Risk management via contemporaneous and temporal dependence structures with applications

Area of Science:

• Econometrics
• Network Analysis
• Time Series Analysis

Background:

• Multivariate time series data often contain complex interdependencies.
• Understanding risk propagation and contagion is crucial in financial and economic systems.
• Existing models may not fully capture the dynamic network structures within time series.

Purpose of the Study:

• To develop and present estimation methods for Bayesian Graphical Vector Auto-regression (BG-VAR(X)) and Bayesian Graphical Systems Equation Modelling (BG-SEM(X)).
• To examine risk network structures embedded in multivariate time series data.
• To analyze risk propagation dynamics and persistence using complex network models.

Main Methods:

• Estimation of BG-VAR(X) and BG-SEM(X) models, with and without exogenous variables.
• Application of network models to identify within-day and across-day risk transmitters and recipients.
• Comparative analysis of models with and without exogenous variables to assess their impact on network structure.

Main Results:

• Both BG-VARX and BG-SEM(X) effectively reveal major risk transmitters and recipients within multivariate time series.
• Models incorporating exogenous variables (BG-VARX and BG-SEM(X)) generate richer network structures compared to those without.
• The methods allow for the estimation of intra-day and inter-day interconnections and their dynamics.

Conclusions:

• The developed Bayesian graphical models provide a robust framework for analyzing risk networks in multivariate time series.
• Exogenous variables play a significant role in depicting influential network structures and risk propagation.
• This approach offers a platform for future research, including extensions to diverse data types and policy implications.