Microeconomics I: Game Theory Lecture 6: The Indifference Principle (see Osborne, 2009, Sect 4.3.3,4.3.4,4.8,4.10,4.11) Dr. Michael Trost Department of Applied Microeconomics November 29, 2013 Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 1 / 37
37
Embed
Lecture 6: The Indifference Principle - Uni Erfurt€¦ · Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 24 / 37 Finding all mixed action Nash equilibria The indifference
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Microeconomics I: Game Theory
Lecture 6:The Indifference Principle
(see Osborne, 2009, Sect 4.3.3,4.3.4,4.8,4.10,4.11)
Dr. Michael TrostDepartment of Applied Microeconomics
November 29, 2013
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 1 / 37
Finding mixed action Nash equilibria
Figuring out directly the set of mixed action Nash equilibria isoften a tedious task.
In the following we present two alternative methods fordetecting mixed action Nash equilibria:
The first method is based on the players’ best responsecorrespondences, the second one on the indifference principle.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 2 / 37
Method of best response correspondence
The method based on players’ best response correspondences(also termed as the two-step method) has been already appliedto figure out the set of pure action Nash equilibria.
As in the case of pure action Nash equilibria, we cancharacterize mixed action Nash equilibria by the players’ bestresponse correspondences.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 3 / 37
B.r.c. and mixed action Nash equilibrium
Consider a vNM strategic game Γ := (I , (Ai)i∈I , (%i)i∈I ) anddenote by Bi the best response correspondence of player i ∈ I .
A profile α∗ := (α∗i )i∈I ∈ ×i∈I∆(Ai) of mixed actions is a mixedaction Nash equilibrium of Γ if and only if, for every player i ∈ I ,
α∗i ∈ Bi(α∗−i)
holds.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 4 / 37
The two-step solution method
The characterization of the mixed Nash equilibrium by players’best response correspondences suggests following two-stepmethod for detecting mixed action Nash equilibria:
1 Find the best response correspondence Bi for each playeri ∈ I .
2 Find all profiles α∗ := (α∗i )i∈I of mixed actions that satisfyα∗i ∈ Bi(α
∗−i) for each player i ∈ I .
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 5 / 37
Example: Matching Pennies
Player Bq 1− q
Head Tail
Player Ap Head 1,-1 -1,1
1− p Tail -1,1 1,-1
In the following, we figure out the mixed action Nash equilibriaof the vNM strategic game MATCHING PENNIES by the two stepmethod.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 6 / 37
Example: Matching Pennies
To derive the players’ best response correspondences, proceedas follows for each player.
1 Calculate the player’s expected utility if both playersrandomize.
2 For each possible randomization by the opponents,describe the set of all mixed actions which maximizes theplayer’s expected utility given this randomization.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 7 / 37
Example: Matching Pennies
Suppose player B chooses head with probability q and tail withprobability 1− q.
If player A chooses head with probability p and tail withprobability 1− p, then
occurs withprobability p ·q p ·(1−q) (1−p)·q (1−p)·(1−q)
and B’sexpected utilityis = p ·q ·(−1) + p ·(1−q)·1 + (1−p)·q ·1 + (1−p)·(1−q)·(−1)
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 12 / 37
Example: Matching Pennies
Suppose player A choose head with probability p and tail withprobability 1− p.
If player B chooses head with probability q and tail withprobability 1− q, then B ’s expected utility is
UB(p, q)
= pq(−1) + p(1− q)1 + (1− p)q1 + (1− p)(1− q)(−1)
= −pq + p − pq + q − pq − 1 + q + p − pq
= −1 + 2p + 2q − 4pq
= −1 + 2p + 4q(12− p)
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 13 / 37
Example: Matching Pennies
Suppose player A choose head with probability p and tail withprobability 1− p.
If player B chooses head with probability q and tail withprobability 1− q, then B ’s expected utility is
UB(p, q) = −1 + 2p + 4q(12− p).
Hence, the best response BB is given by
BB(p) =
{1} if p < 1
2,
[0, 1] if p = 1
2,
{0} if p > 1
2.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 14 / 37
Example: Matching PenniesGraph of B ’s best response correspondence
BB(p)
p0
00.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0.5
0.6
0.6
0.7
0.7
0.8
0.8
0.9
0.9
1.0
1.0
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 15 / 37
Example: Matching PenniesThe best response correspondences of A and B in one figure.
q
p0
00.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0.5
0.6
0.6
0.7
0.7
0.8
0.8
0.9
0.9
1.0
1.0
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 16 / 37
Example: Matching Pennies
The plot of the best response correspondences of A and B
suggests that the mixed action
((p, 1− p), (q, 1− q)) =((12, 12), (1
2, 12))
is a Nash equilibrium of MATCHING PENNIES.
Indeed, the best response correspondences of A and B satisfy
(12, 12) ∈ BA(1
2, 12) and (1
2, 12) ∈ BB(1
2, 12).
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 17 / 37
Exercise: Battle of Sexes
Player Bq 1− q
Bach Stravinsky
Player Ap Bach 2,1 0,0
1− p Stravinsky 0,0 1,2
EXERCISE: For above vNM strategic game BATTLE OF SEXES,determine the expected utility of A and B and their bestresponse correspondences! Plot them in one figure. Find the setof all mixed action Nash equilibria.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 18 / 37
Exercise: Battle of SexesGraph of the best response correspondences of A and B
q
p0
00.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0.5
0.6
0.6
0.7
0.7
0.8
0.8
0.9
0.9
1.0
1.0
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 19 / 37
Finding mixed action Nash equilibria
Another method to figure out the set of mixed action Nashequilibria is to apply the so-called indifference principle, whichis stated next.
The indifference principle is an alternative characterization ofmixed action Nash equilibria.
Before stating this principle some useful notation is introduced.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 20 / 37
Mixed action profiles containing pure actions
Consider vNM strategic game Γ := (I , (Ai)i∈I , (%i)i∈I ) and letai ∈ Ai be a definite action of player i .
The mixed action profile (ai , α−i) ∈ ×i∈I∆(Ai) means that
player i chooses the definite action (or synonymously, thedegenerated mixed action) ai ,
every player j different to i chooses the mixed action αj .
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 21 / 37
Mixed action profiles containing pure actions
The expected utility of player i obtained by mixed action profile(ai , α−i) is
Ui(ai , α−i) =∑a′∈A
(∏j∈I
αj(a′j)
)ui(a
′i , a′j)
=∑
a′−i∈A−i
∏j∈I\{i}
αj(a′j)
ui(ai , a′j)
where the second row follows from the fact that αi is thedegenerated lottery that ascribes probability of 1 to action ai .
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 22 / 37
Indifference principle
Theorem 6.1Consider a vNM strategic game Γ := (I , (Ai)i∈I , (%i)i∈I ).
A profile (α∗i )i∈I of mixed actions is a Nash equilibrium in mixedactions of Γ if and only if, for each player i ∈ I , the following twoconditions are satisfied:
1 Ui(ai , α∗−i) = Ui(a
′i , α∗−i) holds for every actions ai , a′i ∈ Ai to
which α∗i assigns positive probability,
2 Ui(ai , α∗−i) ≥ Ui(a
′i , α∗−i) holds for every action ai ∈ Ai to which
α∗i assigns positive probability and for every action a′i ∈ Ai towhich α∗i assigns zero probability.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 23 / 37
Indifference principle
The indifference principle states that a profile α∗ := (α∗i ) ofmixed actions is a Nash equilibrium in mixed actions if and onlyif, for each player i ∈ I ,
1 all definite actions realizable under mixed action αi yieldthe same expected payoff for player i ,
2 all definite actions not realizable under mixed action αi
yield an expected payoff for player i that does not exceedthat of definite actions realizable under mixed action αi
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 24 / 37
Finding all mixed action Nash equilibria
The indifference principle opens up the following method tofind all mixed actions Nash equilibria of a vNM strategic game.
1 For each player i ∈ I , choose a non-empty subset Si of Ai
2 Check whether there is a mixed action profile (αi)i∈I where(1) supp(αi ) = Si holds for every player i ∈ I
(2) the two conditions of the indifference principle hold forevery player i ∈ I .
3 Repeat the analysis for every collection of non-emptysubsets of the players’ sets of actions.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 25 / 37
Example: Matching Pennies
Player Bq 1− q
Head Tail
Player Ap Head 1,-1 -1,1
1− p Tail -1,1 1,-1
QUESTION: Figure out all mixed action Nash equilibria of thevNM strategic game MATCHING PENNIES by the procedurebased on the indifference principle.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 26 / 37
Example: Matching Pennies
The following combinations of supports of mixed actions mustbe checked.
Support of player B ’smixed action
{H} {T} {H,T}
Support of playerA’s mixed action
{H}{T}
{H,T}
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 27 / 37
Example: Matching Pennies
CASE: The supports of mixed actions αA and αB aresuppA(αA) := {Head} and suppB(αB) := {Head}, respectively.
REASONING: Obviously, αA = Head and αB = Head apply.Since
UB(Head,Head) = −1 < 1 = UB(Head,Tail)
holds, condition 2 of the indifference principle is violated forplayer B .
Hence, no mixed action Nash equilibrium with such supportsexists.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 28 / 37
Example: Matching Pennies
CASE: The supports (suppA(αA), suppB(αB)) of mixed actionsαA and αB belong to one of the combinations ({Head}, {Tail}),({Tail}, {Head}) or ({Tail}, {Tail}).
REASONING: Due to arguments similar to that of the previouscase condition 2 of the indifference principle is violated by oneplayer.
Hence, no mixed action Nash equilibrium with such supportsexists.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 29 / 37
Example: Matching Pennies
CASE: The supports of mixed actions αA and αB aresuppA(αA) := {Head,Tail} and suppB(αB) := {Head}, resp.
REASONING: Obviously, 0 < p < 1 and αB = Head apply. Since
UA(Head,Head) = 1 > −1 = UA(Tail,Head)
holds, condition 1 of the indifference principle is violated forplayer A.
Hence, no mixed action Nash equilibrium with such supportsexists.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 30 / 37
Example: Matching Pennies
CASE: The supports (suppA(αA), suppB(αB)) of mixed actionsαA and αB are either ({Head,Tail}, {Tail}), ({Tail}, {Head,Tail})or ({Tail}, {Head,Tail}).
REASONING: Due to arguments similar to that of the previouscase condition 1 of the indifference principle is violated by oneplayer.
Hence, no mixed action Nash equilibrium with such supportsexists.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 31 / 37
Example: Matching Pennies
CASE: The supports of mixed actions αA and αB aresuppA(αA) := {Head,Tail} and suppB(αB) := {Head,Tail}),respectively.
REASONING: For player A, condition 1 of the indifferenceprinciple implies that
UA(Head, αB) = UA(Tail, αB) ,
or equivalently,
q · 1 + (1− q) · (−1) = q · (−1) + (1− q) · 1,
which results in q = 1
2.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 32 / 37
Example: Matching Pennies
CASE: The supports of mixed actions αA and αB aresuppA(αA) := {Head,Tail} and suppB(αB) := {Head,Tail}),respectively.
REASONING: For player B , condition 1 of the indifferenceprinciple implies that
UB(αA,Head) = UB(αA,Tail) ,
or equivalently,
p · (−1) + (1− p) · 1 = p · 1 + (1− p) · (−1),
which results in p = 1
2.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 33 / 37
Example: Matching Pennies
CASE: The supports of mixed actions αA and αB aresuppA(αA) := {Head,Tail} and suppB(αB) := {Head,Tail}),respectively.
REASONING: Summing up the previous discussion, the mixedaction profile
((αA(Head), αA(Tail)), (αB(Head), αB(Tail))) :=((
1
2, 12
),(1
2, 12
))is a mixed action Nash equilibrium.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 34 / 37
Example: Matching Pennies
The following combinations of supports of mixed actions areconsistent with a mixed Nash equilibrium.
Support of player B ’smixed action
{H} {T} {H,T}
Support of playerA’s mixed action
{H} × × ×{T} × × ×
{H,T} × × ((1
2, 12
),(1
2, 12
))
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 35 / 37
Exercise: Rock Paper Scissors
Player BqR qP qS
Rock Paper Scissors
Player ApR Rock 0,0 -1,1 1,-1pP Paper 1,-1 0,0 -1,1pS Scissors -1,1 1,-1 0,0
QUESTION: Figure out all mixed action Nash equilibria of thevNM strategic game ROCK PAPER SCISSORS by the procedurebased on the indifference principle.
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 36 / 37
Exercise: Rock Paper Scissors
The following combinations of supports of mixed actions mustbe checked.
Support of player B ’s mixed action{R} {P} {S} {R,P} {R,S} {P,S} {R,P,S}
Supp
orto
fpla
yerA
’sm
ixed
acti
on
{R}{P}{S}
{R,P}{R,S}{P,S}
{R,P,S}
Dr. Michael Trost Microeconomics I: Game Theory Lecture 6 37 / 37