Summary and Introduction

As baby boomers retire and draw on their pensions and individual retirement accounts (IRAs), increasing amounts of taxable income will be generated. Contributions to those retirement plans are largely untaxed, tending to reduce taxable income and tax receipts when the contributions are made. If both the structure of existing retirement plans and the tax law remain unchanged, taxable distributions from the plans will become relatively more important with the aging of the population, tending to boost future taxable income and receipts, not only in nominal terms but also as a percentage of gross domestic product (GDP). In effect, some potential tax revenues have been shifted from the past and the present to the future.

The question is, "How much?" The answer has a direct bearing on long-term budget projections. This report examines the extent of the shift in the pattern of income tax receipts that is in store as these deferred taxes are realized over the coming decades. The Congressional Budget Office (CBO) estimates that deferred taxes will modestly increase income tax receipts as a percentage of GDP. At the end of 75 years, the effect is to make receipts about 0.5 percent of GDP higher than in 2003, with about one-half of the rise in the next 25 years (see Figure 1). Although a substantial stock of such potentially taxable funds has been accumulated, their effect on future budgets will be modest because the tax-deferred status of new contributions to retirement plans and of the investment returns on existing assets will offset much of the anticipated taxes on withdrawals.
 

Figure 1.

The Effects of Tax-Deferred Retirement Plans on Income Tax Receipts, 2003 to 2078

(Percentage of GDP)

Source: Congressional Budget Office.

Notes: Data are normalized to zero in 2003.

The initial dip in revenues is associated with restoring pension plans to a fully funded status and has no effect on the long-term results.

Those results hold true under a variety of assumptions about GDP growth, inflation, earnings in the plans, and other factors. The estimates are most sensitive to assumptions about the growth of new contributions and the degree to which taxpayers switch to Roth IRAs and "back-loaded" 401(k) accounts, which receive a different form of favorable tax treatment.

A key feature of this analysis is that it examines budget projections--specifically, the effects of existing tax-deferred retirement plans on the pattern of future tax receipts. For this exercise, projections of GDP used to estimate future revenues already implicitly reflect the savings response resulting from the existence of tax-deferred retirement plans, as well as other impacts on corporate capital, profitability, corporate income taxes, and so forth.

In contrast, another type of study examines the overall effect of introducing tax-deferred retirement plans, including those on GDP. In such an exercise, the analysis must separately calculate the tax receipts that would have been generated in the absence of those plans and the presence of them and then compare the two. As a result, the conclusions of such studies would tend to be highly sensitive to the assumption made about the additional saving stimulated by the plans' tax-deferred status. Not surprisingly, the difference in the conclusions from those two types of studies derives from the difference in the questions addressed (see Box 1).
 

Box 1. Projections Versus Cost Estimates of Deferred Tax Revenues Two important questions have arisen regarding the effects on revenues of deferring taxes on retirement plans. First, how would actual revenues differ from those in a world in which such plans did not exist? Second, how will tax deferral affect projected revenues in the future relative to the present, given that such tax treatment exists? The first question is closely related to a policy analysis or revenue cost estimate: how would revenues be different if tax-deferred plans were introduced into an economy that did not have them? It applies to decisions about creating, expanding, or contracting the scope of deferral in the tax system. If comprehensive, such an estimate would take into account the full feedback effects of the plans on saving, investment, and the growth of gross domestic product (GDP). Those changes would, in turn, affect other tax sources within and in addition to the individual income tax. The answer to the question compares predicted revenues to those in a "counterfactual"--a parallel situation in which that tax treatment does not exist--and indicates how the overall existence of tax deferral affects revenues. The result reveals the revenue loss or gain (if any) that would be associated with the change in law. The second question is associated with a revenue projection: what will the stream of revenues be in the future? It applies strictly to the issue of projecting future revenues. The analysis starts with the existing tax system, including tax-deferred plans, and measures how much the taxes associated with deferral change from one year to the next. The analysis does not require the construction of a counterfactual because the tax law is unchanged. Instead, it is a comparison across time. The result tells how the pattern of receipts will differ in the future from what they are now because the environment in which tax deferral operates is changing. Answering the two questions requires different methods. For the sake of illustrating the revenue-projection approach, imagine a relatively simple tax system in which taxes are levied on total output at a single rate with only one exception: contributions to retirement plans are tax deferred, so that income is taxed not when it is earned but when it is distributed in retirement later on. To further simplify matters, ignore the buildup of earnings in tax-deferred plans and consider only the contributions and distributions. Finally, assume that GDP is growing smoothly over time. If an exceptionally large cohort of workers such as the baby-boom generation is advancing through time in the workforce, tax deferral moves some revenues from the early years to the later ones. Instead of growing steadily, like GDP, the stream of revenues would at first grow more slowly, as large numbers of workers made deductible contributions, but then would grow faster, as they retired and took large distributions. Hence, as the figure shows, the revenue pattern looks like the solid, curved line. In 75 years, revenues would be at point A.   In contrast, if revenues grew at the same smooth rate as GDP (the dashed line), they would be at point B. Consequently, the distance between A and B represents the extra receipts resulting from tax deferral when the large cohort of workers has retired. Alternatively, analyzing the issue from the perspective of cost estimating would take into account that some of the saving that went into contributions might never have occurred in the absence of tax deferral. Any such additional saving would increase the capital stock and, eventually, the taxable income of the owners of that capital. Consequently, the appropriate comparison would be between point A and point C, that lower level of receipts associated with the lower growth rate of income (represented by the dotted line)--a comparison between the economy with the retirement plans and a world without them.1 The figure also illustrates the pitfall of mixing the methods associated with the two questions. In particular, one should not introduce the effects of new saving into a revenue projection based on an existing system of tax deferral. The dashed line and point B already incorporate additional saving and the growth effects of tax-deferred saving--the distance from C to B. Adding that effect to the difference from B to A (shown as point D) would double-count and, therefore, significantly overstate the revenue gains. 1.  The figure as drawn assumes that the plans raise saving and boost economic growth. That assumption is for illustration alone and is not meant to suggest that new saving is substantial. If little new saving is associated with the plans, in their absence receipts would be higher.

 

Tax-Deferred Retirement Plans and Their Fiscal Significance

Tax deferral is a common characteristic of many private retirement plans. The Congress has enacted a variety of statutes that establish tax incentives for retirement saving and define the conditions under which they may be used. Those incentives, by their very nature, are designed to be of financial benefit to the taxpayers who use them. At the same time, however, those benefits directly affect the federal budget.

How Tax Deferral Works

The mechanics of tax deferral are straightforward. Contributions to an employment-based retirement plan are not included in the employee's taxable income; similarly, most contributions to traditional IRAs are excluded from taxable income. Investment income accrues within those plans tax free, and taxes are due only when the funds are withdrawn. Because the reduction in tax liability comes primarily at the front end of the process, when contributions are made, this type of tax incentive is known as a "front-loaded" plan.

Consequently, a taxpayer can reduce his or her overall tax liability by contributing to a front-loaded retirement plan instead of to a taxable account. Consider a 45-year-old taxpayer in the 15 percent tax bracket who wants to save $1,000 in before-tax earned income to use in retirement. One option would be to pay the 15 percent tax on the earned income and deposit the remaining $850 in a taxable account that earned 6 percent interest annually. Because interest on that deposit is taxed at 15 percent each year, the effective (after-tax) interest rate on the deposit is only 5.1 percent--the 6 percent before-tax interest rate reduced by the 15 percent tax. By age 60, the taxpayer will have accumulated $1,793, which can be withdrawn without incurring any further taxes.⁽¹⁾

Alternatively, the taxpayer could contribute the full $1,000 to a tax-deferred retirement account, where it would compound at the full 6 percent interest rate. At age 60, the taxpayer will have accumulated $2,397 but must pay a 15 percent tax--or $359--upon withdrawing the savings from the account. After taxes, there will be $2,037, or $245 more than with a regular savings account.

Sometimes, no up-front tax benefit is available in a retirement plan, but interest is still allowed to accumulate tax free until withdrawal. In that case (referred to here as a deferred-liability incentive), the initial deposit would be only $850, but it would compound at the full 6 percent interest rate, yielding $2,037 at age 60. The participant would then pay a tax of $178 on the $1,187 of accumulated interest and have $1,859, or $67 more than with a regular savings account.

The above examples all assume that the individual's tax rate remains constant throughout the cycle of contributions and withdrawals. The tax benefits are even greater if the tax rate declines after retirement. Conversely, a higher tax rate after retirement reduces the benefits. For a taxpayer in the 15 percent bracket prior to retirement, the benefit of deferring taxes does not disappear until the postretirement tax rate reaches approximately 25 percent.

Types of Tax-Deferred Plans

The tax code recognizes a variety of retirement plans that offer either front-loaded or deferred-liability benefits. Employment-based plans offering front-loaded benefits are of two types: defined-benefit and defined-contribution plans. Defined-benefit plans (which are based on various factors that may include years of service and earnings history) promise specific benefits in retirement, and the employer is responsible for accumulating sufficient funds to pay the pension. Workers receive fixed benefits regardless of the returns on invested funds in the retirement plan. Defined-contribution plans, by contrast, specify only how much the employer will contribute annually; payments in retirement depend on what happens to the invested funds. Poor returns on investment yield lower income in retirement, while higher returns result in increased retirement income. Many defined-contribution plans allow employees to make pretax contributions on their own behalf. Such plans are typically known as 401(k) plans.⁽²⁾

Traditional IRAs offer either front-loaded or deferred-liability benefits depending on the income of participants and whether they (or their spouses) are covered by an employment-based plan. All lower-income participants and those higher-income participants who are not covered by an employment-based plan are eligible to claim a deduction for their IRA contributions, thus receiving front-loaded benefits. All others can still contribute but cannot claim a deduction, thus receiving deferred-liability benefits only on the account's investment earnings.

Since 1998, back-loaded benefits have also been available to certain taxpayers through Roth IRAs. Like an IRA with deferred-liability benefits, a Roth IRA has nondeductible contributions, and investment income accrues tax free. But unlike the nondeductible IRA, a Roth IRA has withdrawals that are altogether nontaxable. As long as tax rates remain constant over the cycle of contributions and withdrawals, the ultimate tax benefits of Roth IRAs relative to regular savings are exactly the same as those of deductible IRAs.⁽³⁾

Individuals can also purchase annuities, which, like defined-benefit plans, remove investment risk from the buyer and place it on the seller (usually a life insurance company).⁽⁴⁾ Such annuities can be purchased within an IRA, but most are purchased with after-tax funds and thus provide only deferred-liability tax benefits. Those not purchased within an IRA are referred to here as "nonqualified annuities" because they do not qualify for front-loaded tax benefits.

As of the end of 2002, $10.1 trillion was in tax-deferred retirement plans,⁽⁵⁾ of which $9.0 trillion was taxable upon withdrawal.⁽⁶⁾ Despite the recent trend toward defined-contribution plans, most of those funds were still in defined-benefit plans: $3.1 trillion in private plans (both employment-based pensions and individual annuities) and $2.9 trillion in government plans.⁽⁷⁾ The remaining $4.1 trillion was in private-sector defined-contribution plans and IRAs.

The Phases of the Retirement Saving System

A tax-deferred retirement plan typically goes through two distinct phases in its "life cycle." The first phase--referred to as the funding phase--corresponds to the participant's working years, when he or she makes contributions to the plan, causing it to steadily increase in value. The second phase--referred to as the depletion phase--corresponds to the participant's retirement years, when contributions cease and the participant draws down the funds in the plan.

Aggregated over all participants, the system of tax-deferred retirement saving experiences something similar to the funding phase but may or may not experience something similar to the depletion phase. When applied to the full system, the funding phase is defined as the period during which aggregate contributions exceed aggregate withdrawals. The excess of contributions over withdrawals increases the level of assets in the system, and the investment returns on those assets furthers their growth.

When aggregate withdrawals exceed aggregate contributions, however, the system typically enters a maintenance phase rather than a depletion phase. During a maintenance phase, assets typically continue to accumulate because investment income more than fills the gap between contributions and withdrawals.

If withdrawals ever become so large that investment income is insufficient to cover the gap between contributions and withdrawals, the system enters a depletion phase and assets begin to decline. That phase can occur temporarily if investment returns are weak, the ratio of retirees to workers is exceptionally large, or retirees begin withdrawing substantially higher percentages of their assets each year than previously. But only if there are no new tax-deferred contributions is the phase inevitable--otherwise, the maintenance phase can, in principle, continue indefinitely.⁽⁸⁾

According to the Employee Benefits Security Administration, private defined-contribution plans were still in a funding phase as of 1998.⁽⁹⁾ Private defined-benefit plans, however, have been in a maintenance phase since 1985. Government retirement plans also entered a maintenance phase during the early 1990s.⁽¹⁰⁾ The only occasions when assets declined in any of those types of plans (signifying a temporary depletion phase) were years in which the stock market declined. Even the Civil Service Retirement System, which has accepted no new participants since 1983, is not projected to enter its depletion phase until 2006, according the Office of Personnel Management.⁽¹¹⁾

Accounting for Tax Deferral

The fiscal implications of tax deferral can be thought of as a series of annual tax differences--or the difference each year between, on the one hand, the gain in revenues from the taxation of withdrawals from retirement plans and, on the other hand, the loss of revenues from deferring the taxes on contributions and the investment income from the plans' assets.⁽¹²⁾ In principle, the pattern of tax differences associated with the three phases of the retirement saving system seem straightforward. The funding phase and the maintenance phase, when contributions and investment income exceed withdrawals, can be expected to be characterized by lower revenues for the government--in other words, a negative tax difference. During the depletion phase, however, the excess of taxable withdrawals over tax-deferred contributions and investment income can be expected to generate higher revenues, or a positive tax difference. As the collective system of retirement plans moves from negative to positive tax differences, revenues increase as a percentage of GDP.

However, moving from the phases of the retirement saving system in principle to an actual series of tax differences is complicated by three issues. First, as explained, the system as a whole may never enter a depletion phase; it may stay in the maintenance phase indefinitely. Second, shifting demographics can profoundly alter the progression of flows and the sequence of phases. Third, the tax treatment of investment income that accrues outside of retirement plans can alter the translation of the progression through phases into an effect on revenue flows. Beginning with a simple case in which only the funding and depletion phases occur, a series of examples illustrates those three key points.

Simple Case. If the pattern for all tax deferral looked like that for a single participant, a progression through the depletion phase could be expected. As a result, one could expect negative tax differences as contributions are made and positive tax differences as withdrawals occur. In the example illustrated in Table 1, retirement plans receive tax-deferred contributions equal to 4 percent of income for two periods, then distribute all of their assets in a third period. Income grows at 10 percent per period, and assets earn a 30 percent return in each period. Each period sees the introduction of a new cohort of workers, each the same size as the first cohort.⁽¹³⁾

In this example, tax deferral is introduced in period 1 (during which only one cohort is allowed to participate) and discontinued after period 2 (after a second cohort has been introduced). Thus, only one cohort gets the full benefit of the tax deferral. Furthermore, investment income accruing outside of tax-deferred retirement plans is taxed at regular rates (in this case, 20 percent).

Periods 1 and 2 represent the funding phase, when contributions exceed withdrawals. The tax difference is negative in both periods. Because the tax deferral is first introduced in period 1, only one cohort is making contributions in that period. In period 2, however, two cohorts are contributing, increasing the tax difference.

Periods 3 and 4 represent the depletion phase, when withdrawals exceed the sum of contributions and investment income. The tax difference is positive in both periods. Because period 3 includes the withdrawals of a cohort that contributed in two periods, whereas period 4 includes the withdrawals of a cohort that contributed in only one period, the tax difference is greater in period 3.

Ongoing Tax Deferral. If deferral is a permanent feature of the tax code, new cohorts begin contributing to retirement plans even as old ones withdraw funds. As a result, even as revenues are generated by withdrawals from retirement plans, new contributions remove income from the taxable base at the same time, offsetting the revenue gain. That scenario raises the prospect of an indefinite maintenance phase in which depletion--for the system as a whole--never occurs. Table 2 shows what happens to the earlier example if tax deferral continues. Periods 1 and 2 still represent the funding phase. Beginning in period 3, however, the system moves into a maintenance phase, as contributions drop below withdrawals but investment income more than makes up the difference.

The example illustrates three important points. First, with steady growth in income and stable patterns of saving for retirement, the depletion phase may never occur. Under such circumstances, it would occur only if the tax deferral was terminated. Second, following from the first, it is very possible that the annual tax difference remains negative. There may never be a period in which deferral results in a positive annual tax difference. Third, even if deferral never results in a positive tax difference in any given year, it may still cause a relative improvement in the fiscal position over time. In the early stages of the system, during initial funding, receipts are depressed more than later during the maintenance phase. As a consequence, a negative tax difference during the funding period will overstate the tax difference that can be expected later during maintenance, so over time the fiscal picture may get better.

Demographic Effects. In contrast to the example in Table 2, the population has not grown steadily over time. Indeed, it is the existence of a demographic bulge--the large cohort of baby boomers--that has partly motivated the discussion of the effects of tax deferral on the budget. As the taxpayers who make up that cohort retire in large numbers and make taxable withdrawals from their retirement plans, a smaller cohort of new workers is taking their place in making contributions--raising the possibility that tax deferral may result in a positive tax difference for a period of years.

Table 3 illustrates the effect of a baby-boom cohort, introduced in period 4, that is 10 percent larger than the other cohorts. In this example, the higher contributions in periods 4 and 5 drive the tax difference more negative, although the system remains in a maintenance phase. (In fact, a cohort could be sufficiently large to trigger a temporary reversion to the funding phase.) In period 6, when the large cohort withdraws its funds, the system still remains in a maintenance phase (although a slightly larger cohort would trigger a temporary depletion phase). The system returns to equilibrium (where it was in period 3) in period 7 after the baby-boom cohort leaves the scene.

In this example, the government can expect an increase in revenues equal to 0.2 percent of income when the economy moves from period 5 to period 6. That transition is analogous to what the real-world economy faces in the near future--a baby-boom cohort that will stop contributing to the retirement system and begin withdrawing from it.

The Taxation of Investment Income. As important as the presence of a baby-boom cohort is in influencing the flows into and out of the tax-deferred retirement plans, another factor potentially influences the effects of tax deferral on revenue projections just as much. Under current law, capital gains and dividends earned outside of tax-deferred plans are taxed at lower rates than regular income is. Withdrawals from retirement plans, however, are taxed at regular rates. Thus, the flows into and out of retirement plans do not translate directly to tax differences. Even unchanging flows of contributions, earnings, and withdrawals do not result in unchanging tax differences.

A lower tax rate on investment income outside of retirement plans has two effects on the revenue flows associated with tax deferral, as shown in Table 4. First, it reduces the negative tax difference during the funding phase. In periods 1 and 2, the revenues not collected (because the earnings accrue in retirement plans) are less than they would be if the tax rate that would otherwise have been imposed was greater. Second, it has a similar effect during the maintenance phase (periods 3 through 5). During the maintenance phase, like the funding phase, balances in the plans continue to grow (earnings in retirement plans exceed the amount by which contributions fall short of withdrawals); but because earnings are taxed at a lower rate, the tax difference is not as great as it would be if the tax rates were the same. Indeed, additional revenues associated with the excess of withdrawals over contributions can be greater than the forgone revenues associated with the earnings. Balances in the plans could be growing, but the tax difference under some circumstances could be positive.

Accounting for the tax rate differential is critical to projecting the effects of deferral over time because its effect on the tax difference tends to change even within the maintenance phase. Even when the net flow into the plans (contributions plus earnings minus withdrawals) remains a stable fraction of GDP, the tax difference can change. If investment earnings and the difference between withdrawals and contributions grow relative to GDP by the same dollar amount, the net flow into the plans remains a constant fraction of GDP. However, a 20 percent tax on the difference between withdrawals and contributions (a positive number in the maintenance phase) exceeds the lower tax on an equal amount of investment income, resulting in a net fiscal improvement over time. (That effect, however, requires a more complex example than illustrated in Table 4.)
 

Projections of Flows into and out of Retirement Plans and the Associated Revenues

As described, projecting the effects of tax deferral on future receipts requires estimates of the flow of contributions, the flow of withdrawals, and the earnings on the balances in the retirement plans. Also needed are the tax rates relevant to those flows, including the rates relevant to earnings outside of the plans. From those figures, it is possible to calculate a measure of the tax difference resulting from deferred taxes: the taxes not collected on contributions and earnings netted against the taxes collected on withdrawals.

Using projections of output for the next 75 years, CBO estimated the flows of contributions to, earnings on, and withdrawals from tax-deferred retirement savings.⁽¹⁴⁾ CBO then applied the relevant tax rates to derive the annual tax differences.⁽¹⁵⁾ The resulting tax difference in any year, expressed as a percentage of GDP, can then be compared to that in the current year to determine how much more in taxes relative to GDP can be expected than currently yielded.

According to those projections, defined-contribution plans and IRAs collectively are still in the funding phase but will enter a maintenance phase in 2012, from which they will not emerge during the 75-year period. During that period, contributions and investment income will exceed withdrawals by increasing amounts in almost every year. Throughout the maintenance phase, the net flow of funds (that is, contributions plus investment income minus withdrawals) into such plans will be around 5 percent of GDP per year, down from 6.4 percent in 2003.

Defined-benefit plans have been in a maintenance phase but because, by law, they must be fully funded, CBO assumed that the plans would temporarily revert to a funding phase between 2004 and 2014 in order to be restored to that status.⁽¹⁶⁾ Thereafter, they will remain in a maintenance phase through the end of the 75-year projection period, with the annual net flow of funds into the plans hovering around 3 percent of GDP, roughly the same level as in 2003.

Applying the relevant tax rates and normalizing to zero in 2003 yields a measure of how the receipts-to-GDP ratio will change over the next 75 years as a result of the plans. Except for the near-term period in which defined-benefit plans must be funded, future years show a growth of receipts relative to GDP. But the effects are relatively modest. By 2078, receipts are projected to be about 0.5 percentage points of GDP higher than in 2003 because of the deferred taxes.

The Structure of CBO's Model

CBO's model simulates the mechanics of defined-contribution and defined-benefit plans, generating different projections for each year for both sexes and for various age groups (see Appendix A for full details). CBO derived age- and sex-specific parameters by examining contributions and withdrawals in a base year, 1997. A key assumption behind the model is that the behavior observed for an age group in the base year will be repeated by other workers when they reach that age in later years. The end product of the model is a set of estimates of tax-deferred contributions, total investment income, and taxable distributions for each year between 2003 and 2078.

The part of the model that simulates defined-contribution plans uses assets in the base year as a starting point, then adds contributions and investment income and subtracts withdrawals to advance to the next year's level of assets. Contributions are linked to wage growth, investment income is determined by the projected rate of return, and withdrawals are a share of assets that varies by age and sex but remains fixed over time. The same modeling approach is appropriate for IRAs.

The part of the model that simulates defined-benefit plans works differently. It simulates the future by identifying new retirees each year, estimating the average benefits that they will receive, and then (after accounting for deaths) projecting the benefits of both existing and new retirees forward to the next year. Having estimated the benefits that have been promised, the model then calculates the contributions and investment income required to meet the resulting obligations. That modeling approach applies to private defined-benefit plans, state and local plans, and nonqualified individual annuities. For federally administered plans, the model uses 75-year projections of contributions, investment income, and distributions produced by the administering agencies--adjusted to conform to CBO's forecast of inflation.

To the extent possible, the model relies on data from 1997 tax returns and information returns such as Forms W-2, 5498, and 1099-R. When necessary, however, it uses supplementary data from other sources. CBO tested the model by comparing its results to actual experience over the period 1998 through 2001. On the basis of that comparison, CBO adjusted certain parameters, particularly those relating to the mix of assets in each type of account and the incidence of rollovers to reflect revealed changes in behavior over that period relative to behavior in 1997.

To calculate the tax difference, CBO used effective marginal tax rates based on its tax projection models. CBO used a marginal tax rate of 20 percent for contributions and withdrawals.⁽¹⁷⁾ That rate approximates the current effective income tax rate under the provisions of the Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA).⁽¹⁸⁾ For earnings, CBO computed the tax receipts forgone by deferral using a lower rate of 12 percent for dividends and realized capital gains, which is consistent with the rates in the Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA).⁽¹⁹⁾ CBO applied the regular tax rate of 20 percent to interest income and a zero rate to unrealized capital gains, which are not taxed.⁽²⁰⁾ The model uses the same rates throughout the 75-year time span in order to isolate the effects of tax deferral from other phenomena that may or may not--depending on policy choices--change the ratio of receipts to GDP.

Defined-Contribution Plans and IRAs

Defined-contribution plans and IRAs are characterized over the projection period by tax-deferred contributions that are flat relative to GDP at roughly 2 percent (see Figure 2). Asset levels are high enough at the start of the period that the return on those assets generates investment income (5.5 percent of GDP in 2003) that exceeds withdrawals (1.4 percent of GDP in 2003). As the pool of retirees increases over time relative to the number of workers, withdrawals increase as a percentage of GDP, reaching 6.0 percent by 2078. At no time, however, do withdrawals ever exceed investment income. Hence, the net flow of funds into tax-deferred IRAs and defined-contribution plans does not reverse but remains stable over time, hovering around 5 percent of GDP from 2004 on.
 

Figure 2.

The Flow of Funds into and out of Defined-Contribution Plans and IRAs

(Percentage of GDP)

Source: Congressional Budget Office.

Because of the lower tax rates on certain types of investment income, the mix of assets in retirement plans matters when applying tax rates to estimate the tax difference. IRAs and defined-contribution plans tend to be more heavily invested in stocks than defined-benefit plans are, which has fiscal consequences. The investment income generated by stocks is--outside of retirement plans--taxed at a lower rate than the investment income generated by bonds. Thus, the negative tax difference from IRAs and defined-contribution plans is ameliorated somewhat when compared with that from defined-benefit plans because of the mix of assets in those plans.

When the tax deferral associated with contributions and investment income is compared with the tax on withdrawals, however, the tax difference still remains negative throughout the 75-year period (see Figure 3). The trend, however, is toward zero. The tax difference from defined-contribution plans and IRAs goes from -0.5 percent of GDP in 2003 to -0.1 percent in 2078.
 

Figure 3.

Tax Differences from Defined-Contribution Plans and IRAs

(Percentage of GDP)

Source: Congressional Budget Office.

Defined-Benefit Plans

The flows of defined-benefit plans differ from those of defined-contribution plans and IRAs over the 75-year projection period in a variety of ways. Unlike defined-contribution plans, defined-benefit plans are projected to have distributions that are relatively flat as a percentage of GDP. From 2.9 percent of GDP in 2003, distributions increase to 3.5 percent by 2021 as the baby boomers retire then level off (see Figure 4).
 

Figure 4.

The Flow of Funds into and out of Defined-Benefit Plans

(Percentage of GDP)

Source: Congressional Budget Office.

The pattern of contributions also differs. While contributions to defined-contribution plans are projected to remain a constant percentage of GDP, contributions to defined-benefit plans are projected to decline over time. After a brief spike in the short run, contributions as a percentage of GDP are projected to drop from 2.5 percent to 1.7 percent by 2078. In a stable economy, investment returns in excess of GDP growth play an increasingly important role over time in funding promised benefits in defined-benefit plans, thereby reducing the need for contributions.

The brief spike in contributions results from the assumption that defined-benefit plans will have to make additional contributions over the next several years to return to a fully funded status. Collectively, both private defined-benefit plans and state and local plans were underfunded at the end of 2002. As a result, some of those plans are making exceptionally large contributions in 2004 just to meet minimum funding standards, and they will have to continue making extra

payments for several years to return to fully funded status. In its model, CBO assumes that those extra contributions will be made over 10 years (although Congressional action beyond a recent measure providing brief relief could either shorten or lengthen that period), resulting in substantially higher projections of tax-deferred contributions as a percentage of wages in those years than in either earlier or later years.⁽²¹⁾

The assumption about the period required for such funding--indeed, if full funding occurs at all--does not appreciably affect the conclusions about the effects of defined-benefit plans on revenues over time. If the plans are not fully funded--reducing the taxes deferred now--they will not pay full benefits later, reducing the distributions that will be taxed later (an issue addressed in more detail below).

The remaining decline in contributions results largely from employers' shift away from defined-benefit plans. Most conspicuous is the assumption of virtually no growth in federal employment. Furthermore, an older workforce participating in state and local plans results in slower employment growth in that sector, where defined-benefit plans are ubiquitous, than in the private sector, where they are not.

For defined-benefit plans, investment income exceeds withdrawals throughout the period. Its rate of growth relative to GDP, however, is much slower than for defined-contribution plans because defined-benefit plans are funded only to the extent needed to pay promised benefits. That structure imposes an implicit cap on plans' assets and on the returns that they can earn. Thus, investment income offsets withdrawals just enough to keep the net income flow relatively flat as a percentage of GDP, fluctuating between 2.7 percent and 3.4 percent except during the period when extra contributions are being made to achieve full funding. During that short period, net flows drop from 4.3 percent of GDP to 3.3 percent.

The patterns do not change significantly when tax rates are applied to estimate the annual tax difference (see Figure 5). Investment income in defined-benefit plans is much more likely to be in the form of interest than is investment income in defined-contribution plans. Hence, the lower tax rate on dividends and realized capital gains and the exemption for unrealized gains play less of a role than they do for defined-contribution plans. Rising rapidly from -0.6 percent of GDP in 2004 to -0.4 percent in 2014, the tax difference then fluctuates between -0.2 percent and -0.4 percent of GDP for the remainder of the period.
 

Figure 5.

Tax Differences from Defined-Benefit Plans

(Percentage of GDP)

Source: Congressional Budget Office.

The Combined Effects on Projected Receipts

Combining all types of plans yields a net flow of funds into all tax-deferred retirement plans that fluctuates between 7.7 percent and 9.6 percent of GDP over the 75-year period (see Figure 6). From 2015 on, the percentage never deviates from 8.0 by more than 0.3 percentage points.
 

Figure 6.

Net Flow of Funds into and out of All Tax-Deferred Retirement Plans

(Percentage of GDP)

Source: Congressional Budget Office.

In contrast to the flow of funds, the tax difference--while always negative--shows a distinct trend over time, namely, toward zero. Between 2003 and 2078, the tax difference changes from -0.8 percent to -0.3 percent of GDP (see Figure 7). In other words, CBO projects that the fiscal improvement resulting from tax-deferred retirement plans will be 0.5 percent of GDP over the next 75 years.
 

Figure 7.

Tax Differences from All Tax-Deferred Retirement Plans

(Percentage of GDP)

Source: Congressional Budget Office.

a. Data are normalized to zero in 2003.

How can the net flow of funds into tax-deferred retirement plans be a fairly constant percentage of GDP, while the tax difference associated with those plans is not? The answer lies in the relationship between withdrawals and investment income and the different tax treatment afforded each. In order for the net flow of funds to remain a constant share of GDP, the difference between withdrawals and investment income must also be a constant share of GDP (as contributions also remain a relatively fixed percentage of GDP). The ratio of withdrawals to investment income, however, changes over time as both flows increase as a share of GDP. Because withdrawals are initially the smaller flow, the same increase in both means that the ratio of withdrawals to investment income increases over time. Because withdrawals are taxed at a higher rate than investment income, the positive tax effect associated with withdrawals grows faster than does the negative tax effect associated with investment income--hence, the forecast of increasing revenues due to tax-deferred retirement plans.
 

Alternative Scenarios

In the base case described above, CBO made certain assumptions concerning the long-term state of the economy, the policy environment during the forecast period, and the future behavior of both individuals and firms. CBO also simulated several alternative scenarios in which it changed those assumptions (see Appendix B for full details). The purpose of the alternative scenarios is to demonstrate how CBO's conclusions about the effects of deferred taxes on federal revenues would vary under different assumed circumstances.

Economic Variables

In the base case, CBO adopted values for economic variables consistent with its long-term economic projections.⁽²²⁾ For the most part, those projections include an average value for each economic variable that applies in every year. In any given year, the economic reality will inevitably be different from the projections--sometimes dramatically. Over time, however, the errors should offset one another, leading to an outcome similar to that implied by the projections. For this analysis of deferred taxes, CBO tested some alternative assumptions about the average values of four key variables: inflation, real wage growth, interest rates, and the return on equities.⁽²³⁾

Each test involved increasing or decreasing the rates used in the base case by 1 percentage point, illustrating how an identical change in each variable affects the model's results differently. For example, a 1 percentage point increase in inflation and a 1 percentage point increase in real wage growth both imply a 1 percentage point increase in GDP, but they do not imply the same thing about receipts resulting from tax-deferred retirement plans. The sensitivity test results thus help to clarify the role of each economic variable in the model and demonstrate the robustness of the model with respect to those variables. One should not, however, extrapolate the robustness of the model to represent the robustness of the retirement system in a changing economy. The model is narrowly focused and cannot reflect all of the interactions between the retirement system and the rest of the economy.

Under any of the economic scenarios tested, a modest increase in receipts as a percentage of GDP over the next 75 years occurs (see Table 5). The results are most sensitive to changes in the return on equities and real wage growth. Furthermore, the return on equities is the only variable for which a higher value leads to higher receipts relative to GDP.

Policy Environment

The policy environment in the base case is roughly consistent with

• Marginal tax rates under JGTRRA extended indefinitely,
  
  
• Restoration of private and state and local defined-benefit plans to a fully funded status over the 2004-2014 period, and
  
  
• No Roth IRAs or equivalent 401(k) plans.

As an alternative to the JGTRRA tax rates, CBO measured the tax difference associated with tax-deferred retirement plans using tax rates consistent with those in place prior to EGTRRA (which, under current law, will become effective again in 2011). The result was a smaller fiscal improvement than in the base case (see Table 6). The increase in receipts to GDP over 75 years was only 0.26 percent instead of 0.50 percent.

As an alternative to the 10-year period for restoring nonfederal defined-benefit plans to a fully funded status, CBO also simulated periods of five years and 20 years. Those simulations demonstrate that the period of time over which full funding is restored has no effect on the long-term fiscal improvement.

Since 1998, workers and their spouses have been allowed to contribute to Roth IRAs. Under EGTRRA, 401(k) participants will also be able to contribute to back-loaded 401(k) plans beginning in 2006. As the data underlying the patterns of contributions and withdrawals in CBO's model are for 1997, they do not contain the effect of allowing back-loaded plans. Had the model accounted for Roth IRAs, the fiscal improvement over 75 years would have been 0.47 percent of GDP instead of 0.50 percent. Because contributions to defined-contribution plans are potentially much larger than to Roth IRAs, however, accounting for back-loaded 401(k)s as well would have resulted in a fiscal improvement of only 0.19 percent of GDP.

Firms' and Individuals' Behavior

The base case generally assumes that the behavior exhibited by firms and individuals in 1997 remains unchanged throughout the 75-year projection period. CBO tested alternatives that allow for different behavior in three respects: firms continuing to phase out defined-benefit plans in favor of more defined-contribution plans, individual contributions growing over time at different rates, and individuals withdrawing funds at a higher or lower rate than in 1997.

The assumption implicit in the base case that participation in private defined-contribution and defined-benefit plans remains unchanged throughout the 75-year period overlooks the trend evident since the early 1980s away from defined-benefit plans and toward defined-contribution plans. As an alternative, CBO simulated the flows into and out of retirement plans as if all nonfederal retirement assets were in defined-contribution plans. That approach represents an upper bound on how the estimates might have differed had the trend toward defined-contribution plans been incorporated into the model. The simulation shows that the results after 75 years, a 0.54 percentage point boost in revenues as a share of GDP, are quite similar to those under the base case (see Table 7).

The base case assumes that contributions to both IRAs and defined-contribution plans increase from their 1997 levels with nominal wage growth. Two alternatives that CBO simulated, one in which contributions remain at 1997 levels and one in which they increase with inflation, would result in a smaller tax difference in each year than does the base case; under those scenarios, revenues in 2078 would be 0.2 percent of GDP higher than under the base case. A third alternative, in which contributions increase in 2008 to reflect higher limits under EGTRRA then increase with wage growth, would always result in a larger tax difference than does the base case; under that scenario, revenues in 2078 would be less than 0.1 percent of GDP lower. A final alternative, in which contributions increase in 2008 to reflect higher limits under EGTRRA then increase with inflation, would have a mixed impact on the tax difference. Initially, the higher contributions would increase the negative tax difference, but by 2078, the negative tax difference would be smaller than in the base case, and revenues as a percentage of GDP would actually be comparable to those under the first two alternatives.

According to empirical evidence, people withdrew significantly different shares of their assets in IRAs and defined-contribution plans in 1998 through 2001 than they did in 1997. To illustrate the fiscal implications of such changes in behavior, CBO simulated two alternative assumptions under which participants withdrew 10 percent less and 10 percent more of their assets than participants did in 1997. In the first instance, tax receipts as a share of GDP would be one-twentieth of a percentage point higher than those in the base case; in the second instance, a similar amount lower.