Product Quality Signaling in Experimental Markets

1
Product Quality Signaling in Experimental Markets

• ROSS M. MILLER CHARLES R. PLOTT Econometrica
  Jul 1985 53, 4
• Presented by Kelly Goldsmith
• 03/02/2005

2
Introduction

• My interest in Quality Signaling originally
  stemmed from the idea of companies Burning
  Money to signal their quality to customers

3
Example Super Bowl Commercials
4
Intuition

• People see ads during the Super Bowl and realize
  how expensive they are
• Even though the ads reveal little to no
  information about the product, consumers view
  them and believe the seller is high quality
• Why? The seller must be successful to afford
  such an ad and to be successful many others
  must have purchased the good!
• Implied Others would not have purchased the
  good, had it not been high quality

5
Product Quality Signaling Through Pricing Miller
Plott

• Generated and manipulated experimental markets
  to explore the possibility that sellers could
  extract information from the buyers by becoming
  aware of Quality Signals

6
The Experiment Sellers

• Possessed commodities exogenously designated as
  regular or super
• Though the grade of the commodity changed, at all
  times half the sellers had regulars and half
  had supers
• Could add units of quality which would be
  observed (and valued) by buyers
• Each unit of quality came at a cost to the seller
• Adding quality to a regular was more costly
  than adding quality to a super
• Sellers had two units to sell, so that total
  quantity was fixed
• After the period was over the grades of the
  sellers commodities were revealed

7
The Experiment Buyers

• Received a bonus for purchasing at least one
  commodity did not have budget restrictions
• They valued supers over regulars
• Had no information prior to purchase as to the
  grade of the commodity
• Had the incentive to look for Quality Signals to
  help them intuit the grade of the good
• All buyers had identical redemption value
  functions

8
The Experiment Buyers
? Redemption Value Function
9
Graph of Unit Demand and Cost
10
Experimental Design
11
Miller Plott Manipulations

• The authors tried a variety of manipulations
• 1) The cost of adding quality to a super at low
  v. high cost to the seller
• 2) The bonus for purchasing one unit
• 3) Prices were in Francs, not dollars the
  manipulated the value of a Franc
• 4) Degree to which the grade of the commodity
  (post sale) was made salient
• With all these different manipulations, several
  models were required to explain their results

12
Theoretical Models

• The Full Information Model
• The Naïve Model
• The Pure Pooling Models
• The Partial Pooling Model
• The Most Efficient Signaling Equilibrium and
  Rothschild-Stiglitz
• Inefficient Signaling Equilibria

13
Theoretical Predictions of Models

• Most models predict
• All regulars will sell at the same
  price/quality
• All supers will sell at the same price/quality

14
Todays Experiment

• Had a low cost for adding quality to a super
• Held the bonus for purchase constant
• Used dollars in stead of Francs (thus constant
  value)
• Made the grade of the commodity (post sale)
  salient

15
Their FindingsParallel Experiment - 2

• Their results found that Signaling Equilibrium
  models apply here
• Studied results as a function of Excess Value and
  Quality of sales

16
Discussion of Excess Value

• Excess Value X (q) P (q) V (R , q)
• The maximum possible loss a buyer can face when
  purchasing a unit of quality, q
• Reflects the buyers confidence that the unit is
  a super
• It is the amount paid over the units value as a
  regular

17
Experiment 2 Results
? Super
? Super
Mean Excess Value
Mean Quality
18
Conclusions from Parallel Experiments

• Quality Signaling occurs when the marginal cost
  of signaling a super is relatively low
• Signaling Models State The Excess Value of
  supers should be near 2.00 and the Excess Value
  of Regulars should be slightly below zero
• 2) Experiment 2 displayed these predicted
  equilibria (p.05)

19
Which Model Best Fits the Data?

• The Inefficient Signaling Equilibrium Model
• Predicts super quality above 27 units
• Direction of movement in Experiment 2 indicates
  quality levels are moving towards the most
  efficient signaling equilibrium

20
Conclusions

• Signaling is a real phenomenon
• The notion of equilibrium is appropriate
• The most efficient signaling equilibrium will
  emerge, if given time

21
Critique of Conclusions

• Essentially, the authors ran several very
  different experiments by using a variety of
  manipulations
• As it stands, it appears they use six different
  models to predict results when you have six
  models to chose from, one is bound to be correct!

22
Improvements on Design

• Relating to the earlier idea of burning money
  it would have been nice to see an analysis of how
  sellers made all their offers rather than just an
  analysis of the successful transactions
• Reputation Building While the paper gives a nod
  to the fact that, in an experiment such as this,
  reputations can be built it would be very
  interesting to run the experiment with anonymous
  sellers and see if the results were maintained

23
THANK YOU!

24
If Youre Interested in Learning More about
Quality Signaling

• Price and Advertising Signals of Product
  Quality by Milgrom Roberts (1986)
• An Introduction to Game Theory (Osborne) p. 350
  357
• The Theory of Industrial Organization (Tirole)
  p. 106 - 115