Offshore Financial Centers and The Five Largest Value Conduits in the World

Public outcry over tax havens has increased in recent years. Journalists have shed light on the users of these offshore financial centres (OFCs), as well as the jurisdictions, banks, accountancy and law firms involved: OffshoreLeaks (2013), LuxLeaks (2014), SwissLeaks (2015), the Panama Papers (2015) and BahamasLeaks (2016).

OFCs are popular instruments for multinational corporations to (legally) reduce their tax bill by moving capital across borders in the form of dividends, royalties and interest by taking advantage of loopholes in the legislation. By playing out one state against another, corporations may reduce their tax rate from around 35% to 15-25% (and some much lower). For instance, Apple uses a combination of subsidiaries in Ireland, the Netherlands and Bermuda to strongly reduce its tax payments in Europe to a stunning 0.005% in 2014 according to the European Comission.

If profits would be accounted for where the economic activity takes place, multinationals would pay at least US$500-650 billion more on taxes, according to estimates by the Tax Justice Network and the International Monetary Fund. From this, around US$200 billion relates to developing countries, which means that developing countries lose more capital in tax avoidance than receive in development aid (US$142.6 billion).

What countries are Offshore Financial Centres?
Given this contested role of OFCs it is surprising that we still lack a broadly accepted definition of what makes a country an OFC. Instead, the identification of OFC jurisdictions has become a politicised and contested issue. International organisations such as the OECD or the IMF have published lists of alleged tax havens (OECD listIMF list), but the chosen criteria remained heavily influenced by politics.

To remedy this lack of transparency, we developed a novel, data-driven approach that identifies OFCs. We simply ask which countries or jurisdictions play a role in corporate ownership chains that is incommensurate with the size of their domestic economies (see Zoromé 2007). Our results show that offshore finance is not the exclusive business of exotic small islands far away. Countries such as the Netherlands and the United Kingdom play a crucial yet previously hidden role as conduits of offshore finance on its way to tax havens.

Using big data to find OFCs
Early attempts at OFC identification have resulted in for instance the Tax Justice Network’s “Financial Secrecy Index” and Oxfam’s list of the worst corporate tax havens. Jan Fichtner’s “Offshore-intensity Ratio” provides a helpful rough yardstick to judge which jurisdictions act as OFCs by describing the proportion between foreign capital (such as FDI) and the size of the domestic economy. However, these measures do not allow us to differentiate if foreign investment reported by Bermuda originates in the Netherlands, or if in contrast it originates in Germany and is routed through the Netherlands. We still don’t know how offshore finance flows across the globe.

To overcome these problems we move from country level statistics to large scale company data. The coming together of political economists and computer scientists in the CORPNET research group at the University of Amsterdam made it possible to study how corporations make use of particular countries and jurisdictions in their international ownership structures.

We analyzed the entire global network of ownership relations, with information of over 98 million firms and 71 million ownership relations. Note that here we are interested in how OFCs cater to the needs of multinational corporations, and not private individuals. Unlike previous attempts at identifying OFCs, this granular firm-level network data allowed us to identify and distinguish what we call “sink-OFCs” and “conduit-OFCs.” With some surprising results.

Introducing sinks and conduits
Sink-OFCs attract and retain foreign capital. Our approach identifies 24 sink-OFCs, including LuxembourgHong Kong, the
British Virgin IslandsBermudaJersey and the Cayman Islands. Indeed this replicates earlier lists of tax havens. But in addition our method confirms for example that Taiwan is indeed an “un-noticed tax haven“.

Figure: Sink Offshore Financial Centers (jurisdictions in blue have been under British sovereignty in the past or are still UK dependencies

Using our method we can now also investigate which jurisdictions are used by corporations en route to sinks. These conduit-OFCs are attractive intermediate destinations and enable the transfer of capital without taxation.

Surprisingly we found that only five big countries act as conduit-OFCs. Together these five conduits funnel 47% of corporate offshore investment from tax havens, according to the data we analysed. The two largest conduits by far are the Netherlands (23%) and the United Kingdom (14%). They are followed by Switzerland (6%), Singapore (2%) and Ireland (1%).

In our detailed research article, “Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership Network” we also show that each conduit jurisdiction is specialised geographically and in industrial sectors.

Figure: Network of relationships between countries. Conduits are marked in green, sinks are marked in red. The size of the country is proportional to the investment flows through the country and the colour to its position as a sink (blue = no sink, red = sink). The size of the arrows is proportional to the investment between two countries and the colour to its importance (blue = lower flow than expected, red = higher flow than expected).

Offshore Financial Centers are often portrayed as small, exotic, far away islands that are difficult if not impossible to regulate. We show that many OFCs are in fact highly developed countries.

18 out of 24 sink-OFCs have a current or past dependence to the United Kingdom, which highlights the central role of London in offshore finance. The finance minister of the United Kingdom has speculated that the UK may become a tax haven of Europe after Brexit if not offered a good deal by the EU. Yet today, many of the UK overseas dependencies (such as Cayman Islands, Bermuda, British Virgin Islands or Jersey) already act as large tax havens.

Since the financial crisis, the EU and the OECD have increased pressure on tax avoidance, with modest effects. We hope that our approach can help regulators target the policy to the sectors and territories where the offshore activity concentrates. In particular, targeting conduit-OFCs could prove more effective than targeting sink-OFCs, since – while new territories with low or no corporate taxes are continuously emerging – the conditions for conduit-OFCs (numerous tax treaties, strong legal systems, good reputation) can only be found in a few countries.

Results and details are available on the dedicated website


J. Garcia-Bernardo, J. Fichtner, F.W. Takes and E.M. Heemskerk, Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership NetworkScientific Reports 7, article 6246, 2017. doi: 10.1038/s41598-017-06322-9

Brexit through the lens of Network Science: The Position of the UK and London in the Global Corporate Network

In this blog post we look at the United Kingdom and London from a network perspective, investigating the connectedness of British corporations in the global network of corporate control. The aim is to give an alternative and/or complementary “social network analysis”-perspective for the lively discussions on a possible Brexit, here focusing on data related to corporate power and control.


Brexit, the possible withdrawal of the United Kingdom from the European Union, is a frequently debated topic. A “yes” as outcome of the June 23 referendum is likely to be of substantial influence on the global economic system. A lot of the debates on Brexit are centered around arguments based on economic impact in terms of trade, manufacturing, import/export, foreign investment, etc. Throughout this blog article, we look at the UK from the perspective of Britain’s corporate elite network, investigating the position of the UK within the global corporate elite network. Firms are not individual market actors, but are typically embedded in dense networks of power and control, for example based on ownership or interlocking directorates. The CORPNET research group at the University of Amsterdam studies these networks as part of a five-year research programme funded by the European Research Council (ERC). The general idea behind the group’s so-called “network science“-approach is that by studying a system of interaction (the global economy) rather than mere sums and averages of the systems’s individuals (economic activity and behavior of corporations/countries), we obtain new insights in the considered system.

Position of the UK and London

Early 2016, our research group studied the organization of the global corporate elite from a network perspective. We visualized the relationships between cities across the globe based on whether the firms in these cities were involved in board interlocks: shared senior level directors between firms. To learn more about the substantial body of literature on the causes and consequences of interlocks, see the excellent survey article by Mark Mizruchi.

We obtain a tightly connected so-called city network consisting of 24,747 cities (the nodes), connected through 874,810 distinct board interlock ties (the ties). The strict center of this network, consisting of the 409 (1.65%) most well-connected cities across the globe, is shown in Figure 1. In this figure, a node (object) represents a city, the color of an object corresponds to the country, and the size of a node is based on how central this node is with respect to the other nodes, based on the structure of the network. Betweenness centrality is a network metric that indicates how often a node is on a shortest path between other nodes. It measures the brokerage position of that node, in our city network corresponding to a city’s contribution to connecting other cities across the globe.


Figure 1: The center of the city network based on board interlocks. Node color corresponds to the city’s country, node size is proportional to the city’s betweenness centrality value.

It is evident that London has by far the most dominant position in this network. Globally, London plays a crucial role in connecting the global corporate elite through short path lengths. Furthermore, we note that in general there is a large number of British cities (albeit as a result of London’s central position) present in the center of this network. More specifically, even though only 723 (2.9%) of the cities in the global network of 24,747 cities are in the United Kingdom, 69 (16.9%) of the cities in the small center of 409 cities is British. This demonstrates how on average the UK cities are nearly 6 times more dominant in the center of the network than an average country’s cities.

The United Kingdom, and in particular London, is indeed at the heart of the world’s corporate elite. London is not only a gatekeeper through which the United Kingdom is connected with the rest of the world, but also connects the European mainland’s corporate elite to for example the United States, the East and the former British Commonwealth countries. It is often argued that in general, The City is in largely favor of staying in the EU (see for example this survey by CSFI). Contrary, Brexit advocates claim that Britain is a well-developed and globally integrated country that can stand on its own within the global economy, without the need of the European Union. But compared to Europe, how divided is the UK itself? Does it indeed have one well-organized and connected corporate structure?

Community detection

To answer the above mentioned question, we will again “consult” the network, and attempt to derive communities from the city network based on board interlocks. We identify groups of nodes (cities) that are more tightly connected with each other, than with the rest of the network. Community detection is a well-known method from the field of network science that allows one to find these densely connected groups of nodes in the network, based only on the structure of the network, so without any prior knowledge rather than the network’s connections. Here we use modularity maximization using the Louvain method, a well-known computationally efficient technique for finding communities that by swapping nodes between communities tries to optimize the modularity value indicating the quality of the division of the network into communities. If we use this algorithm on the aforementioned global city network and focus on Europe, we get the division of cities into communities as indicated by the colors in Figure 2.

citynetwork-europeFigure 2: The European part of the global city network. Colors correspond to communities. Ties are omitted for readability of the figure.

From Figure 2, we can observe that a number of communities spanning multiple countries are found by the algorithm: there is a community of Germany, Switzerland, Austria, Hungary and former Yugoslavian countries, a community of cities in the “Benelux” countries (Belgium, Netherlands, Luxembourg), a community of Southern-European countries including France, Spain and Italy (but not Portugal, which is more strongly connected to Brazil and Latin America, not visualized here). The Czech Republic and Slovakia form a tight community, and so do the Scandinavian countries, except for Finland which is apparently more tightly connected in a community with Estonia. Poland’s cities do not clearly belong to any particular community, and are mixed between the communities around Russia, Germany and Scandinavia. Interesting and worthy to note here is that no information on the geography of Europe was put into the community detection algorithm. The results are purely derived from the structure of the corporate elite through board interlocks.

Turning to the UK; in Figure 2 we see how British cities appear to be in a community together with Ireland, seemingly posing as a whole. It should be noted that a fair share of the board interlocks also occurs between firms within cities, resulting in so-called self-loops in the networks. In particular, London has an extraordinary strong self-loop, as the City houses a large number of firms that interlock with each other, creating a tie that is stronger than any other tie. One way of looking at these self-loops is to say that they somehow indicate the internal power structure of a city. Interestingly, if in the global city network we search for communities in the network with these self-loops, the British business community falls apart into a a total of 19 separate communities, as shown in Figure 3, separating London from the rest of the UK. The resulting communities have a clear regional character, and are no longer as well-connected as they were with London’s integrative power. This suggests that there may be substantial consequences for the international connectedness of certain regions of the UK, should the Brexit somehow lead to a particular separation of these regions from The City.

ukFigure 3: The British part of the global city network. Colors correspond to communities found without the integrative power of London.

Final remarks

Based on these results, one could say that internally, the UK is perhaps as divided as the European Union itself, i.e., it may not want to overestimate its unity. At least the corporate elite in the UK itself is strongly regionally oriented, albeit overshadowed by the integrative power of The City. Of course it remains to be seen what the precise consequences are of a Brexit on a corporate system such as the one considered in this blog post. Nevertheless, given the key position of London and the UK in the global network, it is likely that local changes in the corporate structure of firms in the UK can significantly influence both the UK as well as the global system as a whole, given the dominant position of, in particular, the city of London.

Apart from the results obtained above for the United Kingdom, we hope that this short blog post provides some insight in how network science is able to reveal patterns that are not directly visible from studying the data objects themselves, but are evident in the network perspective. Within the CORPNET group of the University of Amsterdam, we are constantly trying to find new patterns, phenomena, methods and ideas to specifically understand corporate network structures. Want to know more about this research, the data and specifics about the city network construction and analysis? See our latest paper (from which the text above was largely derived):

  • E.M. Heemskerk, F.W. Takes, J. Garcia-Bernardo and M.J. Huijzer. Where is the global corporate elite? A large-scale network study of local and nonlocal interlocking directorates. Forthcoming in Sociologica, 2016. arXiv:1604.04722

Interested in learning more about the CORPNET group? Do feel free to contact us!

franktakes_largeThis blog post was written by Frank Takes, postdoctoral researcher in the CORPNET group at the University of Amsterdam.

Sunbelt 2016 presentations available online

The CORPNET team enjoyed the Sunbelt 2016 conference, attending many great presentations and meeting a number of interesting new and old colleagues and friends. Slides of the presentations given by our group members can be found below.

See you at Sunbelt 2017!

High Performance Computing grant received

The CORPNET project has received a grant of EUR 41,500 for high performance computing facilities from the university’s High Performance Computing and Networking (HPCN) fund. The HPCN fund is available for non-standard eScience computing facilities and ICT infrastructure contributing to research and education.

The grant will be spent on a server architecture for the CORPNET project, including an (online) data sharing platform and an architecture with over 768GB of RAM for real-time in-memory network processing and modelling. In the coming weeks, more information will be announced on this website.