Diversification benefits of cat bonds: An in‐depth examination

• DOI: http://doi.org/10.1111/fmii.12134
• Date: 01 December 2020
• Author: Van Son Lai,Karl Demers‐Bélanger
• Published date: 01 December 2020

DOI: 10.1111/fmii.12134

ORIGINAL ARTICLE

Diversification benefits of cat bonds:

An in-depth examination

Karl Demers-Bélanger1Van Son L ai 2,3

1Analyst-Quantitative Strategies,iA Financial

Group (Industrial Alliance), Quebec, Canada

2Professor of Finance at Laval University,

Quebec, Canada

3A Senior Research Fellowat IPAG Business

School, Paris,France

Correspondence

VanSon Lai, Laval University,2325, rue de

laTerrasse,Quebec (Québec), G1V 0A6,

CANADA.

Email:VanSon.Lai@fsa.ulaval.ca

Abstract

We investigate whether the inclusion of Cat Bonds in port-

folios composed of traditional assets and common factors

is beneficial to investors. Various mean-variance spanning

tests performed for the period of 2002 to 2017 show that

under different market conditions, theaddition of Cat Bonds

gives rise to previously unattainable portfolios. Using the

Engle (2002) Dynamic Conditional Correlation (DCC) model,

we find that including Cat bonds increases significantly the

time-varying Sharpe ratio and the Choueifaty and Coignard

(2008) maximum diversification ratio. Cat Bonds provide

needed diversification during critical times particularly dur-

ing episodes of crisis and of high volatility.Under the second-

order stochastic dominance efficiency (SDE) tests, the null

hypothesis that portfolios without Cat Bonds are efficient

cannot be rejected. Out-of-sample analyses indicate that the

performance of portfolios with Cat Bonds included varies

depending on the performance measures employed, the

portfolio construction techniques used and the assets or fac-

tors considered.

KEYWORDS

asset allocation, catastrophe bonds, diversification, dynamic corre-

lation, factor investing, mean-variance spanning, portfolio optimiza-

tion,regime switching, stochastic dominance efficiency, time varying

© 2020 New YorkUniversity Salomon Center and Wiley Periodicals LLC

Financial Markets,Inst. & Inst. 2020;29:165–228. wileyonlinelibrary.com/journal/fmii 165

166 DEMERS-BÉLANGER ANDLAI

1INTRODUCTION

With average annual temperaturesrising and frequencies of natural disasters increasing in all parts of the world, dire

environmental impacts, vital socio-political and important economic problems associated to global warming are grow-

ingly alarming. In its 2007 special report (Bernstein et al., 2008), the Intergovernmental Panel on Climate Change

(IPCC) describes the effects of global warming that are already being felt and the IPCC expects these effects to grow

in magnitude and costs in the future.1Catastrophic risks associated with climate change are of great importance for

financial markets and institutions, for all quarters and constituencies as well as governmentswhose stakes are greatly

affected by natural disasters.

Increases in the frequency and severity of naturalcatastrophes in recent years have also propelled the use of alter-

nativerisk-transfer (ART) instruments for managing catastrophic risks. Tocover against disaster risk and transfer some

of the risks they do not want to retain, insurance and reinsurancecompanies typically use reinsurance or retrocession

(reinsurance for reinsurers). However,for catastrophic natural disasters, the margins charged by reinsurers are often

prohibitive. Therefore, insurance-linkedsecurities (ILS) have been launched to cover these types of losses at relatively

lowercosts. Among these ART instruments, the most important asset class is catastrophe bonds (Cat Bonds hereafter),

see Barrieu and Albertini (2010), Bouriaux and MacMinn (2009), Cummins (2008), Cummins (2012), Lai, Parcollet, and

Lamond (2014), Smack (2016), Trottier,Lai, and Charest (2018). The size of the outstanding catastrophe bond and ILS

marketsat the end of 2019 is $41 billion.2Naturally, Cat Bonds returns incorporate risk premiums rewarding investors

who are willing to assume a hardly diversifiable and hedgeable risk, see for instance Godin, Lai, and Trottier(2019). In

fact, among ILS, Cat Bonds are the only financial instrument securitized and traded in secondary markets. As a com-

plement to reinsurance, Cat Bonds have been used to transfer the risk attached to the highest layersof reinsurance.

While both reinsurance and Cat Bonds offer companies a means to transfer disaster risk, only Cat Bonds use the cap-

ital markets for this purpose, see e.g., Canter, Cole, and Sandor (1996), Kish (2016), Krutov (2010), Trottier and Lai

(2017).

There are two main benefits from using Cat Bonds. First, their presumed zero or very low correlation with other

financial assets allows for additional diversification to a portfolio, see e.g., Hoyt and McCullough (1999), Kish (2016),

Litzenberger, Beaglehole, and Reynolds (1996), Sterge and van der Stichele (2016), and second, their historical risk-

adjusted returns are attractive to investors, see Kusche (2013), Schöchlin (2002) among others. Cat Bonds are con-

sidered as a high-yielding fixed income asset class with returns independent from macroeconomic risks and cycles. In

addition, Cat Bonds show a high level of intra-class diversificationprovided by their correlation to different and inde-

pendent risk factors arising in different parts of the planet (i.e., tsunami, hurricanes, floods, etc.). Hence, Cat Bonds,

exhibiting a low correlation to other traditionalasset classes (e.g., equities, bonds and commodities) offer institutional

investors a great portfolio diversificationwith an appealing risk-return profile. Institutional investors (mainly pension

funds and hedge funds) looking for a steady,relatively high-yield and exotic asset class in the current low-interest rate

environment are said to be lining up capital to support the increased appetite in the ILS marketfor Cat Bonds (see e.g.,

Carhart, Cheah, De Santis, Farrell, and Litterman (2014), Sterge and van der Stichele (2016), and Kish (2016)).

This paper examines in an in-depth fashion whether the addition of Cat Bonds to an investor portfolio does effec-

tively provide him the benefits of diversification. Weconduct an analysis of the time-varying performance associated

with the inclusion of Cat Bonds under different financial market conditions overthe period of 2002 to 2017. We con-

sider two different benchmark universes. The first is formed from traditional asset classes and the second one from

common factors. Well-known factors which have been considered as the underlying drivers of risk and return across

assets and asset classes, can be macro oriented (e.g., economic growth, inflation) or style oriented (e.g.,value, momen-

tum, quality). Further, factors are not directly investable,but factor exposures are an offshoot of investing in assets.

With relatively more stable returns over time than those obtained from asset classes, depending on the market envi-

ronment some factors perform better than others, see Ang (2014), Cerniglia and Fabozzi (2018), Dimson, Marsh, and

Staunton (2017), Naik, Devarajan, Nowobilski, Page,and Pedersen (2016).

DEMERS-BÉLANGER ANDLAI 167

The asset classes are U.S. equities, global equities ex-U.S., emerging markets equities, real-estate, U.S. treasury

bonds, U.S. corporate bonds, U.S.high-yield bonds and commodities. The factors consist of equity market, value, size,

momentum, volatility,mortgage, default, term, high-yield and commodity curves. To represent Cat Bonds, we use five

different Swiss Reinsurance indexes.3

Standing out from previous works, we conduct our study using four different approaches. First, we perform a bat-

tery of mean-variance spanning tests based on the methodology developed in Kan and Zhou (2012). Furthermore, to

split the sample into two economic regimes, we use three approaches 1- the NBER business cycle chronology,2- the

turbulence index à la Kritzman and Li (2010), and 3- a Markov-Switchingmodel applied to the US stock market index

andthe US bond market index à la Hardy (2001). We conduct our tests using the full sample as well as the regime-based

periods. We find that the addition of Cat Bonds leads to portfolios previouslyunattainable regardless of the regimes.

Second, by means of the Engle (2002) dynamic conditional correlation model (DCC), we study the time-varying

effects of including Cat Bonds in constructing portfolios. We start by estimating the correlations between Cat Bonds

and other assets. We then use these correlations to obtain the maximum Sharpe ratio and maximum diversification

à la Choueifaty and Coignard (2008) portfolios with and without Cat Bonds. Portfolios with Cat Bonds yield higher

Sharpe ratios and diversification ratios than those without these. Third, to determine whether a portfolio of tradi-

tional assets stochastically dominates a portfolio created from the same universe with Cat Bonds added, we employ

a two-step method as in Daskalaki, Skiadopoulos, and Topaloglou(2017) to perform stochastic dominance efficiency

tests à la Scaillet and Topaloglou(2010).4We run the tests on the full sample under the different market conditions

mentioned above. Based on the second-order stochastic dominance criterion, we find that the null hypothesis that

portfolios that do not include Cat Bonds are efficient cannot be rejected. This means that adding Cat Bonds might not

achieve diversificationwhen we consider higher moments of the return distribution.

Finally,by means of a rolling window method, we analyze the out-of-sample performance of the portfolios of max-

imum Sharpe ratio, those of maximum diversification and those constructed from the second-order dominance crite-

rion. Totake into account the non-normal distribution of returns and the portfolios’ turnovers, we use the conditional

Sharpe ratio (CSR) à la Maillard (2018) and the Omega ratio à la Keating and Shadwick (2002) as metrics of perfor-

mance.The results differ depending on the performance measure used, the portfolio construction technique employed

to obtain the optimal portfolio and the assets or factors universe considered. In this context, we conclude that adding

Cat Bonds is not always desirablefor investors.

We make a number of contributions to the current Cat Bond literature. To consider the whole distribution of Cat

Bond returns, we use the second-order stochastic dominance efficiency (SSDE) criterion. In addition to studying the

dynamic aspect of the correlation, we also quantify the variation over time of the effects of including Cat Bonds in a

portfolio. Furthermore, we also consider a universe of factors instead of simply asset classes. Finally,we use out-of-

sample performance measures which allow for deviations from normality. To the best of our knowledge, we are the

first, not only to use both mean-variance spanning and stochastic dominance approaches, but also to takeinto account

different market conditions to study the diversification effects of Cat Bonds in a frameworkof portfolios constituted

of both asset classes and factors.

The rest of the paper is organized as follows. We provide a review of the literature in Section 2.InSection3,we

detail our methodological approaches and in Section 4, we describe our data. After revealing the results in Section 5,

we conclude in Section 6.

2LITERATURE REVIEW

Catastrophe bonds work liketraditional ones, i.e., a Cat Bond pays coupons at a regular interval and a notional amount

at maturity. A major difference with the usual bond case is that these payments are made contingent to the nonoc-

currence of a natural disaster.The definition of disaster is specified in the contract and can take many forms, see for

instance, Barrieu and Albertini (2010), Cummins (2008), Cummins (2012), Smack (2016). For a recent and concise