March 3 NYU Tax Policy Colloquium (with Adam Rosenzweig)

Last Thursday at the colloquium, Adam Rosenzweig presented his paper, Thinking Outside the (Tax) Treaty. The paper, which is still at an early stage, combines a fairly broadly pitched discussion of the need for international tax policy analysis to make greater use of game theory (which is claimed to have large potential payoffs) with a relatively narrow proposal.

With regard to the broader use of game theory, clearly it's true that countries are interacting strategically over time. While in principle there's repeat play and most of the action takes place in public, limited responsiveness for internal political reasons may in some cases create the equivalent of countries, as in the classic prisoner's dilemma, acting to a degree unilaterally.

But in some of the settings that the paper emphasizes, I think a simple Coase set-up also helps one to think clearly. Suppose the U.S. is very unhappy about taxpayers' use of the Cayman Islands to avoid U.S. tax liability. If the issue is information reporting, we actually need cooperation from the Caymans to address the problem. But in many cases (e.g., use of Caymans corporations to avoid U.S. residence, move taxable income to a zero-rate jurisdiction, etc.), we actually could just change our own law, to impose the taxes we ostensibly want, without needing the Caymans to do anything. But despite this, suppose we actually need their cooperation, whether it's information reporting or not.

Suppose further that U.S. companies are taking advantage of Caymans opportunities to avoid $1 billion in U.S. taxes, but are paying the Caymans (through incorporation fees, amounts paid to local lawyers, etc.) only $1 million. The size of the disparity between what we're losing and what they're getting seems to suggest that the U.S. and the Caymans could in principle make a deal leaving both better off.

Let's say the status quo's benefit to the Caymans is $1 million (though presumably it's actually less since they have to provide at least a modicum of services to get this money). But how much is the lost $1 billion of tax revenue hurting us? The answer is less than $1 billion, if those are taxes that U.S. individuals ultimately should have paid, since at least "we" (i.e., all Americans) still have the money one way or the other. Rather, in principle the social welfare cost to the U.S. of failing to collect this $1 billion in tax depends on the difference between the "marginal efficiency and equity cost of funds" (MEECF) from two alternative revenue sources: (a) this $1 billion if we could have gotten it, versus (b) the MEECF of getting the funds by whatever next-best means we are forced to resort to instead.

To make MEECF more intuitive, let's ask why we would have preferred to get the revenue this way rather than some alternative way. Perhaps, if we fail to get the billion dollars this way, there is in effect a tax subsidy for using multinational operations purely as an income-shifting device, rather than due to its non-US tax advantages. Perhaps U.S. businesses have a tax incentive to engage in otherwise unprofitable multinational operations and to do some costly paper-shuffling once they are so engaged. And suppose these machinations undermine our ability to use the entity-level corporate tax to backstop the individual income tax in assuring that high-income individuals pay current tax on their business profits when earned through corporate entities.

In principle, this means we would reasonably have paid $X to get this billion dollars of revenue, rather than having to do something else. Suppose that X here equals $10 million. The Caymans income-shifting is therefore benefiting the Caymans by only $1 million, and yet costing us (in welfare terms) $10 million.

This implies that a Pareto-improving deal - so far as we and the Caymans are concerned - is in principle possible. We simply pay them between $1 million and $10 million not to accommodate the income-shifting, and both sides are better off. And if we get them to recede by threatening them rather than compensating them, then in Coasean bargain terms the outcome is the same, only we've done better and they've done worse distributionally.

If one accepts this setup, then the fact that the deal doesn't take place, presumably reflecting transaction costs (e.g., the fact that, if we pay off the Caymans, U.S. taxpayers might simply park their income somewhere else instead) evidences global inefficiency. Game theory comes into the picture once one starts thinking about why the deals are hard to make with multiple players. But in the context of the paper, it wasn't clear at this point how much of an intellectual payoff one gets by explicitly relying more on game theory to model the problem and possible solutions.

The paper's main suggestion, which it analogizes to an underlying economic literature about using lotteries to pay for public goods, is that countries in a non-treaty relationship (like the U.S. and either the Caymans or Brazil) make greater use of one-off arbitration deals to settle disputes where they think they would benefit from cooperating, without having to commit to broader and more systemic cooperation that they do not consider to be in their mutual interest. It will be interesting to see how later drafts of the article develop this potentially useful idea, although tax haven countries would have to think about whether cooperating in this manner, albeit just sporadically, would undermine their credibility with the suppliers of inbound (actual or paper) capital.