The Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) and the rapid spread of the COVID-19 disease that began at the dawn of 2020, have had a dramatic impact on financial markets around the globe. The COVID-19 crisis has created an unprecedented level of risk, causing investors to suffer significant losses in a very short period of time. The outbreak of the pandemic, including the associated public health measures for containment and suppression, have led to an economic and financial downturn. The goal of the proposed research is to study the multifaceted economic and social consequences of the COVID-19 on financial markets interdependence and on the most important economic indicators. To achieve an in-depth understanding of the market reactions to this shock, we develop a graphical modeling framework that allows us to investigate how and to what extent the interconnections among several stock markets, have been impacted, modified and reshaped, because of the COVID-19 pandemic outbreak. In particular, to handle both continuous and discrete variables we will use Mixed Graphical Models (MGMs) which are able to learn the graph structure between heterogeneous variables in the network. To capture how interactions between the variables in the network change over time during the pandemic period, we introduce a dynamic MGM by considering state-dependent parameters evolving over time according to a latent semi-Markov process. Moreover, since the high level of integration of international financial markets highlights the need to accurately assess contagion and systemic risk under different market conditions, we also propose a dynamic quantile graphical model to identify the tail conditional dependence structure in multivariate data across different quantiles of the variables in the network. We will apply the proposed models to a large financial and economics datasets from several countries and develop tools to be includeded in freely availabe R packages.
We extend and improve the existing literature in a number of ways. As opposed to GGMs, the proposed quantile graphical model has several advantages. Firstly, we can assess contagion and systemic risk during different market conditions focusing on specific parts of the distributions of variables, without relying on the restrictive assumption of normally distributed data. Secondly, we can construct a collection of graphs indexed by the quantile level, which allows us to evaluate whether the financial network is more interconnected and more vulnerable to contagion during periods of financial and economic crises. This in turn provides a flexible approach for the computation of several risk measures (i.e. Value at Risk and Expected Shortfall) in a risk management framework. Furthermore, the generalization to time-varying graphical models allows to: (i) better understand the relationships between the different entities and how these relationships evolve over time, i.e. analyze how the financial market connectedness evolves during the COVID-19 pandemic (ii) associate the states of the hidden semi-Markov process with specific levels of stress in the system (for example high helth restrictions, and low helt restrictions ).
The second part of this project constitutes a significant effort to construct a general class of mixed graphical models that permits each node to belong to a potentially different variable type, thus broadly generalizing the applicability of classical graphical models. The idea of considering a semi-Markov switching process allows us to account for the dynamic evolution of the whole system.
From an empirical standpoint, our newly developed methods will provide risk managers and policy-makers an in-depth understanding of the evolution of the COVID-19 crisis and of its economic consequences. The analytical tools and knowledge acquired by applying the proposed models will be helpful to curb the long-term effects of the pandemic and prevent the damages of future crises.