Markets as ConversationsMaking the Invisible Hand Visible
Robert LuschDaniel Zeng
Hope Jensen Schau
Markets As …
Main Thesis
“Invisible hand” assumes a machine-like economy where unseen co-ordination seeks an equilibrium for supply and demand.
Web 2.0 technology (e.g., social bookmarking, social networks, blogs) makes market actors’ coordination efforts manifest in certain contexts.
Our research reveals examples of technologically mapping market conversations at particular critical moments in time, and glimpsing the no longer invisible hand.
All Actors are Resource Integrators
Service-Dominant Logic’s (Vargo and Lusch 2004, 2008) foundational premise #9 states "All social and economic actors are resource integrators” meaning that market actors gather together resources from disparate sources to satisfy needs.
So, with a critical event we would expect to see even more resource integration.
Critical EventsExogenous Events: outside the systemMultiple Stressors: high-impact Domain: Regional, National, Global
Influence a Market System
Theoretical FoundationChaos theory suggests how certain
conditions can change and permanently alter a system. Like the butterfly effect, critical events change the course of a system because of the initial conditions that underlie the system. Piotrowski 2006: Hurricane Katrina and
the implications of chaos theoryAlesch 2004: Theory of Disaster Recovery
This would suggest that each change is independent initially but that it disrupts the system.
Theoretical FoundationPrior literature on macroeconomics
reveals that critical events drive short term and even lasting impacts on people’s interests and discourse surrounding a domain (social production).Hall 2005: adverse macroeconomic effects
of oil price increasesZivot and Andrews 2002: examine short
and long term impacts of Great Crash of 1929 and the of Oil Price Shock of 1973
This suggests that that a critical economic event impacts exchange systems in the short run and changes market discourse in the long run.
Assumption 1Web 2.0 platforms such as
social bookmarking and community question and answer can illuminate the trajectory of market discourse.Web 2.0 provides multiple
modes of market discourse, which captures the integration of resources by market actors.
Assumption 2Mapping the impact of critical events through
social bookmarking and community question and answer is a Natural Experiment Instrument Tags used/posted by and the questions/answers
from the users provide an unbiased representation of their information-seeking needs and their resource integration efforts.
This means that we can trace the impact of a critical event on resource integration through market discourse observed on Web 2.0 platforms.
Alexa Web Traffic Ranking: Dec. 2005 vs. March 2010
World-wide Share of Online Time
35%
22%
16%
8%
14% 6%
All Other Communications
Social Connections Shopping & Travel
Entertainment & Leisure Work, Business & Education
Category didn’t exist three years ago
11
• In August 2009, 17 percent of all time spent on the Internet was at social networking sites, up from 6 percent in August 2008.
• Estimated online advertising spending on the top social network and blogging sites increased 119 percent, from approximately $49 million in August 2008 to approximately $108 million in August 2009.
Case Study I: Social Bookmarking
• A widely-adopted Web 2.0 technology
“Delicious is a social bookmarking website, which means it is designed to allow you to store and share bookmarks on the web, instead of inside your browser.” － delicious.com
Share Links
13
Delicious Social Networks
14
Delicious Dataset UsedTagging history of 43.5K users covering the period
1/1/2009 --- 11/5/2009, with 3.2M URLs, 0.6M tags, and 17M tagging activities
Key data elementsTags used; Web pages/URLs tagged; timing of tagging; user id
Events studied:H1N1 outbreak (April 2009)Michael Jackson’s death (June 25, 2009)Windows 7 launch (October 22, 2009)
Analysis I: Weekly Measures Studied
– Intensity: # of tagging activities– Reach: # of users involved– Depth: # of related URLs– Impact:
• Indicators for peaks in measurement• Repercussions after peaks
H1N1
Michael Jackson’s Death
Windows 7 Launch
Analysis II: 2-day Measures
– Burstiness： % change in interest / time needed to go from ½ (peak+average) to peak
– Half-life： time needed to go from peak to ½ (peak+average)
– Width： time between two adjacent ½ (peak+average) points
– Persistency： time in which interest in an event is higher than the average level
– Lag： time between occurrence of an event and responses on the tagging site
FindingsEvent Bursti-
ness (%/hour)
Width (day)
Half-life（ day）
Persistency (day)
Lag (day)
H1N1 7.2 6 3 14 ~15
Michael Jackson
0.7 14 3 20 1~2
Windows 7
2.9 8 3 180 -180
Findings• All these curves follow cyclic trends with
different periodicities• More cycles indicate richness of the events
– Michael Jackson’s death involved several sub-topics and resulted in ups-and-downs in interest
• For unexpected events such as H1N1 and Michael Jackson’s death, show clear spikes immediately
• For predictable events such as the Windows 7 launch, the spikes occur in cycles before event and spike with event
Case Study II: Community Question and Answer (CQA)
• Another widely-adopted Web 2.0 technology
Yahoo! Answer Statistics
Research Agenda• For critical incident and market disruptions,
how can we capture the coordination and market conversation?– How does a user community’s interest of a
product (community topic) change over time?– How do the community topics serve as an eWOM
branding mechanism?
Dataset Used
• All the resolved questions about iPhone (a market disruption) and their associated answers during the period of May 26, 2009 to Aug. 20, 2009, from Yahoo! Answers, with more than 16,000 question threads.
• From each question thread, we extracted the following fields– asker, question context, question timestamp, and
associated answers– each associated answer covering answer content,
answerer who posted the answer, and timestamp
# of Questions over Time
0 10 20 30 40 50 60 70 800
100
200
300
400
500
600
time offset (days)
num
ber
of q
uest
ions
Jun. 17th
Jun. 9th
May 26th
June 9th is the Apple Worldwide Developer Conference: 3G iPhone is announcedJune 17th: iPhone software upgrade
Google Trend Data for the Same Time Period
Constructing CQA Interactive Network
• Each unique user id in the dataset as a vertex• An edge <i,j> indicating that a question posed
by user i was answered by userj • Each edge <i,j> in the network weighted by
the number of i ’s questions that were answered by j
Example Communities
• User u3 contributed to the community through answering while u5 never answered questions from other users. Other users performed both activities.
Identifying Overlapping Communities
• After training with LDA, the weighed user adjacent matrix is split into the asker-community matrix and answerer-community matrix – Elements of the asker-community matrix
indicating the probability distribution over asker i given community k
– Elements of the answer-community matrix indicating the probability distribution over answerer j given community k.
Identified Threads• Thread a: syncing w/ iTunes• Thread b: contract and pricing• Thread c: releasing of new iPhone OS• Thread d: comparison w/ other smart phones• Thread e: data plan and jailbreak• Thread f: signing and upgrading service
contract• Thread g: signing contract for iPhone 3GS• Thread h: basic usage
Community Evolutionary Pattern Identified by LDA
5 Stages of iPhone Discussions• Value-in-Exchange Discussions May 26th—Jun.
11th: Discussions focused on signing contract and the price of upgrading iPhone. Keywords such as “contract”, “price”, “upgrade”, “data”, “plan” ranked highly in most communities.
• Use Discussions Jun. 12th—Jun. 28th: Discussions focused on transferring video and music from computer to iPhone and the scheduled release of iPhone OS 3.0. Keywords such as “release”, “June”, “check”, “iTunes”, “video”, “transfer”, ranked highly in communities.
5 Stages (Cont’d)
• Competitive Value Propositions Jun. 29th—Jul. 15th: Discussions focused on comparing the newly released iPhone with other smart phones, particularly with Palm Pre. The AT&T mobile network was also often compared with those from other service providers such as Verizon, and received many complaints. Keywords such as “network”, “verizon”, “palm”, ranked highly.
5 Stages (Cont’d)• Specific Device Attributes Discussions Jul. 16th—
Aug. 01th: Discussions focused more on usage and performance. Keywords such as “touch”, “keyboard”, “battery”, “network” ranked highly.
• Resource Integration Discussions Aug. 02th—Aug 20th: Discussions focused on more advanced usage of iPhone, e.g., converting videos or music in any format to iPhone. Also many users started to exchange free softwares for jailbreak iPhone. Keywords such as “convert”, “jailbreak”, “unlock” ranked highly.
SummaryObservable market discourse allows us to
empirically study the coordination efforts involved in the “invisible” market-making processMarkets as Conversations
Mapping market discourse through examining social bookmarking and CQA is a form of Natural Experimentation.
Initial empirical findings suggest rich and meaningful patterns can be automatically mined from a range of Web 2.0 platforms
Ongoing ResearchExtensive econometric testingNon-trivial linkage to existing theory and
possible development of new theoryIdentifying and classifying different types
and patterns of impact of eventsCase studies in marketing with specific
managerial implications
