Vasilis Sarafidis

What's new?

November 2026: New study on "Leverage Dynamics in Product-Market Networks" (with S. Asimakopoulos). We examine whether corporate leverage persistence reflects firms’ own adjustment frictions or also the past financing decisions of economically connected firms. Using time-varying product-market networks, we estimate a heterogeneous dynamic panel model that accounts for latent common factors. We find significant network dependence, with lagged network leverage accounting for approximately 45% of total estimated leverage persistence. Accounting for these effects reduces the persistence attributed to firms’ own financing histories and increases their implied speed of adjustment. We also uncover systematic product-market clustering in firms’ financing behaviour and show that allowing network connections to evolve over time substantially affects estimated spillovers.

October 2026: New work on "The Anatomy of Commodity Risk: Micro, Market, and Economy-Wide Resources" (with N. Aslanidis, A. Bariviera, G. Kapetanios, and A. Ventouri). We develop a two-stage “divide-and-conquer” framework for decomposing commodity risk into micro, market-level, and economy-wide sources, allowing sensitivities to vary across commodities while treating economy-wide risk as latent. We introduce Risk Intensity Indices (RIIs), to quantify the relative importance of each risk source on a common scale. We find that market risk is the largest component on average and particularly important for energy commodities. Micro and market risk are highly concentrated across commodities, whereas economy-wide risk is more broadly dispersed. Risk composition also varies substantially across sectors and over time, while micro and market RIIs contain information about future volatility and absolute returns. 

September 2026: A substantially revised version of our paper "IV Estimation of Heterogeneous Spatial Dynamic Panel Models with Interactive Effects" (with Jia Chen, Guowei Cui and Takashi Yamagata) is now available. The paper develops a Mean Group Instrumental Variables (MGIV) approach for spatial dynamic panel data models with interactive effects, allowing slope coefficients to vary across cross-sectional units. The revised theoretical framework accommodates serial dependence and heteroskedasticity in the idiosyncratic errors, establishes feasible robust inference for the individual-specific IV estimators, and strengthens the asymptotic theory for MGIV under weaker restrictions on the relative growth rates of (N) and (T). The estimator remains linear and computationally attractive, while avoiding the need for asymptotic bias correction.

August 2026: Delighted to share our new work on "Nonlinear Boosting with Multiple Testing in High-Dimensional Generalised Linear Models with Binary Responses" (with C. Grivas, G. Kapetanios, Z. Psaradakis, M. Vavra, and A. Ventouri). We develop a variable selection procedure for high-dimensional generalised linear models that consistently identifies the true sparse model and enjoys the oracle property, yielding post-selection maximum likelihood estimates that are asymptotically equivalent to those obtained if the correct model were known in advance. Monte Carlo experiments show that BMT outperforms competing methods in both variable selection and parameter estimation. An empirical application demonstrates excellent out-of-sample performance in predicting whether U.S. inflation exceeds a given threshold over a 12-month horizon. 

July 2026: I am pleased to share a new working paper “Estimation and Inference for Latent Dual Networks Using High-Dimensional IV Screening” (with Arturas Juodis and George Kapetanios). We develop a novel methodology for estimation and inference in high-dimensional panel network models with latent dual structures. The framework allows outcomes to be affected simultaneously by positive and negative interaction channels, accommodating settings in which some interactions reinforce outcomes while others generate competition and displacement effects. The proposed method identifies and estimates the network directly from the structural model using observed data without the need to pre-specify the network. Network recovery is achieved through a sequential instrumental-variable screening procedure. We establish exact support recovery and oracle-equivalent post-selection inference.

Previous news items..