Javier García-Bernardo

University of Amsterdam

June 12th, 2019

computational social science and power

The CORPNET research group uncovers, investigates and aims to understand global networks of corporate power in contemporary global capitalism

What are we doing?

How?

  • Using "big data" and bringing new methods from complex systems
  • For the analysis of corporate ownership and corporate elites

 

PI: Eelke Heemskerk

starting point: relational understanding of power

Power: the capacity of actor A to make actor B do something that B would not otherwise do or as the exercise of such a capacity (Mark Haugaard).

  • Consciously (e.g. lobbying, agenda-setting) or unconsciously (e.g. ideology).

  • Direct (e.g. lobbying to the government), or indirect (e.g. lobbying to the unions, which then have power over the government).     

  • All power is relational.

As such, we model actors as part of networks, and use methods from network science to analyze who is important in the network, and why.

what type of networks? (1) Corporate ownership

Nodes: Companies

Edges: Ownership relations

Nodes: Countries

Edges: Ownership relations

Vitali, Glattfelder and Battitson, PlosOne (2011)

 

Babic, Garcia-Bernardo and Heemskerk (forthcoming)

Typically directed networks

what type of networks? (2) Corporate elites

Edges: Shared board membership

 

Edges: Shared director

Typically undirected networks

Nodes: Countries

Edges: Move to a new country

what type of networks? (2) Corporate elites

structure of the talk

  • PART 1: methods and applications

  • PART 2: Data

part 1: methods and applications

 

method 1: network centrality

method 1: network centrality

Marriage network of families in Florence.

 

method 1: network centrality

Find the most powerful nodes in a network.

Main question: How do you define powerful?

 

  1. The one with the most contacts [Degree centrality]
    • Connect to many people directly
    • Less dependent on other actors
  2. The one able to influence most people [Eigenvector centrality]
    • More visible: Acts as a `reference point'
  3. The one able to broker social contacts [Betweeness centrality]
    • Connect different actors, able to extract value from it

Endless definitions of centrality...

How to calculate centralities? General idea

https://www.r-bloggers.com/from-random-walks-to-personalized-pagerank/

methods from complex systems (1): NEtwork centrality

Example undirected networks

  • Degree
  • Betweenness ~ How many times a node is in the middle of a random walk
  • Eigenvector ~ How many random walkers are in a node

project 1: Big Three Passive Asset Managers

 Jan fichtner & EELKE HEEMSKERK

@fichtner_jan

what's the network?

Nodes: Companies

Edges: Ownership relations

Since the financial crisis --> Three trillions transferred from active to passive funds (cheaper)

project 1: Big Three Passive Asset Managers

Passive managers buys all the stocks in an index. Do not try to beat the market.

  • Do they have substantial voting power?
    • Condition 1: One or few companies dominating the market
    • Condition 2: Enough shares to make a difference
    • Condition 3: Coordinate the voting of all funds
  • Do they use it? --> Not yet: Fichtner,  Heemskerk. SSRN  (2018)

Usually, shareholders have three ways for exerting power/influence over companies: (1) Private engagements,  (2) Voting,  (3) Exit

 

  • Large active asset managers in the early twenty-first century, such as Fidelity, preferred to sell their shares (easier).  (Davis 2008)
  • Passive investors have no option for (3) Exit, only options (1) and (2) are available. Regarding (2):

project 1: Big Three Passive Asset Managers

Do they have power? Condition 1: concentration

Three firms combined have >90% of the market

Node size proportional to eigenvector centrality.

Fichtner, J; Heemskerk, E M; Garcia-Bernardo, Business and Politics (2017)

Do they have power? Condition 2: voting rights

Do they have power? Condition 2: voting rights

A centralized voting strategy is a fundamental prerequisite to using their shareholder power effective

 

They coordinate!

Do they have power? Condition 3: coordination

Fichtner, J; Heemskerk, E M; Garcia-Bernardo, Business and Politics (2017)

 

project 2:  the rise of transational state capital

milan babic

@mbabic_1

what's the network?

Nodes: Countries

Edges: Ownership relations

Babic,  Fichtner, Heemskerk. The International Spectator (2017)

States not only regulate, enable and constrain corporate power; they are also actors in the global
economy as shareholders of corporations.

As owners of and investors in corporations, states use the transnational agency space to compete for returns on investments, thereby possibly creating (geo-)political ties.

 

This is not to say that state power is identical with corporate and class power; but states are 'juxtaposed' to those actors.

project 2:  the rise of transational state capital

They have become important economic actors, playing juxtaposed to corporations

Babic, Heemskerk and Garcia-Bernardo (2019)

>100,000 investments

Node size:

Degree centrality

Babic, Heemskerk and Garcia-Bernardo (2019)

The transformation of state power (from regulators to actors) is not a one-way route

Clauset, Arbesman and Larremore. Science Advances (2015)

Morgan, Economou, Way and Clauset. EPJ Data Science (2018)

Ranking institutions

Transmission of information

Possible application: Understand the power of institutions,  scientific journals, corporations

method 2: ranking

Community detection: which nodes form a cohesive group?

https://www.slideshare.net/ErikaFilleLegara/community-detection-with-networkx-59540229

method 3: community detection

project 3: structure of the interlock network

 frank takes & EELKE HEEMSKERK

(Javier Garcia-Bernardo, Jouke Huijzer)

@franktakes

@JoukeHuijzer

what's the network?

Nodes: Cities

Edges: Board interlocks between companies in the two cities

Heemskerk, Takes, Garcia-Bernardo and Huijzer ‘ Sociologica (2016)

  • Dominant approach when looking at interlocking directorates: National = local, transnational = nonlocal.

 

  • In some regions the business communities are organized along national borders, in others at the city level

 

  • Elites from Asia, Latin-America are not yet fully integrated in the corporate elite network that transcend local communities.

5 million firms

Clustering: what are similar sequences?

Power: How does the global elite comes to be? What are important institutions and pathways?

method 4: sequence analysis + clustering

project 4: emergence of the GLOBAL ELITE

diliara valeeva

@diliara_valeeva

what's the network?

Nodes: Countries

Edges: Directors' careers

2017

Telstra Corporation Limited (AU)

 

Career sequence:

NZ

|

IT

|

US

|

CH

|

AU

|

US

1996-2007

Sun Microsystems (NZ)

2007-2016

Juniper Networks Inc (IT)

2014

Tesla (US)

2016-2017

ABB Ltd (CH)

Robyn Denholm, chair of the Tesla board since Nov 2018

2018

Tesla (US)

~10k largest companies and 10k executive

New insights:

  • Understand flows of people, ideas, practices:                                                                              
    • There is a global labor market for executives.
    • But it's highly hierarchical, directed, complex
    • There is a core and periphery

 

  • Competition between elites: There are different ways of integrating into the global power elites and each 'national elite' has their own way of doing so

 

  • Emergence of new global centers: Asian elites interact with the Anglo-American world while trying to establish their own power elite, maybe alternative to the traditional and established ones.

project 5: tax avoidance and offshore finance

javier garcia-bernardo

@javiergb_com

who is the largest investor in germany?

who is the largest investor in brazil?

who is the largest investor in south africa?

why?

Tax avoidance

crash course in tax avoidance

Crash course on "tax avoidance" (base erosion / profit shifting)

Caveat:  Arm-length principle: Intra-group payments need to be priced at "market prices"

However: What's the market price of a mermaid on a coffee cup?

  • Step 1: Sell things in a country
  • Step 2: Make intra-group payments to holdings in low-tax jurisdictions
    • Intellectual property / brand / know-how
    • Interest on loans
    • Services/goods
  • Bonus points if:
    • Sign a secret APA with a government
    • Take advantage of mismatches in tax treaties

drivers of tax avoidance

A: rising power of corporations

the great fragmentation of the firm

Reurink and Garcia-Bernardo (2018)

the great fragmentation of the firm

  • 1980s:
    • Switch to the shareholder value paradigm                                                                                  
    • Reductions in trade and invest barriers
    • Advances in Information and Communication technologies

 

  • Firms began to unbundle:
    • Operationally: Increase efficiency                                                                                                  
    • Legal-financial: Capture the value
      • Intermediate holdings
      • Rearrangement of intangibles
      • Intra-group financing
  • However:
    • High withholding tax rates
    • Multinationals were still mainly domestic                                                                                        

the great fragmentation of the firm

1990s:

- FDI started rising

- Double tax treaties and Investment treaties reduced withholding taxes

- Financialization of the firm: Financial profits started to have a higher weight

 

1) Increase the power of corporations over States: Ability to exit, and large size

 

Example: Shell

- British-Dutch oil and gas company headquartered in the Netherlands and incorporated in the United Kingdom.

 

B: Rising power of ofcs

the great fragmentation of the firm

1990s:

- FDI started rising

- Double tax treaties and Investment treaties reduced withholding taxes

- Intangibles started to have a higher weight

 

1) Increase the power of corporations over States: Ability to exit

 

2) Increased the power of some States:

- Attracting highly mobile parts of the firm

- Tax sovereignty

- BEPS

 

But which States?

Data provider: Orbis
 ~300 million companies
 ~100 million ownership links

sink-OFFshore financial centers

15 companies per capita

Garcia-Bernardo, Fichtner, Takes, Heemskerk (2017)

conduit-OFFshore financial centers

Garcia-Bernardo, Fichtner, Takes, Heemskerk (2017)

Garcia-Bernardo, Janský and Tørsløv  (forthcoming)

main Reason

Small countries (OFCs) have large incentives and little restrictions to give de facto advantages to multinational corporations

That reduces the autonomy of other countries (especially in the EU) on taxation, and force them to decrease corporate tax rates.

Blocking EU legislation

rising power of offshore financial centers

Low tax rates for interests and intellectual property

c: rising power of the professionals

@lajdacic

Lena Ajdacic

Talk at 12:30!

D: rising power of wealth

Source: World Inequality Database

Wealthy people have access to better investments and lower taxation

part 2: data

 

Computational methods allow you to find new approaches to answer questions of power

classification

Supervised machine learning

- Training data: Give titles with labels (important position / not important)

- Prediction: Give unlabeled titles

 

Mr Charles Ko - Quantitative Research Analyst, Global Quantitative Equity

Mr Colin D. Meadows - Senior Managing Director and Chief Administrative Officer

name matching

 Database 1 

  • Mr. John Smith
  • John Smith
  • John Smith
  • John Smith

Database 2

  • John Smith
  • Jonh Smith
  • John Smit
  • Jon Smit

Now do it with 100 million names, you have

10,000,000,000,000,000 possible pairs

data scraping

We collected a set of 8.6 million votes on 2.7 million unique proposals at AGMs worldwide. These votes are cast through 3,545 funds that are part of a set of 131 Asset Managers.

project 6: mapping tax professionals

SAILA Stausholm & javier garcia-bernardo

@thesailaway_CPH

Data: LinkedIn ads

Computational social science gives new methods useful for the understanding of power

  • Network centrality: Importance of actors in networks
  • Community detection: Cohesion, belonging
  • Sequence analysis/Clustering: Similar individuals, find common trajectories
  • Sequence analysis/Markov chains: Find brokers

 

Also new opportunities to automatize work:

  • Classification
  • Name matching / Combine data
  • Data scraping: https://github.com/jgarciab/web_scraping

 

corpnet.uva.nl

@javiergb_com

@uvaCORPNET

javiergb.com

corpnet@uva.nl

garcia@uva.nl

This presentation: slides.com/jgarciab/murten19

FINAL THOUGHTS

Computational Social Science and Power

By Javier GB

Computational Social Science and Power

  • 951