Each edition of Tax Matters consists of free-flowing responses by three tax practitioners to a question regarding a current issue in tax law and policy. Tax Matters commentaries provide insightful perspectives on a broad range of topics, making important contributions to the dialogue within the tax bar about cutting-edge issues. Although the commentaries are certainly of interest to the academic community, they are primarily directed toward tax professionals and their clients.
Section 871(m) and Delta: When Should a Dividend Equivalent Be Treated like a Dividend?
If certain derivative exposures to U.S. stocks present an inappropriate opportunity to avoid U.S. withholding taxes with respect to “dividend equivalents,” what is the nature of those inappropriate derivatives? Is “delta” a useful tool for identifying them?
To begin, is it appropriate to “buy” a known dividend amount without withholding tax? You must pay for that dividend amount. That is, with or without withholding tax, nobody will give you the economics of a $1 future dividend for less than $1 (time value and tax effects aside). So have you “avoided” anything by taking “new” derivative exposure to known – or even highly expected – dividends? In this circumstance, if you suffer a withholding tax, you will lose money (unless you can credit the tax elsewhere); but avoiding the tax will not help you make any money – it’s a zero-sum game (minustransaction costs).
On this logic, you could avoid U.S. withholding tax only if you were already exposed to the underlying stock, and you could – cheaply – avoid an impending tax by switching the form of your exposure to one with no withholding tax. From this perspective, Congress’s focus on “crossing into and out of” total return swaps makes sense.
But the purpose of the dividend withholding tax is to make a $1 dividend worth less than$1 to a non-U.S. person (ignoring credits). Which means even if you don’t make any money via the avoidance, avoiding the tax avoids the intended economic loss that the withholding tax represents.
So perhaps, instead, avoidance occurs when the “short” party is the “tax owner” of, but is not itself effectively subjected to tax on, a dividend. For example, a securities dealer might hold stock, “swap” that stock to a non-U.S. person, mark both positions to market and be “flat” (tax-wise) except for the spread built into its pricing. In that event, if the dividend equivalents under the swap go untaxed, the dividends are effectively never taxed.
So asking if the counterparty holds the underlying stock makes sense as a policy matter. It also ensures we don’t treat as a dividend something that has no actual connection to the underlying stock. (Incidentally, it would eliminate issuer-issued instruments – convertible debt and warrants – from Section 871(m), as they should be. We already have rules that apply to those, which treat all investors consistently – as well as the issuer; Section 871(m) does not affect E&P.) But asking if the counterparty owns the underlying probably isn’t practical, especially considering that the counterparty might have lent the underlying to someone – which should change the result, policy-wise, because that stock presumably ends up in the hands of a “normal” taxpayer (after being sold short by its borrower), resulting in “proper” taxation of the relevant dividend.
Nonetheless, the “logic” of delta seems consistent with the goal of taxing the amount of dividends that the counterparty can be expected to “own” as a U.S. tax matter (a dealer in securities will generally hedge a short exposure by being long the delta at any given point). Why then tax only if delta was high when the long party entered into the transaction? Perhaps for the same reason the proposed regulation exempts indices (as defined): There are typically feasible alternative hedges for the counterparty when the delta is low (e.g., listed options) or when the underlier is an index (e.g., futures contracts, or listed options), so perhaps the thought is that it isn’t appropriate to assume that the counterparty will own any underlying stock(s). A more directly probative question than delta might be whether the counterparty could cheaply hedge otherwise than by holding (a substantially fixed amount of) the underlying.
One theme that threads through this is “cheapness.” With few exceptions, transaction costs for equity swaps are extremely low. There is typically little or no credit risk (at least from the counterparty’s perspective; the long party typically posts adequate collateral), and these transactions are done on standardized documentation, so there’s essentially nothing to negotiate, and no basis to charge meaningful fees. And swaps can be priced to allow the long party seamless transition into and out of existing positions, so no “gap risk” (another form of transaction cost) is created. “Equity finance” is done on razor-thin margins. Indeed, many end users can buy and sell swaps electronically, as they can the underlying.
As delta gets lower than one, the derivative gets more expensive. True options cost money, because the counterparty is taking meaningful risk, for which it must be compensated. The more option-like the exposure, the higher the transaction costs (“risk premia”). If transaction costs exceed the value of the avoided tax, the transaction ceases to be explainable as tax avoidance. And the further out in time the potential (but increasingly uncertain, as dividend policies change) benefit, the less rational it will be to pay meaningful transaction costs for the prospect. So, for example, nobody will use either “structured notes” (generally, prepaid forwards with often complex economics) or combinations of low-delta positions (with unrelated, non-accommodating counterparties) to avoid U.S. withholding tax; it’s just too expensive.
Delta might be a proxy for whether the counterparty should be expected to own the underlying (potentially leaving dividends effectively untaxed), or whether expected avoidance value might be expected to exceed transaction costs. But it is in both cases a very rough proxy. And delta is an unwieldy and imprecise concept (indeed, one that simply can’t be applied coherently in not uncommon fact patterns, e.g., digital options), making its functionality debatable in any event.
One approach might be to establish a presumption that an instrument or group of instruments is subject to Section 871(m) if it reflects or includes on a net basis a high-delta exposure to dividend-paying U.S. stock, a presumption which could be rebutted by showing either material transaction costs or that the counterparty does not own the underlying in connection with the transaction on a given dividend date.
Taxation Without Authorization: The Proposed “Dividend Equivalent” Withholding Regulations Under Section 871(m)
The currently proposed regulations under section 871(m) of the Internal Revenue Code threaten to impose “dividend withholding” on a broad range of swaps, options, forward contracts, futures contracts, debt, and other financial contracts that reference dividend-paying U.S. stock, even if the contracts do not in fact reference dividends, and even if no party to the contracts ever owned the stock. The proposed regulations create a new withholding regime that is overly broad, lacks clarity, will be difficult to comply with, and may violate the “nondiscrimination” provisions contained in many tax treaties to which the United States is a party. There is no statutory authorization for this new withholding regime.
Section 871(m) was enacted to curtail a specific type of tax-avoidance transaction. In the early 2000s, several banks touted “yield enhancement strategies” that allowed their non-U.S. clients to use derivatives to eliminate or reduce dividend withholding. According to a 2008 report by the Senate’s Permanent Subcommittee on Investigations, under these strategies, a non-U.S. client typically transferred U.S. stock to the bank shortly before the stock’s ex-dividend date, received a substitute dividend free of withholding, and reacquired the stock shortly after the dividend payment. This report catalyzed section 871(m).
Section 871(m) imposes withholding on payments made under “specified” swaps that are contingent upon, or determined by reference to, U.S.-source dividends. Consistent with the 2008 Senate report’s description of perceived tax abuses, the statute defines a “specified” swap to include only swaps where the short party is effectively required to acquire the underlying U.S. stock in connection with the transaction. In these situations, section 871(m) recasts the swap as an acquisition of the underlying stock by the short party as an agent of the non-U.S. investor, and treats any substitute dividend paid by the short party as an actual dividend subject to U.S. withholding tax.
The “quasi-agency” test codified in section 871(m) extends dividend withholding to a limited group of transactions that bear indicia of actual stock ownership through an agent. By contrast, the proposed regulations create a brand-new withholding regime. Under the proposed regulations, an instrument that references dividend-paying U.S. stock is subject to withholding if the ratio of the change in the instrument’s value to the change in the stock’s value—referred to as “delta”—is at least 0.70.
The proposed regulations extend the statute by imposing dividend withholding on instruments that do not have payments that are contingent upon actual dividend payments. For example, a “price return only” contract that exposes a non-U.S. investor to IBM’s upside and downside, but not its dividends, is subject to withholding under the proposed regulations, even if the short party never owned any IBM stock in connection with the transaction, and even though the non-U.S. investor is not entitled to receive more or less under the contract if IBM’s dividends are greater or less than expected. The “dividend equivalent payment” that is subject to withholding is the difference between the contract’s actual price and the (presumably higher) price that the investor would have paid for a contract that exposed her to IBM’s dividends as well.
The United States has never before imposed tax on an investor’s foregone costs in choosing one investment over another. It is unimaginable that Congress intended the statutory language of section 871(m) to introduce such a tax. Section 871(m) imposes a tax on “payments” that are contingent upon, or determined by reference to, U.S.-source dividends. There is no suggestion in the statute, or in the corresponding legislative history, that Congress was leaving the word “payment” to the Treasury Department’s interpretation.
Moreover, the proposed regulations’ broad definition of “payment” violates the nondiscrimination provisions of many income tax treaties to which the United States is a party. These provisions generally prohibit the United States from imposing a more burdensome tax on residents of the other signatory state than it imposes on U.S. residents, and U.S. residents are not taxed on their foregone costs in choosing to invest in a “price return only” contract over a “total return” contract.
By adopting a test that would impose withholding on a broad spectrum of transactions that bear little to no resemblance to the “yield enhancement strategies” of the early 2000s, the proposed regulations ignore the 2008 Senate report’s admonition not to condemn transactions with “legitimate business purposes such as facilitating capital flows, reducing capital needs, and spreading risk.” Ironically, the delta test also fails to impose withholding on some transactions that would have been subject to withholding under the quasi-agency test, even though the statute contemplates withholding on these transactions.
Section 871(m) permits the Treasury Department to impose withholding on dividend equivalent payments under swaps that “have the potential for tax avoidance,” and on “substantially similar payments.” The proposed regulations read too much into this language. If Congress had wanted its quasi-agency test to be replaced by a far-reaching delta test that contradicts the 2008 Senate report and flouts the traditional definition of “payment,” it would have said so. The proposed regulations amount to a new and unauthorized withholding regime.
 Staff, U.S. Senate Permanent Subcommittee on Investigations, “Dividend Tax Abuse: How Offshore Entities Dodge Taxes on U.S. Stock Dividends” (Sep. 11, 2008).
The Most Recent Proposed Regulations Under Section 871(m): The Perfect Is the Enemy of the Good Andrew Walker
While the recent proposed regulations under section 871(m) raise many technical issues, three features of the proposed regime are perhaps most controversial: (1) The use of a “delta” threshold of 0.7 to identify those derivatives that should be subject to section 871(m), (2) requiring that delta threshold be tested whenever a party acquires a long position not merely at original issuance and (3) requiring withholding on estimated dividend amounts (i.e., even if the derivative does not provide for payments linked directly or indirectly to actual dividend payments). These three features of the regulation, particularly in combination, may cause the regulations to capture a broader class of transactions than Congress perhaps intended.
The prior proposed regulations focused on identifying “abusive” indicia common to transactions used to avoid dividend withholding. The new proposed regulations largely abandon the attempt to identify factual indicia common to abusive transactions (like “crossing in” or out, where the stock is transferred by an actual owner to the short derivative counterparty pre-dividend and reacquired thereafter). Instead, they effectively treat withholding as appropriate where the economic effect of a derivative transaction is reasonably equivalent to ownership of the underlying stock. The proposed regulations adopt a test based on “delta,” which generally reflects the correlation between changes in the value of the derivative and changes in the value of the underlying U.S. stock.
Using this “bright line” test to identify positions that economically resemble stock ownership has the advantage of being more objective than a test based on factual indicia of abuse and therefore is potentially more easily administered. Derivatives dealers typically compute delta to establish an appropriate hedge of their positions and therefore should be in a position to provide this information to counterparties. To prevent easy avoidance of the rules, it is sensible that positions economically very similar to stock ownership should be caught even if the correlation is less than perfect (i.e., delta is less than 1). Nevertheless, although selecting an appropriate bright line threshold is necessarily somewhat arbitrary, most market participants view a derivative with a 0.7 delta as substantially different economically from a “delta one” transaction which is closely equivalent economically to stock ownership.
The delta test is difficult to apply to derivatives with more complex economics than a simple total return swap referencing an equity security. It may be unclear how many shares of stock should be considered in the denominator of the delta test (and this may be determinative in some cases). For example, assume the swap requires the counterparty to pay any depreciation in the value of 100 shares of reference stock in exchange for the appreciation in 200 shares of stock in excess of a threshold that is substantially higher than the current share price. Should changes in the value of the derivative be compared to changes in value of 100 shares or changes in the value of 200 shares? Presumably the position should be disaggregated in some way to test delta but the regulations do not provide guidance on how to do this.
More fundamentally, delta is an imperfect proxy for dividend equivalence (which is the focus of the statutory provision). A position may be closely correlated with changes in the stock’s value but relatively insensitive to changes in the actual dividend return, or vice versa. Thus, the test may be too broad in some cases but too narrow in others. For example, a taxpayer could in theory enter into a combination of positions that effectively pays it amounts linked to the actual dividend return on a specified number of a U.S. company’s shares for some period but does not provide for any returns related to changes in the stock’s value. The delta of such a position could be lower than 0.7 even though the transaction effectively passes through the dividend. Because delta is an imperfect proxy, the regulation drafters therefore need to include “anti-abuse” rules that can override the objective delta test, although this may undermine the advantages of having a bright line rule.
The proposed regulations compound these difficulties by testing the delta not only at issuance but whenever a party acquires its long position. This heightens the possibility that that the rules will catch derivatives that were not entered into to avoid dividend withholding. For example, a typical convertible bond typically would be issued with a strike price that is out of the money and would not meet the delta test at issuance. However, if the stock later has substantially appreciated above the strike price, the delta may later exceed 0.7 and a secondary market purchaser after that date may have acquired an instrument subject to 871(m) withholding. Further, dealers may not recalculate delta with sufficient regularity to provide the necessary information and, in cases where the derivative takes the form of a traded instrument (like a structured note), the dealer may not even be aware the derivative instrument has changed hands. More generally, applying section 871(m) withholding to instruments that reference the issuer’s own equity like convertible bonds may conflict with other provisions of the Code, like section 305, which may also impute dividends on such instruments in certain (sometimes inconsistent) circumstances.
A third controversial aspect of the new proposed regulations is to adopt withholding on estimated dividends. The new proposed regulations expressly treat an instrument as subject to withholding if it meets the delta test even if it provides for no actual payments linked to dividends (on the theory that expected dividends must necessarily have been priced into the contract). It is questionable whether such an adjustment in the pricing based on an anticipated dividend (which may not in fact correspond to any actual dividend adjustment) is in any sense the payment of a dividend equivalent amount for income tax purposes.
The regulatory drafters deserve sympathy, however. Coming up with administrable rules that do not impede at least some transactions that are not motivated by avoidance of dividend withholding is an impossible challenge. However, changes to the three features of the current proposed regulations above– for example, increasing the delta threshold to 0.8, and limiting substantially situations in which delta is tested post-issuance and withholding is applied to estimated dividends– will probably be necessary to make the rules workable.