Milan van den Heuvel

Data Science - Causality - Complexity - Socioeconomics


Post Doctoral researcher at Ghent University
Full CV available on request

Download Resume

About


Ph.D. Physics and Economics. Currently doing a PostDoc researching the synergies between Machine Learning and economics to identify and quantify causal links in data at scale to better inform (policy) decisions. I am part of a research group that conducts interdisciplinary research in the fields of physics, machine learning, and computational social sciences (economics, socioeconomics, complex networks, and political science) often combining methods from several of these.

I am specialized in processing and analyzing data from an econometrics as well as from a data science point-of-view. I like challenging myself to learn new skills that will help solve problems or better the workflow within the team. I take initiative to keep track of goals, and work to improve communication to provide everybody with the information needed to further these goals.

  Education

2019/now - Post-Doctoral researcher

  @ Department of Economics, Ghent University

2015/2019 - Predoctoral fellow Physics and Economics

  @ Department of Physics and Astronomy, Department of Economics, Ghent University

2013/2015 - M.Sc. Physics and Astronomy

  @ Faculty of Sciences, Ghent University; Thesis

2010/2013 - B.Sc. Physics and Astronomy

  @ Faculty of Sciences, Ghent University

Research

Published papers

Loan maturity aggregation in interbank lending networks obscures mesoscale structure and economic functions

with Marnix Van Soom (VUB), Koen Schoors (UGent, HSE), and Jan Ruckebusch (UGent).

Scientific Reports, 9, 12512

August 2019     Link    

Show Outline

Since the 2007-2009 financial crisis, substantial academic effort was dedicated to improving our understanding of interbank lending networks (ILNs). Because of data limitations, the literature largely lacks loan maturity information. We employ an interbank loan contract dataset to investigate whether maturity details are informative of the network structure. Applying the layered stochastic block model of Peixoto (2015)~\cite{peixoto2015} and other tools from network science on a time series of bilateral loans with multiple maturity layers in the Russian ILN, we find that collapsing all such layers consistently obscures mesoscale structure. The optimal maturity granularity lies between completely collapsing and completely separating the maturity layers and depends on the development phase of the interbank market, with a more developed market requiring more layers for optimal description. Closer inspection of the inferred maturity bins associated with the optimal maturity granularity reveals specific economic functions, from liquidity intermediation to financing. Collapsing a network with multiple underlying maturity layers, common in economic research, is therefore not only an incomplete representation of the ILN’s mesoscale structure, but also conceals existing economic functions. This holds important insights and opportunities for theoretical and empirical studies on interbank market contagion, stability, and on the desirable level of regulatory data disclosure.

Social Stability and Extended Social Balance-Quantifying the Role of Inactive Links in Social Networks

with Andres M. Belaza (UGent), Kevin Hoefman (UGent), Jan Ryckebusch (UGent), Aaron Bramson (RIKEN, UGent, UNC Charlotte), Koen Schoors (UGent, HSE), Corneel Casert (UGent), and Benjamin Vandermarliere (UGent).

Physica A: Statistical Mechanics and its Applications; 518, 270-284

March 2019     Link    

Show Abstract

Structural balance in social network theory starts from signed networks with active relationships (friendly or hostile) to establish a hierarchy between four different types of triadic relationships. The lack of an active link also provides information about the network. To exploit the information that remains uncovered by structural balance, we introduce the inactive relationship that accounts for both neutral and nonexistent ties between two agents. This addition results in ten types of triads, with the advantage that the network analysis can be done with complete networks. To each type of triadic relationship, we assign an energy that is a measure for its average occupation probability. Finite temperatures account for a persistent form of disorder in the formation of the triadic relationships. We propose a Hamiltonian with three interaction terms and a chemical potential (capturing the cost of edge activation) as an underlying model for the triadic energy levels. Our model is suitable for empirical analysis of political networks and allows to uncover generative mechanisms. It is tested on an extended data set for the standings between two classes of alliances in a massively multi-player on-line game (MMOG) and on real-world data for the relationships between countries during the Cold War era. We find emergent properties in the triadic relationships between the nodes in a political network. For example, we observe a persistent hierarchy between the ten triadic energy levels across time and networks. In addition, the analysis reveals consistency in the extracted model parameters and a universal data collapse of a derived combination of global properties of the networks. We illustrate that the model has predictive power for the transition probabilities between the different triadic states.

Statistical Physics of Balance Theory

with Andres M. Belaza (UGent), Kevin Hoefman (UGent), Jan Ryckebusch (UGent), Aaron Bramson (RIKEN, UGent, UNC Charlotte), and Koen Schoors (UGent, HSE).

PLoS one, 12(8): e0183696

Aug 2017     Link    

Show Abstract

Triadic relationships are accepted to play a key role in the dynamics of social and political networks. Building on insights gleaned from balance theory in social network studies and from Boltzmann-Gibbs statistical physics, we propose a model to quantitatively capture the dynamics of the four types of triadic relationships in a network. Central to our model are the triads’ incidence rates and the idea that those can be modeled by assigning a specific triadic energy to each type of triadic relation. We emphasize the role of the degeneracy of the different triads and how it impacts the degree of frustration in the political network. In order to account for a persistent form of disorder in the formation of the triadic relationships, we introduce the systemic variable temperature. In order to learn about the dynamics and motives, we propose a generic Hamiltonian with three terms to model the triadic energies. One term is connected with a three-body interaction that captures balance theory. The other terms take into account the impact of heterogeneity and of negative edges in the triads. The validity of our model is tested on four datasets including the time series of triadic relationships for the standings between two classes of alliances in a massively multiplayer online game (MMOG). We also analyze real-world data for the relationships between the “agents” involved in the Syrian civil war, and in the relations between countries during the Cold War era. We find emerging properties in the triadic relationships in a political network, for example reflecting itself in a persistent hierarchy between the four triadic energies, and in the consistency of the extracted parameters from comparing the model Hamiltonian to the data.

Submitted/Working papers

Consumption heterogeneity

with Benjamin Vandermarliere (UGent), Koen Schoors (UGent, HSE).

TBA

2019     Link    

Show Outline

We investigate the consumption response to positive and negative income changes conditional on the presence of liquid wealth. We home in on a population which holds minimal consumer and mortgage debt, and holds most of its wealth, if any, as liquid wealth. We find an asymmetric consumption response to income changes, with a higher response to income increases than to income decreases. We further find that this asymmetry can be explained by a higher consumption smoothing effect of liquid wealth for income decreases than for income increases.

Other Published Work

Measuring Propagation with Temporal Webs

with Aaron Bramson (RIKEN, UGent, UNC Charlotte), Koen Schoors (UGent, HSE), Kevin Hoefman (UGent), Benjamin Vandermarliere (UGent).

Temporal network epidemiology, 57-104 (Springer)

2017     Link    

Show Summary

This book covers recent developments in epidemic process models and related data on temporally varying networks. It is widely recognized that contact networks are indispensable for describing, understanding, and intervening to stop the spread of infectious diseases in human and animal populations; “network epidemiology” is an umbrella term to describe this research field. More recently, contact networks have been recognized as being highly dynamic. This observation, also supported by an increasing amount of new data, has led to research on temporal networks, a rapidly growing area. Changes in network structure are often informed by epidemic (or other) dynamics, in which case they are referred to as adaptive networks. This volume gathers contributions by prominent authors working in temporal and adaptive network epidemiology, a field essential to understanding infectious diseases in real society.

Work In Progress

Perpetuation of wealth inequality through income mobility: drivers of early career performances.

with Tarik Roukny (MIT, KUL, UGent), Benjamin Vandermarliere (UGent), Koen Schoors (UGent, HSE), and Jan Ruckebusch (UGent).

Show Outline

We propose and evaluate three possible mechanisms behind the observed positive relationship between wealth inequality and income immobility. To this end, we combine de-identified, client-level financial and demographic data on career starters from a large European bank in Belgium, one of the most equal countries in the world.

Who benefits from gentrification?

with Koen Schoors (UGent, HSE)

Show Outline

Does living in an economic diverse neighbourhood benefit the children of the rich, the poor, or both? Is this effect diminished by assortative mixing mechanisms by the parents?

Want to get in touch?


I will get back to you as soon as possible!

Campus Tweekerkenstraat
Gebouw Hoveniersberg
Tweekerkenstraat 2
BE-9000 Gent