Universiteit Leiden

nl en

Network analysis reveals unexpected societal patterns

Thanks to data science, we can chart and search enormous quantities of related information. This generates all kinds of new insights, for example in complex global financial structures or such societal problems as loneliness.

Assistant professor Frank Takes is a specialist in network data. He creates models of gigantic and complex datasets such as social relations between people, global corporate data or communication networks. He then develops algorithms to search these datasets. ‘How long that takes depends on the complexity of the datasets, and on the question you’re trying to answer. There are times when a set of 128 parallel processors will be running for up to four days.’ As Takes himself says, he works at the computational frontier of the field of network analysis.

He uses algorithms to try to unravel patterns that can be distilled from the data. These patterns generate some surprising insights that can be highly useful for politicians or policymakers, for example.

Tax havens

One of Takes’ projects, which he carried out together with the University of Amsterdam, was the study of tens of millions of companies worldwide. ‘The aim was to chart global flows of money and to identify tax havens, not on the basis of the subjective reputation of a country, but founded on detailed flows of finances among millions of companies.’ The study, published in 2017, threw up some remarkable findings. Yes, there are countries – take, for instance, the Cayman Islands or Bermuda – where companies pile up money and then channel it elsewhere to avoid paying tax on it. Takes and his colleagues refer to these countries as ‘sinks’, drain holes where money ‘disappears’. But there are also countries that have a completely different reputation, including the Netherlands, and act as conduits for money and they, too, can play an important role in tax evasion. The findings of the study had some significant effects. Leading newspapers such as The Guardian have published articles on the findings, and in the Netherlands questions have been asked in Parliament about our country’s role as a financial conduit. Related research also produced other interesting images, such as the existence of ‘business communities’ where directors hold positions in different companies. If you look at the resulting networks on a global scale, you see that there are still huge differences between business life in the West and the East. 

‘The method our study used to chart financial networks has now become well established,’ says Takes.  

Strength of social networks

Takes is now applying the kinds of network algorithms that were used to search financial data in a unique new network study. This study aims to unravel patterns in social relations and societal phenomena. Together with the Social Planning Bureau, he will carry out searches of the anonymised social relations of 17 million Dutch people. ‘We want to look at how people interact. Where they work, who they know, where they went to school. That way we can chart the quality of their social network. We are hoping to discover patterns that will help to identify particular – as yet to be decided – groups with different kinds of issues, such as social problems. Which groups of elderly people experience loneliness? Who would benefit from social safety nets? These are very concrete pointers that can be used by policymakers.’  

Predictions

The current project, which will run until 2025, is a snapshot of social networks in the Netherlands. But Takes is already writing a proposal for follow-up research where the networks can be tracked over a longer period. ‘You can probably use patterns from that timeframe to make predictions. For example, you can identify people with a particular social profile who could in future suffer from such problems as loneliness.’ It could generate enormously valuable information for researchers in the social sciences and humanities, fields with which Takes specifically wants to collaborate.

The intention is also to make the infrastructure (the hard- and software) built for this research available for use at national level at some point in the future. Different types of datasets can them be imported and other research questions can be asked about the Dutch population.

This website uses cookies.  More information.