In most states it is permitted to write in the names on the ballot, and these write-in candidates can win if they assemble enough votes. Thus, in all elections in which I might conceivably participate, there are at least three candidates: the Republicans, the Democrats, and my own favorite candidate for president—Gordon Tullock. Indeed, using the notation of the Ferejohn and Fiorina article, assuming that we are talking about 1972, I would find k having a value of about .001. Under the circumstances, if I understand the article correctly, I should always vote, and the vote should always take the form of writing in my own name. Further, as far as I can see, this advice can be generalized. Everyone who would really like to be president should vote and write in his own name, because the minimum regret that they produce for three-candidate elections is also correct for the elections in which there are 30 million candidates. The only problem here is that it would, of course, amount to participating in a lottery, and my possible gain from writing in my own name might turn out to be less than the cost of writing it in. Granted that people are willing to buy lottery tickets, even when the various states offering them take very substantial rakeoffs, I do not imagine this would be much of a disadvantage.
This same conclusion could have been deduced from the Casstevens article, which Ferejohn and Fiorina quite properly criticize as being mathematically incorrect.