Taxing Our Wealth*

Florian Scheuer

University of Zurich

Joel Slemrod

University of Michigan

May 2020

Abstract

This paper evaluates recent prominent proposals for an annual wealth tax. We point

out why the former conventional wisdom—that an optimal tax system would feature no

taxes on capital—has been overturned, and present a series of arguments that justify pro-

gressive taxes on wealth accumulation both in the short and long run. We also discuss

under which conditions they should take the form of a wealth tax versus alternative poli-

cies that achieve similar objectives. While a dozen OECD countries levied wealth taxes in

the recent past, now only three retain them, with only Switzerland raising a comparable

fraction of revenue as the U.S. proposals. Studies of these taxes often find a substantial

behavioral response, which may indicate a large excess burden. To predict the conse-

quences of the U.S.-style proposals, however, one needs to take into account that their

design features—especially the rate schedule, broadness of the base, and enforcement

provisions—are very different from any previous wealth tax. This makes it difficult to

learn from experience, but we can gain insights from closely related taxes, such as the

property and the estate tax.

1 Introduction

During the 2019/20 Democratic presidential nomination campaign, two prominent candi-dates, Bernie Sanders and Elizabeth Warren, proposed that the United States enact an an-nual wealth tax. Warren proposed a 2% rate on net worth in excess of $50 million and a 6%rate above $1 billion,1 while Sanders’s proposal featured graduated rates starting at 1% on

*We are grateful to Paul R. Organ and Gabriele Patete for exceptional research assistance.1Her first proposal featured a top rate of 3%, which was subsequently raised to 6% on November 1, 2019.

1

net worth above $32 million for a married couple, rising to a marginal tax rate of 8% on networth above $10 billion. The candidates (Warren and Sanders, respectively) claimed theselevies would raise $3.75 trillion and $4.35 trillion over 10 years, or approximately 1.3% and1.6% of GDP and 7.9% and 9.1% of federal revenues.2

The wealth tax also has prominent academic adherents. Piketty (2014) called for a globalprogressive wealth tax. Emmanuel Saez and Gabriel Zucman of UC Berkeley advised Sena-tor Warren regarding her proposal, and have written a detailed explanation and defense ofit (Saez and Zucman, 2019a).3

In this paper, we assess the implications of levying such a tax.4 At first blush, this isinescapably a highly speculative undertaking. The United States has never levied an annualwealth tax and, of those countries that have done so, none has applied rates anywhere nearthe 6% or 8% rates in the prominent proposals nor established such a broad base. All isnot lost, however. We can gain some insight from experience with taxes that are similarto a wealth tax and have been in place in the United States. We can also learn from theexperience other countries have had with wealth taxes. And we can apply the insights fromoptimal tax analysis. Before embarking on those tasks, as background we clarify how awealth tax works.

2 What is a Wealth Tax?

2.1 Base and Rates

In principle, the base of the tax is net worth—the value of assets minus debts. Like all taxes,in practice the base to which tax rates are applied could be narrowed by exemptions, de-ductions, or preferential treatment (e.g., discounted valuation) of certain components of netwealth.5 As we will see below, past and current wealth taxes feature many such base nar-rowing features, but both Warren and Sanders have proposed unprecedently broad bases,including for example assets held in trust, and (in Warren’s case) including retirement as-sets and assets held by minor children. To this base, both proposals apply an extremelygraduated rate structure, with no tax due until wealth reaches a very high amount. UnderWarren’s plan 75,000 households would be subject to the tax, while Sanders’ plan wouldapply to 180,000 households. The very top rates would apply to an even smaller subset:

2Both candidates proposed other changes to taxing affluent taxpayers. For example, Sanders proposedimplementing a top estate tax rate of 77% (it is currently 40%).

3Kaldor (1956) called for a wealth tax for developing countries.4We do not address whether a wealth tax would be constitutional in the United States, a subject of some

controversy. For the two views, compare Johnson and Dellinger (2018) and Jensen (2019).5In practice, the base of a tax could also be expanded beyond the principled base.

2

Warren’s top rate kicks in at $1 billion, applying to 621 people as of October 2019 (accordingto Forbes), while Sanders’ top rate kicks in at $10 billion, applying to 47 people.

2.2 Objectives

The extreme graduation of the proposed rate structure leads naturally to considering theobjectives of introducing such a tax into the federal tax policy arsenal. The primary goalis to redistribute resources from the most affluent households to households in the middleand working class. This is evident in the base itself, wealth, which is considerably more un-equally distributed than the most common tax bases, income and consumption. Wolff (2017)reports that in 2016, the top 1% of households ranked by net worth held 40% of the UnitedStates’ wealth, while the top 1% ranked by income earned 24% of income. It is also evidentin the very high wealth thresholds before any tax is due and the graduated rate structure thatapplies above the thresholds. Finally, it is evident in the types of government expenditurethe revenues would fund.6 Warren proposed to allocate the revenue to higher education,public K-12 education, universal childcare, and her Medicare for All plan. Sanders wouldhave used the revenue to fund affordable housing, universal childcare, and to help fundMedicare for All.

The argument for levying a wealth tax is often made in the context of unacceptably high,and growing, inequality of income and wealth, although how much inequality is too muchinevitably involves value judgments. The details about the increase in inequality over thepast four decades are disputed, but, in our view, the conclusion that it has increased non-trivially is not.7

The case is also often couched in claims that effective income tax progressivity is not highenough and its progressivity potential is constrained by structural problems, particularlyhaving to do with the taxation of capital gains, about which we have more to say below.Capital gains are taxed at a preferentially lower rate than other income, are taxed uponrealization rather than accrual, and are excused at death due to the “step-up” of tax basis forbequeathed assets that have appreciated in value.8 Realized capital gains represent a veryhigh fraction of the reported income of the superrich. For example, IRS data shows that intax year 2014 realized capital gains represented 60% of total adjusted gross income (AGI) for

6Neither plan would use any of the wealth tax revenue for deficit reduction.7We do not have the space here to review this debate; interested readers should refer to Bricker et al. (2016),

Kopczuk (2015), Saez and Zucman (2016), and Smith, Zidar and Zwick (2020). The controversy about how tomeasure the extent and distribution of wealth is related to the practical problems of measuring wealth for thepurposes of tax assessment. Consider, for example, the issue of what discount rate to apply to a projected flowof income, or how to disentangle the capitalized value of partnerships and human capital (which is typicallynot included in wealth).

8At the same time, the income that gives rise to the appreciation of some capital assets, such as corporatestock, is subject to taxation at the corporate level.

3

the 400 highest-AGI Americans. In tax year 2016, those with AGI over $10 million report netcapital gains corresponding to 46% of their AGI, whereas it is a negligible fraction for thoseearning less than $200,000 (Scheuer and Slemrod, 2019).

2.3 Interpreting the Rate of Tax

A wealth tax at a rate of tw is roughly equivalent to a tax rate of tw/r on capital incomewhen the interest rate is r. So, for an asset whose rate of return is 2%, a 3% wealth tax is cor-responds to an annual 150% capital income tax and an 8% wealth tax is roughly equivalentto a 400% capital income tax. Of course, not all assets have the same rate of return, whichvaries over time for most assets. Given the large amount of political capital that goes intoadvocating, and resisting, changes in the capital income tax between, say, 15% and 20%, therecent proposals, which involve changes up to 80 times higher, certainly represent a majorstep.

We say “roughly” equivalent because, even if the rates were adjusted appropriately, awealth tax differs from a capital income tax in an important way. For a given amount ofwealth, the tax liability of a wealth tax does not depend on the flow of income the wealthactually produces, while in contrast a capital income tax take is related to that flow. In theextreme, if all of one’s wealth were held in a zero-interest demand deposit, a capital incometax would generate no tax liability while a wealth tax would; one can think of the wealthtax as a tax on the “normal” rate of return to capital at a rate of tw/r. We will return to theimplications of this aspect of a wealth tax below.

Just as does a capital income tax, a wealth tax affects the rate of return to saving and,equivalently, the relative price of consumption in different periods. The after-tax rate ofreturn to saving over n periods is (1 + r − tw)n. Thus, an 8% wealth tax turns a 2% rate ofreturn before tax into a negative 6% rate of return. For example, whereas at a 2% returnone dollar becomes $1.811 (= (1 + .02)30) after 30 years, after tax one dollar becomes 15.6cents (= (1− 0.06)30). Tax rates that might sound low in the income or sales tax context areactually much higher when compounded over time.

3 What Other Taxes is it Like?

Although the United States has never had an annual wealth tax, it has a long experiencewith two closely related taxes.

4

3.1 Property Taxes

Local governments rely heavily on an annual tax on one form of wealth, often called “im-movable” property, in the form of property taxes. Estimates suggest that property taxesaccount for about one-third of local government revenues.

The rate of tax levied on immovable property varies widely. Harris and Moore (2013)report that, for the period 2007-2011, over 60% of counties had property tax bills that wereless than 1% of their median home value. An additional 37% of counties had property taxbills between 1% and 2%, while 3% had property tax bills in excess of 2% of their homes’value. The mean property tax burden as a share of house prices was 1.15%. Thus, the rate oftax is slightly below the lowest rate of tax that would be applied by the Sanders and Warrenplans.

Whether a property tax would have similar effects as an equivalent-rate wealth tax issubject to several qualifications. The first is the Tiebout (1956) theory of property tax. If localgovernments can enforce a minimum property tax base per household and households canchoose among many different communities with varying levels of local public goods, thenthe property tax becomes a non-distorting price for local services. The other is that the prop-erty tax base is only one component of wealth, and the relative size of this component variesdramatically across net worth levels. For example, using the 2016 Survey of Consumer Fi-nances, Wolff (2017) estimates that, while principal residences account for 62% of the grossassets of individuals in the middle three quintiles of net worth, they comprise only 8% ofgross assets for the top 1%, so the property tax is definitely not well-targeted at the verywealthy. Third, most local property taxes in the U.S. feature a fixed rate, perhaps with an ex-emption. This need not be. For example, the council tax in the U.K. implements a graduatedrate structure based on the property value, as do some Swiss cantonal property taxes. ManyU.S. states levy a surcharge on the highest-value homes or have a progressive bracket struc-ture through their real estate transfer tax system, sometimes referred to as mansion taxes.Finally, property taxes do not allow for a deduction for debt.

What is the evidence on avoidance and evasion of property taxes? Kelly (2012) assertsthat the coverage ratio, defined as the ratio of the amount of taxable property captured inthe fiscal cadaster divided by the total taxable property in the jurisdiction, is close to 100%in most OECD countries, although between 40% and 60% in most developing economies.Chirico et al. (2016) find that the national average compliance rate in the U.S. is 95%, wherethe compliance rate is defined as the percent of taxes levied in the collection year that arepaid in the year due; this estimate does not include any properties that evade the assessor’sattention or whose value is illegally understated.

5

3.2 Estate and Inheritance Taxes

The U.S. federal government has levied a tax on a base close to net worth for over 100 years,in the form of an estate tax. It implemented an estate tax in 1916, and appended a gift tax in1924. It requires a valuation of taxable wealth at death, coinciding with the probate process,which locates and determines the value of the assets of the deceased. The U.S. estate tax alsohas a sizeable exemption level, now $11.2 million for singles and $22.4 million for marriedcouples, and features a flat rate of 40% over the exemption. The revenue it generates haseroded over time; in fiscal year 2019 it raised $16.7 billion, or slightly less than 0.5% offederal revenues, from about 0.06% of decedents.

In principle, this should mean that the superrich are exposed to taxation substantiallyin excess of the income tax, but in practice, many additional exceptions allow them to sig-nificantly reduce their exposure. Notably, capital gains unrealized at death (typically in theform of closely-held stocks in private businesses) are excluded from taxation due to step-upprovisions. Moreover, the effective estate tax rate is reduced by extensive undervaluation ofwealth transfers via, for example, family limited partnerships.9

As of 2017, 26 of the 35 OECD countries levied some kind of tax on wealth transfers.With the prominent exception of the U.S., which levies an estate tax such that the heirs’circumstances and relationship to the deceased (other than regarding spouses) do not affecttax liability, most countries levy an inheritance tax, where the liability lies with the recipientand the rate of tax depends on their relationship of the heir to the deceased.10

What is known about the consequences of an estate tax? Kopczuk (2013) surveys the ev-idence and concludes that the literature suggests an elasticity of reported estates with respectto the net-of-tax rate between 0.1 and 0.2. Eller et al. (2001) analyze estate tax evasion basedon data from a stratified random sample of federal estate tax returns as filed and audit as-sessments. They estimate the estate tax underreporting gap to be 13%, but this figure maysubstantially understate the true magnitude of the gap, in part because it does not accountfor any noncompliance not detected during the IRS examination process.

4 Wealth Taxes Elsewhere

We are not limited to studying taxes that are sort of like annual wealth taxes, because actualannual wealth taxes have existed, and still exist, in places sort of like the United States.

9A family limited partnership is a holding company owned by two or more family members created toretain a family’s business interests as well as real estate, publicly-traded and privately-held securities. Due tothe lack of control and lack of marketability that limited partners possess, these interests can be transferred tofuture generations at a discount to market value.

10As of 2019 six U.S. states levy an inheritance tax on top of the federal estate tax.

6

4.1 European Wealth Taxes

While in 1990 twelve OECD countries levied an annual tax on net wealth, by 2018 onlyfour—France, Norway, Spain, and Switzerland— levied such a tax, with Switzerland rais-ing more than three times as much revenue as a fraction of total revenues (3.9%) as any ofthe other three countries (OECD, 2018). In 2018, France replaced its annual wealth tax witha tax only on immovable property. Italy levies an annual tax on financial assets. The Nether-lands has a hybrid system with similarities to an annual wealth tax, imputing an asset-type-specific rate of return to assets and assessing a 30% tax on those imputed returns.11

Table 1 provides some summary statistics about the OECD countries’ wealth taxes, andcompares them to the Sanders and Warren proposals. A few aspects are especially worthnoting. First, the top rates of both the Sanders and Warren proposals are far higher than anytop rate of the OECD wealth taxes. The average top rate was about 1%, and the highest ofall is that of the Spanish region Extremadura, at 3.75%. But the Spanish system and certainSwiss cantons feature a cap on the sum of wealth and income taxes as a fraction of taxableincome, which is a feature of neither the Sanders nor Warren proposals. Such a cap limitsthe liquidity problem of a high ratio of tax liability to disposable income (and imposes azero marginal tax on wealth for those at the cap), but it may provide an additional incentiveto reduce reported taxable income; it also exacerbates the issue that the income tax basedoes not well measure the income of the very rich.12 Second, many of the OECD wealthtaxes featured exemption or preferential treatment of some forms of assets, notably mainresidence, life insurance proceeds, pension wealth, and business assets. The Warren andSanders proposals have no such exemptions, and extend the base to certain assets that fewcountries in the OECD included, such as assets held in trusts, retirement assets, and assetsheld by minor children. Third, the exemption level of the other countries’ wealth taxes aremuch lower than Sanders and Warren proposed, averaging just about €500,000 for marriedcouples.

At first blush, it does not bode well for wealth taxes that, of the dozen OECD countriesthat have had them in the last three decades, only a quarter of them still do. Why did theother three-quarters abandon them? A 2018 OECD report refers to efficiency costs, risk ofcapital flight, failure to meet redistributive goals, and concerns about high administrativecosts. It also notes that revenues collected from net wealth taxes were “very low” (p. 11),

11Several non-OECD countries have had wealth taxes, including Argentina, Bangladesh (more recently a networth-triggered income tax surcharge or net wealth tax, whichever is higher), Colombia, India (repealed in2015), Indonesia (abolished in 1985), Pakistan (removed in 2003 and reinstated in 2013), and Sri Lanka (1959-1993).

12With a cap, a wealth tax increases the ratio of total tax to taxable income. Depending on the degree of grad-uation in the income tax, though, a cap that is a fixed fraction of income might reduce the effective marginaltax on income.

7

Table 1: Warren and Sanders wealth tax proposals vs. European wealth taxes.

8

but the benchmark for such a judgement is not clear. In Germany, the Federal ConstitutionalCourt deemed the wealth tax unconstitutional in 1995 on the grounds that the tax’s discrim-ination of property and financial assets was an infringement against the fiscal principle oftax equality (Drometer et al., 2018). While this could be fixed in principle, it has not been at-tempted so far. In Sweden, it was argued that the special treatment of business equity madethe wealth tax regressive, taxing middle-class wealth (housing, financial assets) and exempt-ing the wealthiest individuals’ assets (large, closely-held firms); in addition, the wealth taxwas blamed for spurring tax avoidance and evasion, including capital flight to tax havens(Waldenström, 2018).

4.2 The Swiss Example

Of the three European countries that still levy a wealth tax, Switzerland is the only one thatraises a non-negligible share of overall tax revenue with it. Total revenue from the Swisswealth tax amounted to 1.1% of GDP in 2018, which is in a similar ballpark as the revenuesprojected for the recent U.S. proposals. Hence, the Swiss example is of particular interest forthe wealth tax debate in the United States.

The wealth tax in Switzerland has a long history and in fact predates the modern incometax. The Swiss tax system is generally structured in three layers: the federal, cantonal andmunicipal level. There is no federal wealth tax, but all cantons must levy a comprehensivewealth tax. Apart from that, cantons have significant freedom in designing wealth taxa-tion. Nine cantons impose flat rates (above some exemption level) and the other 17 featureprogressive schedules. Each municipality then chooses a multiplier that is applied propor-tionally to the cantonal tax rate schedule. Hence, an individual’s overall tax liability dependson both the canton and municipality of residence. Unlike the recent proposals in the U.S.,which all involve a federal wealth tax, this highly decentralized system induces local taxcompetition.

In 2018, the (combined cantonal and municipal) marginal wealth tax rates in the topbracket ranged between 0.1% (canton of Nidwalden) and 1.1% (canton of Geneva). In 16 ofthe 26 canton capitals, the annual top wealth tax rate was below 0.5%. There is also somevariation in the tax-exempted amounts, although they are generally relatively low, rangingin 2018 from $55,000 in the canton of Jura to $250,000 in the canton of Schwyz (for marriedcouples). Hence, even though it raises similar overall revenue as some U.S. proposals, theSwiss wealth tax is much less progressive and targeted at a larger share of the population.

The base of the Swiss wealth tax is broad: in principle, all assets, including those heldabroad, are taxable. Only foreign real estate, usual household assets, and accumulated pen-sion wealth are exempt. The tax liability is based on net wealth, so taxpayers can deduct

9

their obligations (such as mortgages or other debt). The annual reporting requirements forassets and liabilities allow the cantonal tax authorities to track the year-to-year evolution ofwealth and cross-check it against reported income (the so-called wealth development test),so the wealth tax serves a supporting role for income tax enforcement.

Several particularities of the Swiss tax system provide justifications for a wealth tax inSwitzerland. First, capital gains on movable assets (e.g. shares) are tax-exempt unless theowner professionally trades with securities. Second, almost all cantons have gradually abol-ished taxes on gifts and inheritances from parents to children. Thus, there is a particularlysevere defect in the income tax with regard to capital gains, and an effective estate or in-heritance tax is missing. The wealth tax therefore serves as a backstop to at least partlysubstitute for these instruments, which are commonly used in other countries but lacking inSwitzerland.13

So far, the Swiss case is the only modern example for a successful wealth tax in an OECDcountry, in the sense that it has been able to generate sizeable and stable revenues in thelong run. It enjoys broad support, as evidenced by the fact that it keeps being reaffirmedby citizens in Switzerland’s direct democracy where most tax decisions must be put directlyto voters. But its design and the role it plays in the overall tax system are quite differentfrom what is currently discussed in the United States. In particular, it is not geared towardsa major redistribution of wealth, and indeed wealth concentration in Switzerland remainshigh in international comparison (Föllmi and Martinez, 2017).

5 Consequences of a Wealth Tax

We turn now to assessing the likely consequences of a wealth tax, beginning with someobservations on the historical record of wealth taxes elsewhere.

5.1 Evidence on the Response to Wealth Taxes

Studies of the OECD wealth taxes, and of other countries, provide what little direct evidencewe have of their consequences. Unsurprisingly, empirical studies of the behavioral responseto wealth taxes are much sparser than for income taxes, largely because the taxes themselvesare much rarer. They are also harder to generalize from, as the tax bases and relevant en-

13As for immovable property, there is a special capital gains tax for real estate at the cantonal level. Moreover,about half of all municipalities also subject real estate to a graduated property tax (with rates up to 0.3%), basedon the gross value of the property. Finally, for income tax purposes, an estimate of the rental value of owner-occupied housing must be reported as taxable income. On the other hand, the valuation rules used by thecantonal tax authorities to assess real estate property imply that it is typically valued below market for taxpurposes, and this latter advantage often outweighs the double taxation.

10

forcement details vary widely. Indeed, some potentially critical enforcement instruments,such as the information exchange agreements discussed below that are designed to con-strain evasion using foreign accounts, post-date all of the studies. And, finally, because thetop rates of the Sanders and Warren proposals are so much higher than the historical expe-rience of any country, at best one can hold one’s breath and extrapolate from what is knownabout wealth taxes featuring much lower rates. Even with these caveats, it is worthwhile toreview what the studies concluded.

We begin with three studies that find taxable wealth to be highly responsive to its taxrate. Brülhart et al. (2019) take advantage of variations in the Swiss wealth tax rate acrosscantons and over time and find that a 1 percentage-point decrease in wealth taxes leads toat least a 43% increase in reported taxable wealth after six years. Administrative tax recordsfrom two cantons with quasi-randomly assigned differential tax reforms suggest that aboutone-fourth of the effect comes from taxpayer mobility and another one-fifth from house pricecapitalization. They argue that savings responses cannot explain more than a small fractionof the remainder, suggesting sizable evasion responses in this setting with no third-partyreporting of financial wealth.

The results of Jakobsen et al. (2020) suggest an even larger response. They examine theDanish wealth tax that was implemented in 1989 and abolished in 1997, taking advantage oftwo tax system design aspects: a doubling of the exemption threshold for married couplesand the cap on the ratio of income, payroll, and wealth taxes as a fraction of income thatrenders the marginal wealth tax equal to zero for those at the cap. For the very wealthy, theyconclude that reducing the wealth tax rate by 1.45 percentage points would raise wealthafter 30 years by 65%, corresponding to an elasticity with respect to the net-of-tax rate ofreturn of 1.15. They argue that, because the estimated effect grows over time, it could not beall a one-time avoidance effect.

Also sizeable, but not quite as large, are the conclusions of Durán-Cabré et al. (2019)about the behavioral response, who study the surprise re-introduction of a wealth tax inCatalonia in 2011. They find no evidence of it reducing wealth accumulation, and that ittriggered substantial tax avoidance via taxpayers changing their asset composition towardexempt assets (mainly company shares) so as to reduce wealth tax liability and reducedtaxable income to take advantage of an income-related cap on the sum of income and wealthtax liability. They calculate an elasticity of taxable wealth with respect to the net-of-tax rateof return of 0.6, implying that a 0.1 percentage point increase in the average wealth tax rateleads to a reduction in taxable wealth of 3.2% over 4 years.

Two other studies of European wealth taxes find smaller effects. Seim (2017) exploitsbunching around a kink in the Swedish wealth tax rate schedule where the rate changesfrom 0% to 1.5% and estimates an elasticity of taxable wealth with respect to the net-of-tax-

11

rate of between 0.09 and 0.27. He concludes that the elasticity mainly represents reportingresponses and finds no evidence of households changing their saving or portfolio compo-sition. Zoutman (2018) studies a major reform to wealth and capital income taxation in theNetherlands that occurred in 2001. Comparing households that were similar in wealth andincome, but treated differently by the reform, he concludes that a 0.1% increase in the wealthtax rate leads to a long-run reduction in accumulated wealth of 1.4%.14

In sum, recent studies of the European experience suggest that the behavioral response towealth taxation can be substantial, and the anatomy of the response—real versus avoidanceversus evasion—varies a lot, probably because of differences in the broadness of the taxbase and enforcement details such as the extent of third-party reporting of wealth. There ismoreover an important interaction among these types of responses. The effective marginaland average tax rate on a real behavior depend on the avoidance and evasion opportunities.In the extreme, a tax that can be costlessly evaded will provide no disincentive to the realbehavior. As Slemrod (2001) details, more generally, the tax disincentive to the real behaviordepends on how the marginal cost of avoidance and evasion is affected by the real decision.

Before applying these results to forecasts about the impacts of a Sanders-Warren-stylewealth tax in the U.S., it is essential to understand how design differences affect the likelyconsequences. Supporters have suggested several reasons why the U.S. proposals might bemore successful than the European experience would indicate: The United States is a muchlarger country, its tax system is citizenship- rather than residence-based, the proposals in-volve much higher exemption thresholds, they are accompanied by plans to enhance tax en-forcement, and their implementation would post-date the adoption of the Foreign AccountTax Compliance Act (FATCA) in 2010. We take up some of these issues below.

5.2 Real Behavioral Responses

For a perfectly enforced wealth tax with a comprehensive base, neither avoidance nor eva-sion pertain. The most important potential real behavioral response is in terms of saving andcapital accumulation because a wealth tax reduces the after-tax return to saving. It could alsoaffect work effort. As Section 2.2 suggests, in the absence of any tax, the wage in terms ofconsumption n years hence is w(1 + r)n, but with a wealth tax it is w [(1 + r)(1− tw)]n, sothat with a 2% tax rate an hour of work today buys only about half as much consumption

14Londoño-Velez and Avila-Mahecha (2019) is the only recent study of the impact of wealth taxation ina non-OECD country—Colombia. They find clear evidence of bunching responses of reported wealth belownotches in the tax rate structure and estimate a short-run elasticity of reported wealth with respect to the net-of-wealth-tax rate of 2. They conclude that these responses reflect predominantly avoidance and evasion, such asmisreporting wealth items subject to less third-party reporting. They also find that wealthy taxpayers increasedcompliance in response to incentives for the disclosure of previously hidden wealth, as well as in response toan exogenous increase in the risk of detection and punishment due to the publication of the Panama Papers.

12

thirty years hence compared to without the tax. There is certainly no consensus on what therelevant labor supply elasticity is, especially as those subject to the wealth tax would be atthe very upper end of the distribution. Indeed, a substantial fraction of the very wealthyare (themselves or descendants of) principals in a rather successful business venture.15 Asa result, the relevant margin is probably not hours of work in the narrow sense. Instead,it presumably pertains mainly to the incentives for entrepreneurial risk-taking and to theownership and control structure of business enterprises. One concern, for example, is that awealth tax might force entrepreneurs to continually reduce their share in a company whosevaluation increases over time.16 Even if such founders are not primarily motivated by mon-etary incentives, but instead are mostly interested in being able to realize their ideas, suchan anticipated dilution of control rights could have discouraging effects ex ante.

One might wonder whether a U.S. wealth tax might induce some people to simply moveout of the country. However, because the U.S. taxes on the basis of citizenship rather thanresidence, moving does not relieve an American citizen of any of its tax obligations—citizen-ship renunciation is required. Although there are a few famous instances, the extent of U.S.citizenship renunciation by the wealthy up to now has been very small.17 Between 2005 and2017, roughly 30,000 individuals dropped their U.S. citizenship, of whom fewer than 100reported net worth greater than $100 million (Organ, 2020). Overall, however, about one-third of those dropping citizenship were millionaires, compared with only about 12% in theU.S. population. An increase in renunciations in the 2010s was probably due to increasedenforcement of offshore financial activity prompting renunciation by dual citizens alreadyresident abroad. All in all, there is no historical precedent to help gauge the renunciationresponse to a substantial wealth tax. What is more, Senator Warren proposed a 40% exittax on the net worth above $50 million of any U.S. citizen who renounces their citizenship;Sanders proposed a 40% exit tax on the net value of all assets under $1 billion, and a 60% exittax for those with wealth exceeding $1 billion. If enforced, these measures would probablygreatly limit any potential exit responses.

5.3 Avoidance

The studies of other countries’ wealth taxes that we discussed in Section 5.1 revealed sub-stantial avoidance responses in many countries. Crucially, the extent of such a response will

15Of the wealthiest Americans on the 2018 Forbes 400 list, 69% were “self-made,” i.e., founders of theirbusiness. See Scheuer and Slemrod (2019) for further details.

16By selling shares, an additional liability for federal and state capital gains taxes is generated, which multi-plies the amount of shares that need to be sold every year in order to pay the wealth tax.

17Two such examples are Kenneth Dart’s attempt to have his Sarasota, Florida residence designated as aconsulate of Belize in the 1990s, and Facebook co-founder Eduardo Saverin dropping his U.S. citizenship infavor of Singapore just prior to the Facebook IPO in 2012.

13

depend on how broad the tax base effectively is.The range of avoidance responses includes substitution across assets if some are prefer-

entially taxed or if all are taxed but differ in their observability, as well as in ways to holdassets (e.g., through trusts). Spain offers a stark example: when it exempted some forms ofclosely-held businesses from its wealth tax base, the share of the exempted stock as a share ofall closely-held businesses increased from 15% to 77% (Alvaredo and Saez, 2009). The broadbases inherent to the U.S. proposals would preclude such effects. Still, high net-worth indi-viduals might shift into assets that are harder to value, such as keeping businesses privaterather than going public. Currently, much wealth of the Forbes 400, for example, is held inpublicly-traded stock, but this feature cannot be taken as unresponsive to a potential wealthtax. Such shifting into less visible assets would also have repercussions for our measures ofwealth inequality: it might look like a wealth tax reduces concentration when in reality itpartly shifts top wealth into forms that are less transparent.

Finally, another avenue of avoidance is inter-vivos gifts. Recall that the U.S. estate taxis actually an estate and gift tax, so that aside from an annual per-donee exemption, inter-vivos gifts are in the tax base, and research suggest that these gifts are tax-sensitive (see, forexample, Bernheim et al., 2004, and Joulfaian and McGarry, 2004).

5.4 Evasion

Another behavioral response to a wealth tax is to simply not report (all of) one’s wealth. Eva-sion of the very wealthy is difficult to uncover (and therefore difficult to measure) throughtraditional means like random audits, as the auditor typically lacks the resources to tracethe sophisticated means of evasion often involving layers of financial intermediaries. High-profile leaks from these intermediaries, such as the 2007 leak from HSBC Bank in Switzer-land and the 2015 “Panama Papers” from the firm Mossack Fonseca, have recently allowedresearchers to gain insights into tax evasion by the richest. Alstadsæter et al. (2019) usedata from these leaks along with administrative data from Norway, Sweden and Denmarkto show that evasion rates rise across the income distribution, and conclude that the top0.01% evade about 25% of the income and wealth taxes they owe. The authors link the ac-count names from the HSBC leak with individual tax data and find that 95% of these foreignaccount holders did not report the existence of the account on their tax forms.

Guyton et al. (2020) combine random audit data with data on offshore bank accounts andshow that U.S. taxpayers’ evasion through offshore financial institutions is indeed highlyconcentrated at the very top of the income distribution, and that random audits virtuallynever detect this form of evasion. According to their calculations, accounting for evasionthrough offshore financial institutions implies that random-audit estimates of the tax gap

14

for very-high-income earners should be revised upwards by 4 to 6 percentage points.Despite this new evidence about evasion among the high-wealth population, we do not

yet know the extent to which a wealth tax would cause more such evasion—the evasionelasticity. This elasticity is endogenous to the enforcement environment and, as Slemrodand Kopczuk (2002) suggest is optimal, advocates of the wealth tax argue that its impo-sition should be accompanied by expanded IRS resources. Warren and Sanders proposeda significant increase in the IRS enforcement budget, a minimum audit rate for taxpay-ers subject to the wealth tax, and systematic third-party reporting that builds on existingtax information exchange agreements adopted after the Foreign Account Tax ComplianceAct. Through threat of a punitive withholding tax for non-compliers, FATCA provides U.S.tax subjects strong incentives to report to the IRS the value and income generated by theiraccounts in foreign financial institutions. Although FATCA holds promise for restrainingtax evasion through the use of foreign accounts, its efficacy has not yet been conclusivelydemonstrated.18

How effective such expanded enforcement would be in restraining evasion was perhapsthe most controversial issue in the debate about the wealth tax proposals. Saez and Zuc-man (2019a) claimed that evasion would shrink the tax base by just 15%. Kopczuk (2019)expressed skepticism, noting that the most effective tax enforcement relies on market trans-actions reported by third parties, and this would be absent for much wealth. This is notpurely an enforcement problem because the valuation of many assets is objectively hard, asis known from estate taxation. Ingenious proposals have been put forward to address thisproblem, for example wealth owners self-setting the value of their assets but then requiringthem to stand willing to sell those assets at the price they self-assess (Posner and Weyl, 2018).Their practical effectiveness, however, remains to be demonstrated.

One difference between the wealth tax and the estate tax is that the former requires re-porting at a much higher frequency. While this potentially raises compliance costs, the up-side is that any evasion strategy must engineer an entire path of reports that is plausible ona yearly basis, notably relative to yearly income, rather than just one end-of-life snapshot.This may make it harder to systematically conceal wealth (or report it again once enforce-ment is stepped up) than in the case of the estate tax, which allows for decades of planningwithout generating much data for tax authorities.

18The international version of FATCA, known as the Automatic Exchange of Financial Account Informationin Tax Matters (the AEOI Standard), began in September 2017 and, by 2019, 94 countries had exchanged infor-mation. Johannesen et al. (2020) provide evidence on the impact of pre-FATCA enforcement policies aimed atforeign accounts.

15

5.5 Administrative and Compliance Costs

Not all of the potential economic costs of a wealth tax would come from the distortionaryeffects of behavioral responses. There also would be the cost of collection, including thecosts borne directly by taxpayers and third parties—compliance costs—and the costs bornedirectly by the government—administrative costs. Leiserson (2019) extrapolates from expe-rience with the U.S. estate tax to estimate the ratio of compliance costs to revenues from a2% wealth tax (with an exemption level of $25 million for married couples and $12.5 millionfor individual filers) at 19%, which is approximately double the conventional wisdom aboutthe U.S. income tax; he estimates administrative costs to be just 0.6% of revenue. However,since his calculation is based on the ratio of compliance costs to the gross wealth reportedon estate tax returns, it gives a mechanical disadvantage to the wealth tax simply becausethe annual wealth tax rate is much lower. Due to the fixed-cost nature of valuation and re-porting efforts, the compliance cost relative to revenue would decline if a higher wealth taxrate were to be applied.

5.6 Incidence

Related to the behavioral response is the issue of incidence. Taxes that are levied “on” thewealthy may be shifted away from them via tax-induced changes in pre-tax prices. Mostimportantly, if a wealth tax reduces capital accumulation, in the long run it may reducewage rates. Such an argument figured prominently in the debate preceding the Tax Cutsand Jobs Act of 2017, when supporters argued that the proposed cut in the rate of corporateincome tax would, via increased business investment and eventually a larger capital stock,increase average annual wages by as much as $9,000. This conclusion is highly controversial,however (see Slemrod, 2018, for the arguments made at the time).

6 Optimal Taxes on Wealth

We now turn to the normative question whether wealth should be taxed and, if yes, how.Motivated by the rise in income inequality over the last few decades, a growing literaturein public economics has started to incorporate more realistic labor markets into models ofoptimal tax policy, accounting for phenomena such as rent-seeking, skill-biased technolog-ical change, and superstar effects, to name just a few advances (see Scheuer and Slemrod,2019, for an overview). This line of work has focused on the optimal design of labor incometaxes in static models that capture recent trends in occupational sorting and wage inequal-ity. It has though stayed silent on whether these trends also shift the tradeoff for the optimal

16

taxation of capital or wealth. This is a natural question because a growing concentration ofearnings affects, through savings, the degree of wealth inequality down the road.

6.1 The Atkinson-Stiglitz Benchmark

We begin with the hypothetical scenario where all wealth inequality is driven by inequalityin labor incomes. In this case, the well-known Atkinson-Stiglitz (1976) theorem provides aclean benchmark. It states that, if a nonlinear labor income tax is available, any distortionof savings is Pareto-inefficient whenever individuals’ preferences satisfy two conditions: (i)they are separable between consumption and labor; and (ii) all individuals have the sameutility over consumption across time. In other words, if individuals only differ in theirlabor productivities, and the recent rise in wealth inequality is the result of changes in labormarkets, such as top earners being overpaid (Rothschild and Scheuer, 2016) or being able toleverage their skills on a larger scale (Scheuer and Werning, 2017), then the policy responseshould be to adjust the progressivity of labor income taxes. The taxation of capital incomeor wealth on top of that is not justified.

While conceptually useful, there are at least two reasons why this benchmark is a poorguide to actual policy. The first objection is that, in fact, a positive correlation between laborproductivities and savings propensities is plausible (Saez, 2002). This breaks the Atkinson-Stiglitz result because it violates condition (ii), for example because individuals differ intheir discount rates in a way that is related to earnings abilities. It lends support to theadditional taxation of capital, in particular when heterogeneity is multi-dimensional, i.e.,when even individuals with the same history of labor incomes differ in their accumulatedwealth (Golosov et al., 2009, Diamond and Spinnewijn, 2011).

In our view, there is a second reason that is at least as relevant in practice. It takes intoaccount that policymakers who now consider a potential tax reform face a situation withpre-existing wealth inequality. In other words, it is not the case that future wealth is entirelydetermined by just future labor incomes. Instead, individuals already differ in the wealththey own, either because they have inherited it from previous generations or because theythemselves have saved in the past. When deciding how to design tax policy going forward,we need to recognize that future wealth inequality will be shaped partly by this initial wealthinequality, which from today’s perspective must be taken as given.19

19Other objections to the Atkinson-Stiglitz benchmark include the notion that disentangling labor and capitalincome can be challenging in practice (Piketty et al., 2014), or that there are structural constraints to reformingthe labor income tax.

17

6.2 Taxing Existing Wealth

In principle, pre-existing wealth inequality could be redistributed in a lump-sum fashionthrough a one-time, unanticipated wealth tax. Indeed, historically, various countries haveused one-time wealth taxes to deal with revenue requirements, such as war-time spendingshocks. In 1999, Donald Trump, then a candidate for the Reform Party presidential nomi-nation, proposed a 14.25% one-time “net worth tax” on individuals and trusts worth morethan $10 million in order to eliminate U.S. national debt in one swoop. More recently, callshave been made for a time-limited, progressive wealth levy to stem the fiscal burden fromthe Coronavirus pandemic (see e.g. Landais et al., 2020).

From an optimal tax perspective, such policies are attractive because they avoid behav-ioral distortions by only touching wealth that has already been accumulated. But this ap-pealing feature critically hinges on policymakers’ ability to implement them on short noticeand on their commitment not to make them permanent or reintroduce them when similartimes come about in the future. In the past, originally one-off war taxes have often turnedinto long-lasting tax policies.

6.3 Taxing Future Wealth Accumulation

Against this background, suppose that an unexpected, distortion-free redistribution of ex-isting wealth is not feasible. One policy option then is to adjust the labor income tax goingforward. For example, if initial wealth and earnings abilities are positively correlated, amore progressive labor income tax could be used to target both determinants of inequality,at least indirectly. The alternative is to introduce a tax on future wealth accumulation, whichwill of course distort individuals’ savings incentives. Will such a wealth tax then be part ofthe optimal tax mix, even if preferences satisfy conditions (i) and (ii)? We briefly lay out thatthe answer is yes and that, in fact, the optimal schedules of the wealth and labor income taxare closely linked.20

Suppose individuals live for two periods t = 0, 1: They choose how much to work,consume and save in the first period and only consume out of their (after-tax) savings in thesecond. This could be interpreted as a parsimonious life-cycle model with the first periodcorresponding to working life and the second to retirement. Individuals differ in their laborproductivity θ, so their preferences are

u(c0)− h(y0/θ) + βu(c1),20Cremer et al. (2001) observe that the Atkinson-Stiglitz uniform commodity taxation result breaks down

when individuals have heterogeneous endowments and offer simulations for a four-type model.

18

where u(ct) is the consumption utility in period t, h(y0/θ) is the disutility of labor in period0 (y0 is labor income), and β is a common discount factor. Individual θ is born with wealthk0(θ) in period 0 and there is a linear savings technology with a common gross return R =1 + r.

If a direct tax on initial wealth is ruled out, the remaining policy instruments are non-linear taxes on period-0 labor income Ty(y0) and on period-1 wealth Tw(Rk1). Individuals’budget constraints are

c0 = y0 − Ty(y0) + k0 − k1

c1 = Rk1 − Tw(Rk1).

One can derive the following formula for the marginal wealth tax in any Pareto optimum:

T′w(Rk1(θ)) =T′y(y0(θ))

1− T′y(y0(θ))

[σ(θ)

α(θ)η(θ)

(1 +

1ε(θ)

)− 1

]−1. (1)

Here, σ is the intertemporal elasticity of substitution, α = k0/c0 measures the share ofperiod-0 consumption financed out of initial wealth, η = k′0(θ)θ/k0(θ) is the elasticity ofinitial wealth with respect to earnings ability, and ε is the Frisch elasticity of labor supply.

The formula is intuitive. First, if there is no initial wealth inequality then η = 0, so wereturn to the Atkinson-Stiglitz benchmark with T′w(Rk1) = 0 and all desired redistribution isachieved through the labor income tax. The same is true when the intertemporal substitutionelasticity σ is infinite (the saving distortions induced by a wealth tax explode) or when theFrisch elasticity ε is zero (inelastic labor supply implies that the labor tax is lump-sum, sothere is no need for an additional wealth tax).

Second, more generally, formula (1) links the shapes of the wealth and labor income taxschedules at any optimum (as both are driven by the redistributive motives of the govern-ment). But the term in square brackets introduces a wedge between the two. For instance,suppose σ and ε are fixed parameters.21 Then any variation in this term is pinned downby how αη varies across the distribution. If αη, which summarizes the importance of ini-tial wealth relative to labor income inequality, is increasing towards the top, the wealth taxshould be more progressive than the labor income tax, and vice versa.22

These results show that, unless there is already a fully equalized wealth distribution, it is

21This would be the case when u(c) exhibits constant relative risk aversion and the disutility of labor isiso-elastic.

22It can also be shown that T′w(Rk1) > 0 under standard redistributive motives. Moreover, the parameters informula (1) can all be connected to empirical statistics; notably, η can be backed out from the joint distributionof wealth and income. For example, the top marginal wealth tax rate tw can be written as

tw =ty

1− ty

[σρw + ερy

εαρy− 1

]−1

19

generally optimal to introduce progressive taxes on future wealth accumulation in additionto labor income taxes, despite their distortive effects, at least for some amount of time. Thereason is that, in the short to medium term, future wealth inequality still partly reflects pre-existing wealth inequality. In the long run, of course, this effect diminishes. Indeed, theinfluential Chamley-Judd result argued that zero capital taxation is optimal in the long run,in Judd (1985) even in the face of extreme wealth inequality.23 However, Straub and Werning(2020) have recently demonstrated that this result is invalid whenever the intertemporalelasticity of substitution is at most one (the empirically relevant case): the long-run tax oncapital should in fact be positive and significant. For higher elasticities, it converges to zeropossibly at a very slow rate—after centuries of high tax rates.24

6.4 Wealth versus Capital Income Taxes

Taking stock of the preceding discussion, we conclude that the modern theory of optimaltaxation lends support to taxing wealth accumulation. However, this statement does notpin down the appropriate tax instruments to use for the purpose. As discussed in Section2.3, in standard models such as the one above, the wealth tax Tw(Rk1) is equivalent to atax on capital income rk1 and to a tax on period-1 consumption c1. It also corresponds to abequest tax when re-interpreting the model in terms of parents living in t = 0 and childrenin t = 1 (Farhi and Werning, 2010).25 Given that most countries already have progressivecapital income taxes, for instance, what might justify levying a wealth tax instead, or inaddition?

Guvenen et al. (2018) provide such an argument when there is heterogeneity in the re-turn to capital, such as when r is interpreted as the return to a real investment. In this case,the capital-income-tax equivalent tw/r of a given wealth tax rate tw is also heterogeneous:the higher is the return r, the lower is the equivalent capital income tax, so that more pro-ductive entrepreneurs face a lower capital-income-tax equivalent. As a result, a wealth taxencourages the reallocation of capital from unproductive to productive entrepreneurs, andthe authors find efficiency gains of 8% compared to a uniform capital income tax.

There is, however, an opposing effect. If heterogeneous returns reflect heterogeneouswindfall gains, rents or excess profits (e.g. due to market power or inside information),rather than actual productivity differences, then taxing those away has well-known effi-

where ty is the top marginal income tax and ρw and ρy are the Pareto tail coefficients of the wealth and incomedistribution, respectively.

23Chamley (1986) instead considered long-run capital taxation in a representative agent framework.24Saez and Stantcheva (2018) show that when wealth enters utility directly, on top of the consumption it

finances, the optimal long-run capital tax is also positive.25A tax on initial wealth k0 is equivalent to a uniform consumption tax in both periods, coupled with a

subsidy on labor income.

20

ciency benefits. But a wealth tax gets this exactly reversed—it taxes the normal rate of returnand leaves the excess returns untouched. For example, if all investors have a real rate of re-turn of 2%, but some earn additional excess profits on their investments, then a 2% wealthtax would not target any of those rents, whereas a capital income tax would. This is becausea wealth tax is equivalent to a unit tax on the rate of return rather than an ad valorem tax.A related issue is that much of what shows up as return to capital on the tax returns of thesuperrich (e.g. in the form of realized capital gains) is arguably compensation to labor. InScheuer and Slemrod (2019), we argue that converting labor income into preferentially-taxedcapital gains is a key margin of behavioral response to taxes at the top, which we refer to asthe plasticity of the tax base. A wealth tax again only taxes some normal return, whereas acapital income tax hits the full extent of such shifted labor compensation.

In sum, the tradeoff between wealth and capital income taxes depends on the source ofdifferences in rates of return. Whereas progress has been made in measuring the extent ofreturn heterogeneity (see e.g. Fagereng et al., 2020), a decomposition into actual produc-tivity differences versus differential rents or shifted labor compensation has not yet beenaccomplished.

7 Political Economy

The recent focus on the extent and growth of inequality has raised the possibility that it maylead to political instability. Two rather different mechanisms for such an effect have beenconsidered. On the one hand, Farhi et al. (2012), Piketty (2014), and Scheuer and Wolitzky(2016) have taken up concerns going back to Karl Marx, who predicted that an increasein the concentration of capital would lead to a political revolution and eventually radicalredistribution. On the other hand, there have been fears that excessive inequality mightallow the rich to capture the political system and tilt it in their favor (see e.g. Gilens, 2014and Bartels, 2016). We briefly consider both concerns and whether they can justify a wealthtax.

7.1 Politically Sustainable Tax Policy

Our analysis in Section 6 has assumed that the government in charge at the beginning oftime can determine the entire future path of taxes. In practice, though, tax policy is not set instone once and for all. In a democracy, for example, it must regularly stand the test of elec-tions. This raises the question whether a course of tax policy—and the resulting economicinequality—is politically sustainable, in the sense that it maintains the support of a majorityof citizens over time. Scheuer and Wolitzky (2016) study the optimal sustainable policy and

21

show that it involves a positive marginal tax on the wealth accumulation of the rich, whilesubsidizing that of the middle class.

Take, for instance, the two-period model from Section 6.3. In the second period, thereis always a temptation to reform whatever policy has been announced originally towards agreater equalization of wealth because, at this point, wealth accumulation is sunk. Policy-makers in the first period realize that letting wealth concentration explode will lead to anunstable situation in the future, where a majority of voters will prefer a progressive reform.Anticipating this, individuals would save very little in the first place, leading eventually toa poor outcome for everyone. Hence, from an ex ante perspective, it is better to design taxpolicy in such a way that it is sustainable, i.e., such that it will defeat a future reform threat interms of popular support. This is achieved by reducing wealth inequality through a tax onthe savings of the rich, and by creating a middle class that accumulates just enough wealthto become unfavorable to more extreme redistribution in the future.

From this perspective, the role of the wealth tax is not to redistribute existing wealth expost, as discussed in Section 6.2. It is about making the system more stable, with limitedinequality, so that it can resist the threat of political upheaval looming in the background.Such threats were important drivers of tax and welfare state policies in 19th and 20th centuryEurope, when the socialist movement gained momentum (Esping-Andersen, 1990), and theyare palpable today in many South American countries (Venezuela being an example that wasnot resilient, with disastrous consequences). Whether they could emerge in other countriesas well, including the U.S., remains an open question. But the above logic shows that thereare political risks in letting inequality grow unchecked and that some compression of thewealth distribution can help mitigate them.26

7.2 Wealth and Political Power

It is a fundamental observation in political economy that wealth can buy political influence.The concern is that, even in a “one-man, one-vote” democracy, billionaires can affect politicsmore than others through campaign contributions, ownership of media outlets, or fundinglobbying activities. Indeed, Gilens (2014) and Bartels (2016) collect evidence that politicaldecisions often are more sensitive to the preferences of the rich than those of the medianvoter. Accordingly, proponents of a progressive wealth tax have argued that reducing thewealth of the superrich is a desirable objective in itself, beyond the revenue it could raise.Bernie Sanders famously said “I don’t think billionaires should exist” (New York Times,

26In both Farhi et al. (2012) and Scheuer and Wolitzky (2016), a wealth tax and a capital income tax areequivalent instruments. However, if political instability is an urgent problem, annual wealth taxes are able tocompress the wealth distribution relatively quickly because they can correspond to capital income taxes of arate higher than 100% and have an immediate effect compared to, say, bequest taxes.

22

2019). In the same vein, Saez and Zucman (2019b) have proposed to set top tax rates abovethe revenue-maximizing rate, destroying part of the tax base in the interest of preventing an“oligarchic drift” that would otherwise undermine democracy.

Even if concerns about an extreme concentration of wealth and political power are war-ranted, the jury is still out on how a wealth tax would help fix the problem. Other in-struments may be better targeted at ensuring a more equal political representation, such asregulating campaign contributions and Super PACs. Some European countries with similardegrees of inequality offer examples of democracies where money plays a smaller role inpolitics than in the United States. One particular concern with the wealth tax is that it mightencourage political donations (as they reduce tax liability) and thereby further stimulatepolitical engagement of wealthy individuals.27

8 Policy Alternatives

Several alternatives to a wealth tax have been proposed to achieve its primary goal of in-creasing the progressivity of the U.S. tax system. As discussed in Section 2, the current taxsystem exhibits various structural deficiencies in the treatment of capital gains. Capital gainsare extremely concentrated and represent a large share of the income of the superrich. Atthe same time, they benefit from various provisions that imply preferential taxation: lowertax rates, taxation upon realization, and step-up of basis at death. It is no surprise that manyhave therefore proposed to fix these defects in order to restore tax progressivity.

Indeed, Joe Biden, the eventual nominee of the Democratic Party, has released plans totax capital gains and dividends at the same rate as ordinary income for taxpayers with in-comes exceeding $1 million and tax unrealized capital gains at death. His plan would alsoincrease income tax rates for taxpayers with incomes over $400,000 from 37% to 39.6%, sub-ject earnings over $400,000 to the Social Security payroll tax, and limit the value of itemizeddeductions for high-income taxpayers. Hence, the top marginal tax rate for capital gainswould increase from 20% to 39.6%. However, his plan does not include a wealth tax. Oneanalysis of Biden’s tax proposals estimated that, for the top 0.1% of taxpayers by income,federal taxes would go up by about $1.8 million per year, on average (Tax Policy Center,2020); this compares to an average additional tax burden for the top 0.1% of householdsunder Warren’s plan of $5 million per year.

While Biden’s plan would eliminate two of the preferential provisions for capital gains,it would retain the current system of taxation based on realization rather than accrual and

27There is also some tension between the claim that the wealth tax is not about collecting revenue but ratherabout reducing very high wealth levels versus the claim that it would raise substantial amounts of revenuethat could be used to fund major redistributive initiatives (see Section 2.2).

23

thereby preserve the within-a-lifetime tax deferral advantage. Moreover, a realization-basedtreatment leads to very low tax burdens for superrich individuals who neither receive muchordinary nor dividend income nor sell many shares of the businesses they own. An exampleis Warren Buffett, who in 2015 released his tax return showing an AGI of $11.6 million, atiny fraction of his wealth of $62 billion in that year according to Forbes. In view of this,calls for the taxation of accrued capital gains have been made. For example, Batchelder andKamin (2019) offer a menu of “incremental” revenue options, including an accrual-basedcapital gains tax consisting of an annual mark-to-market tax on publicly-traded assets plusa retrospective accrual tax for illiquid assets.28

9 Conclusion

Concerns about inequality and its consequences have led in 2019 to widely-discussed pro-posals for highly progressive annual wealth taxes from two leading presidential candidatesof the Democratic Party. As their candidacies faltered, so did attention to their proposals,although the coronavirus pandemic has revived calls for a one-time wealth tax to help fundthe massive fiscal interventions it has engendered. These ideas certainly warrant a closelook.

Although related taxes—annual taxes on immovable property and taxes on transfers ofwealth at death—are a staple of U.S. public finance, the U.S. has no experience with com-prehensive annual wealth taxes on which to draw. While a dozen OECD countries levieda wealth tax in the recent past, now only three have them, with Switzerland raising a com-parable fraction of revenue as the U.S. proposals, albeit with much lower rates along witha much lower level of wealth exempted. A handful of studies have addressed the conse-quences of these wealth taxes. Several of them find a substantial behavioral response sug-gesting a large excess burden, with the anatomy of the response—real versus avoidanceversus evasion—varying a lot, likely because of differences in the broadness of the tax baseand enforcement details such as the extent of third-party reporting of wealth.

The former conventional wisdom—that an optimal tax system would feature no tax oncapital income, much less a substantial wealth tax—has been overturned, and we presentvarious theoretical arguments that justify progressive taxes on wealth accumulation bothin the short and long run. Whether they should take the form of a wealth tax depends onthe costs of implementing it and the attractiveness of alternative policies aimed at the same

28Under such a scheme, tax is assessed upon realization, but the statutory tax rate rises as the holding periodlengthens, effectively charging interest on past gains when realization occurs. This eliminates the need tovalue assets except when sold while minimizing liquidity problems and the incentive to defer such realization(Auerbach, 1991).

24

objective. The available evidence suggests that the costs might be substantial. Supportersof the U.S.-style proposals discount the relevance of this evidence on the grounds that theproposal details, especially the rate schedule, broadness of the base, and enforcement provi-sions, are very different from any previous wealth tax. This is true, but it is a double-edgedsword, as its unprecedented design features also make it difficult to learn from experienceand to predict its consequences with much confidence.

References

Alstadsæter, A., N. Johannesen, and G. Zucman (2019), “Tax Evasion and Inequality,”American Economic Review 109, 2073–2103.

Alvaredo, F. and E. Saez (2009), “Income and Wealth Concentration in Spain from a His-torical and Fiscal Perspective.” Journal of the European Economic Association 7, 1140-1167.

Atkinson, A. and J. Stiglitz (1976), ”The Design of Tax Structure: Direct versus IndirectTaxation.” Journal of Public Economics 6, 55-75.

Auerbach, A. (1991), “Retrospective Capital Gains Taxation,” American Economic Review81, 167-178.

Bartels, L. (2016), Unequal Democracy, Political Economy of the New Gilded Age, PrincetonUniversity Press.

Batchelder, L. and D. Kamin (2019), "Taxing the Rich: Issues and Options," Mimeo.Bernheim, B. D., R. Lemke, and J. K. Scholz (2004), “Do Estate and Gift Taxes Affect the

Timing of Private Transfers?” Journal of Public Economics 88, 2617-2634.Bricker, J., A. Henriques, J.A. Krimmel, and J.E. Sabelhaus (2016), “Measuring Income

and Wealth at the Top Using Administrative and Survey Data,” Brookings Papers on EconomicActivity 1, 261-331.

Brülhart, M., J. Gruber, M. Krapf, and K. Schmidheiny (2019), “Behavioral Responses toWealth Taxes: Evidence from Switzerland,” NBER Working Paper 22376.

Chamley, C. (1986), “Optimal Taxation of Capital Income in General Equilibrium withInfinite Lives,” Econometrica 54, 607-622.

Chirico, M., Inman, R. P., Loeffler, C., MacDonald, J., and Sieg, H. (2016), “An Experimen-tal Evaluation of Notification Strategies to Increase Property Tax Compliance: Free-Ridingin the City of Brotherly Love,” Tax Policy and the Economy, 30(1), 129-161.

Diamond P. and J. Spinnewijn (2011) “Capital Income Taxes with Heterogeneous Dis-count Rates” American Economic Journal: Economic Policy 3, 52-76.

Drometer, M., M. Frank, M. Hofbauer Pérez, C. Rhode, S. Schworm, and T. Stitteneder(2018), “Wealth and Inheritance Taxation: An Overview and Country Comparison,” Ifo DICEReport 16.

25

Duran-Cabré, J., A. Esteller-Moré, and M. Mas-Montserrat (2019), “Behavioural Responsesto the (Re-)Introduction of Wealth Taxes. Evidence From Spain,” IEB Working Paper.

Eller, M., B. Erard, and C.-C. Ho, (2001), “Noncompliance with the Federal Estate Tax,”Rethinking Estate and Gift Taxation, 375-410.

Esping-Andersen, G. (1990), The Three Worlds of Welfare Capitalism, Princeton UniversityPress.

Fagereng, A., L. Guiso, D. Malacrino, and L. Pistaferri (2020), “Heterogeneity and Persis-tence in Returns to Wealth,” Econometrica 88, 115-170.

Farhi, E. and Werning, I. (2010), “Progressive Estate Taxation,” Quarterly Journal of Eco-nomics 125, 635-673.

Farhi, E., C. Sleet, I. Werning, and S. Yeltekin (2012), “Non-linear Capital Taxation With-out Commitment,” Review of Economic Studies 79, 1469–1493.

Föllmi R. and I. Martinez (2017), “Volatile Top Income Shares in Switzerland? Reassess-ing the Evolution between 1981 and 2010,” Review of Economics and Statistics 99, 793-809.

Gilens, M. (2014), Affluence and Influence, Economic Inequality and Political Power in America,Princeton University Press.

Golosov, M., M. Troshkin, A. Tsyvinski, and M. Weinzierl (2013), “Preference Hetero-geneity and Optimal Capital Income Taxation,” Journal of Public Economics 97, 160-175.

Guvenen, F., G. Kambourov, B. Kuruscu, S. Ocampo, and D. Chen (2018), “Use It or LoseIt: Efficiency Gains from Wealth Taxation,” Working Paper, University of Minnesota.

Guyton, J., P. Langetieg, D. Reck, M. Risch, and G. Zucman (2020), “Tax Evasion by theWealthy: Measurement and Implications,” in Measuring and Understanding the Distributionand Intra/Inter-Generational Mobility of Income and Wealth, R. Chetty, J. Friedman, J. Gornick,B. Johnson, and A. Kennickell, editors.

Harris, B.H. and B.D. Moore (2013), “Residential Property Taxes in the United States,”Urban-Brookings Tax Policy Center Working Paper.

Herd, R. and T. Thorgeirsson (2001), “Increasing Simplicity, Neutrality and Sustainabil-ity: A Basis for Tax Reform in Iceland,” OECD Economics Department Working Paper No.292.

Jakobsen, K., K. Jakobsen, H. Kleven, and G. Zucman (2020), “Wealth Taxation andWealth Accumulation: Theory and Evidence From Denmark,” Quarterly Journal of Economics,135(1) 329-388.

Jensen, E. (2019), “Is a Tax on Wealth Constitutional?” Journal of Taxation of Investments,Spring 36(3), 79-86.

Johannesen, N., P. Langetieg, D. Reck, M. Risch and J. Slemrod (2020), “Taxing HiddenWealth: The Consequences of U.S. Enforcement Initiatives on Evasive Foreign Accounts,”forthcoming, American Economic Journal: Economic Policy.

26

Johnsen, D., and W. Dellinger (2018), “The Constitutionality of a National Wealth Tax,”Indiana Law Journal 93.

Joulfaian, D., and K. McGarry (2004), “Estate and Gift Tax Incentives and Inter-vivosGiving,” National Tax Journal, 429-444.

Judd, K. (1985), “Redistributive Taxation in a Simple Perfect Foresight Model,” Journal ofPublic Economics 28, 59–83.

Kaldor, N. (1956), Indian Tax Reform, Ministry of Finance, Government of India.Kelly, R. (2012), “Property Tax Collection and Enforcement,” In A primer on property tax:

Administration and policy, W. McCluskey, G. Cornia, and L. Walters (eds.), Blackwell Publish-ing, 141-171.

Kopczuk, W. (2013), “Taxation of Intergenerational Transfers and Wealth,” in A. Auer-bach, R. Chetty, M. Feldstein, and E. Saez, eds., Handbook of Public Economics, Vol. 5, Elsevier.

Kopczuk, W. (2015), “What Do We Know About the Evolution of Top Wealth Shares inthe United States?” Journal of Economic Perspectives 29(1), 47-66.

Kopczuk, W. (2019), “Comment on ‘Progressive Wealth Taxation’ by Saez and Zucman,”Brookings Papers on Economic Activity, forthcoming.

Landais, C., E. Saez and G. Zucman (2020), “A Progressive European Wealth Tax to Fundthe European COVID Response,” https://voxeu.org/article/progressive-european-wealth-tax-fund-european-covid-response

Leiserson, G. (2019), “Taxing Wealth,” Mimeo, Washington Center for Equitable Growth.Londono-Velez, J. and J. Ávila-Mahecha (2019), “Can Wealth Taxation Work in Develop-

ing Countries? Quasi-Experimental Evidence from Colombia,” Mimeo.New York Times (2019), “Bernie Sanders Proposes a Wealth Tax: ‘I Don’t Think That Bil-

lionaires Should Exist’,” https://www.nytimes.com/2019/09/24/us/politics/bernie-sanders-wealth-tax.html

OECD (2018), The Role and Design of Net Wealth Taxes in the OECD.Organ, P. (2020), “Expatriation and the U.S. Tax System,” Mimeo, University of Michigan.Piketty, T. (2014), Capital in the Twenty-first Century, Harvard University Press.Piketty, T., Saez, E. and Stantcheva, S. (2014), “Optimal Taxation of Top Labor Incomes:

A Tale of Three Elasticities,” American Economic Journal: Economic Policy 6, 230-271.Posner, E., and E. G. Weyl (2018), Radical Markets: Uprooting Capitalism and Democracy For

a Just Society, Princeton University Press.Rothschild, C. and F. Scheuer (2016), “Optimal Taxation with Rent-Seeking,” Review of

Economic Studies 83, 1225– 1262.Saez, E. (2002), “The Desirability of Commodity Taxation Under Non-Linear Income Tax-

ation and Heterogeneous Tastes,” Journal of Public Economics, 83, 217-230.

27

https://voxeu.org/article/progressive-european-wealth-tax-fund-european-covid-responsehttps://voxeu.org/article/progressive-european-wealth-tax-fund-european-covid-responsehttps://www.nytimes.com/2019/09/24/us/politics/bernie-sanders-wealth-tax.htmlhttps://www.nytimes.com/2019/09/24/us/politics/bernie-sanders-wealth-tax.html

Saez, E. and S. Stantcheva (2018), “A Simpler Theory of Optimal Capital Taxation” Journalof Public Economics 162, 120-142 .

Saez, E. and G. Zucman (2016), “Wealth Inequality in the United States Since 1913: Evi-dence From Capitalized Income Tax Data,” Quarterly Journal of Economics 131, 519–578.

Saez, E. and G. Zucman (2019a), “Progressive Wealth Taxation,” Brookings Papers on Eco-nomic Activity, forthcoming.

Saez, E. and G. Zucman (2019b), The Triumph of Injustice: How the Rich Dodge Taxes andHow to Make Them Pay, Orell Fuessli.

Scheuer, F. and J. Slemrod (2019), “Taxation and the Superrich,” forthcoming, AnnualReview of Economics.

Scheuer, F. and I. Werning (2017), “The Taxation of Superstars,” Quarterly Journal of Eco-nomics 132, 211-270.

Scheuer, F. and A. Wolitzky (2016), “Capital Taxation under Political Constraints,” Amer-ican Economic Review 106, 2304–2328.

Seim, D. (2017), “Behavioral Responses to Wealth Taxes: Evidence from Sweden,” Amer-ican Economic Journal: Economic Policy 9, 395–421.

Slemrod, J. (2001), “A General Model of the Behavioral Response to Taxation,” Interna-tional Tax and Public Finance 8, 119-128.

Slemrod, J. (2018), “Is This Tax Reform, or Just Confusion?” Journal of Economic Perspec-tives, 32(4), 73-96.

Slemrod, J. and W. Kopczuk (2002), “The Optimal Elasticity of Taxable Income,” Journalof Public Economics, 84, 91–112.

Smith, M., O. Zidar, and E. Zwick (2020), “Top Wealth in America: New Estimates andImplications for Taxing the Rich,” Mimeo.

Straub, L., and I. Werning (2020), "Positive Long-Run Capital Taxation: Chamley-JuddRevisited," American Economic Review 110, 86-119.

Tax Policy Center (2020), “An Analysis of Former Vice President Biden’s Tax Proposals,”https://www.taxpolicycenter.org/publications/analysis-former-vice-president-bidens-tax-proposals/full

Tiebout, C. M. (1956), "A Pure Theory of Local Expenditures,” Journal of Political Economy64, 416-424.

Waldenström, D. (2018), “Inheritance and Wealth Taxation in Sweden,” Ifo DICE report16.

Wolff, E. (2017), A Century of Wealth in America, Harvard University Press.Zoutman, F. (2018), “The Elasticity of Taxable Wealth: Evidence from the Netherlands,”

Working Paper, Norwegian School of Economics.

28

https://www.taxpolicycenter.org/publications/analysis-former-vice-president-bidens-tax-proposals/fullhttps://www.taxpolicycenter.org/publications/analysis-former-vice-president-bidens-tax-proposals/full

IntroductionWhat is a Wealth Tax?Base and RatesObjectivesInterpreting the Rate of Tax

What Other Taxes is it Like?Property TaxesEstate and Inheritance Taxes

Wealth Taxes ElsewhereEuropean Wealth TaxesThe Swiss Example

Consequences of a Wealth TaxEvidence on the Response to Wealth TaxesReal Behavioral ResponsesAvoidanceEvasionAdministrative and Compliance CostsIncidence

Optimal Taxes on WealthThe Atkinson-Stiglitz BenchmarkTaxing Existing WealthTaxing Future Wealth AccumulationWealth versus Capital Income Taxes

Political Economy Politically Sustainable Tax PolicyWealth and Political Power

Policy AlternativesConclusion