The Population of Workers Covered by the Auto IRA:Trends ...€¦ · Auto IRAs with an expanded...

AARP Public Policy Institute

years

C

E LE B R AT I N G

Economic Effects of Automatic Enrollment in

Individual Retirement AccountsBenjamin H. Harris

The Brookings Institution1

Rachel M. JohnsonThe Urban Institute

1 Harris is on leave from the Brookings Institution while serving as a senior economist with the Council of Economic Advisers. He was a research economist with the Brookings Institution when this paper was written.

Research Report

Economic Effects of Automatic Enrollment in

Individual Retirement Accounts

Benjamin H. HarrisThe Brookings Institution

Rachel M. JohnsonThe Urban Institute

AARP’s Public Policy Institute informs and stimulates public debate on the issues we face as we age. Through research, analysis and dialogue with the nation’s leading experts, PPI promotes development of sound, creative policies to address our common need for economic security, health care, and quality of life.

The views expressed herein are for information, debate, and discussion, and do not neces-sarily represent official policies of AARP.

#2012-04February 2012©2012, AARPReprinting with permission only

AARP Public Policy Institute601 E Street, NW, Washington, DC 20049http://www.aarp.org/ppi

http://www.aarp.org/ppi

Acknowledgment

The authors thank Bill Gale, David John, Surachai Khitatrakun, Gary Koenig, Eric Toder, and Bob Williams for helpful comments. All errors or omissions are the authors’ own.

v

Table of Contents

Acknowledgment............................................................................................................iv

Executive Summary ........................................................................................................1Results............................................................................................................................1

Introduction......................................................................................................................3Automatic Enrollment’s Promise ..................................................................................3Reversing the “Upside-Down” Nature of Saving Incentives.........................................4Results in Brief ..............................................................................................................5

Description of the Proposals..........................................................................................7Auto IRA........................................................................................................................7Saver’s Credit.................................................................................................................7

Methodology ....................................................................................................................9

Results............................................................................................................................ 11Revenue Effects ...........................................................................................................11Distributional Effects ...................................................................................................12National Saving Effects ...............................................................................................14

Conclusion.....................................................................................................................15

References ....................................................................................................................17

Appendix: Methodology................................................................................................24Eligibility Imputation ..................................................................................................24Take-Up Rates .............................................................................................................25Contribution Levels ....................................................................................................26Choice of Account Type ..............................................................................................27

List of TablesSummary Table of Revenue, Distributional, and National Saving Effects on Automatic IRAs and Saver’s Credit Expansion ....................................................................................................... 2

Table 1. Simulation Parameters, by Scenario ............................................................................................. 9

Table 2. Revenue Loss Due to Automatic IRAs and Saver’s Credit ExpansionImpact on Tax Revenue ($ billions), 2011–20 ............................................................................ 19

Table 3. Automatic IRAs Under Low Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 20

Table 4. Automatic IRAs and Saver’s Credit Expansion Under Low Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 20

Table 5. Automatic IRAs Under Medium Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 21

Table 6. Automatic IRAs and Saver’s Credit Expansion Under Medium Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 21

Table 7. Automatic IRAs Under High Revenue Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 22

Table 8. Automatic IRAs and Saver’s Credit Expansion Under High Cost AssumptionsDistribution of Federal Tax Change by Cash Income Percentile, 2015 ...................................... 22

Table 9. Average Change in IRA Contributions Per Affected Taxpayer, 2015......................................... 23

Table 10. Aggregate Change in IRA Contributions ($Millions), 2015....................................................... 23

vi

1

Economic Effects of Automatic Enrollment in Individual Retirement Accounts

executive SummAry

Automatic enrollment in individual retirement accounts (IRAs) is an innovative proposal designed to increase retirement saving among millions of Americans. The plan—which we refer to as the “Auto IRA”—aims to improve retirement security for workers without employer-provided retirement plans. The Auto IRA would require firms without retirement savings plans to automatically enroll each worker in an IRA unless the worker opts out.

The idea behind automatic enrollment is simple: By removing administrative barriers to saving, automatic enrollment can increase the likelihood that workers contribute to their retirement. The Auto IRA takes advantage of “financial inertia” by making saving the default option if a worker takes no action, rather than requiring an explicit action to save. This policy appears to improve the saving behavior of workers who find the saving process difficult or intimidating.

Auto IRAs have the potential to influence the saving behavior of millions of households. In a companion study to this paper, Harris and Fischer (2012) estimate that between 19.7 percent (24.4 million) and 32.0 percent (43.1 million) of workers would be eligible for the Auto IRA.

Results

This study analyzes the revenue and distributional effects of recent Auto IRA proposals alone, including two congressional bills and a proposal in the administration’s FY 2012 budget, and in combination with an expanded Saver’s Credit. It also provides an estimate of the effects on national savings. The primary findings are as follows:

� The 10-year revenue loss resulting from the Auto IRA —excluding the expansion of the Saver’s Credit—is estimated to range from $3.8 billion under low-cost behavioral assumptions to $65.8 billion under high-cost assumptions.

� With the expansion of the Saver’s Credit in place, the revenue cost would rise substantially. We find the 10-year cost of the Auto IRA and an expanded Saver’s Credit to range from $63.4 billion under low-cost assumptions to $146.4 billion under high-cost assumptions.

� The gain in after-tax income resulting from the Auto IRA alone—our preferred measure of a tax reform’s progressivity—would be higher for workers in the middle three quintiles than for those in the top and bottom quintiles. For example, in the intermediate-cost scenario, those in the middle three quintiles would receive an after-tax income gain of 0.08 or 0.09 percent, while those in the top and bottom quintiles would receive an increase of 0.05 percent.

� Under the Auto IRA in isolation, we estimate that between 4.8 percent and 15.2 percent of taxpayers would receive a tax cut; essentially no taxpayers would experience a tax increase.

� Combining automatic enrollment and the Saver’s Credit expansion would benefit between 17.2 percent and 30.2 percent of taxpayers, while increasing taxes for less than 1 percent of taxpayers. The average changes in after-tax income are substantially greater for the bottom two quintiles relative to the others, and are almost zero for the

2


top quintile. These patterns indicate that the combination of automatic enrollment and expansion of the Saver’s Credit is a strongly progressive reform.

� For the taxpayers who would benefit from the Automatic IRA, the gains can be significant. Automatic enrollment in IRAs, in isolation, would increase IRA contributions by between $1,280 and $1,623 for each automatically enrolled taxpayer. The average contribution generally increases with income, but the average increase in contributions is still substantial for lower-income households—approaching $1,000 per year.

� Under the intermediate-cost scenario, the Auto IRA without an expanded Saver’s Credit would boost new contributions to retirement saving accounts by 0.14 percent of gross domestic product (GDP) ($26.9 billion) in 2015. These contributions would be offset by a 0.02 percent of GDP reduction in government revenue and other reductions in private saving, which we do not estimate. A policy that implemented Auto IRAs with an expanded Saver’s Credit would raise IRA contributions by 0.19 percent of GDP ($34.6 billion) under our intermediate-cost scenario, but would be offset by a 0.06 percent of GDP reduction in government saving, plus reduced private saving in other accounts.

These results indicate that the Auto IRA could transform the saving landscape in the United States. The policy could set millions of workers on a path toward more adequate retirement saving, particularly workers with low income and without current access to employer-sponsored retirement saving accounts. Moreover, these benefits can be achieved at low cost in terms of lost revenue relative to the billions of dollars devoted to existing retirement saving incentives. Since increased personal saving would outweigh the modest amount of lost federal revenue, this policy has the potential to moderately increase national saving. This increase would lead to a better standard of living not only for those households directly affected by the policy, but for the country as a whole.

Summary Table of Revenue, Distributional, and National Saving Effects on Automatic IRAs and Saver’s Credit Expansion

10-Year Revenue Change

($Billions)

Change in After-Tax Income (Percent), 2015

Aggregate Change in IRA Contributions

($Billions), 2015

1st Quintile

2nd Quintile

3rd Quintile

4th Quintile

5th Quintile All

Baseline: Current Policy

1. Low Revenue Cost Assumptions

-3.8 0.02 0.03 0.04 0.03 0.02 0.02 10.1

2. Low Cost Assumptions, Saver’s Credit Expansion

-63.4 0.25 0.25 0.19 0.06 0.02 0.08 16.4

3. Medium Cost Assumptions -28.2 0.05 0.09 0.09 0.08 0.05 0.06 26.9

4. Medium Cost Assumptions, Saver’s Credit Expansion

-100.8 0.35 0.36 0.27 0.11 0.05 0.14 34.6

5. High Cost Assumptions -65.8 0.06 0.12 0.13 0.11 0.07 0.09 40.5

6. High Cost Assumptions, Saver’s Credit Expansion5,7

-146.4 0.42 0.42 0.32 0.14 0.07 0.17 49.1

(1) See text and table 2 for notes and definitions.

3


introduction

Retirement income is inadequate for many U.S. workers. One recent study estimated that nearly half of workers—47.2 percent—have insufficient savings to meet their consumption needs during retirement (VanDerhei and Copeland 2010). Another study puts the percentage of Americans with inadequate savings at 43 percent and finds this proportion to be rapidly growing over time (Munnell, Webb, and Golub-Sass 2007). In response, policy makers are searching for innovative policies to increase private saving and wealth accumulation at retirement.

This study analyzes the revenue and distributional effects of recent proposals—including two congressional bills and a proposal in the administration’s FY 2012 budget—aimed at boosting national saving and improving retirement saving adequacy. These proposals center around an initiative called the Automatic IRA (Auto IRA), which would require firms without a retirement savings plan to automatically enroll their workers in individual retirement accounts (IRAs), unless the workers opt out.2

Auto IRAs have the potential to influence the saving behavior of millions of households. In a companion study to this paper, Harris and Fischer (2012) estimate that between 19.7 percent (24.4 million) and 32.0 percent (43.1 million) of workers would be eligible for the Auto IRA.

Automatic Enrollment’s Promise

The Auto IRA is an extension of an effort to encourage firms with defined-contribution plans with a cash or deferred arrangement, such as 401(k)s, to automatically enroll workers in the company plan. Although employers sponsoring 401(k) plans are not required to enroll their employees automatically, automatic enrollment in 401(k) plans is becoming increasingly popular in the pension universe; the proportion of defined-contribution plans offering automatic enrollment jumped from 17 percent in 2005 to 42 percent in 2010.3 The percentage of workers covered by automatic enrollment is even higher in large plans—those with 5,000 or more enrollees—with 54 percent offering the automatic enrollment feature (Plan Sponsor Council of America 2011).

Automatic enrollment has proven effective in bringing more workers into retirement saving programs. One study found that an automatic enrollment plan boosted initial enrollment from 37 percent to 86 percent in one firm (Madrian and Shea 2001). Other studies have noted the policy’s potential to increase aggregate saving (Iwry, Gale, and Orszag 2006), improve retirement preparedness (Gale, Iwry, and Orszag 2005a), and increase tax progressivity (Geissler and Harris 2009).

2 This paper is an update and expansion of Harris and Johnson (2009), using recent estimates of the Auto IRA-eligible population from Harris and Fischer (2012). Harris and Johnson (2009) calculated the revenue and distributional effects of the Auto IRA under a point estimate of the Auto IRA-eligible population. This paper updates earlier estimates under a new range of assumptions about the proportion of workers eligible for the Auto IRA, and adds a discussion of the potential impacts of the Auto IRA on national saving.

3 Gale, Iwry, and Walters (2007) note that the Pension Protection Act of 2006 removed many of the practical barriers to automatic enrollment, including concern among pension administrators over their liability for how contributions are invested, whether employees could demand a refund of their contributions, the application of pension nondiscrimination laws, and violation of state antigarnishment laws.

4


The idea behind automatic enrollment is simple: By removing administrative barriers to saving, automatic enrollment can increase the likelihood that workers will contribute to retirement. Taking advantage of “financial inertia,” automatic enrollment makes saving the default option if a worker takes no action, rather than requiring an explicit action to participate. Such a policy appears to be an effective way of improving the saving behavior of workers who find the process of saving difficult or intimidating. The 78 million workers without employer-sponsored retirement plans have less incentive to contribute to retirement accounts. Unlike their counterparts enrolled in employer-sponsored retirement savings plans, they do not receive an employer match for contributions. Furthermore, since the tax benefit for retirement saving is larger for taxpayers in higher tax brackets, low-income workers—who are more likely to be in firms without any form of company retirement plan—often have only modest tax incentives to save for retirement.

Reversing the “Upside-Down” Nature of Saving Incentives

For traditional IRAs or 401(k)-type retirement savings plans, tax incentives to save are related to a taxpayer’s income through the deduction or exclusion of the initial contribution from taxable income. Taxpayers contributing to these types of retirement saving accounts can usually deduct or exclude all or a portion of the contribution from taxable income; the initial tax benefit is thus the marginal tax rate multiplied by the amount deducted or excluded.4 This benefit is larger for taxpayers with higher income and higher marginal tax rates than it is for taxpayers in lower tax brackets. Hence, the tax benefit from saving in these tax-preferred accounts is often higher for wealthier taxpayers.

For taxpayers contributing to Roth IRAs or Roth 401(k)s, the tax benefit takes the form of exemption of withdrawals from taxation rather than deferral of taxation on initial contributions. Unlike traditional IRAs or 401(k)s, Roth accounts do not allow a deduction for the initial contribution. However, Roth accounts do permit the tax-free buildup of earnings on contributions (like other accounts) and, unlike traditional IRAs or 401(k)s, allow for qualified distributions to be taken tax-free. The net effect of this structure is that taxpayers with high marginal tax rates in retirement—typically higher-income individuals—have the strongest incentive to save.

Some analysts have argued that these mismatched incentives mean that IRAs are not particularly effective at stimulating additional saving (Attanasio and DeLeire 2002; Engen, Gale, and Scholz 1996; Gale and Scholz 1994), and instead primarily serve as a mechanism for wealthier taxpayers to shift saving from taxable accounts to nontaxable accounts, making IRAs a windfall for upper middle-income taxpayers. Others disagree (Hubbard and Skinner 1996; Poterba, Venti, and Wise 1996), arguing that the tax incentive is an effective way to induce additional net saving.

Regardless of their net effect on aggregate saving, it is clear that IRAs alone are a weak incentive to save for lower-income taxpayers. Partially combating this effect is the Saver’s Credit, a tax provision that gives lower-income taxpayers a tax credit for contributions to retirement accounts. The limited evidence on these types of policies has shown that they can be effective in raising saving rates among lower-income households (Duflo et al. 2006).

4 Sometimes, the contribution can push a taxpayer into a lower marginal tax bracket. In that case, the tax benefit is the average marginal tax rate on contributions times the amount of contribution.

5


Recognizing the potential to help bolster saving rates among low-income households, the administration’s 2011 budget proposed changes to the Saver’s Credit that would make it simpler, fully refundable, and more available to middle-income households.5 The proposal would also lower the maximum contribution eligible for the matching tax credit and raise the match rate for some households. In combination, these proposals would make the saving incentive from the Saver’s Credit more like employers’ matching contributions in 401(k) plans; we explain the details more fully in the next section.

The expansion of the Saver’s Credit and the implementation of the Auto IRA complement each other. Expanding the Saver’s Credit would generally increase the tax incentives to save, while automatic enrollment in IRAs increases the number of workers with an easy way to do so. For this reason, it is important to consider the two policies in conjunction with one another.

Results in Brief

We estimate that requiring businesses to offer automatic enrollment in IRAs would reduce federal revenues by between $3.8 billion and $65.8 billion over 10 years, depending on various factors, most notably whether individuals take up IRAs in the form of traditional or Roth IRAs. Adding the cost of expanding the Saver’s Credit increases the 10-year revenue cost to between $63.4 billion and $146.4 billion. Our intermediate 10-year revenue estimate for the Auto IRA is $28.2 billion (a reduction of federal revenues of only 0.07 percent) without an expanded Saver’s Credit and $100.8 billion with an expanded Saver’s Credit (a reduction of federal revenues of only 0.26 percent).

These revenue losses are small in relation to total tax expenditures for retirement saving. Tax expenditures for retirement saving represent the lost tax revenue (i.e., the tax benefit of the exclusion or deduction and the deferral of tax on earnings net of taxes paid on distributions) that results from the preferential treatment of retirement saving. For example, the Joint Committee on Taxation (2010) estimates that tax expenditures for retirement saving will amount to $170.0 billion in 2014. Under the intermediate-cost assumptions, the cost of the Auto IRA in isolation represents just 1.6 percent of the tax expenditure cost for retirement saving, and the Auto IRA combined with the Saver’s Credit expansion represents just 6.4 percent of the tax expenditure cost.6

The revenue estimates are, to a certain extent, necessarily artificial, since they are constrained to a 10-year horizon.7 Both Roth and traditional IRA accounts alter the realization of tax benefits substantially after the 10-year window, although the change is markedly different for Roths compared to traditional IRAs.8 The revenue costs of

5 The expansion of the Saver’s Credit was not included in the administration’s FY 2012 budget. 6 Tax expenditures estimates are not directly comparable to revenue estimates, since tax expenditures

do not take into account particular behavioral responses. Still, we offer this comparison to illustrate the small magnitude of the costs of the Auto IRA policy relative to the costs of other retirement saving provisions.

7 Revenue estimates produced by the Congressional Budget Office and the Joint Committee on Taxation are typically presented within the 10-year time frame as well.

8 See Burman, Gale, and Weiner (2001) for a discussion of the difference between front- and back-loaded accounts.

6


Roth IRAs are “back-loaded,” or realized after a worker has reached retirement age; this is due to the nontaxation of distributions from a Roth account. Conversely, the costs of traditional IRAs are “front-loaded” because the deduction is realized in the year of contribution, but distributions from the account are taxed. This difference in timing accounts for the dramatically higher cost of the policy under the assumption that most workers take up traditional IRAs.

The benefits of the Auto IRA alone would be more pronounced for taxpayers in the middle three quintiles, with lesser benefits for the top and bottom quintiles. Under our intermediate-cost scenario, about 4 percent of taxpayers in the bottom quintile and 12 to 16 percent of those in the other four quintiles would get a tax reduction. After-tax incomes would increase more for taxpayers in the middle three quintiles relative to other groups, although the average change in taxes is small for all groups.

The combined distributional effects of the Auto IRA and the Saver’s Credit would be more progressive. Under our intermediate-cost scenario, the policies would increase relative after-tax incomes for the bottom three quintiles by about 0.3 percent, while the top two quintiles would increase between 0.05 to 0.11 percent. Approximately one-third of taxpayers in the second and third quintiles would benefit from the policy, with smaller proportions benefiting in other quintiles. A few taxpayers, less than 1 percent, would see their after-tax incomes fall because of changes in the maximum value of the Saver’s Credit.

For the taxpayers who would benefit from the Auto IRA and expansion of the Saver’s Credit, the gains can be significant. Automatic enrollment in IRAs, in isolation, would increase IRA contributions by between $1,280 and $1,623 for each automatically enrolled taxpayer. A policy that included an expanded Saver’s Credit in conjunction with automatic enrollment would increase average contributions even more.

These changes would represent a prominent shift in the national saving landscape. Between 7.9 million and 24.9 million additional taxpayers would have retirement saving accounts to which they regularly contributed. This increase in active saving behavior could have spillover effects to other workers, generating additional gains in contributions.9 Moreover, for those households brought into the retirement saving universe by automatic enrollment, the gains from saving could extend beyond increased retirement security; prior studies have found that asset accumulation—particularly among low-income households—can lead to changes in self-confidence, goal-setting, and community participation (Grinstein-Weiss and Irish 2007).

National saving would increase. Our intermediate-cost estimates of the Auto IRA without an expanded Saver’s Credit show additional contributions to retirement saving accounts amounting to 0.14 percent of GDP ($26.9 billion) in 2015. These contributions would be offset by a reduction to government revenue equal to 0.02 percent of GDP and

9 Prior studies have found evidence of spillover effects for worker attitudes toward retirement saving. For example, Duflo and Saez (2003) conducted an experiment by sending a letter to randomly selected university employees in randomly selected departments inviting them to attend a retirement fair; the letter informed recipients that they would be modestly compensated for attendance. The payment had a significant impact on attendance—28 percent of employees receiving a letter attended, compared to only 5 percent in departments where no one received a letter offering compensation. Importantly, Duflo and Saez observed a “social networking” effect, as evidenced by a 15 percent attendance rate for employees who did not receive a letter but were in the same department as someone who did.

7


offsets to other private saving, which we do not estimate. A policy that implemented Auto IRAs with an expanded Saver’s Credit would raise IRA contributions by 0.19 percent of GDP under our intermediate-cost scenario, but would be offset by a 0.06 percent of GDP reduction in government saving, plus reduced private saving in other accounts.

The proposals are described in the following section. The methodology and results are presented in subsequent sections, followed by conclusions.

deScription of the propoSAlS

Auto IRAThere are multiple proposals for automatic enrollment in IRAs. This paper focuses

on two related proposals introduced in the U.S. Senate and the House of Representatives. In the Senate, Senator Jeff Bingaman (D-NM) introduced the Automatic IRA Act of 2010 (S. 3760),10 which would require most private firms without employer-sponsored retirement plans to automatically enroll employees in a Roth IRA; to automatically direct a preset proportion of the employee’s wages to the IRA on behalf of the employee; and to automatically invest these contributions in a designated low-cost investment.

The Senate version of the legislation does not require all workers to be automatically enrolled in an IRA. Firms that offer retirement plans are generally exempt from the requirement, but firms that exempt particular divisions from the retirement plan are required to automatically enroll ineligible workers in an IRA. Other classes of exempt workers include those who have not satisfied the minimum age and job tenure requirements; government workers and church employees; employees with fewer than three months of service at a company; employees of newly established firms; and workers under 18 years of age. In the first year of enactment, firms with fewer than 100 employees are exempt from the requirement; this exemption is gradually phased down over four years to firms with fewer than 10 employees. For these purposes, an “employee” is counted as a worker earning more than $5,000 in annual wages.

A nearly identical companion bill with the same title (H.R. 6099) was introduced in the House by Rep. Richard Neal (D-MA). With respect to automatic enrollment, the primary difference between the bills is the default account type for eligible workers. The default account type in the House bill is a traditional IRA, while the default account type in the Senate bill is a Roth IRA. The bills also differ on the threshold for the size of company subject to the provisions. While the Senate bill gradually reduces the minimum size of the firm subject to the requirement from 100 to 10, the House bill immediately requires firms with 10 or more employees to automatically enroll workers in an Auto IRA.11

Saver’s CreditThe Saver’s Credit—enacted in 2001 and made permanent in 2006—provides a

nonrefundable tax credit for contributions to retirement saving accounts, such as 401(k)s or other employer-provided plans, IRAs, Simplified Employee Pension (SEP) plans,

10 On September 14, 2011, Senator Bingaman reintroduced the Auto IRA bill as S. 1557, the Automatic IRA Act of 2011.

11 For the purposes of measuring the proportion of workers exempt from automatic enrollment, we ignore the phase-in provision in the Senate bill.

8


and Savings Incentive Match Plan for Employees (SIMPLE). The maximum eligible contribution allowed by the Saver’s Credit is $2,000 per individual, or $4,000 for married couples filing jointly.

The rate at which tax benefits accrue varies by adjusted gross income (AGI) level. In 2009, married couples with an AGI up to $33,000 received a 50 percent credit; couples with an AGI between $33,301 and $36,000 received a 20 percent credit; and couples with an AGI between $36,001 and $55,500 received a 10 percent credit. The corresponding thresholds for heads of household and single taxpayers are reduced by 25 percent and 50 percent of the married thresholds, respectively. The 50 percent credit corresponds to a 100 percent after-tax match rate, while the 20 percent credit corresponds to a 25 percent after-tax match rate. The 10 percent credit corresponds to just an 11 percent after-tax match rate.12

The administration’s fiscal year 2011 budget proposal would expand the existing Saver’s Credit to make it refundable, simplify the matching structure, and improve the incentives for making retirement contributions. The combined set of reforms would make the incentives for contributing to a retirement account for low- and middle-income taxpayers more analogous to the incentives for upper-income taxpayers.

The administration proposed to make the credit fully refundable. Under the existing system, the credit is nonrefundable, meaning that the Saver’s Credit can only benefit taxpayers to the extent that it will reduce income tax liability to zero. For the 65 million taxpayers with no income tax liability, the Saver’s Credit provides little or no incentive to contribute to a retirement account; many more taxpayers receive only a partial benefit because their incentive is greater than their tax liability.

The administration’s proposal would simplify the existing three-tier matching structure to a 50 percent credit on eligible contributions up to a maximum benefit of $250 per individual. Thus, a $1,000 contribution would generate a $500 benefit for low- and moderate-income married taxpayers. This change in structure increases the maximum potential benefit for some while reducing it for others. In addition, workers who participate in an Auto IRA or 401(k)-type retirement savings plan would get their Saver’s Credit automatically deposited into their retirement saving account.

Lastly, the administration’s proposal would expand the eligibility limits for taxpayers. The new eligibility limits for the maximum 50 percent match would rise from $33,300 to $65,000 for married couples filing jointly, from $24,750 to $48,750 for heads of households, and from $16,500 to $32,500 for single taxpayers.13 These limits would propel the Saver’s Credit from a provision that primarily affected low-income households to one that benefits both low- and middle-income households.

12 For example, Gale, Iwry, and Orszag (2005b) show that a taxpayer with AGI in a range that would qualify contributions for a 10 percent credit might contribute $2,000 to an account and receive a $200 credit. Thus, the $1,800 after-tax contribution would generate a credit of $200, or 11 percent (i.e., $200 ÷ $1,800).

13 The amount of contributions eligible for the 50 percent match would be phased out at a 5 percent rate of AGI over these thresholds, indicating that the phase-out range for the match would be $20,000 over the respective thresholds for each type of filing status.

9


methodology

This study utilizes the Urban-Brookings Tax Policy Center (TPC) microsimulation model to derive revenue and distributional estimates of the Auto IRA and Saver’s Credit proposals. 14 Modeling the revenue and distributional estimates is a two-step process. First, we impute the proportion of workers who would be eligible for the Auto IRA; these estimates were derived in Harris and Fischer (2012). Second, given imputed eligibility, we utilize the TPC microsimulation model to produce estimates of the change in tax liability due to the proposals. This procedure is described briefly below and in detail in the appendix.

To account for variation in model parameters, we calculate low-, intermediate-, and high-cost scenarios. Parameter values that differ between scenarios include the take-up rate for eligible workers, the proportion of workers eligible for the Auto IRA, the proportion of workers investing in Roth IRAs versus traditional IRAs, and the contribution rates among participating workers. The parameter values for each scenario are shown in table 1.

The TPC tax model uses two data sources: the 2004 public-use file (PUF) produced by the Statistics of Income (SOI) Division of the Internal Revenue Service, and the 2005 Current Population Survey (CPS). The PUF contains 150,047 income tax records with detailed information from federal individual income tax returns filed in the 2004 calendar year. It provides key data on the level and sources of income and deductions, income tax liability, marginal tax rates, and use of particular credits, but it excludes most information about pensions and IRAs, as well as demographic information such as age. Additional information is mapped onto the PUF through a constrained statistical match with the March 2005 CPS of the U.S. Census Bureau.15

14 Rohaly, Carasso, and Saleem (2005) provide a complete documentation of the TPC model.15 The statistical match provides important information not reported on tax returns, including measures of

earnings for head and spouse separately, their ages, the ages of their children, and transfer payments. The statistical match also generates a sample of individuals who do not file income tax returns (“nonfilers”). By combining the data set of filers with the data set of estimated nonfilers from the CPS, we are able to carry out distributional analysis on the entire population rather than just the subset that files individual income tax returns.

Table 1 Simulation Parameters, by Scenario

Low CostIntermediate

Cost High CostYear of Simulation 2015 2015 2015Baseline Current Policy Current Policy Current PolicyMean Take-Up Rate by Eligible Workers 31.0% 46.5% 62.0%Imputed Eligibility 19.7% 32.0% 32.0%Roth IRAs (as a Share of Total Auto IRAs) 90% 50% 10%Regular Contribution Rates 3.0% 3.0% 3.0%High-Income Contribution Rates 3.0% 3.75% 4.5%

10


To model retirement saving incentives, we supplement the PUF and CPS data described above with information from the 2004 Federal Reserve Board of Governors’ Survey of Consumer Finances (SCF) and the Survey of Income and Program Participation (SIPP). Our principal data source for type of pension, pension participation, and contributions by employers and employees is the SCF, a stratified sample of about 4,400 households with detailed data on wealth and savings. The SCF has the best available data on pensions for a broad cross-section of the population, but does not report enough information to determine eligibility for deductible IRA contributions; for this we use data from the SIPP.16

Imputations for Auto IRA eligibility were based on estimates from Harris and Fischer (2012). Using data from the SCF and the CPS, Harris and Fischer estimated that between 19.7 percent (24.4 million) and 32.0 percent (43.1 million) of workers would be eligible for the Auto IRA. We imputed Auto IRA eligibility so that the proportion of eligible workers equaled 19.7 percent in the low-cost scenario and 32.0 percent in the intermediate- and high-cost scenarios.

Modeling the revenue and distributional effects of the Automatic IRA also requires making particular assumptions about taxpayer behavior, including assumptions about the proportion of taxpayers who take up the IRA after being automatically enrolled (i.e., the proportion who do not opt out), the contribution levels of those workers who remain enrolled, and the proportion of workers choosing a traditional IRA versus a Roth IRA.

These behavioral assumptions are important factors in the results. The assumption concerning the proportion of taxpayers who do not opt out of automatic enrollment is subject to a high degree of uncertainty, given the lack of precedent. Prior studies have measured the behavior of individuals after being automatically enrolled in a 401(k) plan. For example, Madrian and Shea (2001) show that automatic enrollment dramatically increases the proportion of workers participating in a company retirement saving plan, while Beshears et al. (2009a) show that this effect continues to hold even in retirement plans without a company match, although at a diminished rate. We use these results as a guide for our assumptions.17

Previous research suggests how workers might contribute to an IRA following automatic enrollment. Prior studies have found that workers frequently contribute the default contribution rate, and that high-income workers are the group most likely to contribute more than the default rate (Beshears et al. 2009b). Following this pattern, we assume that all low- and middle-income workers contribute the proposed default contribution rate of 3 percent included in the congressional bills. In our low-cost scenario, we assume that high-income workers also contribute the default rate, while in the high-cost scenario we assume that high-income workers contribute 1.5 times the default rate.

16 Burman et al. (2004) provide a more complete description of the data and methods used in modeling the revenue and distributional effects of retirement saving accounts.

17 We assume that the absence of an employer match in Auto IRAs will increase opt-out rates relative to Automatic 401(k)s (while controlling for income and other demographic characteristics). See the appendix for more details.

11


Lastly, the assumption concerning account-type choice (i.e., traditional or Roth IRA) drives the revenue results (but not the distributional estimates18). Since there is no analogous choice for automatic enrollment in 401(k)s,19 it is difficult to use prior studies to predict how workers might choose between traditional and Roth IRAs. Since workers tend to adhere to the default option, it is likely that the account-type choice would largely be determined by the default option specified by law. To account for potential variation in this factor—the default account type differs across the House and Senate bills—we vary the proportion of taxpayers taking up a Roth versus a traditional IRA across our low-, intermediate-, and high-cost scenarios.

reSultS

In this section we present revenue, distributional, and national saving effects of the Auto IRA and Saver’s Credit expansion. The results are presented in detail in tables 2 through 10.

Revenue EffectsWe estimate the 10-year revenue losses resulting from the Auto IRA excluding the

expansion of the Saver’s Credit to range from $3.8 billion for the low-cost estimate to $65.8 billion under high-cost assumptions (see table 2); the intermediate 10-year cost is $28.2 billion. With the expansion of the Saver’s Credit, the revenue cost would rise substantially. We find the 10-year cost of the Auto IRA and an expanded Saver’s Credit to range between $63.4 billion under low-cost assumptions and $146.4 billion under high-cost assumptions. The intermediate cost of the Auto IRA and expanded Saver’s Credit is $100.8 billion.

Two factors primarily drive the 10-year revenue costs of automatic enrollment in IRAs: behavioral assumptions concerning workers eligible for the Auto IRA, and whether the administration’s Saver’s Credit expansion is in place. The most important factor affecting behavioral assumptions is workers’ choice between traditional and Roth IRAs. Taxpayers currently contribute to Roth and traditional IRA accounts in roughly equal proportions, but these taxpayers all actively established accounts, and this precedent is not necessarily a sound predictor of behavior under automatic enrollment. We expect that under automatic enrollment, account-type choice would be largely dependent on which account is statutorily designated as the default option.

The importance of account type in revenue estimates results from the structure of the Roth and traditional IRAs. As described earlier in the text, the tax preference for traditional IRAs is realized earlier than for Roth IRAs. These differing treatments mean that more of the revenue cost of traditional IRAs is borne up front, while the lost tax revenue from Roth IRAs is deferred into the future. In addition, since changing the tax treatment of retirement savings can affect tax revenue decades into the future, the 10-year

18 To compare the distributional effects of Roth and traditional IRAs, the TPC model calculates the present value of IRA benefits for both Roth and traditional accounts. Such a calculation requires several assumptions, including years until retirement and annual rate of return on assets. See Burman et al. (2004) for more detail.

19 While Roth 401(k) plans do exist, very few if any plans with automatic enrollment use them as the default account (Plan Sponsor Council of America 2011).

12


scenarios presented here do not capture the total revenue effect of the Auto IRA and Saver’s Credit expansion.

Revenue estimates are also sensitive to the take-up rate (i.e., the proportion of workers who do not “opt out” of the plan) and the contribution rate among participants. Our low-cost assumption assumes lower take-up rates and lower contribution rates relative to the high-cost scenario. Both factors affect how much workers contribute to retirement accounts, which in turn affects the aggregate amount of income exempt from taxation.

Including the costs associated with the expansion of the Saver’s Credit raises the revenue loss in each scenario substantially and consequently reduces the relative difference between the low-cost and high-cost estimates. Part of the cost increase is due to the structure of the Saver’s Credit expansion, which makes the credit refundable, increases eligibility thresholds, and indexes the credit amounts. Since the credit is targeted toward lower-income individuals who tend to have little or no tax liability, making it refundable increases the effective match rate for participants.

To be clear, the revenue effects of a policy that introduces the Auto IRA and expands the Saver’s Credit can be decomposed into three parts: the revenue lost under the Auto IRA in isolation, the revenue effects of the expanded Saver’s Credit in isolation, and the revenue effects of the interaction of the two policies.20 We are not aware of a revenue score for the president’s proposed expanded Saver’s Credit in isolation, but the score is likely to comprise a substantial portion of the total cost of the combined proposal.21

Distributional EffectsSeveral factors affect the progressivity of automatic enrollment. The policy is

progressive because it incorporates workers without employer-sponsored retirement plans into tax-preferred accounts, and these workers tend to have less income than workers with a workplace retirement account. The Auto IRA also has a regressive aspect, in that higher-income workers benefit more from IRAs since their initial deduction is worth more because of their high tax bracket.22 Similarly, while the Saver’s Credit is generally considered a progressive measure since it typically benefits savers with lower incomes, the administration’s proposed restructuring of the credit rate and maximum contribution match means that a small proportion of taxpayers will receive fewer benefits than under current law.23

20 This paper presents revenue estimates for the first part (i.e., Auto IRA in isolation) and all three factors combined.

21 Gale, Iwry, and Orszag (2005b) provide revenue estimates of various Saver’s Credit expansions, but none perfectly match with the president’s proposed expansion. The U.S. Treasury (2010) offers a combined revenue estimate for the expanded Saver’s Credit in conjunction with the Auto IRA, but not for the expanded Saver’s Credit in isolation.

22 Toder, Harris, and Lim (2011) show that retirement saving tax expenditures, including IRAs and defined-contribution plans, are regressive overall.

23 The administration’s proposal would increase the matching rate for all contributions to 50 percent and would raise the eligibility thresholds for matching contributions, but would also reduce the maximum contribution eligible for a match. Most taxpayers benefiting from the Saver’s Credit would receive a net tax cut under the revised parameters, but some taxpayers would receive a net tax increase because of the lower maximum contribution eligible for a match (i.e., the maximum tax credit would be reduced from $1,000 to $500).

13


The distributional effects are less sensitive to behavioral assumptions than revenue estimates because we measure tax benefits on a present-value basis; this treatment equalizes the value of equal after-tax contributions to a Roth or traditional IRA. In sharp contrast, including the expansion of the Saver’s Credit in our simulation alters the distributional effects substantially.

Several factors underlie the patterns of the distributional effects under the various scenarios. (1) Middle-income taxpayers tend to be more likely than upper-income taxpayers to work at jobs without retirement benefits, and thus are more likely to be eligible for the Auto IRA. (2) Middle- and upper-income taxpayers tend to benefit from the initiative more than low-income taxpayers because they tend to have substantial positive tax liability, while low-income taxpayers have negative or low income tax liability. (3) Higher-income taxpayers tend to benefit more from automatic enrollment because they are expected to accrue higher contributions because of their higher wage levels (although this effect is tempered by contribution limits) and they tend to have higher marginal tax rates. (The latter effect is mitigated by the existing Saver’s Credit, which rewards contributions by moderate-income taxpayers with a partial match.) The net effect of these combined factors would be a policy that benefits middle-income taxpayers slightly more than those at the top and bottom of the income distribution.24

Combining the Auto IRA with expansion of the Saver’s Credit would greatly increase the benefits for the bottom three quintiles compared to the Auto IRA proposal alone. High-income taxpayers would receive virtually nothing from the expansion of the Saver’s Credit, while middle-income taxpayers would generally benefit from the credit’s higher eligibility limits and higher matching rates. Taxpayers in the bottom two quintiles, however, would benefit the most from making the credit refundable. As a result, these two groups experience the largest increase in after-tax income relative to the Auto IRA proposal alone and compared to other income groups.

To approximate the proportion of eligible workers contributing to an Auto IRA, we assume constant take-up rates across scenarios. In practice, expanding the Saver’s Credit could affect take-up rates by changing saving incentives. Expanding the Saver’s Credit would both magnify the impact of automatic enrollment for middle-income taxpayers and generate other benefits for those taxpayers who would have contributed to a retirement account even without auto-enrollment. At the same time, some taxpayers would be net losers under these policies since the expanded Saver’s Credit would reduce the maximum tax credit for certain taxpayers, but the effect would be modest; less than 1 percent of taxpayers would experience a tax increase.

When modeling automatic enrollment without the Saver’s Credit expansion, we estimate that between 4.8 percent (table 3) and 15.2 percent (table 7) of taxpayers would receive a tax cut; essentially no taxpayers would experience a tax increase. Substantially

24 These results differ slightly from the results obtained for automatic enrollment in 401(k)s. Under the automatic 401(k), Geissler and Harris (2009) found that the bottom four quintiles experienced greater increases in after-tax income than the top quintile; this result was driven by the already high 401(k) participation rates among higher-income workers. Thus, the incremental effect of the automatic 401(k) was limited for the group that already had high rates of participation, although the participation effect was balanced against the increased benefit for high-income taxpayers owing to their higher marginal tax rate. The increased participation effect is mitigated in the Auto IRA proposal, since we assume take-up rates equal to 40 percent to 80 percent of those under an automatic 401(k).

14


more taxpayers in the top four quintiles would receive a tax cut than those in the bottom quintile. The gain in after-tax income—our preferred measure of a tax reform’s progressivity—would be higher for those in the middle three quintiles than for those in the top and bottom quintiles. For example, in the intermediate-cost scenario, taxpayers in the middle three quintiles would receive an after-tax income gain of 0.08 or 0.09 percent, while those in the top and bottom quintiles would receive a gain of 0.05 percent (table 5).

Combining automatic enrollment and the Saver’s Credit expansion would benefit between 17.2 percent (table 4) and 30.2 percent (table 8) of taxpayers, while increasing taxes for less than 1 percent of taxpayers. Middle-income taxpayers in the second and third quintiles are more likely to benefit than those in other quintiles. In particular, fewer taxpayers in the top quintile would receive a tax cut. The average changes in after-tax income are substantially greater for the bottom two quintiles relative to the others, and are almost zero for the top quintile. For example, in the intermediate-cost scenario, taxpayers in the first two quintiles would see their after-tax incomes rise by about 0.35 percent, while taxpayers in the fourth and fifth quintiles would see their after-tax incomes rise by just 0.11 percent and 0.05 percent, respectively. These patterns indicate that the combination of automatic enrollment and expansion of the Saver’s Credit is a strongly progressive reform.

For the taxpayers who would benefit from the Auto IRA and expansion of the Saver’s Credit, the gains can be significant. Automatic enrollment in IRAs, in isolation, would increase IRA contributions by between $1,280 and $1,623 for each taxpayer (table 9). The average contribution generally increases with income, but the average increase in contributions is still substantial for lower-income households—approaching $1,000.25

National Saving EffectsDetermining the net effect of a tax incentive on saving is a challenging exercise.

For tax-preferred savings accounts such as IRAs, the net effect on national saving is the gross saving within the account adjusted for the private saving offset (i.e., saving shifted from taxable accounts to IRAs) and the reduction in government saving (i.e., lower tax receipts). This paper provides estimates of both the gross change in saving resulting from new contributions and the reduction in government revenue, but does not estimate the offset in private saving.

Iwry, Gale, and Orszag (2006) provide “rough, ballpark estimates” of the effect of the Auto IRA on national saving. They find that new contributions to IRAs through automatic enrollment would amount to 0.11 percent of GDP in the long run, but that the reduction to other private savings would amount to 0.04 percent of GDP. Government revenue would reduce saving by another 0.02 percent of GDP, making the net contribution to GDP approximately 0.06 percent of GDP.

Our intermediate estimates of the saving effects of Auto IRAs closely match those of Iwry, Gale, and Orszag. The Auto IRA without an expanded Saver’s Credit would boost

25 Under a policy that included an expanded Saver’s Credit in conjunction with automatic enrollment, the average change in contributions would be less as a result of the higher number of taxpayers affected by the Saver’s Credit expansion, but not automatic enrollment. Still, the average change in contributions is significant; affected taxpayers would see their average contribution increase by between $541 and $972. The average change in contributions for those exclusively affected by automatic enrollment would exceed the $1,280 to $1,623 change cited above, but the exact change is not modeled here.

15


new contributions to retirement saving accounts by 0.14 percent of GDP ($26.9 billion) in 2015 (table 10).26 These contributions would be offset by a 0.02 percent of GDP reduction in government revenue and other reductions in private saving, which we do not estimate. A policy that implemented Auto IRAs with an expanded Saver’s Credit would raise IRA contributions by 0.19 percent of GDP ($34.6 billion) under our intermediate-cost scenario, but would be offset by a 0.06 percent of GDP reduction in government saving, plus reduced private saving in other accounts.

concluSion

We estimate the economic effects of recent congressional proposals that would require most firms without retirement plans to automatically enroll their workers in IRAs, referred to as the Auto IRA. Our estimates are presented under a scenario in which the Auto IRA is implemented in isolation, and also under a scenario in which the Automatic IRA is coupled with an expanded Saver’s Credit. Due to uncertainty about worker response to such a plan, we model the effects of the policy under low-, intermediate-, and high-cost assumptions.

We find substantial variation in revenue costs resulting from the impact of taxpayers taking up traditional versus Roth IRAs. Under our intermediate-cost scenario, which assumes that taxpayers take up traditional and Roth IRAs in equal proportions, the 10-year revenue cost of the Auto IRA is $28.2 billion. The cost of this policy rises to $100.8 billion if the Saver’s Credit is expanded concurrently with the Auto IRA. These revenue estimates should be interpreted carefully since the total revenue impact of the Auto IRA is expected to last well beyond the 10-year window presented in this paper.

The benefits of the Auto IRA in isolation would be more heavily directed toward taxpayers in the middle-income quintiles, with smaller proportional benefits going to taxpayers at the top and bottom of the income distribution. If the Saver’s Credit is expanded along with the Auto IRA, the policy becomes decidedly progressive, with larger proportional gains going to low-income taxpayers.

For taxpayers brought into the retirement saving universe through the Auto IRA, the increase in contributions could be significant. Average contribution increases for automatically enrolled workers amount to between $1,300 and $1,600; an expanded Saver’s Credit would increase these contributions even further.

The Auto IRA has the potential to boost national saving. Gross new contributions to IRAs would represent 0.11 percent of GDP if the Auto IRA is implemented without the Saver’s Credit and 0.19 percent of GDP if the Saver’s Credit is expanded. This gross increase in private saving would be offset by losses in government revenue and reduced private saving.

These calculations embody a high level of uncertainty owing to several unknown responses by participants, including the proportion of workers who opt not to contribute to a plan, the contribution rates of those who do participate, and the extent to which workers choose back-loaded versus front-loaded type plans. As research sheds light on the behavioral effects of automatic enrollment, we expect to gain precision in our estimates of the policies examined in this paper.

26 The Congressional Budget Office (2010) estimates nominal GDP in 2015 to be $18.6 trillion.

16


Despite these caveats, the Auto IRA is a promising initiative that could transform the saving landscape in the United States. The policy could set millions of workers on a path toward more adequate retirement saving, particularly workers with low income and currently without access to employer-sponsored retirement saving accounts. Moreover, these benefits can be achieved at a low cost in terms of lost revenue relative to the billions of dollars devoted to existing retirement saving incentives. Since increased personal saving would outweigh the modest amount of lost federal revenue, this policy has the potential to moderately increase national saving. This increased national saving would lead to a better standard of living not only for those households directly affected by the policy, but for the country as a whole.

17


referenceS

Attanasio, Orazio P., and Thomas DeLeire. 2002 (July). “The Effect of Individual Retirement Accounts on Household Consumption and National Saving.” The Economic Journal 112, no. 481: 504–8.

Beshears, John, James J. Choi, David Laibson, and Brigitte C. Madrian. 2009a. “The Impact of Employer Matching on Savings Plan Participation under Automatic Enrollment.” In Research Findings in the Economics of Aging, ed. David A. Wise. Chicago, IL: University of Chicago Press.

Beshears, John, James J. Choi, David Laibson, and Brigitte C. Madrian. 2009b. “The Importance of Default Options for Retirement Savings Outcomes: Evidence from the United States.”In Social Security Policy in a Changing Environment, ed. Jeffrey R. Brown, Jeffrey R. Liebman, and David A. Wise. Chicago, IL: University of Chicago Press.

Burman, Leonard E., William G. Gale, Matthew Hall, and Peter R. Orszag. 2004. “Distributional Effects of Defined Contribution Plans and Individual Retirement Accounts.” Tax Policy Center Discussion Paper No. 16. Washington, DC: Tax Policy Center

Burman, Leonard E., William G. Gale, and David Weiner. 2001. “The Taxation of Retirement Saving: Choosing Between Front-Loaded and Back-Loaded Options.” National Tax Journal 54, no. 3: 689–702.

Congressional Budget Office. 2010. The Budget and Economic Outlook: An Update. Washington, DC: Congressional Budget Office.

Duflo, Esther, William G. Gale, Jeffrey Liebman, Peter R. Orszag, and Emmanuel Saez. 2006 (November). “Saving Incentives for Low- and Middle-Income Families: Evidence from a Field Experiment with H&R Block.” Quarterly Journal of Economics 121, no. 4: 1311–46.

Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118, no. 3: 815–42.

Engen, Eric M., William G. Gale, and John Karl Scholz. 1996. “The Illusory Effects of Saving Incentives on Saving.” Journal of Economic Perspectives 10, no. 4: 113–38.

Gale, William G., J. Mark Iwry, and Peter R. Orszag. 2005a. “The Automatic 401(k): A Simple Way to Strengthen Retirement Saving.” Retirement Security Project Policy Brief 2005-1. www.brookings.edu/~/media/Files/rc/papers/2005/03saving_gale/20050228_401k.pdf.

Gale, William G., J. Mark Iwry, and Peter R. Orszag. 2005b. “The Saver’s Credit: Expanding Retirement Savings for Middle- and Lower-Income Americans.” Retirement Security Project Policy Brief 2005-2. www.brookings.edu/~/media/Files/rc/papers/2006/11_proposal_effects/11_proposal_effects.pdf.

Gale, William G., J. Mark Iwry, and Spencer Walters. 2007. “Retirement Saving for Middle- and Lower-Income Households: The Pension Protection Act of 2006 and the Unfinished Agenda.” Retirement Security Project Policy Brief 2007-1. www.brookings.edu/papers/2007/04_pension_protection_gale_iwry.aspx.

Gale, William G., and John Karl Scholz. 1994 (December). “IRAs and Household Saving.” American Economic Review 84, no. 5: 1233–60.

Geissler, Christopher, and Benjamin H. Harris. 2009. “The Automatic 401(k): Revenue and Distributional Estimates.” In Automatic: Changing the Way America Saves, ed. William G. Gale, J. Mark Iwry, David John, and Lina Walker. Washington, DC: Brookings Institution Press.

http://www.brookings.edu/~/media/Files/rc/papers/2005/03saving_gale/20050228_401k.pdf

http://www.brookings.edu/~/media/Files/rc/papers/2006/11_proposal_effects/11_proposal_effects.pdf

http://www.brookings.edu/~/media/Files/rc/papers/2006/11_proposal_effects/11_proposal_effects.pdf

www.brookings.edu/papers/2007/04_pension_protection_gale_iwry.aspx

www.brookings.edu/papers/2007/04_pension_protection_gale_iwry.aspx

18


Grinstein-Weiss, Michal, and Katie Irish. 2007. “Individual Development Accounts: Frequently Asked Questions.” Center for Social Development Perspectives no. 07-09. St. Louis, MO: Washington University Center for Social Development.

Harris, Benjamin H., and Rachel M. Johnson. 2009. “Automatic Enrollment in Individual Retirement Accounts: Revenue and Distributional Estimates.” Tax Notes 124, no. 9: 903–14.

Harris, Benjamin H., and Christopher Geissler. 2008. “Taxpayer Eligibility for IRAs.” Tax Notes 118, no. 8: 739.

Harris, Benjamin H,. and Ilana Fischer. 2012. The Population of Workers Covered by the Auto IRA: Trends and Characteristics. Washington, DC: AARP.

Hubbard, R. Glenn, and Jonathan S. Skinner. 1996. “Assessing the Effectiveness of Saving Incentives.” The Journal of Economic Perspectives 10, no. 4: 73–90.

Iwry, J. Mark, William G. Gale, and Peter R. Orszag. 2006. “The Effects of Retirement Security Project Proposals on Private and National Saving: Exploratory Calculations.” Retirement Security Project Policy Brief 2006-2. www.brookings.edu/papers/2006/11_proposal_effects.aspx.

Joint Committee on Taxation. 2010. “Estimates of Federal Tax Expenditures for Fiscal Years 2010–2014.” Joint Committee on Taxation Document JCS 03-10. Washington, DC: Joint Committee on Taxation.

Madrian, Brigitte C., and Dennis F. Shea. 2001 (November). “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior.” The Quarterly Journal of Economics 116, no. 4: 1149–87.

Munnell, Alicia H., Anthony Webb, and Francesca Golub-Sass. 2007. “Is There Really a Retirement Savings Crisis? An NRRI Analysis.” Boston College Center for Retirement Research Brief 7-11. Boston, MA: Boston College Center for Retirement Research.

Poterba, James M., Steven F. Venti, and David A. Wise. 1996. “How Retirement Saving Programs Increase Saving.” The Journal of Economic Perspectives 10, no. 4: 91–112.

Plan Sponsor Council of America. 2011. “54th Annual Survey of Profit Sharing and 401(k) Plans.” Chicago, IL: Plan Sponsor Council of America.

Rohaly, Jeffrey, Adam Carasso, and Mohammed Adeel Saleem. 2005. “The Urban-Brookings Tax Policy Center Microsimulation Model: Documentation and Methodology for Version 0304.” www.taxpolicycenter.org/UploadedPDF/411136_documentation.pdf.

Toder, Eric J., Benjamin H. Harris, and Katherine Lim. 2011. “Distributional Effects of Tax Expenditures in the United States.” In Tax Expenditures: State of the Art, ed. Lisa Philipps, Neil Brooks, and Jinyan Li. Toronto: Canadian Tax Foundation.

U.S. Treasury. 2010. General Explanation of the Administration’s Fiscal Year 2011 Revenue Proposals (2010 Green Book). Washington, DC: Government Printing Office.

VanDerhei, Jack, and Craig Copeland. 2010. “The EBRI Retirement Readiness Rating: Retirement Income Preparation and Future Prospects.” EBRI Issue Brief 344. Washington, DC: Economic Benefits Research Institute.

www.brookings.edu/papers/2006/11_proposal_effects.aspx

http://www.taxpolicycenter.org/UploadedPDF/411136_documentation.pdf

19


Tabl

e 2

Rev

enue

Los

s D

ue to

Aut

omat

ic IR

As

and

Save

r’s C

redi

t Exp

ansi

on

Impa

ct o

n Ta

x R

even

ue ($

bill

ions

), 20

11–2

01

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2011

–202

Bas

elin

e: C

urre

nt P

olic

y3

1. L

ow R

even

ue C

ost A

ssum

ptio

ns4

-0.1

-0.3

-0.4

-0.4

-0.4

-0.4

-0.4

-0.4

-0.5

-0.5

-3.8

2. L

ow C

ost A

ssum

ptio

ns, S

aver

’s C

redi

t Exp

ansi

on4,

5-3

.0-7

.3-7

.1-6

.9-6

.8-6

.8-6

.6-6

.4-6

.3-6

.3-6

3.4

3. M

ediu

m C

ost A

ssum

ptio

ns6

-0.9

-2.5

-2.7

-2.8

-2.9

-3.0

-3.2

-3.3

-3.4

-3.6

-28.

24.

Med

ium

Cos

t Ass

umpt

ions

, Sav

er’s

Cre

dit E

xpan

sion

5,6

-4.4

-10.

9-1

0.8

-10.

8-1

0.7

-10.

8-1

0.7

-10.

6-1

0.5

-10.

7-1

00.8

5. H

igh

Cos

t Ass

umpt

ions

7-2

.2-5

.7-6

.2-6

.5-6

.8-7

.1-7

.4-7

.7-7

.9-8

.3-6

5.8

6. H

igh

Cos

t Ass

umpt

ions

, Sav

er’s

Cre

dit E

xpan

sion

5,7

-6.0

-15.

0-1

5.2

-15.

3-1

5.5

-15.

7-1

5.7

-15.

8-1

5.9

-16.

2-1

46.4

Sou

rce:

Urb

an-B

rook

ings

Tax

Pol

icy

Cen

ter M

icro

sim

ulat

ion

Mod

el (v

ersi

on 0

509-

5).

(1) E

stim

ates

incl

ude

a m

icro

dyna

mic

beh

avio

ral r

epso

nse

with

an

elas

ticity

of t

axab

le in

com

e w

ith re

spec

t to

(1 –

mar

gina

l rat

e) o

f 0.2

5.

(2) N

umbe

rs m

ight

not

add

due

to ro

undi

ng.

(3) B

asel

ine

is c

urre

nt p

olic

y w

hich

ext

ends

200

9 es

tate

tax

law

and

all

indi

vidu

al in

com

e ta

x pr

ovis

ions

in th

e 20

01 a

nd 2

003

tax

cuts

as

wel

l as

a pe

rman

t 200

9 A

MT

patc

h, in

dexe

d fo

r infl

atio

n. A

ssum

es a

40

-60

fisca

l yea

r spl

it.

(4) P

ropo

sal i

mpl

emen

ts a

utom

atic

IRA

s an

d as

sum

es 9

0 pe

rcen

t of t

hose

elig

ible

for b

oth

Rot

h an

d tra

ditio

nal I

RA

s ta

ke u

p R

oth

IRA

s; c

ontri

butio

ns to

aut

omat

ic p

lans

are

3 p

erce

nt o

f wat

ges;

of t

he

popu

latio

n of

wor

kers

, 19.

7 pe

rcen

t are

elig

ible

for a

utom

atic

IRA

s; a

nd a

utom

atic

IRA

take

-up

rate

s ar

e 40

per

cent

of t

hose

est

imat

ed b

y M

adria

n an

d S

hea

(200

1).

(5) P

ropo

sal i

mpl

emen

ts th

e ex

pans

ion

of th

e S

aver

’s C

redi

t des

crib

ed in

Pre

side

nt O

bam

a’s

FY20

11 b

udge

t.

(6) P

ropo

sal i

mpl

emen

ts a

utom

atic

IRA

s an

d as

sum

es 5

0 pe

rcen

t of t

hose

elig

ible

for b

oth

Rot

h an

d tra

ditio

nal I

RA

s ta

ke u

p R

oth

IRA

s; c

ontri

butio

ns to

aut

omat

ic p

lans

are

3 p

erce

nt o

f wag

es a

nd

3.75

per

cent

for h

igh

earn

ers;

of t

he p

opul

atio

n of

wor

kers

, 32

perc

ent a

re e

ligib

le fo

r aut

omat

ic IR

As;

and

aut

omat

ic IR

A ta

ke-u

p ra

tes

are

60 p

erce

nt o

f tho

se e

stim

ated

by

Mad

rian

and

She

a (2

001)

.

(7) P

ropo

sal i

mpl

emen

ts a

utom

atic

IRA

s an

d as

sum

es 1

0 pe

rcen

t of t

hose

elig

ible

for b

oth

Rot

h an

d tra

ditio

nal I

RA

s ta

ke u

p R

oth

IRA

s; c

ontri

butio

ns to

aut

omat

ic p

lans

are

3 p

erce

nt o

f wag

es a

nd 4

.5 p

erce

nt

for h

igh

earn

ers;

of t

he p

opul

atio

n of

wor

kers

, 3 p

erce

nt o

f wag

es a

nd 4

.5 p

erce

nt fo

r hig

h ea

rner

s; o

f the

pop

ulat

ion

of w

orke

rs, 3

2 pe

rcen

t are

elig

ible

for a

utom

atic

IRA

s; a

nd a

utom

atic

IRA

take

-up

rate

s ar

e 80

per

cent

of t

hose

est

imat

ed b

y M

adria

n an

d S

hea

(200

1).

20


Table 4 Automatic IRAs and Saver’s Credit Expansion Under Low Cost Assumptions

Distribution of Federal Tax Change by Cash Income Percentile, 2015

Cash Income Percentile

Percent of Tax Units Percent

Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)

Average Federal Tax Rate

With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal

Lowest Quintile 13.2 0.9 0.25 13.0 -29 -0.2 4.4Second Quintile 26.1 1.5 0.25 27.7 -69 -0.2 10.0Middle Quintile 25.2 0.4 0.19 32.8 -90 -0.2 16.6Fourth Quintile 15.2 0.1 0.06 15.0 -49 -0.1 19.8Top Quintile 6.7 0.1 0.02 11.2 -42 0.0 25.5All 17.9 0.7 0.08 100.0 -55 -0.1 21.2

Addendum80–90 7.4 0.0 0.03 5.0 -37 0.0 22.490–95 6.5 0.1 0.02 2.6 -39 0.0 23.395–99 5.2 0.0 0.02 2.5 -46 0.0 25.2Top 1 percent 5.4 0.0 0.01 1.1 -81 0.0 29.1Top 0.1 percent 3.9 0.0 0.00 0.1 -87 0.0 31.7

Source: Urban-Brookings Tax Policy Center Microsimulation Model (version 0509-6).

Table 3 Automatic IRAs Under Low Cost Assumptions




Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)


With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal

Lowest Quintile 1.4 0.0 0.02 3.1 -2 0.0 4.6Second Quintile 4.7 0.0 0.03 12.9 -9 0.0 10.2Middle Quintile 6.3 0.0 0.04 22.2 -18 0.0 16.7Fourth Quintile 7.2 0.0 0.03 26.0 -25 0.0 19.8Top Quintile 6.1 0.0 0.02 35.8 -40 0.0 25.5All 4.8 0.0 0.02 100.0 -16 0.0 21.2

Addendum80–90 6.8 0.0 0.03 15.8 -35 0.0 22.490–95 5.8 0.0 0.02 8.1 -37 0.0 23.295–99 4.9 0.0 0.02 8.2 -46 0.0 25.2Top 1 percent 4.9 0.0 0.01 3.6 -80 0.0 29.1Top 0.1 percent 3.8 0.0 0.00 0.4 -87 0.0 31.7


21


Table 5 Automatic IRAs Under Medium Cost Assumptions




Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)


With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal


Addendum80–90 15.3 0.0 0.08 16.2 -91 -0.1 22.490–95 12.9 0.0 0.06 8.9 -103 -0.1 23.395–99 11.4 0.0 0.05 9.5 -137 0.0 25.2Top 1 percent 10.5 0.0 0.01 3.6 -204 0.0 29.1Top 0.1 percent 7.7 0.0 0.00 0.3 -191 0.0 31.7


Table 6 Automatic IRAs and Saver’s Credit Expansion Under Medium Cost Assumptions




Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)


With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal


Addendum80–90 15.9 0.0 0.08 7.8 -94 -0.1 22.490–95 13.5 0.1 0.06 4.3 -106 -0.1 23.395–99 11.7 0.0 0.05 4.5 -138 0.0 25.2Top 1 percent 11.0 0.0 0.01 1.7 -205 0.0 29.1Top 0.1 percent 7.8 0.1 0.00 0.2 -191 0.0 31.7


22


Table 7 Automatic IRAs Under High Revenue Cost Assumptions




Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)


With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal

Lowest Quintile 5.4 0.0 0.06 3.1 -7 -0.1 4.6Second Quintile 16.2 0.0 0.12 12.4 -34 -0.1 10.2Middle Quintile 19.7 0.1 0.13 19.8 -59 -0.1 16.6Fourth Quintile 21.1 0.0 0.11 23.6 -85 -0.1 19.7Top Quintile 18.2 0.0 0.07 41.1 -168 -0.1 25.4All 15.2 0.0 0.09 100.0 -61 -0.1 21.2

Addendum80–90 20.0 0.0 0.11 16.6 -135 -0.1 22.390–95 17.6 0.0 0.10 10.0 -166 -0.1 23.395–99 15.3 0.0 0.08 10.9 -224 -0.1 25.2Top 1 percent 13.9 0.0 0.02 3.6 -293 0.0 29.1Top 0.1 percent 10.5 0.0 0.00 0.3 -266 0.0 31.7


Table 8 Automatic IRAs and Saver’s Credit Expansion Under High Cost Assumptions




Change in After-Tax

Income

Share of Total

Federal Tax

Change

Average Federal

Tax Change

($)


With Tax Cut

With Tax Increase

Change (% Points)

Under the Proposal

Lowest Quintile 25.0 0.9 0.42 10.8 -48 -0.4 4.3Second Quintile 38.6 1.5 0.42 23.3 -118 -0.4 9.9Middle Quintile 37.6 0.4 0.32 26.7 -150 -0.3 16.4Fourth Quintile 28.5 0.1 0.14 16.9 -113 -0.1 19.7Top Quintile 18.7 0.0 0.07 22.3 -171 -0.1 25.4All 30.2 0.7 0.17 100.0 -114 -0.1 21.1

Addendum80–90 20.6 0.0 0.12 9.0 -137 -0.1 22.390–95 18.2 0.1 0.10 5.5 -169 -0.1 23.395–99 15.6 0.0 0.08 5.8 -225 -0.1 25.2Top 1 percent 14.5 0.0 0.02 1.9 -294 0.0 29.1Top 0.1 percent 10.6 0.1 0.00 0.2 -266 0.0 31.7


23


Table 9 Average Change in IRA Contributions Per Affected Taxpayer, 2015

Low Cost Intermediate Cost High CostWithout

Saver’s CreditWith

Saver’s CreditWithout




Saver’s CreditLowest Quintile 972 287 991 337 960 357Second Quintile 824 350 840 491 881 563Middle Quintile 1,074 546 1,164 767 1,276 901Fourth Quintile 1,369 800 1,566 1,168 1,764 1,402Top Quintile 2,109 1,948 2,684 2,591 3,292 3,204All 1,280 541 1,436 800 1,623 972

Addendum80–90 1,718 1,604 2,070 2,010 2,492 2,43390–95 2,037 1,843 2,730 2,640 3,391 3,27495–99 2,885 2,719 3,899 3,796 4,915 4,819Top 1 percent 4,829 4,397 6,170 5,86 7,152 6,875Top 0.1 percent 6,164 5,988 7,228 7,125 8,138 8,052


Table 10 Aggregate Change in IRA Contributions ($Millions), 2015

Low Cost Intermediate Cost High CostWithout






Saver’s CreditLowest Quintile 586 1,680 1,576 3,041 2,158 3,842Second Quintile 1,410 3,539 3,738 6,444 5,247 8,289Middle Quintile 2,248 4,621 5,747 8,507 8,329 11,320Fourth Quintile 2,710 3,376 6,904 7,635 10,293 11,086Top Quintile 3,122 3,172 8,926 8,981 14,511 14,568All 10,076 16,425 26,892 34,644 40,537 49,142

Addendum80–90 1,436 1,464 3,881 3,913 6,099 6,13390–95 710 725 2,106 2,122 3,574 3,59095–99 685 690 2,145 2,150 3,613 3,619Top 1 percent 292 293 794 795 1,225 1,227Top 0.1 percent 29 29 69 69 106 106


24


Appendix: methodology

In this section we describe our methodology for imputing eligibility for the Automatic IRA, take-up rates among those eligible, and contribution rates and choice of account type among those participating.

Eligibility Imputation To measure the effect of the Automatic IRA, it is necessary to impute eligibility for

this program into the PUF/CPS database. We use estimates from Harris and Fischer (2012) to impute Auto IRA eligibility. Harris and Fisher analyzed data from the SCF and CPS to estimate the proportion of workers who would be eligible for the Auto IRA under the congressional proposals to be between 19.7 percent and 32.0 percent. Part of the variation in estimates was due to differences in data sets, and part was due to different assumptions about program characteristics. The low estimate of 21.3 percent is derived from the CPS under a scenario in which no employees were voluntarily enrolled; the high estimate of 32.0 is derived from the SCF under a scenario in which half of small business workers are voluntarily enrolled by their employers.

To impute eligibility, we adopted techniques similar to those used in previous TPC analysis. Using a probit model, we assume coverage is observed if and only if X1β1 + ε1 > 0, where ε1 is assumed to be a standard normal random variable with mean 0 and variance 1, X is a vector of explanatory variables, and β is a vector of parameters to be estimated. We only assign eligibility for the program to those taxpayers who were deemed ineligible for an employer-sponsored pension in prior imputation procedures.27

The list of exogenous variables for each probit and censored regression is designed to be an exhaustive set of relevant variables that exist on both the SCF and the PUF. These variables include number of dependents, age (included as 10-year bracket dummies), income (as defined for purposes of the SCF), and the following components of income: income from a farm or business, tax-exempt interest income, taxable interest income, rental income from schedule E, pension income, taxable dividends, and realized capital gains (all defined as the natural logarithm of the income item plus one). We also include dummies for zero values of each income item; dummies for negative overall income, negative income from a business or farm, and negative capital income; as well as interactions between the negative income dummies and the appropriate negative income amount (defined as the natural logarithm of the absolute value of the income item plus one). In addition, we include dummies for whether the individual itemizes deductions on his or her federal tax return, and dummies for whether certain federal tax schedules are filed (C for business income, E for rental income, and F for farm income). The list of explanatory variables is identical for each equation. Equations are estimated separately for head of household and spouse, but are based on household-level values for the explanatory variables with the exception of age and earnings.28

27 Geissler and Harris (2009) describe the imputation methodology for eligibility. 28 The SCF is a household-based survey that records only total income and wealth items for all individuals

in the “primary economic unit” (PEU); it does not attribute shares of those amounts to individuals within the PEU. This provides a slight complication for those PEUs that consist of two unmarried individuals living together (with or without other financially interdependent members of the PEU). These individuals will show up in the income tax file as two single tax returns but will show up in the

25


Given the estimates of coverage and eligibility from the SCF, we impute values to tax filing units in the PUF/CPS database in two steps. First, we simulate whether the taxpayer is eligible for the Auto IRA using criteria described above.29 Using the estimated coefficients from the probit estimation and values of explanatory variables in the tax model database, we calculate Xb1 (where b1 refers to the probit estimate for β1). We then calculate the threshold probability, z = Φ-1(X1b1), where Φ is the cumulative standard normal probability distribution, and draw a uniform random number, p, between 0 and 1. If z < p, we assign a nonzero value for the item.30 Second, we adjust the imputed aggregates to match the aforementioned totals from Harris and Fischer. After the adjustment, the number of eligible taxpayers for automatic enrollment approximately matches the totals reported by Harris and Fischer.

It should also be noted that the statutory eligibility in the model complements the imputed eligibility calculations discussed above. That is, a taxpayer may be deemed “eligible” to automatically be enrolled in an IRA, but is still subject to the statutory limits of the tax benefits associated with those pensions.

Take-Up Rates We model take-up rates for eligible taxpayers using estimates derived from Madrian

and Shea (2001). Madrian and Shea studied the experience of employees at a company that adopted auto-enrollment in a 401(k) plan, and measured the program’s effects on take-up rates by gender, race, age, and compensation. Since the PUF/CPS data do not contain information on gender or race, we only used estimates of the age and compensation effects on take-up rates. We adjust the Madrian and Shea estimates to calculate each taxpayer’s probability of taking up a pension under auto-enrollment given that the taxpayer has already failed to “opt in” to a 401(k) or IRA account. Specifically, where Madrian and Shea calculated:

pi = the probability of an employee in age group i participating prior to auto-enrollment pj = the probability of an employee in compensation group j contributing prior to auto-

enrollmentp'i = the probability of an employee in age group i contributing post auto-enrollmentp'j = the probability of an employee in compensation group j contributing post auto-

enrollmentp = the average participation rate prior to auto-enrollmentp' = the average participation rate post auto-enrollment

SCF as one unit. We assume that an unmarried couple living together with shared finances behaves like a married couple and thus include them in the married category when running the regressions. The results do not change significantly if these individuals are dropped from the analysis.

29 Note that each taxpayer with positive wages falls into one of four groups for the purposes of automatic enrollment: covered by an employer pension, eligible but not participating in an employer pension (automatic 401(k)), ineligible for an employer pension but eligible for the Auto IRA, and ineligible for all programs.

30 Without adjustment, this process can produce too many or too few individuals with pension contributions on the PUF data set. We force the numbers to match published totals by shifting the threshold probabilities so the simulated number of contributors matches the estimates from the SCF.

26


we calculate for each taxpayer:

Then, in a technique similar to that utilized in the imputation process, we calculate the threshold probability, z = Φ-1(pij), where Φ is the cumulative standard normal probability distribution, and draw a uniform random number, p, between 0 and 1. If z < p, we model that taxpayer has having adopted a pension.

With no comparable study of the effects of auto-enrollment on IRA participation, we examine multiple sets of take-up rates for the Auto IRA, each a percentage between 30 percent and 50 percent of the rates calculated by Madrian and Shea for automatic enrollment in a 401(k)-type plan. Lacking observed take-up rates for the Auto IRA, we have no empirical basis for choosing specific parameters. We therefore estimate the effects of a range of values accounting for the decreased incentives of participating in an IRA relative to a company-provided 401(k).

Contribution Levels Madrian and Shea estimate auto-enrollment’s impact on average contribution rates

among 401(k) participants by gender, race, age, and compensation. Since they show only group averages and not regression-adjusted means, however, we cannot apply their estimates to our model. Still, their study provides some valuable insight into the trend of contribution rates to a retirement plan post auto-enrollment.

Madrian and Shea find that the average contribution rate among participants drops with the initiation of an auto enrollment program with a 3 percent default rate;31 employees at the lower end of the compensation scale reduce their contributions proportionately more.32 The authors find that although the average contribution rate for employees with higher compensation remained well above the default rate, the average contribution rate for employees with less compensation approached the default rate. Since auto-enrollment is more likely to affect lower-income employees, the results imply that it is common for employees with less compensation to contribute the default rate after auto-enrollment.

Given the lack of empirical evidence measuring the impact of automatic enrollment on IRA participation rates, we examine two scenarios of contribution rate trends. One scenario assumes that all automatically enrolled taxpayers contribute the default 3 percent of wages proposed by the administration. The second scenario assumes 3 percent contribution rates for all participants except high-income workers, who contribute 1.5 times the default rate in response to the increased incentives for high-income taxpayers. All contribution levels are subject to the statutory contribution limits.

31 While the study does not address aggregate contribution levels, the results imply that auto-enrollment increases aggregate contribution levels by about 40 percent if the income profile of participants remains approximately the same.

32 Beshears et al. (2009b) discuss the impact of default rates on saving behavior.

.

27


Choice of Account Type Modeling the type of account participants choose is important for revenue estimates.

Since the revenue cost of a traditional IRA occurs in the year the contribution is made and the revenue costs of a Roth IRA are deferred until retirees withdraw funds tax free, the assumption that most taxpayers contribute to a Roth IRA results in smaller current revenue losses than the assumption that most contributions go into traditional accounts. Distributional results, on the other hand, are much less sensitive to taxpayers’ assumed choice of account type. Since the TPC model accounts for the present value of future retirement benefits, the type of IRA taken up by savers has little effect on the distribution of benefits by income.

We present results for two broad scenarios, one in which 90 percent of automatic enrollees contribute to a Roth IRA and another similar 90 percent contribute to a traditional IRA.

The Population of Workers Covered by the Auto IRA:Trends ...€¦ · Auto IRAs with an expanded...

Documents

Transcript of The Population of Workers Covered by the Auto IRA:Trends ...€¦ · Auto IRAs with an expanded...