Solutions to Chapter 15 Exercises

SOLVED EXERCISES

S1. Recall that an object has an objective value if all bidders have the same value for it and know the
value.

(a) A $20 gift card at Store X is worth $20 to someone who is already going to be shopping
at Store X, but worth less to others who would not normally shop there. For that reason, gift cards may
not have an objective value. However, because Amazon is so ubiquitous, most students in a game-theory
class are likely to be using Amazon in the near future to buy something; so, the gift card is worth $20 to
most students, and all of them know this.

(b) Students will naturally differ in how much they are willing to pay to have lunch with the
professor; so, the “object” here clearly does not have an objective value. (Whether bidders have private or
common values is unclear and will depend on why students vary in their willingness to pay for lunch. For
instance, if students vary in how much they like the professor, and don’t care what others think about her,
then the auction will have private values. On the other hand, if students are unsure how interesting it will
be to have lunch with the professor, they will naturally care about the opinions of others and the auction
will not have private values.)

(c) Students’ desire for a bottle of water will depend on how thirsty they are; so, the “object”
here also clearly does not have an objective value. (Because one’s thirst does not depend on how thirsty
others are, bidders in an auction for a bottle of water will naturally have private values.)

S2. The painter can compare her estimated cost with a job’s true cost only when she does the job. But
the painter does a job only when she agrees (through the bidding process) to do it for less than anybody
else would charge. The fact that she submitted the lowest bid suggests that this is a job for which the
painter has likely underestimated the real cost. This is therefore a winner’s curse; the painter only wins
contracts that tend to cost more than she expected.

Games of Strategy, Fifth Edition Copyright © 2021 W. W. Norton & Company

S3. If you turn out to be the lowest bidder and therefore fail to get the object, this must be because all
of the others got a higher estimate of the value of the object than you did. Therefore, you have reason to
believe that you got an exceptionally low estimate—one with a large and negative error. This is a “loser’s
curse,” just like the winner’s curse that occurs when only one object is auctioned among many bidders,
and you are the highest bidder only if you get an exceptionally high estimate. You will correct for this
loser’s curse by bidding somewhat more aggressively than would be justifiable on the basis of your own
estimate alone. The precise calculation of course requires more information on the probability distribution
of the errors, and so on. (See Wolfgang Pesendorfer and Jeroen Swinkels, “The Loser’s Curse and
Information Aggregation in Common Value Auctions,” Econometrica, vol. 65, no. 6 [November 1997],
pp. 1247–1281.)

S4. (a) If you offer $3,000, the current owner sells only if the car’s true value to him is less than
$3,000. Since all values between $1,000 and $3,000 are equally likely, a car that you’ll buy has an
expected value of $2,000. On average, such a car will be worth (4/3)  $2,000 = $2,667 to you. If you
offer $3,000, therefore, you can expect the transaction to cause you to lose $333 in value.

(b) Suppose that your offer equals B. The current owner will sell you his car if its value to
him is between $1,000 and B; the expected value of such a car is 1,000 + (B – 1,000)/2 = 500 + B/2. The
expected value to you of a car that is sold to you is thus (4/3)(500 + B/2), and your net gain is (4/3)(500 +
B/2) – B. If you are not going to lose money, this expression must equal 0, and this occurs at B = 2,000.

S5. (a) Bob wins the auction with probability v B /12 and, when winning, gets value v B but pays
2
v B /2 on average. Overall, then, Bob’s expected surplus S( v ¿¿ B)= v B × v B = ( B ) ¿. For the specific
v
12 2 24
27 3 3
values mentioned, S(12)=6 ; S ( 9 )= =3.375 ; S ( 6 )= =1.5 ; and S ( 3 )= =0.375 .
8 2 8

(b) By part (a), S ( 12 )=6>12 /3 and S ( 9 )=3.375> 9/3 ; so, Bob prefers having an auction
for the whole cupcake rather than getting half for free given value v B=$ 12 or v B=$ 9 . On the other
hand, since S ( 6 )=1.5< 6/2 and S ( 3 )=0.375< 3/3, Bob prefers getting half for free given value v B=$ 6
or v B=$ 3.

Games of Strategy, Fifth Edition Copyright © 2021 W. W. Norton & Company

S6. (a) If your opponent always bids half her value, and her value is uniformly distributed on [0,
1], then her bid will be uniformly distributed on [0, 0.5].

If you bid b = 0.1, the probability of winning is 0.1/(0.5 – 0) = 0.2.

If you bid b = 0.4, the probability of winning is 0.4/(0.5 – 0) = 0.8.

If you bid b = 0.6, you cannot be outbid by your opponent, because 0.6 is greater than her
maximum possible bid. Your probability of winning is 1.

(b) Pr(win) = 2b for 0  b  0.5

=1 for b > 0.5

(c) π = (v – b) * Pr(win) + 0 * Pr(lose) = (v – b) * 2b = 2bv – 2b2 for 0  b  0.5

=v–b for b > 0.5

(d) The first-order condition for π when 0  b  0.5 is: ∂π/∂b = 2v – 4b = 0, which implies
that b* = v/2. Because v ≤ 1, the solution b is ≤ 0.5. Also, the profit π decreases as b increases beyond 0.5.
Therefore b = v/2 gives the global maximum of π.

(e) We see that the best response to an opponent’s bidding half her value is to bid half your
value. Likewise, when you bid half your value, your opponent’s best response is to bid half her value. The
bidding strategy b(v) = v/2 is a mutual best response, or Nash equilibrium, for this auction.

S7. (a) Suppose all the other n−1 bidders are using the given bidding-strategy function b (V ),
and your value is X. You have to choose your bid y. Let Y be the solution to y = b(Y). So your bid is
equivalent to pretending to have a value Y and applying the bidding-strategy function b(Y) to it. You will
win if all other n−1 bidders’ values are such that their b ( V ) <b ( Y ) , that is, V <Y . Since each V is
uniformly and independently distributed in the interval [0,1], the probability of all n−1 values being less
than Y is Y n−1. You get X (your true value) if you win. You pay b (Y ) (using the pretend value in your
bidding strategy) whether or not you win (probability 1). So your expected payoff, denoted by Π (Y ), is

n−1 n
Π ( Y )=X Y n−1−b (Y )=X Y n−1−[ ]Y .
n

Then

Games of Strategy, Fifth Edition Copyright © 2021 W. W. Norton & Company

 Π ' ( Y )=X ( n−1 ) Y n−2− [ ]
n−1
n
nY n−1 =(n−1)Y n−2 ( X −Y ).

So Π ' ( Y )> 0 for Y < X , and <0 for Y > X .Therefore Y = X maximizes your expected payoff, that is,
using the bidding function b(V) for your true value X is your optimal strategy. This function is your best
response when all the other n−1 players are using it. That is, all n players using the strategy b(V) is a
Nash equilibrium.

(b) b 3 ( V ) >b 2 ( V ) when ( 23 ) V >( 12 ) V , or V > 34 . And:

2

3
b 4 (V )> b3 ( V ) when ( )V 2 >
4
2
3 () 8
V , or V > .
9

For the general case, we have the following:

2
n n +1 n−1 n (n+1)(n−1) n −1
b n+1 ( V )> bn (V ) when V > V or V > = 2 .
n+1 n n
2
n

That is, with a larger number of players, bids will be lower for most players, but higher only for those
who happen to have private values at the very top end of the range.

Games of Strategy, Fifth Edition Copyright © 2021 W. W. Norton & Company