Uncovering Systemic Risk in ASEAN Corporations: A Framework Based on Graph Theory and Hidden Models

Abstract

In the context of an ever-evolving global economy, ASEAN companies face dynamic systemic risk that reshapes their financial interrelationships. This study examines the transmission of these risks using advanced graph theory techniques, particularly the measurement of eigenvector centrality based on Euclidean distances, combined with a hidden model that incorporates macroeconomic variables, such as GDP. The research focuses on identifying critical nodes within the corporate network, evaluating their contagion potential—both in terms of reinforcing resilience and amplifying vulnerabilities—and analyzing the influence of external factors on the network’s structure and behavior. The findings offer an innovative framework for managing systemic risk and provide strategic guidelines for the formulation of economic policies in emerging ASEAN markets.

1. Introduction

In today’s era of rapid globalization and economic interdependence, emerging markets have assumed a critical role in shaping the global financial landscape. Among these, the Association of Southeast Asian Nations (ASEAN) stands out as a uniquely dynamic economic bloc. Comprising ten member states—Indonesia, Malaysia, the Philippines, Singapore, Thailand, Brunei, Vietnam, Laos, Myanmar, and Cambodia—ASEAN has experienced sustained economic growth, with a combined GDP exceeding USD 3 trillion and a population of over 650 million (ASEAN Secretariat 2022). This diversity in economic development, regulatory frameworks, and market maturity creates an environment where the transmission of systemic risk warrants rigorous examination.

Historically, much of the research on systemic risk has concentrated on well-known emerging markets such as the BRICS nations (Chinazzi and Fagiolo 2015; Martínez-Jaramillo et al. 2010). However, ASEAN-listed companies present a distinctive context due to their rapid industrialization, heterogeneous financial systems, and increasing integration into global capital flows (Swenson and Woo 2019; Segura and Jing 2020). The region’s strategic geographic location and its pivotal role in global supply chains further heighten its relevance in studies of financial contagion and risk propagation. As ASEAN economies continue to evolve, understanding the mechanisms through which financial shocks traverse corporate networks becomes indispensable for both policymakers and market participants.

Recent methodological advancements have paved the way for a more nuanced analyses of complex financial networks. Graph theory, a mathematical framework for examining network structures, has emerged as a powerful tool to decipher the intricate web of inter-firm connections. By representing each company as a node and the financial dependencies between them as edges, graph theory allows researchers to capture both the direct and indirect pathways of risk transmission. Among the various measures available, eigenvector centrality—when computed using Euclidean distance metrics—provides critical insights into the influence of individual nodes. This measure accounts not only for the quantity of connections but also for the quality, thereby identifying those firms whose influence is magnified by their connections with other highly influential entities (Fantazzini et al. 2020; Chen et al. 2019). Such a refined approach is essential in an environment like ASEAN, where the heterogeneous nature of financial institutions may lead to non-uniform patterns of risk propagation.

Complementary to network analysis, hidden models have recently gained prominence for their ability to uncover latent dynamics within financial systems. Hidden Markov models (HMMs), in particular, are adept at identifying unobservable states of market stability and vulnerability by integrating macroeconomic variables—most notably, gross domestic product (GDP). In the ASEAN context, where GDP fluctuations can significantly influence market sentiment and corporate performance, HMMs provide a dynamic framework for assessing the transitions between stable and unstable states. By coupling macroeconomic insights with detailed network structures, hidden models enable a more comprehensive assessment of systemic risk, capturing both exogenous shocks and endogenous network vulnerabilities (Bikas and Glinskytė 2021).

The integration of graph theory and hidden models offers a novel methodological framework that bridges micro-level financial interdependencies and macro-level economic forces. This study leverages such an integrated approach to examine systemic risk transmission among publicly traded companies in the ASEAN region. Our analysis unfolds in two distinct yet interrelated stages. First, we construct a detailed corporate network using financial data from ASEAN-listed firms. In this network, nodes represent individual companies, and edges are weighted based on Euclidean distances calculated from key financial indicators. By applying eigenvector centrality measures, we identify critical nodes that serve as potential conduits for risk—either by absorbing shocks or by propagating financial distress throughout the network (Fantazzini et al. 2020). Second, we incorporate a hidden Markov model that utilizes GDP as a proxy for external economic conditions, allowing us to capture latent states of systemic risk and assess how these external factors influence the network’s stability (Bikas and Glinskytė 2021).

ASEAN’s significance in the global economy is underscored by its impressive growth trajectory over the past two decades. The region has consistently registered annual growth rates often exceeding 5%, driven by robust domestic demand, increased foreign direct investment, and significant infrastructural improvements (ASEAN Investment Report 2021). Moreover, ASEAN’s strategic position as a hub for international trade—facilitated by its membership in various free trade agreements—renders its financial system both a magnet for investment and a potential channel for global financial contagion. The interplay between rapid economic expansion and inherent market vulnerabilities makes the ASEAN region particularly susceptible to systemic risk. Episodes such as the Asian Financial Crisis of the late 1990s and the more recent economic disruptions resulting from the COVID-19 pandemic have highlighted the region’s sensitivity to both internal shocks and external disturbances (Duncan and Kabundi 2013).

The academic literature has increasingly recognized the importance of advanced quantitative techniques in capturing the multifaceted nature of systemic risk. For example, studies published have demonstrated the efficacy of graph-theoretic approaches in quantifying risk propagation within complex financial networks. Bikas and Glinskytė (2021) introduced a novel method for measuring systemic risk using eigenvector centrality, showing that firms with high centrality values play a disproportionate role in influencing network stability. Similarly, Cerqueti et al. (2021) employed a complex network approach to empirically analyze interbank markets, highlighting the potential for contagion in highly interconnected financial systems. These contributions, among others, underscore the necessity of incorporating sophisticated analytical tools to unravel the dynamics of systemic risk in emerging markets.

In parallel, the application of hidden models in financial risk analysis has garnered substantial attention. Magner and Valle (2022) demonstrated the utility of hidden Markov models in capturing regime shifts in financial markets, while Newman (2018) extended this approach to assess systemic risk within financial networks. By integrating macroeconomic indicators into the HMM framework, these studies provide valuable insights into the latent states of financial stability and the conditions under which systemic risk is likely to escalate. Such methodological innovations are particularly pertinent for the ASEAN region, where the confluence of rapid economic growth and market heterogeneity creates complex risk dynamics that traditional models may fail to capture.

The present study seeks to contribute to this growing body of literature by offering an integrated framework that combines graph theory and hidden Markov models to analyze systemic risk transmission among ASEAN-listed companies. This approach not only facilitates a detailed mapping of inter-firm relationships but also allows for the dynamic assessment of external economic influences on network stability. Our empirical strategy involves constructing a corporate network based on financial data, followed by the computation of eigenvector centrality measures to identify key nodes. The subsequent integration of GDP into a hidden Markov model enables us to capture latent risk states and quantify the impact of macroeconomic fluctuations on the transmission of systemic risk.

The implications of this research are manifold. For policymakers, the identification of critical nodes within the ASEAN corporate network can inform the design of targeted regulatory interventions aimed at mitigating systemic risk. By preemptively identifying firms that have the potential to trigger widespread contagion, regulators can implement measures to enhance market resilience and prevent financial crises. For investors, understanding the intricate pathways through which risk propagates offers valuable guidance for portfolio diversification and risk management. The ability to pinpoint firms that serve as risk amplifiers provides a more nuanced basis for investment decisions, ultimately contributing to the stability of financial markets in the region.

Furthermore, the methodological innovations presented in this study have broader applicability beyond the ASEAN context. The integrated framework that combines graph theory with hidden Markov models offers a versatile tool for analyzing systemic risk in other emerging markets, where similar dynamics of financial interdependence and economic volatility may be at play. Future research could extend this framework by incorporating additional macroeconomic and microeconomic variables—such as environmental, social, and governance (ESG) criteria—to further enhance the predictive power of systemic risk models in today’s increasingly interconnected global economy.

In conclusion, as ASEAN continues to ascend as a pivotal player in the global economic arena, the need to understand and manage systemic risk has never been more pressing. This study provides a rigorous, integrated framework that bridges advanced network analysis and dynamic macroeconomic modeling to elucidate the transmission mechanisms of systemic risk in ASEAN-listed companies. By drawing on state-of-the-art methodologies and leveraging empirical evidence from leading publications (Chen et al. 2019; Fantazzini et al. 2020; Bikas and Glinskytė 2021), our research offers both theoretical and practical contributions that are essential for fostering resilient financial systems in emerging markets worldwide.

2. Literature Review

Financial integration, while a catalyst for economic growth through enhanced capital mobility and risk diversification, also magnifies exposure to external shocks. This duality is particularly pronounced in emerging markets, where rapid integration can precipitate unforeseen systemic vulnerabilities. Early studies (Sifat et al. 2022; Martínez-Jaramillo et al. 2010) predominantly examined the BRICS economies, establishing a baseline understanding of liquidity enhancements and risk-sharing mechanisms. However, recent investigations argue that ASEAN’s distinctive economic and regulatory milieu necessitates a specialized analytical framework.

ASEAN comprises economies at varying stages of development, with regulatory infrastructures that range from highly sophisticated to rudimentary. Newman (2018) argues that these regulatory disparities intensify systemic fragilities, especially during periods of accelerated economic integration. Despite robust growth rates—as documented in the ASEAN Investment Report (2021), where annual growth frequently exceeds 5%—the region’s financial architecture remains susceptible to external disturbances. Rabiner (1989) further contends that, while cross-border capital flows serve as engines for growth, they simultaneously act as transmission channels for volatility, thereby demanding more nuanced risk management strategies. Rabiner (1989) extend this argument by highlighting the phenomenon of multichannel contagion, wherein simultaneous shocks across various asset classes—equities, bonds, and currencies—challenge conventional risk assessment models.

Graph theory has emerged as an indispensable methodological framework for dissecting the complex interdependencies that characterize modern financial systems. By representing financial institutions as nodes and their interrelationships as edges, network analysis transcends the limitations of bilateral correlation measures. A key construct within this domain is eigenvector centrality, which evaluates the systemic importance of a node not solely based on its number of direct connections but also by incorporating the centrality of its neighbors. Bikas and Glinskytė (2021) demonstrated that this metric provides a spatially nuanced understanding of risk propagation, revealing that nodes with elevated eigenvector centrality can serve as critical conduits for systemic shocks.

Building on this framework, Cerqueti et al. (2021) applied complex network methodologies to interbank markets and identified that a small cluster of highly interconnected nodes is instrumental in shock transmission. Such findings, further supported by Wibawa et al. (2024) suggest that traditional risk metrics may significantly underestimate the systemic implications of network topology. The practical relevance of these methodologies is underscored by their growing adoption among regulatory bodies and financial institutions, which increasingly rely on network-based analyses to inform real-time systemic risk monitoring and mitigation strategies.

Hidden Markov models (HMM) represent a significant methodological advancement in the quantification and prediction of systemic risk. Unlike conventional econometric models that rely exclusively on observable variables, HMMs capture latent regimes—periods characterized by either pronounced vulnerability or relative stability. (Magner and Valle 2022) provides compelling evidence that HMMs can detect subtle regime shifts, offering early warning signals that precede financial distress. This capacity is particularly critical in volatile markets, where traditional models may overlook emerging patterns of systemic risk.

Further advancing the field, Newman (2018) integrated HMMs with financial network data to capture the temporal evolution and spatial dissemination of risk. In ASEAN markets, where the macroeconomic indicators such as GDP growth, inflation, and exchange rates exert significant influence, HMMs have been shown to dynamically assess how these external conditions modulate internal network vulnerabilities (Linnenluecke et al. 2016). Moreover, Wibawa et al. (2024) demonstrated that the predictive prowess of HMMs facilitates the anticipation of regime transitions, thereby offering a quantifiable basis for preemptive policy and investment decisions. This integration of time-series analysis with network dynamics constitutes a powerful tool for forecasting systemic risk and underscores the need for its broader application in emerging market contexts.

The confluence of network theory and hidden Markov models marks a significant methodological breakthrough in systemic risk research. This integrative approach effectively bridges the gap between micro-level interfirm interactions and macro-level economic dynamics. Empirical investigations by Cerqueti et al. (2021) and Newman (2018) illustrate that hybrid models can robustly predict systemic risk episodes by capturing both the structural vulnerabilities inherent in financial networks and the dynamic shifts in macroeconomic regimes.

In the ASEAN context, where market structures and regulatory frameworks are characterized by pronounced heterogeneity, such integrated methodologies are especially pertinent. Cerqueti et al. (2021) argue that synthesizing network metrics with HMM-derived regime analyses provides a more holistic view of systemic risk. Their findings suggest that early warning indicators derived from network analysis—such as shifts in eigenvector centrality or clustering coefficients—can be effectively linked to the macroeconomic stress signals identified by HMMs. This dual analytical perspective not only enhances the precision of risk assessments but also informs the design of targeted regulatory interventions.

The promise of these integrative frameworks lies in their potential to reconcile theoretical insights with practical risk management strategies. As global financial systems become increasingly interconnected and complex, the ability to dynamically monitor and predict systemic risk is of paramount importance. Integrated methodologies thus offer a pathway for both advancing academic theory and guiding empirical practices that enhance financial stability.

Moreover, the incorporation of Environmental, Social, and Governance (ESG) factors into systemic risk analysis represents a critical extension of the existing framework, particularly in the context of ASEAN-listed firms. As financial integration accelerates across emerging markets, the structural interdependencies captured by eigenvector centrality and hidden Markov models (HMMs) must be complemented by non-financial risk indicators to fully account for systemic vulnerabilities. ESG metrics—such as carbon emissions, labor practices, and board diversity—offer a lens through which to assess how sustainability-related risks propagate through corporate networks. For instance, firms with high eigenvector centrality in sectors like energy or financial services may exhibit elevated ESG-related risks, which could amplify their role as conduits for systemic shocks during periods of environmental or governance-related stress (Cerqueti et al. 2021). This integration is especially pertinent in ASEAN, where regulatory frameworks for sustainability disclosure remain uneven, potentially obscuring latent risks that traditional financial metrics fail to capture (Wibawa et al. 2024). By embedding ESG factors into the network-based risk model, researchers can achieve a more comprehensive understanding of how non-financial drivers interact with macroeconomic fluctuations to shape systemic stability.

From a practical standpoint, the inclusion of ESG considerations in systemic risk assessments has profound implications for both investment strategies and regulatory policies in emerging markets. Investors, equipped with insights into how ESG performance modulates a firm’s systemic importance, can make more informed decisions regarding portfolio diversification and risk hedging. For example, firms with strong ESG profiles but high eigenvector centrality may present lower long-term risk, as their sustainability practices could mitigate exposure to regulatory or reputational shocks (Magner and Valle 2022). Conversely, firms with poor ESG scores in critical network positions may warrant reduced exposure or targeted hedging strategies. On the regulatory front, policymakers could leverage ESG-augmented risk models to design more precise interventions, such as mandating enhanced ESG disclosures for firms identified as systemically important through network analysis (Redondo Alamillos and de Mariz 2022). This approach not only strengthens financial stability but also aligns with broader sustainability goals, fostering resilience in ASEAN’s rapidly evolving economic landscape.

3. Methodology

3.1. Study Predictions

This study posits that the integration of graph-theoretic measures and hidden Markov models (HMMs) can effectively elucidate the mechanisms underlying systemic risk transmission in ASEAN-listed companies. Grounded in the literature on financial networks and dynamic risk modeling, we developed a set of interrelated hypotheses that bridge micro-level network structures with macroeconomic dynamics.

Hypothesis 1. 

Eigenvector Centrality as a Predictor of Systemic Influence.

Building on the premise that a firm’s influence within a corporate network is determined not only by its number of direct connections but also by the prominence of its neighbors, we hypothesize that firms exhibiting high eigenvector centrality scores act as pivotal nodes for risk propagation. Specifically, when eigenvector centrality is computed via Euclidean distance metrics, it encapsulates both the intensity and quality of inter-firm linkages (Fantazzini et al. 2020; Chen et al. 2019). Consequently, we predict that:

• Prediction 1a: ASEAN-listed firms with eigenvector centrality values above a critical threshold (e.g., >0.7) will exhibit a significantly higher propensity to transmit financial shocks throughout the network.
• Prediction 1b: The influence of these central nodes will be observable not only in terms of direct contagion but also via indirect effects that amplify systemic risk across multiple sectors.

This hypothesis aligns with the network theory literature, which suggests that highly central nodes can serve as both stabilizing agents and potential conduits for systemic collapse under adverse conditions (Cerqueti et al. 2021).

Hypothesis 2. 

Centrality Dispersion and Sectoral Vulnerability.

The second hypothesis examines the distribution of centrality within industry sectors. We posit that a concentrated centrality—where a few firms dominate the network—will be associated with heightened systemic vulnerability. In contrast, a more equitable distribution of centrality among firms within a sector should confer greater resilience against localized shocks. Thus, we make the following predictions:

• Prediction 2a: Sectors characterized by a high concentration of central nodes will demonstrate elevated volatility and a higher likelihood of contagion during periods of economic stress.
• Prediction 2b: Conversely, sectors with a more dispersed centrality structure will exhibit lower systemic risk, as the impact of a shock is more likely to be absorbed by the network rather than amplified.

Empirical studies in both interbank markets and broader corporate networks support this view, demonstrating that network homogeneity can play a critical role in dampening or exacerbating financial instability (Fantazzini et al. 2020; Chen et al. 2019).

Hypothesis 3. 

Macroeconomic Modulation of Latent Risk States via Hidden Markov Models.

The third hypothesis integrates external macroeconomic forces into our risk transmission framework using hidden Markov models. We posit that aggregate GDP—the key indicator of ASEAN regional macroeconomic health—plays a crucial role in modulating the latent states of systemic risk within the corporate network. Specifically, we make the following predictions:

• Prediction 3a: During periods of robust economic growth, as indicated by rising GDP, the probability of the network transitioning into a high-risk state will be reduced due to improved liquidity and investor confidence.
• Prediction 3b: In contrast, during periods of economic contraction or instability, a decline in GDP will correlate with an increased probability of the network entering a latent high-risk state, thereby heightening the potential for systemic contagion.

This hypothesis is supported by the literature on dynamic risk modeling, where hidden Markov models have been successfully used to capture regime shifts in financial markets (Bikas and Glinskytė 2021). The integration of GDP into the HMM framework allows us to infer the latent risk conditions that are not immediately observable for micro-level data alone.

3.1.1. Operationalization and Empirical Strategy

To empirically test these hypotheses, our study follows a two-stage analytical strategy:

• Network Construction and Centrality Analysis:

We first construct a comprehensive corporate network for ASEAN-listed companies using financial indicators that the capture key aspects of firm performance and inter-firm connectivity. Euclidean distance metrics are applied to quantify the similarity between firms, and eigenvector centrality is computed to identify the most influential nodes within the network. This approach not only measures the direct connectivity of each firm but also weights its importance by the centrality of its neighbors, thus providing a nuanced view of systemic influence (Fantazzini et al. 2020; Chen et al. 2019).

• Dynamic Risk Modeling via Hidden Markov Models:

Next, we incorporate macroeconomic data—specifically, the aggregate GDP of ASEAN—into a hidden Markov model to capture the latent states of systemic risk. This model is designed to detect transitions between stable and unstable regimes, thereby linking macroeconomic fluctuations to changes in network vulnerability (Bikas and Glinskytė 2021). The HMM framework allows for the estimation of transition probabilities between latent states, which can then be correlated with changes in the network’s centrality distribution and other risk metrics.

3.1.2. Implications for Theory and Practice

The integration of these methodologies is anticipated to yield several theoretical and practical contributions. Theoretically, our approach extends the current understanding of systemic risk by demonstrating how micro-level network properties interact with macroeconomic conditions to influence the stability of financial systems. Practically, the findings will offer valuable insights for regulators and investors by identifying critical nodes and vulnerable sectors within the ASEAN market, thereby informing strategies for risk mitigation and portfolio diversification.

By rigorously testing these hypotheses with robust empirical techniques, our study aims to provide a comprehensive framework for assessing systemic risk in ASEAN-listed companies. The anticipated outcomes will not only contribute to the academic literature on financial contagion and network theory, but will also have direct implications for policy formulation and investment decision making in emerging markets.

3.2. Variables Studied

Building on the classic risk-return tradeoff, our analysis posits that higher profitability is intrinsically linked to an elevated level of investment risk. In this framework, several financial indicators are interdependent, serving as critical measures for assessing this tradeoff. Specifically, metrics such as Return on Assets (ROA), Return on Equity (ROE), Sales Ratio, Working Capital, Accounts Receivable Turnover, Cash Conversion Cycle, Supplier Payment Period, Cost of Sales Ratio, and Debt Ratio collectively provide a comprehensive picture of a company’s financial health and risk profile.

For example, ROA evaluates how efficiently a company generates profits relative to its total assets:

$$ROA={\displaystyle \frac{Net Income}{Total Assets}}$$

(1)

Similarly, ROE measures a firm’s ability to produce profits from its shareholders’ equity, reflecting the effectiveness with which invested capital is utilized:

$$ROE={\displaystyle \frac{Net Profit}{Net Equity}}$$

(2)

The Sales Ratio, defined as the proportion of total sales to total assets, serves as an indicator of a company’s operational efficiency. A higher Sales Ratio suggests a more effective utilization of assets to generate revenue, while a lower ratio may indicate inefficiencies or underutilized assets:

$$Sales Ratio={\displaystyle \frac{Total Sales}{Averge Total Assets}}$$

(3)

In terms of risk assessment, liquidity indicators such as Working Capital and the Cash Conversion Cycle are particularly useful for evaluating short-term liquidity risk, as they reflect a firm’s ability to meet its immediate financial obligations. However, to capture broader dimensions of financial risk—including profitability, leverage, and systemic exposure—these measures are complemented by other indicators such as ROA, ROE, and the Debt Ratio. Furthermore, in the context of the network analysis, these firm-level metrics help identify potential nodes of vulnerability within the market structure, thereby contributing to a better understanding of systemic risk propagation under macroeconomic stress scenarios.

Working Capital, calculated as the difference between current assets and current liabilities, provides insights into a company’s ability to finance its day-to-day operations:

$$Working Capital = Current Assets-Current Liabilities$$

(4)

A robust Working Capital indicates ample resources to meet short-term obligations, thereby reducing investment risk. Conversely, low or negative Working Capital may signal financial distress and higher risk

Additionally, the Accounts Receivable Turnover ratio, computed as:

$$Accounts Receivable Turnover = {\displaystyle \frac{Net Credit Sales}{Average Accounts Receivable}}$$

(5)

offers a measure of the efficiency in managing credit and collection processes, which is critical for assessing liquidity and credit risk.

The Cash Conversion Cycle (CCC), defined as the difference between Days Payable Outstanding and Days Sales Outstanding, quantifies the time taken to convert investments in inventory and other resources into cash flows from sales:

$$Cash Conversion Cycle=Days Payable Outstanding-Days Sales Outstanding$$

(6)

Another essential metric is the Supplier Payment Period, which measures the average time a company takes to settle its accounts payable. It is calculated as:

$$Supplier Payment Period={\displaystyle \frac{Average Accounts Payable}{Cost of Sales}} \mathrm{*} 365$$

(7)

This period reflects the firm’s cash management practices—an extended period might indicate the strategic use of trade credit, while a shorter period could signal aggressive payment policies aimed at strengthening supplier relationships.

The Cost of Sales Ratio is pivotal for assessing operational efficiency, as it indicates the proportion of direct production costs relative to total sales:

$$Cost of Sales Ratio={\displaystyle \frac{Cost of Goods Sold}{Total Sales}}$$

(8)

Finally, the Debt Ratio, defined as:

$$Debt Ratio={\displaystyle \frac{Total Debt}{Average Total Assets}}$$

(9)

serves as a key indicator of financial leverage. A higher Debt Ratio implies greater reliance on debt financing, thereby increasing a company’s risk profile, whereas a lower ratio suggests a more conservative and stable financial structure.

3.3. Empirical Methodology

The integration of Eigenvector Centrality (EC) with Euclidean distance methodologies within the context of financial neural networks represents a sophisticated approach to elucidating complex interdependencies and influence patterns in financial markets. This methodology leverages the strengths of EC to assess influence based on the quality of connections within the network, while utilizing Euclidean distances to quantify the similarity between financial entities, thus providing a multidimensional perspective on market dynamics. The application of this integrated approach in neural network models facilitates a nuanced analysis of financial systems, where nodes represent financial entities and edges denote the relationships or flows between them, informed by both transactional interactions and attribute-based similarities.

3.3.1. Construction of the Corporate Network and Similarity Measurement

Each firm i is represented by a vector of financial attributes:

$$x_i=\left[x_{i1,}{x}_{i2},\dots .,x_{im}\right],$$

(10)

where each $x_{ik}$ corresponds to a relevant financial indicator (e.g., leverage, profitability, and market capitalization). The dissimilarity between two firms $i$ and $j$ is quantified using the Euclidean distance:

$$d_{ij}=\sqrt{{\sum }_{k=1}^m{\left(x_{ik}-x_{jk}\right)}^2}$$

(11)

This metric is chosen for its ability to directly capture differences in magnitude and for its intuitive geometric interpretation within the space ${\mathbb{R}}^m$. The quadratic accumulation of discrepancies emphasizes substantial differences in critical indicators, which is essential for distinguishing firms with varying risk profiles.

To transform this measure of dissimilarity into a similarity scale, we apply the function:

$$S_{ij}={\displaystyle \frac{1}{1+d_{ij}}},$$

(12)

ensuring that $S_{ij}\in \left(0,\left.1\right]\right.$, where values close to 1 indicate high similarity between firms. This transformation is particularly advantageous due to its simplicity and robustness against variations in data scales. In comparison to alternative metrics (e.g., Manhattan or Mahalanobis distances), the Euclidean distance provides a direct and less parametrically dependent solution, assuming that the data are properly scaled.

In addition, a transactional interaction matrix $T$ is incorporated to capture direct financial relationships (such as equity stakes and inter-firm loans). The weighted adjacency matrix $A$ s defined by the Hadamard product of $S$and $T$

$$A=S\circ T$$

(13)

where each element $A_{ij}=S_{ij} \circ T_{ij}$reflects both the similarity in financial profiles and the intensity of the interactions. This dual weighting provides a robust representation of the corporate network, forming a solid foundation for subsequent analysis of systemic risk transmission.

3.3.2. Eigenvector Centrality

To identify critical nodes within the network, eigenvector centrality is employed. This measure is based on solving the eigenvalue problem for the matrix A

$$A_V=\lambda v$$

(14)

and, according to the Perron–Frobenius theorem, there exists a unique, real, and positive principal eigenvector $v_{max}$ associated with the largest eigenvalue $\lambda max$for nonnegative matrices. The centrality for each firm $i$ is then defined as:

$$EC\left(v_i\right)=v_{max,i}$$

(15)

This measure is preferred over alternatives (such as degree or betweenness centrality) because it incorporates the global influence of each node, accounting for both its direct connections and the impact of its neighbors. Empirical studies have demonstrated that eigenvector centrality is a robust indicator of a node’s potential to amplify or mitigate the spread of shocks across the network (Fantazzini et al. 2020; Newman 2018). Additionally, threshold values (e.g., EC > 0.7 for critical nodes) are established through sensitivity analysis and align with empirical evidence in the financial network literature.

3.3.3. Dynamic Risk Modeling Using a Hidden Markov Model (HMM)

Given that systemic risk evolves over time, it is essential to capture the temporal dynamics of latent states. A hidden Markov model (HMM) is implemented to infer these unobserved states from macroeconomic variables (such as GDP growth). The HMM is specified as follows:

• Latent States: A sequence of states $\mathcal{Q}=\left\{q_1,q_2, \dots .., q_T\right\}$ is defined, where each $q_t$ belongs to the set $S=\left\{stable, vulnerable\right\}$
• Transition probabilities: The probability of transitioning from state $s_i$ at time $t$ to state $s_j$ at time $t+1$is given by:

  $$a_{ij}=P\left(q_{t+1}=s_j|q_t=s_i\right)$$

  (16)
  With ${\sum }_{j}aij=1$
• Emission Distribution: The observed variable ${\mathcal{O}}_t$ (e.g., GDP growth) is modeled using a Gaussian distribution:

  $$b_j\left({\mathcal{O}}_t\right)=P\left({\mathcal{O}}_t|q_t=s_j\right)=\mathcal{N}\left({\mathcal{O}}_t\right|{\mu }_j,{\delta }_j^2)$$

  (17)
• Initial Distribution:

The probability of starting in state $s_i$ is defined as:

$${\pi }_i=P\left(q_1=s_i\right)$$

(18)

Ensuring ${{\sum }_i\pi }_i=1$

Parameters $\lambda =\left\{A,B,\pi \right\}$ are estimated using the Baum–Welch algorithm (Rabiner 1989). A threshold is set for the “vulnerable” state (e.g., $P \left(q_t=vulnerable |\mathsf{O},\lambda \right)>0.8$, which, from both statistical and economic perspectives, indicates that the system is operating in a high-risk regime.

3.3.4. Integration of HMM Dynamics with Network Structure

The dynamic information obtained from the HMM is integrated with the network structure by adjusting each node’s centrality based on its latent state. This integration is formalized as:

$${EC}_{adjusted}\left(i,t\right)=EC\left(i\right)\cdot f\left(P(q_t=vulnerable |\mathsf{O},\lambda \right))$$

(19)

where $f(\cdot )$is a monotonically increasing function, such as $f\left(p\right)=1+{\beta }_p$ with $\beta >0$. This transformation implies that, during periods of high vulnerability, the relative importance of critical nodes is amplified, reflecting the system’s increased sensitivity to external shocks. This integration coherently combines the static structure of inter-firm relationships with the temporal dynamics of systemic risk.

3.4. Defense of Methodological Choices: A Geometric and Risk Management Perspective

3.4.1. Geometric Foundations and Implications of Euclidean Distance

The selection of the Euclidean distance is grounded in its well-established geometric properties and its suitability for high-dimensional financial data:

Definition and Fundamental Properties:

The Euclidean distance between two points $x_i$ and $x_j$ in ${\mathbb{R}}^m$ is defined as:

$$d_{ij}=\sqrt{{\sum }_{k=1}^m{\left(x_{ik}-x_{jk}\right)}^2}$$

(20)

This metric satisfies the key properties of any metric space:

○

Non-negativity: $d_{ij}\ge 0$and $d_{ij}=0$ if and only if $x_i=x_j$;

○

Symmetry:$d_{ij}=d_{ji}$;

○

Triangle Inequality: For any three points $x_i, x_j$ and $x_k$, we have $d_{ij}\le d_{ik}+d_{kj}$, ensuring that the direct route is the shortest.

Intuitive Geometric Interpretation

The Euclidean distance offers a straightforward geometric interpretation: it measures the length of the shortest path between two points in a multidimensional space, where each dimension corresponds to a financial indicator. Within the context of risk analysis, this enables an intuitive visualization of the dissimilarity between firms based on their financial profiles. Marked deviations in any single indicator translate into proportionally greater distances, potentially signaling higher relative exposure to risk.

Sensitivity to Data Variations:

A key feature of Euclidean distance is its quadratic sensitivity: by squaring the differences, the metric disproportionately emphasizes large deviations. In financial applications, this sensitivity is particularly desirable, as significant divergences in specific indicators—such as leverage, solvency, or liquidity—may reflect heightened vulnerability. Consequently, the measure is well suited for detecting early warning signals in systemic risk assessments.

Transformation to a Similarity Scale:

$$S_{ij}={\displaystyle \frac{1}{1+d_{ij}}},$$

(21)

normalizes the distances to a $\left(0,\left.1\right]\right.$ scale, ensuring that firms with nearly identical financial profiles yield similarity scores close to 1. This normalization facilitates a consistent comparison across different pairs of firms and integrates smoothly into the construction of the adjacency matrix, further enhancing its utility in network analysis.

Normalization and Methodological Considerations

Given that financial indicators often differ substantially in scale and units, all variables are standardized to zero mean and unit variance prior to computing distances. This normalization step is essential to ensure that each variable contributes equally to the distance metric, thereby preserving the interpretability and internal consistency of the resulting similarity measures and clusters.

Comparison with Alternative Metrics:

Although Euclidean distance is widely adopted for its simplicity and interpretability, it is not universally optimal. Cosine similarity, for instance, captures the angle between vectors and is insensitive to magnitude. This makes it suitable in contexts where the direction of variation is more informative than its absolute size—yet it may fail to flag critical risk differences when firms exhibit similar financial ratios but vastly different scales.

On the other hand, the Mahalanobis distance accounts for correlations among variables and adjusts distances accordingly. While methodologically appealing in theory, it requires a reliable estimation of the covariance matrix, which becomes problematic in high-dimensional settings or when sample sizes are limited, leading to instability and overfitting risks.

In contrast, the Euclidean distance applied to standardized data offers a pragmatic balance: it is robust, interpretable, and sensitive to financially meaningful variations without the computational burden or parametric assumptions of more complex alternatives. This makes it a sound choice for similarity-based clustering and network construction in the context of financial risk analysis.

3.4.2. Implications for Risk Management

From a risk management perspective, the geometric interpretation of similarity offers several advantages:

○

Visualization and Pattern Recognition: Managers can visualize clusters of firms that share similar financial characteristics. Such clusters may indicate sectors or groups with common vulnerabilities, enabling targeted mitigation strategies and tailored stress-testing procedures.

○

Identification of Critical Nodes: The similarity measure directly informs the construction of the adjacency matrix, which underpins the calculation of eigenvector centrality. This integration is vital for pinpointing nodes that, due to their geometric proximity to other high-risk firms, may act as critical conduits for risk contagion.

○

Adaptability and Scalability: The simplicity and clarity of the Euclidean distance make it highly adaptable across various contexts and scales, facilitating its incorporation into comprehensive risk analysis platforms. This adaptability is crucial for extending the methodology to incorporate additional variables or to analyze different markets and sectors.

4. Results

4.1. Sector-Level Centrality and Risk Profiles

Table 1. Centrality values by sector.

	Centrality Value	Risk Level	Risk Probability
Agriculture	0.51	Transition	0.42
Construction	0.63	Critical	0.7
Energy	0.73	Critical	0.82
Financial Services	0.83	Critical	0.9
Industrial	0.56	Critical	0.55
Minning and Chemical Industry	0.63	Critical	0.68
Retail	0.68	Critical	0.7
Health	0.51	Transition	0.35
Services	0.73	Critical	0.88

Prepared by the authors.

(Sectors) summarizes the centrality values, corresponding risk levels, and estimated risk probabilities for nine key sectors: Agriculture, Construction, Energy, Financial Services, Industrial, Mining and Chemical Industry, Retail, Health, and Services. The centrality values reflect each sector’s overall prominence within the ASEAN corporate network, while the risk level and risk probability capture the likelihood of transitioning to or remaining in a vulnerable state, as indicated by the HMM.

4.1.1. High-Centrality, High-Risk Sectors

Financial Services (centrality = 0.83, risk level = Critical, risk probability = 0.90) emerges as the most systemically influential sector. Its elevated centrality suggests strong interconnections with other sectors, magnifying the potential for contagion. The hidden Markov model indicates a 90% probability of this sector being in or entering a high-risk state, underscoring the pivotal role of financial institutions in systemic stability.

Services and Energy each register centrality scores of 0.73, coupled with critical risk levels and high-risk probabilities (0.88 and 0.82, respectively). These findings suggest that disruptions in these sectors could propagate swiftly through the network, given their extensive interdependence and pronounced macroeconomic sensitivity.

4.1.2. Moderate-Centrality, Elevated-Risk Sectors

Industrial (centrality = 0.56) and Mining and Chemical Industry (0.63) both exhibit critical risk levels, albeit with somewhat lower probabilities of high-risk states (0.55 and 0.68, respectively) compared to financial services or energy. The hidden Markov model thus indicates that while these sectors are structurally important, their risk probabilities, though significant, do not reach the extremes observed in the more interconnected sectors.

4.1.3. Transition and Stable Sectors

Agriculture (centrality = 0.51, risk level = transition, risk probability = 0.42) and Health (0.51, transition, 0.35) occupy transitional risk states. Their moderate centrality values imply moderate interconnections within the corporate network, resulting in less acute contagion pathways relative to sectors such as Financial Services or Energy. Nonetheless, the HMM probabilities suggest that under adverse macroeconomic conditions, these sectors may still shift into higher-risk regimes.

4.2. Firm-Level Analysis by Sector

To further elucidate the sources of systemic risk, we analyze the firm-level centrality values within each sector. These results offer insights into how specific companies may serve as nodes of contagion or resilience.

4.2.1. Agriculture

Within the Agriculture sector (

Table 2. Agriculture.

Companies	Centrality Value	Risk Level	Country
Charoen Pokphand Foods PCL.	0.997230388	Critical	Thailand
Charoen Pokphand Indonesia Tbk	0.099417474	Stable	Indonesia
FAP Agri Tbk	0.142523057	Stable	Indonesia
Golden Agri-Resources Ltd.	0.257261181	Stable	Singapore
Japfa Ltd.	0.139964845	Stable	Singapore

Prepared by the Authors.

: Agriculture), Charoen Pokphand Foods PCL (Thailand) stands out with a centrality value of 0.997, classified as Critical. This firm’s elevated centrality suggests it is highly interconnected and thus capable of significantly amplifying risk if faced with financial distress. By contrast, companies such as Charoen Pokphand Indonesia Tbk (0.099), FAP Agri Tbk (0.143), Golden Agri-Resources Ltd. (0.257), and Japfa Ltd. (0.140) are deemed Stable, reflecting lower systemic influence. Although Agriculture, as a whole, is in a Transition state, the presence of a single dominant node could alter the sector’s vulnerability under deteriorating macroeconomic conditions.

4.2.2. Construction

The Construction sector (

Table 3. Construction.

Companies	Centrality Value	Risk Level	Country
Mapletree Logistics Trust	0.997230388	Critical	Singapore
Frasers Centrepoint Trust	0.036573141	Stable	Singapore
Keppel DC REIT	0.022059562	Stable	Singapore
Pakuwon Jati Tbk	0.022055804	Stable	Indonesia
CapitaLand Ascendas REIT	0.021289794	Stable	Singapore
Ciputra Development Tbk	0.019164411	Stable	Indonesia
IREIT Global	0.014028383	Stable	Singapore
Land And Houses PCL	0.010241329	Stable	Thailand
Gamuda	0.008246703	Stable	Malaysia
Frasers Property Limited	0.007940636	Stable	Singapore

Prepared by the Authors.

: Construction) is dominated by Mapletree Logistics Trust (Singapore) with a centrality of 0.997, classified as critical. All other firms in this sector exhibit substantially lower centrality values (ranging from 0.036 to 0.007), categorized as stable. This stark disparity underscores a concentrated risk structure, where a single high-centrality entity may serve as a conduit for contagion. Given the sector’s overall critical status (centrality = 0.63, risk probability = 0.70), any adverse developments at Mapletree Logistics Trust could cascade through the broader ASEAN network.

4.2.3. Energy

In the energy sector (

Table 4. Energy.

Companies	Centrality Value	Risk Level	Country
Adaro Energy Indonesia Tbk	0.99999997	Critical	Indonesia
Tenaga Nasional	0.001275761	Stable	Malaysia
Pertamina Geothermal Energy Tbk	0.000744977	Stable	Indonesia
PTT PCL.	0.000504717	Stable	Thailand
Aboitiz Power Corporation	0.000482299	Stable	Philippines
Electricity Generating PCL.	0.000233787	Stable	Thailand
BANPU PCL.	0.000142819	Stable	Thailand
Thai Oil PCL.	0.000122236	Stable	Thailand
Dian Swastatika Sentosa Tbk	0.0000796	Stable	Indonesia
IRPC PCL.	0.0000707	Stable	Thailand
Harum Energy Tbk	0.0000433	Stable	Indonesia
Petrindo Jaya Kreasi Tbk	0.0000391	Stable	Indonesia

Prepared by the authors.

: Energy), Adaro Energy Indonesia Tbk (centrality = 0.9999, critical) emerges as a pivotal node. The remaining companies, including Tenaga Nasional (0.0013), Pertamina Geothermal Energy Tbk (0.0007), and PTT PCL (0.0005), exhibit minimal centrality, categorized as stable. Despite the overall sector being classified as critical (centrality = 0.73, risk probability = 0.82), the risk is highly concentrated in a dominant single firm. This configuration indicates that while most energy firms maintain lower connectivity, systemic contagion could be triggered if Adaro Energy experiences financial distress, especially in periods of adverse macroeconomic shifts.

4.2.4. Financial Services

The Financial Services sector (

Table 5. Financial services.

Companies	Centrality Value	Risk Level	Country
Saratoga Investama Sedaya Tbk	0.999975567	Critical	Indonesia
GT Capital Holdings	0.023567408	Stable	Philippines
BIDV	0.022561929	Stable	Vietnam

Prepared by the authors.

: Financial Services) features Saratoga Investama Sedaya Tbk (Indonesia) as the predominant node (centrality = 0.9999, Critical). With the sector showing a risk probability of 0.90, this finding confirms the pivotal role of financial institutions in systemic risk propagation. Other entities, such as GT Capital Holdings (0.0236) and BIDV (0.0226), exhibit relatively low centrality, highlighting the extent to which a single institution can dominate the sector’s overall risk profile.

4.2.5. Industrial

Firm-level results for the Industrial sector (

Table 6. Industrial.

Companies	Centrality Value	Risk Level	Country
Cemindo Gemilang Tbk	0.026287308	Stable	Indonesia
Sime Darby	0.008219951	Stable	Malasia
Aboitiz Equity Ventures	0.006749372	Stable	Philippines
JG Summit Holdings	0.006749372	Stable	Philippines
TPI Polene PCL	0.005587594	Stable	Thailand
Indah Kiat Pulp & Paper Tbk	0.004499133	Stable	Indonesia
Indofood Sukses Makmur Tbk	0.003184904	Stable	Indonesia
Thai Union Group PCL.	0.00251433	Stable	Thailand
Impack Pratama Industri Tbk	0.00235672	Stable	Indonesia
Siam City Cement PCL.	0.002118254	Stable	Thailand
Thai Beverage	0.001681672	Stable	Thailand
Indofood CBP Sukses Makmur Tbk	0.001677425	Stable	Indonesia
United Tractors Tbk	0.001333389	Stable	Indonesia
Keppel Ltd.	0.001239617	Stable	Singapur
Multistrada Arah Sarana Tbk	0.000903324	Stable	Indonesia
MD Pictures Tbk	0.000721271	Stable	Indonesia
Hoa Phat Group	0.000602652	Stable	Vietnam
Mayora Indah Tbk	0.000588541	Critical	Indonesia
Sembcorp Industries Ltd.	0.000493816	Critical	Singapur
Nusantara Sejahtera Raya Tbk	0.000411707	Critical	Indonesia
Nestlé Malaysia)	0.000276031	Critical	Malasia

Prepared by the authors.

: Industrial) reveal a predominantly stable landscape, with most companies—e.g., Cemindo Gemilang Tbk (0.0263), Sime Darby (0.0082), Aboitiz Equity Ventures (0.0067)—displaying low centrality values and stable risk classifications. However, four firms—Mayora Indah Tbk, Sembcorp Industries Ltd., Nusantara Sejahtera Raya Tbk, and Nestlé Malaysia—are classified as Critical despite lower centrality values, suggesting that their vulnerability is influenced by factors beyond network connectivity alone (e.g., leverage, liquidity constraints). The sector’s moderate overall centrality (0.56) but critical risk status (probability = 0.55) underscores the interplay between structural influence and broader macroeconomic pressures modeled by the HMM.

4.2.6. Mining and Chemical Industry

In the mining and chemical industry (

Table 7. Minning and chemical industry.

Companies	Centrality Value	Risk Level	Country
Amman Mineral Internasional Tbk.	0.516273508	Transition	Indonesia
Merdeka Copper Gold Tbk	0.473263462	Transition	Indonesia
PTT Global Chemical PCL	0.467120467	Transition	Thailand
Indorama Ventures PCL	0.450671715	Transition	Thailand
Indorama Ventures PCL.	0.450671715	Transition	Thailand
Trimegah Bangun Persada Tbk	0.430942263	Transition	Indonesia
Vale Indonesia Tbk	0.189161174	Stable	Indonesia
Adaro Minerals Indonesia Tbk	0.035682498	Stable	Indonesia
Bumi Resources Tbk	0.009137841	Stable	Indonesia
Prima Andalan Mandiri Tbk	0.008317333	Stable	Indonesia
Bumi Resources Minerals Tbk	0.005367967	Stable	Indonesia
Chandra Asri Pacific Tbk	0.00235585	Stable	Indonesia
Bayan Resources Tbk.	0.001088032	Stable	Indonesia

Prepared by the authors.

: Mining and Chemical), firms such as Amman Mineral Internasional Tbk (0.5163), Merdeka Copper Gold Tbk (0.4733), and PTT Global Chemical PCL (0.4671) occupy “Transition” risk states. These relatively higher centrality values suggest an intermediate level of network influence, aligning with the sector’s overall transitional status. In contrast, companies with markedly lower centralities, e.g., Chandra Asri Pacific Tbk (0.0024), Bayan Resources Tbk (0.0011)—are deemed stable, indicating limited structural impact on the network.

4.2.7. Retail

The retail sector (

Table 8. Retail.

Companies	Centrality Value	Risk Level	Country
Sumber Alfaria Trijaya Tbk	0.523914042	Transition	Indonesia
Berli Jucker PCL	0.280102993	Stable	Thailand
Emperador Inc.	0.168424372	Stable	Philippines
Masan Group	0.145109086	Stable	Vietnam
Century Pacific Food	0.081165407	Stable	Philippines
SM Prime Holdings	0.081165407	Stable	Philippines
San Miguel Food and Beverage	0.081165407	Stable	Philippines
Wilcon Depot	0.081165407	Stable	Philippines
CP All PCL.	0.035163249	Stable	Thailand
Berli Jucker PCL.	0.033342946	Stable	Thailand
CP All PCL	0.026874721	Stable	Thailand
Unilever Indonesia Tbk	0.007455203	Stable	Indonesia

Prepared by the authors.

: Retail) displays a mix of transitional and stable risk classifications. Sumber Alfaria Trijaya Tbk (Indonesia) stands out with a centrality of 0.5239, categorized as transition, mirroring the sector’s moderate connectivity (0.68) and critical risk level (probability = 0.70). Companies such as Berli Jucker PCL (0.2801), Emperador Inc. (0.1684), and Masan Group (0.1451) exhibit stable classifications and lower centralities, suggesting that they exert less influence on network-wide contagion dynamics.

4.3. Interpretation and Implications

The empirical findings derived from our integrated approach—merging eigenvector centrality with a hidden Markov model—provide compelling evidence of a dual pattern in systemic risk distribution among ASEAN-listed firms. Specifically, our results demonstrate that certain sectors, notably financial services, energy, and services, are characterized by highly concentrated risk, wherein a few dominant nodes possess significantly elevated eigenvector centrality scores. These nodes not only exert considerable direct influence on their immediate neighbors, but also engender cascading effects throughout the network. The hidden Markov model corroborates these structural findings by indicating a high probability of these sectors existing in a critical risk state, thereby underscoring their vulnerability to systemic shocks (Fantazzini et al. 2020; Chen et al. 2019).

In contrast, sectors such as agriculture, retail, and mining and chemical industry exhibit a more heterogeneous risk profile. Although these sectors display transitional risk states, as evidenced by moderate centrality values and lower HMM-derived risk probabilities, their systemic risk appears more contingent upon broader macroeconomic fluctuations. In these contexts, no single firm dominates the risk transmission process; rather, the diffuse network structure suggests that external economic variables—such as GDP fluctuations—play a more pivotal role in shifting these sectors from stability to heightened vulnerability (Bikas and Glinskytė 2021).

From a theoretical perspective, the present findings substantiate the proposition that the structural configuration of a corporate network is a critical determinant of systemic risk. Eigenvector centrality, as a recursive measure of influence, enables a nuanced assessment of both direct and indirect inter-firm linkages (Cerqueti et al. 2021). Our study further advances this literature by dynamically linking network structure to macroeconomic conditions via a hidden Markov model, thereby capturing the temporal evolution of risk states. This dual-faceted approach offers a more holistic understanding of risk transmission, aligning with contemporary views that emphasize the importance of both static network metrics and dynamic external factors in shaping systemic vulnerabilities (Bikas and Glinskytė 2021).

Practically, the implications of our findings are significant for both regulatory policy and strategic investment. Regulatory bodies should prioritize the monitoring of firms with exceptionally high eigenvector centrality within critical sectors such as financial services and energy, as these entities possess the potential to serve as conduits for widespread contagion. Targeted regulatory interventions, such as imposing higher capital buffers or more frequent stress-testing protocols, may prove effective in mitigating systemic risk (Fantazzini et al. 2020). Furthermore, the dynamic risk signals provided by the hidden Markov model suggest that policymakers should adopt flexible regulatory frameworks that can respond to rapid shifts in macroeconomic conditions.

For investors, our integrated model offers a sophisticated tool for risk assessment and portfolio management. By identifying both the structural vulnerabilities inherent in the corporate network and the dynamic shifts induced by macroeconomic variables, investors can more effectively diversify their portfolios and hedge against potential systemic shocks. For instance, in sectors where risk is highly concentrated, investors may consider reducing exposure or employing hedging strategies to offset potential losses during economic downturns. Conversely, in sectors with a more distributed risk, investment strategies might focus on capitalizing on the relative stability afforded by a diversified network structure.

In conclusion, our study provides a robust and academically rigorous framework for understanding systemic risk in ASEAN-listed companies. By integrating eigenvector centrality measures with dynamic hidden Markov modeling, we capture both the static architecture of financial networks and their evolution in response to external economic conditions. This comprehensive approach not only enriches the academic discourse on systemic risk and financial contagion, but also offers actionable insights for enhancing regulatory oversight and informing strategic investment decisions in emerging markets.

5. Conclusions

This study presents a novel integrated framework that combines advanced graph theory with dynamic hidden Markov models (HMMs) to investigate systemic risk transmission among ASEAN-listed companies. By coupling micro-level network metrics—specifically eigenvector centrality computed from Euclidean distance matrices—with macroeconomic dynamic modeling, our research offers a multifaceted perspective on how intrinsic corporate network structures interact with external economic shocks to shape systemic vulnerability.

Our empirical findings largely support the primary hypothesis that firms with high eigenvector centrality act as critical nodes capable of amplifying systemic risk. In sectors such as financial services, energy, and services, dominant firms exhibit exceptionally high centrality values, and the HMM consistently assigns these sectors high probabilities of being in a critical risk state. This convergence of structural and dynamic evidence confirms that high-centrality nodes are key conduits for risk propagation (Fantazzini et al. 2020; Chen et al. 2019). However, our analysis also revealed significant nuances. In sectors with a more dispersed network structures such as agriculture, retail, and mining and chemical industry—the anticipated buffering effect of a distributed network was only partially observed. Although these sectors generally show transitional risk states with moderate risk probabilities, our HMM results indicate that even modest adverse shifts in macroeconomic conditions (e.g., a GDP contraction) can trigger abrupt transitions into critical risk states. This nuanced outcome suggests, that while a diffuse network can attenuate the impact of any single node’s failure, the overall resilience is still heavily contingent upon favorable external economic conditions.

The second hypothesis posited that an even distribution of centrality within a sector would mitigate systemic risk. Our findings provide mixed support for this assertion. In sectors where centrality is more evenly distributed, the risk profile tends to be transitional rather than critical under stable economic conditions. However, these same sectors remain vulnerable to external shocks, as demonstrated by the sensitivity of the HMM-derived risk states to fluctuations in ASEAN’s aggregate GDP. In this context, the protective effect of network dispersion appears to be conditional rather than absolute. Thus, while our research confirms that network dispersion can dampen localized shock amplification, it also highlights that the diffusion effect is insufficient to fully offset the impact of macroeconomic downturns (Bikas and Glinskytė 2021).

Our third hypothesis—that macroeconomic conditions significantly modulate the latent risk states of the corporate network—was robustly supported. The HMM analysis reveals a clear correlation between GDP fluctuations and transitions between stable and vulnerable states. During periods of economic expansion, the probability of the network maintaining a stable state increases, whereas economic contractions precipitate a marked shift towards high-risk regimes. This dynamic interaction underscores the critical importance of incorporating external economic variables into systemic risk models, particularly in a region as economically diverse and rapidly evolving as ASEAN (Swenson and Woo 2019).

The novelty of our research lies in its integrated approach. While previous studies have typically employed either static network analysis or dynamic macroeconomic modeling in isolation, our methodology bridges these two paradigms. By dynamically adjusting network metrics in response to macroeconomic signals, our framework captures the evolving nature of systemic risk in real time—a significant advancement over traditional models that often assume static conditions. This innovation not only enhances predictive accuracy but also provides a more robust tool for both policymakers and investors. For regulators, our findings suggest that the targeted oversight of high-centrality firms could be a potent strategy to preempt cascading failures. Investors, meanwhile, can utilize our model to better diversify portfolios and hedge against systemic shocks by understanding the conditions under which risk transmission is the most acute (Cerqueti et al. 2021; Fantazzini et al. 2020).

In summary, this study confirms that systemic risk in ASEAN-listed companies is a product of both internal network structures and external macroeconomic forces. Although high-centrality nodes in sectors such as financial services, energy, and services substantially elevate systemic risk, even sectors with more evenly distributed networks remain vulnerable under adverse economic conditions. Our integrated approach represents a significant methodological advancement, providing actionable insights for risk mitigation in emerging markets. Future research could build on our framework by incorporating additional macroeconomic variables—such as interest rates, inflation, and ESG factors—to further refine risk assessment models and extend their applicability to other regions. Ultimately, our study deepens the theoretical understanding of financial contagion and offers practical strategies for enhancing market stability in an increasingly interconnected global economy.

Additionally, our study provides detailed and actionable recommendations tailored to specific sectors, thereby enhancing the practical application of our findings for systemic risk management informed by network structures. For highly vulnerable sectors—particularly financial services, energy, and services—it is crucial that regulatory bodies establish specialized monitoring protocols and early-warning systems precisely calibrated to firms exhibiting exceptionally high eigenvector centrality. These measures should be supported by comprehensive, scenario-based stress testing, which considers both firm-specific and network-wide vulnerabilities. Moreover, it is essential to develop and regularly update sector-specific contingency plans and crisis management frameworks that enable rapid and effective responses to potential systemic disruptions. Furthermore, promoting the diversification of inter-firm financial exposures and implementing rigorous capital adequacy standards aligned with centrality metrics could substantially mitigate the risk of cascading failures.

For sectors characterized by more distributed network structures, such as agriculture, retail, and the mining and chemical industry, the study recommends strengthening resilience through collaborative sector-wide initiatives. These initiatives might include the development of collective risk-sharing instruments, such as sector-specific insurance schemes or mutual guarantee funds, and the establishment of centralized information-sharing platforms to enhance transparency and coordination during periods of heightened economic uncertainty. Policymakers should facilitate cooperative frameworks that enable these sectors to collectively address vulnerabilities, thereby bolstering their capacity to withstand external shocks.

Despite its contributions, our study faces several important limitations that must be transparently addressed. Primarily, the application and generalization of our integrated analytical framework beyond the ASEAN region may be constrained by inherent differences in economic environments, regulatory contexts, market maturity, and corporate governance standards across diverse geographical regions. The unique structural characteristics and institutional frameworks of other economies could influence the applicability and predictive accuracy of our findings. Additionally, while the analysis predominantly employs eigenvector centrality and GDP fluctuations as the primary variables, future research could substantially enrich the analytical scope by incorporating additional macroeconomic indicators and variables such as interest rates, inflation rates, unemployment levels, political stability indicators, and ESG (environmental, social, and governance) factors. Exploring how technological disruptions, digital transformations, and industry-specific regulatory shifts impact systemic risk dynamics could further refine the model’s precision and robustness.

Future studies could also benefit from extending the empirical testing and validation of our proposed framework across different economic regions and diverse industrial contexts, including comparative analyses between emerging and developed markets. Such comprehensive investigations would not only enhance the generalizability and applicability of our findings but also provide deeper insights into the universality or specificity of systemic risk propagation mechanisms. Ultimately, these extensions would significantly enrich the toolkit available to policymakers, regulators, and financial institutions, empowering them to more effectively anticipate, prevent, and mitigate systemic financial crises in an increasingly interconnected global economic landscape.