Agent Based Models and Common Value Auctions

Preview:

DESCRIPTION

Agent Based Models and Common Value Auctions. B. Wade Brorsen. Agent based models (ABM) simulate the actions and interactions of autonomous agents to determine effects on the system as a whole. Agent-Based Model: Bottom-up Method Carry out the experiments with programmed agents - PowerPoint PPT Presentation

Citation preview

Agent Based Models and Common Value Auctions

B. Wade Brorsen

Agent based models (ABM) simulate the actions and interactions of autonomous agents to determine effects on the system as a whole.

3

• Agent-Based Model: Bottom-up Method– Carry out the experiments with programmed agents – Interactive agents with simple rules

Initial condition

Observed Market Equilibrium

Applications of ABM

• Biology• Land Use• Urban Planning• Oligopoly• Auctions

Contrast with Traditional TheoryFully Rational

a. Maximize profitb. Maximize expected utility

Boundedly RationalDo what worked last time

Game Theory

Small number of playersIndustrial organizationInternational tradeNash equilibrium

Oligopoly Models

CournotBertrandStackelberg

Auctions

Private valueCommon valueAffiliated value

Meyers (1988)

Matagorda County rice auctionsBids increased as # buyers increasedWinner’s curseBid shading

Relevant Tools to Study Auctions

TheoreticalHuman experimentsAgent-based models

Agent Based Models

Computerized agentsFollow specific rulesSimulation or optimization?

Stochastic Global Optimization

Evolutionary AlgorithmsGenetic algorithmsParticle swarm

Random Restarts

13

Market 1

particle 1 particle K Agent 1

Agent 2

Strategy 1

Local Best 1 Local Best K

Market K

Local Best 1 Local Best K

Global Best

Global Best

Profit

particle 1 particle K

Select New

strategy

Select New

strategyStrategy K

Convergence to Equilibrium

Software

• Many canned packages - Mostly land use• We use Java• Can be slow

Revisiting Mandatory Price Reporting Act

of 1999 Using an Agent-Based Model

Objective

• Develop an agent-based common-value auction to determine expected effect of Livestock Mandatory Price Reporting Act on packer market power

MPR Literature

• Previous theoretical work disagrees

• Recent econometric work shows beef packers’ market power increased after MPR

Theoretical Framework

• Previous theoretical work follows Cournot theory

• Auction model is closer to how cattle are actually purchased

Auction Structure

• First-price common-value auction with a reserve price

• Common value for the packers is 0.5

Post MPR Results

Category Full Noise

Seller Reduced

Noise

Buyer Reduced

Noise

Average winning bid price 0.02 0.11 0.06

Reserve price -0.31 -0.08 -0.43

Buyer 1 expected revenue 0.17 0.10 0.20

Buyer 2 expected revenue 0.16 0.10 0.18

Seller expected revenue 0.01 0.06 0.06

Application to Cattle Sales

• Reducing either the seller’s or buyers’ uncertainty through USDA-AMS reports created by MPR benefits the seller

Recommended