Professor Marco Barassi at the Centre for Big Data in Economics and Finance at HSE University

The Centre for Big Data in Economics and Finance at HSE University hosted Professor Marco Barassi of the University of Birmingham. During his visit he delivered a plenary talk at a special session of the 8th Applied Econometrics conference and also presented a paper at the International Research Seminar of the Faculty of Economic Sciences.

The Centre for Big Data in Economics and Finance at HSE University hosted Professor Marco Barassi of the University of Birmingham.

Marco Barassi works in the field of applied time series econometrics. His research interests span non-stationary time series models, structural change detection, time-varying parameter techniques, and non-linear time series models. A substantial part of his research focuses on the estimation of long memory models and related non-linear specifications, with applications in environmental economics and financial economics.

During his visit, Professor Barassi delivered a plenary talk at a special session of the 8th Applied Econometrics conference held on April 24–25, 2026. His presentation, titled «Single-index continuous threshold regression in heterogeneous panel data with interactive fixed effects», introduced a novel methodology for continuous-index threshold regression in heterogeneous panels with interactive fixed effects.

On April 28, 2026, Professor Barassi also presented a paper at the International Research Seminar of the Faculty of Economic Sciences, titled «Panel Vector Autoregression with Latent Group Structures». This study develops a panel VAR model with heterogeneous coefficients that allows for an unknown number of latent groups and unknown group membership. The proposed specification remains parsimonious and computationally tractable. The authors derive the asymptotic distribution of the estimator and establish its consistency as both the cross-sectional (N) and time (T) dimensions grow large.

The empirical component of the paper draws on Chinese data and shows that accounting for latent group structures yields conclusions that differ substantively from those produced by a standard panel VAR. Notably, the latent-group approach delivers a superior regional classification compared to conventional geographic partitioning. The discussions that took place during the conference and following the seminar laid the foundation for further joint research and academic collaboration.