Globalization, through international trade and production, has significantly increased economic growth and reduced global income inequality. However, this interconnectedness also creates channels for economic shocks to propagate across borders. Previous research has demonstrated the propagation of shocks through production networks, citing examples such as the impact of natural disasters and the Great East Japan Earthquake. These studies, however, often focus on aggregated data, neglecting the intricate topology of the firm-level global supply network. This paper addresses this gap by examining how economic shocks spread at the firm level across international supply chains, focusing on the unequal distribution of risk and its implications for global inequality. The core research question is to quantify and analyze the unequal distribution of systemic risk across countries due to the structure of the global supply network, and to assess whether this risk is compensated by economic benefits.

The authors review existing literature on the effects of globalization, highlighting both its positive impacts on economic growth and income reduction and its negative consequences, such as environmental pollution and exploitation of labor in developing countries. They also examine previous research on shock propagation in production networks, focusing on studies using input-output analysis and agent-based models. A key point is the limitation of prior work which often uses highly aggregated data, obscuring the crucial role of the detailed firm-level network topology in shock propagation. The authors discuss the application of network centrality measures, particularly DebtRank, initially developed for financial networks, to analyze systemic risk in supply networks. They note the differences between financial and production networks, emphasizing the focus on production networks and systemic risks in their study.

The study uses a global firm-level dataset from Standard & Poor's Capital IQ platform, focusing on supplier-customer relationships for the year 2017. The data includes firm identification, location, industry, and sector information. The authors preprocess the data, removing firms lacking key information and those from countries with too few firms or known for having significant numbers of offshore firms. This results in a network of 230,970 firms and 660,701 links. They employ a modified DebtRank algorithm to quantify a country's exposure to economic shocks originating from firm defaults in other countries. This involves simulating the spread of shocks through the network, considering both direct and indirect effects. Two key metrics are defined: Country-Firm Exposure (Eᵢᵈ), representing the fraction of a country's production lost due to a specific firm's failure; and Country-Country Exposure (Eᶜᵈ), representing the average exposure of country *d* to shocks from country *c*. The authors distinguish between upstream and downstream cascades. They analyze the exposure matrix, considering different levels of aggregation (geographic regions, income groups). Additionally, they investigate the relationship between exposure and economic indicators (GDP per capita, GDP growth).

The analysis reveals a geographically clustered structure of country-country exposures, with strong connectivity and exposure within regions. Richer countries demonstrate a significantly higher propensity to export systemic risk to poorer countries, an asymmetry not reciprocated. While the number of links between countries partially explains exposure, the network topology at the firm level significantly influences the spread of shocks. The study finds a strong negative correlation between a country's total exposure and its GDP per capita, indicating that poorer countries are disproportionately vulnerable. Contrary to expectations, higher exposure is not associated with higher GDP growth rates, suggesting that higher risk does not translate to higher returns. The distribution of exposure is highly unequal, with a Gini coefficient of 0.83 (compared to 0.59 for GDP), highlighting a new dimension of global inequality. Analysis of upstream cascades shows similar patterns, with high inequality to the disadvantage of poorer countries.

The findings demonstrate that the global distribution of exposure to economic shocks is highly unequal and structured, with richer nations disproportionately exporting risk to poorer nations. This inequality is not simply a reflection of economic size or trade volume but is largely determined by the intricate structure of the firm-level global supply network. The lack of a risk premium contradicts the common assumption that higher risk is compensated by higher returns. This has important implications for development policy, suggesting that the benefits of globalization may be unequally distributed and that poorer countries bear a disproportionate share of the risk. The high Gini coefficient for exposure underlines the systemic nature of this inequality, exceeding that of GDP inequality and highlighting a new area for policy intervention.

This paper provides a novel method for quantifying international economic exposures using firm-level supply network data, revealing a highly unequal distribution of systemic risk globally. The findings highlight the need for global efforts to collect and monitor supply network data, to promote fairer risk distribution through network restructuring, and to consider systemic risk management strategies for international supply networks. Future research could explore how network structure perpetuates existing inequalities and develop strategies to mitigate systemic risks.

The study acknowledges several limitations. The dataset used, while substantial, represents only a fraction of the global firm-level network, potentially leading to sample bias. The absence of information on transaction volumes or the type of goods traded limits the precision of the shock propagation analysis. The model simplifies shock propagation by assuming a complete firm failure and neglecting factors like inventories and substitution dynamics. Future work should incorporate more granular data, richer model dynamics, and empirical default probabilities.