Game Theoretic Analysis

Part 1

Ajaz Hussain, Economics, University of Toronto (STG)

 Top 𝑛
Recall: Market Structure Metrics 𝐶𝑅𝑛 = σ𝑖=1 𝑠𝑖 and
𝐻𝐻𝐼 = σ𝑇𝑖=1 𝑠𝑖2

Loose Oligopoly
40 ≤ 𝐶𝑅4 ≤ 60
Competitive Monopoly
𝐶𝑅4 < 40 𝐶𝑅1 ≥ 90
Tight Oligopoly
𝐶𝑅4 > 60

“Un-concentrated” “Moderately Concentrated” “Concentrated”

𝐻𝐻𝐼 ≤ 1,500 1,500 < 𝐻𝐻𝐼 < 2,500 𝐻𝐻𝐼 ≥ 2,500

To be discussed Week 23 and Discussed Week 21 and Week 22

Discussed Week 20 Lecture
Week 24 Lectures Lectures

Game-theoretic situations arise when players’ “payoffs” are determined by a player’s own actions and other
players’ actions
2
ECO 204 Game Theory: Assumptions and Type of Games

Game Theory in ECO204

“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

One-Shot Games
(“Static”)
L Repeated Games
(“Dynamic”)

Games with Simultaneous Moves

t Games with Sequential Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium

“Mixed Strategies”
Players “play” each of their actions with probability ≥ 0
(such that total probability of playing all actions is 1)
ECOII “Pure Strategies”
Players “play” one of their actions with probability 1 and all their
other actions with probability 0
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies

3
 Game Theoretic Analysis
Part 2

Ajaz Hussain, Economics, University of Toronto (STG)

ECO 204 Game Theory: Assumptions and Type of Games

Game Theory in ECO204

“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

One-Shot Games Repeated Games

(“Static”) (“Dynamic”)

Games with Sequential Moves

Games with Simultaneous Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium

“Mixed Strategies” “Pure Strategies”

Players “play” each of their actions with probability ≥ 0 Players “play” one of their actions with probability 1 and all their
(such that total probability of playing all actions is 1) other actions with probability 0
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies

5
Example: Representing 2-Player Simultaneous Move Games
Payoff Matrix of a 2-Player Simultaneous Move Game

Player 1’s actions are across rows

Player 2 can play “Left” Player 2 can play “Right”
Player 2’s actions are down columns

Player 1 can play “Top” $2 $1 $0 $0

Player 1 can play “Bottom” $0 $0 $1 $2

If player 1 plays “bottom” and player 2 plays “left” then

player 1’s payoff is $0 and player 2’s payoff is $0

Game-theoretic situations arise when players’ “payoffs” are determined

by a player’s own actions and other players’ actions
6
 Game Theoretic Analysis
Part 3

Ajaz Hussain, Economics, University of Toronto (STG)

Simultaneous Move Games: Nash Equilibrium in Pure Strategies
Game Theory in ECO204
AUDIO
“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

l
One-Shot Games Repeated Games
(“Static”) (“Dynamic”)

Games with Sequential Moves

Games with Simultaneous Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium
Nightbefore thinking

“Mixed Strategies” “Pure Strategies”

Any pair of mutually
Players “play” each of their actions with probability ≥ 0 Players “play” one of their actions with probability 1 and all their
best responses is a
(such that total probability of playing all actions is 1) other actions with probability 0 reasonable
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies “solution” to the
game (a “Pure
Strategies Nash
“Interesting Case” “Interesting Case”
Equilibrium (or
Multiple Equilibria (multiple NE) Prisoner’s dilemma (NE is sub-optimal)
Equilibria))”.

8
Example 1: Simultaneous Move Game -- Pure Strategies Nash Equilibrium
Player 1’s action set is 𝑈, 𝐷 and Player 2’s action set is {𝑆, 𝐹}

Player 2’s best response to Player 1’s action

thinks thoo
all
Player 1’s best response to Player 2’s action
q O g
Payoff Matrix

Player 2 Plays “Slow” Player 2 Plays “Fast”

o p o Player 1 Plays “Up” $2 $0 $0 $1

Player 1 Plays “Down” $0 $0 $1 $2

pI
This game has a unique pure strategies Nash equilibrium: 𝐷, 𝐹

9
Example: Simultaneous Move Game -- Pure Strategies Nash Equilibrium
Sony’s (Player 1) action set is $299, $399 and Microsoft’s (Player 2) action set is $299, $399

Projected prices and annual sales of Sony Playstation PS3 and Microsoft Xbox 360 Elite in Xmas 2009
PS3 Projected Sales Xbox 36 Elite Projected Sales
PS3 Price Xbox 360 Elite Price
(millions of units) (millions of units)
$299 $299 11.25 11.5
$299 $399 11.75 7.0
$399 $299 8.25 12.5
$399 $399 8.75 8.0

Sony Playstation PS 3 Production Variable Costs per Unit (2009)

Units produced
8.25 8.75 9.25 9.75 10.25 10.75 11.25 11.75 12.25 12.75
(millions)
𝐀𝐕𝐂 306 281 262 240 228 219 212 198 197 200

Microsoft Xbox 360 Elite Production Variable Costs per Unit (2009)
Units produced
7.00 7.50 8.00 8.50 9.00 9.50 10.00 10.50 11.00 11.50 12.00 12.5
(millions)
𝐀𝐕𝐂 308 291 279 257 248 241 235 232 230 219 218 218

gross I p Ave Q
10
Example: Simultaneous Move Game -- Pure Strategies Nash Equilibrium
Sony’s (Player 1) action set is $299, $399 and Microsoft’s (Player 2) action set is $299, $399

Microsoft’s best response to Sony’s action

Sony’s best response to Microsoft’s action

Payoff Matrix of Sony PS 3 vs. Microsoft Xbox 360 Pricing Game

Microsoft charges $399 Microsoft charges $299

Entries are Gross Margins (Xmas 2009)
(“High Price”) (“Low Price”)

Sony charges $399

$1,032.5m $960m $767.25m $1,012m
(“High Price”)

Sony charges $299

$1,186.75m $637m $978.75m $920m
(“Low Price”)

The pure strategies Nash Equilibrium is $299, $299

11
Console Makers Now Have Bigger Game to Play
Sony and Microsoft cool their console price war with the new PlayStation and Xbox, as games and services take on greater importance
Microsoft and SONY have cooled their long-running price war over videogame consoles. But that is because a much bigger
contest looms.

The two companies have dueled over price since 2006. That is when Sony launched its PlayStation 3 for $100 more than
Microsoft was asking for its Xbox 360, which hit the market the year before. That cost Sony some valuable ground, as the
PlayStation 3 ultimately sold a little over half the amount of the previous PlayStation model. Microsoft made the same mistake
in the next cycle, initially charging $100 more for the Xbox One than the PlayStation 4 when both made their debut in 2013. The
Xbox One likewise is estimated to have sold just a little over half the units of its predecessor—and less than half of what the
competing PlayStation 4 has sold to date.

Both companies seem to have tired of this particular fight. The new flagship models of the Xbox Series X and PlayStation 5
coming out this fall will carry the same price for the first time since 2001, when Microsoft’s first Xbox came on the scene to
challenge Sony’s game business. The flagship version of each console will cost $499 when they go on sale in early November.

But ultimately, a war over unit sales becomes less important as the game business evolves. Now entering their third decade of
competition, both Sony and Microsoft have enormous bases of players with established game libraries—a growing portion of
which is digital games run as a service. Such players are less likely to switch over, given their investment. In its last earnings call
in July, Microsoft reported that its Xbox Live membership has hit nearly 100 million players. Notably, Xbox Live membership has
more than tripled during the lifespan of the Xbox One, even given that console’s relatively poor performance.

The success of Xbox Live reflects Microsoft’s larger corporate goal of driving use of its software and cloud services. The
company even sells PlayStation versions of Minecraft—the popular world-building game it spent $2.5 billion on in 2014. Sony, by
contrast, is far more dependent on its games business, which made up more than a quarter of the company’s revenue for the
trailing 12-month period ended in June, compared with about 8% for Microsoft for the same period.

Microsoft’s overall cloud strategy has garnered the company a market value of more than $1.5 trillion—up nearly fivefold since
the start of the most recent console cycle in late 2013 and more than 16 times that of Sony’s. PlayStation may keep winning the
console war, but Microsoft will be very hard to catch in the much bigger game.
Source: Wall Street Journal, September 18th, 2020 12
Example: Why is NE is an “Equilibrium”?
Sony’s (Player 1) action set is $299, $399 and Microsoft’s (Player 2) action set is $299, $399

Microsoft’s best response to Sony’s action

Sony’s best response to Microsoft’s action

Payoff Matrix of Sony PS 3 vs. Microsoft Xbox 360 Pricing Game

Microsoft charges $399 Microsoft charges $299

Entries are Gross Margins (Xmas 2009)
(“High Price”) (“Low Price”)

Sony charges $399

$1,032.5m $960m $767.25m $1,012m
(“High Price”) e

Sony charges $299

$1,186.75m $637m $978.75m $920m
(“Low Price”)

Always agave back to N E

The pure strategies Nash Equilibrium is $299, $299

13
Simultaneous Move Games: Nash Equilibrium in Pure Strategies
Game Theory in ECO204
“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

One-Shot Games Repeated Games

(“Static”) (“Dynamic”)

Games with Sequential Moves

Games with Simultaneous Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium

“Mixed Strategies” “Pure Strategies”

Any pair of mutually
Players “play” each of their actions with probability ≥ 0 Players “play” one of their actions with probability 1 and all their
best responses is a
(such that total probability of playing all actions is 1) other actions with probability 0 reasonable
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies “solution” to the
game (a “Pure
Strategies Nash
“Interesting Case” “Interesting Case”
Equilibrium (or
Multiple Equilibria (multiple NE) Prisoner’s dilemma (NE is sub-optimal)
Equilibria))”.

14
Example: Simultaneous Move Game -- Pure Strategies Nash Equilibria
Player 1’s action set is 𝑇, 𝐵 and Player 2’s action set is {𝐿, 𝑅}

Player 2’s best response to Player 1’s action

Player 1’s best response to Player 2’s action

Payoff Matrix

Player 2 Plays “Left” Player 2 Plays “Right”

Player 1 Plays “Top” $2 $1 $0 $0

Player 1 Plays “Bottom” $0 $0 $1 $2

This game has multiple pure strategies Nash equilibria: 𝑇, 𝐿 , 𝐵, 𝑅

15
Example: Simultaneous Move Game -- Pure Strategies Nash Equilibria
TV networks (Player 1) action set is 𝐻𝐷𝑇𝑉, 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑇𝑉 Microsoft & Fox Consortium
Microsoft’s & Fox (Player 2) action set is 𝐻𝐷𝑇𝑉, 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑇𝑉
Backed Standard Definition
Adopt Standard
Adopt HDTV Format
Definition Format

Adopt HDTV Format 100 50 30 20

Oo
TV Networks
Backed HDTV
Standard Adopt Standard
0 0 60 90
Definition Format

Based on figures from “Ready or Not: Here Comes HDTV (New York Times)”

Unsurprisingly, “Tech-Standard” games tend to have multiple pure strategies Nash equilibria

16
Simultaneous Move Game Pure Strategies: Prisoner’s Dilemma (NE suboptimal)

Game Theory in ECO204

“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

One-Shot Games Repeated Games

(“Static”) (“Dynamic”)

Games with Sequential Moves

Games with Simultaneous Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium

“Mixed Strategies” “Pure Strategies”

Any pair of mutually
Players “play” each of their actions with probability ≥ 0 Players “play” one of their actions with probability 1 and all their
best responses is a
(such that total probability of playing all actions is 1) other actions with probability 0 reasonable
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies “solution” to the
game (a “Pure
Strategies Nash
“Interesting Case” “Interesting Case”
Equilibrium (or
Multiple Equilibria (multiple NE) Prisoner’s dilemma (NE is sub-optimal)
Equilibria))”.

17
Example: Simultaneous Move Game Pure Strategies: Prisoner’s Dilemma
(NE suboptimal)

Superbowl Ads
Payoffs are $ billions profits 8 weeks after Super bowl
on
egg p

9014
Coke No Ad Coke Ad

Pepsi No Ad 1.182 1.647 1.060 1.647

Pepsi Ad 1.525 1.209 1.103

too 1.568

Source: Table 6 in Superbowl Ad, Wesley R. Hartmann and Daniel

Klapper, Stanford Business School Working Paper No. 2139, March 2015
18
Example: Simultaneous Move Game Pure Strategies: Real-Life Prisoner’s
Dilemma (NE suboptimal)
Tanya

Nancy

Negative payoffs represent years in prison (which is a very bad place)

Suspect # 2 Suspect # 2
Better than Nash Equilibrium
“Doesn’t Confesses” “Confesses”
Suspect # 1
−1 −1 −10 0
“Doesn’t Confess”
Suspect # 1
0 −10 −6 −6
“Confesses”

Nash Equilibrium

19
 Game Theoretic Analysis
Part 4

Ajaz Hussain, Economics, University of Toronto (STG)

Simultaneous Move Games: Nash Equilibrium in Mixed Strategies
Game Theory in ECO204
“Two rational players” with “common knowledge” of all players’ set of actions
“Rational Players”: play to maximize their payoff (“never hurt themselves”)
“Common Knowledge”: Everybody Knows ..

One-Shot Games Repeated Games

(“Static”) (“Dynamic”)

Games with Sequential Moves

Games with Simultaneous Moves
“Follower” moves after observing “Leader’s” move
Players “move” without knowing rivals’ actions
“Solution Technique”: Sub-Perfect Nash Equilibrium

“Mixed Strategies” “Pure Strategies”

Players “play” each of their actions with probability ≥ 0 Players “play” one of their actions with probability 1 and all their
(such that total probability of playing all actions is 1) other actions with probability 0
“Solution Technique”: Nash Equilibrium in Mixed Strategies “Solution Technique”: Nash Equilibrium in Pure Strategies

Mutual “beliefs” is a reasonable “solution” to the game (a “Mixed Strategies Nash Equilibrium).
At play, the players will in fact “play” their strategies with Mixed strategies NE probabilities

See “Mixed Strategies Nash Equilibrium” Calculator

21
Example: “Pure” vs. “Mixed” Strategies

C C
100% 0% 0% 10% 0% 90%

Example: Pure Strategies Example: Mixed Strategies

Play 𝑳, 𝑪, 𝑹 with probabilities {𝟏, 𝟎, 𝟎} Play 𝑳, 𝑪, 𝑹 with probabilities {𝟎. 𝟏, 𝟎, 𝟎. 𝟗}

Playing Mixed strategies amounts to choosing non-zero probabilities for at least two actions
Playing Pure strategies amounts to choosing non-zero probability for exactly one action
22
“Mixed” Strategies in Penalty Shoot Outs

Penalty shoot-outs are basically still crap-shoots. Economist, June 29th, 2018
23
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium
𝑎+𝑏=1 Player 2 Plays “Left” Player 2 Plays “Right”
𝑝+𝑞 =1 with probability 𝒑 with probability 𝒒
Player 1 Plays “Top” with probability 𝒂 $2 $1 $0 $0
Player 1 Plays “Bottom” with probability 𝒃 $0 $0 $1 $2

We already know this game has multiple pure strategies Nash equilibria: 𝑇, 𝐿 , 𝐵, 𝑅 or 𝑎 = 1, 𝑝 = 1 , {𝑏 = 1, 𝑞 = 1}

max 𝐸𝑉 Player 1 = 𝑎 𝐸𝑉 Player 1 plays top + 𝑏 𝐸𝑉 Player 1 plays bottom 𝑠. 𝑡. 𝑎 + 𝑏 = 1, 𝑎 ≥ 0, 𝑏 ≥ 0

𝑎,𝑏

max L = 𝑎 𝐸𝑉 Player 1 top + 𝑏 𝐸𝑉 Player 1 bottom − 𝜆1 𝑎 + 𝑏 − 1 + 𝜆2 𝑎 + 𝜆3 𝑏

𝑎,𝑏,𝜆1 ,𝜆2 ,𝜆3

max L =𝑎 2𝑝 + 0𝑞 +𝑏 0𝑝 + 1𝑞 − 𝜆1 𝑎 + 𝑏 − 1 + 𝜆2 𝑎 + 𝜆3 𝑏
𝑎,𝑏,𝜆1 ,𝜆2 ,𝜆3
𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑜𝑓 𝑎,𝑏 𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑜𝑓 𝑎,𝑏

𝜕L
= 𝐸𝑉 Player 1 top − 𝜆1 + 𝜆2 = 0
𝜕𝑎

𝜕L
= 𝐸𝑉 Player 1 bottom − 𝜆1 + 𝜆3 = 0
𝜕𝑏

𝜕L
=− 𝑎+𝑏−1 =0⇒𝑎+𝑏 =1
𝜕𝜆1

𝜆2 ≥ 0, 𝑎 ≥ 0, 𝜆2 𝑎 = 0
See “Mixed Strategies Nash Equilibrium” Calculator 𝜆3 ≥ 0, 𝑏 ≥ 0, 𝜆3 𝑏 = 0 24
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium

𝜆3 ≥ 0, 𝑏 ≥ 0, 𝜆3 𝑏 = 0

𝑏=0 𝜆3 = 0
If so, then 𝜆3 ≥ 0 If so, then 𝑏 ≥ 0

𝜆2 ≥ 0, 𝑎 ≥ 0, 𝜆2 𝑎 = 0 𝜆2 ≥ 0, 𝑎 ≥ 0, 𝜆2 𝑎 = 0

𝑎=0 𝜆2 = 0 𝑎=0 𝜆2 = 0
If so, then 𝜆2 ≥ 0 If so, then 𝑎 ≥ 0 If so, then 𝜆2 ≥ 0 If so, then 𝑎 ≥ 0
Case A Case B Case C Case D
𝑎=0 If so, then 𝑎 ≥ 0 ⇒ 𝑎 = 1 𝑎=0 If so, then 𝑎 ≥ 0
𝑏=0 𝑏=0 If so, then 𝑏 ≥ 0 ⇒ 𝑏 = 1 If so, then 𝑏 ≥ 0
If so, then 𝜆2 ≥ 0 𝜆2 = 0 If so, then 𝜆2 ≥ 0 𝜆2 = 0
If so, then 𝜆3 ≥ 0 If so, then 𝜆3 ≥ 0 𝜆3 = 0 𝜆3 = 0

We already know this game has multiple pure strategies Nash equilibria: 𝑇, 𝐿 , 𝐵, 𝑅 or 𝑎 = 1, 𝑝 = 1 , {𝑏 = 1, 𝑞 = 1}
Impossible Pure strategy Nash Equilibrium {𝑇, 𝐿} Pure strategy Nash Equilibrium {𝐵, 𝑅} Mixed Strategies requires 𝑎, 𝑏 > 0 (why?)

plus top 10000 play bot 10090

25
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium
Case D
When is/is 𝑎 ≥ 0?
𝑎+𝑏=1 Player 2 Plays “Left” Player 2 Plays “Right”
When is/is 𝑏 ≥ 0? 𝑝+𝑞 =1 with probability 𝒑 with probability 𝒒
𝜆2 = 0 Player 1 Plays “Top” with probability 𝒂 $2 $1 $0 $0
𝜆3 = 0
Player 1 Plays “Bottom” with probability 𝒃 $0 $0 $1 $2
Mixed Strategies requires 𝑎, 𝑏 > 0 (why?)

𝜕L
= 𝐸𝑉 Player 1 top − 𝜆1 + 𝜆2 = 0 ⇒ 𝜆1 = 𝐸𝑉 Player 1 top
key 𝜕𝑎

𝜕L

as
= 𝐸𝑉 Player 1 bottom − 𝜆1 + 𝜆3 = 0 ⇒ 𝜆1 = 𝐸𝑉 Player 1 bottom
𝜕𝑏

“Trick” for calculating Mixed

Strategies Equilibrium
𝐸𝑉 Player 1 top = 𝐸𝑉 Player 1 bottom EV top p 2 po
2𝑝 = 𝑞
ZP
t EV bottom
p0 t q1
Use the fact that 𝑝 + 𝑞 = 1 ⇒ 2𝑝 = 1 − 𝑝 a
1 2
𝑃𝑙𝑎𝑦𝑒𝑟 1 𝐵𝑒𝑙𝑖𝑒𝑣𝑒𝑠 𝑡ℎ𝑎𝑡 𝑃𝑙𝑎𝑦𝑒𝑟 2 𝑤𝑖𝑙𝑙 𝑝𝑙𝑎𝑦 𝐿𝑒𝑓𝑡, 𝑅𝑖𝑔ℎ𝑡 𝑎𝑠 𝑀𝑖𝑥𝑒𝑑 𝑆𝑡𝑟𝑎𝑡𝑒𝑔𝑖𝑒𝑠 𝑤𝑖𝑡ℎ 𝑝 = , 𝑞 =
3 3
26
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium
𝑎+𝑏=1 Player 2 Plays “Left” Player 2 Plays “Right”

DO
𝑝+𝑞 =1 with probability 𝒑 with probability 𝒒
Player 1 Plays “Top” with probability 𝒂 $2 $1 $0 $0
Player 1 Plays “Bottom” with probability 𝒃 $0 $0 $1 $2

By a similar argument for Player 2:

“Trick” for calculating Mixed
Strategies Equilibrium
𝐸𝑉 Player 2 Left = 𝐸𝑉 Player 2 Right Key
⇒ 1𝑎 + 0𝑏 = 0𝑎 + 2𝑏

Use the fact that 𝑎 + 𝑏 = 1

2 1
𝑃𝑙𝑎𝑦𝑒𝑟 2 𝐵𝑒𝑙𝑖𝑒𝑣𝑒𝑠 𝑡ℎ𝑎𝑡 𝑃𝑙𝑎𝑦𝑒𝑟 1 𝑤𝑖𝑙𝑙 𝑝𝑙𝑎𝑦 𝑇𝑜𝑝, 𝐵𝑜𝑡𝑡𝑜𝑚 𝑎𝑠 𝑀𝑖𝑥𝑒𝑑 𝑆𝑡𝑟𝑎𝑡𝑒𝑔𝑖𝑒𝑠 𝑤𝑖𝑡ℎ 𝑎 = ,𝑏 =
3 3
This game has the following three Nash equilibria (two pure, one mixed):

2 1 1 2
Play 𝑇, 𝐵 and 𝐿, 𝑅 with: 𝑎 = 1, 𝑏 = 0 , 𝑝 = 1, 𝑞 = 0 , 𝑎 = 0, 𝑏 = 1 , 𝑝 = 0, 𝑞 = 1 , 𝑎 = ,𝑏 = , 𝑝 = ,𝑞 =
3 3 3 3

See “Mixed Strategies Nash Equilibrium” Calculator

27
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium
𝑎+𝑏=1 Player 2 Plays “Left” Player 2 Plays “Right”
𝑝+𝑞 =1 with probability 𝒑 with probability 𝒒
Player 1 Plays “Top” with probability 𝒂 $2 $1 $0 $0
Player 1 Plays “Bottom” with probability 𝒃 $0 $0 $1 $2

We already know this game has multiple pure strategies Nash equilibria: 𝑇, 𝐿 , 𝐵, 𝑅 or 𝑎 = 1, 𝑝 = 1 , {𝑏 = 1, 𝑞 = 1}

28
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium

29
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium

30
Example 1: A Game with Pure and Mixed Strategies Nash Equilibrium

31
Example 2: A Game with Pure (But No Mixed) Strategies Nash Equilibrium

Payoff Matrix of Sony PS 3 vs. Microsoft Xbox 360 Pricing Game

Microsoft charges $399 (“High Price”) Microsoft charges $299 (“Low Price”)
Entries are Gross Margins (Xmas 2009)
With probability 𝒑 With probability 𝒒
Sony charges $399 (“High Price”)
$1,032.5m $960m $767.25m $1,012m
With probability 𝒂
Sony charges $299 (“Low Price”)
$1,186.75m $637m $978.75m $920m
With probability 𝒃

If there is a Mixed Strategies NE, then:

“Trick” for calculating Mixed
Strategies Equilibrium 𝐸𝑉 Sony charges $399 = 𝐸𝑉 Sony charges $299 key
𝑝 1,032.50 + 𝑞 767.24 = 𝑝 1,186.75 + 𝑞 978.75 between
No prob
𝑝+𝑞 =1
O 1
Confirm there is no solution to the equations above
S1 1
If there is a Mixed Strategies NE, then:
“Trick” for calculating Mixed
Strategies Equilibrium 𝐸𝑉 Microsoft charges $399 = 𝐸𝑉 Microsoft charges $299 key
𝑎 960 + 𝑏 637 = 𝑎 1012 + 𝑏 920

𝑎+𝑏=1

Confirm there is no solution to the equations above.

32
Example 3: A Game with a Mixed (But No Pure) Strategies Nash Equilibrium

Paper ≻ Rock ≻ Scissors ≻ Paper (note: intransitive!)

“If the two players say the same word”  “Draw”

“If the players say different Words”  “Winner” receives $1 from “Loser” (zero sum game)
Payoff Matrix (All payoffs in dollars). Best Responses underlined. TES
This is an example of a 0
Player # 2 says “Rock” 100Yo
Player # 2 says “Paper”
0
Player # 2 says AYE
zero-sum game “Scissors”
0 43 3 Es O plaber I
Player # 1 says “Rock” 0 0 −1 +1 +1 −1
Infinitely lose
Player # 1 says “Paper” +1 −1 0 0 −1 +1 if he has
0 o n States's

Player # 1 says “Scissors” −1 +1 +1 −1 0 0

0 t
Intuitively, this game can’t have a pure strategies equilibrium (why not?)
Look for a Mixed Strategies Equilibrium
See “Mixed Strategies Nash Equilibrium” Calculator
33
Example 3: A Game with Mixed (But No Pure) Strategies Nash Equilibrium

Player # 2 says “Rock” Player # 2 says “Paper” Player # 2 says “Scissors”

with probability = 𝒑 with probability = 𝒒 with probability = 𝒓
Player # 1 says “Rock”
0 0 −1 +1 +1 −1
with probability = 𝒂
Player # 1 says “Paper”
+1 −1 0 0 −1 +1
with probability = 𝒃
Player # 1 says “Scissors”
−1 +1 +1 −1 0 0
with probability = 𝒄

If there is a Mixed Strategies NE, then:

“Trick” for calculating Mixed

Strategies Equilibrium
𝐸𝑉Player 1 Rock = 𝐸𝑉Player 1 Paper = 𝐸𝑉Player 1 Scissors

𝑝 0 + 𝑞 −1 + 𝑟 1 = 𝑝 1 + 𝑞 0 + 𝑟 −1 = 𝑝 −1 + 𝑞 1 + 𝑟 0
I Inewas

𝑝+𝑞+𝑟 =1

1
⇒ 𝑝∗ = 𝑞 ∗ = 𝑟 ∗ =
3
See “Mixed Strategies Nash Equilibrium” Calculator
34
Example 3: A Game with Mixed (But No Pure) Strategies Nash Equilibrium

Player # 2 says “Rock” Player # 2 says “Paper” Player # 2 says “Scissors”

with probability = 𝒑 with probability = 𝒒 with probability = 𝒓
Player # 1 says “Rock”
0 0 −1 +1 +1 −1
with probability = 𝒂
Player # 1 says “Paper”
+1 −1 0 0 −1 +1
with probability = 𝒃
Player # 1 says “Scissors”
−1 +1 +1 −1 0 0
with probability = 𝒄

If there is a Mixed Strategies NE, then:

“Trick” for calculating Mixed
Strategies Equilibrium 𝐸𝑉Player 2 Rock = 𝐸𝑉Player 2 Paper = 𝐸𝑉Player 2 Scissors

𝑎 0 + 𝑏 −1 + 𝑐 1 = 𝑎 1 + 𝑏 0 + 𝑐 −1 = 𝑎 −1 + 𝑏 1 + 𝑐 0 = 0

𝑎+𝑏+𝑐 =1

1
⇒ 𝑎∗ = 𝑏 ∗ = 𝑐 ∗ =
3
See “Mixed Strategies Nash Equilibrium” Calculator
35
 pea by
MR a
zbay
your actions as siren
I take
my
compute reaction

Game Theoretic Analysis: Oligopolies

Part 5

Ajaz Hussain, Economics, University of Toronto (STG)

Oligopoly Models
Models of Oligopolistic Rivalry/Cooperation

Cournot Rivalry Stackelberg Rivalry

𝑁 ≥ 2 Firms compete in a one-shot game by choosing 𝑁 = 2 Firms compete in a one-shot game over two periods in
outputs simultaneously to maximize their respective profits which the leader-follower choose their outputs sequentially
to maximize their respective profits
Examples:
∎ Samsung and LG ∎ Boeing and Airbus Examples:
∎ European car makers (except Volkswagen) ∎ Apple and Kindle ∎ Amazon-Whole Foods vs. Grocers

Bertrand Rivalry Cartel Cooperate

𝑁 = 2 Firms compete in a one-shot game by choosing prices 𝑁 ≥ 2 Firms collude by choosing outputs simultaneously to
simultaneously to maximize their respective profits maximize the joint cartel profit

Examples: Examples:
∎ Coke & Pepsi ∎ Japanese & American car makers, and VW ∎ OPEC ∎ Saline ∎ Finnish Cartels

37
Ahead: One-Shot, Pure Strategy, Oligopoly Models
There are 𝑵 ≥ 𝟐 “symmetric” profit-maximizing firms with ample capacity and constant 𝑴𝑪 = 𝑨𝑽𝑪
producing/offering a “homogeneous” product/service with a Linear Market Demand Curve:
Marketoutput firmoutput
𝑃 = Market Price = 𝑎 − 𝑏 Market Output = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏 𝑞1 + ⋯ + 𝑞𝑖 + ⋯ + 𝑞𝑗 + ⋯ + 𝑞𝑁

Now, imagine the following sequence of “events” – we are going to complete the “Table” below
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
Monopoly
Cournot
“Competition”
Cartel
Stackelberg
Bertrand
38
Benchmark: Monopolist
Suppose there is only one firm behaving as a monopolist (𝑵 = 𝟏) with
ample capacity and constant 𝑴𝑪
Benchmark: “Naïve” Monopoly

I
𝑃 = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏𝑞t
Cournot Rivalry

Cournot Rivalry with 𝑁 → ∞: “appears to be perfect

competition”

Collusion in a Cartel

Monopolist’s optimal output is:

Stackelberg Rivalry

Bertrand Rivalry
𝑀𝑅 = 𝑀𝐶

𝑎 − 2𝑏𝑞 = 𝑀𝐶

𝑎 − 𝑀𝐶
𝑞 𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦 =
2𝑏
𝑎 + 𝑀𝐶
𝑃𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦 =
2

39
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶
Monopoly 1
2𝑏 2𝑏 2
Cournot

“Competition”

Cartel

Stackelberg

Bertrand

40
Cournot Rivalry (All firms compete by choosing outputs simultaneously)
Firm 𝒊 = 𝟏, 𝟐, . . , 𝑵 chooses its profit maximizing output by holding all other firms’ output as “given” (unknown
Benchmark: “Naïve” Monopoly numerical value)

Cournot Rivalry 𝑁−1

𝑃 = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏 𝑞1 + ⋯ + 𝑞𝑖 + ⋯ + 𝑞𝑗 + ⋯ + 𝑞𝑁 − 𝑏𝑞𝑖 = 𝑎 − 𝑏 ෍ 𝑞𝑗 − 𝑏𝑞𝑖
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
competition” a 𝐹𝑟𝑜𝑚 𝑓𝑖𝑟𝑚 𝑖 ′ 𝑠 𝑝𝑒𝑟𝑠𝑝𝑒𝑐𝑡𝑖𝑣𝑒 𝑡ℎ𝑖𝑠 𝑖𝑠 𝑎 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑗≠𝑖

Collusion in a Cartel
intercept Cournot Firm 𝑖 ′ 𝑠 optimal output is:
MR same
Stackelberg Rivalry
a
𝑀𝑅𝑖 = 𝑀𝐶
twice slope of
one
Bertrand Rivalry

𝑁−1

𝑎 − 𝑏 ෍ 𝑞𝑗 − 2𝑏𝑞𝑖 = 𝑀𝐶
𝑗≠𝑖

𝑎 − 𝑏 σ𝑁−1
𝑗≠𝑖 𝑞𝑗 + 𝑀𝐶 𝑎 − 𝑏 𝑁 − 1 𝑞𝑖𝑐𝑜𝑢𝑟𝑛𝑜𝑡 + 𝑀𝐶 2 𝑎 − 𝑀𝐶 2
𝑞𝑖 = 𝑞𝑖𝑐𝑜𝑢𝑟𝑛𝑜𝑡 = = = = 𝑞𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦
2𝑏 2𝑏 𝑁 + 1 2𝑏 𝑁+1
Firm 𝑖 ′ 𝑠 reaction function Using symmetry

𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
⇒ 𝑄 𝐶𝑜𝑢𝑟𝑛𝑜𝑡 = 𝑁𝑞𝑖𝑐𝑜𝑢𝑟𝑛𝑜𝑡 = ⇒ 𝑃𝑐𝑜𝑢𝑟𝑛𝑜𝑡 =
𝑁+1 𝑏 𝑁+1
41
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶
Monopoly 1
2𝑏 2𝑏 2
2 𝑎 − 𝑀𝐶 𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
Cournot ≥1
𝑁 + 1 2𝑏 𝑁+1 𝑏 𝑁+1
“Competition”

Cartel

Stackelberg

Bertrand

42
“Cournot” Rivalry (𝑁 → ∞ firms compete by choosing outputs simultaneously)
Firm 𝒊 = 𝟏, 𝟐, . . , 𝑵 chooses its profit maximizing output by holding all other firms’ output as “given” (unknown
Benchmark: “Naïve” Monopoly numerical value)

Cournot Rivalry 𝑁−1

𝑃 = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏 𝑞1 + ⋯ + 𝑞𝑖 + ⋯ + 𝑞𝑗 + ⋯ + 𝑞𝑁 − 𝑏𝑞𝑖 = 𝑎 − 𝑏 ෍ 𝑞𝑗 − 𝑏𝑞𝑖
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
competition”
𝐹𝑟𝑜𝑚 𝑓𝑖𝑟𝑚 𝑖 ′ 𝑠 𝑝𝑒𝑟𝑠𝑝𝑒𝑐𝑡𝑖𝑣𝑒 𝑡ℎ𝑖𝑠 𝑖𝑠 𝑎 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑗≠𝑖

Collusion in a Cartel

Stackelberg Rivalry 𝑵 → ∞ Cournot firms choosing outputs simultaneously:

Bertrand Rivalry

1 𝑎 − 𝑀𝐶

u
𝑞𝐶𝑜𝑢𝑟𝑛𝑜𝑡 𝑤𝑖𝑡ℎ 𝐼𝑛𝑓𝑖𝑛𝑖𝑡𝑒 𝑅𝑖𝑣𝑎𝑙𝑠 = lim 𝑞𝑖𝑐𝑜𝑢𝑟𝑛𝑜𝑡 = lim
n
𝑁 →∞ 𝑁 →∞ 𝑁 + 1 𝑏
→0
O
“aka Atomistic Competition”

N700 Finite bomb

at 70 𝑁 𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶

a p Qi
𝑄 𝐶𝑜𝑢𝑟𝑛𝑜𝑡 𝑤𝑖𝑡ℎ 𝐼𝑛𝑓𝑖𝑛𝑖𝑡𝑒 𝑅𝑖𝑣𝑎𝑙𝑠 = lim 𝑄 𝑐𝑜𝑢𝑟𝑛𝑜𝑡 = lim
𝑁 →∞ 𝑁 →∞ 𝑁 + 1 𝑏
≈

B
𝑏
M
00 0 𝑁
𝑃𝐶𝑜𝑢𝑟𝑛𝑜𝑡 𝑤𝑖𝑡ℎ 𝐼𝑛𝑓𝑖𝑛𝑖𝑡𝑒 𝑅𝑖𝑣𝑎𝑙𝑠 = lim 𝑃𝑐𝑜𝑢𝑟𝑛𝑜𝑡 = lim 𝑎 − 𝑎 − 𝑀𝐶 = 𝑀𝐶
𝑁 →∞ 𝑁 →∞ 𝑁+1

𝑃 = 𝑀𝐶 “looks like” (but isn’t) “perfect competition”

43
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶
Monopoly 1
2𝑏 2𝑏 2
2 𝑎 − 𝑀𝐶 𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
Cournot ≥1
𝑁 + 1 2𝑏 𝑁+1 𝑏 𝑁+1
𝑎 − 𝑀𝐶
“Competition” →∞ 0 𝑀𝐶
𝑏
Cartel

Stackelberg

Bertrand

44
Cartel Collusion
to
(All firms cooperate by choosing outputs simultaneously)
Firm 𝒊 = 𝟏, 𝟐, . . , 𝑵 chooses its output to maximize cartel profit by holding all other firms’ output as “given”
(unknown numerical value)
Benchmark: “Naïve” Monopoly

max Gross Π𝐶𝑎𝑟𝑡𝑒𝑙 = 𝑅𝐶𝑎𝑟𝑡𝑒𝑙 − 𝑇𝑉𝐶𝐶𝑎𝑟𝑡𝑒𝑙

Cournot Rivalry 𝑞1 ,..,𝑞𝑁

Cournot Rivalry with 𝑁 → ∞: “appears to be perfect

competition”
Each cartel member produces at 𝑀𝑅𝑖 = 𝑀𝐶 . But a cartel member’s 𝑀𝑅 is tricky:
Collusion in a Cartel

𝑁 𝑁
Stackelberg Rivalry
𝑅𝐶𝑎𝑟𝑡𝑒𝑙 = 𝑃 × 𝑄 = 𝑎 − 𝑏𝑄 × 𝑄 = 𝑎 − 𝑏 ෍ 𝑞𝑖 × ෍ 𝑞𝑖 = 𝑃 𝑄 𝑞𝑖 × 𝑄 𝑞𝑖
w u 4 𝑖=1 𝑖=1

go
Bertrand Rivalry

f y
𝑑𝑅𝐶𝑎𝑟𝑡𝑒𝑙 𝑑𝑃 𝑄 𝑞𝑖 × 𝑄 𝑞𝑖 𝑑𝑃 𝑄 𝑞𝑖 𝑑𝑄 𝑞𝑖 𝑑𝑄 𝑞𝑖
𝑀𝑅𝑖 = = = 𝑄 𝑞𝑖 + 𝑃 𝑄 𝑞𝑖 = −𝑏 1 𝑄 + 𝑎 − 𝑏𝑄 1
𝑑𝑞𝑖 𝑑𝑞𝑖 𝑑𝑄 𝑑𝑞𝑖 𝑑𝑞𝑖
Product and Chain Rules

𝑁 𝑁−1

𝑀𝑅𝑖 = 𝑎 − 2𝑏𝑄 = 𝑎 − 2𝑏 ෍ 𝑞𝑖 = 𝑎 − 2𝑏 ෍ 𝑞𝑗 − 2𝑏𝑞𝑖

𝑖 𝑗≠𝑖
45
Cartel Collusion
(All firms cooperate by choosing outputs simultaneously)
Each cartel member produces at 𝑀𝑅𝑖 = 𝑀𝐶
Benchmark: “Naïve” Monopoly
𝑁−1
Cournot Rivalry 𝑀𝑅𝑖 = 𝑎 − 2𝑏 ෍ 𝑞𝑗 − 2𝑏𝑞𝑖 = 𝑀𝐶
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
𝑗≠𝑖
competition”

Collusion in a Cartel 𝑎 − 2 σ𝑁−1

𝑗≠𝑖 𝑏𝑞𝑗 + 𝑀𝐶 𝑎 − 2𝑏 𝑁 − 1 𝑞𝑖𝑐𝑎𝑟𝑡𝑒𝑙 + 𝑀𝐶
𝑞𝑖𝑐𝑎𝑟𝑡𝑒𝑙 = =
Stackelberg Rivalry 2𝑏 2𝑏
Cartel Member 𝑖 ′ 𝑠 reaction function Using symmetry
Bertrand Rivalry

1 𝑎 − 𝑀𝐶 1 𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦
𝑞𝑖𝑐𝑎𝑟𝑡𝑒𝑙 = = 𝑞𝑖
𝑁 2𝑏 𝑁

𝑎 − 𝑀𝐶
𝑄 𝑐𝑎𝑟𝑡𝑒𝑙 = 𝑁𝑞𝑖𝑐𝑎𝑟𝑡𝑒𝑙 = = 𝑄 𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦
2𝑏

𝑎 + 𝑀𝐶
𝑃𝑐𝑎𝑟𝑡𝑒𝑙 = 𝑎 − 𝑏𝑄 𝑐𝑎𝑟𝑡𝑒𝑙 = = 𝑃𝑚𝑜𝑛𝑜𝑝𝑜𝑙𝑦
2
46
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶

7
Monopoly 1
2𝑏 2𝑏 2
2 𝑎 − 𝑀𝐶 𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
Cournot ≥1
𝑁 + 1 2𝑏 𝑁+1 𝑏 𝑁+1
𝑎 − 𝑀𝐶
“Competition” →∞ 0 𝑀𝐶
𝑏

D I
1 𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶

D
Cartel >1
𝑁 2𝑏 2𝑏 2
Stackelberg

Bertrand

47
Stackelberg Rivalry
(Two Firms compete by choosing outputs sequentially)
For simplicity, analyze Stackelberg Model with TWO firms

Firm 2 (Follower) observes 𝑞1 and then

Benchmark: “Naïve” Monopoly

Firm 1 (Leader) chooses output 𝑞1 chooses its output 𝑞2

Cournot Rivalry

Cournot Rivalry with 𝑁 → ∞: “appears to be perfect

competition”
As in decision making under uncertainty, solve this problem backwards: given
Collusion in a Cartel
“leader’s output” 𝑞1 , the “follower”, firm 2, will choose 𝑞2 by setting:
Stackelberg Rivalry
𝑀𝑅𝑓𝑜𝑙𝑙𝑜𝑤𝑒𝑟 = 𝑀𝐶
Bertrand Rivalry

Starting backwards from the follower’s perspective:

NI 𝑃 = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏 𝑞1 + 𝑞2 = 𝑎 − 𝑏𝑞1 − 𝑏𝑞2
From
follower′ s
perspective this
is a constant

AYY 𝑀𝑅𝑓𝑜𝑙𝑙𝑜𝑤𝑒𝑟 = 𝑀𝑅2 = 𝑎 − 𝑏𝑞1 − 2𝑏𝑞2 = 𝑀𝐶

48
Stackelberg Rivalry
(Two Firms compete by choosing outputs sequentially)
Firm 2 (Follower) observes 𝑞1 and then
Firm 1 (Leader) chooses output 𝑞1 chooses its output 𝑞2
Benchmark: “Naïve” Monopoly

Cournot Rivalry
Stackelberg “follower” chooses its output by setting:

𝑀𝑅2 = 𝑎 − 𝑏𝑞1 − 2𝑏𝑞2 = 𝑀𝐶

Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
competition”

Collusion in a Cartel

𝑆𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑎 − 𝑏𝑞1 + 𝑀𝐶
𝑞2 =
Stackelberg Rivalry
2𝑏
Bertrand Rivalry
Stackelberg Follower′ s reaction function

Contrast this with 𝑁 = 2 Cournot Reaction Function:

D
𝑎 − 𝑏𝑞1 + 𝑀𝐶
𝑞𝑖𝑐𝑜𝑢𝑟𝑛𝑜𝑡 =
2𝑏
Cournot Firm 𝑖 ′ 𝑠 reaction function

Why is the Follower’s Stackelberg Reaction Function Identical to the Cournot Reaction
Function?
49
Stackelberg Rivalry
(Two Firms compete by choosing outputs sequentially)
Firm 2 (Follower) observes 𝑞1 and then
Firm 1 (Leader) chooses output 𝑞1 chooses its output 𝑞2
Benchmark: “Naïve” Monopoly

Cournot Rivalry

The Stackelberg leader knows that the follower’s output depends on the leader’s output”:
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
competition”

𝑆𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑎 − 𝑏𝑞1 + 𝑀𝐶
Collusion in a Cartel 𝑞2 =
2𝑏
Stackelberg Rivalry Stackelberg Follower′ s reaction function

Bertrand Rivalry The Stackelberg Leader chooses its output by (from the Leader’ perspective):

𝑎 − 𝑏𝑞1 + 𝑀𝐶 𝑎 + 𝑀𝐶 𝑏
𝑃 = 𝑎 − 𝑏𝑄 = 𝑎 − 𝑏𝑞1 − 𝑏𝑞2 = 𝑎 − 𝑏𝑞1 − 𝑏 = − 𝑞1
2𝑏 2 2

𝑀𝑅𝑙𝑒𝑎𝑑𝑒𝑟 = 𝑀𝐶

𝑎 + 𝑀𝐶 𝑏
− 2 𝑞1 = 𝑀𝐶
2 2
50
Stackelberg Rivalry
(Two Firms compete by choosing outputs sequentially)
Firm 2 (Follower) observes 𝑞1 and then
Firm 1 (Leader) chooses output 𝑞1 chooses its output 𝑞2
Benchmark: “Naïve” Monopoly

𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑎 − 𝑀𝐶
Cournot Rivalry
𝑞𝑙𝑒𝑎𝑑𝑒𝑟 = 𝑞1 =
2𝑏
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect
competition”

Collusion in a Cartel
𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔
𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑎 − 𝑏𝑞1 + 𝑀𝐶 𝑎 − 𝑀𝐶
𝑞𝑓𝑜𝑙𝑙𝑜𝑤𝑒𝑟 = 𝑞2 = =
2𝑏 4𝑏
Stackelberg Rivalry

Bertrand Rivalry Follower′ s reaction function

𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 𝑆𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 3 𝑎 − 𝑀𝐶
𝑄 𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 = 𝑞1 + 𝑞2 =
4𝑏

𝑎 + 3 𝑀𝐶
𝑃 𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 = 𝑎 − 𝑏𝑄 𝑠𝑡𝑎𝑐𝑘𝑒𝑙𝑏𝑒𝑟𝑔 =
4

51
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 outputs 𝑷

Monopoly 1
𝑎 − 𝑀𝐶
2𝑏 fymounolb
𝑎 − 𝑀𝐶
2𝑏
𝑎 + 𝑀𝐶
2
2 𝑎 − 𝑀𝐶 𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
Cournot ≥1
𝑁 + 1 2𝑏 𝑁+1 𝑏 𝑁+1
𝑎 − 𝑀𝐶
“Competition” →∞ 0 𝑀𝐶
𝑏
1 𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶

I
Cartel >1
𝑁 2𝑏 2𝑏 2

Imitate
𝑎 − 𝑀𝐶
𝑞𝑙𝑒𝑎𝑑𝑒𝑟 =
2𝑏 3 𝑎 − 𝑀𝐶 𝑎 + 3 𝑀𝐶
Stackelberg 2
𝑎 − 𝑀𝐶 4𝑏 4
𝑞𝑓𝑜𝑙𝑙𝑜𝑤𝑒𝑟 =
4𝑏
Bertrand

52
Bertrand Rivalry
(Two Firms compete by choosing prices simultaneously)
For simplicity, analyze Bertrand Model with TWO firms
Benchmark: “Naïve” Monopoly

Cournot Rivalry
“Bertrand Price Competition Rules”
Cournot Rivalry with 𝑁 → ∞: “appears to be perfect

If the two firms are charging different prices: then the firm with the
competition”

Collusion in a Cartel
lower price point captures the entire market
Stackelberg Rivalry

Bertrand Rivalry
If the two firms are charging the same price: then the two firms split
the market 50:50

If 𝑃𝑖 < 𝑃𝑗 ⟹ 𝑞𝑖 = 𝑄, 𝑞𝑗 = 0
If 𝑃𝑖 > 𝑃𝑗 ⟹ 𝑞𝑖 = 0, 𝑞𝑗 = 𝑄
Bertrand Firm i′ s Demand given 𝑃𝑗 =
𝑄 𝑄
If 𝑃𝑖 = 𝑃𝑗 ⟹ 𝑞𝑖 = , 𝑞𝑗 =
2 2

53
Bertrand Rivalry
(Two Firms compete by choosing prices simultaneously)
Benchmark: “Naïve” Monopoly ❶ ❷
$ Suppose initially 𝑃1 = 𝑃2 > 𝑀𝐶 $ Now suppose 𝑃1 < 𝑃2
Companies split market Firm 1 takes entire market
Cournot Rivalry

Cournot Rivalry with 𝑁 → ∞: “appears to be perfect

competition” 𝑃1 = 𝑃2 1 𝑃2

Collusion in a Cartel 𝑃1 2

Stackelberg Rivalry 𝑀𝐶 𝑀𝐶
Market Demand Market Demand
Bertrand Rivalry
𝑞1 = 𝑞2 𝑞 𝑞2 𝑞1 𝑞

❸ ❹
If 𝑃𝑖 < 𝑃𝑗 ⟹ 𝑞𝑖 = 𝑄, 𝑞𝑗 = 0 Ultimately: 𝑃1 = 𝑃2 = 𝑀𝐶
Firm 2 responds: 𝑃2 < 𝑃1
If 𝑃𝑖 > 𝑃𝑗 ⟹ 𝑞𝑖 = 0, 𝑞𝑗 = 𝑄 $ $
Now, Firm 2 takes entire market Companies split market
Bertrand Firm i′ s Demand =
𝑄 𝑄
If 𝑃𝑖 = 𝑃𝑗 ⟹ 𝑞𝑖 = , 𝑞𝑗 =
2 2

𝑃1

𝑃2 3
𝑀𝐶 𝑃1 = 𝑃2 4 𝑀𝐶
Market Demand Market Demand
𝑞1 𝑞2
𝑞 𝑞1 = 𝑞2 𝑞

54
Evolution of a Theoretical Industry
“Perfect Competition”
𝑁 → ∞ firms compete by choosing outputs
Cournot Rivalry simultaneously
Firms compete by choosing outputs simultaneously Cartel
𝑄 = 𝑞1 + 𝑞2 + … + 𝑞𝑁 Firms cooperate by choosing outputs
simultaneously
𝑃 = 𝑎 − 𝑏𝑄
Stackelberg Rivalry
Monopoly
Firms compete by choosing outputs sequentially

𝑇𝑉𝐶𝑖 = 𝑀𝐶 𝑞𝑖 Bertrand Rivalry

Firms compete by choosing prices simultaneously

“Industry” 𝑵 𝒒𝒊 𝑸 𝑷
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶
Monopoly 1
2𝑏 2𝑏 2
2 𝑎 − 𝑀𝐶 𝑁 𝑎 − 𝑀𝐶 𝑎 + 𝑁 𝑀𝐶
Cournot ≥1
𝑁 + 1 2𝑏 𝑁+1 𝑏 𝑁+1
𝑎 − 𝑀𝐶
“Competition” →∞ 0 𝑀𝐶
𝑏
1 𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶 𝑎 + 𝑀𝐶
Cartel >1
𝑁 2𝑏 2𝑏 2
𝑎 − 𝑀𝐶
𝑞𝑙𝑒𝑎𝑑𝑒𝑟 =
2𝑏 3 𝑎 − 𝑀𝐶 𝑎 + 3 𝑀𝐶
Stackelberg 2
𝑎 − 𝑀𝐶 4𝑏 4
𝑞𝑓𝑜𝑙𝑙𝑜𝑤𝑒𝑟 =
4𝑏
𝑎 − 𝑀𝐶 𝑎 − 𝑀𝐶
Bertrand 2 𝑀𝐶
2𝑏 𝑏

55
 Secret Appendix
(For next ECO204 Cohort)

56
 Player B
𝑟 𝑠
1 2
Player A 𝑝 1 𝑎 𝑏 𝑎 𝑏
𝜋11 , 𝜋11 𝜋12 , 𝜋12
𝑞 2 𝑎 𝑏 𝑎 𝑏
𝜋21 , 𝜋21 𝜋22 , 𝜋22
𝑎 𝑏 𝑎 𝑏 𝑎 𝑎 𝑎 𝑏 𝑎 𝑏 𝑎 𝑎
𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋11 > 𝜋21 𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋22 > 𝜋12
𝑏 𝑏 𝑏 𝑏
𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋11 > 𝜋12 𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋11 < 𝜋12
𝑏 𝑏 𝑏 𝑏
𝜋21 > 𝜋22 𝜋21 < 𝜋22
𝑎 𝑏 𝑎 𝑏 𝑎 𝑎
𝑎
𝜋11 𝑏
, 𝜋11 𝑎
𝜋12 𝑏
, 𝜋12 𝑎
𝜋11 𝑎
< 𝜋21 𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋12 < 𝜋22
𝑏 𝑏
𝑎 𝑏 𝑎 𝑏
𝑏
𝜋21 𝑏
> 𝜋22 𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋21 < 𝜋22
𝜋21 , 𝜋21 𝜋22 , 𝜋22 𝑏 𝑏
𝑏
𝜋11 𝑏
> 𝜋12 𝜋11 < 𝜋12
𝑎 𝑏 𝑎 𝑏 𝑎 𝑎
𝑎
𝜋11 𝑏
, 𝜋11 𝑎
𝜋12 𝑏
, 𝜋12 𝑎
𝜋11 𝑎
> 𝜋21 𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋11 = 𝜋21
𝑏 𝑏
𝑎 𝑏 𝑎 𝑏
𝑏
𝜋11 𝑏
= 𝜋12 𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋11 > 𝜋12
𝜋21 , 𝜋21 𝜋22 , 𝜋22 𝑎 𝑎
𝜋12 > 𝜋22
𝑏 𝑏
𝑏
𝜋21 𝑏
> 𝜋22 𝜋21 > 𝜋22
𝑎 𝑏 𝑎 𝑏 𝑎 𝑎
𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋11 > 𝜋21
𝑏 𝑏
𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋11 > 𝜋12
𝑎 𝑎
𝜋12 < 𝜋22
𝑏 𝑏
𝜋21 < 𝜋22 57
 Player B
𝑟 𝑠
1 2
Player A 𝑝 1 𝑎 𝑏 𝑎 𝑏
𝜋11 , 𝜋11 𝜋12 , 𝜋12
𝑞 2 𝑎 𝑏 𝑎 𝑏
𝜋21 , 𝜋21 𝜋22 , 𝜋22

No Pure Strategy
𝑎 𝑏 𝑎 𝑏 𝑎 𝑎
𝜋11 , 𝜋11 𝜋12 , 𝜋12 𝜋11 < 𝜋21
𝑏 𝑏
𝑎
𝜋21 𝑏
, 𝜋21 𝑎
𝜋22 𝑏
, 𝜋22 𝜋11 > 𝜋12
𝑎 𝑎
𝜋12 > 𝜋22
𝑏 𝑏
𝜋21 < 𝜋22
𝑏 𝑏
𝜋22 − 𝜋21
𝑝= 𝑏 𝑏 𝑏 𝑏
𝜋11 + 𝜋22 − 𝜋12 − 𝜋21
𝑎 𝑎
𝜋12 − 𝜋22
𝑟= 𝑎 𝑎 𝑎 𝑎
𝜋21 + 𝜋12 − 𝜋11 − 𝜋22

58