Firms lobby for a single license, rival laboratories race for one patent, and agencies prepare costly pitches for an account that only one of them will win. What each contestant spends improves its own chance and is gone whatever the result.
The game tests how much a group spends in competing for a fixed prize, and how the total compares with the value of the prize. It shows that competition can consume resources without creating anything, and that people tend to spend more than the equilibrium prescribes. The same structure describes litigation, advertising wars, promotion tournaments and elections.
Tullock (1967) argued that the social cost of a monopoly or a tariff includes the resources spent on obtaining it, and Krueger (1974) gave the activity the name of rent seeking. In the contest model of Tullock (1980), each contestant's chance of winning is the contestant's own spending divided by the total spending of the group, with spending raised to a power that measures how decisive it is.
With a group of a given size and a power of one, each contestant who cares about expected earnings spends the prize multiplied by the group size minus one and divided by the square of the group size. The group as a whole then spends the fraction of the prize given by the group size minus one over the group size, which approaches the whole prize as the group grows. With the default settings, four contestants and a prize of 100, the equilibrium spending is 18.75 each and 75 in total. The group would do best by spending nothing and letting chance decide. When the prize is divided in proportion to spending, expected earnings are the same as under the lottery and the element of risk is removed.
Spending above the equilibrium is the regular finding of the experimental literature. In an early experiment by Millner and Pratt (1989), two contestants competing for a prize of 8 dollars spent 2.24 dollars each on average where the equilibrium was 2. In the survey of Sheremeta (2013), 28 of 30 studies found significant overspending, with a median excess of 72 percent over the equilibrium level, and in some studies participants on average lost money by taking part. Dechenaux, Kovenock and Sheremeta (2015) review the wider literature on contests, tournaments and all-pay auctions.
Behavior varies widely across individuals, with some contestants spending nothing and others most of their endowment, a difference that the average conceals. Explanations that have been proposed include a value placed on winning in itself, mistakes, and a concern with doing better than the others.