Research Philosophy

My research is driven by an iterative dialogue between theory, methodology, and real-world applications. I often begin with a theoretical question—sometimes motivated by challenges encountered in the analysis of real data, and other times deeper curiosity about underlying probabilistic structures. Theoretical insights arising from this process naturally inform the development of statistically principled methods, which I then bring back to applied problems.

With a background in quantitative finance and economics, the practical questions I’m drawn to often arise in finance and insurance, especially those involving systemic risk driven by the tail behavior of risk variables. I’m also interested in climate and environmental issues, particularly the risks they pose to insurance and financial systems.

Research Interests

• Dependence Modeling
• Extreme Value Theory
• Multivariate Time Series Analysis
• Quantitative Risk Management

Recent Publications

• A factor-copula latent-vine time series model for extreme flood insurance losses [DOI] [Github]
  Journal of the American Statistical Association, 2025.
  This paper introduces a novel factor-copula latent-vine model for capturing complex spatiotemporal dependence in extreme flood insurance losses, with applications to the U.S. National Flood Insurance Program (NFIP) data.
• Properties of CoVaR based on tail expansions of copulas [DOI]
  Journal of Multivariate Analysis, 2025.
  This work proposes a new thoeretical framework for analyzing the conditional Value-at-Risk (CoVaR) using tail expansions of copulas.It studies the asymptotic behavior of the CoVaR under various tail dependence structures, providing insights into systemic risk measurement.
• Multivariate directional tail-weighted dependence measures [DOI]
  Journal of Multivariate Analysis, 2024.
  This paper introduces a new class of multivariate directional tail-weighted dependence measures. We study their theoretical properties, derive asymptotic expansions, and develop sample estimators with established asymptotic guarantees. These measures fill an important gap in assessing the goodness-of-fit of multivariate extreme value models.