Systemic Risk and the Interconnectedness Between Banks and Insurers: An Econometric Analysis

• Author: Krupa S. Viswanathan,J. David Cummins,Hua Chen,Mary A. Weiss
• DOI: http://doi.org/10.1111/j.1539-6975.2012.01503.x
• Date: 01 September 2014
• Published date: 01 September 2014

© The Journal of Risk and Insurance, 2013, Vol. 81, No. 3, 623–652

DOI: 10.1111/j.1539-6975.2012.01503.x

623

SYSTEMIC RISK AND THE INTERCONNECTEDNESS

BETWEEN BANKS AND INSURERS:ANECONOMETRIC

ANALYSIS

Hua Chen

J. David Cummins

Krupa S. Viswanathan

Mary A. Weiss

ABSTRACT

This article uses daily market value data on credit default swap spreads

and intraday stock prices to measure systemic risk in the insurance sec-

tor. Using the systemic risk measure, we examine the interconnectedness

between banks and insurers with Granger causality tests. Based on linear

and nonlinear causality tests, we find evidence of significant bidirectional

causality between insurers and banks. However, after correcting for condi-

tional heteroskedasticity, the impact of banks on insurers is stronger and of

longer duration than the impact of insurers on banks. Stress tests confirm

that banks create significant systemic risk for insurers but not vice versa.

INTRODUCTION

Systemic risk is often triggered by financial institutions that are too big to fail or

too interconnected to fail. From a statistical perspective, systemic risk involves the

comovement of key financial variables measuring the health or stability of financial

institutions; it has also been described as the potential for multiple simultaneous

defaults of major financial institutions. Traditional measures such as correlation co-

efficients generally are not adequate to measure systemic risk because systemic risk

tends to involve tail behavior, which is not captured by conventional measures. Con-

ventional measures also rely on balance sheet or accounting information that is only

available on a relatively low-frequency basis and often with a lag.

There is a growing literature on systemic risk in banking, as banks have long been

known to be a source of systemic risk. However, little research has been conducted

on measuring systemic risk in the insurance industry.1Inthe past, researchers argued

The authors are from Temple University. The authors can be contacted via e-mail:

hchen@temple.edu, cummins@temple.edu, krupa@temple.edu, and mweiss@temple.edu,

respectively.

1There have been several studies of systemic risk in insurance using aggregate data or a case

study approach. Swiss Re (2003) and the Group of Thirty (2006) conclude that the reinsurance

industry is not a significant source of systemic risk. Harrington (2009) analyzes the role of AIG

624 THE JOURNAL OF RISK AND INSURANCE

that insurers werenot systemically risky because they lacked the “special character” of

banks, primarily the susceptibility to bank runs due to the liquidity of bank deposits

(Swiss Re, 2003). However the near collapse and government bailout of American

International Group (AIG) has challenged this traditional view of insurance.

Two prior papers measure systemic risk in banking and insurance using market

data. Billio et al. (2011) propose several econometric measures of systemic risk to

capture the interconnectedness among the monthly stock returns of hedge funds,

banks, brokers, and insurance companies based on principal components analysis

and linear Granger causality tests. They find that all four sectors have become highly

interrelated over the past decade, increasing the level of systemic risk in the finance

and insurance industries. Acharya et al. (2010) model systemic risk using daily stock

price data. An institution’s contribution to systemic risk, denoted systemic expected

shortfall (SES), is its propensity to be undercapitalized when the system as a whole

is undercapitalized. They provide empirical evidence of the ability of SES to predict

emerging systemic risk during the financial crisis of 2007–2009.

This article adds to the sparse information on the interconnectedness of the banking

and insurance industries. The purpose of the article is to create and implement a robust

systemic risk measure for the insurance sector and investigate the interconnectedness

between the banking and insurance industries during the financial crisis using this

measure. The systemic risk measure is based on Huang et al. (2009) and relies on daily-

frequency market data on credit default swap (CDS) spreads as well as intraday

trading data on stock prices.2This measure is risk neutral, forward looking, and

economically intuitive. Also, the direction of any interconnectedness found between

banking and insurance is investigated using linear and nonlinear Granger causality

tests, providing information on whether the insurance industry is the source or victim

of systemic risk.

By way of preview, we find linear and nonlinear causal effects between systemic

risk measures of banks and insurers in both directions. However, heteroskedasticity

is present in the data. After adjusting for heteroskedasticity and reconducting the

Granger tests, the impact of banks on insurers is found to be stronger and of longer

duration than the impact of insurers on banks. Further, stress testing indicates that

banks create significant systemic risk for insurers but not vice versa. Thus, the results

in the 2007–2009 financial crisis and concludes that systemic risk is relatively low in insurance

markets. Weiss(2010) and Cummins and Weiss (2012) argue that the core activities of insurers

are not a significant source of systemic risk but that banking functions such as derivatives

trading are potential sources of systemic risk. Similar conclusions are supported in Geneva

Association (2010).

2Other approaches have been proposed to measure systemic risk, including the nth–to–default

probability,probability of joint defaults, and credit value-at-risk (e.g., Chan-Lau and Gravelle,

2005; Inui and Kijima, 2005; Yamai and Yoshiba, 2005; Avesani, Pascual, and Li, 2006). The

nth-to-default probability is not satisfactory and sometimes misleading, as it is increasing

in the probability of default but decreasing in the correlation of defaults. The probability

of joint defaults treats all companies as equal and does not take into account differential

impacts of failures of financial institutions of different sizes. In addition, these measures are

typically regarded as physical rather than risk-neutral measures and hence more likely to be

misinterpreted by users.