Systemic risk and centrality: The role of interactions

• Published date: 01 November 2022
• Author: Hossein Asgharian,Dominika Krygier,Anders Vilhelmsson
• Date: 01 November 2022
• DOI: http://doi.org/10.1111/eufm.12340

DOI: 10.1111/eufm.12340

ORIGINAL ARTICLE

Systemic risk and centrality: The role of

interactions

Hossein Asgharian

1

|Dominika Krygier

2

|Anders Vilhelmsson

1

1

Department of Economics, Lund

University, Lund, Sweden

2

Financial Stability, Sveriges Riksbank,

Stockholm, Sweden

Correspondence

Hossein Asgharian, Department of

Economics, Lund University, Box 7082,

22007 Lund, Sweden.

Email: Hossein.asgharian@nek.lu.se

Abstract

We analyze to what extent the contribution of banks

to systemic risk depends on their centrality in fi-

nancial networks. We find that centrality is an im-

portant determinant of systemic risk, but not

primarily by its direct effect. Its main influence is to

make other risk measures, such as probability of

default, more important for highly connected banks.

Neglecting the indirect effect of centrality may se-

verely underestimate or overestimate the systemic

risk of banks. We also show that, even though size

and centrality are related, the inclusion of centrality

provides valuable information when assessing the

systemic importance of banks.

KEYWORDS

CoVaR, loan syndication, network centrality, systemic risk

JEL CLASSIFICATION

G21, G18

Eur Financ Manag. 2022;28:1199–1226. wileyonlinelibrary.com/journal/eufm

|

1199

EUROPEAN

FINANCIAL MANAGEMENT

This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits

use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or

adaptations are made.

© 2021 The Authors. European Financial Management published by John Wiley & Sons Ltd.

The authors would like to thank an anonymous referee, Jens Forssbaeck, Erik Hjalmarsson, Martin Strieborný as well

as participants at the CFF‐KWC 2018 conference, the EEA ESEM 2019 conference and the IRMC 2019 conference.

1|INTRODUCTION

Systemic risk is the risk of a crisis in the financial sector with consequential negative spillover

effects on the real economy. To understand and manage systemic risk, it is important to

understand both macro and micro determinants of systemic risk. Macro determinants focus on

the overall structure of the financial system, whereas the micro approach focuses on the

marginal contributions of individual actors to systemic risk. Our paper is primarily focused on

the micro level, but it includes the macro level by studying the banking network. Our basic idea

is simple. We suggest that a bank's centrality should not be considered a separate cause of

systemic risk. Rather, we suggest that centrality affects how much a bank's ‘riskiness’con-

tributes to systemic risk. Statistically, this means we should treat centrality as a moderator

variable. We, therefore, investigate how the contribution to systemic risk of standard bank‐level

risk measures varies depending on the bank's centrality. We suggest that risky banks contribute

extensively to systemic risk only if they are centrally placed in the financial network.

Somewhat surprisingly, the impact on systemic risk from the interaction effect of centrality

and bank characteristics has not been investigated before, though both the impact of firm

characteristics (Adrian & Brunnermeier, 2016; Brunnermeier et al., 2020; Saunders et al., 2019)

and that of network centrality (Cai et al., 2018; Martinez‐Jaramillo et al., 2014) are studied

without allowing for interactions. The treatment of centrality and other bank risks as separate

and independent sources of systemic risk is also reflected in current systemic‐risk regulation

(Basel Committee on Banking Supervision [BCBS], 2018).

To calculate systemic risk, we use Adrian and Brunnermeier's (2016)ΔCoVaR,andweobtain

centrality using the network of banks participating in the loan syndication market. The paper most

closely related to ours is that of Cai et al. (2018), who also calculate an interconnectedness measure

from syndicated loans and use it to explain different measures of systemic risk, including ΔCoVaR.

Our paper differs from Cai et al. (2018) in several aspects: they measure the commonality of asset

holdings, whereas our paper considers network connections between banks. Hence, our paper

complements Cai et al. (2018) by focusing on the centrality of a bank rather than on balance sheet

overlap. Further and most importantly, Cai et al. (2018) do not interact firm characteristics with their

commonality measure, treating it instead as a separate source of risk.

The data on loan syndication is obtained from the Thomson Reuters DealScan database,

which provides historical information on the terms and conditions of deals in the global

commercial and industrial loan market. It has been used by, for example, Ivashina et al. (2015)

and Sufi (2007). We consider two banks to be linked when they participate in the same loan

syndicate and calculate six different centrality measures based on the loan syndication data.

UsingpaneldataregressionstoexplainΔCoVaR, our main finding is that centrality is an im-

portant determinant of systemic‐risk contribution, for all but the smallest banks, but primarily not by

its direct effect. Rather, its main influence is to make risk measures such as the probability of default

(PD) much more important for highly connected firms. A bank's contribution to systemic risk from

PD is about three times higher for a bank with two standard deviations above‐average centrality

compared to a bank with average network centrality. Current systemic‐risk regulation takes centrality

into account as one of five categories used for calculating systemic importance, but it does so as a

standalone component. By giving each of the five categories that contribute to systemic risk equal

weight, current regulation cannot capture that the importance of firm characteristics varies with

centrality while underestimating the importance of centrality for risky banks and overestimating the

effect for ‘safer’banks. Our results are robust to different specifications of centrality. Centrality

remains important after simultaneously allowing also size to act as a moderator variable. This

1200

|

EUROPEAN

FINANCIAL MANAGEMENT

ASGHARIAN ET AL.