Research Proposal submitted to the 2008 FMA annual meeting

Modeling and Pricing Longevity Risk:

Permanent Jump Effects and Extreme Value Approach

Hua Chen and J. David Cummins

Hua Chen (Corresponding author)

Department of Risk Management and Insurance
Georgia State University
P.O. Box 4036
Atlanta, GA 30302-4036
Phone: (405) 370-5899
Fax: (404) 413-7499
Email: hchen10@student.gsu.edu

J. David Cummins
Department of Risk, Insurance, and Health Management
Temple University
410 Ritter Annex
Philadelphia, PA 19122
Phone: (215) 204-8468
Fax: (215) 204-4712
E-mail: cummins@temple.edu

1
1. Introduction and Motivation
The purpose of this study is to analyze securitization of longevity risk with emphases on the
longevity risk modeling and longevity payoff pricing. Although various longevity derivatives
have been proposed and revised so far, pitfalls still exist. A typical example would be the
unsuccessful attempt of the EIB longevity bond in 2004. After analyzing the pros and cons of
previous securitizations, we present our hypothetical longevity bond with the payoffs as a put
option spread. We integrate the parametric model with permanent jump effects and the extreme
value theory for mortality fitting and forecasting. Our method is fresh new in longevity risk
securitization and has the advantages of both capturing the mortality improvements within
sample and extrapolating the rare longevity events out of sample. Next, we value the payoffs of
the longevity bond via the Wang transform in an incomplete market, and illustrate the effects of
extreme value approach on the accuracy of security valuation.

1.1. Longevity Risk and the Capital Market

Longevity risk is defined as the uncertainty of mortality improvement in the future. It exists in
both individual and aggregate levels (MacMinn, Brockett, and Blake, 2006). Individuals may
face the longevity risk of outliving their resources. However, they can insure against the risk
through the social security systems, defined benefits plans, and private annuity products. Pension
plans and insurers can diversify the longevity risk at the individual level following the law of
large numbers. In the aggregate sense, longevity risk is referred to as the fact that people of some
certain population might live longer, on average, than expected. For instance, Hardy (2005)
points out that the life expectancy for men aged 60 is 5 years’ longer in 2005 than it was
anticipated in the mortality projection made in the 1980s. Such risk breaks down the risk pooling
mechanism and becomes non-diversifiable, making the provision of risk management tools
increasingly difficult.

Due to the retirement of baby-boomers and the reduction of death rate among the elderly, both
social security system and pension plans are facing an “aging-population tsunami” and
significant future imbalances in consequence. The shift of pension plans from defined benefit
(DB) to defined contribution (DC) raises the demand for private annuity products even further.
As demand for individual annuity increases, insurers will have to manage the potential longevity
risk in issuing new annuity policies. The Alternative Risk Transfer (ART) can provide theoretical
fundamentals for a capital market solution. The insurers and reinsurers are stepping away from
the traditional risk warehousing function towards risk intermediation function, which enables
them to operate more efficiently as well as increase underwriting capacity (Froot, 2001; Cowley
and Cummins, 2005; Cox and Lin, 2007). In the past few decades, insurance companies have
successfully transferred the catastrophic risk in the property-liability business to financial
markets by issuing CAT bonds (Cummins, Lalone, and Phillips 2004). More recently, various
mortality bonds (or longevity bonds) have been designed and/or issued to transfer mortality risk
(or longevity risk) into the capital market.

1.2. The Design of Longevity Bond

The idea of longevity bonds was first proposed by Blake and Burrows (2001). The bonds are
designed so that coupon payments are contingent on the percentage of a certain population
cohort surviving some further period. The longevity bonds allow annuity providers to hedge
against aggregate longevity risk, that is, when annuitants live longer than anticipated, the

2
insurance companies would incur losses due to longer payment periods, but they would also
receive greater coupon payments for holding longevity bonds to offset the losses (Denuit,
Devolder, and Goderniaux, 2007). A more extensive discussion of mortality-linked derivatives is
provided by Blake, Cairns and Dowd (2006), who discuss the various forms of longevity bonds,
swaps, futures, and options, and investigate their potential uses.

Ever since the Swiss Re successfully issued the first pure mortality bond in December 2003,
insurers or reinsurers have made another four mortality bond transactions in order to reduce their
extreme mortality exposures, with a total volume exceeding $1.2 billion (Bauer and Kramer,
2007). Possibly inspired by the successful securitization of mortality risks, the European
Investment Bank (EIB) offered the first longevity bond in November 2004 to hedge the longevity
risk for pension planners and annuity providers. Although this particular deal failed to be
launched because of insufficient demands, it did attract public attention and provided an
instructive case study (Blake, Cairns, and Dowd, 2006). Lin and Cox (2006) propose a mortality
securitization model in an incomplete market framework, and calculate the market prices of
mortality risk for the Swiss Re mortality bond and the EIB longevity bond, respectively. They
find that the Swiss Re deal offers a higher risk premium to investors than the property-linked
catastrophe bonds, while the EIB bond charges a very high risk premium to take longevity risks
in the UK pension plans, which explains the opposite market outcomes of these two securities.

However, in order to understand why the EIB bond did not sell, we need to explore further
reasons underneath the intrinsic bond design. First, the principal of the EIB bond is provided by
hedgers (annuity providers), which impose a substantial capital requirement for the hedgers. In
contrast, in the case of the Swiss Re deal, the hedgers (insurer or reinsurers) pay only an annual
premium to hedge the extreme mortality risk. Second, the longevity risk embedded in the EIB
bond is transferred to a reinsurer, Partner Re. However, the mortality risk in the Swiss Re is
passed onto the whole financial market, thus bringing more capital and creating new risk-bearing
capacity. Third, the hedgers of the EIB bonds are exposed to the direct credit risk of EIB and
indirect credit risk of BNP Paribas and Partner Re, while the Swiss Re deal is fully collateralized.
Additionally, the EIB bond provides “ground up” protection, covering the entire annuity
payment (Lin and Cox, 2007). The Swiss Re mortality bond is more attractive because it only
covers payments in excess of some strike level.

Therefore, the simplest way to securitize longevity risk would be to make it parallel to the design
of the Swiss Re deal. Payoffs of the longevity bond to the hedgers would be structured as a put
option spread with an original upper strike and a lower strike. Once the survival index falls
below the upper strike, parts of the principal should be withdrawn to help the hedgers make
annuity payments, and the upper strike would be reset to the index value. The longevity bond
terminates at the earlier time of a predetermined maturity date (20 years, for example) and the
time when the revised upper strike reaches the lower strike.

1.3. Mortality Modeling: Permanent Jump Effects and the Extreme Value Theory
Mortality models are critical in quantifying mortality/longevity risks and providing the basis of
pricing and reserving. An ideal mortality model should be able to clearly convey the nature of
risk transfer to investors, reflect the characteristics of available data, and provide scenario
analyses. People have exerted a lot of efforts in exploring such an ideal model. Cairns, Blake and

3
Dowd (2006a) provide a detailed overview and categorization of stochastic mortality models,
most of which fall in the framework of short-rate models. Lin and Cox (2006) argue that
mortality jumps must be taken into account, because the rationale behind mortality securitization
is to hedge extreme mortality risks. Biffis (2005) uses affine jump-diffusions to address the risk
analysis and market valuation of life insurance contracts. Cox, Lin and Wang (2006) propose a
geometric Brownian motion with jumps for mortality modeling, and price the Swiss Re mortality
bond. Chen and Cox (2007) build a jump-diffusion process in the Lee-Carter framework and
explore the models with permanent versus transitory jump effects. They find that the model with
transitory jump effects is more appropriate for mortality risk securitization, because most of
mortality jumps are caused by catastrophic events and only have short-term effects.

Likewise, rare longevity events need to be modeled explicitly in the securitization of longevity
risk. We believe that the model with permanent jump effects is applicable in the longevity
market, since the main concern of investors of longevity bonds is the aggregate mortality
improvement risk. For example, heart disease is one of the leading causes of death in the United
States. It accounts for 28.5 percent of the total death in 2002 according to the report of National
Center of Health Statistics (Vol. 53, No. 5, October 2004). If there were a breakthrough of
medicine protecting people from heart attack, the mortality rates would decrease substantially
and the effects of mortality improvement would last for a long period, even forever.

Another challenge in longevity modeling is the imprecise knowledge on rare longevity risk. The
mortality improvement evolutes slow, but continuously. This procedure is influenced by various
factors in socioeconomic, biological, environmental, and behavioral aspects. It is hard for us to
account for all factors in a comprehensive model framework. Sometimes the change of one
factor can affect the mortality improvement dramatically. The problems associated with available
data also make the modeling difficult. For example, big sampling errors and highly volatile crude
morality rates occur at higher ages, due to the diminishing number of deaths and exposure-to-
death (Li, Boyle, Hardy, and Tan, 2007). The inaccuracy and unavailability of mortality data at
the very old ages has also been discussed (Buettner, 2002). Furthermore, we have very few
extreme mortality improvement cases, which constrained us from learning from past experience.

Recent development in extreme value theory (EVT) offers an attractive solution to this problem.
For detailed discussions of extreme value theory and its application to insurance and finance, we
refer interested readers to the book by Reiss and Thomas (2001) (see also Kluppelberg and
Mikosch, 1997; McNeil, 1997; McNeil and Frey, 2000; Kotz and Nadarajah, 2000; Sanders,
2003; Li, Boyle, Hardy, and Tan, 2007). Beelders and Colarossi (2004) apply EVT to the Swiss
Re mortality bond issued in 2003 and review some of the benefits of this approach. In this study,
we extend the traditional parametric mortality modeling and nicely integrate it with EVT. The
Pickands-Balkema-De Haan Theorem in EVT tells us, under certain conditions, the limiting
distribution of the exceedances is the generalized Pareto distribution, as the threshold approaches
to the right endpoints. Therefore, we can use the model with permanent jump effects for
mortality rates before the threshold, and assume mortality exceedances over the threshold follow
a generalized Pareto distribution. By this means, we can not only model the mortality data within
our samples, but also rationally extrapolate more extreme, out of sample longevity events. The
threshold can be obtained through an iterative method of maximum likelihood estimation.

4
1.4. Valuation the Longevity Payoffs via the Wang Transform
Longevity securities involve significant valuation problems. In a complete market, we can
replicate the payoffs of financial securities dynamically by creating a portfolio which consists of
the underlying asset and the money market account. The necessary condition requires the
underlying asset be tradable in the market. However, this is not the case with longevity securities.
Since the underlying survivor index is not an existing tradable asset, it introduces the market
incompleteness. In an incomplete insurance market, there are mainly two approaches for security
valuation. One is to adapt the arbitrage-free pricing framework of interest-rate derivatives to the
valuation and securitization of mortality risk. Cairns, Blake and Dowd have a detailed discussion
on this issue and give as an example the pricing of the EIB longevity bond (see Cairn, Blake and
Dowd 2006a, 2006b). The second method is to use a distortion operator to create an equivalent
risk-adjusted distribution, and obtain the fair value of the security under this risk-adjusted
measure. Examples of this approach, based on the Wang transform (Wang 2000, 2002), include
Lin and Cox (2006), Dowd, Blake, Cairns and Dawson (2006), Blake, Cairns, Dowd, and
MacMinn (2006), Denuit, Devolder, and Goderniaux (2007). We prefer the Wang transform
because it is simple to be implemented in practice. In addition, the Wang transform has an
important feature: it preserves the normal and lognormal distributions. Therefore it can be used
to replicate the CAPM if the return for an underlying asset has a normal distribution and recover
the Black-Scholes formula if the return for the underlying is lognormally distributed.

1.5. Agenda
The rest of this paper proceeds as follows. In Section 2, we will provide an overview of the
longevity risk. In Section 3, we will point out the design problems of the EIB bond by comparing
it with the Swiss Re mortality bond, and propose our hypothetical longevity bond parallel to the
Swiss Re deal. In section 4, we are going to model the mortality dynamics as the combination of
EVT and the parametric model with permanent jump effects. In section 5, we will value the
longevity payoff as a put option spread via the Wang transform in an incomplete market
framework. Lastly, conclusion and discussion come in section 6.

References:
Bauer, D., and F. W. Kramer. 2007. Risk and Valuation of Mortality Contingent Catastrophe
Bonds. Working Paper, Ulm University.
Beelders, O., and D. Colarossi. 2004. Modelling mortality risk with extreme value theory: The
case of Swiss Re's mortality-indexed bond. Global Association of Risk Professionals, 4
(July/August), 26-30
Biffis, E.. 2005. Affine Processes for Dynamic Mortality and Actuarial Valuations. Insurance:
Mathematics and Economics. 37(3):443–468.
Blake, D., and W. Burrows. 2001. Survivor Bonds: Helping to Hedge Mortality Risk. Journal of
Risk and Insurance. 68(2):339–348.
Blake, D., A. J. G. Cairns, and K. Dowd. 2006. Living with Mortality: Longevity Bonds and
Other Mortality-linked Securities. British Actuarial Journal. 12: 153-197
Blake, D., A. J. G.. Cairns, K. Dowd and R. MacMinn. 2006. Longevity Bonds: Financial
Engineering, Valuation and Hedging. Journal of Risk and Insurance, 73: 647-672.

5
Buettner, T.. 2002. Approaches and Experiences in Projecting Mortality Patterns for the Oldest
Old. In Living to 100 and Beyond: Survival at Advanced Ages Symposium. Society of
Actuaries.
Cairns, A. J. G., D. Blake and K. Dowd. 2006a. Pricing Death: Frameworks for the Valuation
and Securitization of Mortality Risk. ASTIN Bulletin. 36(1): 79-120.
Cairns, A. J. G., D. Blake and K. Dowd. 2006b. A Two-Factor Model for Stochastic Mortality
with Parameter Uncertainty: Theory and Calibration. The Journal of Risk and Insurance.
73(4): 687-718
Chen, H., and S. H. Cox. 2006. Modeling Mortality with Jumps: Transitory Effects and Pricing
Implications to Mortality Securitization. Working paper, Georgia State University.
Cowley, A., and J. D. Cummins. 2005. Securitization of Life Insurance Assets. Journal of Risk
and Insurance, 72(2):193–226.
Cox, S. H., and Y. Lin. 2007. Natural Hedging of Life and Annuity Mortality Risks. North
American Actuarial Journal, 11(3):1–15.
Cox, S. H., Y. Lin, and S. S. Wang. 2006. Multivariate Exponential Tilting and Pricing
Implications for Mortality Securitization. The Journal of Risk and Insurance. 73(4): 719-
736
Cummins, J. D.. 2006. The Securitization of Life Insurance and Longevity Risks. The 2nd
International Longevity Risk and Capital Market Solutions Symposium, Chicago.
Cummins, J. D., D. Lalonde, and R. Phillips. 2004. The Basis Risk of Catastrophic-Loss Index
Securities. Journal of Financial Economics. 71(1): 77-111.
Denuit, M., P. Devolder, and A. Goderniaux. 2007. Securitization of Longevity Risk: Pricing
Survivor Bonds with Wang Transform in the Lee-Carter Framework. The Journal of Risk
and Insurance. 74(1): 87–113.
Dowd, K., D. Blake, A. J. G. Cairns, and P. Dawson. 2006. Survivor Swaps. The Journal of Risk
and Insurance. 73(1): 1-17.
Froot, K. A.. 2001. The Market for Catastrophe Risk: A Clinical Examination. Journal of
Financial Economics, 60(2-3):529–571.
Hardy, M. 2005. A Matter of Life and Death. Financial Engineering News. Available at
http://www.fenews.com/fen46/topics act analysis/topics-act-analysis.htm.
Kluppelberg C, T. Mikosch. 1997. Large deviation of heavy-tailed random sums with
applications in insurance and finance. Journal of Applied Probability. 34: 293—308
Kotz. S., and S. Nadarajah. Extreme Value Distributions Theory and Applications. Imperial
College Press.
Li, S. H., P. P. Boyle, M. R. Hardy, and K. S. Tan. 2007. On Pricing and Hedging the No-
Negative-Equity-Guarantee in Equity Release Mechanisms. Working paper, the University
of Waterloo.
Lin, Y., and S. H. Cox. 2006. Securitization of Catastrophe Mortality Risks. Available at SSRN:
http://ssrn.com/abstract=994575
Lin, Y., and S. H. Cox. 2007. Longevity Risk, Rare Event Premia and Securitization. Available
at SSRN: http://ssrn.com/abstract=1070421
MacMinn, R., P. Brockett, and D. Blake. 2006. Longevity Risk and Capital Markets. Journal of
Risk and Insurance. 73 (4): 551-557.
McNeil, A.. 1997. Estimating the Tails of Loss Severity Distributions Using Extreme Value
Theory. ASTIN Bulletin. 27: 117–137.

6
McNeil, A. J., R. Frey. 2000. Estimation of Tail-Related Risk Measures for Heteroscedastic
Financial Time Series: An Extreme Value Approach. Journal of Empirical Finance. 7: 271–
300
National Center of Health Statistics. October 2004. Vol. 53, No. 5.
Reiss, R. D., and M. Thomas. 2001. Statistical Analysis of Extreme Values. 2nd Edition. Basel;
Boston; Berlin: Birkhauser. 2001.
Sanders, D.E.A. 2003. Extreme Events Part 2. Financial Catastrophes. The Overthrow of Modern
Financial Theory. GIRO 2003.
Wang, S, S. 2000. A Class of Distortion Operators for Pricing Financial and Insurance Risks.
The Journal of Risk and Insurance. 67(1):15-36.
Wang. S. S. 2002. A Universal Framework for Pricing Financial and Insurance Risks. ASTIN
Bulletin. 32(2): 213-234.