Dieter Wang

PhD Candidate in Financial Econometrics
Tinbergen Institute and Vrije Universiteit Amsterdam

NEWS Currently working as Quantitative Analyst in Green Finance at the World Bank, Washington DC.

This, in a nutshell, is the guiding question for my research. As a visiting researcher at De Nederlandsche Bank, I'm able to find answers to this crucial question with the help of detailed supervisory datasets. Since contagion is unobserved and inherently dynamic, I develop a new class of spatial econometric state-space models to extract policy-relevant signals of the network. However, because the estimation of such highly nonlinear time series models with stochastic volatilty is challenging, I use simulation-based methods such as importance sampling or particle filters. My research is programmed in Python and I use JavaScript for interactive apps.

Chain of contagion
http://dieter.wang/contagionchain

International banking statistics
http://dieter.wang/internationalbanking

Contagion through asymmetric portfolio overlap networks and the credit spread puzzle
This paper explores the credit spread puzzle in banks from a network perspective. The puzzle refers to the limited ability of structural regressions to explain credit risk, characterized by low R2 and uncaptured residual factors. These models, however, treat banks as independent entities, which stands in stark contrast to the contagious nature of credit risk. We therefore augment the structural approach with networks to allow for contagion. In our networks, banks are prone to contagion if they follow similar business models, leading to common asset exposures. We proxy business models similarities through portfolios overlaps, since banks with similar portfolios likely follow similar business models. We construct the overlap network of the 24 largest banks in the Eurozone, using their complete asset holdings from a confidential supervisory dataset. We then embed the portfolio overlap network into the structural regression to estimate the network effects over time. The result is the Dynamic Network Effects (DNE) model. Our key finding is that the time-varying network effects explain the puzzle by capturing the residual factors and increasing the R2. This is because structural regressions without the overlap networks lack the means to capture common dynamics of credit risk, leaving them for the residuals. At the same time, when network effects are absent we likely overestimate the importance of the structural regressors.

Graduate level

Lecturer, Financial Econometrics in Python, Northeast Normal University
TA/Lecturer, Financial Markets and Institutions, VU Amsterdam
TA, Empirical Finance and Accounting (Stata), VU Amsterdam
TA, Advanced Asset Management, Amsterdam Business School
TA, Mathematics I, Tinbergen Institute
TA, Introduction to Programming + LaTeX course (Matlab, Ox), Tinbergen Institute

Undergraduate level

Supervisor, Bachelor Thesis, VU Amsterdam
TA, Finance I, VU Amsterdam
TA, Quantitative Methods of Economics, Eberhard Karls Universität Tübingen
TA, Intermediate Microeconomics, Eberhard Karls Universität Tübingen
TA, Risk and Probability, Eberhard Karls Universität Tübingen
TA, Mathematical Methods of Economics, Eberhard Karls Universität Tübingen

2017, VU Innovation Prize (€50,000) with Ines Lindner and Bernd Heidergott
2015, DAAD Graduate Scholarship
2014, Tinbergen Institute Scholarship
2012, DAAD Full-Year Exchange Scholarship

Regional Science and Urban Economics

Vrije Universiteit Amsterdam
Department of Finance
7A-47 De Boelelaan 1105
1081HV Amsterdam.

d.wangvu.nl