You are on page 1of 3

EC 246: Decisions & Games Term IV, 2020

Problem Set VII – Repeated Games

Problem 1: Tacit Collusion among asymmetric Firms

Consider the ”Supply-Choice Game” in Section 5.1 of Chapter 5 in the Lecture Notes in
which two whole-sellers – B and G – supply rice (of identical quality) to a market near
Kolkata. Consider the following modification of that Cournot duopoly game: The firms
simultaneously decide on their day’s supply volume at 6 am each morning, but they can
change their supplies from day to day. The daily patience factor for each firm is 0.99; and
every day, they know that there is a 70% chance that the game will be played following day.
Thus, the daily discount factor for each firm is 0.693.
Consider the asymmetric collusion target {10, 4}; i.e., the whole-sellers attempt to support
the daily outcome in which the more-efficient firm B supplies 10 quintals of rice and the less-
efficient firm G supplies 4 quintals. Can they “rationally sustain” this collusion level (as an
SPNE outcome) by each firm pursuing the appropriate ‘perpetual Nash reversion’ strategy?

Problem 2: Cheating against “trigger strategy”


Two firms Row and Column compete in “advertising budgets” from the beginning of the
current year. At the beginning of every year, each firm can choose a “low” advertising
budget, or a “medium” budget, or a “high” budget for the year. The yearly profits (in
millions of rupees) for each strategy combination are given by the following payoff matrix:

low medium high


low 50, 50 32, 60 25, 55
medium 60, 32 40, 40 26, 42
high 55, 25 42, 26 30, 30

The two firms play this game repeatedly. In every year, they know that there is 2/3 chance
that the game will be played in the next year. Further, the annual ‘discount factor’ for each
firm is (3/4) [i.e., one rupee a year from now is valued the same as (3/4) rupees today].
To avoid a fierce advertising battle and to sustain self-enforcing collusion, the firms consider
playing the following ‘trigger strategy’:
“A firm will choose ‘low advertising budget’ on January 1 of the current year, and keep on
choosing the low budget as long as both firms have chosen the low budget in previous years.
If any firm chooses any other budget in any year, the two firms will choose high budget for
the next two years, after which they will go back to choosing low budgets every year (with an
identical two-year punishment threat for any subsequent deviation).” This is in effect a
collusion strategy with a “two-period Nash reversion threat”, rather than the “grim strategy”
which specifies a threat of “perpetual Nash reversion”.
Prove that the proposed trigger strategy rationally sustains collusion on ‘low advertising’. Is
the resultant SPNE renegotiation-proof?

1
EC 246: Decisions & Games Term IV, 2020

Finally, consider a change in the probability of game continuation: Suppose that every year,
the firms know that there is 4/9 chance that the game will be played in the next year. If
Column expects Row to play the ‘trigger strategy’ described above, what will be Column’s
“subgame-perfect (i.e., credible) best response”?

Problem 3: Tacitly sustaining collusive payoffs in the convex hull of stage-game payoffs
Two firms Micro and Macro compete in “marketing outlays” from the current year onwards.
At the beginning of every year, each firm chooses a “low” marketing outlay, or a “medium”
outlay, or a “high” outlay for the year. Yearly profits (in thousand rupees, net of marketing
outlays) for each outlay combination are given by the following payoff matrix:
Macro

low medium high


.
low 18 , 18 14 , 30 10 , 23
Micro
medium 30 , 14 15 , 15 8 , 13
high 23, 10 13 , 8 5, 5

The firms play this game repeatedly every year, and have a common annual discount factor δ.
(i) In pre-play communication on January 1 of the current year, the firms attempt to sustain
the ‘cooperative outcome’ of {low, low} every year by using the threat of ‘perpetual Nash
reversion’. For what values the common δ  (0, 1) will they succeed?
(ii) Next, suppose that δ = 0.9, and that in pre-play communication on January 1 of the current
year, the firms try to devise history-dependent strategies that will sustain cooperation under
which each firm will get at least 210 thousand rupees in present value profits. As a strategy-
consultant to the two firms, suggest a renegotiation-proof SPNE strategy profile that will
enable the firms to attain this objective, and that has the following structure: each simple
strategy prescribes a “cooperation code”, and an “optimal penal code for each player” that
specifies that cheating in the cooperation path or in any punishment path by any player i will
initiate the penal code for player i (which will continue till a further instance of cheating).

Problem 4: Abreu’s Example


Consider two symmetric firms playing a Cournot super-game in supply volumes, with the
restriction that the only permissible choices are ‘high’ volume, ‘medium’ volume, and ‘low’
volume. This is Dilip Abreu’s example presented in the posted Notes on Repeated Games:
Firm 2

low medium high


.
low 10 , 10 3 , 15 0,7
Firm 1 medium 15 , 3 7,7 –4 , 5
high 7, 0 5 , –4 –15, –15

2
EC 246: Decisions & Games Term IV, 2020

Abreu shows that when the firms’ common discount factor is δ = 4/7, the following
symmetric (but non-stationary) penal code rationally sustains the cooperative outcome (low,
low) in every period of the infinite super-game:
“Deviant firm is required to play medium in the first round after deviation, and low thereafter
and ever after; the ‘non-deviant’ firm is required to play high in the first round after
deviation, and medium thereafter and ever after.”
Verify that the above penal code is “optimal” and “strongly renegotiation-proof” for each of
the firms, and sustains collusion on (low, low) forever.

You might also like