Credit and Systemic Risks in the Financial Services Sector: Evidence From the 2008 Global Crisis

• DOI: http://doi.org/10.1111/jori.12210
• Author: Jean‐François Bégin,Geneviève Gauthier,Mathieu Boudreault,Delia Alexandra Doljanu
• Date: 01 June 2019
• Published date: 01 June 2019

263

©2017 The Journal of Risk and Insurance (2017).

DOI: 10.1111/jori.12210

Credit and Systemic Risks in the Financial Services

Sector: Evidence From the 2008 Global Crisis

Jean-François Bégin

Mathieu Boudreault

Delia Alexandra Doljanu

Geneviève Gauthier

Abstract

Wedevelop a portfolio credit risk model that includes firm-specific Markov-

switching regimes as well as individual stochastic and endogenous recovery

rates. Using weekly credit default swap premiums for 35 financial firms,

we analyze the credit risk of each of these companies and their statistical

linkages, putting emphasis on the 2005–2012 period. Moreover, we study

the systemic risk affecting both the banking and insurance subsectors.

Introduction and Review of the Literature

The financial crisis of 2008 shed light on how the interconnectedness of large financial

institutions can seriously affect their solvency. Yet,even if the financial crisis is behind

us, we still need to comprehend its aftereffects. In this spirit, we assess the evolution

of major determinants of financial crises during the 2005–2012 period, namely, the

roles of leverage, losses, and linkages.

Jean-François Bégin is at the Department of Statistics and Actuarial Science, Simon Fraser Uni-

versity,8888 University Drive, Burnaby, British Columbia, V5A 1S6, Bégin can be contacted via

e-mail: jbegin@sfu.ca. Mathieu Boudreault, director of Quantact and member of GERAD, is at

the Department of Mathematics, UQAM, 201 Président-Kennedy Avenue, Montreal, Quebec,

Canada, H2X 3Y7. Boudreault can be contacted via e-mail: boudreault.mathieu@uqam.ca.Delia

Alexandra Doljanu is at National Bank of Canada, 600 De La GauchetièreOuest Street, Montreal,

Quebec, Canada, H3B 4L2. Doljanu can be contacted via e-mail: deliaalexandra.doljanu@bnc.ca.

Geneviève Gauthier, member of the GERAD, is at the Department of Decision Sciences, HEC

Montréal, 3000 Côte-Sainte-Catherine Road, Montreal,Quebec, Canada, H3T 2A7. Gauthier can

be contacted via e-mail: genevieve.gauthier@hec.ca. Bégin would like to acknowledge the fi-

nancial support of the National Science and Engineering Research Council of Canada (NSERC),

HEC Montréal, the Society of Actuaries, and Montréal Exchange (M-X). Boudreault wishes to

acknowledge the financial support of NSERC. Doljanu wishes to thank the Institut de Finance

Mathématique de Montréal (IFM2) and M-X. Gauthier would like to acknowledge the support

of NSERC and HEC Montréal.

Vol. 86, No. 2, 263–296 (2019).

2The Journal of Risk and Insurance

264

This article offers new insights into the role of credit and systemic risks affecting both

the insurance and the banking subsectors. Among others, we investigate the changes

in correlation through time and the contribution of insurance and banking firms in

the risk of collapse of an entire financial system. To this end, we construct a multi-

variate credit risk model that accounts for firm-specific financial health. The frame-

work embeds oft-cited stylized facts such as leverage volatility (modeled via statisti-

cal regimes), recovery rates negatively related to default probabilities, and pairwise

regime-dependent correlations. We also propose a consistent and reliable method to

estimate the multivariate framework.

Variouscredit risk models have been proposed in the literature. They have been histor-

ically divided into two categories: structural and reduced-form models.1Even though

the reduced-form approach provides a better fit to market data than the structural ap-

proach does, it lacks the economic and financial intuition of the structural framework.

To overcome the limitations of both traditional approaches while retaining the main

strengths of each, hybrid credit risk models have emerged in the literature.2In this ar-

ticle, we adopt a credit risk framework that belongs to this last class of models, linking

the default intensity to the capital structure of the firm through its leverage ratio. More

precisely, to model the leverages, losses, and linkages adequately,a regime-switching

extension of the multivariate hybrid credit risk model of Boudreault, Gauthier, and

Thomassin (2014) is proposed; it allows for firm-specific statistical regimesthat accom-

modate for changes in the leverage volatility and an endogenous stochastic recovery

rate that is negatively related to the default probabilities,and therefore impacts the loss

distribution. Regime-switching dynamics are required to capture the various changes

in behavior through time, and more particularly during crises.

Generally,studies of individual firms’ solvency have mostly focused on balance sheet

information (Allen et al., 2002), credit ratings (Gupton, Finger, and Bhatia, 2007), or

distance to default (Bharath and Shumway, 2008). The financial services sector is no

exception to the rule. Indeed, Harrington (2009) employs, among other things, balance

sheet information to assess the role of AIG and the insurance subsector in the recent

crisis. Milne (2014) uses the distance to default to investigate the solvency of European

banks, concluding that the distance to default measure performs poorly as a market-

based signal for bank risk. In our study,we employ weekly single-name credit default

swap (CDS) premiums of 35 major financial institutions over 2005–2012. The use

of market data is worthwhile; CDS premiums contain forward-looking information

and are frequently updated by market participants as the information becomes avail-

able. Accordingly, they are more appropriate to detect sudden changes in solvency or

1Structural models link the credit events to the firm’s economic fundamentals by assuming that

default occurs when the firm’s value falls below some boundary. Reduced-form models con-

sider the surprise element of the default trigger exogenously given through a default intensity

process.

2For instance, Duffie and Lando (2001), Giesecke and Goldberg (2003), Çetin et al. (2004), and

Giesecke (2006) use incomplete information models in a way that firm assets and the default

barrier are not observable by investors. Another segment of the literature focuses on modeling

the default time as the first jump of a Cox process for which the intensity depends on the firm’s

fundamentals (e.g., Madan and Unal, 2000).

Credit and Systemic Risks in the Financial Sector 3

265

occurrence of crises.3In particular, we find that AIG’s 1-year default probability(PD)

spikes to 42 percent on September 10, 2008, a week before its near default. On average,

the banking subsector’s 1-year PD increases from 0.5 to 4.6 percent during the crisis,

whereas the insurance subsector’s PD increases from 0.4 to 4.2 percent.

Although numerous single-firm approaches exist for measuring credit risk, finan-

cial institutions are intertwined, and therefore, credit risk assessment of the financial

services sector requires an examination of the interconnectedness of its institutions.

There are several ways to look at the interconnectedness of companies: correlation

in the firm’s assets or default intensity through copulas or common factors (e.g., Li,

2000; Frey and McNeil, 2003; Hull, Predescu, White, 2010; Meine et al., 2016), expo-

sure to other common risks such as jumps (e.g., Duffie and Gârleanu, 2001), or other

contagion mechanisms (e.g., Davis and Lo, 2001) such as network approaches (e.g.

Nier et al., 2007; Billio et al., 2012; Markose, Giansante, and Shaghaghi, 2012). This

study models dependence through pairwise regime-dependent correlations of lever-

age comovements. We link the regimes to the firm-specific correlation coefficients

as one of our main goals is to capture the increase in pairwise correlation during

the last financial crisis. Our empirical results show that linkage varies over time. We

find evidence of larger correlations between firm leverage comovements during the

high-volatility regime that suggests the existence of greater interconnectedness dur-

ing the last crisis. Moreover, the regime-dependent linkage structure varies across

subsectors.

Since the financial crisis, many multivariate credit risk frameworks have been used

to investigate systemic risk in the financial sector. Notably, Huang, Zhou, and Zhu

(2009, 2012) construct a systemic risk measure inferred from CDS spreads and equity

price comovements.4Using a network approach, principal component analysis, and

Granger-causality networks, Billio et al. (2012) quantify the interdependence among

four groups of financial institutions during the recent crisis. Their empirical results

suggest that the banking and insurance subsectors are more important sources of

interconnectedness than other financial institutions. Another contribution in that field

is the systemic expected shortfall proposed by Acharya et al. (2010) that measures the

expected loss to each institution conditional on the undercapitalization of the entire

financial system. Other measures of systemic risk applied to financial institutions are

proposed by Adrian and Brunnermeier (2009) and Sald´

ıas (2013).5

3Moreover, these are superior to ratings-based methods because rating revisions tend to lag

behind the market and default probabilities based on the latter depend on aggregated default

counts (i.e., not firm specific).

4Huang, Zhou, and Zhu (2009) propose the use of the so-called “distress insurance premium.”

This theoretical price of insurance against distressed losses is calculated as the risk-neutral

expectation of portfolio credit losses that equal or exceed a minimum share of the sector’s total

liabilities.

5Adrian and Brunnermeier (2009) introduce the concept of CoVaR that measures the value at

risk (VaR) of the financial system conditional on the distress of a specific firm. Sald´

ıas (2013)

develops a forward-looking measure based on the gap between portfolio and average distance

to default series to monitor systemic risk in Europe.