The modern financial system is highly interconnected. Institutions engage in a diverse range of transactions and contracts with each other. Examples include the interbank lending market and the over-the-counter derivatives market. Interconnections facilitate risk sharing among institutions, but they also allow risks to be propagated from one institution to many others. The insights are evident in the recent financial crisis: initial losses caused the financial distress of a few institutions, which then spread via linkages to otherwise healthy institutions, resulting in systemic failures. For this reason, several recent studies have identified interconnectedness as a major source of systemic risk.

While the buildup of the interconnections is a collection of individual bank’s decisions, most existing studies analyze contagion in predetermined network structures and do not consider financial institutions' strategic behavior in forming links and choosing counterparties.

This paper highlights a novel channel of systemic risk due to negative externalities created when institutions strategically choose counterparties and decide to form linkages. I demonstrate this idea using a model of network formation between financial institutions under incomplete contracting. Namely, bank A can write contracts with bank B contingent on who bank B contracts with, but not on who B's counterparties contract with. Due to this realistic contracting friction, inefficient networks emerge in equilibrium, thereby increasing systemic risk.  

The theory has unique predictions on the composition of links and its effect on stability. More precisely, links between financially healthy banks can create value from risk-sharing, so they are socially ‘good links’; links between a healthy and a distressed bank have high social cost because they can cause system-wide failures through contagion. The equilibrium network is characterized by an inefficient composition. It features too many ‘bad links’, and too few ‘good links’. Both lead to excessive systemic risk and increase the probability of joint bank failures.

The intuition is as follows. Contracting with a financially distressed bank may enable that distressed bank to avoid liquidation. Such a link is ex ante profitable for a liquid bank because it can extract high surplus from its distressed counterparty. However, when a bank is severely distressed, linking with it can be socially costly. This is because the interconnections across banks allow counterparty risks to propagate, and thus the distressed assets would eventually contaminate the balance sheets of other banks in the network due to contagion. Nonetheless, a liquid bank still has strong incentive to link with such a distressed bank because of a profit-making motive, despite that connecting with this risky counterparty creates negative spillovers to the rest of the system. Furthermore, this network externality in turn reduces the participation to share risks among non-distressed banks. There are two forces at play: the transmission of distressed assets that should have been isolated, and the insufficient risk-sharing among non-distressed banks. The two mechanisms reinforce one another and lead to inefficiency.

While the prior literature largely focuses on the average soundness of the financial sector, the theory also identifies a novel indicator for the network inefficiency: the distress dispersion across financial institutions. An increase in the distress dispersion leads to a higher level of inefficiency. This positive relation arises from the wedge between individual and social incentives to link with a distressed counterparty. When distress dispersion is higher, the distressed institutions get more distressed and the liquid ones become more liquid. It is precisely then that a liquid institution has a stronger private incentive to form the socially ‘bad link’.

Policies on financial stability should not analyze institutions in isolation. In particular, this study provides new insights on financial regulation. First, regulators should aim to reduce network inefficiencies by overseeing the composition of financial linkages. While interconnections are important, this study concludes that financial regulation must consider the composition of links in addition to the degree of connectivity and the network structure. We should not simply control and reduce interconnections, but we should scrutinize the financial system, supervise and remove the ‘bad links’, and preserve and encourage the ‘good links’ that can create social value. Second, while detailed data on the precise linkages among financial institutions are limited, this paper draws a relation between the degree of network inefficiency and the cross-sectional distribution of fundamentals, thus contributing to the measurement of systemic risk.   

Jessie Jiaxu Wang is an Assistant Professor of Finance at the Arizona State University, W. P. Carey School of Business.