Finance Research Seminar by Prof. Ruijun Bu

The distinguished speaker is Prof. Ruijun Bu who is a Professor of Econometrics at the University of Liverpool Management School (ULMS). He is also the current Director of Research of the Economics Group at ULMS, and from 2012 to 2021 he was Founder and Director of the Econometrics and Big Data research cluster at ULMS. Prof. Bu has been also a Visiting Research Fellow of the Bendheim Centre for Finance at Princeton University.

Prof. Bu’s research interests are in the areas of financial economics and econometrics, quantitative finance, and time series analysis. He is particularly interested in non- and semi-parametric methods for modelling nonlinear time series models and continuous-time stochastic processes with macroeconomic and financial applications.

His excellent publications have appeared in the Journal of Econometrics, Econometric Theory, the European Journal of Operational Research, the Journal of Financial Econometrics, the Econometrics Journal, Econometric Reviews, the Journal of Time Series Analysis, the International Journal of Forecasting, and the Journal of Empirical Finance.

For this seminar, Prof. Bu will present a paper titled "Nonparametric Estimation of Large Spot Volatility Matrices for High-Frequency Financial Data”.

Abstract

We develop a technique for estimating spot volatility matrices for large-dimensional high-frequency financial data. We first integrate kernel smoothing with a generalized shrinkage method to estimate the large spot volatility matrix from noise-free data and establish the uniform consistency of our estimator. For data contaminated by microstructure noise, we introduce a localized pre-averaging method to reduce the effective magnitude of the noise and consider the estimation of the time-varying volatility matrix of the high-dimensional noise vector. We then extend our framework to estimate large spot volatility matrices in a time-varying factor model with observable risk factors. Corresponding uniform convergence results are also derived. Numerical studies, including simulations and an empirical application, assess the finite-sample performance of the proposed estimators. Finally, we demonstrate how our technique can be adapted for estimating large spot precision matrices, integrated volatility matrices and spot volatility matrices of stochastic factor loadings in our framework.