In defense of tax shelters.

I. INTRODUCTION

Tax shelters do not have many defenders. The lawyers who have to defend them on behalf of a client tend to simply appeal to the court's duty not to engage in law-making. Often that will win the case, but it does not help with the policy question that centers on what the law-maker, rather than the court, ought to do. Ought he, for instance, to go after tax shelters with expansive anti-abuse rules? Ought he to treat all exclusively tax-motivated transactions as the exploitation of a loophole that should be closed, where practicable--if not through anti-abuse rules, then by more detailed legislation specifically banning the offending transaction? The prevailing answers to both questions I take to be "yes" and "yes." In this article I will argue for "no" and "no."

I may be overstating things a little, in part because I am no tax scholar and my reading of the literature is very limited. (1) However, the most extended defense of aggressive tax planning I have found (I take tax shelters to be synonymous with aggressive tax planning) is the following tepid expression of support in a well-regarded (indeed, to my mind superb) book on financial planning:

Although the deadweight costs associated with time spent in tax

planning may seem socially wasteful, the relevant question is how much

waste would exist using alternative means to achieve the same social

goals. In other words, how does the net benefit of the altered

economic activity brought about by the tax system compare with the net

benefits of the next best alternative? Obviously, if we could

implement social policy through a mechanism that would result in zero

waste, we would do so, but such a goal is not realistic. (2)

Not exactly a ringing endorsement of zealous tax lawyering, nor something that will make the architect of ambitious tax shelters feel terribly good about the enterprise.

In this article I hope to do better than that, even though this may seem like an immodest objective for someone without any expertise in tax law. What I do however have is a broader-than-usual perspective on tax shelter-like activities in other, non-tax domains, consideration of which I think serves to make tax shelters appear in a somewhat different light. Let me give a small example. Suppose we tried to transpose the tax scholar's strong desire to reduce deadweight loss, which he thinks should be the main underpinning of our tax policies, into the criminal law. We might then be led to focus on the fact that a great deal of what private citizens do to avoid becoming victims of crime does not actually serve to reduce crime but simply diverts it onto other, easier prey. That is a deadweight loss. And yet, does it seem at all appealing to deal with this deadweight loss by prohibiting people from taking precautions that seem to have that effect--by banning, for instance, the sale of sturdy doors and heavy-duty locks in the same way that we ban the sale of certain kinds of toilet flushes deemed too wasteful of water? If it does not seem appealing, should that not give us pause about analogous attempts to reduce deadweight loss through tax shelter regulation? Much here will turn on how significantly different the two realms, tax law and criminal law, really are. But it seems like an analogy worth thinking about in trying to figure out to what extent we should tolerate tax shelters.

But it is not the analogy to the criminal law that I intend to pursue in this article. It is a different, to my mind more interesting analogy that I intend to explore, the analogy to voting--or more precisely, the analogy between tax shelters and agenda manipulation. Here is how I plan to proceed. In Part II, I will offer a few useful reminders about the nature and source of agenda manipulation. In Parts III and IV, I will transplant some of the strongest arguments against tax shelters into the domain of voting and see what they look like from that new vantage point. I am hoping that the reader will come away from this exercise with some serious doubts about those traditional arguments. Finally in Part V, I will make the argument that there is no meaningful difference between the two domains and that if it is hard to quarrel with agenda manipulation--as will turn out to be the case--it is equally hard to quarrel with aggressive tax planning.

My source for the arguments against tax shelters--the arguments that I intend to transplant into the domain of voting--is an article by David Weisbach, Ten Truths About Tax Shelters, (3) which strikes me as a kind of manifesto, a brilliant polemic that expresses in especially forceful terms the case against tax shelters.

II. SOME REMINDERS ABOUT THE NATURE AND SOURCES OF AGENDA MANIPULATION

Voting we all know is subject to manipulation. There are different forms of manipulation. The one most familiar to lawyers and most relevant for this article is agenda manipulation. The possibility for agenda manipulation is a direct implication of Arrow's famous theorem on which the theory of social choice is founded. There is, mercifully, no need to lay out the full argument for, and meaning of, Arrow's theorem, but a slight review and adaptation of the argument behind it will be helpful to understanding what it is about voting that makes agenda manipulation so inevitable.

The core message of Arrow's theorem I take to be that there is an irresolvable tension between two very basic substantive desiderata of a voting mechanism: On the one hand, the Pareto principle, namely that if all voters prefer a certain alternative to another, then that alternative should be ranked ahead of the other; versus, on the other hand, a more esoteric-sounding principle, the Independence of Irrelevant Alternatives. (4) The Independence of Irrelevant Alternatives asserts that in deciding which of two alternatives ought to be ranked more highly than the other we need only worry about the relationship of those two alternatives to each other, not their relationship to some third alternative. To use a well-worn example, in deciding whether to order steak or chicken for the group, we should not have to ask about their feelings about fish.

By adapting the original proof of Arrow's theorem to the very special case of two voters and three alternatives, we can quickly see how the tension between the Pareto principle and the Independence of Irrelevant Alternatives arises. Suppose Voter 1 prefers alternative A to alternative B, whereas Voter 2 feels the opposite (i.e., he would rank B ahead of A). Suppose further that Voter 2 would rank alternative B ahead of alternative C, whereas Voter 1 feels the opposite (i.e., he would rank C ahead of B). If we want the aggregation of their desires to be a collective enterprise, in other words, to take into account to some extent both of their desires, we might let Voter 1 prevail in the ranking of A and B and Voter 2 in the ranking of B and C. The overall ranking we then obtain is A, B, and C, in that order. The problem is that there is nothing in what I have said so far that would prevent it from being the case that both Voter 1 and Voter 2 prefer C to A. Thus if we adopt the A-B-C ranking we are violating the Pareto principle. If we want to restore coherence, something has to give. Assuming we do not want to relinquish transitivity or the collective nature of the decision-making process, or the Pareto principle, what is left? We could make it so that we override the wishes of one of the voters only in those cases in which failing to do so would get us into a cycle. But that makes the decision about what to do about two alternatives dependent on the voters' feelings about some third alternatives--in other words, it makes it dependent on an irrelevant alternative.

It is this feature of Pareto optimal voting systems that opens them up to agenda manipulation. Someone who wants to derail an outcome he finds undesirable has the opportunity of injecting some hitherto unconsidered alternative into the agenda and might thus suddenly cause the outcome to shift from the one he disliked to another one--not the one he injected, that would not be so strange--but one that was previously on the table but remained unchosen. Even once one understands the inevitability of this, its appearance cannot help but be startling. A much-loved example from the 2000 Olympics involved the rivalry between two figure skaters Nancy Kerrigan and Oksana Bayul. Bayul ended up winning the gold medal, Kerrigan the silver. Crucial to this outcome was the participation in the competition of a Chinese skater, but for whom Kerrigan would have won the gold and Bayul the silver--holding constant, that is, the performances of each skater.