the under-appreciated potential and misunderstood failings ... · getting americans to save: the...

41
VERY PRELIMINARY DRAFT: NOT FOR CITATION 1 Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin * DRAFT: 3/23/16 Note to readers: This is a very early draft, and I appreciate your wading through! Much of this is tentative, but Parts III and IV are especially so. Among other issues, Parts III and IV need to be better tied to Parts I and II, and, as I describe below, I’m aiming to reorganize Part IV at the least. I look forward to your feedback and thanks! Introduction........................................................................................................................................... 2 I. Under-Saving and the Rise of Tax-Preferred Savings Accounts ......................................... 4 A. Under-Saving: Empirical Evidence ................................................................................................ 4 B. Under-Saving: Behavioral Models ................................................................................................. 6 C. Rise of Tax-Preferred Saving Accounts .......................................................................................... 7 II. How Tax Subsidies Can Increase Saving ............................................................................ 10 A. Empirical Evidence Before the Denmark Study .......................................................................... 11 B. The Denmark Study: What Happens in Denmark Doesn’t Stay in Denmark ......................... 12 1. Design and Results of the Denmark Study .................................................................................... 12 2. Active Savers in the Denmark Study ............................................................................................ 14 4. Passive Savers in the Denmark Study ........................................................................................... 17 5. Comparison to Mandates ............................................................................................................... 18 C. The Importance of Subsidy Design ............................................................................................... 19 1. Location of the Subsidy ................................................................................................................. 20 2. Timing of Taxation ........................................................................................................................ 20 3. The Case of Roth-Style Accounts ................................................................................................. 21 III. Why Not Tax Subsidies: A Tentative Case ....................................................................... 24 A. Increasing Saving Most of Those Who Need It Least.................................................................. 24 B. Detrimental Distributional Consequences .................................................................................... 25 C. Defaults and Floors As Alternatives.............................................................................................. 28 1. Defining a Nudge and a Floor ....................................................................................................... 28 2. Nudges, Floors, and “Passive” and “Active” Savers..................................................................... 29 D. Tax Subsidies as Complements to Nudges? ................................................................................... 30 1. Goldin and Lawson Framework .................................................................................................... 30 2. Relative Insensitivity to Tax Subsidies and, Maybe, Nudges ....................................................... 31 3. Redistribution ................................................................................................................................ 32 4. A Tentative Case Against Tax Incentives ..................................................................................... 33 IV. Three Reform Models ......................................................................................................... 34 A. Improving Tax Subsidies................................................................................................................ 35 1. Structuring for Passive Savers ....................................................................................................... 35 2. Eliminating the “Upside Down” Structure .................................................................................... 36 3. Eliminating Disproportionate Benefits by Sophisticated Players ................................................. 37 4. The Model of Matching Credits .................................................................................................... 38 B. Better Nudges and Expanded Floors ............................................................................................. 39 1. Better Defaults ............................................................................................................................... 39 2. Expanded Floors ............................................................................................................................ 40 Conclusion ................................................................................................................................... 40 * Associate Professor of Law, New York University School of Law.

Upload: others

Post on 13-Jun-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

1

Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies

David Kamin* DRAFT: 3/23/16

Note to readers: This is a very early draft, and I appreciate your wading through! Much of this is tentative, but Parts III and IV are especially so. Among other issues, Parts III and IV need to be better tied to Parts I and II, and, as I describe below, I’m aiming to reorganize Part IV at the least. I look forward to your feedback and thanks!

Introduction...........................................................................................................................................2I. Under-Saving and the Rise of Tax-Preferred Savings Accounts ......................................... 4

A. Under-Saving: Empirical Evidence ................................................................................................ 4 B. Under-Saving: Behavioral Models ................................................................................................. 6 C. Rise of Tax-Preferred Saving Accounts .......................................................................................... 7

II. How Tax Subsidies Can Increase Saving ............................................................................ 10 A. Empirical Evidence Before the Denmark Study .......................................................................... 11 B. The Denmark Study: What Happens in Denmark Doesn’t Stay in Denmark ......................... 12

1. Design and Results of the Denmark Study .................................................................................... 12 2. Active Savers in the Denmark Study ............................................................................................ 14 4. Passive Savers in the Denmark Study ........................................................................................... 17 5. Comparison to Mandates ............................................................................................................... 18

C. The Importance of Subsidy Design ............................................................................................... 19 1. Location of the Subsidy ................................................................................................................. 20 2. Timing of Taxation ........................................................................................................................ 20 3. The Case of Roth-Style Accounts ................................................................................................. 21

III. Why Not Tax Subsidies: A Tentative Case ....................................................................... 24 A. Increasing Saving Most of Those Who Need It Least .................................................................. 24 B. Detrimental Distributional Consequences .................................................................................... 25 C. Defaults and Floors As Alternatives .............................................................................................. 28

1. Defining a Nudge and a Floor ....................................................................................................... 28 2. Nudges, Floors, and “Passive” and “Active” Savers ..................................................................... 29

D. Tax Subsidies as Complements to Nudges? ................................................................................... 30 1. Goldin and Lawson Framework .................................................................................................... 30 2. Relative Insensitivity to Tax Subsidies and, Maybe, Nudges ....................................................... 31 3. Redistribution ................................................................................................................................ 32 4. A Tentative Case Against Tax Incentives ..................................................................................... 33

IV. Three Reform Models ......................................................................................................... 34 A. Improving Tax Subsidies ................................................................................................................ 35

1. Structuring for Passive Savers ....................................................................................................... 35 2. Eliminating the “Upside Down” Structure .................................................................................... 36 3. Eliminating Disproportionate Benefits by Sophisticated Players ................................................. 37 4. The Model of Matching Credits .................................................................................................... 38

B. Better Nudges and Expanded Floors ............................................................................................. 39 1. Better Defaults ............................................................................................................................... 39 2. Expanded Floors ............................................................................................................................ 40

Conclusion ................................................................................................................................... 40

*Associate Professor of Law, New York University School of Law.

Page 2: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

2

Introduction There is considerable evidence that Americans under-save,1 and one of the primary systems for getting them to save more—a system of tax-preferred retirement accounts—is fundamentally broken, according to the recent literature. This system of 401(k)s, Individual Retirement Accounts (IRAs), and other tax-preferenced accounts cost the government about $80 billion per year relative to a true income tax system,2 and influential new research suggests that tax incentives like these are a rather expensive mistake, as they do very little to increase saving.3 However, the case against tax subsidies is not as simple as it is sometimes made out to be, and, in fact, misinterpretation of the latest research has led to an unhelpful dead end in the literature. This article describes how tax subsidies, in fact, can increase saving; the recent empirical results do not, in fact, suggest otherwise. Tax subsidies may still not be a first-best policy tool but that case is tentative—much more tentative than this recent literature would suggest. Further, whether tax subsidies are in theory a first best or second best tool, the current structure of tax subsidies in the United States is deeply flawed, and this article also seeks to reinvigorate the project of reforming these tax subsidies if they are retained as policy tools. That is because they are not irredeemably broken as some of the recent literature might suggest. For purposes of this article, the key characteristic of a tax subsidy for saving is that it reduces tax liability for those who save more versus those who save less. Empirical studies over the last few decades arrived at varying results as to the effectiveness of such tax subsidies at increasing private saving, but all of these were limited by the availability of data.4 A recent article published by the economist Raj Chetty, John Friedman, and a set of co-authors overcame many of these data problems by looking at Denmark (hereinafter referred to as the “Denmark study”), and the analysis has been widely lauded and cited as a result.5 Denmark features remarkably complete data on savings and assets that allows the economists to study the effects of tax incentives and other interventions in a way that has so far not been possible. Based on this, the authors conclude that “each $1 of government expenditure on [tax] subsidies increases total

1 See infra Part I.A. The magnitude of this problem is subject to some debate. However, 2 This is Treasury’s estimate of the net present value cost of tax-preferred retirement saving accounts relative to a baseline of pure income taxation. If the accounts were actually eliminated, the actual cost may be less as people shift into other tax-preferenced forms of investment. See Table 4, https://www.treasury.gov/resource-center/tax-policy/Documents/Tax-Expenditures-FY2017-Tables-11-13-2015.xlsx. 3 See, e.g., Ryan Bubb & Richard H. Pildes, How Behavioral Economics Trims Its Sails and Why, 127 HARV. L. REV. 1593, 1631 (2014) (describing how retirement tax incentives in the United States represent “a vast amount of public money spent for a minor increase in total private retirement savings.”); RAJ CHETTY ET AL., CTR. FOR RETIREMENT RESEARCH AT BOSTON COLLEGE, SUBSIDIES VS. NUDGES: WHICH POLICIES INCREASE SAVING THE MOST? (2013) (“The findings of this study call into question the large tax expenditure currently used to induce individuals to save…. Automatic saving, which relies on passive choice, is an attractive alternative.”). 4 See infra Part II.A. 5 Raj Chetty et al., Active vs. Passive Decisions and Crowd-Out in Retirement Savings Accounts: Evidence from Denmark, 129 Q.J. ECON. 1141 (2014). The study on savings behavior, for instance, is among the work described by the American Economics Association in explaining its award of the John Bates Clark Medal for best economist under 40 to Raj Chetty. See https://www.aeaweb.org/honors_awards/bios/Raj_Chetty.php.

Page 3: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

3

saving by only 1 cent.”6 In other work, they cite to this as showing why policymakers should pursue alternatives like nudges and mandates.7 And, other scholars have reached similar conclusions based on these findings.8 However, the evidence on the effects of tax subsidies has been widely misinterpreted. Tax subsidies can increase saving through at least two channels. First, for those who appear to be attentive to changes in incentives (what Chetty and others call “active savers”), the Denmark study suggests that they do end up increasing their saving somewhat due a tax subsidy. This represents an increase in private saving.9 Second, for those who are inattentive (what Chetty and others call “passive savers”), the structure of the subsidy matters. So, while the subsidy studied in Denmark did not change private saving among this group, it is possible to construct tax subsidies—for instance, by having any subsidy go directly into a savings account—that should increase saving of passive savers.10 In fact, one type of saving account now used in the United States—a Roth-style account—may be doing exactly that.11 Thus, tax subsidies do not so clearly fail as a way of increasing saving, either among “active savers” or “passive savers.” As a result, the literature making the case against tax subsidies has inappropriately “dead ended” based on the contention that tax subsidies simply do not increase saving by much at all. That conclusion is not supported by the most recent empirical evidence. However, even if tax subsidies can increase saving, they may still not be a first-best policy tool, and this article lays out a tentative case against them, though one different from the recent focus of the literature. First, such subsidies will tend to most increase the savings of those who were saving more to begin with. This seems relatively perverse—under-saving seems likely to be more of a problem for those who save less relative to their incomes.12 Second, such subsidies—even if restructured—will necessarily have redistributional consequences, transferring resources from non-savers to savers in a way that could very well be deleterious. Tax subsidies could potentially be combined with the use of other nudges—such as default saving rates—to make them more effective in combination. A case for such a combination has recently been made by the scholars Jacob Goldin and Nicholas Lawson.13 However, while 6 Chetty et al., supra note 5, at 1141. 7 See, e.g., RAJ CHETTY ET AL., CTR. FOR RETIREMENT RESEARCH AT BOSTON COLLEGE, SUBSIDIES VS. NUDGES: WHICH POLICIES INCREASE SAVING THE MOST? (2013) (“The findings of this study call into question the large tax expenditure currently used to induce individuals to save…. Automatic saving, which relies on passive choice, is an attractive alternative.”) 8 [Add in cites, including to Brian Galle.] 9 See infra Part II.A.3. 10 See infra Part II.B. 11 See infra Part II.B.3. 12 Ryan Bubb makes a similar point in laying out his case against employer matching incentives, noting how they perversely help those most who are already saving more than the typical worker. See Ryan Bubb & Patrick Corrigan, A Behavioral Contract Theory Perspective on Retirement Savings. [get additional citation info.] 13 See generally Jacob Goldin & Nicholas Lawson, Nudges, Mandates, and Taxes: Policy Design with Active and Passive Decision-Makers (Aug. 30, 2015) (unpublished manuscript).

Page 4: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

4

certainly an improvement over using tax subsidies alone, there remain serious questions whether they should be deployed even in this context. In particular, there is still a question whether nudges and subsidies can effectively change the behavior of many savers, and there is a further question whether the redistributive effects—from non-savers to savers—is worth the additional saving the subsidies can promote. In short, there is a case to be made against tax subsidies—and for the alternative tools of defaults and mandates—even if that case is not clear-cut. However, whether subsidies are first- or second-best tools in theory, it is clear that the current system of tax subsidies in the United States is poorly designed. And, by focusing on how tax subsidies could actually help people save, this article is meant to revive the project of reforming them, if they continue to remain an important part of the retirement saving system. For instance, the tax subsidies should be restructured so that they go directly into the retirement accounts and so that sophisticated taxpayers cannot game the system to achieve large tax windfalls, as they now can do. Part I starts by briefly reviewing the evidence that Americans under-save and describing the system we have to try to encourage greater saving, including the system of tax-preferenced accounts. Part II explains why tax subsidies can be more effective at increasing saving than much of the recent literature gives them credit for. Part III describes why, nonetheless, tax subsidies may not be a first best policy tools for responding to under-saving. Finally, Part IV lays out paths forward for reform, including ways to significantly improve the performance of tax subsidies and, perhaps even better, ways to move away from them. I. Under-Saving and the Rise of Tax-Preferred Savings Accounts

A. Under-Saving: Empirical Evidence While there is some real dispute over the degree to which Americans under-save for retirement, most estimates show a significant share of the population with too little in savings.14 Further, those estimates suggest that the problem may be getting worse with time. To say that Americans are under-saving implies that there is a normative model for determining how much people should save. One set of widely-used models describe a rational decision-maker distributing consumption over a lifetime in such a way that maximizes welfare.15 The notion is that a person’s welfare will be maximized if the person saves and spends so that the 14 As Daniel Shaviro summarizes the evidence, “In both public policy debate and the academic literature, there iswidespread, though not universal, agreement that millions of Americans are saving too little for their own retirements.” Daniel Shaviro, “Multiple Myopias, Multiple Selves, and the Under-Saving Problem,” 47 CONN. L. REV. 1215, 1215 (2015). For a broad review of this literature reaching this same conclusion, see KEITH MILLER ET AL., CTR. FOR AM. PROGRESS, THE REALITY OF THE RETIREMENT SAVING CRISIS (2015). 15 These models are based on the related theories laid out by Milton Friedman (often called the “permanent income hypothesis”) and Franco Modigliano and Richard Brumberg (often called the the “life-cycle model of consumption and saving”). See, e.g., MILTON FRIEDMAN, A THEORY OF THE CONSUMPTION FUNCTION 20-37 (1957); Franco Modigliani & Richard Brumberg, Utility Analysis and the Consumption Function: An Interpretation of Cross-Section Data, in Post-Keynesian Economics 388 (Kenneth K. Kurihara ed., 1954). For further explanation of these theories, see Lee Anne Fennell & Kirk J. Stark, Taxation Over Time, 59 TAX L. REV. 1 , 6-9 (2005); [add cite to Shaviro, Beyond Pro-Consumption Tax Consensus].

Page 5: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

5

marginal utility of an additional dollar of consumption is the same at all points across that person’s lifetime. Under certain strong assumptions, the models call for strict consumption smoothing—the same amount spent in every year and irrespective of the amount of income earned in that particular year. Strict consumption smoothing is almost certainly not optimal in practice—for instance, because the marginal utility of consumption may be higher with children in the home—and the models tend to include some adjustments like these. But, the basic conclusion from these models remains the same—that consumption should be spread so that it does not rise and fall from year to year with volatile income. Using these models, a number of economists have tried to model optimal saving and consumption paths, and then compare that to what is actually occurring. One source for such estimates is the National Retirement Risk Index (NRRI), produced out of the Center for Retirement Research at Boston College.16 The NRRI calculates the share of the population currently on track to be unable to maintain living standards in retirement—falling more than 10 percent below their target income replacement rate.17 The NRRI suggest that roughly half of the working-age population is under-saving.18 Importantly, that figure has risen over time—from about thirty percent of the working-age population in the early 1980s.19 Notably, the numbers vary widely depending on whether workers have access to an employer retirement and the type of plan. The figure is lowest, about one-fifth, for workers with access to a plan with a “defined benefit”—as in a traditional pension plan. It rises to about half for those with only access to a “defined contribution” plan—that is contributions to savings accounts. Finally, it is highest for those with access to neither, with nearly 70 percent falling short among this population.20 The NRRI suggests there is under-saving across the income spectrum, ranging from 40 percent under-saving among those with higher incomes to about 50 percent for those in the middle of the income spectrum to 60 percent for those with low-incomes.21 Finally, the needed increase in savings rate to hit the calculated replacement rate targets varied greatly depending on the age of the savers. But, for those aged 30-39 for instance, the necessary increase was about seven to eight percent of their incomes for each income class.22

16 For the latest NRRI estimates, see ALICIA H. MUNNELL ET AL., CTR. FOR RETIREMENT RES. AT B.C., NRRI UPDATE SHOWS HALF STILL FALLING SHORT (2014) [hereinafter “NRRI UPDATE”]. For an explanation of the methodology used, see Alicia H. Munnell et al., Ctr. for Retirement Res. at B.C., How Much Should People Save? (2014) [hereinafter “HOW MUCH TO SAVE?”]. 17 MUNNELL ET AL., HOW MUCH TO SAVE?, supra note 16, at 3. 18 MUNNELL ET AL., NRRI UPDATE, supra note 16, at 1. 19 Id. at 3 fig.2. 20 Id. at 5 tbl.3. 21 Id. at 5 tbl.2. 22 MUNNELL ET AL., HOW MUCH TO SAVE?, supra note 16, at 4 tbl.4.

Page 6: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

6

These statistics have been disputed, with some suggesting that the savings shortfall is even worse than suggested by the NRRI,23 and others indicating that the saving shortfall is significantly smaller.24 Some of the dispute relates to exactly how to adjust for their departure of children from the home, among other factors. However, in all of these models, a significant share of the population is seen to be under-saving from retirement.25 In other words, that empirical fact seems relatively well supported, across a variety of studies, even if the magnitudes do differ.

B. Under-Saving: Behavioral Models The very notion of “under-saving” assumes either: (1) a market failure that leads otherwise rational and informed actors to under-save; (2) a behavioral model in which people, for one reason or other, choose saving rates that are too low despite market incentives being properly aligned. The most recent economics literature tends to focus, for the most part, on behavioral models to explain under-saving.26 As Daniel Shaviro summarizes, this literature identifies “multiple myopias” that might lead to under-saving—that is, multiple mechanisms that might lead people to consume more sooner and save less than they should.27 Shaviro chronicles the various possible mechanisms, and I will not repeat that full accounting here—but, rather, give a few leading examples. For instance, one possible mechanism is “hyperbolic discounting.” This is a behavior in which people apply “a much higher discount rate between the current time and any future times than between future times.”28 Thus, for hyperbolic discounters, there is always a thumb on the scale toward current consumption in the moment they are making the decision as to whether to save or consume—and more than they would prefer looking ahead. Preferences, in this case, are inconsistent across time in such a way that favors one set of these preferences (consuming more now). Another possible cause of under-saving is the use of “mental accounts.”29 In this case, there is evidence that people use certain rough rules of thumb to decide how much to spend and save. And, people seem to code certain types of receipts as “current income” to be used for current consumption, in contrast to assets and future income which are coded for future consumption. 23 See NARI RHEE, NATIONAL INSTITUTE ON RETIREMENT, THE RETIREMENT SAVINGS CRISIS: IS IT WORSE THAN WE THINK? (2013) (estimating that between 65 and 90 percent of households fall short in terms of retirement savings). 24 William G. Gale et al., Are All Americans Saving 'Optimally' for Retirement?, (Working Paper, The Brookings Institution and the University of Wisconsin-Madison, 2009), available at http://www.ssc.wisc.edu/~scholz/Research/Are_All_Americans_v6.pdf. 25 For instance, the more optimistic of these studies suggests that 28 percent of those born early on in the baby boom were under-saving (the latest cohort addressed in that study). The figure was somewhat less for earlier cohorts addressed in the study. Id. at 24 tbl.4. 26 See, e.g., Hersh M. Shefrin & Richard H. Thaler, The Behavioral Life-Cycle Hypothesis, 26 ECON. INQUIRY 609 (1988) (cataloguing some of the cognitive difficulties people face in smoothing consumption). [build out citations] 27 See generally Shaviro, supra note 14. 28 Id. at 1243. [Add in additional cites.] 29 Id. [Add in additional cites.]

Page 7: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

7

But, this can lead to too much current income going towards consumption (and too close a relationship between current income and current consumption).30 Yet another possible mechanism is one of inattentiveness or passivity.31 Examples of this are discussed at greater length below. Passivity is not necessarily irrational.32 After all, given limited mental capacity, it can be rational to not pay immediate attention to new information or choices. However, to the extent limited attention can be overcome in a way that improves outcomes, it should be. Such passivity also does not necessarily in itself generate low savings. It can if underlying defaults and other parts of the decision framework are designed in such a way as to produce under-saving due to inattentiveness. In other words, with passivity, the decision framework matters hugely and can theoretically either lead to over- or under-saving. The concern is that the decision framework in the United States now leads passive decision-makers to under-save.

C. Rise of Tax-Preferred Saving Accounts The United States has a tripartite structure for facilitating retirement saving—long referred to as a “three-legged stool.”33 The three legs of the stool are a mandatory system of saving combined with a lifetime annuity at retirement through the Social Security system; tax-preferred private pensions through employers (often called a defined-benefit plan); and, finally, contributions to private accounts (often called a defined-contribution plan when done through the employer), facilitated by tax-preferred savings accounts. Table 1 briefly summarizes the array of retirement saving vehicles in the United States. In recent years, the tax-preferred saving accounts have taken on a much greater role in the system, and they are the focus of much of this article. The question is whether tax subsidies like that delivered through these accounts—that makes saving cheaper as compared to what it would otherwise be—increases savings or not and, even if it does, whether tax incentives are a first-best tool. There are a thicket of details and requirements when it comes to each of these types of saving vehicles. The defined-benefit (DB) plans, defined contribution (DC) plans, and IRA accounts (often set up independent of any employer) have contribution limits,34 though the amounts vary widely. For instance, as of 2016, individuals can contribute no more than $18,000 per year to a 401(k) plan—one type of defined-contribution plan—and combined employer and employer contributions to these accounts cannot exceed $53,000.35 Further, there are a complex set of anti-discrimination rules that apply to both defined benefit and defined contribution plans that

30 Id. 31 Id.at 1249-51. [Add in additional cites.] 32 Id. at 1253. 33 For a history of the origins of the “three-legged stool” metaphor, see here: https://www.ssa.gov/history/stool.html. 34 Note that some IRAs can be established by employers and receive employer contributions. [Cite to SEP IRA.] 35 [Add in cites.]

Page 8: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

8

restrict them from being too skewed toward highly compensated employees within a firm.36 Each of these accounts has some form of restriction on withdrawal prior to retirement. For instance, most of these accounts include a 10 percent early withdrawal penalty, with some exceptions, and some have additional barriers to early withdrawal beyond that.37 Tax preferred-saving accounts—in the form of both DB plans and IRAs—have grown considerably over recent decades. IRAs were first introduced in 1974,38 and the modern employer retirement account—in the form of the 401(k)—was first allowed in the early 1980s when the IRS clarified regulatory guidance.39 These accounts—and especially IRAs—have expanded as DC plans have dissipated in importance. The chart below shows the composition of retirement assets by type of tax-preferred retirement savings vehicle, based on calculations from the President’s Council of Economic Advisers. Whereas in 1978, the combination of defined contribution plans and IRAs composed just over 20 percent of such assets, that had risen to over 50 percent by 2013.40

36 [Add in cites.] 37 Cite here: https://www.irs.gov/Retirement-Plans/Retirement-Plans-FAQs-regarding-Hardship-Distributions. 38 Cite to ERISA. 39 U.S. Internal Revenue Service, “Certain Cash or Deferred Arrangements Under Employee Plans,” 46 Fed. Reg. 55544 (November 10, 1981). 401(k) had been added into the code in 1978. But, until this regulatory guidance was issued in 1981, it was not clear whether 401(k) contributions could only come from a profit-sharing bonus as opposed to ordinary salary. In the guidance, the IRS indicated that it could come from ordinary salary, and the modern 401(k) was born. See here: https://www.ici.org/pdf/per12-02.pdf. 40 COUNCIL OF ECONOMIC ADVISERS, THE EFFECTS OF CONFLICTED INVESTMENT ADVICE ON RETIREMENT SAVINGS 5 (2015).

Page 9: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

9

The subsidy provided by the tax-preferred retirement account is to exclude the normal return on saving from taxation—a deviation from the principles of an income tax (though not a consumption tax, as described further below). The traditional accounts provide for an exclusion or deduction upfront of the amounts contributed to the account; an exclusion for earnings on those assets as they accrue; and, finally, taxation at the point of distribution on all amounts withdrawn. The effect this has on the timing of taxation is further explored below,41 but, in total, the result is to exclude from taxation the normal return to saving. Since the late 1990s, an alternative structure has also been available that will, under certain circumstances, provide the same net subsidy over time. These are Roth-style accounts that allow contribution of after-tax dollars but then exclude the gains on those assets from any further taxation, including at withdrawal.42 In addition to this subsidy, there is also a very narrow tax credit for low- to moderate-income Americans—the Saver’s Credit—that provides additional incentive to save. For those filing

41 See infra Part II.B.2. 42 [Add in cites on Roths.]

Figure 1: Distribution of Tax-Preferred Retirement Saving by Type

Source: Council of Economic Advisers, 2015.

Page 10: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

10

jointly, the credit is equal to 50 percent of amounts saved in a qualified retirement account up to $4,000 for those making under $37,000 in 2016. The credit rate steps down to zero above that income threshold, hitting zero at $61,500 of income in 2016.43 Importantly, the tax credit only goes to those without positive income tax liability, which means that most of those in relevant income ranges are ineligible.44 As a whole, tax preferred saving accounts provide significant tax subsidies toward retirement. The latest tax expenditure estimates from the Treasury Department suggest that the government each year spends around $80 billion subsidizing defined contribution plans and various forms of IRAs.45 (Importantly, this cost calculation is done in net present value and so captures the net subsidy over the life of the account. Further, it is only a subsidy relative to an income tax system. Some would argue that the fairest and most efficient system is a consumption tax, in which case this brings the system closer to that baseline.) The question is how successful this growing part of the retirement saving system is at increasing retirement saving. The very recent literature has been highly skeptical of tax subsidies as an effective tool. But, as the next section describes, this conclusion is over-drawn, and the leading study in recent years coming to this result shows why that is. Tax subsidies can increase saving. II. How Tax Subsidies Can Increase Saving In reaching the conclusion that tax subsidies do not increase saving, much of the most recent literature points to the ground-breaking work of Chetty, Friedman, and co-authors in the Denmark Study. In the words of Ryan Bubb and Rick Pildes—citing to the Denmark Study—there is a “a vast amount of public money spent for a minor increase in total private retirement savings,”46 and, based on this, they—like Chetty and his co-authors—go on to condemn tax subsidies relative to other tools like nudges and mandates. However, that broad interpretation of the Chetty paper’s finding is misguided. That is because tax subsidies can increase private saving in two ways, at least. For “active” savers, the Denmark Study itself showed tax subsidies doing so. When it comes to those who are

43 For parameters of the Saver’s Credit, see: https://www.irs.gov/Retirement-Plans/Plan-Participant,-Employee/Retirement-Savings-Contributions-Savers-Credit. 44 The Saver’s Credit provides only about $1 billion per year in subsidies—very small relative to the total retirement saving system. Table 1, https://www.treasury.gov/resource-center/tax-policy/Documents/Tax-Expenditures-FY2017-Tables-11-13-2015.xlsx. 45 Table 4, https://www.treasury.gov/resource-center/tax-policy/Documents/Tax-Expenditures-FY2017-Tables-11-13-2015.xlsx. 46 Bubb & Pildes, supra note 5, at 1631. For a similar interpretation in the popular press, see Annie Lowrey, Study Questions Tax Break’s Effect on Retirement Saving, N.Y. TIMES: ECONOMIX (Nov. 25, 2012, 10:43 PM), http://economix.blogs.nytimes.com/2012/11/25/study-questions-tax-breaks-effect-on-retirement-savings/ (“Every year, the government spends more than $100 billion on tax breaks to encourage Americans to save more for retirement. But a new study suggests such provisions may have little effect on the amount Americans save.”) See also Renu Zaretsky, TaxVox blog, http://taxvox.taxpolicycenter.org/2015/11/04/the-case-of-the-questionable-tax-incentive-women-and-retirement-savings/ (citing to Denmark Study as suggesting that tax incentives do not increase retirement saving).

Page 11: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

11

passive in their saving decisions, the Denmark Study finds no effect on private saving from the tax incentive. However, that result is very much dependent on the structure of the subsidy involved. As other empirical work has shown, it is possible for tax subsidies to increase private saving rates through passive channels, depending on their structure. This does not in fact mean that tax subsidies are necessarily a first-best tool for increasing saving. They come with some serious drawbacks, as described in Part III; however, the condemnation they have received recently—as doing close to nothing at all when it comes to private saving—seems misguided.

A. Empirical Evidence Before the Denmark Study There is a considerable empirical literature from previous decades and especially the 1980s and 1990s on whether tax-preferenced accounts increase saving. There were sharp disagreements in that literature, though all of the studies suffered from significant shortcomings in terms of the available data. These studies faced numerous analytical challenges including: calculating the degree to which savings in the accounts represented transfers from taxable accounts versus new savings, especially with limited data on savings (and borrowing) outside retirement savings system; distinguishing the effects of tax incentives versus underlying differences in saving preferences in comparing those participating in the accounts versus those who did not; and understanding the role of different motivations and approaches to saving.47 Some of the most pessimistic estimates (in terms of the effects of these subsidies) found that tax subsidies did not increase private saving much at all. For instance, Eric Engen and co-authors report regression results that “are consistent with the view that 401(k) plans have not raised private saving.”48 They also found “little evidence that IRAs substantially raised private saving.”49 However, other studies found significant, positive effects of tax subsidies on private saving—some, quite large. For instance, a study by James Poterba and co-authors found that “contributions to 401(k) plans represent largely net additions to personal saving.”50 Others argued that a middle ground between these pessimistic and optimistic findings might be closer to the right answer. Glenn Hubbard and Jonathan Skinner wrote, based on their own empirical work, that “there is good reason to believe that the truth lies between the extremes of ‘no new

47 See B. Douglas Bernheim & John Karl Scholtz, Saving, Taxes and, in ENCYCLOPEDIA OF TAXATION AND TAX POLICY, 353, 355(Joseph J. Cordes et al., 2nd ed., 2005) (explaining the challenges of calculating the effects of tax subsidies on saving). See also CHETTY ET AL., supra note 7, at 1 (“The ability to answer these questions has been limited by inadequate U.S. data on household saving. In particular, it is hard to know whether tax subsidies encourage families to save more, or simply shift money they would otherwise save into tax-advantaged retirement accounts.”) 48 Eric M. Engen et al., Do Saving Incentives Work?, BROOKINGS PAPERS ECON. ACTIVITY 85, 150 (1994). 49 Id. 50 James M. Poterba et al., Do 401(k) Contributions Crowd Out Other Personal Savings?, 58 J. Publ. Econ. 1, 27 (1995).

Page 12: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

12

saving’ and ‘all new saving’” in the tax-preferenced accounts.51 And, yet another study—by Orazio P. Attanasio and Thomas DeLeire—reached findings along somewhat similar lines. Attanasio and DeLeire concluded that just under half of contributions to IRAs represented new private saving. Of that amount, they found that most represented the participants saving the tax subsidy provided by the IRA and a small portion was induced saving above and beyond that.52 Notably, a number of these studies separately discussed the effects of tax subsidies on private saving and national saving (with national saving taking into account the possible ramifications of tax subsidies for government deficits). As described further below, it is the effect on private saving that is most relevant for determining whether tax incentives can help correct any biases in decision making that lead people to misallocate resources over their lifetimes. The issue of national saving goes to other questions, even if important ones.53 Thus, before the Denmark Study, there were a range of empirical results, with some suggesting that tax-preferenced accounts did not increase private saving and others suggesting very significant effects on private saving and others in between. But, all of these results were constrained by the limited data available.

B. The Denmark Study: What Happens in Denmark Doesn’t Stay in Denmark

1. Design and Results of the Denmark Study Denmark may not be a large country by most measures, but it looms large in the recent retirement saving literature. In the words of Jonathan Skinner (who had contributed himself to the earlier literature on retirement tax subsidies), the new work by Chetty, Friedman, and co-authors on saving incentives in Denmark represents a “landmark study.”54 This is for several reasons. First, Denmark features remarkably complete data on its citizens’ saving behavior. Not only does Denmark have data available on saving in retirement accounts; it has data on saving and borrowing in all other forms as well (except, apparently, home mortgages). Further, there were a series of policy interventions in the 1990s that could allow researchers to relatively cleanly identify the effects of changes in tax subsidies as well as, separately, changes in mandatory saving regimes. Denmark, thus, represented a golden opportunity for researchers to build on the earlier—and divided—empirical literature on retirement saving incentives. While only published in its final form in 2014, the Denmark Study is now widely cited,55 and has formed the basis of an 51 R. Glenn Hubbard & Jonathan S. Skinner, Assessing the Effectiveness of Tax Incentives, 10 J. ECON. PERSP. 73, 74 (Fall 1996). 52 Orazio P. Attanasio & Thomas DeLeire, The Effect of Individual Retirement Accounts on Household Consumption and National Saving, 112 Econ. J. 504, 531 tbl.9 (2002). 53 See infra note 71-72 and accompanying text. 54 Center for Retirement Research at Boston College, 401(k) Tax Break May Be Weak Incentive, SQUARED AWAY BLOG (Aug. 23, 2012), http://squaredawayblog.bc.edu/squared-away/research/401k-tax-break-may-be-weak-incentive/ (quoting Jonathan Skinner). 55 For instance, Google scholar suggests that it has been cited 131 times. [Get proper cite.]

Page 13: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

13

increasingly accepted conventional wisdom, despite the earlier disagreement on the topic—that tax subsidies do not increase saving to any significant degree. The Denmark Study specifically looks at the effects of three different policy interventions in Denmark. When it comes to tax incentives, the paper studies a reduction in the value of “capital pension” subsidies for those in Denmark’s top tax bracket in 1999. “Capital pensions” are similar—though not the same—as traditional tax-preferenced accounts in the United States. Danes could contribute pre-tax income to the accounts, earnings in the accounts were subject to a reduced tax rate, and distributions were taxed as “ordinary income” (though at a reduced tax rate). Thus, the value of the initial exclusion per Danish kroner (DKr) contributed was initially equal to the marginal tax rate, and, as of 1999, the top marginal rate was 59 percent. The policy intervention in 1999 capped he value of the exclusion at 45 for those in the top bracket—a reduction of 14 cents (ø re in Danish) per Danish kroner (DKr) in the value of the tax incentive. Or, to put it differently, contributions began to be taxed at 14 percent instead of their previous zero-rate.56 Based on the effect of this policy, Chetty, Friedman, and co-authors reach the following dramatic conclusion: each $1 of government tax expenditure increases saving by only one cent. The study describes how there are two distinct parts of the population in responding to the change in incentives. First, 81 percent of those affected did not change their pension contributions as a result.57 These are “passive savers,” who, if they had been active decision-makers, should have adjusted their pension contributions.58 Instead, they continued to contribute the same amount to the pensions as they otherwise would have. Second, there are the “active savers,” making up 19 percent of the affected population.59 The active savers reduced their pension contributions by shifting into a combination of other, non-taxable accounts and taxable accounts. Even when it comes to these “active savers,” the tax change has only a very small effect on saving, according to the Denmark Study’s authors.60 Summarizing these results, the authors write, “[t]he response in terms of increased household saving is quite limited because most individuals are passive savers and the evidence suggests that the small percentage of active savers tend to take advantage of the subsidy without increasing the total amount they save.”61 Chetty, Friedman, and co-authors compare this to the effects of certain automatic and mandatory saving arrangements. Specifically, they look at the effects when employees switch between employers which have different levels of employer retirement contributions and also the effect of a mandatory saving program that required Danes to contribute one percent of income to a mandatory savings plan. The net result in both cases was a substantial increase in wealth

56 Chetty et al., supra note 5, at 1185. [Check: The Denmark Study generally refers to the top rate as 60 percent, but it, in one places, describes it as 59 percent.] 57 Id. at 1144. 58 Id. (“[U]tility maximization would call for some nonzero change in contributions when prices change at an interior optimum….”) 59 Id. 60 Id. 61 CHETTY ET AL., supra note 7, at 5.

Page 14: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

14

according to the article.62 For instance, the mandatory intervention—requiring workers to contribute 1 percent of earnings to retirement saving accounts—was found to increase total saving by nearly that same amount, even for those who were already saving more than that.63 The findings above would seem to be more than enough to condemn tax subsidies as being almost entirely ineffective at actually increasing savings—with other tools being clearly superior. After all, what purpose is there to using tax subsidies if they increase saving by only 1 cent per dollar of tax expenditure, especially when other interventions can have such significant, positive effects? However, the case is not that clear cut. When it comes to active savers, the Denmark study finds only that the change in tax incentive did not change national saving—including the gain in revenue to the government. However, the study in fact does find that the active savers significantly reduced their private saving due to the reduced tax subsidy—meaning that there was real sensitivity to the incentive effect. Further, when it comes to the passive savers, the Denmark Study’s results are, in part, dependent on the structure of the particular tax subsidy studied, which is especially important in terms of the effects on passive savers.

2. Active Savers in the Denmark Study Chetty and co-authors characterize the “active savers” in their study as “tak[ing] advantage of the subsidy without increasing the total amount they save.”64 However, that is true in only a very particular sense. For the active savers, the change in tax subsidies apparently did not change national saving in Denmark by very much (taking into account revenue loss to the government). It did significantly change the amount they saved for themselves in terms of private saving. And, it is this latter metric that should matter in judging the degree to which tax subsidies can improve private saving decisions and offset bias in decision-making.

i. How Active Savers in Denmark Were Sensitive to Tax Subsidies To understand exactly how active savers were sensitive to tax subsidies, it is useful to walk through exactly what happened after the imposition of the new tax in Denmark on one type of tax-preferenced saving. This illuminates how the authors of the Denmark Study conclude that there is little to no change in national savings, even as there is a very real effect on private savings—with the latter perhaps being most relevant to retirement policy. The policy change imposed a 14 percent tax on contributions to “capital pensions,” whereas contributions could previously be made tax-free. According to the Denmark Study, the vast majority of active savers entirely exited the “capital pensions” as a result, shifting some funds to

62 Chetty et al., supra note 5, at 1143. 63 Id. 64 CHETTY ET AL., supra note 7, at 5.

Page 15: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

15

another tax-preferenced saving vehicle not subject to the tax (“annuity pensions”) and some funds to fully taxable accounts.65 Specifically, the following shifts occurred:66

• Average contributions to “capital pensions” fell by DKr 2,449 per person (from an average of 5,113).

• Of the average DKr 2,449 drop in capital pensions, DKr 1,195 shifted into annuity pensions (the other tax preferenced saving vehicle) and DKr 1,295 left tax-preferenced saving entirely.

• There was an average increase of DKr 515 in taxable saving.

• Due to these shifts, government revenues increased by DKr 777, as the DKr 1,295 shifted

out of the accounts was subject to the normal top tax rate of about 60 percent. Thus, there is a considerable reduction in private saving in the year the tax is imposed. Due to the 14 percent tax, average tax-preferenced saving falls by DKr 1,295 to be offset only by an increase in taxable savings of DKr 515—an average drop of DKr 780. Or, to put this somewhat differently, nearly 40 percent of the drop in tax-preferenced saving is offset by an increase in taxable saving and just over 60 percent is not. These figures, however, overstate the actual effect on the saving available for future consumption (as well as the increase in government revenue over time). The reason is that the saving in the taxable accounts would not be subject to tax at withdrawal, since these funds were taxed upfront. By contrast, the saving in the tax-preferenced account would be subject to tax at withdrawal, though at a reduced rate (of 40 percent). Thus, there is, in part, a temporary timing effect. According to Chetty and co-authors, the total tax subsidy for each dollar contributed to the capital pensions—combining both the lower tax rate at withdrawal and the lower tax imposed on gains in the meantime—was worth about 33 cents per DKr.67 Thus, in terms of the future consumption that can be financed, the real reduction in saving is about DKr 430. This is a real reduction in resources available for future consumption by these active savers (and a real increase in revenue for the government). The active savers choose to give up future consumption rather than now pay the 14 percent tax on contributions to capital pensions. Importantly, this measured effect should represent a “substitution” rather than an “income” effect—capturing the effect of a change in the relative price of future and current consumption rather than total resources available to the savers. As Chetty and co-authors argue, the vast

65 Chetty et al., supra note 5, at 1196 (“The vast majority of these 19.3% of individuals respond by completely exiting capital pensions.”) 66 For these effects, see id. at 1191 tbl.V, 1201 tbl.VI, 1209. 67 Chetty and co-authors estimate that the capital pension account provides a subsidy, in net present value, of 19.4 cents per DKr as compared to fully taxable accounts—taking into account the 14 percent tax imposed on contributions starting in 1999 for those in the top tax bracket. Thus, before the 14 percent tax was imposed, the subsidy was worth about 33 cents per DKr contributed. See id. [Check this—may not be a correct interpretation of Chetty.]

Page 16: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

16

majority of the active savers shifted their saving so as to entirely avoid paying the 14 percent tax. As a result, the change in behavior comes from the marginal incentive rather than an effect on resources available.68

ii. Why the Denmark Study Suggests Active Savers Did Not Change Total Savings However, despite these effects, the authors of the Denmark Study indicate that active savers are not “increasing the total amount they save” due to tax subsidies. That is because the authors focus on the effects of the policy intervention on national saving in Denmark rather than private saving.69 Due to the 14 percent tax, the active savers choose to reduce their future consumption (rather than pay that tax upfront). But, the reduction in future consumption happens to be nearly offset one-for-one by an increase in government revenue. In the year of the intervention, the shift from tax preferenced saving to taxable saving means that private saving drops by an average drop of DKr 780 (though some of this is a timing effect as described above). At the same time, government revenues increase by DKr 777 (though some of this is a timing effect as well). In terms of national saving, there is very little change assuming, as the authors do, that the government revenue goes toward reducing the deficit (which may or may not be true). The active savers have effectively given up a tax subsidy that could finance future consumption, and that goes directly into the government’s coffers. As a result, Chetty and co-authors suggest that there is a vast amount of public resources spent on achieving very little increase in saving. They calculate the revenue gain to the government from the 14 percent tax and behavioral responses (shifting out of tax-preferenced accounts) as generating an average of DKr 883 in net present value terms across the treatment group. They then note that national saving drops by only DKr 3 (as private saving drops by DKr 780, offset by a revenue increase of DKr 777). Dividing the DKr 3 by DKr 883, they conclude that total saving fell by “less than 1 cent per DKr reduction in tax expenditure on the subsidy.”70 However, that headline finding is presented in terms of national saving. In terms of private saving, the effects are different. Another way to present what happened is that a 14 percent tax on what were originally DKr 5,113 of contributions—a tax that on a static basis would raise DKr 715 from both active and passive savers (with about half of that coming from active savers)—caused an effective reduction in private saving of DKr 430 among the active savers.

68 See Chetty et al., supplemental appendix, at 5 (“The 19.3% of individuals who respond to the subsidy change do so primarily by exiting capital pensions entirely, and thus their tax liabilities are unaffected by the change in the tax rate on capital pensions.”). [Find the other place they say this too.] 69 See Chetty, supra note 5, at 1202 (explaining that they study the degree to which tax-preferenced saving crowds out other saving in terms of national saving rather than in terms of private saving). 70 Id. at 1209.

Page 17: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

17

iii. Why Private Saving Matters Most for Retirement Policy For purposes of determining the effectiveness of policy interventions on correcting misallocation of resources across people’s lifetimes, the right metric is private saving rather than national saving. This in fact has led some scholars to misinterpret the Denmark Study, assuming that the authors were looking at private saving. For instance, Bubb and Pildes—in relying on the Chetty findings to criticize tax subsidies—explicitly frame the Chetty findings in terms of private saving when that is not what the study purports to measure.71 The issue of how the nation should distribute its consumption (as opposed to how particular individuals should distribute their consumption) is a deep and important question, but it is separate and apart from the issue of whether and how much to correct private saving.72 For instance, it is entirely consistent to believe that savers are misallocating resources across their lifetimes by not saving enough for retirement (and other events) and to believe that national consumption is still being appropriately divided among generations. The point is that there is not a necessary relationship between these policy judgments. Further, in assessing the effect of a policy on national saving, a series of additional questions must be engaged including whether policymakers would use revenue to reduce the deficit or, for instance, finance government consumption or other tax cuts with any revenue saved. In sum, in assessing the degree to which tax subsidies can correct behavior and offset the effects of biases, what matters is the degree to which private saving is sensitive to marginal incentives. And, in the case of the active savers in the Denmark Study, the results suggest some sensitivity to tax subsidies at the margin.

4. Passive Savers in the Denmark Study For the passive savers, by contrast, there was a very different result. For this four-fifths of the treated population, there was no change in private saving due to the limitation on the value of the deduction for contributing to capital pensions. This was even as the government generated more revenue than it did before from this population. In other words, private saving remained the same, even as government revenues increased.73 Notably, in the case of passive savers, Chetty and co-authors may actually understate the degree to which the government expends revenue for no increase in private saving in their summary measures. In particular, they describe their summary figures as focused on the effects of tax incentives on national saving.74 And, that is exactly what they do for the active savers.

71 Bubb & Pildes, supra note 5, at 1631 (interpreting the Denmark Study as suggesting that there “is a vast amount of public money spent for a minor increase in total private retirement savings.”) [Get additional cites.] 72 See, e.g., Daniel Shaviro, The Long-Term Fiscal Gap: Is the Main Problem Generational Inequity, 77 Geo. Wash. L. Rev. 1298, 1319-28 (2009) (discussing key issues in judging intergenerational equity and concluding that it is difficult to know how much earlier generations should save for future generations). 73 Chetty et al., supra note 5, at 1209 (noting how, for passive savers, the subsidy “is an inframarginal transfer that has no effect on the behavior of passive savers.”) 74 Id. at 1206.

Page 18: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

18

However, somewhat surprisingly, they do not do the same for the passive savers (in the very same calculation). That is because there is apparently no change in private saving, even as these workers paid more taxes to the government than they would have otherwise. The Denmark Study does not say this, but their results imply that this must have come from reduced consumption by those passive workers (since the taxes do not reduce either retirement saving or fully taxable saving according to their estimates). That suggests the reduction in the tax subsidy in Denmark increased national saving (assuming the revenue goes to reducing the deficit)—and did not merely hold national savings constant. In terms of the broader policy implications, the results in Denmark illustrate how tax subsidies can be completely ineffective at raising private saving rates among passive savers. This is an important finding, and rightly raises questions whether tax subsidies are effective at all, at least with regard to a significant segment of the population. Of course, the contention of this paper is that the broader conclusion—that tax subsidies do not increase private saving—has been overstated. Unlike with active savers, this is not because of a misinterpretation of what happened in Denmark; the tax subsidies were in fact ineffective at getting the passive savers to save more according to the results in the Denmark Study. Rather, it is because the results in Denmark for passive savers depend on a particular structure for the tax benefit—in which much of the benefits of the tax incentive are delivered outside the saving vehicle rather than directly into the vehicle. Part II.B. explains how such tax incentives can, using a different structure, potentially increase saving, even among savers. The caveat to this—also explored in that part—is that current saving subsidies may perform even worse when it comes to passive savers than in the Danish study because of the combination of where tax savings are placed and because of how they change the timing of taxation. In short, when it comes to passive savers, structure matters and the findings of the Denmark Study are dependent on that.

5. Comparison to Mandates As described above, the Denmark Study not only looked at the effects of tax incentives but also compared these to mandatory approaches to increasing saving, with the authors concluding that the mandates are significantly more effective. However, this conclusion, with regard to active savers, is not a correct reading of their own findings and, with regard to passive savers, may be dependent on the specific structure of the tax incentive being studied. The Denmark Study specifically looked at two other instruments, both of which involved forms of at least partially mandated saving. The paper studied variance in employer retirement contributions when employees switched jobs and a mandatory saving program that required Danes to contribute one percent of income to a savings plan. Both instruments were found to be effective at generating higher private saving. The co-authors found that “individuals’ total saving rises by approximately 80 cents…when they move to a firm that contributes 1 Danish kroner…more to their retirement account even if they could have fully offset the increase.”75 For

75 Id. at 1143.

Page 19: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

19

the government imposed saving requirement, they found that requiring contribution of one percent of earnings caused private savings to rise by nearly that amount even for those already saving more than that and, thus, had an ability to offset the effect. In other words, 1 DKr of resources put toward saving accounts using these mandatory mechanisms generated approaching 1 DKr of additional saving.76 However, tax subsidies can potentially have a comparable effect. For active savers, a 14 percent tax on retirement saving led them to reduce the value of their private saving by even more than that—as they were willing to give up the still-valuable subsidy provided by tax-preferenced accounts. Thus, there is no clear advantage of a mandate for this group when it comes to increased saving. With regard to passive savers, these mandates are clearly more successful in Denmark in raising private saving—since the tax incentive appeared to have no effect whereas the mandates did. However, as the next part describes, this result is likely a function of the structure of the particular tax subsidy. It is imaginable that tax subsidies could be structured to have similar effects on saving as the mandate—increasing saving by the amount of the subsidy devoted.

C. The Importance of Subsidy Design For all savers but perhaps especially for passive or unsophisticated savers, the design of tax subsidies can affect the amount saved. First, it might increase the amount saved if the subsidy were entirely directed into the saving account or otherwise directly increased the value of those savings. With this design, the subsidy provided could “stick” in the saving account either due to inattentiveness or confusion of the savers. Thus, the result in Denmark—certainly among passive savers—is not generalizable; it is specific to the type of tax incentive provided, since that tax incentive relied on active and informed decision-making to increase saving. A tax subsidy delivered directly into the saving account would not require any active decision-making to increase saving. Second, and relatedly, it is possible that saving incentives could be even less effective than suggested in the Denmark Study. That is because traditional tax-preferenced accounts both deliver tax savings outside the savings account and because these incentives change the timing of taxation by decreasing taxes upfront and increasing taxes in the future. The combination of the location and timing effect is not tested in the Denmark Study, and it could plausibly lead passive savers with less resources in the future than they otherwise would have. Thus, the Denmark Study’s results are not entirely generalizable. There are ways that tax subsidies can be structured to increase saving rates among passive savers, and there are ways that they could in fact be even less effective than suggested by the Denmark Study.

76 Id.

Page 20: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

20

1. Location of the Subsidy The specific tax subsidy at issue in Denmark was delivered outside the saving account. When Denmark reduced the value of the subsidy for contributing to capital pensions by 14 cents per DKr, that tax was paid by funds held outside the retirement accounts. For passive savers, it apparently made no difference in the amount that they saved. In an alternative structure, that additional tax could have been paid by funds within the retirement accounts. For instance, there could have been a 14 percent withholding tax imposed on any contributions to the retirement accounts to be collected by the account administrator—so that a $100 pre-tax contribution would, after-tax, become an $86 contribution. Of course, that may have been administratively messy to do within the Danish system, but the point is that it is possible. And, in fact, it is entirely imaginable that a retirement account system could be explicitly designed to have such a direct subsidy—in fact, one type of account in the United States essentially does, as described in the next section. If people were active, rational, and well-informed in making saving decisions, this design choice should make no difference. Money is fungible. And, irrespective of whether the tax incentive is delivered outside or inside the account, there should be no difference in the amount of saving. If delivered inside the account, people should adjust and reduce the amount they put in as compared to a system in which the tax savings are delivered outside the account. But, of course, the framing of saving decisions does make a significant difference in the amounts people save. That is among the central findings of the retirement literature in recent years, including the Denmark Study. The Denmark Study does not directly shed light on what would have happened if the tax subsidy reduction came from the retirement account rather than outside it. But, it is possible to conjecture that many passive savers may have seen their saving in capital pension accounts fall by 14 percent if the tax had been taken directly from the savings accounts. If those savers were truly passive, they would have continued to contribute the same amount as they had before on a pre-tax basis—with the actual amount of saving falling on a post-tax basis. To be clear, the effect would not have resulted through a traditional price signal but through a lump sum effect, as savers would not have been paying attention to the amount of subsidy delivered into the retirement account.

2. Timing of Taxation Another key and related factor in the design of tax subsidies is the timing of taxation. The Denmark Study also looked at a very particular change in terms of the timing of the tax subsidy. Specifically, the Denmark Study explored what happened if the value of the tax-preference were reduced by increasing taxes upfront (financed outside the savings account), at the time the contribution was made to the account. However, in their basic structure, traditional tax-preferenced accounts not only involve a tax subsidy; they also involve a change in the timing of taxation, with potential detrimental effects on saving especially for passive savers.

Page 21: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

21

As described in Part I, traditional tax-preferenced retirement accounts in the United States allow an upfront deduction for saving to this account; returns on saving then accrue tax-free in the account; and, finally, all qualified withdrawals are taxed as income. Compare this to a fully taxable account, where there is no upfront deduction; returns are taxed as they are realized; and there is no taxation at withdrawal from the account, except to the extent gains are then realized. Relative to the fully taxable account, there is an overall reduction in taxation over time on amounts that are saved; this is the incentive. This subsidy comes from, first, any differential between the tax rate upfront and the back-end—so, to the extent the rate at withdrawal is lower than the rate upfront (as in Denmark, for instance), this generates a subsidy. Second, the subsidy comes from no taxation of the returns on saving as they accrue. In addition to there being a subsidy, however, there is a change in the timing of taxation. Relative to a taxable account, traditional tax-preferenced accounts feature less taxes paid upfront than otherwise would be the case and more taxes paid at the back-end upon withdrawal. One way to conceptualize this is as a loan. In particular, relative to a taxable account and assuming constant tax rates, the government makes an interest-free loan to the saver equal to the amount of taxes that would have been owed on income deposited in or earned in the savings account for the year—with the loan then due upon withdrawal. The fact that the loan is interest free generates a subsidy; the fact that there is a loan changes the timing of taxation.77 This change in timing may affect the amount of resources available later depending on what savers do with the proceeds of the loan—and it is plausible that this could lead to a reduction in resources for later consumption relative to a world without such accounts. The problem, in part, goes back to where dollars are delivered. In traditional tax preferenced accounts, the initial proceeds of the loan are delivered outside the savings account. Each dollar contributed to a traditional tax-preferenced account can reduce the taxes that must be paid throughout the year—in the form of withholding on wages and estimated tax payments—and, then, results in a lower final tax bill at filing time. The lower taxes—the equivalent to an interest free loan—are not delivered directly into the savings account. And, if a passive saver does not put a significant share of those proceeds toward saving, it is in fact possible that the interest-free loan leads to a reduction in resources available in the future—since the loan has to be paid back. The Denmark Study does not address the effects of this shift in timing for passive savers. To repeat the conditions which it studies: It looks at what happens if those savers must pay more taxes upfront on their saving with those taxes being financed outside the savings account.

3. The Case of Roth-Style Accounts The importance of both place and time when it comes to the effects of tax incentives is illustrated by a recent study of Roth-style savings accounts in the United States.78 Roth-style saving accounts differ from traditional tax-preferenced saving accounts in that they effectively increase

77 [Get citations of sources comparing to interest-free loan.] 78 John Beshears et al., Does Front-Loading Taxation Increase Savings? Evidence from Roth 401(k) Introductions (Nat’l Bureau of Econ. Research, Working Paper No. 20738, 2014).

Page 22: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

22

the value of amounts saved in the accounts (like a direct subsidy would) and do not involve trading off lower taxation upfront for higher taxation later. Probably because of this structure, the introduction of Roth-style accounts is found to increase private saving relative to a system without them.79 As described previously, in terms of the basic structure, Roth-style accounts do not provide any upfront deduction for savings. The contributions to the savings accounts are made on a post-tax basis. Returns to saving then accrue tax-free in the account, and, finally, all withdrawals from the accounts are tax-free. Under a number of conditions, traditional and Roth-style accounts are exactly equivalent in terms of the value of the subsidy that they provide.80 Among other things, this assumes tax rates are constant and that the investments in the retirement accounts are not expected to earn above the ordinary rate of return. Congress created and expanded Roth-style accounts not so much for any laudable policy reason, but, instead, largely as a budget gimmick. Since the tax benefit provided by the Roth account is backloaded, it would appear to cost much less within the normal 10-year budget window used to estimate the cost of legislation than a traditional retirement account, despite delivering a similar subsidy, and Congress has repeatedly taken advantage of this scoring quirk.81 However, Roth accounts deliver the tax subsidy differently. Relative to a system that taxes realized gain, the subsidy effectively increases the amount of future consumption that a given amount of saving can finance. To the degree that the realization coincides with withdrawal from the account, it is akin to a direct contribution by the government in the amount of the taxes saved increasing the value of those withdrawals from the account. So, while Roth accounts do not technically involve the government depositing funds into the saving accounts, it is a close parallel. In terms of timing, Roth accounts do not—unlike traditional tax-preferenced accounts—involve any upfront reduction in taxes and increase in future tax liability. There is not the equivalent of a loan being made. Instead, the tax savings are entirely back-loaded. This is important because this means that it eliminates the possibility that people who are inattentive or confused would spend the proceeds of the loan upfront not realizing that there is essentially a loan to be paid back later.

79 Id. at 2 (describing how workers contributed the same amounts to Roth-style accounts as they had to traditional accounts, despite the fact that the Roth savings was greater on an after-tax basis). 80 The equivalence between a Roth-style and a traditional tax-preferred savings account is a version of the Cary Brown theorem, which describes how allowing immediate expensing of an investment would—under certain conditions—be equivalent to not allowing the upfront deduction but then exempting any further gain on the asset from taxation. See E. Cary Brown, Business-Income Taxation and Investment Incentives, in INCOME, EMPLOYMENT AND PUBLIC POLICY: ESSAYS IN HONOR OF ALVIN E. HANSEN 330-416 (1948). 81 See Cheryl D. Block, Budget Gimmicks, in FISCAL CHALLENGES: AN INTERDISCIPLINARY APPROACH TO BUDGET POLICY, 36-67 (Elizabeth Garrett et al. eds., 2007) (describing budget gimmick of Roth-style accounts). [Get additional cites on budget gimmick.]

Page 23: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

23

These differences in structure appear to make a difference. A new study by John Beshears and a number of co-authors documents how workers effectively save more when Roth-style savings accounts are introduced than with traditional saving accounts alone being available.82 Beshears and co-authors compare the contribution rates of newly hired employees at eleven firms before and after introduction of an option to contribute to Roth-style savings accounts (specifically, a 401(k) Roth-style saving account). They conclude that “introducing a Roth option to the 401(k) does not significantly change total contributions to the 401(k). An unchanged total contribution rate translates into higher after-tax retirement consumption if some of those contributions are directed to the Roth and the balances are kept in the Roth for a long enough period of time.”83 In other words, the introduction of the Roth-style account effectively increased private saving and the resources available in retirement, despite the fact that it provides a similar, if not the same, subsidy value as a traditional account. That is because the employees continued to contribute the same dollar amount, despite the fact that the government would be kicking in a significant subsidy upon withdrawal, as the returns would go untaxed at that point. There are some possible rational explanations for why people might react this way, but Beshears and co-authors dismiss most based on the empirical data and suspect that it has more to do with confusion, neglect, and non-rational heuristics.84 Looking at the behavior of the new employees and surveys that they conducted, they conclude that one contributing factor is likely some combination of “ignorance and/or neglect of the 401(k) tax rules.”85 Another factor—which has little or nothing to do with there being a tax subsidy—may be a partition effect. In particular, there tends to be a bias toward dividing resources equally among options. By creating one more option for saving, this may tend to increase the total amounts that people save.86 The possibility of confusion and neglect of the tax rules is what is particularly relevant for this paper. The Roth accounts are structured in such a way that the “confusion and/or neglect” leads the tax subsidy to be essentially saved for future consumption—in contrast to much of the subsidy for traditional tax-preferenced saving accounts. The broader point is that Roth-style accounts demonstrate that tax incentives can have a positive effect on private saving, if they are designed to do so. The subsidy may not have an effect by making saving cheaper than it otherwise would be, but, instead, by delivering the subsidy in a way that it is saved through “confusion and/or neglect.” And, it shows how the result in the Denmark Study—that tax subsidies do not increase the private saving of “passive savers”—is not generalizable to all structures.

82 See generally Beshears, supra note 78. 83 Id. at 9. 84 Id. at 5 85 Id. at 15. 86 Id. at 17-18.

Page 24: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

24

III. Why Not Tax Subsidies: A Tentative Case The previous part describes how tax subsidies can in fact be effective at increasing saving, both among active and passive savers. So, a blanket condemnation of these instruments as entirely ineffective at their desired aim is not backed by the empirical evidence. However, there is a case to be made against using tax subsidies as a first-best policy tool, whatever their structure—comparing them to the alternatives of defaults or saving floors. But, the case is more tentative than is sometimes made out, especially if they are combined with widely applied (and well-considered) defaults. It is, in other words, a “close call,” and the current literature—by wrongly concluding that the empirical evidence demonstrates that tax subsidies do not increase private saving at all, is not focused on the right trade-offs. There are two key problems with tax subsidies: First, tax subsidies, however structured, will tend to do more to increase saving among those who save more than those who save less—and so are mis-targeted in terms of how they redistribute resources across people’s lifetimes. Second, tax subsidies will tend to redistribute resources across the population—that is, redistributing resources between people—in ways that could be deleterious. With that said, tax subsidies come with important possible benefits: they can potentially help to correct the behavior of active savers (unlike a default) and can do so without the welfare loss associated with imposing a single saving solution on a possibly heterogeneous population (unlike a saving floor). To the extent that large tax subsidies are needed to correct the behavior of active savers—which seems plausible based on the empirical evidence, this

A. Increasing Saving Most of Those Who Need It Least Even if a tax subsidy is structured in such a way as to increase saving among passive savers, there remains a fundamental problem: the subsidy is based on the amount being saved. So, those who save more, get more put into their savings account (if the subsidy is directed in that way). And, those with greater savings are less likely to need the behavioral intervention than those who save less.87 The degree of the problem is illustrated by a simple fact: 45 percent of working age households do not own any assets in a retirement account.88 The ratio is not much lower even for older adults. For those aged 55-64, the percent with retirement accounts stood at 41 percent.89 Retirement account savings are particularly relevant since any tax subsidy presumably would only go into a specialized account—like a retirement account—where there are restrictions on

87 For a similar point made in the context of employer matching credits, see Ryan Bubb & Patrick Corrigan, A Behavioral Contract Theory Perspective on Retirement Savings. [get additional citation info.] 88 NARI RHEE & ILANA BOIVIE, NATIONAL INSTITUTE ON RETIREMENT, THE CONTINUING RETIREMENT SAVINGS CRISIS 8 fig.5 (2015). 89 Id.

Page 25: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

25

withdrawal. Otherwise, the system would be subject to gaming as people could take advantage of the subsidy and then withdraw the funds immediately to be consumed. But, even broadening the definition of savings to all savings, balances are very low or non-existent for many families, outside of the Social Security system.90 So, however defined, there is a significant share of the population with very low or non-existent cash savings. Of course, a tax subsidy—even if structured to take advantage of passive behavior—can do nothing for families with no savings and little for families with very low savings. There is nothing to match after all, and so a tax subsidy could do little to change the outcomes for them. The tax subsidy would increase saving more for those who save more. But, even this could prove problematic. Even as most people appear to be completely unresponsive to tax incentives, there is substantial variance in how much people save, even within the same income class.91 By increasing the saving most of those who already save more, it is possible that a matching incentive could produce over-saving for some. Saving, after all, is not an unalloyed good. In fact, if the tax incentive were optimized for low savers—to the extent that is even possible given their very low saving—the incentive might be expected to in fact produce just such over-saving among those who already save more. The bottom line is that a tax subsidy will, by its nature, be poorly targeted. Given the correlation between low saving rates and low income,92 the system can be adjusted to provide more generous tax subsidies toward those at the bottom, but, even that has its limits given variation within income classes and given that many may simply have little to no savings to match.

B. Detrimental Distributional Consequences The second problem with tax subsidies is that they are necessarily redistributional—redistributing resources from those who save less to those who save more. So, it is not just a question of the tax subsidies being a poor intervention to adjust people’s behavior and, thus, failing to appropriately redistribute resources within a person’s lifetime—the problem discussed in the prior section. The tax incentives also redistribute resources across people and in ways that may be deleterious as a distributional matter and that cannot be readily reversed (without reversing the subsidy itself). The tax subsidies for saving will provide a net benefit to those who save more as compared to those who save less. This is unless the tax subsidies are financed with an offsetting tax on saving, which would essentially undo the tax subidy and make this a wash. Whether this is good or bad as a distributional matter depends both on the overall treatment of saving in the tax system and one’s view of whether saving should or should not lead to heavier taxation across a person’s lifetime. 90 [Get cites.] 91 See infra note 99 and accompanying text. 92 See generally Karen E. Dyan et al., Do the Rich Save More?, 112 J. Pol. Econ. 397 (2004) (finding association between low saving rates and low lifetime income).

Page 26: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

26

The current tax system imposes greater taxes on those who save than those who do not—since it taxes, if somewhat inconsistently, the return to saving. This has potential consequences for both efficiency and distribution. As an efficiency matter, there is a preference for current consumption over future consumption that could distort decision making by discouraging saving, though the effect may not be all that large.93 As a matter of distribution, the tax system distinguishes based on this choice with regard to saving—which may or may not be appropriate from a “fairness” perspective and depends, in part, whether saving should be taken as an independent indicator of ability to pay. There is a long and storied literature on this particular tax policy choice; whether to tax the return to saving is the core distinction between income tax and consumption tax systems—a topic of considerable debate.94 Those who support income taxation—this author included—believe that there are good reasons to impose a heavier tax burden on those who save more. For instance, saving may be a useful marker of ability and, thus, should prompt heavier taxation for that reason—even if there is a trade-off with efficiency.95 Those who support consumption taxation tend to believe the opposite, and argue that those with the same reported lifetime income, irrespective of when they choose to consume it, should be taxed the same for reasons of both efficiency and fairness. In an important way, both income and consumption tax advocates should be able to agree that adding tax incentives as a way to encourage saving in a system that is otherwise optimal would produce negative distributional consequences. Income tax advocates believe that there should be a positive tax rate imposed on returns to saving, and deviating from that would come with negative distributional consequences. Consumption tax advocates believe that there should be no positive tax on ordinary returns to saving, and deviating from that—as in providing an additional tax incentive beyond that to save—would also have negative distributional consequences. Table 2 summarizes. The point is that a consumption tax advocate might support eliminating any taxation of returns to saving—providing a tax subsidy relative to the current system—since, from their perspective, this is right as a distributional matter, and they probably would support doing so irrespective of whether this is an effective means for increasing saving. But, the consumption tax advocate would also see any additional tax subsidy beyond that as a distributional harm (even if it might improve decision-making by correcting a bias). And an income tax advocate would see any tax subsidy that shifts the tax burden from savers to non-savers, relative to an income tax system, also as a distributional harm. So, from the perspective of this income tax supporter, tax subsidies are not just poor ways of correcting the behavior of people—and redistributing resources within a person’s lifetime. They

93 Cite to Chetty and others. But see Feldstein response to these types of arguments. 94 For examples of this, see Joseph Bankman & David A. Weisbach, The Superiority of an Ideal Consumption Tax Over an Ideal Income Tax, 58 Stan. L. Rev. 1413 (2006) (making case for superiority of a consumption tax over an income tax); Daniel Shaviro, Beyond the Pro-Consumption Tax Consensus, 60 Stan. L. Rev. 745 (2007) (questioning the conclusion that consumption tax is necessarily superior to an income tax). 95 See Shaviro, supra note 94, at 784-86 (describing how saving may provide information about ability).

Page 27: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

27

are also detrimental in terms of the distribution of resources across people, as they shift the tax burden from savers to non-savers. It is important to note that this redistributional consequence cannot be offset. Louis Kaplow famously shows that, in correcting an externality, optimal taxes (or subsidies) should be set putting to the side any distributional concerns, under certain conditions.96 And, the same is true for correcting “internalities”—that is, self-defeating preferences like when it comes to retirement savings (as opposed to market failure).97 That is because the income tax schedule can be adjusted to offset distributional harms, again in certain circumstances. However, as Kaplow notes, that finding is contingent on the condition that there is not heterogeneity at each income level with regard to relevant preferences; if there is heterogeneity, then that could bear on the desirability of a corrective tax—since the distributional consequences within an income class could not be offset.98 Here, it should bear on the wisdom of tax subsidies since those subsidies would favor those who tend to save more relative to those who save less. Just to emphasize the point: there is substantial heterogeneity in savings controlling for income, and so the distributional harm has the potential to be significant. For instance, looking at families with a head of household aged fifty-one to sixty-two in 1992, Steven Venti and David Wise find vast disparities in wealth even for families with similar lifetime incomes. For instance, for those in the middle of the lifetime income distribution (the fifth decile), those in the 90th percentile of wealth had thirty-five times the wealth of those in the 10th percentile.99 And, just to emphasize, this controls for lifetime income.

Table2DistributionalEffectofTaxIncentivetoSave

WhatIstheExistingSystem?NormativePerspective IncomeTax ConsumptionTax

IncomeTaxSupporter WrongDirection WrongDirection

ConsumptionTaxSupporter RightDirection WrongDirection To be clear, the potentially detrimental distributional consequences of tax incentives are not alone reason to condemn them. They have to be weighed against the welfare benefits that result from correcting saving decisions in this way. However, there is reason to think that the trade-off with regard to saving incentives may not be a good one. That is because the relevant behavior is not all that sensitive to the tax incentives—even if it is more sensitive than many now say. And,

96 See Louis Kaplow, On the (Ir)Relevance of Distribution and Labor Supply Distortion to Government Policy, J. ECON. PERSPECTIVES, Fall 2004, at 159, 164-66. 97 See Donald B. Marron, Should We Tax Internalities Like Externalities? 10 (Urban-Brookings Tax Policy Center Working Paper, 2015). 98 Kaplow, supra note 96, at 166. 99 Venti & Wise, supra note Error! Bookmark not defined., at 34. [See if updated data and check whether it includes Social Security.]

Page 28: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

28

that means that there would be a considerable amount of redistribution for any given change in behavior. Thus, it should be a significant concern for that reason.

C. Defaults and Floors As Alternatives It seems clear that tax subsidies alone are not the best way to correct saving behavior—certainly as a prime tool for doing so for most Americans. While they can affect savings, they remain poorly targeted tools even in that case—helping those save more who need it least and redistributing resources from non-savers to savers. The question then is what other tools could be used to increase saving among most of the population if not tax subsidies, and this brings us to nudges and floors.

1. Defining a Nudge and a Floor Nudges have been the recommended approach—or at least the focus—of much of the recent economics literature. Retirement saving is one of the touchstone fields in behavioral law and economics.100 Behavioral law and economics scholars have tended to advocate policies intended to change framing of decisions, but in ways that preserve choice for decision-makers and do not significantly change economic incentives. This is what Sunstein and Thaler, the originators of the term, define as a “nudge.”101 In the retirement context, nudges tend to focus on: (1) increasing saving rates and (2) improving investment choices.102 While both are clearly important for augmenting resources available later in life, this article is focused on the first of these challenges. By contrast to a nudge, a floor sets a minimum below which saving cannot fall, or cannot fall easily for many. The United States already has a version of this in the form of Social Security. Social Security, for instance, replaces forty to fifty percent of pre-retirement earnings for those in the middle quintile of lifetime earnings (with a higher replacement rate for those with lower incomes and a lower rate for those with higher incomes).103 And Social Security does so automatically. There is no ready way to opt out of Social Security payroll taxes and, later, benefits. A floor like Social Security is a type of mandate, but it is asymmetric. To the extent savers want to exceed the floor, they can, but, to the extent they want to go below the floor, they cannot as easily. In contrast to this, a full mandate would set a saving rate to which all must adhere; saving could neither exceed nor drop below that floor. Of course, a symmetric mandate is a theoretically possible tool—mandating that people save a certain amount and no more or no less. But, there are relatively few concerns about people over-saving based on self-defeating irrationality, and so it is hard to develop a strong justification for a symmetric mandate. Further, 100 Bubb & Pildes, supra note 3, at 1599 (noting how the shift toward automatic enrollment is “the single most celebrated achievement of BLE [behavioral law and economics]”). 101 RICHARD THALER & CASS SUNSTEIN, NUDGE XX (2009). 102 Bubb & Pildes, supra note 3, at 1614-15. 103 CONG. BUDGET OFFICE, 2015 LONG-TERM PROJECTIONS FOR SOCIAL SECURITY: ADDITIONAL INFORMATION 16 exhibit 10 (2015).

Page 29: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

29

it would be administratively challenging to do so. After all, while the government can with relative ease require people to save at least a certain amount of money in an account or system which it tracks, it is much more challenging—and beyond any information-reporting systems the U.S. government now has in place—to track all saving behavior by a person across all possible saving (or dissaving) vehicles. To be clear, even a floor will, as an administrative matter, not be entirely firm. If the government mandates that a person save “X” in a government monitored account or system, it is possible that a person still saves less than “X” by borrowing—since the government is not able to easily regulate that either. But, many Americans are credit constrained, at least to some degree, and so a floor set for a specific saving vehicle will be hard to overcome for many.

2. Nudges, Floors, and “Passive” and “Active” Savers For completely passive savers, both nudges and mandates should have similar effects. If a person is completely passive, then a default set to encourage a certain level of savings in a given saving vehicle will be little different from setting a floor that requires at least that amount of savings in the same vehicle. The passive saver will save that same amount in either system. And, the Denmark Study suggests that most savers are passive at least in the face of the types of changes studied in that article. The study concludes that, in the face of the three interventions studied in Denmark, at least four-fifths were passive.104 To be clear, and as Bubb and Pildes so convincingly argue, optimally setting the default or mandate for these passive savers is both important and challenging.105 It requires analyzing what would maximize welfare across a heterogeneous population. Nonetheless, in the case of passive savers, the analysis is the same, whether implementing a default or a floor. Both will result in truly passive decision-makers saving that amount. So, the differences arise when it comes to active decision-making. A nudge does not as dramatically affect active decision-makers as it does passive decision-makers; to be clear, it still might change their decision-making through an “anchoring” effect. But, a mandate can have greater influence. And a mandate can help active decision-makers to overcome biases to a degree that nudges cannot. It can also hurt them by forcing them to choose an option that is misaligned with their actual self-interest, putting biases to the side. The losses become greater to the extent the population is more heterogeneous in its underlying saving preferences. Of course, to the degree that average saving rates are too low for the active population as a whole, this suggests an intervention like a floor would come with a significant benefit—that of helping to address the bias. The trade-off is that a floor is likely to lead some to save more than they should, even as it leads others to save more as they should. There is good reason to think that savings behavior is widely plagued not just by passivity but by biases among active savers. There is evidence that a set of biases lead to under-saving among

104 Chetty et al., supra note 5, at 1143-44. 105 Bubb and Pildes, supra note 3, at 1621-25.

Page 30: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

30

those actively considering the choice of how much to save.106 This presents a plausible case for a floor and not just a nudge. A default cannot do much to correct these biases when they are proceeding through active decision-making. This is not enough to absolutely conclude that a floor, like Social Security, is necessary or should be expanded. But, it does suggest that floors may have important roles to play in encouraging additional saving; it is the only instrument of these two that can correct the behavior of biased active savers.

D. Tax Subsidies as Complements to Nudges? Tax subsidies alone are not a first-best tool for increasing saving. But, this does not in itself demonstrate that tax subsidies could not work as complements to other tools and perhaps especially nudges as a first-best correction to behavior. The scholars Jacob Goldin and Nicholas Lawson have recently put forward a compelling theoretical framework to justify the combination of nudges and tax incentives as a general matter for correcting bias—and as an alternative to imposing mandates.107 However, in the retirement context at least, there are reasons to question this combination—echoing earlier objections to tax subsidies: first, even active savers are relatively insensitive to changes in tax subsidies, meaning that any such incentives would have to be large to offset bias; and, second, tax subsidies could lead to a potentially detrimental redistribution of resources. However, this is a “close call”—and deserving of additional exploration—especially in light of a correct reading of the evidence with regard to the effects of tax subsidies on saving behavior.

1. Goldin and Lawson Framework To arrive at their finding, Goldin and Lawson put forward a framework detailing the optimal role of nudges, mandates, and taxes. In particular, they assume a world featuring two distinct and separate types, very similar to that described in the Denmark Study—passive and active decision-makers. Those decision-makers face a binary choice, and, to the extent they make an active decision, they may be biased to some varying degree. In their model, the passive decision-makers go where the nudge takes them and nowhere else (if nudged toward Option B over Option A, Option B is what they always choose). By contrast, active decision-makers do not follow the nudge but could respond to changes in price (i.e. a tax imposed on choosing Option A instead of Option B). Finally, both passive and active decision-makers would choose whatever option is mandated, as the mandate is defined as the equivalent to an infinite tax favoring one option over another.108 From this model, Goldin and Lawson conclude that nudges and taxes are complements and should often be deployed together in correcting bias, rather than a mandate—with this leading to higher welfare in a world where active decision-makers have heterogeneous preferences. The 106 See supra notes 26-32 and accompanying text. [build out] 107 Goldin & Lawson, supra note 13. 108 Id. at 6-8.

Page 31: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

31

basic intuition is that the nudge corrects the behavior of passive decision-makers and does this exactly the same as a mandate would. Further, the nudge has no effect on active decision-makers. The tax then corrects the bias for the active decision-makers but, unlike a mandate, still allows this heterogeneous population to choose their preferred option but with the bias offset on average for those on the margin choosing between the two options. The nudge plus tax is then seen as superior to a mandate (or floor) that would impose that solution on all (with Goldin and Lawson identifying some exceptions to this general finding).109 But, the retirement context shows why that important theoretical conclusion is limited in ways above and beyond those that Goldin and Lawson identify. This is in light of the combination of the actual effect of tax subsidies on saving and the effect of such subsidies on the distribution of resources, the latter being a factor that Lawson and Goldin do not consider.

2. Relative Insensitivity to Tax Subsidies and, Maybe, Nudges First, to the degree that active savers are biased, tax subsidies—at least in the ranges that have been traditionally tested—have only limited effect. That is described in Part II. The empirical work, including from the Denmark Study, suggest that marginal incentives can increase private saving. But, to the extent that effect is modest (as many of those studies still seem to suggest), offsetting any significant bias in saving behavior among active savers could require substantial tax subsidies.110 Further, there is a question of whether nudges would be as effective as the Goldin and Lawson analysis assumes in concluding that the combination of nudges and tax incentives are likely to be optimal in many circumstances. In particular, their analysis—following on that from the Denmark Study—neatly divides the world into two categories, “passive” and “active” savers, with the former entirely following defaults. However, there is strong evidence to indicate that the population does not divide so simply. As Dan Shaviro notes surveying recent evidence, “not all possible defaults are comparably sticky.”111 For instance, one recent experiment showed the vast majority of savers—in stark contrast to the results in the Denmark Study—not choosing the default but rather engaging in active decision making to select otherwise.112 In particular, in 2010, the IRS gave taxpayers the option of receiving refunds in the form of U.S. saving bonds, a low-risk and liquid asset. The experiment was conducted at sites used by low-income Americans to help fill out their tax returns. In an earlier experiment, it was found that, when the saving bonds were not the default, the take up of them was extremely low, around 6 percent. However, when the default was changed so that the savings bonds were the default, the take-up rate did not budge significantly. In other words, the

109 Id. at 8-19. 110 [Further build out/support.] 111 Shaviro, supra note XX, at 52. 112 See generally Erin Todd Bronchetti et al., When a Nudge Isn't Enough: Defaults and Saving Among Low-Income Tax Filers, 66 NATIONAL TAX JOURNAL 609 (2013).

Page 32: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

32

vast majority of these taxpayers actively chose to save less than the default or, at least, in another form.113 Similarly, while among the most famous findings in behavioral economics is the significant effect on participation and saving rates from defaulting people into 401(k) plans, fewer people “passively” chose that default initially than what the Denmark Study alone might suggest would occur, and the effects of the default diminished with time. For instance, one set of researchers—who pioneered this area—looked at the effects of the 401(k) default at two companies over a worker’s tenure. Taking one of their companies as an example: When the default was non-participation, that was the choice of 69 percent in the first three to five months of their tenure, but that fell to 47 percent by 24 to 26 months of tenure. And, when the default was contributing 3 percent of income, 72 percent chose that in the first 3 to 5 months of tenure, but that fell to 41 percent by 24 to 26 months of tenure, with the others either choosing lower or higher saving rates.114 The point is that passivity is variable. Sometimes people will engage with active decision-making, and sometimes they will not. Of course, that alone does not suggest that the Goldin and Lawson framework is faulty. After all, perhaps these savers—who are sometimes passive and sometimes active—will be sensitive to tax subsidies when they make active decisions. This is an empirical question to which there is not a firm answer: how do “semi-active” decision-makers respond to tax subsidies? The Denmark Study suggests that the less than one-fifth of the Danish population who were active in the face of modest changes in tax incentives saved somewhat less due to a reduction in a tax subsidy. How would “semi-active” decision-makers respond? It seems plausible that they would be no more sensitive and quite possibly less than this other group. Of course, the subsidy could be delivered in such a way that it is saved even by passive decision-makers—but, in that case, the subsidy is not increasing saving at the margin but, instead, having a lump-sum effect only, disconnected from any active decision-making (and not offsetting bias through an incentive effect).

3. Redistribution Even in combination with defaults, tax subsidies would continue to be redistributive in nature. Those who save less would, as a result of these interventions, pay more taxes than those who save more—all else being equal. And, again, the distributional effects could not be offset within the tax system, without withdrawing the subsidies themselves. In other words, an offset of this distributional effect is not just practically challenging, it is actually impossible. Floors or nudges alone would not come with this same redistributional consequence. To be clear, the extent of the redistribution should be less—and less problematic—than if done via tax subsidies alone. That is because passive savers should see their saving increased via the nudge. And the tax incentive would be layered on top of that. 113 Id. at 611. 114 James J. Choi et al., For Better or for Worse: Default Effects and 401(k) Savings Behavior,

in PERSPECTIVES ON THE ECONOMICS OF AGING 81, 98 tbl.2.2 (David A. Wise ed., 2004).

Page 33: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

33

The nudge, as a result, reduces the objection. Those who are entirely oblivious to tax subsidies and responsive to the nudge will all save the amount set by the default. And, assuming that nudge increases their saving rates, that should reduce the degree of redistribution between them and higher savers. In other words, there should be considerably less heterogeneity by income than exists now, and so less potentially problematic redistribution. Further, when it comes to active savers, there is still something to the basic intuition that Goldin and Lawson express. Low savers will have to pay a penalty for being a low saver, and to the benefit of higher savers. But, would they be better off if they were simply forced to save more rather than given the option to pay the penalty? Perhaps not. In fact, the low savers choose to be low savers and, essentially, pay the penalty to the benefit of the high savers; arguably, both are better off relative to a mandate. The low savers are, after all, choosing to save low, despite the incentive to save more—with that incentive correcting bias, on average. But, there remain two inter-related problems: First, tax subsidies could continue to redistribute from passive savers to those who save more than them—even if the redistribution is less. And, the passive savers are not actively choosing to pay a penalty relative to those who save more than them. Thus, the passive saver would be better off without a system that so rewarded those who save more—and there is at least some reason to think that such redistribution is not optimal from the larger perspective of society. Second, and relatedly, there is the question of the semi-passive savers—those who may not stick with a default but could be even less sensitive to traditional tax tools than the fully active population. Even as they move off the default, they may not be actively taking into account the tax subsidy in considering how much to save. They too may be better off without a choice, and it could be a distributional harm to shift resources from them to higher savers. Again, the fact that there may be distributional detriments does not mean that tax subsidies should not be used. But, it certainly complicates the case for doing so—and is a harm, which Goldin and Lawson do not address, relative to either nudges or floors alone.

4. A Tentative Case Against Tax Incentives There is a strong case against using tax subsidies as a primary way to increase saving. But, the case against doing so in tandem with nudges is more tentative. To review, the key concerns are that: (1) even fully active savers—as in those who react to modest changes in saving incentives such as those studied in Denmark—may not be all that sensitive to tax subsidies, implying that corrective measures would have to be large; (2) there may be a population that is “semi-active” in its decision-making and that moves off of defaults but could be even less sensitive to tax subsidies than fully active decision-makers (though this is an empirical question to which there is no good answer as of now); (3) tax subsidies could still have detrimental redistributional effects that would not occur either with nudges or mandates alone—and that could not be reversed. To be clear, these are not separate concerns. They are interrelated. The redistributional challenge is particularly significant given a possible lack of sensitivity to tax subsidies.

Page 34: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

34

Significant subsidies may therefore be needed to increase saving substantially and the redistributional effects rise in tandem. This does not necessarily mean that nudges alone or floors are superior to deploying a nudge with a tax subsidies. They avoid the negative distributional consequences, and that is significant. However, a nudge alone will fail to adjust the behavior of both active and semi-active savers. And, while a floor can vary to some degree (such as by income), it cannot accommodate the full heterogeneity of the population, which can produce its own welfare losses. So, what is presented here is a tentative case against the combination of nudges plus tax subsidies rather than a definitive one—and pointing to a key factor that is not directly addressed in the Goldin and Lawson framework, that of distribution. It is a case suggesting that nudges and floors alone or in combination may be superior to using tax subsidies, and that, when tax subsidies are deployed, they should be done in ways that minimize the detrimental distributional consequences. IV. Three Reform Models [Note: I’m planning to reorganize part IV and do so by focusing on what it would concretely take—in terms of reforming the current system—to achieve three reform models. As the previous sections help to illuminate, all of these models would be significant improvements over the current system, though models (2) and (3) would be clearly superior to (1). In particular, the models are: (1) A model focused on reforming subsidies alone—where subsidies are reformed to try to increase saving among both passive and active savers and to do so with as little distributional harm as possible. This includes the “universal credit” model from the 1990s. I’ll explain how—while this would be an important step forward—there are clearly better approaches, both in terms of efficiency and fairness. (2) A model that combines reformed subsidies with well-considered and widely applicable defaults. It’s a much closer call whether this is a first-best solution, but I’ll reiterate some skepticism based on the trade-offs described in Part III. (3) Finally, expanded floors, possibly combined with defaults—and abandoning tax subsidies entirely. I will lean into this as the first-best, consistent with the prior part, though express this as tentative—with a set of empirical assumptions driving that conclusion that deserve further inquiry.] This part focuses on concrete steps forward for savings policy based on the lessons so far identified in this article and the actual structure of the U.S. saving system. This does not put forward a new solution that has yet to be introduced. The retirement saving literature is, after all, replete with proposals of many different flavors for reforming the system. Rather, this part is meant to reframe reforms that have been discussed in light of the new lessons that this article highlights. One key lesson is that tax tools can be effective ways for improving saving, even if they may not be first-best tools. This then can reinvigorate efforts to reform these incentives to improve them—if they are not altogether eliminated. In particular, they can be reformed to increase saving via passive channels. Further, they can be redone to reduce—but only to some degree—the distributional harms being generated by the current system, especially by complex rules that

Page 35: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

35

allow sophisticated players to grab particularly large tax savings. This, in fact, brings us back to an old solution put on the table by President Bill Clinton but that lost some cache because of the recent literature like the Denmark Study: universal matching tax credits. A second key lesson is that tax subsidies—even reformed ones—are certainly not a first best tool standing alone. If they are not eliminated, they should be combined with carefully designed defaults that are applied widely and drive higher saving especially among “passive savers.” However, the current system of defaults is applied inconsistently across employers and at rates that are too low—so low in fact that they have the potential to be counterproductive, as scholars like Bubb and Pildes have claimed. The final lesson is that there is a real attraction to using floors to a greater degree than we do today to correct the savings behavior of active savers—whether in the form of expanded Social Security or an alternative regime like mandatory contributions to a universal saving account. Among other reasons, there is more confidence in the likely effects on low savers who may be neither sensitive to defaults nor tax subsidies, and, further, the system would not generate a detrimental redistribution of resources from low to higher savers that necessarily results from systems based on tax incentives.

A. Improving Tax Subsidies There are ways to reform tax subsidies to improve their operation significantly—both in terms of their effect on savings and the distribution of resources. In other words, and contrary to what one might take away from some of the current literature, tax tools do not necessarily have to be abandoned as a way of increasing saving; they can be improved.

1. Structuring for Passive Savers For the most part, the current tax incentives are structured to only increase saving among fully active savers. The Roth-style accounts described earlier are the exception rather than the rule in the current tax system. The traditional accounts suffer from two distinct flaws that make them ineffective at increasing saving among passive savers. As described in Part II, in designing saving incentives, there are key choices related to both the placement of such an incentive and its timing. First, the upfront tax savings occur outside the accounts, so that passive savers would have to actively choose to save those amounts rather than them being automatically deposited in the specified savings account. Second, these accounts involve a trade-off in the timing of taxation relative to fully taxable accounts—with lower taxation upfront (outside the saving account) and higher taxation at the back end (for funds distributed from the savings account). While the net effect of this timing trade-off is to provide a subsidy, it is easy to imagine how this could reduce savings to the degree people simply spend the amounts saved upfront, not understanding that part of it is a loan. So, improving these tax incentives involves some combination of: (1) directing the tax savings into the savings account as the default; and (2) eliminating the timing trade-off in the current system.

Page 36: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

36

Roth-style accounts, for instance, do this better by delivering their tax incentive later and not involving a timing trade off. It is not surprising then that the introduction of Roth-style accounts are associated with higher private saving relative to traditional accounts. Still, even the Roth-style could be improved in terms of the timing and placement of the subsidy. That account generates a subsidy at the time gains are realized, and this subsidy is essentially located outside the account. A passive saver may simply spend those savings at the time. While that may often align with the liquidation of the account (and thus the savings will finance consumption in retirement for instance), it will not necessarily. This stands in contrast to an incentive that would go directly into the account or be delivered only at the point of qualified withdrawal from the account. For a passive saver, this may result in more savings available at the actual point at which funds are being withdrawn for retirement consumption. However, the greater shortcoming in the Roth-style structure is not the placement and timing of the incentive; it is the targeting, along two separate dimensions.

2. Eliminating the “Upside Down” Structure The current tax incentives—including for Roth-style accounts—have two major targeting problems, the first being well known and the second somewhat less so. The first targeting problem is that the tax incentives—with the notable exception of the very limited Saver’s Credit—are proportional to one’s marginal tax rate, and marginal tax rates tend to rise with one’s income level. This is often referred to as an “upside down” system of savings incentives.115 As a result of this upside down system, middle class families get a significantly smaller incentive to save as compared to those with higher incomes. The effect is to some degree blunted by the annual limits on contributions to saving accounts, but it does not change the fact that, for the qualified saving, those with higher incomes get more subsidy per dollar of saving. The main problem with this upside down system is that there is no reason to think bias rises with income—and quite plausibly the opposite. As a result, the incentive is fundamentally mistargeted. If bias does not rise with income, then structuring any incentive in this way is inefficient. At the least, the same tax incentive should be available to encourage saving up and down the income spectrum. Many commentators also focus on the fact that the “upside down” system might reduce the overall progressivity of the tax system. That, too, could be a problem, but it is one that at least can be offset by changes in other parts of the tax code by making the rest of the income tax code more progressive. So, the fundamental problem of the upside structure—one that cannot be overcome in any other way—is one of efficiency, since a more efficient system would better align the incentive with the actual biases against saving.116 115 [Add in cites to the a few of the many articles describing the incentive system as upside down.] 116 Lily Batchelder, Fred Goldberg, and Peter Orszag have laid out the case for using tax credits versus deductions or exclusions as a way of incentivizing behavior when correcting for externalities. See generally Lily Batchelder et al., Efficiency and Tax Incentives: The Case for Refundable Tax Credits, 59 STAN. L. REV. 23 (2006). In their words, “Under the most

Page 37: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

37

3. Eliminating Disproportionate Benefits by Sophisticated Players

The second targeting problem in the current tax system—and one that, again, cannot be overcome by adjusting other parts of the tax system—is the degree to which sophisticated taxpayers are able to shield disproportionate amounts of income. While there are caps on annual contributions to tax-preferenced accounts, these are gameable, and some sophisticated taxpayers appear to be doing so, especially with Roth-style accounts. For instance, the Government Accountability Office (GAO) has specifically studied very large IRAs.117 Their 2014 study of this one form of retirement account found that between about 6,830 and 11,420 people had IRA balances in excess of $5 million, and between 115 and 650 people had balances in excess of $25 million.118 The GAO concludes that, to achieve such large balances, some are likely “us[ing] alternate strategies involving investments not available to most taxpayers.”119 In particular, GAO speculates that some of the account holders are investing in non-publicly traded assets—that are under-valued and thus fit under the IRA limit. GAO describes two such assets that, based on its interviews, appeared to be driving the extraordinary returns: namely, investment in “founders stock” by founders of start-ups and investment in partnership interests by hedge fund and private equity managers. GAO documents how one of the founders of a successful technology company used a Roth IRA to purchase 4 million shares of his company for a total price of $4. And, GAO is investigating the tip of the iceberg.120 What sophisticated taxpayers have done in Roth IRAs is now apparently also happening in other types of Roth style saving vehicles that have been introduced—like Roth 401(k)s.121 The problem is particularly severe in Roth-style accounts. First, that is because such strategies allow taxpayers to not only stick assets into the accounts that are worth much more than is being claimed but also assets with large, unrealized gains. In the case of a Roth account, it means that not only is further appreciation exempt from taxation, but the previous appreciation is as well—a particularly large tax windfall for those who do the maneuver. Second, and relatedly, taxpayers can invest in assets using a Roth-style account where they expect outsized returns (and thus outsized tax subsidy), whether because the return reflects a combination of returns to labor and

reasonable set of default assumptions, a tax incentive provision correcting for positive externalities should apply uniformly across the income distribution and different lifetime earnings patterns. Refundable credits are the only straightforward way to achieve such uniform application.” Id. at 27-28. A similar logic should apply to the correction of “internalities.” 117 GOVERNMENT ACCOUNTABILITY OFFICE, INDIVIDUAL RETIREMENT ACCOUNTS: IRS COULD BOLSTER ENFORCEMENT ON MULTIMILLION DOLLAR ACCOUNTS, BUT MORE DIRECTION FROM CONGRESS IS NEEDED (2014). 118 Id. at 17 tbl.1. 119 Id. at 26. 120 Id. at 26-34. 121 [Get cites.]

Page 38: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

38

saving or because they know they have a monopoly on a new technology (with both possibly being true in start up companies). Importantly, such planning is not uniform among high-income Americans. These abuses are almost certainly exclusive to that income group, but some do it and some do not. In part, this is because only some have access to the types of assets and tax advice that allow these aggressive maneuvers. As a result, the effects cannot be perfectly offset given the heterogeneity, other than simply directly stopping such maneuvers. One direct way to end such planning is to base any tax subsidy on the amounts contributed to accounts rather than final balances, since tax administrators can most easily monitor and limit amounts being contributed—and can avoid subsidizing outsized returns. Or, to put it differently, it would avoid a Roth-style structure that bases the amount of subsidy on the reported gain achieved. The fundamental point being made here is that it is possible to restructure tax saving incentives in such a way as to increase saving via passive mechanisms—as the Roth-style accounts appear to do. And, it is possible to do so without the mistargeting of current Roth-style accounts, both in terms of the upside down incentives and providing disproportionate benefit to certain savers.

4. The Model of Matching Credits There have been attempts at reforms in this vein, even if some upsides (such as structuring them to take account of passive savings) were not fully understood at the time. For instance, President Bill Clinton in 1999 proposed a system of matching credits that would have gone directly into savers’ accounts. However, this was not in place of the current tax preferenced accounts but rather a discrete add-on for lower to middle-income Americans rather than in replacement of the current system of tax preferred accounts.122 Since then, more expansive versions have been suggested for instance by Gene Sperling (who also helped develop the version proposed by President Clinton). Sperling proposes a system in which lower- and moderate-income Americans would get a dollar-for-dollar credit on up to $4,000 of savings per year, and the upper-middle class would get a 60 percent credit rate.123 Again, the credit would go directly into the savings account. These matching credits represent perhaps the best possible version of a tax subsidy alone to encourage saving. They can increase saving via passive decision mechanisms; they are not upside-down; and they close down most avenues by which sophisticated taxpayers might manipulate the system. And, they cannot simply be dismissed—as some have—as entirely ineffective at their desired goal of increasing saving. Still, they are fundamentally different from the way most of the current tax incentives work, since, again, the subsidy goes directly into the savings account.

122 See here: http://clinton4.nara.gov/WH/Work/041499.html. 123 See here: http://www.nytimes.com/2014/07/23/opinion/a-401-k-for-all.html?_r=0.

Page 39: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

39

Nonetheless, they suffer from the problems of any tax-based tool described in Part III. While better targeted than the existing tax-incentive system, such matching credits would still do more to help those who save than those who do not, as well as redistribute resources toward savers.

B. Better Nudges and Expanded Floors Thus, tax subsidies—even a reformed set that would be a significant step forward relative to the current system—should not be relied on as a primary tool for increasing saving. Rather, they should at least be complemented by a set of substantially improved defaults and potentially displaced entirely by these better defaults and expanded floors.

1. Better Defaults While automatic enrollment in employer saving plans is on the rise in the United States, the system’s defaults remain poorly designed. Among other problems, auto-enrollment is applied inconsistently on an employer-by-employer basis; decisions on where to set the savings defaults remain woefully under-informed; and it is even possible that defaults have had a counterproductive effect on saving, as Bubb and Pildes explain. In sum, if they are to be used as a key tool for increasing saving, such nudges must be not only be expanded but also carefully reformed so that they enhance saving, rather than possibly the opposite. The Pension Protection Act of 2006 (PPA) aimed to encourage auto-enrollment in employer-based retirement plans. The PPA did so by providing legal safe-harbors for 401(k) plans that include certain types of defaults, shielding the plans from a number of rules that would otherwise apply. To fit within the applicable safe-harbor, all newly eligible employees had to be automatically enrolled in a 401(k) plan with an initial contribution rate of at least 3 percent but no higher than 10 percent. The PPA then escalates the minimum, default contribution rates that automatic enrollment plans must feature in order to be eligible for the applicable safe-harbor. While the default contribution rate can be set as low as 3 percent in an employee’s first year of participation, this escalates to 4 percent in the second year, 5 percent in the third year, and 6 percent in the fourth year and thereafter.124 The PPA has corresponded to a rise in auto-enrollment, but it is still inconsistent across the population. First, such defaults are only in place for those employer-sponsored saving plans, and half of American workers do not have access to such plans.125 Second, only some employers offer such defaults. The share has been rising, but is still only about 50 percent of employer-sponsored plans—up from 8 percent as of 2003 and just over one-third in 2007.126 Finally, these automatic default plans tend to feature relatively low saving defaults, in part following the guidance of the PPA. And, there is a concern that these low defaults could, in fact, be counterproductive on net, leading some who would save more absent auto-enrollment to save

124 Pension Protection Act of 1996 [add full citation]. 125 http://www.pewtrusts.org/~/media/assets/2016/01/retirement_savings_report_jan16.pdf 126 Profit Sharing/401k Council of America, Annual Survey of Profit Sharing and 401(k) Plans [get cite

information].

Page 40: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

40

less. Bubb and Pildes emphasize this last concern,127 and, remarkably, there is not good evidence as of yet on whether automatic-enrollment has been such in such a ways as to increase saving or not. The key point is that, while the United States has been sliding toward a system that combines automatic enrollment and tax incentives, it is not just the tax incentives that are poorly designed—but also the automatic enrollment. If automatic-enrollment is to be either a complement or substitute for tax incentives, the system should be redesigned. Among other things, defaults should not vary depending on one’s employment status or employer—and, to the extent the defaults do vary, they should do so based on factors that should affect the optimal default like age. They should be set at higher rates and with greater confidence that they are enhancing saving and people’s welfare rather than the opposite. [More TK]

2. Expanded Floors Even then, properly designed defaults are probably insufficient given that they will not help biased, active savers. For active decision makers, the choice is between adjusting their behavior through tax incentives or increasing the existing floor. This returns to this article’s prior discussion—whether tax subsidies should be deployed in combination with a reformed set of defaults. The case here is not definitive, but the concern laid out in this article is that tax subsidies would have to be very large to properly correct behavior—and, in doing so, would have negative distributional consequences. These same problems do not come with a higher savings floor—though that would come with its own detriment of forcing those who, from the perspective of their enlightened self-interest should save less, to save more. This article does not definitively settle the debate between using tax incentives and floors to affect active decision makers. In fact, it—to an important degree—rejuvenates that debate by showing that tax incentives do increase saving at least to some degree among active savers. Nonetheless, it lays out a renewed case for using floors instead. And, a number of scholars and policymakers have identified ways to expand floors in the United States—even if some of their reasons for dismissing tax incentives may have been misguided. Their proposals generally involve either some combination of enhanced Social Security benefits or mandatory contributions to separate add-on accounts. [TK—will briefly discuss these proposals.] Conclusion Recent scholarship has argued that tax incentives are deeply flawed as a tool for increasing retirement saving. That is the takeaway from the most recent retirement literature including the groundbreaking Denmark Study. The point of this article is that they are not that flawed. Or, to

127 Bubb & Pildes, supra note 3, at 1622-24.

Page 41: The Under-Appreciated Potential and Misunderstood Failings ... · Getting Americans to Save: The Under-Appreciated Potential and Misunderstood Failings of Tax Subsidies David Kamin*

VERY PRELIMINARY DRAFT: NOT FOR CITATION

41

put it differently, they can actually increase retirement saving and contrary to many recent claims that they fail entirely at that task. The distinction is important. First, there is a real case to be made against using tax incentives to encourage retirement saving as this article describes. And, it is a tentative case, where additional research could help better illuminate the key tradeoffs discussed here, including the distributional detriments of tax incentives versus some of the welfare losses that would using other tools like enhanced saving floors. Second, it means that tax incentives can be improved; they still cannot be turned into a first-best tool, at least from the perspective of this author—but they can be made much better than what they are today. In short, this article comes to revive tax incentives as a way of getting Americans to save and then to bury them. But, in reviving and burying them, it aims to push the debate forward—and lay out both important questions for research that deserve additional attention and, based on the information we have, the next steps that should be taken to reform the retirement saving system and address the lack of saving that now plagues millions of Americans.