The Inner Circle Revisited

1
The Inner Circle Revisited

• Using exponential random graph models (ERGM) to
  study the political activity of corporate elites.

Nicholas Harrigan Political Science
Program Research School of Social
Sciences Australian National University Visitor C
entre for Research Methods in the Social
Sciences Department of Politics and International
Relations University of Oxford.
2
Introduction

• Michael Useem (1984) The Inner Circle
• Linked Politics and Interlocking Directorates
• Multiple Directorships Political Leadership
• Limitations
• statistics cross tabs
• no controls
• The Puzzle Multiple directorships correlate with
  a lot of important characteristics of
  corporations and directors, but which
  characteristics most important?

3
Purpose of this paper

• Untangle the correlations in Inner Circle What
  are the most important forces driving corporate
  interlocking?
• Apply p model/exponential random graph (ERG)
  model to corporate network.

4
Structure of this paper

• Overview of The Inner Circle theory (and
  related hypotheses)
• Brief introduction to ERG (p) models
• Preliminary results

5
The Inner Circle What drives interlocks/multiple
directorships?

• Not collusion, power/influence, cooptation.
• Instead desire for business scan
• Scan Information on business environment
  (including politics)
• business scan gives directors/corporations
• class-wide perspective
• status within corporate community
• Inner Circle Directors with multiple
  directorships

6
The Inner Circle What drives interlocks/multiple
directorships?

• Directors with multiple positions (Inner Circle)
  overrepresented in
• political donations
• business associations
• media
• arts bodies
• Entry to Inner Circle increased by
• Exclusive schools upper class, family, networks
• Businessmen's Clubs upper class, networks
• Wealth/Family upper class, stockholding

7
Main Hypotheses

• H1 Politics Politically active firms have
  greater interlocking.
• Think tanks
• Business Association Executives
• Government Boards
• Political Donations
• H2 Status Firms with high status directors have
  greater interlocking.
• Directors in Whos Who

8
Main Hypotheses

• H3 School Firms with exclusive school graduates
  have greater interlocking.
• School attendance from Whos Who
• H4 Clubs Firms with members of top
  businessmens clubs have greater interlocking
• Club membership from Whos Who
• H5 Wealth/Family Firms with wealthy directors
  have greater interlocking
• Rich 200 List (publication)

9
Secondary Hypotheses

• H6 Geography
• H6a Firms in same region have greater ties to
  firms in the same region (homophily)
• State of firm headquarters
• H6b Firms in dominant economic region have
  greater interlocking (activity).
• H7 National/Foreign
• H7a Local firms have greater interlocking with
  other local firms (homophily).
• H7b Local firms have greater interlocking
  (activity).

10
Secondary Hypotheses

• H8 Corporate Liberalism Theory Politically
  moderate firms have greater interlocking.
• Donation to both major parties (bipartisan
  donors)
• H9 Political Factions Politically moderate
  (conservative) firms have greater interlocking
  with other politically moderate (conservative)
  firms (homophily)
• Donation to conservative parties only vs
  bipartisan donors

11
Secondary Hypotheses

• H10 Regulation Companies which are highly
  regulated have greater interlocking.
• H11 Publicly Listed Companies which are
  publicly listed have greater interlocking.
• H12 Revenue/Size Larger firms have greater
  interlocking.

12
Exponential Random Graph (ERG or p) models

• Also called p models or ERGMs
• For more
• Harrigan 2007. An introduction to ERG models for
  corporate networks. On Politics and Interlocking
  Directorates Research Community website
• http//sna.pl/politicsandinterlocking/

13
ERG (p) models

• Key ideas of ERG models
• We can model the fundamental building building
  blocks of graphs (generative).
• Building blocks are small sub-graphs. For
  example
• edge
• triangle
• edge between two attributes
• Sub-graph is called a configuration.
• Relationship between each sub-graph
  (configuration) and an observed graph (our
  network) can be expressed as a number (parameter)

14
Specifications for Experiment

• Software PNet (Robins, Pattison and Wang
  University of Melbourne). Both user friendly and
  efficient (handles large networks, lots of
  functionality).
• See also Harrigan 2007, PNet for Dummies.

15
Specifications for Experiment

• Non-directed network Ties formed by
  Non-Executive Directors serving on two or more
  board of the 250 Largest Corporations in
  Australia (2006)
• Structural Configurations
• star effects rich get richer
• triangle effects friend of a friend
• isolates tendency for isolated nodes
• 27 Attributes with two configurations for each
  activity and homophily
• Covariate network Executive Directors network

16
Interlocking Directorates
17
Exclusive School Attendance
18
Revenue
19
Whos Who Listing
20
Results
21
Results
22
Results

• Structural Parameters
• edges interlocks rare
• isolates large number of isolates
• k-star marginally significant and ve
• popularity is not an organising principle
• reverse unpopularity is an organising principle
• k-triangle significant and ve
• artefact of 2-mode gt 1-mode (control)
• core-periphery driven by transitivity
• small number of highly networked directors An
  Inner Circle

23
Results

• Activity Effects
• Revenue ve marginally significant
• information and prestige for directors
• large corporations have the means to recruit
  highly interlocked directors
• Business Association Execs ve marginally
  significant
• inner circle tied to the leadership of business

24
Results

• Activity Effects
• Whos Who ve significant
• inner circle tied to high status individuals
• School ve significant
• inner circle tied to upper class
• Club ve significant
• inner circle tied to other important social
  networks

25
Results

• Activity Effects
• No effect on activity of
• Australian vs foreign
• Regulated vs unregulated
• Listed vs private
• Geography
• Think Tanks
• Govt Boards
• Executive Directors Network

26
Results

• Activity Effects
• Political Donations -ve significant
• donations require autonomy
• donations crude form of influence
• donations liability as well as benefit
• highly networked firms more to loose
• less networked firms more to gain by donations
• Eg. National Australia Bank vs Ethanol Producer

27
Results

• Activity Effects
• Wealthy directors -ve significant
• Similar pattern in British data
• Why?
• wealth as independent source of power, interlocks
  not necessary to gain social capital or
  influence?
• interlocks interfere with power of wealthy?
  Scrutiny and unwanted influence over family
  business?
• rich lists sensitive to new wealth with fewer
  interlocks?

28
Results

• Homophily Effects
• Listed Firms Community? Expertise?
• No effect of
• Political donations/factions
• Australian vs Foreign
• Geography
• Regulation
• Wealth/Director capital
• Heterophily (opposites attract/hierarchy)
• School
• Government board
• Number of NEDs on board
• ( tendency for clubs, revenue, etc.)

29
Heterophily School Attendance
30
Results

• Why heterophily?
• Low status firms gain from upwards
  interlocking,
• But why do High status firms engage in these
  interlocks?
• avoiding conflicts of interest?
• limited number of positions and directors
• ambitious directors need low status boards
  because so few positions at the top
• top corporations need to recruit directors from
  lower status firms because so few top directors.

31
Conclusions

• Interlocking driven by
• Director characteristics (vs corporate)
• Small number of high status directors
• Upper class network effects are strong
• Centre of business politics and policy
• business associations.
• Heterophily High and Low status firms
• limited board size and directorships
• Political donations
• crude form of power/influence that is avoided by
  the most networked corporations
• Wealthy directors
• wealth opens its own doors?
• Interlocks constrain the power of wealthy?

32
Further Research

• Bipartite networks (director/corporation)
• Longitudinal (causality)
• Social Influence p models (attributes)
• Covariate networks
• stockholdings
• industry input/output monopoly/concentration
  (constraint)
• debt, banks
• Attributes
• profit
• lobbying
• Full network 2000 corporations, 10,000 directors
• Anglo-saxon and/or European comparisons