A recent proposal by Andrew Biggs and Alicia Munnell suggests repealing the tax exemption for employer‐sponsored retirement plans and IRAs and using the new revenue to address the majority of Social Security’s long‐term funding gap. The authors’ argument rests on their conclusion that the almost $190 billion annual tax increase on Americans’ retirement savings would not meaningfully change individuals’ incentive to save.
There are many reasons to object to the Biggs‐Munnell analysis, its impact on Social Security, and the implications for the federal budget. However, the crux of their argument—that tax‐advantaged retirement accounts “do little to increase retirement saving”—is an overconfident misinterpretation of the academic literature that does not acknowledge the broader economic benefits of private saving.
The overwhelming evidence is that tax‐advantaged accounts significantly increase private savings. Over time, even small increases in private savings can contribute to a larger capital stock, additional labor supply, and a bigger economy. The private benefits to additional saving, combined with the broader economic benefits, outweigh any temporary government losses, even if the lower revenue does not induce a one‐for‐one increase in private savings.
The balance of the evidence suggests that raiding Americans’ private retirement savings to prop up Social Security would significantly reduce private retirement savings and slow capital accumulation necessary to sustain a growing economy.
Saving in Theory
The US income tax levies multiple layers of tax on an individual’s savings. Wages are first taxed by the income and payroll taxes. The return on any income saved is then taxed again through taxes on corporate income, dividends, capital gains, and transfers at death. These taxes on investment returns reduce the after‐tax value of delaying current consumption for future consumption.
Tax‐advantaged saving accounts, such as 401(k)s and IRAs, reduce some of the income tax system’s savings penalty by eliminating capital gains and dividend taxes on funds invested for retirement. By raising after‐tax returns, tax‐advantaged accounts could reduce personal savings by allowing people to consume the same amount in retirement at a lower savings rate (called the income effect), or the accounts could induce additional savings and less pre‐retirement consumption to take advantage of the larger returns (substitution effect). Empirical estimates of these effects attempt to measure how much 401(k) balances represent “new savings,” as opposed to simply transfers of assets that would have been saved regardless of the tax benefit.
The empirical measurement of the personal savings response is only one part of the analysis. Policymakers should also care about a tax’s total economic cost or deadweight loss. Taxes on investment returns do not have to affect current savings to reduce future consumption. An individual’s well‐being is not determined by their savings per se but by the consumption the savings afford them. If the income and substitution effects cancel out and savings remain unchanged, a tax on capital gains that lowers a saver’s after‐tax return still mechanically reduces future consumption and thus has a deadweight cost.
Harvard economist Martin Feldstein explains that without changing personal savings, such taxes on capital can also reduce “labor supply broadly defined, including not only the number of hours worked but, more importantly, the incentive to acquire human capital, the choice of occupation, and the amount of effort.” By reducing lifetime consumption, capital taxes can have significant economic costs without a measurable effect on the amount saved. In other words, even if the observed effect of 401(k) accounts is as small as critics claim, repealing the tax treatment would be economically damaging.
Complicating matters further, tax‐advantaged accounts with contribution limits only change marginal decisions for savers who save less than the annual account limits and would otherwise face capital gains taxes. The first capital gains income threshold kicks in at $47,025 in 2024 for single taxpayers ($94,050 for married filing jointly). According to Vanguard account data, only 15 percent of Americans hit their contribution limits in 2022. At least one‐third, and likely many more, of Vanguard participants who are subject to the capital gains tax contribute less than the annual maximum and thus benefit from the marginal saving incentive. If policymakers are worried about diminished marginal investment incentives, the solution is simple: eliminate contribution limits.
The final theoretical challenge is that reducing taxes on retirement savings comes at a fiscal cost to the government, which could offset the gains from increased personal savings. In the absence of other changes, the lower tax revenue requires additional government borrowing. The savings literature often assumes that any increase in private savings is offset by the government’s dissaving, which crowds out domestic investment. While this assumption is partly true, in an open economy with easy access to global credit markets, a significant share of federal debt is financed with international investment and thus does not crowd out domestic investment. In a fully open economy, a decrease in the after‐tax return for domestic savers would also not significantly affect the capital stock but would increase domestic ownership of US assets. While the degree of domestic investment crowd out and international capital openness are the subject of ongoing debate, simply comparing tax expenditure costs with increases in private savings is not sufficient to learn anything meaningful about how tax‐advantaged accounts affect national or net savings.
Empirical literature shows significant increases in personal saving
The empirical literature is plagued with theoretical, methodological, and data issues that leave most research summaries to conclude there is no consensus on the impact of tax‐advantaged accounts on national savings. Despite the surface‐level disagreements, the literature clearly shows that tax‐advantaged accounts increase personal savings. Modeling suggests that tax‐advantaged accounts can also contribute to a larger capital stock over time. However, the way in which the results are extended to measure the effects on national savings is not informative for policymaking because the increase in private savings is not directly comparable to official expenditure estimates.
In the 1990s, two groups of authors published a series of papers that used different econometric approaches and assumptions to estimate the savings effect of 401(k) and IRA contributions. The two methods came to conflicting results, finding under one specification that the accounts had “not stimulated private savings” and the other concluding that “most 401(k) contributions represent net new saving.” A robust academic debate on the topic continues to this day.
More recent research by Alexander Gelber looks at how individuals change their savings after becoming newly eligible for a 401(k), after completing a year at a new employer. Some of the results are imprecise, but using a within‐individual change helps control for unobserved differences between savers. Gelber finds that 401(k) eligibility results in a net increase in savings, raising 401(k) balances “substantially,” with no significant decreases in other financial assets. Looking at a much narrower cohort of only college and university professor behavior, David Card and Michael Ransom report that about 30 percent of supplemental savings in tax‐deferred plans could represent new personal savings. Framing additional retirement contributions as provided by the employer significantly increases the personal savings rate, and as much as 70 percent of account balances represent new savings.
Harvard economist Daniel Benjamin improves on research from the 1990s with a better matching technique to control for unobserved characteristics, estimating that roughly one‐quarter of 401(k) balances represent new national savings. A second quarter represents tax savings. Combined, Benjamin’s estimates imply that about half of 401(k) balances are new private savings.
In their summary of the literature, Biggs and Munnell rely on a study using Danish data by Raj Chetty and four co‐authors, claiming the results “have been well received and broadly accepted.” As Joshua Rauh and Stan Veuger recently pointed out, it is unclear if Denmark’s more rigid, government‐mandated retirement system and the observed change in the relative value of different account types provides a useful analogy to the US system or a policy that eliminates retirement accounts altogether. In other words, it’s doubtful that the findings in Denmark are generalizable to the United States. Still, the paper’s results confirm that tax‐advantaged accounts can substantially increase personal savings.
The Chetty et al. results primarily investigate how nontax factors that make savings easier can increase savings by focusing on active vs. passive savers. They find that about 80 percent of the increase in employer retirement contributions are new personal savings, and about 10 percent (and as much as 50 percent) of individual retirement contributions are new personal savings. Only after accounting for the fiscal cost of the savings accounts do the researchers conclude that tax‐advantaged account balances are about 1 percent (and as much as 10 percent) new national savings.
Even small increases in total savings compound over time. Eric Egen, Willam Gale, and John Karl Scholz—whose empirical work is generally cited to argue that tax‐advantaged accounts only induce a small savings response—model the long‐run effects of retirement savings accounts on national savings rates. Their results show that 401(k)s can increase the long‐term national savings rate by between 8 percent and 17 percent, and other accounts, such as IRAs, can increase it further.
Synthesizing available evidence in the mid‐1990s, Glenn Hubbard and Jonathan Skinner similarly concluded that a “conservative estimate of the effect of IRAs on personal saving” shows that 26 cents of every dollar of IRA contributions represent new savings. They suspect the true effect “is actually somewhat larger.” Hubbard and Skinner then show that “even for quite conservative measures of the saving effects of IRAs or 401(k)s…the incremental gains in capital accumulation per dollar of lost revenue are generally large.” After acknowledging the strong assumptions necessary for multi‐decade models, one summary of these results from the Federal Reserve Bank of New York concluded that the estimates imply retirement accounts “would eventually increase the capital stock $4 and $16, respectively, for every dollar increase in the government debt.”
While beyond the scope of this brief literature review, it is worth noting that related bodies of economic research generally find large positive economic effects when taxes on capital income are reduced through other mechanisms, such as lower headline capital gains and corporate income taxes.
Tax‐advantaged savings accounts increase personal savings and can contribute to a larger capital stock, which increases the productive economy’s size and in turn can also increase tax revenues.
A note on tax expenditures
The overwhelming finding in the savings literature is that tax‐advantaged accounts increase private savings. Thus, their effect on national savings turns on the definition of tax expenditure and estimates of the revenue reduction from the preferential tax treatment. The tax expenditure estimates used by most of the literature, including Biggs and Munnell, omit any additional tax revenue from a larger economy and more fundamentally rely on a biased definition of the tax expenditures.
In 1995, Feldstein modeled the effect of tax‐advantaged accounts on total government revenues after accounting for higher corporate tax revenues from the additional personal investment in retirement accounts. The modeling is subject to debated assumptions—including the degree of openness to international capital—but Feldstein concludes that “the increase in corporate tax collections will eventually outweigh the loss of personal tax revenue regardless of the [models] parameter values.” His preferred specification shows that the additional corporate tax revenue makes up about two‐thirds of the decline in personal income tax revenue within ten years. If the higher after‐tax rate of return on savings induces additional labor supply and work effort, there should also be an increase in income and payroll tax revenue. These effects do not need to be as large as Feldstein claimed or make up all the lost revenue to make 401(k)-type accounts a net‐benefit to saving.
Lastly, the definition of tax expenditure matters when thinking about budget deficits. Tax expenditure lists must first pick a reference tax base from which to measure any deviations. By using an income tax base that implicitly assumes capital income should be taxed at a higher rate than wage income, government scorekeepers bias the way researchers and policymakers think about revenue loss. If measured from a tax base that treats current and future consumption equally, 401(k)s and other similar accounts would not be counted as a tax expenditure. Instead, the tax on capital gains would be a negative tax expenditure, bringing in additional revenue beyond what the normal tax base would collect.
While some may consider the definition of tax expenditure to be a semantic point, it should make policymakers searching for additional revenue consider how language and modeling decisions distort policy choices. For policymakers looking to increase revenue, hundreds of billions of dollars of true tax loopholes could be eliminated with smaller deadweight costs than raising taxes on savers. However, any transfer of income from individuals to the government through higher taxes is, at best, zero‐sum. In the case of retirement savings, the government has clearly demonstrated that private savers are better stewards of their money than Congress.
None of this analysis denies that the Social Security system is in crisis. The trust fund will become insolvent in less than a decade, and the typical newly‐retired beneficiary will face automatic benefit cuts of around $17,000 in 2033. The proposal from Biggs and Munnell is correct that Social Security is in dire need of reformers with the courage to make difficult tradeoffs that can ensure America’s current and future retirees don’t face indiscriminate benefit cuts as a result of congressional inaction. However, increasing taxes to prop up the current broken Social Security system at the expense of economic growth should be avoided. Increasing taxes that directly reduce personal savings and undermine the private alternative to the broken government‐run system would be particularly destructive to America’s future retirees.
 Individuals with annual savings above the combined tax‐advantaged account threshold have no substitution effect but still face an income effect. Americans with no annual savings or assets face a positive substitution effect but no income effect. Net debtors and individuals whose net present value of future income is greater than current assets benefit from mutually reinforcing income and substitution effects that increase savings.
 Chetty et al. do not observe investment changes in some durables, such as cars. This could be a significant omission given that Gelber finds car values fall substantially in response to 401(k) eligibility.