The results are based on a exible Stochastic Block model for multilayer networks (DSBMM) proposed in Baltodano López and Casarin (2023) and applied to trade flows and public debt in Baltodano López et al. (2024). The topological properties of the debttoGDP and trade ow networks are captured by the groups of countries sharing similar connectivity properties in the system.
The composition of the clusters and their evolution are inferred using Bayesian methods and the forecast of the probability of belonging to each cluster is estimated as a tool for tracking composite risk. In other words, the DSBMM allows us to jointly consider risks across financial systems and/or between unbalances in the scale and external sectors due to a potential twin decit phenomenon.
In order to extract the debt network and to construct the multilayer structure with trade flows, we use quarterly government debt in percentage of the GDP series and import values of Goods (CIF) in euros aggregated by quarter for all pair of countries in the EU countries during the period 20032024 from Eurostat (see Table 1 for further details).
The results show negative and positive correlations between countries' public debts, with a coreperiphery network structure that emerges during crisis periods evidencing high systemic risks. Comparing 2024QI and 2024Q3 there has been an increase in the network density for some countries, departing from an assortative behavior and getting closer to a coreperiphery topology, which is more prone to contagion and spillover effects. In particular, after COVID19 Poland has persistently remained in a high systemic risk group, while countries such as Italy, Finland, and Greece had changed to more stable communities during the first part of 2024, but they changed back to more riskier clusters by the end of 2024.
Other countries belong to more stable groups, but a high membership uncertainty is observed in the postpandemic period. For instance, Italy during the two first quarters of 2023 was presenting lower correlations with the rest of the countries in terms of debttoGDP trends, but with a posterior probability just above 50%. Still, there are few negative correlations between the EU members, which start to resemble the preCOVID period.
Therefore, it can be concluded that most of the countries present a higher risk compared to the previous results for 2024Q1 and the system is still in transition toward a lower-risk structure after the COVID19 shock.
