Transcript Social network analysis in business and economics

Social network analysis in
business and economics
Marko Pahor
Agenda
• What is social network analysis
• Short overview of social network analysis
techniques
• Applications of social network analysis in
business and economics
• Learning networks
• Ownership relations
Why do we need (social) network
analysis?
• Different types of data:
• Attribute (properties, opinions, behavior,…)
• Ideational (meanings, motives, definitions,…)
• Relational (contacts, ties, connections…)
• Different data needs different analysis
• Variable analysis for attribute data
• Typological analysis for ideational data
• Network analysis for relational data
What is social network analysis
• Network analysis is a series of techniques
(mathematical, statistical,…) designed to
analyze relation data
• Mathematically funded in graph theory
• Social network analysis is the application of
network analysis in the social sciences context
What are networks
• Imagine a closed set of units, call them actors or
nodes
• For example people, companies, web pages,
• Networks are a set of actors or nodes connected
by one or more relations
Social networks
• Thinking:
• … well, yes, but not really what social network
analysis is about
• Social networks is any network with social
entities (persons, groups, companies, social
events,…) as actors
Social network analysis techniques
• “Descriptive statistics” of networks
• Actors’ properties
• Network properties
• Statistical methods
• Blockmodelling
• Probability models
• One-time networks
• Dynamic networks
Descriptive statistics
• Properties of actors
• Measures, that describe the position and importance of
individual actors in the network
• E.g. Degree, betweeness,...
• Properties of network
• Describe the entire network
• Centalization, degree distribution, triadic census,...
Statistical methods of social network
analysis
• Blockmodelling
• A clustering methods
• Permutations of the adjacency matrix in order to find
some apriori expected blocks
• Probability models for one-time networks
• Modeling the probability of existence of a tie given
parameters
• Network parameters (e.g. reciprocity), covariates (e.g.
gender) and dyadic covariates (e.g. other relation)
• Probability models for dynamic networks
• Modeling the probability of creation or dissolution of a tie
given parameters
Applications of social network analysis
in business and economics
• Example 1: Organizational learning and through
learning networks
• Example 2: The evolution of the cross-ownership
network in Slovenia
The Network Perspective to
Organizational Learning – A
Comparison of Two
Companies
Organizational learning and learning
networks
• Organizational learning: individuals’ acquisition
of information and knowledge, analytical and
communicative skills
• Twodivergent perspectives for organizational
learning
• the acquisition perspective
• the participation perspective
• Elkjaer’s (2004) ‘third way’ - a synthesis of the
participation perspective and communities of
practice
• Critisim: too much emphasis on the participation
perspective and neglects some vital aspects of the
acquisition perspective
The learning network perspective
• The individual is recognized as the primary
source and destination for learning
• Learning takes place primarily in social
interaction
• The network perspective helps develop an
organizational learning culture
Learning networks
• External
• an extended enterprise model and comprise relationships
that a firm has with its customers, suppliers and other
stakeholders
• Internal
• a set of internal relationships among individual members
of the firm and other constituencies such as
product/service divisions and geographical units
• Components of learning networks
• learning processes
• learning structures
• actors
Propositions
P1: Learning in the network will mostly occur in relatively dense
clusters.
P2a: More experienced employees will be more sought after to learn
from.
P2b: More experienced employees will have less of a need to learn
from others.
P3a: People higher up the hierarchical ladder will be more sought after
to learn from.
P3b: People higher up the hierarchical ladder learn as much or even
more than those on lower levels.
P4a: An opportunity (working in the same location or in the same
business unit) will increase the probability of learning.
P4b: Homophily has an effect; it is more probable you will learn from
those who are similar in terms of gender, position, tenure...
Data – first company
• a software company
• 93 employees in three geographical units
• 81 employees participated in the study
• 59 from Ljubljana (Slovenia), 11 in Zagreb (Croatia) and
another 11 in Belgrade (Serbia)
• 56.7% of the respondents have a university
degree or higher (even one PhD)
• 74% of the respondents are male
• average tenure 38.9 months
Learning network in the first company
Data – second company
• main business engineering and production of
pre-fabricated buildings
• 860 employees, 470 of which on the main
location
• One production and several sales subsidiaries
• 348 employees from the main location
participated
• 59 % of respondents have finished high school,
29 % have a university degree
• 79% of the respondents are male,
• average tenure is 12.7 years
Learning network in the second
company
Methodology
• Network analysis is concerned with the structure
and patterning of these relationships
• Logistic model for social networks known as the
exponential random graph model (Snijders, 2002,
Snijders et al., 2004)
• What makes a learning tie more probable?
• Structural effects
• Actor covariate effects
• Dyadic covariate effects
Model 1
Model 2
Model 3
Model 4
Model 5
reciprocity
1.32 (0.24)
0.75 (0.29)
0.53 (0.26)
0.54 (0.3)
0.54 (0.31)
alternating out-k-stars, par. 2
-0.63 (0.23)
-0.6 (0.2)
-0.74 (0.22) -0.83 (0.26)
-0.8 (0.21)
alternating in-k-stars, par. 2
0.46 (0.13)
0.23 (0.15)
0.23 (0.14)
0.21 (0.16)
0.22 (0.17)
direct + indirect connections
1.24 (0.13)
0.15 (0.17)
0.2 (0.18)
0.26 (0.21)
0.17 (0.2)
Clu
steri
ng
alternating k-triangles, par. 2
1.14 (0.11)
0.8 (0.12)
0.76 (0.14)
0.79 (0.14)
alternating independent twopaths, par. 2
-0.2 (0.03)
-0.19 (0.02) -0.21 (0.03)
-0.2 (0.03)
Opp
ortu
nity
location (centered)
1.68 (0.26)
1.63 (0.26)
1.72 (0.23)
sector (centered)
0.62 (0.11)
0.81 (0.13)
0.72 (0.11)
Exp
erie
nce
tenure ego
0 (0.02)
-0.02 (0.02)
tenure alter
0.02 (0.01)
0.05 (0.02)
hierarchy ego
-0.04 (0.06)
-0.06 (0.07)
hierarchy alter
-0.12 (0.04)
-0.15 (0.06)
Gen
der
gender ego
0.26 (0.15)
gender alter
0.07 (0.13)
Homop
hily
Structural
effects
Effect
Seni
orit
y
Results – first company
tenure similarity
0.47 (0.16)
hierarchy similarity
0.78 (0.41)
gender identity
0.47 (0.22)
Model 1
Model 2
Model 3
Model 4
Model 5
reciprocity
1.38 (0.27)
1.4 (0.26)
1.33 (0.34)
1.27 (0.31)
1.22 (0.35)
alternating out-k-stars, par. 2
-0.47 (0.12)
-0.33 (0.12)
-0.55 (0.11)
-0.57 (0.13)
-0.6 (0.13)
alternating in-k-stars, par. 2
0.7 (0.08)
0.79 (0.08)
0.86 (0.09)
0.74 (0.09)
0.72 (0.09)
direct + indirect connections
1.63 (0.11)
0.77 (0.24)
0.92 (0.31)
0.75 (0.26)
0.93 (0.24)
0.73 (0.18)
0.66 (0.26)
0.82 (0.18)
alternating independent twopaths, par. 2
-0.13 (0.02)
-0.17 (0.03)
-0.17 (0.02)
-0.16 (0.02)
1.14 (0.11)
1.23 (0.11)
1.18 (0.12)
tenure ego
0.1 (0.09)
0.08 (0.1)
tenure alter
-0.05 (0.04)
-0.03 (0.07)
hierarchy ego
0.16 (0.21)
1.09 (0.93)
hierarchy alter
-0.02 (0.13)
0.86 (0.89)
Clus
terin
g
0.79 (0.3)
alternating k-triangles, par. 2
Exp
erie
nce
Effect
Seni
orit
y
Structural
effects
Results – second company
Homop
hily
Gen
der
Opportunity sector (centered)
gender ego
-0.31 (0.18)
gender alter
0.02 (0.09)
education similarity
0.04 (0.27)
hierarchy similarity
-1.84 (1.8)
gender identity
0.05 (0.15)
Discussion of results
• findings offers support for the network
perspective to organizational learning
• learning often occurs in project settings and
mainly involves the transfer of tacit knowledge
through participation
• a particular learning setting is dependent on
corporate culture and it is hard to capture all of
its parameters
Paths of Capital: The Creation and
Dissolution of the Slovenian
Corporate Network
Corporate networks
• Networks between corporate entities (companies)
• Different types of links
•
•
•
•
•
•
Interlocking directorates
Financial links
Strategic alliances
Cross-ownership
Multilink
…
Corporate networks configurations
• Corporate networks evolve through time
• Self-organized or guided
• The configuration of a network is a reflection of
the current situation (and historical path)
• Some configurations:
• Groups around financial centers (US, Mizruchi, 1982)
• Pyramidal structures (Belgium, Renneboog, 1997 and
1998)
• Cross-owned groups (keiretsu system in Japan, Gerlach,
1992)
• Sparse (“dismanteled”) network (Hungary, Stark, 2001)
The Slovenian corporate network
• Basically no corporations before 1992
• Socially (not state!) owned companies
• “Ownership allocation” (privatization)
• Began in 1992
• Was over by 1998
• “Voucher” privatization
• In 1998 almost no connection between (nonfinancial) corporation
• By the year 2000 a rather dense network and
growing
Data
• Ownership relations
• Owns a share in a public limited company
• Only non-financial companies
• 476 public limited companies
• Companies that existed in 2000 and their legal
successors
• Were connected at least once in the observed period
• 10 years (2000-2009), two observations per year
Changes in the network
• Network is evolving
• Changing links
• Changing composition
• Two distinct periods are visible
• Network creation period
• Dismantlement period
Network in 2000
Network in 2002
Network in 2004
Network in 2006
Network in 2008
Number of ties
ties
1000
900
800
700
600
500
400
300
200
100
0
Disconnected companies
350
300
250
200
150
100
50
0
Size of the largest strong component
45
40
35
30
25
20
15
10
5
0
How it happened?
• Some patterns can be observed
• Two phased process
• Preparatory phase
• Execution phase
• Coincides with changes in network
• Examples
• 2000 – 2004 comparisons for the first phase
• 2005 – 2009 comparisons for the second phase
First phase: Building a portfolio
Second phase: Cashing out
First phase: making a group
Second phase: Closing the deal
Findings
• Slovenian corporate network is dissolving after a
rise early in the decade
• Reason: ownership changes
• Shows how networks are used to gain control
Conclusions
• Social network analysis is an emerging
technique for the analysis of relations data
• Networks are everywhere
• Many possibilities for applications in business
and economics
• Interpersonal relations
• Interorganizational relations
• Marketing applications: products and customers networks