Research Interests:

• Decision analysis by stochastic analysis, statistics, and Bayesian machine learning in the framework of Expected Utility Theory and Cumulative Prospect Theory. 

  • Stochastic analysis and convex analysis for portfolio optimization

Expected utility theory is one underlying hypothesis of many models for decision analysis in finance and economics. Considering a non-concave utility function allows us to incorporate "irrational" behavioral observations beyond classical rational agent assumptions. We have written two papers investigating the impacts of risk measures, e.g., Value-at-Risk and Expected Shortfall, on surplus-driven investors, which technically corresponds to non-concave optimization under constraints.  Link: paper 1 and paper 2. 

• Machine learning for public finance

Public finance and social welfare models are essential to study new public policies, e.g., pension reform, climate mitigation policy, and tax policy. However, these models usually have a complex nature and long horizon, which brings challenges in obtaining closed-form and computational solutions. Machine learning can help speed up complicated models' computation and simulation procedures. We have written one paper that, for the first time, applies Bayesian optimization, a global blackbox optimization algorithm based on nonparametric Gaussian process regression, to study the benefit of intergenerational risk sharing in pension reform. Link.