Pensions IsNow Given Low Priority,” The New York Times,February 26.

Auerbach, Alan, Leonard E. Burman, and JonathanSiegel. 2000. “Capital Gains Taxation and Tax Avoid-ance: New Evidence From Panel Data,” in Joel Slem-rod, ed., Does Atlas Shrug?: The Economic Conse-quences of Taxing the Rich, Russell Sage Foundationand Harvard University Press.

Benjamin, Daniel J. 2000. “Does 401(k) Eligibility In-crease National Savings?” Mimeo., London Schoolof Economics.

Burman, Leonard E., Richard W. Johnson, and DeborahI. Kobes. 2003. “Pensions, Health Insurance, and TaxIncentives,” The Urban Institute.

Burman, Leonard, William Gale, and David Weiner.2003. “The Taxation of Retirement Saving: ChoosingBetween Front-Loaded and Back-Loaded Options.”

Campbell, David, Michael Parisi, and Brian Balkovic.2000. “Individual Income Tax Returns, 1998,” SOIBulletin, Fall, pp. 8-46.

Carroll, Robert. 2000. “IRAs and the Tax Reform Act of1997,” Office of Tax Analysis, Department of theTreasury, January.

Clark, Robert L., Janemarie Mulvey, and Sylvester J.Schieber. 2000. “The Effects of NondiscriminationRules on Pension Participation.” Forthcoming inPrivate Pensions and Public Policies, ed. by WilliamG. Gale, John B. Shoven, and Mark J. Warshawsky.Brookings.

Congressional Budget Office. 2003. The Economic &Budget Outlook: Fiscal Year 2004-2013.

Copeland, Craig. 2002 “IRA Assets and Characteristicsof IRA Owners,” EBRI Notes, December.

Dusseault, Brianna and Jonathan Skinner. 2000. “DidIndividual Retirement Accounts Actually RaiseRevenue?” Tax Notes, Feb. 7, p. 851.

Employee Benefit Research Institute. 2002. “The 2002Small Employer Retirement Survey (SERS): Sum-mary of Findings.

En gelhardt , Gary. 1999. “Have 401(k)s RaisedHousehold Saving? Evidence from Health andRetirement Study.” Mimeo. (October).

Engelhardt, Gary V. and Anil Kumar. 2002. “EmployerMatching, Non-Linear Budget Sets, and Employee401(k) Saving: Evidence from the Health and Retire-ment Study.” Syracuse University. December 31.

Engen, Eric and William Gale. 2000. “The Effects of401(k) Plans on Household Wealth: DifferencesAcross Earnings Groups.” NBER Working Paper No.8032 (October).

Engen, Eric M., William G. Gale, and John Karl Scholz.1994. “Do Saving Incentives Work?” Brookings Paperson Economic Activity, 1994:1, 85-151.

Engen, Eric M., William G. Gale, and John Karl Scholz.1996. “The Illusory Effect of Saving Incentives onSaving.” Journal of Economic Perspectives 10(4): 113-38.

Gale, William and Peter Orszag. 2003a. “The Adminis-tration’s Proposal to Cut Dividend and CapitalGains Taxes.” Tax Notes, Jan. 20, p. 415.

Gale, William and Peter Orszag. 2003b. “Private Pen-sions: Issues and Options.” Forthcoming in Agendafor the Nation, edited by Henry J. Aaron, JamesLindsay, and Pietro Nivola, Brookings.

Gravelle, Jane. 1993. “Estimating Long-Run RevenueEffects of Tax Law Changes.” Eastern Economic Jour-nal, vol. 19, no. 4, Fall.

Halperin, Daniel. 1998. “I Want a Roth IRA for Xmas.”Tax Notes, Dec. 21, 1998, p. 1567.

Orszag, Peter and Stein, Norman. 2001. “Cross-TestedDefined Contribution Plans: A Response to Profes-sor Zelinsky,” Buffalo Law Review, 49:2, Spring/Sum-mer pp. 629-674.

Poterba, James M., Steven F. Venti, and David A. Wise.1995. “Do 401(k) Contributions Crowd Out OtherPersonal Savings?” Journal of Public Economics 58:1-32.

Poterba, James M., Steven F. Venti, and David A. Wise.1996. “How Retirement Programs Increase Savings.”Journal of Economic Perspectives 10(4): 91-112.

President’s Commission to Strengthen Social Security.2001. Strengthening Social Security and Creating Per-sonal Wealth for All Americans, December, p. 97.

Rojas, Warren. 2003. “House Taxwriters CompletingBipartisan IRA Reform Plan,” Highlights and Docu-ments, Feb. 7, p. 1627.



(C) T

ax Analysts 2003. A




Richardson, David and David Joulfaian. 2001. “WhoTakes Advantage of Tax-Deferred Saving Programs?Evidence from Federal Income Tax Data,” Office ofTax Analysis, US Treasury Department.

Steuerle, C. Eugene. 1997. “Back-Loaded IRAs: HeadTaxes Replace Income and Consumption Taxes.” TaxNotes, Oct. 6, pp. 109-110.

U.S. Department of Treasury. 2003a. “President’sBudget Proposes Bold Tax-Free Savings and Retire-

ment Security Opportunities for All Americans(press release),” January 31.

U.S. Department of Treasury. 2003b. “General Explana-tions of the Administration’s Fiscal Year 2004 Reve-nue Proposals,” February.

Vanderhei, Jim. 2003. “GOP Not Backing SavingsChanges; Bush Plan Won’t Pass, Leaders Say.”Washington Post, Feb. 7, p. A1.



(C) T

ax Analysts 2003. A




Documents

tax break - Brookings · president’s proposal to cut dividend and capital gains taxes and with proposals for fundamental tax reform. Section IX closes with comments on reform of