No More Tax-Free Lunch for Billionaires: Closing the Borrowing Loophole

I. Introduction

Larry Ellison, co-founder of Oracle Corp. and the world’s fourth richest person, is worth $142 billion.¹ Ellison lives a lavish lifestyle: He built a yacht for $270 million, just one of his many homes is worth $200 million, and he famously bought a Hawaiian island for $300 million.² Yet he has been taxed relatively little.³ This is a general problem. Between 2014 and 2018, Warren Buffett, Jeff Bezos, Michael Bloomberg, and Elon Musk collectively got richer by $160 billion and paid just $1.7 billion in taxes — 1.1 percent of their increase in wealth.⁴ In a time of large budget deficits and deep concern about inequality in the United States, how is it possible that these billionaires pay so little in taxes?

Ellison can pay little in taxes because he, like many billionaires, takes advantage of a simple loophole in the tax code: He borrows against his billions of dollars in assets — which does not require paying any income tax. As of August 2023, Ellison had pledged Oracle stock worth over $30 billion to secure “personal indebtedness.”⁵ Ellison is not unique in this practice. A 2021 Forbes analysis concluded that about 20 percent of Forbes 400 billionaires whose primary wealth was in publicly traded firms had pledged substantial amounts of stock to secure borrowing.⁶ As of November 2021, that group had pledged stock worth $185 billion.

We propose a simple fix to this borrowing loophole: tax billionaires’ (and hundred-millionaires’) borrowing. That way, billionaires like Ellison will pay tax on their income just like ordinary Americans do.

Ellison’s borrowing strategy takes advantage of the typical requirement in the U.S. tax code that assets — like the billions of dollars in highly appreciated Oracle stock held by Ellison — must be sold or exchanged for gains to be taxed. This rule results in an average tax rate of 8.2 percent among the country’s richest 400 families when counting gains in asset value as income.⁷ The rule is based partly on the notion, correct or incorrect, that gains aren’t real until they become cash. But that argument does not apply when billionaires borrow against their gains and use the money. Ellison hasn’t just gotten richer on paper when he borrows against his stock to buy a Hawaiian island; he’s used that income just as if he’d sold the stock.⁸ Indeed, under current law, Ellison — despite using the stock appreciation to purchase his island — will never pay income tax on those gains if he holds the stock until his death. This is a tax avoidance strategy known as “buy, borrow, die,” and it can allow the wealthiest Americans to live lavishly while paying little income tax.⁹

Americans understand this argument. In a nationally representative survey we conducted, most respondents supported taxing appreciated assets used in borrowing by the rich, even if the tax applies to those with “only” $10 million in assets.¹⁰ The idea is intuitive enough that Trevor Noah argued on The Daily Show that the current treatment of billionaires’ borrowing is unjust.¹¹

Taxing the borrowing of the superrich would be very progressive, and we estimate that it could raise about $100 billion over 10 years.

This article shows how the “billionaire borrowing tax” could be designed and effectively administered, explains how it is a relatively economically efficient tax, relates the tax to other proposals, responds to potential critiques, and discusses various design features.

II. Main Proposal

Turning back to Ellison, we use illustrative numbers to show how the tax would work. Suppose for simplicity that all his $140 billion wealth takes the form of Oracle stock. Let’s also suppose that Ellison originally paid $1 billion for his earliest-purchased stock, now worth $10 billion. In other words, for those shares, 90 percent of the value is gains that have not been taxed. Now suppose that he borrows $10 billion, which is only possible because banks know he has the (appreciated) assets to repay. We propose that Ellison pay income tax on his borrowing by realizing gains on his assets.

The tax would apply even if Ellison did not pledge his stock (or other assets) as a security interest. Recourse borrowing implicitly is always backed by all a taxpayer’s assets, regardless of any security interest. Moreover, if the borrowing tax only applied to secured borrowing, covered taxpayers could easily avoid it by borrowing without pledging assets, even as their hundreds of millions of dollars of assets would often still back up the unsecured borrowing sufficiently for the lender. Similarly, the borrowing tax would not look only at gains on the assets securing a given loan. If it did, covered taxpayers could avoid the tax by pledging only high-basis assets, even as the other assets continued to implicitly backstop the loan.¹² Instead, we propose starting with the pool of all taxpayers’ assets.¹³

The tax would be calculated as follows. Since Ellison is getting $10 billion in cash through his stock ownership, the tax code would conclude that $10 billion of the stock has been realized (that is, become taxable). When income is realized, the tax code subtracts the basis, which is the amount that the owner paid for the asset and has already paid tax on. We propose imposing a first-in, first-out assumption. We thus suggest using taxpayers’ basis in their oldest assets to calculate gain under the tax. Compared with using the basis in all assets, and thus having to value all assets (as would be the case if our proposal used a taxpayer’s average basis), the FIFO proposal has considerably lower compliance costs because it requires only valuing the portion of assets whose total value equals the borrowing amount. In the example, because Ellison’s oldest tranche of stock was purchased for $1 billion and is now worth $10 billion, he would get a basis offset of $1 billion on his loan. Thus, he would pay income tax on $9 billion of gains when he borrows $10 billion ($10 billion realized minus $1 billion of basis).

In return, when or if (in light of “buy, borrow, die”) Ellison sells his stock, he pays less income tax. That $9 billion of income that is taxed becomes new basis allocated to his oldest stock. Going forward, when Ellison borrows additional amounts, his basis offset is calculated based on his oldest assets not yet revalued under the borrowing tax.

We would limit the valuation of gains on assets to “major assets.” We define major assets as significant shares in business interests or other major holdings (for example, land). The implication is that taxpayers would skip over assets like homes and art that are more costly to value as taxpayers apply FIFO in first valuing their earliest-purchased assets and then moving on to their latest-purchased ones. Applying a major assets limit reduces compliance costs by not requiring the valuation of every asset, no matter how small. In other words, it avoids the “every teaspoon” problem, which can raise compliance costs for estate taxes.¹⁴

We would apply this tax to borrowing for any purpose by households with net assets above $100 million. The tax would be phased in for households with $100 million to $200 million in net assets, with each increase of $1 million of net assets beyond $100 million corresponding to a 1 percentage point increase in the share of borrowing covered by the tax.¹⁵

The tax would apply to all future borrowing as well as borrowing presently on the books. For the tax on existing borrowing, taxpayers would have five years to pay, with the ability to pay in five equal installments. In future years, taxpayers would be taxed on increases¹⁶ in borrowing.

To reduce compliance and administrative costs, covered households would have lifetime and annual exemptions, similar to the gift tax. Using the lifetime exemption, covered households could borrow up to $1 million without triggering tax. The lifetime exemption would apply directly to the tax on existing borrowing, with borrowing exceeding $1 million being taxable. In addition, each year, covered households would be able to borrow up to $200,000 without using a part of their lifetime exemption.¹⁷ These figures were selected to minimize avoidance opportunities while avoiding costly revaluations of taxpayers’ major assets.

To administer the tax, covered taxpayers would report their borrowing in the same way that those with mortgage interest report their borrowing. Third-party reporting from banks on the lending could back this up, but audits could be sufficient to ensure compliance.

Everything else in the tax system stays the same. Covered taxpayers would still get interest deductions if they borrow for business purposes (for example, if Elon Musk borrows to buy Twitter).

In the Appendix, we describe the scoring for the borrowing tax. We estimate that the proposal would raise $102 billion in the decade following enactment.¹⁸

To reach this figure, we assume that, following the current green book proposals, the highest-income taxpayers — including those $100 millionaires and billionaires covered by the proposal — will pay ordinary income tax rates on their capital gains. We also assume that, as proposed, the top marginal rate returns to 39.6 percent and that the net investment income tax increases to 5 percent, producing a tax rate of 44.6 percent. It is helpful to have the rate for the borrowing tax equal to the tax rate on capital gains so that closing the borrowing loophole essentially closes off the usefulness of borrowing to avoid selling and paying taxes on capital gains. Relatedly, closing the loophole likely pushes up the revenue-maximizing capital gains tax rate by shutting down this avoidance technique, thus complementing the proposed rate increase.

We estimate that covered taxpayers will have borrowed about $260 billion as of April 2024.¹⁹ This borrowing is explicitly or implicitly backed by assets, roughly 55 percent of which are unrealized gains. Put differently, ultrawealthy Americans have likely accessed roughly $140 billion of their gains via borrowing without paying any income tax. The proposal would tax this use of unsold gains as current income, raising about $56 billion in income tax from that source. We estimate that new borrowing over the decade following enactment would generate an additional $46 billion of income taxes on new borrowing, as seen in the table.²⁰

Four important things are worth emphasizing about this revenue projection. First, a majority of the revenue comes from existing borrowing, so a proposal that was just forward-looking would raise far less. Second — and reinforcing the first point — our estimate is more speculative, and possibly more upward-biased for revenue over the next decade, because it is possible to avoid taxes on future behavior, but not past behavior — and ours is a static estimate that does not incorporate behavioral changes.²¹ Third, exactly half of the estimated revenue comes from billionaires, and 90 percent of that is from the richest 400 taxpayers, making the tax sensitive to the behavior of a very small number of people. Fourth, if current tax rates (rather than those in the Biden administration’s budget) are used, the revenue projection is about half that shown in the table below.

Revenue Estimate Breakdown

	Revenue From Existing Borrowing ($ billions)	Revenue From Borrowing Over Next 10 Years ($ billions)	Total Revenue ($ billions)
Net wealth $100 million to $200 million	3.58	2.99	6.58
Net wealth $200 million to $1 billion	24.15	20.18	44.32
Non-top-400 billionaires	2.70	2.26	4.96
Forbes top 400 billionaires	25.09	20.96	46.05
Total	55.52	46.39	101.91
Note: Method is explained in the Appendix.

III. Economic Effects

The proposed tax reform would raise considerable revenue in an equitable and relatively economically efficient way. The billionaire borrowing tax would:

• Eliminate the tax bias favoring borrowing instead of selling appreciated assets and reduce wasteful tax planning. As noted, individuals today have a strong tax incentive to borrow against appreciated assets instead of selling them, even if they’d prefer to sell absent taxes. As The Wall Street Journal observed: “For borrowers, the calculation is clear. . . . The more they can borrow, the longer they can hold appreciating assets. And the longer they hold, the bigger the tax savings.”²² This tax bias has helped to build up a huge business in which the wealth management divisions of large banks lend to wealthy clients against their securities portfolios. We document over $175 billion worth of these loans as of the second quarter of 2023.²³ This figure has grown rapidly over the last decade²⁴ as financial professionals at well-heeled firms pitch borrowing rather than selling because “the tax benefits are stunning.”²⁵ Eliminating this bias for covered taxpayers would eliminate many economically wasteful transactions whose primary purpose is tax avoidance.²⁶

• Reduce the tax bias favoring assets that do not regularly produce realized income. The tax code today strongly encourages owning assets like growth stocks or land, which do not regularly produce realized income, over assets like bonds or stocks, which pay a regular dividend, even if all the assets have the same risk and produce the same pretax return. (We’re ignoring for simplicity the ways these tax advantages are capitalized into asset prices, but accounting for that does not change the upshot.) Part of the reason that taxpayers prefer assets that do not regularly produce realized income is that, under current tax law, taxpayers can access some or all of the unsold gains on their assets by borrowing without triggering income tax. By reducing this bias for covered taxpayers, they will be less likely to choose assets with lower pretax returns simply because of their tax advantages, likely improving economic outcomes.

• Equalize the tax treatment of income used to fund consumption. Ordinary Americans must pay income tax on their wages and can use only what’s left over to fund their consumption. Even when they borrow, they repay that borrowing primarily using wages that have already been subject to income tax. By contrast, the current treatment of borrowing allows wealthy Americans to consume their income without paying taxes on it. Wealthy Americans earn economic income as their assets appreciate. They then fund consumption by borrowing, using that appreciation either explicitly or implicitly to obtain the loan, without triggering any tax. Likewise, they never pay income tax on that appreciation, despite having used it to fund consumption, if they hold their assets until death. The borrowing tax equalizes the taxation of consumed income between rich and ordinary Americans.

• Likely have little effect on incentives for entrepreneurship. Economic theory suggests that the efficiency costs of this proposal in terms of a reduced incentive to save or invest for entrepreneurs will be small because they will only become covered by the tax if they become enormously wealthy through their venture. Entrepreneurs will know that if they do eventually become incredibly wealthy and thus covered by the borrowing tax, they will have a low marginal utility of additional after-tax funds.²⁷ Further, this tax would discourage the practice of mischaracterizing (higher-taxed) labor income as (lower-taxed) capital income by more effectively taxing entrepreneurial capital income for covered individuals.

• Have modest effects on covered taxpayers’ incentives to save or work. This proposal, like other ways of raising revenue by taxing capital income, would likely have some efficiency costs by modestly increasing the effective marginal rate on capital income. Similarly, there could be small effects on work effort, as workers anticipate slightly increased taxes on the returns to saving earnings from working. Any effects, however, are likely to be small because borrowing represents only a small portion of wealth (about 1 to 2 percent).²⁸ Further, the resulting disincentive to save is mitigated by the fact that the alternative to saving — consuming — is also taxed a little more by closing the borrowing loophole.

• Minimize distortion by taxing existing borrowing. Importantly, much of the revenue (55 percent) would be raised from funds that have already been borrowed, reducing any potential distortion, because taxing an action that has already happened cannot directly cause any distortion.

IV. Relation to Other Proposals

The billionaire borrowing tax has upsides and downsides compared with four related proposals: mark-to-market and similar proposals, a wealth tax, taxing gains at death, and a progressive consumption tax.

The proposal is most closely related to proposals to tax unsold gains. For example, a mark-to-market tax would simply tax gains each year as they accrue. President Biden’s billionaire minimum income tax would be a withholding tax on unsold gains that is paid over several years.

Holding the tax rate and wealth threshold fixed, the borrowing tax would raise less money than a mark-to-market tax for the simple reason that hundred-millionaires’ and billionaires’ borrowing makes up only a relatively small fraction of their gains. However, we think that the borrowing tax is more politically palatable than at least a pure mark-to-market tax. In our survey, taxing the borrowing of multimillionaires is 19 percentage points more popular than taxing all unsold gains by those multimillionaires mark-to-market.

The borrowing tax could be viewed as a complement to the minimum income tax for billionaires. For example, the president’s proposed tax rate for the minimum income tax is 25 percent, alongside a tax rate of 44.6 percent on sold capital gains. Our suggestion that the billionaire borrowing tax rate be the full capital gains tax rate would complement the existing tax proposal by taxing the portion of gains that has been turned into cash at a higher rate.

Compared with a wealth tax, the borrowing tax is, in our view, likely to be on firmer constitutional ground. It is also potentially more likely to be enacted as a small fix to the current income tax, rather than a new tax on a new base. Taxing borrowing would also raise fewer liquidity concerns compared with a mark-to-market tax on unrealized gains or a wealth tax because borrowers have received cash before being taxed. Still, it is difficult to compare the borrowing tax to these much more ambitious proposals like a mark-to-market tax and a wealth tax, which would have larger revenue potential (and larger administrative costs).

A different way to deal with the problem of the income tax system’s failure to tax unsold appreciated assets would be to reform the treatment of those assets at death. Eliminating basis step-up, which entirely forgives any income tax that would be due on those assets when an individual dies, would raise more revenue than the current proposal. Doing so would also substantially reduce the incentive for taxpayers covered by the borrowing tax to borrow to avoid taxation since they would have to pay tax on the gains at death.²⁹

An important advantage of the borrowing tax over ending basis step-up, though, is that the government gains the revenue now, rather than waiting for the taxpayer’s death, by which point the government has lost the time value of money or the tax change could have been repealed. The political instability of ending basis step-up reduces the expected revenue and otherwise salutary incentive effects. Further, the politics of the borrowing tax would not be tied up in the politics around the death tax. In any case, the borrowing tax could be adopted alongside the elimination of basis step-up without double taxing that income. Rather, the gains would be taxed upon borrowing, not upon death.

Finally, the borrowing tax is broadly related to arguments for adopting a progressive consumption tax.³⁰ One concern motivating the borrowing tax is that, under the status quo, billionaires can borrow and effectively consume unrealized gains on assets without ever paying tax if the assets are held until death. Similarly, progressive consumption taxes are intended to tax all consumption, regardless of borrowing. Among consumption taxes, the borrowing tax is closest to those implemented using a base of all real and financial flows (R+F).³¹ That cash flow tax would tax cash received and deduct cash that leaves unless spent on consumption. Under that R+F consumption tax, all borrowing would be immediately taxable, while repayment of loan principal would be deductible, without regard to basis in existing assets. There are important differences between the borrowing tax and an R+F consumption tax when it comes to borrowing, such as our allowance for basis recovery.³² Most importantly, the borrowing tax slots more easily into the existing income tax, while instituting a progressive consumption tax with an R+F base would require a series of wholesale changes.

V. Responding to Potential Objections

Critics of the tax could raise at least four objections, that: (1) the tax is unconstitutional, (2) it is bad or even unconstitutional to tax existing borrowing, (3) compliance costs would be high, and (4) covered taxpayers would sell high-basis assets to avoid the tax. We respond to each in turn.

A. Constitutionality

Some might argue that the proposal taxes unrealized income and is thus unconstitutional. However, the tax should be upheld as an excise tax on the activity of borrowing under long-standing precedents. For example, in Flint,³³ the Supreme Court upheld the corporate income tax as an excise on the privilege of doing business as a corporation, even though the amount owed under the tax was based on the income of the corporation. Similarly, here we are proposing an excise on borrowing that sets tax liability in part based on the borrower’s previously unrealized income on assets.

In the alternative, the borrowing tax should be upheld as a constitutional income tax. The Supreme Court has repeatedly said the realization rule is “founded on administrative convenience,”³⁴ and the code taxes apparently unrealized income in a variety of areas, including in partnership and international taxation, as well as the taxation of some financial products.³⁵ Nevertheless, the Court has granted certiorari in Moore³⁶ on whether realization is constitutionally required for economic income to constitute income under the 16th Amendment. If the court finds that the answer is “yes” — an unjustified holding, in our view — our proposal should still pass constitutional muster. In particular, the petitioners in Moore rely on Macomber, a Lochner-era case, which finds that realized income is a gain “received or drawn by the recipient (the taxpayer) for his separate use, benefit and disposal” (emphasis in original).³⁷ When taxpayers borrow against gains on assets and in return receive cash or property for their own “separate use, benefit and disposal,” they have, in our view, realized income derived from those gains for purposes of the 16th Amendment.³⁸

Putting aside realization, the proposal defines covered individuals based on their assets (over $100 million), similar to the president’s billionaire minimum tax. One can argue that by basing a taxpayer’s income tax liability on their wealth or assets, the proposal is no longer a tax on income, but instead one on wealth, and thus plausibly unconstitutional. We disagree with this argument, but the proposal could easily be enacted with a savings clause that uses income, possibly including unrealized gains, as the threshold for whether individuals will be covered, rather than assets.

B. Retroactivity

Some may find it problematic (despite the efficiency benefits) to tax existing borrowing because it accumulated under a different regime. However, tax legislation frequently changes taxpayers’ liability today based on their past actions. Take two examples. First, whenever a law raises the capital gains rate, it increases the tax on gains accrued under the prior regime, but not yet realized. Second, and more directly analogous to our proposal to tax existing borrowing, is the mandatory repatriation tax in the Tax Cuts and Jobs Act of 2017. That provision was a one-time tax on income accrued in years past by foreign subsidiaries of U.S. companies, which previous law had not recognized as realized by the parent until repatriated. In both examples, as well as in our proposal, taxpayers have chosen to defer paying taxes — by not selling assets, by not repatriating earnings, or by borrowing instead of selling appreciated assets. If taxpayers choose to defer, it is reasonable for the government to choose to change how it taxes those activities.

For these reasons and more, the tax should also not cause constitutional problems as impermissibly retroactive under the Fifth Amendment. The Supreme Court has “repeatedly upheld retroactive tax legislation against a due process challenge.”³⁹ Likewise, the mandatory repatriation tax was held not to violate the Fifth Amendment by the Ninth Circuit.⁴⁰

C. Compliance Costs

The proposal would impose compliance and administrative costs. We expect these costs to be manageable in part because of the small number of taxpayers involved — there are only about 35,000 covered households, and fewer than 14,000 with over $1 million in existing debt.⁴¹ Further, the tax code has the apparatus for doing these calculations because of the estate tax. These methods will likely improve with time, as well — for example, with the use of big data techniques to estimate harder-to-value assets.

Available estimates suggest that the compliance costs would be modest. Many of the compliance cost issues in the proposal overlap with those that would occur under a wealth tax or a periodic tax on unrealized gains because both require a valuation of taxpayers’ major assets. Estimates of the cost of that valuation vary widely. Greg Leiserson estimates compliance costs, using data from the U.S. estate tax for estates worth over $20 million, as 0.2 percent of wealth subject to tax.⁴² Under the proposal (using FIFO), compliance costs for taxing the existing borrowing of the roughly 13,000 affected households would be quite modest, at about $500 million.⁴³ That part of the tax would raise about $55.7 billion, with compliance costs thus being about 0.8 percent of revenue, which is likely an overestimate for two reasons: First, because of our limitation on easier-to-value major assets; and second, because our proposal will encourage less gaming than under the estate tax. In particular, undervaluing for the estate tax reduces taxes by the full tax rate, whereas doing so for the borrowing tax just means that more assets (with a slightly higher basis/value ratio) are drawn into the tax, yielding significantly lower gains.⁴⁴

That ratio of compliance costs to revenue should stay roughly steady in subsequent years because the amount of assets that need to be valued will continue to equal the amount of borrowing — and the assets will continue to have to count as major assets.⁴⁵

D. Sale of High-Basis Assets

Taxpayers may react to the proposed tax in part by selling assets that under current law they would have borrowed against to enjoy the tax benefits. That is a feature, not a bug, of this proposal. Still, to the extent that taxpayers selectively sell assets with a high basis (that is, with little unrealized gains built in or with built-in losses)⁴⁶ that will reduce the revenue generated by the proposal. This will likely happen to some degree. Still, there are a variety of real-world factors that limit how strong this effect is likely to be. This incentive already exists for sales of assets, and yet empirically taxpayers frequently sell publicly traded assets in which there is a large built-in gain even when they have one with a large built-in loss, perhaps partly because they feel better selling winners.⁴⁷ In addition, higher-basis assets may be things like houses, which the taxpayer is disinclined to sell; those assets might be recently invested and subject to lock-ups; or the taxpayer may simply believe the higher-basis asset will outperform the low-basis asset.

VI. Design Features

We have provided a baseline proposal for taxing borrowing. However, the tax could potentially be varied along various dimensions: (1) basis calculation; (2) overall design, including taxing only secured borrowing as a realization, or taxing all borrowing by covered households as gross income; (3) treatment of related entities; (4) addressing other avoidance issues; and (5) thresholds.

A. Basis Calculation

While in our main proposal we suggest that basis be calculated using FIFO, there are other ways to calculate basis. Instead, the taxpayer’s average basis could be used. This would likely raise less revenue than using FIFO because the taxpayer’s average asset likely has a higher basis than their oldest assets. Average basis might well be thought of as fairer, given that recourse borrowing is implicitly backed by all the taxpayer’s assets. The downside to using average basis is that it likely has substantially higher compliance costs because it requires the valuation of all the taxpayer’s assets, rather than just an amount equal to the borrowing (approximately 2 percent of wealth on average for existing borrowing). Under the assumptions used above, the compliance costs are roughly 50 times greater for using average basis than FIFO.

Another possibility would attempt to roughly replicate the results of using average basis while avoiding valuing all the taxpayer’s major assets. For example, instead of using FIFO (which starts with the taxpayer’s oldest assets) or last-in, first-out (which starts with their newest assets), the tax could start with their assets that are in the middle in age — so middle-in, first-out. When a taxpayer triggers the tax for the first time, she could calculate the age of her oldest major asset — say, 30 years. This would lead to an assumed average asset age of 15 years. She would then calculate her basis under the borrowing tax using the basis of her assets that are 15 years old, and if those are not sufficient, then use her assets that are next closest in age (so 14 and 16 years old, etc.), working toward her newest and oldest assets.

As noted, we think that a system that just looked at the basis of the specifically secured assets would not be workable because of the incentive to secure a loan with assets with the highest basis and (depending on design) the potential incentive to not use any specific asset to secure loans at all.

B. Alternative Overall Designs

The billionaire borrowing tax is a middle path between two other versions of taxing borrowing. One more minimal version, as just discussed, would be to tax only secured borrowing. We think this is not workable for the reasons noted above. Another version, which is more expansive than our proposal, would be to treat all borrowing as gross income and all repayment as deductible. In terms of direct tax application, we instead propose changing the tax treatment of the underlying assets (and their deemed realization) rather than the borrowing itself. Importantly, there is basis offset with our proposal. This has the benefit of following the underlying logic offered for the tax, in which the taxpayer has not paid tax except to the extent of her basis. Moreover, our proposal is easier to implement since it would not require complicated deductions upon loan repayment.⁴⁸

C. Treatment of Related Entities

The tax would include antiavoidance provisions. Especially for business borrowing, related entities may be involved.⁴⁹ For example, suppose that Ellison formed Ellison Partnership, contributed assets, borrowed based on those assets, and then distributed cash to Ellison. For passthroughs, the tax treatment (including borrowing) flows through to individuals, and we would not disrupt that. We would distinguish between passthroughs created to avoid taxes on borrowing from passthroughs borrowing in the normal course of non-financial-portfolio business by imposing an economic substance test on the partnership and the borrowing.⁵⁰ Therefore, borrowing by passthroughs owned by covered households would trigger the tax for borrowing that is not in the ordinary course of a partnership’s business, for reasons other than avoiding the borrowing tax. (This would leave open the possibility that borrowing to purchase a company in which one will be an active manager would not trigger the tax.) Personal expenses paid for by the passthrough would be covered under existing hobby loss rules. C corporations should not be a viable avoidance technique because when corporations pay the borrowed funds to the shareholder, that will usually trigger taxes.⁵¹

D. Addressing Avoidance: Leasing, Derivatives

One potential avoidance or evasion opportunity is for covered households to switch to leasing instead of borrowing. For example, instead of purchasing his Hawaiian island, Ellison could have gotten a 99-year lease on it on terms that resembled him paying the “lessor” what he would have paid to the bank had he borrowed against his stock to fund the purchase. This is a familiar problem in tax. In some cases in which a lease is in practice a seller lending the purchaser the funds to purchase the property, the courts have characterized these transactions according to their economic substance. And doing so here would trigger the borrowing tax.⁵² This is not a panacea, however, as even true leases share many economic properties with borrowing to purchase the asset. We could combat this avoidance by characterizing leases of more than five years by covered taxpayers as borrowing.

An additional goal of borrowing by covered taxpayers might be to diversify their portfolios. Ellison might do that by borrowing against his Oracle stock and buying shares in, say, an S&P 500 index fund. Instead of borrowing, to avoid the borrowing tax, he could just buy options on the S&P 500 (which are implicitly leveraged). There are, however, widely agreed-on methods for measuring the leverage implicit in market-traded options, and we could calculate the leverage implicit in the option (for example, using the Black-Scholes model) and tax that as borrowing.

E. Thresholds

Our proposal has a variety of thresholds for being subject to the tax: a wealth threshold, a total borrowing threshold, and an annual borrowing threshold. While we offer these as starting points, any of these could be changed. Lowering the thresholds would raise more money and, in some cases, would affect progressivity.

VII. Conclusion

We face high inequality and strong revenue needs in the United States. Some object to taxing the wealthy unless they use their gains. But many of the wealthiest do use their gains through the borrowing loophole — they get rich, borrow against those gains, consume the borrowing, and do not pay any tax. This is unfair. Closing the borrowing loophole on the superrich would raise over $100 billion in an administrable, highly progressive, and reasonably efficient way.

Appendix: Details of the Score

We estimate revenues by dividing taxpayers into three groups:

• those with net wealth greater than $100 million and less than or equal to $1 billion;

• those who are included in the Forbes 400⁵³ (wealthiest 400 Americans); and

• billionaires who are not included in the Forbes 400.

We estimate revenue for each group as:

We must rely on different computations and different data sources to compute total borrowing and the ratio of unrealized capital gains to total wealth for the three groups. Thus, we calculate revenues for each group separately.

We first describe how we estimate existing borrowing and the other parts of the equation above for the three groups. Second, because our data come from different years and the tax would not be enacted immediately, we describe how we update estimates of borrowing to correspond to April 2024. Third, we turn to the taxes on new borrowing over the subsequent 10 years.

Tax on Existing Stock of Borrowing

Existing borrowing of taxpayers with net wealth greater than $100 million and less than or equal to $1 billion.

To estimate revenues from taxpayers with net assets greater than $100 million and less than or equal to $1 billion, we use the 2019 Survey of Consumer Finances (SCF) summary public extract. This group is defined using the SCF variable for net worth, which is computed as the difference between total assets and total debt. We first estimate total assets, total borrowing, and total unrealized capital gains using the SCF variables asset (total value of assets), debt (total value of debt held), and kgtotal (total unrealized capital gains or losses), as well as SCF-provided sample weights. For this group, total borrowing refers to borrowing that incorporates the phase-in for households with net wealth between $100 million and $200 million and incorporates the $1 million exemption. These values are in 2019 dollars. The SCF interviews were largely conducted between May and December 2019. When aging the SCF estimates with other data sources to the revenue window, as described below, we use April 2019 as our reference date for the SCF data.⁵⁴

Existing borrowing of the Forbes 400.

The SCF excludes families in the Forbes 400 and other families whose finances are deemed too unique for public data disclosure.⁵⁵ Therefore, we estimate revenue from this group using publicly available data from Forbes computed in November 2021. Forbes provides a table of 32 billionaires in the Forbes 400 who had pledged public stock of companies listed on the New York or Nasdaq stock exchanges.⁵⁶ This table displays each billionaire’s value of pledged shares.

We use information on (unrealized) capital gains as a share of wealth from Emmanuel Saez, Danny Yagan, and Gabriel Zucman.⁵⁷ They estimate capital gains as a share of wealth using the 2019 SCF combined with the 2020 Forbes billionaires. We assume that capital gains as a share of wealth for the Forbes 400 are equivalent to their estimate for households with net wealth above $100 million and that this ratio is constant over time. We also assume that:

We base this on reporting in The Wall Street Journal that “banks typically will lend a borrower at least 50 percent of a diversified portfolio’s value.”⁵⁸ Since billionaires typically do not have diversified portfolios, perhaps even for the portion of their wealth against which they are borrowing, we lower this ratio from 1:2 to 1:3.

Taxes on existing borrowing for those with primarily public stock are then:

We only have data on pledged assets from the Forbes 400 for publicly traded pledged assets. This neglects Forbes 400 individuals whose wealth is primarily concentrated in nonpublic stock. Therefore, we estimate taxes on existing borrowing separately for those with primarily public stock and primarily nonpublic stock. We assume that the tax revenue from borrowing by those with primarily nonpublic stock or other non-traded assets is equal to the taxes from public stock times the ratio of privately held unrealized gains to publicly traded unrealized capital gains. Following Saez, Yagan, and Zucman,⁵⁹ we assume that the unrealized gains of Forbes billionaires have the following composition: 60 percent publicly traded stock, 35 percent private stock, and 5 percent real estate. So, this ratio is 35:60.

Existing borrowing of billionaires not in the Forbes 400.

The 2019 SCF produced an unreasonably low estimate of the number of households that are billionaires.⁶⁰ Thus, following Saez, Yagan, and Zucman,⁶¹ we disregard the SCF billionaires and use the Forbes billionaires list instead. Our data on the Forbes billionaires contains the name of each American billionaire on August 17, 2021, and their net worth on that date.⁶² We use this data to estimate the total wealth of the non-top-400 billionaires in November 2021.⁶³ We assume that average borrowing as a share of wealth for the non-top-400 billionaires is the midpoint between borrowing as a share of wealth for the Forbes 400 and borrowing as a share of wealth for tax households with a net worth of $200 million to $1 billion.⁶⁴ Like the calculations for the Forbes 400, we assume that capital gains as a share of wealth for the non-top-400 billionaires are equivalent to the estimate by Saez, Yagan, and Zucman for households with net worth above $100 million and that this ratio is constant over time.⁶⁵

Aging Data to 2024

To convert taxes on existing borrowing estimated in Section II into taxes on existing borrowing as of April 2024, we multiply those sums by an aging factor.

The aging factor is computed in two steps. First, we assume that total borrowing, total assets, and total unrealized capital gains for each group grew at the same rate as total household wealth until April 2023; this comes from the Federal Reserve’s Financial Accounts of the United States.⁶⁶ The SCF data are aged from April 2019 to April 2023. Data from Forbes is aged from November 2021 to April 2023. We do not have data on total household wealth from November 2021; therefore, we approximate November 2021 using total household wealth from October 2021 for the Forbes 400 and non-top-400 billionaire groups.⁶⁷ Then, to age the data from April 2023 to April 2024, we assume that total borrowing, total assets, and total unrealized gains will grow at the same rate as the historic annual growth rate of total wealth for the top 10 percent of households. This comes from analysis done by the Congressional Budget Office⁶⁸ using data ultimately from 1989-2019.⁶⁹

Taxes on Growth in Borrowing Over 10 Years

We assume that total assets, total borrowing, and total unrealized capital gains will grow according to the historic annual household wealth growth rate.⁷⁰ Cumulative taxes over 10 years are then:

taxes on existing borrowing * (historic annual wealth growth rate)¹⁰