際際滷

際際滷Share a Scribd company logo
GAME THEORY
Easy College Study
Lecture 5: Bertrand Duopoly, Crime
Report and Auctions
 In this lecture, we are going to see applications of static game theoretical model in real
world, to understand production decision making, moral dilemma, and auction process
 Review: definition of games, strategies, NE and Cournot Duopoly
 Bertrand Duopoly: continuous and discrete pricing
 Reporting a crime:
 How can we understand the inefficiency in social moral dilemmas like this ?
 Auctions
 Complete information
 Second-price sealed-bid auction
 First-price sealed-bid auction
 Spectrum Auctions: some facts
Review: Static games
Review: Best response
 The best response captures the best choice of individual i when given other players
actions a-i
Review: Pure strategy Nash Equilibrium
(PNE)
Review: Mixed strategy
 Denote  基 = {腫}, then  基 is a (|基|  1)-simplex
 Denote   the set of lottery over A, then   is a (|A|  1)-simplex
 Then, each prole of mixed strategies  = (1,  ,  ), maps to a lottery on   ,
with assumption of independence, we have
Review: Mixed strategy Nash Equilibrium
Cournot Duopoly
 Two firms i and j, compete by producing quantities   0, and    0. under price
function  ,   =   ( +  ), ,  > 0. Marginal cost of production c>0
 Each firm maximizes profit given the other's production. For firm i :
 First-order condition:   2      = 0
 Solving for  gives firm i's best response:
 Likewise, we can derive firm js best response
Bertrand Duopoly
 While Cournot Duopoly is competing in quantity, Bertrand Duopoly, however, is
competing in pricing strategy
 Two firms i and j compete by setting prices   0 and   0, resp. Consumers buy
only the cheapest good; firm i faces demand function:
Bertrand Duopoly
 Each firm maximizes profit  given the other's price. For firm i :
Bertrand Duopoly
Bertrand Duopoly
Bertrand Duopoly
Bertrand Duopoly: discrete pricing
 Assume firms can set prices equal to some multiple of the smallest denomination
cent (e.g., 1 penny USD)?
 Refine 

to maximize (, ) when   0 subject to 

dividable by cents.
Bertrand Duopoly: discrete pricing
Bertrand Duopoly: discrete pricing
Reporting a crime
 Players: n > 1 bystanders.
 Actions: Ai = {(R)eport, (D)ont Report}.
 Pareto efficient outcome: exactly one bystander reports.
 Best response:
 Report if no one else reports, Don't Report if anyone else reports.
 Is existence of symmetric MNE guaranteed?
Reporting a crime
Reporting a crime
 Symmetric MNE: p = 1 - (c/v )1/(n-1) for each i  N.
 Probability i reports, p, is decreasing in n: increasing incentives to free load" as
more bystanders witness the crime (as n ).
 Probability no one reports, (1 - p)n = (c/v )n/(n-1), is also decreasing in n(!): as more
bystanders witness the crime, the public good" of reporting the crime is more under
provided.
 Equilibrium welfare? What inefficiencies are there?
Auctions: complete information
 In this class: we dene an auction" as a market mechanism in which a single good is
sold to a player (bidder") who submits a highest bid.
 More generally: auctions may include multiple units, and more complicated actions
than a single bid, e.g., bidding fee" auctions, and Simultaneous Multiple-Round"
(SMR) auctions (more later).
Auctions: complete information
 Types of auctions (we study):
 Second-price sealed-bid auction (SPA): static (simultaneous move) auctions where
highest bidder pays the second highest bid. SPA approximates dynamic assenting price
English" auction.
 First-price sealed-bid auction (FPA): static (simultaneous move) auctions where highest
bidder pays their bid. FPA approximates dynamic descending price Dutch" auction.
 Auctions can either be under complete information (covered today), or
incomplete/private information
 Both SPA and FPA under complete information yield multiple PNE...
Auctions: Second-price sealed-bid auction (SPA)
 Players: n > 1 bidders.
 Actions: bids bi  0 for each i  N.
Auctions: Second-price sealed-bid auction (SPA)
Auctions: Second-price sealed-bid auction
(SPA)
Auctions: Second-price sealed-bid auction (SPA)
Auctions: First-price sealed-bid auction
(FPA)
 Players: n > 1 bidders.
 Actions: bids bi  0 for each i  N.
Auctions: First-price sealed-bid auction (FPA)
Auctions: First-price sealed-bid auction
(FPA)
Auctions: First-price sealed-bid auction
(FPA)
Auctions: First-price sealed-bid auction (FPA)
Auctions: First-price sealed-bid auction (FPA)
Auctions: First-price sealed-bid auction (FPA)

More Related Content

L5

  • 2. Lecture 5: Bertrand Duopoly, Crime Report and Auctions In this lecture, we are going to see applications of static game theoretical model in real world, to understand production decision making, moral dilemma, and auction process Review: definition of games, strategies, NE and Cournot Duopoly Bertrand Duopoly: continuous and discrete pricing Reporting a crime: How can we understand the inefficiency in social moral dilemmas like this ? Auctions Complete information Second-price sealed-bid auction First-price sealed-bid auction Spectrum Auctions: some facts
  • 4. Review: Best response The best response captures the best choice of individual i when given other players actions a-i
  • 5. Review: Pure strategy Nash Equilibrium (PNE)
  • 6. Review: Mixed strategy Denote 基 = {腫}, then 基 is a (|基| 1)-simplex Denote the set of lottery over A, then is a (|A| 1)-simplex Then, each prole of mixed strategies = (1, , ), maps to a lottery on , with assumption of independence, we have
  • 7. Review: Mixed strategy Nash Equilibrium
  • 8. Cournot Duopoly Two firms i and j, compete by producing quantities 0, and 0. under price function , = ( + ), , > 0. Marginal cost of production c>0 Each firm maximizes profit given the other's production. For firm i : First-order condition: 2 = 0 Solving for gives firm i's best response: Likewise, we can derive firm js best response
  • 9. Bertrand Duopoly While Cournot Duopoly is competing in quantity, Bertrand Duopoly, however, is competing in pricing strategy Two firms i and j compete by setting prices 0 and 0, resp. Consumers buy only the cheapest good; firm i faces demand function:
  • 10. Bertrand Duopoly Each firm maximizes profit given the other's price. For firm i :
  • 14. Bertrand Duopoly: discrete pricing Assume firms can set prices equal to some multiple of the smallest denomination cent (e.g., 1 penny USD)? Refine to maximize (, ) when 0 subject to dividable by cents.
  • 17. Reporting a crime Players: n > 1 bystanders. Actions: Ai = {(R)eport, (D)ont Report}. Pareto efficient outcome: exactly one bystander reports. Best response: Report if no one else reports, Don't Report if anyone else reports. Is existence of symmetric MNE guaranteed?
  • 19. Reporting a crime Symmetric MNE: p = 1 - (c/v )1/(n-1) for each i N. Probability i reports, p, is decreasing in n: increasing incentives to free load" as more bystanders witness the crime (as n ). Probability no one reports, (1 - p)n = (c/v )n/(n-1), is also decreasing in n(!): as more bystanders witness the crime, the public good" of reporting the crime is more under provided. Equilibrium welfare? What inefficiencies are there?
  • 20. Auctions: complete information In this class: we dene an auction" as a market mechanism in which a single good is sold to a player (bidder") who submits a highest bid. More generally: auctions may include multiple units, and more complicated actions than a single bid, e.g., bidding fee" auctions, and Simultaneous Multiple-Round" (SMR) auctions (more later).
  • 21. Auctions: complete information Types of auctions (we study): Second-price sealed-bid auction (SPA): static (simultaneous move) auctions where highest bidder pays the second highest bid. SPA approximates dynamic assenting price English" auction. First-price sealed-bid auction (FPA): static (simultaneous move) auctions where highest bidder pays their bid. FPA approximates dynamic descending price Dutch" auction. Auctions can either be under complete information (covered today), or incomplete/private information Both SPA and FPA under complete information yield multiple PNE...
  • 22. Auctions: Second-price sealed-bid auction (SPA) Players: n > 1 bidders. Actions: bids bi 0 for each i N.
  • 26. Auctions: First-price sealed-bid auction (FPA) Players: n > 1 bidders. Actions: bids bi 0 for each i N.