Expert Identification via Virtual Stock
Markets: Finding Lead Users in Consumer
Product Markets
Martin Spann, Holger Ernst,
Bernd Skiera, Jan Henrik Soll
DIMACS Workshop on Markets as Predictive Devices
Rutgers University, February 3rd 2005
Goethe-University, Frankfurt am Main (Germany)
WHU-Otto Beisheim GSM, Vallendar (Germany)
Objective of the Presentation
 Application of virtual stock markets to marketing
research
 Outline the rationale for virtual stock markets as an
efficient tool for expert identification
 Present results of an empirical study
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
2
Marketing Research Applications
 Business Forecasting
•
•
•
Company revenues and product sales
Market shares
Return of strategic investments (e.g. technology)
 New Product Development
•
•
Evaluation of Product Concepts
Identification of lead users
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
3
Marketing Research via VSM
Information
Source VSM
Traders' individual portfolios
Stock Prices
Forecasting
Events in near Future
Forecasting
Alternatives
• Chen/Plott (2002)
• Chan et al. (2002)
• Forsythe et al. (1992)
• Hanson (1992)
Expert
Identification
• This study
• Forsythe/Rietz/Ross (1999)
• Pennock et al. (2000)
• Spann/Skiera (2003)
• Wolfers/Zitzewitz (2004)
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
4
Lead User Concept
 von Hippel (1986): Lead Users
•
•
face future product needs months or even years
earlier than normal customers
often try to find solutions to these needs by
themselves
 Analysis of lead users can detect these future
needs and obtain new product ideas to satisfy
these needs
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
5
Rationale for Virtual Stock Markets
as a Tool for Lead User Identification
 Idea:
Participant's performance at a virtual stock market is
an indicator of knowledge about event to be predicted
Two effects permit identification:
 Self-selection effect
Attraction of participants who display higher
involvement with the product
 Performance effect
•
•
Successful participants are more knowledgeable,
because they detect and exploit inefficient prices
Inefficient prices = incorrect predictions
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
6
Empirical Study



Goal: Analyze feasibility of VSMs for lead user
identification
Methodology: Analyze participants at an VSM
according to performance and lead user
characteristics
Application to success forecasting of movies:
•
•
•
•
Relevant for producers and exhibitors
Each movie is a new product with high failure rates
Movies increasingly rely on branding (e.g. sequels)
Value chain: theaters, rental, sale and merchandize
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
7
Design of Empirical Study:
Two Phases
1. Setup of VSM for the prediction of the box-office
success of movies in Germany
•
•
•
•
Forecast of number of movie visitors
6 rounds with total number of 350 participants
70 movies (release between May and October 2001)
Participant's performance: Mean Portfolio increase in
active rounds
2. Online-survey for lead user characteristics (after
end of VSM; lottery of gift vouchers as incentive):
•
•
•
•
Opinion leadership
Expertise
Expected Benefit
Survey response rate of 29.2% (n=102)
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
8
Proportion of Lead Users among
Traders
 Identification of lead users by threshold levels
(=sample mean) of each factor: opinion leadership,
expertise and expected benefit
 Result: 20.6% (=21) of respondents fulfill required
level of the three lead user criteria
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
9
Are Top Traders more likely to be
Lead Users? (1/2)
 Factor scores of top and bottom traders
Mean Scores
Bottom 80%
Best 20%
Opinion Leadership
3.41
3.73*
Expertise
2.57
3.03*
Expected Benefit
3.16
3.40
Notes: * Differences significant at .1-level (t-test for independent samples).
Mean values per group are given in the table.
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
10
Are Top Traders more likely to be
Lead Users? (2/2)
Not a lead user
Lead user
Total
Frequency
Proportion
Expected Frequency
Residual
Frequency
Proportion
Expected Frequency
Residual
Frequency
Bottom 80%
of Traders
Top 20%
of Traders
Total
69
85.19%
65.12
3.88
13
61.90%
16.88
-3.88
82
12
14.81%
15.88
-3.88
8
38.10%
4.12
3.88
20
81
100%
81
21
100%
21
102
Note: Fisher-Test for group differences is significant at the .05 level.
 Significant relationship between performance and
frequency of lead users
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
11
How differ successful and non
successful Lead Users?
 Factor scores of top and bottom performing lead
users
Lead Users among
Bottom 80% of Traders
Lead Users among
Best 20% of Traders
Opinion Leadership
4.04
3.94
Expertise
3.31
3.88*
Expected Benefit
3.85
4.06
Mean Scores
Notes: * Differences significant at .1-level (t-test for independent samples).
Mean values per group are given in the table.
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
12
How do Lead Users achieve higher
Performance?
 Hypothesis: Lead users exploit every perceived
price inefficiency
 conduct significantly more orders and trades than
non lead users:
•
•
•
No. of trades per active round: 47.71 for lead users
(20.90 for non lead users)
No. of orders per active round: 45.00 for lead users
(19.35 for non lead users)
Differences significant at 1% level (t-test for
independent samples and ANOVA)
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
13
Discussion and Limitations
 Not all lead users perform well at VSM
•
•
not all lead users can translate assessment of unmet
needs into success forecast of product
VSM selects those lead users with better market
understanding  most desired ones to integrate into
new product development process
 Only one product category
 Limited availability of benchmark studies
(proportion of lead users in consumer markets)
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
14
Managerial Implications
 Results show feasibility of VSM for efficient lead
user identification
 Possible double benefit of VSM: forecasting and lead
user identification
 Identified lead users can be used for in-depth
studies:
•
•
•
Interviews
Idea generation
Concept testing
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
15
Contact
 Martin Spann
School of Business and Economics
Johann Wolfgang Goethe-University
60054 Frankfurt am Main (Germany)
Phone: +49-(0)69-798-22380
Fax: +49-(0)69-798-28973
E-Mail: spann@spann.de
www.virtualstockmarkets.com
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
16
Items to Measure Lead User
Characteristics
Construct
Measurement item
Opinion
Leadership:
- When you talk to your friends about movies, do you give:
very little information – a great deal of information (OL1)
- In a discussion of movies, you would most likely:
listen to your friends' ideas – convince your friends of your ideas (OL2)
- In discussions of movies, which of the following happens most often?
Your friends tell you about movies – you tell your friends about movies (OL3)
- Overall in all of your discussion with friends and neighbors, are you:
not used as a source of advice – often used as a source of advice (OL4)
Expertise:
- Movies are very important to me compared to my other hobbies. (E1)
- Movies consume a large portion of my free time in relation to other hobbies. (E2)
- I frequently go to the movies. (E3)
Expected
Benefit:
- I am dissatisfied with the movies screened recently at theatres. (EB1)
- I would go to the movies more frequently if they would better meet my
expectations. (EB2)
Note: All items except for the opinion leadership construct are Likert-type five point scales using "fully agree" and "do
not agree" as anchors. The individual anchors for the opinion leadership scale are listed above.
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
17
Dimensions of Lead User
Characteristics
 Exploratory FA
Items
Opinion Leadership
OL1
.81
OL2
.88
OL3
.71
OL4
.61
E1
.28
E2
.27
E3
.10
EB1
.05
EB2
-.08
Variance explained
.42
Cronbach’s 
.86
Expertise
.25
.04
.39
.27
.75
.85
.91
-.03
-.08
.16
.84
Expected Benefit
.01
.09
-.07
-.35
-.10
-.06
-.04
.87
.78
.13
.59
Communality
.71
.78
.67
.56
.65
.79
.83
.77
.63
Note: Principal Component Analysis (Eigenvalues >1), Varimax-Rotation, N=102.
 Satisfactory results of confirmatory FA:
GFI = .88, CFI = .89, RMSEA = .13
"Expert Identification via VSMs"
Spann/Ernst/Skiera/Soll, Goethe-University,
Frankfurt am Main & Vallendar, Germany © 2005
18