Title: Financial network discovery from restricted vector autoregression: New model and application
Authors: Massimiliano Caporin - University of Padova (Italy)
Deniz Erdemlioglu - IESEG School of Management (France)
Stefano Nasini - IESEG School of Management (France) [presenting]
Abstract: A new VAR model is developed that uncovers the interconnectedness among financial assets by aggregating realized trading information from intraday data. With a limited number of parameters, the presented model can accommodate the dynamic interactions in large panels of financial assets and realised measures (such as returns, volatilities, etc.). We propose an alternating direction method for maximum likelihood estimation of this type of VAR family. Using daily cross-sectional data on 1095 individual firms over fifteen years, we show that the full model estimation is reliable and computationally feasible. Our empirical results support the practical effectiveness of our model for ranking systemically important financial institutions.