Domino Effects When Banks Hoard Liquidity: The French Network

Abstract

Résumé: Afin d'analyser les effets systémiques de la thésaurisation des banques sur la stabilité d'un réseau financier, cet article propose un nouveau modèle de contagion bancaire. Cette dernière estétudiée suivant deux canaux : les expositions bilatérales directes entreétablissements de crédit et les difficultés de financementà court-terme auxquelles les banques peuventéventuellement faire face si une crise de confiance vientà se matérialiser (comportement préventif de thésaurisation). En s'inspirant du rôle majeur joué par la thésaurisation pendant la crise financière de 2007-2009, le modèle développé dans cet article se distingue du traditionnel algorithme séquentiel de calcul de défauts en cascade largementévoqué et employé dans la littérature sur le risque systémique pour mesurer les effets d'un choc de marché sur l'ensemble d'un systme bancaire, en introduisant le comportement de thésaurisation des banques. Au-delà de la simple contagion via les expositions bilatérales, un tel phénoméne initié par certaines banques peut entraîner des problèmes de financementà court-terme pour d'autres. En s'appuyant sur les données d'expositions bilatérales des banques francaises collectées par l'Autorité de Contrôle Prudentiel, le secteur bancaire francais sembleêtre relativement robuste lorsqu'il està la fois soumisà un risque de marché (pertes sur les actifs de marché détenus par les banques dans leur portefeuille) et aux effets induits de la contagion par insolvabilité et par difficulté de financementà court-terme. Les résultats obtenus en termes de poids relatifs de chacun des canaux de contagionétudiés sur les pertes totales estimées mettent en exergue le caractère fondamental et complexe des effets de la thésaurisation.Mots-clés: Thésaurisation, contagion par insolvabilité et par difficulté de financement, réseaux financiers, risque systémique JEL classification: G01, G21, G28Abstract: We investigate the consequences of banks' liquidity hoarding behaviour for the stability of the financial system by proposing a new model of banking contagion through two channels, bilateral exposures and funding shortage. Inspired by the key role of liquidity hoarding in the 2007-2009 financial crisis, we incorporate banks' hoarding behaviour in a standard Iterative Default Cascade algorithm to compute the propagation of a common market shock through a banking system. In addition to potential solvency contagion, a market shock leads to banks liquidity hoarding that may generate problems of short-term funding for other banks. As an empirical exercise, we apply this model to the French banking system. Relying on data on banks bilateral exposures collected by France' Prudential Supervisory Authority, the French banking sector appears resilient to the combination of an initial market shock (losses on marked-to-market assets) and the resulting solvency and liquidity contagion. Moreover, the model gauges the relative weight of the various factors in the total loss.

Control of Interbank Contagion Under Partial Information

“…Our goal is to bring insight into the relation between a regulator's control, the overall connectivity in the network, and the banks' position in the network, either at the periphery or at the core. It is beyond our scope to construct a theoretical model of liquidity hoarding, and we take as given that liquidity hoarding is a meaningful channel of contagion, as documented by studies conducted using central bank data; see, e.g., [30,20]. 1 We study the optimal intervention policy of a lender of last resort that injects liquidity into the system either in the core banks or in the peripheral banks after the initial shock.…”

The Influence of Information Diffusion on Interbank Risk Contagion

“…Many scholars studied the motivation, mechanism, and impact of rollover risk and liquidity hoarding and found that they could lead to market freezes and collapses when rollover risk contagion occurs and banks consecutively hoard liquidity because of a precautionary or speculative motivation [2-5, 15, 26, 36-39]. e distinction between long-term and short-term interbank network in this paper is similar to that of Fourel et al [33] and the dynamics of the model are based on the model by Gai et al [4], except that we analyze only unsecured lending and add the possibility of liquidity hoarding even in the absence of a liquidity shortage. Even if the methodologies are different, this paper's enrichment is similar to that of Acharya and Skeie [2], where banks withdraw funding for precautionary motives depending on information about the counterparties.…”

“…erefore, many studies analyzed the relationship between assets and liabilities in the interbank funding network from the perspectives of both counterparty risk and rollover risk [31][32][33]. In addition, many studies focus on the contagion of both rollover risk and common asset holdings risk [34,35].…”

Optimal Monitoring and Mitigation of Systemic Risk in Financial Networks

“…In this section, we introduce the projected stochastic gradient descent method to solve (26). This is an online algorithm, which allows us to handle one sample at a time, without building a huge linear program.…”