Scribd
Upload
39 views3 pages

Common Knowledge of Rationality and Successive Elimination of Strictly Dominated Strategies

This document discusses game theory and the concept of common knowledge of rationality. It provides examples to illustrate how assuming players have common knowledge of rationality up to different degrees allows equilibrium strategies to be determined through successive elimination of strictly dominated strategies. In some games, common knowledge of only individual rationality is enough, while in others it requires assuming deeper levels of common knowledge for players to deduce the optimal strategies. The method breaks down if no strategies are strictly dominated, as in the final example game presented.

Uploaded by

Adnan Muridi Baba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
39 views3 pages

Common Knowledge of Rationality and Successive Elimination of Strictly Dominated Strategies

This document discusses game theory and the concept of common knowledge of rationality. It provides examples to illustrate how assuming players have common knowledge of rationality up to different degrees allows equilibrium strategies to be determined through successive elimination of strictly dominated strategies. In some games, common knowledge of only individual rationality is enough, while in others it requires assuming deeper levels of common knowledge for players to deduce the optimal strategies. The method breaks down if no strategies are strictly dominated, as in the final example game presented.

Uploaded by

Adnan Muridi Baba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Game Theory and Economic Applications – 6BUS1140

Common knowledge of rationality and successive elimination of strictly dominated


strategies
In the “Prisoners’ Dilemma”, the only assumption that we needed in order to find the
equilibrium was that the players were rational in the mainstream economics sense (i.e.
each player tried to maximise his/her utility). By use of this assumption, we were able
to eliminate the strategy “non-confession” from both players’ sets of available
strategies, because it was strictly dominated by “confession”. Nevertheless, the
assumption of rationality does not suffice for finding an equilibrium in most static
games. Consider, for example, the game below:
Player C
C1 C2
R1 1,0 3,1
Player R
R2 0,2 2,1

In this game, strategy R2 is strictly dominated by R1. Hence, we know that a


rational player R would never choose R2, and therefore, in equilibrium, player R
chooses R1. However, player C does not have a strictly dominated strategy: neither C1
nor C2 fare strictly worse than the other strategy; if R chooses R1, then C2 is better
than C1, but if R chooses R2, then C1 is better than C2. This means that the assumption
that player C is rational cannot lead us to a conclusion as to what C will select. If,
however, we assume that player C knows that player R is rational, then C knows
that R2 would never be played, and this means that C ought to choose C2 (for, given
that R chooses R1, C gets a utility payoff equal to 0 when (s)he chooses C1, and a utility
payoff equal to 1 when (s)he chooses C2). This additional assumption that we made
(i.e. that C knows R is rational) is referred to as common knowledge of rationality, and,
in this particular game, it was necessary (along with the standard assumption of
individual rationality) for finding the equilibrium of this game, according to which,
player R chooses R1 and player C chooses C2.

Formally, the assumption of common knowledge or rationality (CKR hereafter)


means that we assume each player to know that the other player is rational, and,
sometimes, this common knowledge may be deeper in that player i knows that player j
knows that i is rational – or even deeper such that player i knows that player j knows
that player i knows that j is rational, and so on.

To illustrate, consider the game below:

Player C
C1 C2 C3 C4
R1 1,1 1,0 4,2 3,1
R2 0,2 0,0 2,1 5,0
Player R
R3 2,2 1,1 3,3 4,0
R4 1,3 5,2 2,2 3,4
In this game each player has four available strategies. Player R does not have a strictly
dominated strategy, but strategy C2 is strictly dominated (by C1) for player C. Thus,
we can eliminate it under the assumption that C is rational. Now that C2 is eliminated,
under the assumption that player R knows that C is rational (1st degree CKR), R4
becomes strictly dominated by R3 (since R knows that C2 would never be played), and
it can be eliminated too. But under the assumption that C knows that R knows that C
himself/herself is rational (2nd degree CKR), C knows that R4 shall never be played,
and this elimination of R4 now makes C4 strictly dominated (by C3) as well. Similarly,
if R knows that C knows that R knows that C is rational (3rd degree CKR), R knows
that C4 will not be played, and hence, R2 now becomes strictly dominated (by both R1
and R3). CKR of 4th degree (C knows that R knows that C knows that R knows that C
is rational) means that C, knowing that R2 has been eliminated, would never choose C1
because it now is strictly dominated by C3. Thus, C can only play C3 (it is the only
strategy that “survived”). Under 5th degree CKR, R knows that C will choose C3, and
this means that R will choose R1 (which strictly dominates over the only other surviving
strategy, R3). Therefore, the equilibrium of this game is that R chooses R1 and C
chooses C3. To arrive at this conclusion, we had to assume individual rationality, as
well as common knowledge of rationality up to the 5th degree.

The reasoning that helped us find the equilibrium in this game is known as the
method of successive elimination of strictly dominated strategies. It uses the
assumption of CKR up to whatever degree is needed, in order to bring the game to a
point where no player has a strictly dominated strategy anymore. In the last two games
we examined, this assumption made strategies “fall like dominoes”, the one after the
other, and along with one additional degree of CKR, until each player remained with
only one strategy, leading us to a unique equilibrium. However, the method of
successive elimination of strictly dominated strategies is not guaranteed to lead to a
resolution where each player is left with just one strategy. In fact, in most games none
of the strategies available to the players are strictly dominated, so this method might
not even take off at all (see for example, the game below).

Player C
C1 C2 C3
R1 2,1 0,0 1,2
Player R
R2 0,0 1,1 0,0
R3 1,2 0,0 2,1

Here we see that neither R nor C have a strictly dominated strategy, and
therefore, it is not possible to eliminate any one of these strategies. Thus, even if we
have assumed players who not only maximise utility, but also have mutual knowledge
of this fact (a knowledge that can theoretically go to infinite depths), there is nothing
that we can say on the resolution of this game, unless we make further assumptions
(we’ll do that on Thursday!). In the absence of strictly dominated strategies, the players
would be rational to make any choice: player R would be rational to choose R1, if (s)he
expected player C to choose C1. Or, (s)he could rationally choose R2, if (s)he expected
C to choose C2. Finally, (s)he would also rationally choose R3, if (s)he expected C to
choose C3. Similarly, C would rationally choose C1, C2 or C3 if (s)he expected player
R to choose R3, R2 and R1 respectively. These strategies are called rationalisable.

Formally, a strategy is rationalisable if it survives the method of successive elimination


of strictly dominated strategies.
In the game above, the method of successive elimination of strictly dominated strategies
cannot be even applied (for there are no strictly dominated strategies to eliminate), and
therefore, all three strategies for both players are rationalisable.

We should not eliminate weakly dominated strategies when we apply the method of
successive elimination of strictly dominated strategies!

Reality check: Is common knowledge of rationality of infinite degree a plausible


assumption to make? Check the additional material resource for this week to see the
results of a related experiment!

You might also like