The second research lecture of the year will be given by Marie Ternes. She is a PhD student at the Department of Quantitative Economics, and she also fought through the bachelor and the master in Econometrics at UM. In fact, she was actively involved in Vectum and was the secretary during the academic year of 2016/2017! Marie will present her paper "Hierarchical Regularizers for Mixed-Frequency Vector Autoregressions“, which was co-written with Alain Hecq and Ines Wilms. Here's a sneak peek into the paper: 

Vector AutoRegressive (VAR) models are a cornerstone for modeling multivariate time series; for studying their dynamics and for forecasting. However, a serious complication of macroeconomic forecasting with VARs is that macro and financial time series are usually recorded at different frequencies. For instance, the gross domestic product (GDP) is collected at quarterly frequency. Other series such as unemployment rate or inflation are released at monthly frequency. The money supply is published weekly and financial data such as stock prices are even available at daily frequency. If the objective is to forecast e.g., quarterly GDP, economic and financial indicators which are published at a monthly, weekly or even daily frequency, may be useful to improve forecast accuracy. While the standard VAR can only be used for time series of the same frequency, mixed-frequency VARs (MF-VAR) form an appealing alternative since they avoid the need of data aggregation and are able to exploit the information available in series recorded at different frequencies. However, for MF-VARs the number of parameters to estimate grows quickly. For these so-called "high-dimensional" models, standard econometric methods fail. In the paper, we address the "curse of dimensionality" through the convex regularization framework. The basic idea is to impose a simplifying structure on the parameter estimates and encourage the estimation towards that structure. I will show how we specifically do this for MF-VARs and how the estimator can be applied to the U.S. economy.

To hear more on the topic, join us coming Tuesday, November 9, at 19:00 via the button below!

After the intellectual stimulation, we will head to the Preuverij for some drinks.