However, if there is more than one Nash equilibrium, there is always the possibility
behavior in early stages of repeated interaction. This good behavior is sustained by the prospect of better
future play than that which follows short-term opportunistic behavior. Again, the general argument is the
same as the one that we saw in the context of the repeated Modified Prisoners' Dilemma. Good behavior in
early
page_218
Page 219
interactions can be rewarded by the play of better Nash equilibria in future subgames, while any deviations
from this behavior can be punished by the play of bad Nash equilibria in future subgames.10
14.3 Case Study: Treasury Bill Auctions
The institutional structure of Treasury bill auctions was explained earlier in this chapter. We will therefore
focus on one key aspect of these auctions, the pricing issue. Recall that for some securities, there is a
single-price auction in which all buyers pay the same price. For some others, there is a multiprice auction in
which different buyers pay different prices. The question that we will now investigate is the following: If the
Treasury wants to maximize the amount that it collects, which of the two auction forms should it use?
In order to keep the analysis tractable, we will make several simplifying assumptions. First, we will assume
that there are two financial institutions, or players, at this auction. Second, the quantity that the Treasury
sells remains fixed from auction to auction; let this amount be 100.11 Third, we will assume that there are
two prices and two quantities that each buyer can offer; call the prices high (h) and low (l) and the amounts
50 and 75. Fourth, buyers care only about profits; denote the profit per security if the price is h as ph and
likewise the profit if the price is l as pl. Suppose that both profit levels are positive (and, of course, pl > ph).
If each buyer wants to buy at a high price, then the total demand at that price is at least 100 and all of the
Treasuries sell at that price. Likewise, if both buyers want to buy at the low price, then the market price is
low. If, however, one of the buyers wants to buy at h while the other wants to buy at l, then the price
outcome depends on the type of auction. In a single-price auction, the market price will be low, while in a
multiprice auction one buyer will pay h and the other will pay l. In either case, the high bidder gets all of the
quantity he asks for and the remaining quantity is allocated to the low bidder. Finally, if the price bids are
the same, then the quantity is allocated in proportion to the quantity demands. For example, if one buyer
wants 75 units and the other wants 50, then the former gets 60 of the available 100 units.12
With these assumptions, the strategic form of a single-price auction is as follows:
Buyer 1 \ Buyer 2 50, h 75, h 50, l 75, l
50, h50ph, 50ph 40ph, 60ph 50pl, 50pl 50pl, 50pl
75, h60ph, 40ph 50ph, 50ph 75pl, 25pl 75pl, 25pl
50, l50pl, 50pl 25pl, 75pl 50pl, 50pl 40pl, 60pl
75, l50pl, 50pl 25pl, 75pl 60pl, 40pl 50pl, 50pl
For example, if buyer 1 bids a high price against 50, l, then he gets all of his quantity at a low price; the price
is low because there is no demand for all 100 units at price h. If he bids 75, l (against 50, l), then he gets 60
units allocated at the low price. Note furthermore
10In the Modified Prisoners' Dilemma there are two Nash equilibria in the stage game, one of
which is preferred by both players to the other equilibrium. In general there may not exist such
unanimity in opinion; that is, there may be two Nash equilibria in the stage game, the first of which
is preferred by player I while the second is preferred by player 2. Nevertheless it is still possible to
sustain good behavior in early stages by an appropriate modification of the arguments discussed in
the text. For details, consult Drew Fudenberg and Jean Tirole, Game Theory (Cambridge, MA:
MIT Press, 1991).
11Note that we are looking at the repeated auction of any one kind of security, say, 26-week Treasury
bills. The assumption that the same number of such bills are sold every week makes it a repeated game.
If the quantity were to vary from week to week, each week's stage game would be a little different from
every other week's. The current analysis can be generalized in that direction. In Chapter 17 we will
look at repeated games with random variations in the stage game (for instance, because the federal
government's financing needs vary randomly from time to time).