Who You Borrow From Matters More Than What You Borrow

Author: Denis Avetisyan

New research reveals that a company’s position within the lending network is increasingly influencing credit access, potentially eclipsing traditional financial metrics.

The analysis of bank and firm networks reveals that bank connectivity is markedly more heterogeneous than firm connectivity, and while a strong correlation exists between network strength and the size of relationships for banks, this relationship is only modest for firms - a disparity likely stemming from incomplete corporate loan data and suggesting fundamental differences in how these entities operate within the financial system. — The analysis of bank and firm networks reveals that bank connectivity is markedly more heterogeneous than firm connectivity, and while a strong correlation exists between network strength and the size of relationships for banks, this relationship is only modest for firms – a disparity likely stemming from incomplete corporate loan data and suggesting fundamental differences in how these entities operate within the financial system.

This paper demonstrates that network topology significantly impacts credit allocation, introducing systemic risks and necessitating a re-evaluation of macroprudential supervision.

Conventional assessments of credit risk prioritize firm fundamentals, yet fail to fully account for the influence of relational information. This paper, ‘Topology as information: Network effects in corporate lending’, investigates how a borrower’s position within the lending network-its network topology-increasingly drives credit allocation decisions. Our analysis of Italian financial data reveals that network connectivity systematically outperforms traditional balance-sheet metrics in predicting both access to credit and loan volume, suggesting a shift toward ‘topological certification’ where reputational collateral substitutes for physical assets. As firms navigate this evolving landscape, potentially fragmenting debt while facing complexity penalties, will macroprudential supervision need to adapt to monitor vulnerabilities hidden within these interconnected lending networks?

Beyond Balance Sheets: Mapping the Network of Credit

Conventional corporate finance typically assesses a borrower’s creditworthiness through firm-specific characteristics-factors like profitability, asset values, and collateral-effectively treating each company as an isolated entity. However, this approach often overlooks the crucial role of systemic influences and the interconnectedness within the broader economic landscape. A singular focus on fundamental attributes neglects the reality that a firm’s access to credit is heavily shaped by its relationships with other businesses, and by the overall structure of the lending network. This simplification can lead to an incomplete understanding of credit allocation, potentially mispricing risk and hindering accurate predictions of financial stability, as the health of one firm can propagate-positively or negatively-throughout the network via lending relationships.

Conventional understanding of corporate credit allocation centers on firm-specific characteristics – profitability, asset quality, and collateral – yet the `Corporate Lending Network` demonstrates that credit isn’t simply granted based on these fundamentals. Instead, the allocation of credit emerges as a property of the intricate web of relationships between companies. This research reveals that a firm’s access to credit isn’t solely determined by its own merits, but also by its position within the broader network – who it borrows from, who it lends to, and the connections of those entities. This challenges the notion that credit decisions are purely based on isolated firm characteristics, suggesting that the structure of interconnectedness itself plays a crucial, and often overlooked, role in shaping the flow of capital within an economy.

The allocation of credit isn’t solely determined by a firm’s inherent financial health; rather, the relationships a company cultivates within the broader corporate network significantly influence its access to capital. Research focusing on the Italian credit market demonstrates a ‘network substitution effect’, whereby strong inter-firm connections can compensate for weaker individual fundamentals. This suggests lenders increasingly leverage information gleaned from a borrower’s network – the creditworthiness of its suppliers and customers – to assess risk and extend credit. Consequently, a firm embedded in a robust network of reliable partners benefits from a richer information environment, effectively signaling its creditworthiness and potentially securing loans it might otherwise be denied based on traditional metrics alone. This highlights a fundamental shift in how credit is evaluated, moving beyond isolated firm characteristics to embrace the systemic importance of interconnectedness.

Inferring Quality: The Logic of Connectedness

Lenders within the corporate lending network utilize observational learning to assess borrower quality by monitoring the lending actions of other institutions. This process involves inferring characteristics about a borrower based on which lenders choose to extend credit, and the terms of that credit. The rationale is that other lenders possess private information or expertise in evaluating risk, and their decisions reveal signals about a borrower’s creditworthiness. Consequently, a borrower receiving loans from multiple, discerning lenders is perceived as having a higher quality profile, even without direct assessment by the observing lender. This behavior suggests that lending decisions are not solely based on independently gathered financial data, but are also influenced by the collective judgments expressed through network participation.

Topological certification utilizes a borrower’s position within the corporate lending network as an indicator of creditworthiness. This method assesses a firm’s connections to other borrowers – specifically, the quality and number of these relationships – to infer its own credit risk. The premise is that firms connected to borrowers with established credit histories are themselves more likely to be creditworthy. This network position is then incorporated into credit scoring models, functioning as a proxy for independent assessment of financial health. Consequently, lenders can utilize this topological data to evaluate loan applications, supplementing or even replacing traditional financial ratios and credit reports.

The issue of screening free-riding – where lenders benefit from the due diligence of others without bearing the full cost – is partially mitigated by utilizing network topology as a credit assessment tool. Analysis demonstrates that a borrower’s position within the corporate lending network provides predictive power exceeding that of conventional financial ratios. Specifically, lenders can infer creditworthiness by observing the lending decisions of firms connected to a prospective borrower, effectively leveraging shared information and reducing the need for independent, costly evaluations. Empirical results consistently show that network-based assessments outperform traditional financial metrics in predicting a firm’s access to credit, suggesting a measurable reduction in the costs associated with borrower screening.

Model predictions of firm and bank degrees closely match empirical data across both network-driven and balance-driven scenarios, and consolidated and unconsolidated samples, as evidenced by the alignment of model means (colored dots and shaded areas) with the identity line and raw data (gray dots) across degree bins.

The Fragility of Connection: Network Effects and Systemic Risk

The structure of the Corporate Lending Network demonstrates preferential attachment, a process where firms with a greater number of existing lending relationships are more likely to attract additional connections. This phenomenon isn’t random; new lending ties disproportionately favor already well-connected entities. Consequently, a small number of firms can accumulate a significant proportion of the total lending connections, creating an imbalanced network topology. This concentration of connections introduces systemic risk because the failure of a highly-connected firm can trigger a cascade of defaults throughout the network, amplifying financial instability beyond what would be expected in a more evenly distributed system. The effect isn’t merely about size; it’s the structure of connections that drives the increased vulnerability.

Information cascades within the corporate lending network occur when lenders base credit decisions primarily on the network position of a firm-specifically, the number and prominence of its connections-rather than on underlying financial fundamentals such as profitability, asset quality, or debt ratios. This behavior arises because lenders may interpret a firm’s strong network ties as a signal of creditworthiness, assuming that other well-connected institutions have already performed due diligence. Consequently, positive feedback loops can develop, where loans are extended to firms with weak fundamentals solely due to their network centrality. This overemphasis on network position diminishes the role of traditional risk assessment, leading to misallocation of capital and increasing systemic vulnerability, as a cascade of loan approvals based on flawed signaling can amplify financial distress throughout the network.

The observed network dynamics within the corporate lending network necessitate the development of novel regulatory frameworks, specifically Topological Supervision, which assesses systemic risk based on network position rather than solely on firm-specific fundamentals. Our modeling demonstrates a strong explanatory power for loan size, achieving an R-squared value of 0.78 when applied to consolidated sample data. This indicates that a substantial portion of the variance in loan amounts can be attributed to a firm’s position within the lending network, suggesting that network topology is a key determinant of credit allocation and, consequently, a critical factor for regulators to consider when evaluating financial stability and potential systemic risk.

Residual analysis from OLS regressions reveals that the consolidated sample exhibits negative skewness and a more normal distribution, while the unconsolidated sample displays higher kurtosis, indicating potentially non-normal error terms.

Benchmarking Connection: Validating Network Insights

The Gravity Model and Maximum-Entropy Counterfactual are statistical techniques utilized to assess credit access and establish a comparative network baseline. The Gravity Model, adapted from international trade analysis, predicts the likelihood of a lending relationship based on the size and characteristics of both borrower and lender; the probability of a connection is proportional to the product of their respective ‘masses’ – typically measured by total assets or loans outstanding. The Maximum-Entropy Counterfactual generates a null hypothesis network, preserving the marginal degree distribution of the observed Corporate Lending Network but randomizing edge connections to remove any systematic structure; this allows for comparison with the actual network to identify statistically significant relationships beyond those attributable to firm size alone. Both methods provide quantifiable metrics for evaluating network effects and disentangling intrinsic creditworthiness from network-induced connections.

Within the Corporate Lending Network, both Network Strength and Degree serve as quantifiable metrics for assessing systemic importance. Network Strength, calculated as the sum of edge weights connected to a node, indicates the total volume of credit a firm both receives and provides. Degree, representing the number of direct connections a firm has within the network, highlights its immediate influence and interconnectedness. Firms with high values for either metric are potentially critical nodes; their failure or disruption could propagate through the network, impacting multiple counterparties. Analysis focuses on identifying firms that exhibit disproportionately high Network Strength or Degree relative to their size or financial characteristics, signaling a level of systemic importance beyond what would be expected based on intrinsic creditworthiness alone.

The application of network analysis tools within the corporate lending network enables differentiation between a firm’s inherent creditworthiness and its position derived solely from network effects, thereby reducing risks related to adverse selection and moral hazard. Specifically, analysis reveals a Spearman correlation coefficient of 0.25 between the degree centrality of firms in the empirical network and a comparable null model – a network generated without firm-specific attributes. This statistically significant, yet moderate, correlation indicates that network topology is not solely attributable to firm size or inherent credit risk, but reflects genuine lending relationships and systemic influence beyond simple scale effects.

Beyond Collateral: The Rise of Reputational Credit

The structure of corporate lending increasingly relies on a firm’s position within the broader network of financial relationships, effectively functioning as ‘reputational collateral’. This means a company’s creditworthiness is not solely determined by its tangible assets, but also by its connections to other firms and banks – a strong network signaling trustworthiness and reducing perceived risk. Consequently, companies with limited physical collateral can still secure financing, as their network affiliations vouch for their reliability and ability to repay. This system introduces a dynamic where reputation, built through interconnectedness, partially substitutes for traditional security, reshaping the landscape of credit access and potentially creating new pathways for both growth and systemic vulnerability within the financial system.

The increasing reliance on reputational collateral within corporate lending networks presents a double-edged sword for financial stability. While firms lacking substantial tangible assets-those traditionally deemed higher risk-can now access credit through strong network connections and established reputations, this practice simultaneously creates new channels for systemic risk. Essentially, credit is extended beyond the capacity of conventional collateral, meaning that a cascade of defaults stemming from a single weakened entity can spread rapidly through interconnected lenders. This phenomenon amplifies exposure, as lenders become reliant not just on the borrower’s financial health, but also on the stability of its entire network, potentially masking vulnerabilities and increasing the likelihood of correlated failures across the lending landscape.

Accurate creditworthiness assessment increasingly demands consideration of both a firm’s tangible assets and its network connections within the corporate lending landscape. Recent analysis demonstrates a significant interplay between these factors, revealing a ‘degree elasticity’ of -1.04 for large corporate groups – meaning that for each additional banking relationship a firm maintains, loan size decreases by approximately one percent. This suggests that robust network connections can, to a degree, substitute for traditional collateral, yet also indicate a potential ceiling on borrowing capacity as lenders moderate exposure through diversified relationships. Understanding this dynamic is critical not only for individual risk assessment but also for preventing systemic crises, as over-reliance on network-based ‘reputational collateral’ could mask underlying vulnerabilities and contribute to widespread instability if network trust erodes.

Traditional credit risk assessment, centered on individual borrower characteristics and balance sheet strength, proves increasingly incomplete when viewed through the lens of interconnected lending networks. This research demonstrates a crucial shift in perspective, revealing that a firm’s position within the broader lending ecosystem significantly impacts access to credit – a network’s strength exhibiting a coefficient of 0.53 in unconsolidated samples. Notably, the study identifies a negative correlation – of -0.09 – between reported balance sheet debt and loan size, suggesting that lenders may be compensating for a lack of tangible assets with increased reliance on network connections and reputational collateral. This dynamic implies that systemic risk isn’t solely concentrated in individual firms, but arises from the complex interplay of relationships, necessitating a more holistic approach to financial stability that considers the interconnectedness of the entire lending landscape.

The study of corporate lending networks reveals a predictable human tendency: prioritizing relationships over raw fundamentals. This isn’t malice, but a deeply ingrained habit of assessing credibility through perceived connections – a form of ‘reputational collateral’ as the paper terms it. As David Hume observed, “It is only from a series of similar instances that we can form any general conclusion.” The increasing weight given to network topology in credit allocation isn’t a failure of economic models, but a manifestation of this human pattern. Every chart is a psychological portrait of its era, illustrating how easily fear and hope, channeled through interconnectedness, can overwhelm rational calculation, even in supposedly objective financial decisions. The paper’s findings confirm that humans aren’t driven by pure logic, but by emotional algorithms, predictably favoring the familiar face within the complex web of finance.

What’s Next?

The assertion that credit flows increasingly resemble rumour – dictated by who lends to whom, rather than demonstrable solvency – isn’t novel, but this work offers a disconcerting level of formalization. It maps the gossiping, the herd instincts, onto the balance sheets. The next step, predictably, will be attempts at prediction. Models will be built to forecast systemic fragility based on network centrality, eigenvector scores, and whatever topological metric proves most aesthetically pleasing. The quiet irony is that these models, however sophisticated, will merely reflect the biases of their creators – the same anxieties and optimistic fictions that already shape the lending landscape.

A more interesting, though likely neglected, avenue lies in understanding the perception of network structure. It isn’t enough to know how firms are connected; it matters how lenders believe they are connected. Misinformation, strategic signalling, and simple miscalculation will warp the perceived topology, creating phantom risks and opportunities. Markets don’t move – they worry, and worry is rarely based on complete information.

Ultimately, the pursuit of ‘topological certification’ feels like an attempt to impose order on a fundamentally chaotic system. It’s a human need to quantify the unquantifiable, to believe that a sufficiently complex algorithm can anticipate the irrationality of others. The real challenge isn’t predicting the next crisis, but accepting that crises are inevitable – a consequence of the fact that lending, like all social interactions, is driven by habit, hope, and the enduring illusion of control.

Original article: https://arxiv.org/pdf/2603.12417.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/