Markets The Economics of Platform Competition and Online Markets

112
Econ 191: The Economics of Platform Competition and Online Markets Preparatory lecture for Prof. John Morgan Issi Romem Introduction Nash equilibrium with pure strategies Nash equilibrium with mixed strategies Experimental design: the importance of randomization Econ 191: The Economics of Platform Competition and Online Markets Preparatory lecture for Prof. John Morgan Issi Romem February 21, 2012

Transcript of Markets The Economics of Platform Competition and Online Markets

Page 1: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Econ 191:The Economics of Platform Competition

and Online MarketsPreparatory lecture for Prof. John Morgan

Issi Romem

February 21, 2012

Page 2: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Introduction

Today’s lecture:

Nash equilibrium with pure strategies

Nash equilibrium with mixed strategies

And also:

The importance of randomization in experimental design.

Page 3: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Introduction

Today’s lecture:

Nash equilibrium with pure strategies

Nash equilibrium with mixed strategies

And also:

The importance of randomization in experimental design.

Page 4: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Introduction

Today’s lecture:

Nash equilibrium with pure strategies

Nash equilibrium with mixed strategies

And also:

The importance of randomization in experimental design.

Page 5: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Introduction

Today’s lecture:

Nash equilibrium with pure strategies

Nash equilibrium with mixed strategies

And also:

The importance of randomization in experimental design.

Page 6: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 7: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.

What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 8: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 9: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 10: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 11: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Game theory is a mathematical tool that can help understandsituations in which decision-makers interact.What constitutes a game?

Decision-makers, called “players”.

For each player, a set of possible “strategies”.

For each possible combination of players’ strategies, an“outcome”, consisting of a “payoff” for each player.

Page 12: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 13: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 14: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 15: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 16: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 17: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

Example:

Jacktv internet

Jilltv 2,2 5,5

internet 6,6 0,0

Two players in this game, Jack and Jill.

Jack has two strategies, tv and internet, and so does Jill.

The outcomes for each combination of players’ strategiesis given.

What will happen?

Page 18: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

In order to say what will happen, we need a “solution concept”.

“Nash equlibrium” is a solution concept, and can be defined asfollows:

Nash equilibrium

An outcome is a Nash equilibrium if, given the strategies of allother players, no player has anything to gain by changing onlyhis own strategy unilaterally.

What outcome/s are a Nash equilibrium in the previous game?

Page 19: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

In order to say what will happen, we need a “solution concept”.

“Nash equlibrium” is a solution concept, and can be defined asfollows:

Nash equilibrium

An outcome is a Nash equilibrium if, given the strategies of allother players, no player has anything to gain by changing onlyhis own strategy unilaterally.

What outcome/s are a Nash equilibrium in the previous game?

Page 20: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

In order to say what will happen, we need a “solution concept”.

“Nash equlibrium” is a solution concept, and can be defined asfollows:

Nash equilibrium

An outcome is a Nash equilibrium if, given the strategies of allother players, no player has anything to gain by changing onlyhis own strategy unilaterally.

What outcome/s are a Nash equilibrium in the previous game?

Page 21: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

In order to say what will happen, we need a “solution concept”.

“Nash equlibrium” is a solution concept, and can be defined asfollows:

Nash equilibrium

An outcome is a Nash equilibrium if, given the strategies of allother players, no player has anything to gain by changing onlyhis own strategy unilaterally.

What outcome/s are a Nash equilibrium in the previous game?

Page 22: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What is a “game”?

In order to say what will happen, we need a “solution concept”.

“Nash equlibrium” is a solution concept, and can be defined asfollows:

Nash equilibrium

An outcome is a Nash equilibrium if, given the strategies of allother players, no player has anything to gain by changing onlyhis own strategy unilaterally.

What outcome/s are a Nash equilibrium in the previous game?

Page 23: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Jack and Jill:

Player 2: Jacktv internet

Player 1: Jilltv 2,2 5,5

internet 6,6 0,0

Page 24: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Coordination game:

AT&T Verizon

AT&T 2,2 0,0Verizon 0,0 1,1

Page 25: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Still a coordination game?

L R

T 2,2 0,0B 0,0 -1,-1

Page 26: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

The prisoner’s dilemma:

be quiet tell

be quiet 2,2 0,3tell 3,0 1,1

Page 27: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Working together:

work hard good off

work hard 2,2 0,3goof off 3,0 1,1

Page 28: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Hawk/Dove:

H D

H 0,0 3,1D 1,3 2,2

Page 29: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 30: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 31: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 32: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 33: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 34: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Examples:

Matching pennies:

Each player has a penny.

Both players simultaneously show each other one face oftheir penny.

Player 1 wins if both pennies show the same face.

Player 2 wins if the pennies show different faces.

Head Tail

Head 1,-1 -1,1Tail -1,1 1,-1

Page 35: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Does “matching pennies” have a Nash equilibrium?

“Matching pennies” does not have a Nash equilibrium withpure strategies.

What if we define players’ possible strategies differently...

Page 36: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Does “matching pennies” have a Nash equilibrium?

“Matching pennies” does not have a Nash equilibrium withpure strategies.

What if we define players’ possible strategies differently...

Page 37: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Does “matching pennies” have a Nash equilibrium?

“Matching pennies” does not have a Nash equilibrium withpure strategies.

What if we define players’ possible strategies differently...

Page 38: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixed strategies

Instead of having just two strategies, Head or Tail, each playercan choose a probability of Head or Tail as his strategy - amixed strategy.

Player 1’s (2’s) strategy is now:Head with probability 0 ≤ p ≤ 1 (0 ≤ q ≤ 1).(Of course, this implies Tail with probability 1− p (1− q).)

This means that each player has a set of strategies that isinfinite!

First, let’s practice finding mixed strategy Nash equilibria.Then, we’ll think about what they actually represent.

Page 39: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixed strategies

Instead of having just two strategies, Head or Tail, each playercan choose a probability of Head or Tail as his strategy - amixed strategy.

Player 1’s (2’s) strategy is now:Head with probability 0 ≤ p ≤ 1 (0 ≤ q ≤ 1).(Of course, this implies Tail with probability 1− p (1− q).)

This means that each player has a set of strategies that isinfinite!

First, let’s practice finding mixed strategy Nash equilibria.Then, we’ll think about what they actually represent.

Page 40: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixed strategies

Instead of having just two strategies, Head or Tail, each playercan choose a probability of Head or Tail as his strategy - amixed strategy.

Player 1’s (2’s) strategy is now:Head with probability 0 ≤ p ≤ 1 (0 ≤ q ≤ 1).(Of course, this implies Tail with probability 1− p (1− q).)

This means that each player has a set of strategies that isinfinite!

First, let’s practice finding mixed strategy Nash equilibria.Then, we’ll think about what they actually represent.

Page 41: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixed strategies

Instead of having just two strategies, Head or Tail, each playercan choose a probability of Head or Tail as his strategy - amixed strategy.

Player 1’s (2’s) strategy is now:Head with probability 0 ≤ p ≤ 1 (0 ≤ q ≤ 1).(Of course, this implies Tail with probability 1− p (1− q).)

This means that each player has a set of strategies that isinfinite!

First, let’s practice finding mixed strategy Nash equilibria.Then, we’ll think about what they actually represent.

Page 42: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixed strategies

Instead of having just two strategies, Head or Tail, each playercan choose a probability of Head or Tail as his strategy - amixed strategy.

Player 1’s (2’s) strategy is now:Head with probability 0 ≤ p ≤ 1 (0 ≤ q ≤ 1).(Of course, this implies Tail with probability 1− p (1− q).)

This means that each player has a set of strategies that isinfinite!

First, let’s practice finding mixed strategy Nash equilibria.Then, we’ll think about what they actually represent.

Page 43: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

If player 1 picks the strategy p = 1 (i.e. play Head forsure) then his payoff is: q · 1 + (1− q) · −1 = 2q − 1.

If player 1 picks the strategy p = 0 (i.e. play Tail for sure)then his payoff is: q · −1 + (1− q) · 1 = 1− 2q.

Generally, if player 1 picks the strategy p then his payoff is:p · (2q − 1) + (1− p) · (1− 2q).

What does player 1 do?

Page 44: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

If player 1 picks the strategy p = 1 (i.e. play Head forsure) then his payoff is: q · 1 + (1− q) · −1 = 2q − 1.

If player 1 picks the strategy p = 0 (i.e. play Tail for sure)then his payoff is: q · −1 + (1− q) · 1 = 1− 2q.

Generally, if player 1 picks the strategy p then his payoff is:p · (2q − 1) + (1− p) · (1− 2q).

What does player 1 do?

Page 45: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

If player 1 picks the strategy p = 1 (i.e. play Head forsure) then his payoff is: q · 1 + (1− q) · −1 = 2q − 1.

If player 1 picks the strategy p = 0 (i.e. play Tail for sure)then his payoff is: q · −1 + (1− q) · 1 = 1− 2q.

Generally, if player 1 picks the strategy p then his payoff is:p · (2q − 1) + (1− p) · (1− 2q).

What does player 1 do?

Page 46: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

If player 1 picks the strategy p = 1 (i.e. play Head forsure) then his payoff is: q · 1 + (1− q) · −1 = 2q − 1.

If player 1 picks the strategy p = 0 (i.e. play Tail for sure)then his payoff is: q · −1 + (1− q) · 1 = 1− 2q.

Generally, if player 1 picks the strategy p then his payoff is:p · (2q − 1) + (1− p) · (1− 2q).

What does player 1 do?

Page 47: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

If player 1 picks the strategy p = 1 (i.e. play Head forsure) then his payoff is: q · 1 + (1− q) · −1 = 2q − 1.

If player 1 picks the strategy p = 0 (i.e. play Tail for sure)then his payoff is: q · −1 + (1− q) · 1 = 1− 2q.

Generally, if player 1 picks the strategy p then his payoff is:p · (2q − 1) + (1− p) · (1− 2q).

What does player 1 do?

Page 48: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 49: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 50: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 51: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 52: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 53: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

⇒ Expected payoff 2q − 1

⇒ Expected payoff 1− 2q

2q − 1 > 1− 2q

4q > 2

q > 0.5

If q > 0.5, player 1 chooses p = 1.

If q < 0.5, player 1 chooses p = 0.

If q = 0.5, player 1 is... indifferent, andmay choose any p!

What does player 2 do? A similarcalculation...

Page 54: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

Player 1: q > 0.5⇒ p = 1; q < 0.5⇒ p = 0;q = 0.5⇒ indifferent, so any p is optimal.

Player 2: p < 0.5⇒ q = 1; p > 0.5⇒ q = 0;p = 0.5⇒ indifferent, so any q is optimal.

⇒ Mixed strategy Nash equilibrium: (p, q) = (0.5, 0.5).

Page 55: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

Player 1: q > 0.5⇒ p = 1; q < 0.5⇒ p = 0;q = 0.5⇒ indifferent, so any p is optimal.

Player 2: p < 0.5⇒ q = 1; p > 0.5⇒ q = 0;p = 0.5⇒ indifferent, so any q is optimal.

⇒ Mixed strategy Nash equilibrium: (p, q) = (0.5, 0.5).

Page 56: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Matching pennies

q 1-qHead Tail

p Head 1,-1 -1,11-p Tail -1,1 1,-1

Player 1: q > 0.5⇒ p = 1; q < 0.5⇒ p = 0;q = 0.5⇒ indifferent, so any p is optimal.

Player 2: p < 0.5⇒ q = 1; p > 0.5⇒ q = 0;p = 0.5⇒ indifferent, so any q is optimal.

⇒ Mixed strategy Nash equilibrium: (p, q) = (0.5, 0.5).

Page 57: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixture implies indifference between pure strategies

A mixture of two (or more) pure strategies can only be optimalif, given the other players’ behavior, the pure strategies providethe same payoff.

Why?Because if one of the pure strategies provides a bigger payoff, itis optimal to place more weight on it.

We can use this insight to solve for mixed strategies.

Page 58: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixture implies indifference between pure strategies

A mixture of two (or more) pure strategies can only be optimalif, given the other players’ behavior, the pure strategies providethe same payoff.

Why?Because if one of the pure strategies provides a bigger payoff, itis optimal to place more weight on it.

We can use this insight to solve for mixed strategies.

Page 59: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixture implies indifference between pure strategies

A mixture of two (or more) pure strategies can only be optimalif, given the other players’ behavior, the pure strategies providethe same payoff.

Why?Because if one of the pure strategies provides a bigger payoff, itis optimal to place more weight on it.

We can use this insight to solve for mixed strategies.

Page 60: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Mixture implies indifference between pure strategies

A mixture of two (or more) pure strategies can only be optimalif, given the other players’ behavior, the pure strategies providethe same payoff.

Why?Because if one of the pure strategies provides a bigger payoff, itis optimal to place more weight on it.

We can use this insight to solve for mixed strategies.

Page 61: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

We saw earlier that this game has two pure strategy Nashequilibria:(p, q) = (1, 1) and (p, q) = (0, 0).

Does this game also have any mixed strategy Nashequilibria?

Try it...

Page 62: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

We saw earlier that this game has two pure strategy Nashequilibria:(p, q) = (1, 1) and (p, q) = (0, 0).

Does this game also have any mixed strategy Nashequilibria?

Try it...

Page 63: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

We saw earlier that this game has two pure strategy Nashequilibria:(p, q) = (1, 1) and (p, q) = (0, 0).

Does this game also have any mixed strategy Nashequilibria?

Try it...

Page 64: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

Player 1:

AT&T exp. payoff

= Verzion exp. payoff

2q = 1− q

3q = 1

q = 1/3

Player 2:

AT&T exp. payoff

= Verzion exp. payoff

2p = 1− p

3p = 1

p = 1/3

⇒ A third Nash equilibrium: (p, q) = (1/3, 1/3).

Page 65: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

Player 1:

AT&T exp. payoff

= Verzion exp. payoff

2q = 1− q

3q = 1

q = 1/3

Player 2:

AT&T exp. payoff

= Verzion exp. payoff

2p = 1− p

3p = 1

p = 1/3

⇒ A third Nash equilibrium: (p, q) = (1/3, 1/3).

Page 66: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

Player 1:

AT&T exp. payoff

= Verzion exp. payoff

2q = 1− q

3q = 1

q = 1/3

Player 2:

AT&T exp. payoff

= Verzion exp. payoff

2p = 1− p

3p = 1

p = 1/3

⇒ A third Nash equilibrium: (p, q) = (1/3, 1/3).

Page 67: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

Player 1:

AT&T exp. payoff

= Verzion exp. payoff

2q = 1− q

3q = 1

q = 1/3

Player 2:

AT&T exp. payoff

= Verzion exp. payoff

2p = 1− p

3p = 1

p = 1/3

⇒ A third Nash equilibrium: (p, q) = (1/3, 1/3).

Page 68: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Coordination game revisited

q 1-qAT&T Verizon

p AT&T 2,2 0,01-p Verizon 0,0 1,1

Player 1:

AT&T exp. payoff

= Verzion exp. payoff

2q = 1− q

3q = 1

q = 1/3

Player 2:

AT&T exp. payoff

= Verzion exp. payoff

2p = 1− p

3p = 1

p = 1/3⇒ A third Nash equilibrium: (p, q) = (1/3, 1/3).

Page 69: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In matching pennies, it makes sense to think of mixedstrategies literally - players might decide on their move bythrowing dice behind their back.

In the coordination game, this seems artificial.

Another way of thinking about mixed strategy Nash equilibriais as a concise description of the pure strategy Nash equilibriain many slightly different versions of the same game(“Harsanyi’s purification”).

Page 70: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In matching pennies, it makes sense to think of mixedstrategies literally - players might decide on their move bythrowing dice behind their back.

In the coordination game, this seems artificial.

Another way of thinking about mixed strategy Nash equilibriais as a concise description of the pure strategy Nash equilibriain many slightly different versions of the same game(“Harsanyi’s purification”).

Page 71: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In matching pennies, it makes sense to think of mixedstrategies literally - players might decide on their move bythrowing dice behind their back.

In the coordination game, this seems artificial.

Another way of thinking about mixed strategy Nash equilibriais as a concise description of the pure strategy Nash equilibriain many slightly different versions of the same game(“Harsanyi’s purification”).

Page 72: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In matching pennies, it makes sense to think of mixedstrategies literally - players might decide on their move bythrowing dice behind their back.

In the coordination game, this seems artificial.

Another way of thinking about mixed strategy Nash equilibriais as a concise description of the pure strategy Nash equilibriain many slightly different versions of the same game(“Harsanyi’s purification”).

Page 73: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In a 1973 article, (Berkeley Nobel laureate) John Harsanyisuggested (roughly) the following:

Suppose the payoffs of a game are “disturbed”, i.e. theyhave small random fluctuations around the payoffs given inthe game.

It can be shown that for any “disturbed” version of thegame there exists a pure Nash equilibrium.

The mixed strategy Nash equilibrium can be thought of asthe average of many “disturbed” versions of the game.

The equilibrium mixture of strategies reflects thedistribution of the disturbances.

Page 74: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In a 1973 article, (Berkeley Nobel laureate) John Harsanyisuggested (roughly) the following:

Suppose the payoffs of a game are “disturbed”, i.e. theyhave small random fluctuations around the payoffs given inthe game.

It can be shown that for any “disturbed” version of thegame there exists a pure Nash equilibrium.

The mixed strategy Nash equilibrium can be thought of asthe average of many “disturbed” versions of the game.

The equilibrium mixture of strategies reflects thedistribution of the disturbances.

Page 75: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In a 1973 article, (Berkeley Nobel laureate) John Harsanyisuggested (roughly) the following:

Suppose the payoffs of a game are “disturbed”, i.e. theyhave small random fluctuations around the payoffs given inthe game.

It can be shown that for any “disturbed” version of thegame there exists a pure Nash equilibrium.

The mixed strategy Nash equilibrium can be thought of asthe average of many “disturbed” versions of the game.

The equilibrium mixture of strategies reflects thedistribution of the disturbances.

Page 76: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In a 1973 article, (Berkeley Nobel laureate) John Harsanyisuggested (roughly) the following:

Suppose the payoffs of a game are “disturbed”, i.e. theyhave small random fluctuations around the payoffs given inthe game.

It can be shown that for any “disturbed” version of thegame there exists a pure Nash equilibrium.

The mixed strategy Nash equilibrium can be thought of asthe average of many “disturbed” versions of the game.

The equilibrium mixture of strategies reflects thedistribution of the disturbances.

Page 77: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In a 1973 article, (Berkeley Nobel laureate) John Harsanyisuggested (roughly) the following:

Suppose the payoffs of a game are “disturbed”, i.e. theyhave small random fluctuations around the payoffs given inthe game.

It can be shown that for any “disturbed” version of thegame there exists a pure Nash equilibrium.

The mixed strategy Nash equilibrium can be thought of asthe average of many “disturbed” versions of the game.

The equilibrium mixture of strategies reflects thedistribution of the disturbances.

Page 78: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In the AT&T vs. Verizon coordination game, think of mixedstrategy Nash eq. as follows:

A specific person can only belong to one network - a purestrategy.

Player 2 describes the average of all other people (a shareq belong to AT&T, the rest to Verizon).

For player 1 the game is “disturbed” in such a way that heeither goes AT&T or Verizon.

If player 1 also describes the average of many people, thenthe mixed strategy Nash eq. is just a concise way ofdescribing all the versions of the game at once.

Page 79: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In the AT&T vs. Verizon coordination game, think of mixedstrategy Nash eq. as follows:

A specific person can only belong to one network - a purestrategy.

Player 2 describes the average of all other people (a shareq belong to AT&T, the rest to Verizon).

For player 1 the game is “disturbed” in such a way that heeither goes AT&T or Verizon.

If player 1 also describes the average of many people, thenthe mixed strategy Nash eq. is just a concise way ofdescribing all the versions of the game at once.

Page 80: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In the AT&T vs. Verizon coordination game, think of mixedstrategy Nash eq. as follows:

A specific person can only belong to one network - a purestrategy.

Player 2 describes the average of all other people (a shareq belong to AT&T, the rest to Verizon).

For player 1 the game is “disturbed” in such a way that heeither goes AT&T or Verizon.

If player 1 also describes the average of many people, thenthe mixed strategy Nash eq. is just a concise way ofdescribing all the versions of the game at once.

Page 81: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In the AT&T vs. Verizon coordination game, think of mixedstrategy Nash eq. as follows:

A specific person can only belong to one network - a purestrategy.

Player 2 describes the average of all other people (a shareq belong to AT&T, the rest to Verizon).

For player 1 the game is “disturbed” in such a way that heeither goes AT&T or Verizon.

If player 1 also describes the average of many people, thenthe mixed strategy Nash eq. is just a concise way ofdescribing all the versions of the game at once.

Page 82: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

In the AT&T vs. Verizon coordination game, think of mixedstrategy Nash eq. as follows:

A specific person can only belong to one network - a purestrategy.

Player 2 describes the average of all other people (a shareq belong to AT&T, the rest to Verizon).

For player 1 the game is “disturbed” in such a way that heeither goes AT&T or Verizon.

If player 1 also describes the average of many people, thenthe mixed strategy Nash eq. is just a concise way ofdescribing all the versions of the game at once.

Page 83: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The conceptual difficulties with mixed strategy Nash eq. aretwofold:

In most situations, a player randomizing his behaviorseems unrealistic.

In a mixed strategy Nash eq. each player is indifferentbetween his mixed strategies, and chooses exactly theprobability that makes the other players indifferent. Whatprevents players from deviating from their strategy?

Page 84: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The conceptual difficulties with mixed strategy Nash eq. aretwofold:

In most situations, a player randomizing his behaviorseems unrealistic.

In a mixed strategy Nash eq. each player is indifferentbetween his mixed strategies, and chooses exactly theprobability that makes the other players indifferent. Whatprevents players from deviating from their strategy?

Page 85: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The conceptual difficulties with mixed strategy Nash eq. aretwofold:

In most situations, a player randomizing his behaviorseems unrealistic.

In a mixed strategy Nash eq. each player is indifferentbetween his mixed strategies, and chooses exactly theprobability that makes the other players indifferent. Whatprevents players from deviating from their strategy?

Page 86: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The previous line of thought following Harsanyi solves both ofthese issues:

Players are not actually randomizing. They each face a“disturbed” game.

In the “disturbed” each player chooses a pure strategythat is strictly optimal, and so does not deviate.

The mixed strategy Nash eq. can be thought of as aconcise description of many such disturbed games.

Page 87: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The previous line of thought following Harsanyi solves both ofthese issues:

Players are not actually randomizing. They each face a“disturbed” game.

In the “disturbed” each player chooses a pure strategythat is strictly optimal, and so does not deviate.

The mixed strategy Nash eq. can be thought of as aconcise description of many such disturbed games.

Page 88: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The previous line of thought following Harsanyi solves both ofthese issues:

Players are not actually randomizing. They each face a“disturbed” game.

In the “disturbed” each player chooses a pure strategythat is strictly optimal, and so does not deviate.

The mixed strategy Nash eq. can be thought of as aconcise description of many such disturbed games.

Page 89: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

What do mixed strategies represent?

The previous line of thought following Harsanyi solves both ofthese issues:

Players are not actually randomizing. They each face a“disturbed” game.

In the “disturbed” each player chooses a pure strategythat is strictly optimal, and so does not deviate.

The mixed strategy Nash eq. can be thought of as aconcise description of many such disturbed games.

Page 90: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In economic experiments, you will often hear that people areassigned to “treatment” and “control” groups randomly.

What is the importance of randomization?

Page 91: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In economic experiments, you will often hear that people areassigned to “treatment” and “control” groups randomly.

What is the importance of randomization?

Page 92: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In economic experiments, you will often hear that people areassigned to “treatment” and “control” groups randomly.

What is the importance of randomization?

Page 93: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 94: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 95: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 96: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 97: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 98: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Let’s think through an example:

Suppose we want to measure the effect of an extra hour ofsection in econometrics.

We set up one section with 3 hrs/wk and another sectionwith 2 hrs/wk, and put this information in the coursecatalog.

At the end of the semester, we compare the averagegrades for students in both sections.

Is the difference between the sections the effect of anextra hour of section time?

No. If better (or worse) students disproportionately select intothe longer section, this may drive (confound!) the difference inaverage grades.

Page 99: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Would randomly assigning students to the two sections help?

Yes.

Why?

One section might still get better students than the other,but the difference will be independent of the extra hour ofsection time, and will disappear as the sections grow larger.

Page 100: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Would randomly assigning students to the two sections help?

Yes.

Why?

One section might still get better students than the other,but the difference will be independent of the extra hour ofsection time, and will disappear as the sections grow larger.

Page 101: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Would randomly assigning students to the two sections help?

Yes.

Why?

One section might still get better students than the other,but the difference will be independent of the extra hour ofsection time, and will disappear as the sections grow larger.

Page 102: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Would randomly assigning students to the two sections help?

Yes.

Why?

One section might still get better students than the other,but the difference will be independent of the extra hour ofsection time, and will disappear as the sections grow larger.

Page 103: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Another example:

Suppose we work at Facebook and want to measure theeffect of a new feature on time spent on the site.

We place a little button for installing the feature in thecorner of the page, and then compare the average timespent on the site for users who did and did not use thenew feature.

Would you interpret the difference between the groups asbeing caused by the new feature? Why?

Page 104: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Another example:

Suppose we work at Facebook and want to measure theeffect of a new feature on time spent on the site.

We place a little button for installing the feature in thecorner of the page, and then compare the average timespent on the site for users who did and did not use thenew feature.

Would you interpret the difference between the groups asbeing caused by the new feature? Why?

Page 105: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Another example:

Suppose we work at Facebook and want to measure theeffect of a new feature on time spent on the site.

We place a little button for installing the feature in thecorner of the page, and then compare the average timespent on the site for users who did and did not use thenew feature.

Would you interpret the difference between the groups asbeing caused by the new feature? Why?

Page 106: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

Another example:

Suppose we work at Facebook and want to measure theeffect of a new feature on time spent on the site.

We place a little button for installing the feature in thecorner of the page, and then compare the average timespent on the site for users who did and did not use thenew feature.

Would you interpret the difference between the groups asbeing caused by the new feature? Why?

Page 107: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In general:

If units are not randomly assigned to treatment andcontrol groups, non-random selection into groups preventscausal interpretation.

Non-random selection likely reflects unobservedcharacteristics that affect outcomes, and may drive groupdifferences in addition to the treatment.

Randomization - randomized assignment to groups - solvesthis problem.

Page 108: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In general:

If units are not randomly assigned to treatment andcontrol groups, non-random selection into groups preventscausal interpretation.

Non-random selection likely reflects unobservedcharacteristics that affect outcomes, and may drive groupdifferences in addition to the treatment.

Randomization - randomized assignment to groups - solvesthis problem.

Page 109: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In general:

If units are not randomly assigned to treatment andcontrol groups, non-random selection into groups preventscausal interpretation.

Non-random selection likely reflects unobservedcharacteristics that affect outcomes, and may drive groupdifferences in addition to the treatment.

Randomization - randomized assignment to groups - solvesthis problem.

Page 110: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Experimental design: the importance ofrandomization

In general:

If units are not randomly assigned to treatment andcontrol groups, non-random selection into groups preventscausal interpretation.

Non-random selection likely reflects unobservedcharacteristics that affect outcomes, and may drive groupdifferences in addition to the treatment.

Randomization - randomized assignment to groups - solvesthis problem.

Page 111: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Thank you for your attention.

Good night!

Page 112: Markets The Economics of Platform Competition and Online Markets

Econ 191:The

Economics ofPlatform

Competitionand OnlineMarkets

Preparatorylecture forProf. JohnMorgan

Issi Romem

Introduction

Nashequilibriumwith purestrategies

Nashequilibriumwith mixedstrategies

Experimentaldesign: theimportance ofrandomization

Thank you for your attention.

Good night!