Note

0.0(0)

Take a practice test

Chat with Kai

Explore Top Notes

Membranes- Chapter 3.1

Studied by 787 people

5.0(1)

Chapter 4:Molecular compounds

Note

Studied by 15 people

5.0(1)

AP Physics 1: Ultimate Guide (copy)

Note

Studied by 170 people

5.0(1)

EWOT - 12. Externalities and conflicting rights

56d ago

Guala 2001

Building Economic Machines: The FCC Auctions Introduction Background of the FCC Auction Mechanism Design The Role of Theory Constructing and Controlling ‘Microeconomic Systems’Testing the Robustness of Alternative Designs Testing the Rules Checking External Validity Rational Choice Technology

Building Economic Machines: The FCC Auctions

Introduction

The FCC auctions, started in 1994 to sell spectrum licenses, are a great example of economic engineering.
This article explains how the FCC auctions were designed, tested, and put into action.
It shows how important game theory and experiments were in the process.
It begins and ends by discussing how we use rational choice models.
It also emphasizes how important it is to use science to understand scientific theories, especially in social sciences.
It focuses on auction theory, which is closely related to game theory and its applications.

Background of the FCC Auction

In the late 1980s, the American economy started to become more decentralized.
Before this, licenses for wireless Personal Communication Systems (PCS) were given out through administrative hearings.
This process was slow, complicated, not transparent, and the licenses were given away for free.
In 1982, Congress changed the system to use lotteries, which randomly assigned licenses.
The lottery system was faster, but companies started reselling licenses for profit, creating a secondary market.
In July 1993, Congress decided to replace the lottery system with a market institution.
The FCC had to find and implement the best auction system within ten months.
Previous experiences in New Zealand and Australia showed that a poorly planned reform could be a disaster.
Economists and game theorists were hired to help design the auction system.
In September 1993, the FCC released a ‘Notice of Proposed Rule Making’ that outlined the goals of the auctions, suggested a design, and asked for feedback.

Aims of the Auction

Efficient allocation: To give spectrum rights to the companies that valued them most.
Prevent monopolies.
Promote small businesses, rural telephone companies, and firms owned by minorities and women.
Maximize revenue for the auctioneer (the FCC).

Mechanism Design

The FCC's situation is a typical ‘mechanism design’ problem.
Mechanisms are structures that generate 'processes'.
This case shows how we use models to create mechanisms and processes, using theory to shape the social world to match a model.
Mechanism design aims to replace outdated regulation systems with better ones.
It involves studying how different institutions work and evaluating them.
Rational choice theory is a useful tool, where institutional rules define a game that agents try to solve rationally.
Ideally, we should be able to predict the outcome of a mechanism by analyzing its equilibrium.

The Role of Theory

The FCC auction was considered a major success because the Federal Government gained a lot of money from it -- $23 billion$ $$23 billion$$ between 1994 and 1997.
It was also claimed to be a triumph for game theory and game theorists.
Auctions like the PCS auctions are examples of what game theory isn't very good at modeling.
Game-theoretic accounts of auction mechanisms date back to the sixties, thanks to William Vickrey (1961).
Vickrey solved an auction game known as the ‘independent private values model’.
Wilson (1977) and Milgrom and Weber (1982) expanded Vickrey’s private value model to other cases.
Auctions are modeled as games where bidders try to maximize their expected utility.
The players use equilibrium strategies, in the standard sense of a Bayes-Nash equilibrium.
After twenty years of theoretical development, auction theory still relies on restrictive assumptions and cannot be applied to all cases.

Complementarities

A key feature of PCS licenses is that they are ‘complementary’ and ‘perfect substitutes’ for each other, meaning the value of one license depends on owning others.
Complementarities mean that the value of a ‘package’ can be different from the sum of the values of its individual items.
Complementarities are problematic for economists because models of competitive markets with such goods usually don't have a unique and stable equilibrium.
No theorem in auction theory can tell you what institution will achieve an efficient outcome in these cases; the theory is incomplete.
In September 1993, the FCC suggested a two-stage auction system (a ‘combinatorial auction’).
Paul Milgrom and Robert Wilson, and Preston McAfee, proposed an alternative mechanism called ‘simultaneous ascending-bid auction’.
In a simultaneous auction, several markets are open at the same time, allowing bidders to operate in all of them at once.
In an ascending auction, bidders continue to make offers until the market closes.
Simultaneity and the ascending form allow bidders to gather valuable information.

Milgrom and Wilson's Arguments

Milgrom and Wilson argued that a combinatorial institution might lead to free-riding.
For example, consider three bidders: One, Two, and Three. One is willing to pay $4$ $$4$$ for item A, Two is willing to pay $4$ $$4$$ for item B, and Three is willing to pay $1 + ε$ $$1 + ε$$ for A, $1 + ε$ $$1 + ε$$ for B, and $2 + ε$ $$2 + ε$$ for AB (with $ε$ $$ε$$ being small and positive).
The payoffs are represented in the following matrix:

	A	B	AB
One	$4$ $$4$$	—-	—-
Two	—-	$4$ $$4$$	—-
Three	$1+ε$ $$1+ε$$	$1+ε$ $$1+ε$$	$2+ε$ $$2+ε$$

The most efficient allocation is to assign A to One and B to Two.
Under the two-stage combinatorial design, One and Two have an interest in raising the total value of A and B by bidding more on at least one of them.
Milgrom and Wilson showed that such free-riding has an inefficient mixed-strategy equilibrium.
Bidders One and Two each face a sub-game represented by the following payoff matrix:

	Raise bid	Don’t raise
Raise bid	2,2	2,3
Don’t raise	3,2	0,0

By backward induction, this sub-game has an equilibrium where each bidder plays ‘Raise the bid’ with probability $2/3$ $$2/3$$ and ‘Don’t raise’ with probability $1/3$ $$1/3$$.
There is a $1/9$ $$1/9$$ probability of Three getting both A and B by paying just $1/4$ $$1/4$$ of what bidders One and Two would jointly pay for them.

- This argument relies on independent theoretical insights and analyses of how players behave when solving tasks in isolation.

Constructing and Controlling ‘Microeconomic Systems’

Experimental economics, which emerged after World War II, involves the idea of a ‘microeconomic system’.
A ‘correctly performed’ experiment is one that the experimenter can control and interpret properly.
In 1993, Pacific Bell hired economists from Caltech, led by Charles Plott, to run experiments.
Initially, the experiments helped choose between the ‘combinatorial’ and ‘continuous ascending’ auctions.
Experiments allow us to check if an institution allocates goods to those who value them most.

Experimental Testbeds

The Caltech team used ‘testbed’ experiments.
These must meet the requirements of ‘Induced Value Theory’, developed by Vernon Smith (1976, 1982, 1987).
Induced Value Theory includes five ‘precepts’ for a valid controlled microeconomic experiment (Smith, 1982, p. 261).
Two of them, saliency and non-satiation, help us assert that we have created a microeconomic system in the laboratory (Smith, 1987, p. 108).
A ‘microeconomic system’ is defined by the characteristics of the agents and the institution regulating their decisions.
The saliency requirement states that the outcomes of an experiment should depend only on the agents’ decisions, and the rewards should increase with the outcomes.
Non-satiation requires setting rewards so that agents always prefer more rewards to less.
Privacy and dominance impose constraints on agents.

Testing the Robustness of Alternative Designs

In early 1994, the Caltech group ran comparative efficiency tests of the simultaneous ascending auction versus a combinatorial sealed-bid plus sequential continuous design.
Mechanisms were tested in the laboratory for the first time.
Experimentalists found problems that theorists had not anticipated.
One problem with a combinatorial procedure is that some bidders stay in the Japanese auction above their reservation price to raise prices and overcome a sealed-bid pre-offer, causing a bubble effect.

Robustness

Robustness: The combinatorial auction is difficult to implement correctly.
Environmental Robustness vs Personality Robustness.
Environmental robustness: An institution’s ability to work properly in different environments.
Personality robustness: Its ability to work with real players who may behave differently from the rational agents in game-theoretic models.
Testbed experiments helped identify moments when subjects needed help from the auctioneer to understand details of the auction.
Laboratory experiments also helped develop the software.
Data from testbed experiments were used as inputs for the final software used in the real FCC auctions.
Trained students were used to investigate the software's properties.

Testing the Rules

Experiments can help move from theoretical ideas to the real world while maintaining the desired properties of a mechanism.
The most important rules concerned increments, withdrawals, eligibility, waivers, and activity.
Each bidder must deposit money proportional to the licenses they want to bid for, establishing their ‘initial eligibility’, and bidders are incentivized to maintain a certain level of activity.
To account for mistakes, bidders have five ‘waivers’ of the activity rules.
Bidders can withdraw a bid but risk paying the difference if the final selling price is lower than their withdrawn bid.
The activity rules were designed to prevent bidders from strategically slowing down the auction.
The length of an auction depends on the number of rounds and the interval between rounds.
Experiments tested different rules with varying intervals between rounds.
Big increments sometimes eliminated bidders too quickly, reducing their eligibility and creating a ‘demand killing’ effect.
The Milgrom–Wilson–McAfee rules also allowed withdrawals, which could cause ‘cycles’.
Experiments were used to create cycles in the laboratory.

Checking External Validity

Consultants monitored a real FCC auction in Washington, DC, in October 1994.
The expertise gained in the lab was invaluable, especially when intervention was needed.
Laboratory tests with similar parameters to the real auction were run beforehand to compare the results afterward.
Data collected in the lab and the real auction were systematically analyzed and compared.
The external validity argument uses background assumptions and observed data to understand the underlying process.

Rational Choice Technology

The FCC auctions were scientifically successful.
This case study shows that economic engineers must design mechanisms considering individuals’ real capacities.
Bidding teams

Note

0.0(0)

Take a practice test

Chat with Kai

Explore Top Notes

Membranes- Chapter 3.1

Studied by 787 people

5.0(1)

Chapter 4:Molecular compounds

Note

Studied by 15 people

5.0(1)

AP Physics 1: Ultimate Guide (copy)

Note

Studied by 170 people

5.0(1)

EWOT - 12. Externalities and conflicting rights

Guala 2001

Building Economic Machines: The FCC Auctions

Introduction

The FCC auctions, started in 1994 to sell spectrum licenses, are a great example of economic engineering.
This article explains how the FCC auctions were designed, tested, and put into action.
It shows how important game theory and experiments were in the process.
It begins and ends by discussing how we use rational choice models.
It also emphasizes how important it is to use science to understand scientific theories, especially in social sciences.
It focuses on auction theory, which is closely related to game theory and its applications.

Background of the FCC Auction

In the late 1980s, the American economy started to become more decentralized.
Before this, licenses for wireless Personal Communication Systems (PCS) were given out through administrative hearings.
This process was slow, complicated, not transparent, and the licenses were given away for free.
In 1982, Congress changed the system to use lotteries, which randomly assigned licenses.
The lottery system was faster, but companies started reselling licenses for profit, creating a secondary market.
In July 1993, Congress decided to replace the lottery system with a market institution.
The FCC had to find and implement the best auction system within ten months.
Previous experiences in New Zealand and Australia showed that a poorly planned reform could be a disaster.
Economists and game theorists were hired to help design the auction system.
In September 1993, the FCC released a ‘Notice of Proposed Rule Making’ that outlined the goals of the auctions, suggested a design, and asked for feedback.

Aims of the Auction

Efficient allocation: To give spectrum rights to the companies that valued them most.
Prevent monopolies.
Promote small businesses, rural telephone companies, and firms owned by minorities and women.
Maximize revenue for the auctioneer (the FCC).

Mechanism Design

The FCC's situation is a typical ‘mechanism design’ problem.
Mechanisms are structures that generate 'processes'.
This case shows how we use models to create mechanisms and processes, using theory to shape the social world to match a model.
Mechanism design aims to replace outdated regulation systems with better ones.
It involves studying how different institutions work and evaluating them.
Rational choice theory is a useful tool, where institutional rules define a game that agents try to solve rationally.
Ideally, we should be able to predict the outcome of a mechanism by analyzing its equilibrium.

The Role of Theory

The FCC auction was considered a major success because the Federal Government gained a lot of money from it -- $23 billion$ between 1994 and 1997.
It was also claimed to be a triumph for game theory and game theorists.
Auctions like the PCS auctions are examples of what game theory isn't very good at modeling.
Game-theoretic accounts of auction mechanisms date back to the sixties, thanks to William Vickrey (1961).
Vickrey solved an auction game known as the ‘independent private values model’.
Wilson (1977) and Milgrom and Weber (1982) expanded Vickrey’s private value model to other cases.
Auctions are modeled as games where bidders try to maximize their expected utility.
The players use equilibrium strategies, in the standard sense of a Bayes-Nash equilibrium.
After twenty years of theoretical development, auction theory still relies on restrictive assumptions and cannot be applied to all cases.

Complementarities

A key feature of PCS licenses is that they are ‘complementary’ and ‘perfect substitutes’ for each other, meaning the value of one license depends on owning others.
Complementarities mean that the value of a ‘package’ can be different from the sum of the values of its individual items.
Complementarities are problematic for economists because models of competitive markets with such goods usually don't have a unique and stable equilibrium.
No theorem in auction theory can tell you what institution will achieve an efficient outcome in these cases; the theory is incomplete.
In September 1993, the FCC suggested a two-stage auction system (a ‘combinatorial auction’).
Paul Milgrom and Robert Wilson, and Preston McAfee, proposed an alternative mechanism called ‘simultaneous ascending-bid auction’.
In a simultaneous auction, several markets are open at the same time, allowing bidders to operate in all of them at once.
In an ascending auction, bidders continue to make offers until the market closes.
Simultaneity and the ascending form allow bidders to gather valuable information.

Milgrom and Wilson's Arguments

Milgrom and Wilson argued that a combinatorial institution might lead to free-riding.
For example, consider three bidders: One, Two, and Three. One is willing to pay $4$ for item A, Two is willing to pay $4$ for item B, and Three is willing to pay $1 + ε$ for A, $1 + ε$ for B, and $2 + ε$ for AB (with $ε$ being small and positive).
The payoffs are represented in the following matrix:

	A	B	AB
One	$4$	—-	—-
Two	—-	$4$	—-
Three	$1+ε$	$1+ε$	$2+ε$

The most efficient allocation is to assign A to One and B to Two.
Under the two-stage combinatorial design, One and Two have an interest in raising the total value of A and B by bidding more on at least one of them.
Milgrom and Wilson showed that such free-riding has an inefficient mixed-strategy equilibrium.
Bidders One and Two each face a sub-game represented by the following payoff matrix:

	Raise bid	Don’t raise
Raise bid	2,2	2,3
Don’t raise	3,2	0,0

By backward induction, this sub-game has an equilibrium where each bidder plays ‘Raise the bid’ with probability $2/3$ and ‘Don’t raise’ with probability $1/3$ .
There is a $1/9$ probability of Three getting both A and B by paying just $1/4$ of what bidders One and Two would jointly pay for them.

- This argument relies on independent theoretical insights and analyses of how players behave when solving tasks in isolation.

Constructing and Controlling ‘Microeconomic Systems’

Experimental economics, which emerged after World War II, involves the idea of a ‘microeconomic system’.
A ‘correctly performed’ experiment is one that the experimenter can control and interpret properly.
In 1993, Pacific Bell hired economists from Caltech, led by Charles Plott, to run experiments.
Initially, the experiments helped choose between the ‘combinatorial’ and ‘continuous ascending’ auctions.
Experiments allow us to check if an institution allocates goods to those who value them most.

Experimental Testbeds

The Caltech team used ‘testbed’ experiments.
These must meet the requirements of ‘Induced Value Theory’, developed by Vernon Smith (1976, 1982, 1987).
Induced Value Theory includes five ‘precepts’ for a valid controlled microeconomic experiment (Smith, 1982, p. 261).
Two of them, saliency and non-satiation, help us assert that we have created a microeconomic system in the laboratory (Smith, 1987, p. 108).
A ‘microeconomic system’ is defined by the characteristics of the agents and the institution regulating their decisions.
The saliency requirement states that the outcomes of an experiment should depend only on the agents’ decisions, and the rewards should increase with the outcomes.
Non-satiation requires setting rewards so that agents always prefer more rewards to less.
Privacy and dominance impose constraints on agents.

Testing the Robustness of Alternative Designs

In early 1994, the Caltech group ran comparative efficiency tests of the simultaneous ascending auction versus a combinatorial sealed-bid plus sequential continuous design.
Mechanisms were tested in the laboratory for the first time.
Experimentalists found problems that theorists had not anticipated.
One problem with a combinatorial procedure is that some bidders stay in the Japanese auction above their reservation price to raise prices and overcome a sealed-bid pre-offer, causing a bubble effect.

Robustness

Robustness: The combinatorial auction is difficult to implement correctly.
Environmental Robustness vs Personality Robustness.
Environmental robustness: An institution’s ability to work properly in different environments.
Personality robustness: Its ability to work with real players who may behave differently from the rational agents in game-theoretic models.
Testbed experiments helped identify moments when subjects needed help from the auctioneer to understand details of the auction.
Laboratory experiments also helped develop the software.
Data from testbed experiments were used as inputs for the final software used in the real FCC auctions.
Trained students were used to investigate the software's properties.

Testing the Rules

Experiments can help move from theoretical ideas to the real world while maintaining the desired properties of a mechanism.
The most important rules concerned increments, withdrawals, eligibility, waivers, and activity.
Each bidder must deposit money proportional to the licenses they want to bid for, establishing their ‘initial eligibility’, and bidders are incentivized to maintain a certain level of activity.
To account for mistakes, bidders have five ‘waivers’ of the activity rules.
Bidders can withdraw a bid but risk paying the difference if the final selling price is lower than their withdrawn bid.
The activity rules were designed to prevent bidders from strategically slowing down the auction.
The length of an auction depends on the number of rounds and the interval between rounds.
Experiments tested different rules with varying intervals between rounds.
Big increments sometimes eliminated bidders too quickly, reducing their eligibility and creating a ‘demand killing’ effect.
The Milgrom–Wilson–McAfee rules also allowed withdrawals, which could cause ‘cycles’.
Experiments were used to create cycles in the laboratory.

Checking External Validity

Consultants monitored a real FCC auction in Washington, DC, in October 1994.
The expertise gained in the lab was invaluable, especially when intervention was needed.
Laboratory tests with similar parameters to the real auction were run beforehand to compare the results afterward.
Data collected in the lab and the real auction were systematically analyzed and compared.
The external validity argument uses background assumptions and observed data to understand the underlying process.

Rational Choice Technology

The FCC auctions were scientifically successful.
This case study shows that economic engineers must design mechanisms considering individuals’ real capacities.
Bidding teams