Hansj¨org Furrer Market-consistent Actuarial Valuation ETH...
Transcript of Hansj¨org Furrer Market-consistent Actuarial Valuation ETH...
Guaranteed Annuity Options
Hansjorg Furrer
Market-consistent Actuarial Valuation
ETH Zurich, Fruhjahrssemester 2008
Guaranteed Annuity Options
Contents
A. Guaranteed Annuity Options
B. Valuation and Risk Measurement
Guaranteed Annuity Options 1
A. Guaranteed Annuity Options
Course material
• Slides
• Boyle, P. and Hardy, M. (2003). Guaranteed Annuity Options. ASTIN Bulletin,Vol. 33, No. 2, pp. 125-152.
• Pelsser, A. (2003). Pricing and Hedging Guaranteed Annuity Options viaStatic Option Replication. Insurance: Mathematics and Economics. Vol. 33,283-296.
• The above documents can be downloaded from
www.math.ethz.ch/~hjfurrer/teaching/
Guaranteed Annuity Options 2
An Introduction into GAO
• Definition:
Under a guaranteed annuity option, the insurer guarantees to convert thepolicyholder’s accumulated funds into a life annuity at a fixed rate g at thepolicy maturity date T
• GAOs provide a minimum return guarantee:
- the policyholder has the right to convert the accumulated funds into a lifeannuity at the better of the market rate prevailing at maturity and theguaranteed rate
- if the annuity rates under the guarantee exceed the market annuity rates, thena rational policyholder will exercise the option. In that case, the insurer mustcover the difference
Guaranteed Annuity Options 3
Origins
• When many of these guarantees were written in the UK in the 1970ies and1980ies
- long-term interest rates were high
- mortality tables did not include an explicit allowance for future mortalityimprovements (longevity)
• Ever since, however, long-term interest rates declined and mortality improvedsignificantly for lives on which these policies were sold. . .
• To summarize, the guarantee corresponds to a put option on interest rates:
- when interest rates rise, the annuity amount per 1000 fund value increases- when interest rates fall, the annuity amount per 1000 fund value decreases
Guaranteed Annuity Options 4
Effect of Mortality Improvement
• The price of the guarantee depends on the mortality:
- 13-year annuity certain corresponds to an interest rate of 5.7% since
1000 =111
1 + 0.057+
111(1 + 0.057)2
+ · · ·+ 111(1 + 0.057)13
- 16-year annuity-certain requires an interest rate of 7.72%:
1000 =111
1 + 0.0772+
111(1 + 0.0772)2
+ · · ·+ 111(1 + 0.0772)16
I when mortality improves, the interest rate at which the guarantee becomeseffective increases
Guaranteed Annuity Options 5
Setting the Scene
• S = {S(t) : 0 ≤ t ≤ T} market value of the accumulated funds
• B = {B(t) : 0 ≤ t ≤ T} with B(t) = exp{∫ t
0ru du} money market account and
{rt : 0 ≤ t ≤ T} instantaneous short rate process
• P (t, T ) time-t price of a zero-coupon bond with maturity T , t ≤ T . Hence,P (t, T ) = EQ[B(t)/B(T )|Ft]
• ax(T ) : market value of an immediate annuity of 1 p.a. (payable in arrear) to alife aged x at T :
ax(T ) =∞∑
k=1
kpx P (T, T + k), (1)
where kpx denotes the conditional probability that a person having attainedage x will survive k years (k-year survival probability)
Guaranteed Annuity Options 6
The Nature of GAO
• g : Conversion rate. Determines the guaranteed annuity payment per annum, e.g.if S(T ) = 1000, then g = 9 implies an annuity payment of 1000/9 (≈ 111) p.a.
• Y (T ) : payoff from exercising the option at maturity T (= value of the guaranteeat maturity T ):
Y (T ) =
0, if S(T )
g ≤ S(T )a65(T )(
S(T )g
)a65(T )− S(T ), if S(T )
g > S(T )a65(T ) .
• Conditional on the survival of the policyholder up to time T , one thus has:
Y (T ) = S(T )(
a65(T )g
− 1)+
. (2)
Guaranteed Annuity Options 7
Valuing the GAO
Assumptions:
• single-premium payments, no expenses
• financial risk is independent from biometric risk
• in a first step, it is even assumed that S and {rt : 0 ≤ t ≤ T} are independent
• there exists an equivalent martingale measure Q, i.e. an arbitrage-free economy(discounted market prices of tradable securities are Q-martingales)
• policyholders behave rationally, i.e. policyholders select the highest annuitypayout
Guaranteed Annuity Options 8
Change of Measure Technique
• For valuing the GAO, a switch from the spot martingale measure Q to theT -forward measure QT turns out to be appropriate.
• Let X be a contingent claim that settles at T . Then
Notation Numeraire Pricing formula State price density
Spot
martingale
measure
Q B(t) πt(X) = B(t) EQ
[X
B(T )
∣∣Ft
]dQdP
∣∣∣Ft
= E[ζ|Ft]
T -forward
measure
QT P (t, T ) πt(X) = P (t, T ) EQT[X|Ft]
dQTdQ
∣∣∣Ft
= P (t,T )P (0,T )B(t)
Guaranteed Annuity Options 9
Main Steps (1/2)
• Let X = Y (T ) =(
S(T )g
)(ax(T )− g)+
• It follows from the above assumptions that the time-t value of the GAO is givenby:
πt(X) = P (t, T ) EQT
[S(T )|Ft
]EQT
[(ax(T )− g
)+∣∣Ft
]T−tpx−(T−t)
g
By the martingale condition, we have that P (t, T )EQT[S(T )|Ft] = S(t). Hence,
πt(X) = S(t) EQT
[(ax(T )− g
)+|Ft
]T−tpx−(T−t)
g(3)
Guaranteed Annuity Options 10
Main Steps (2/2)
• Using the definition of ax(T ) from (1), one obtains
πt(X) = S(t) EQT
[( J∑k=1
kpx P (T, T + k)− g)+∣∣∣Ft
]T−tpx−(T−t)
g(4)
• The expression inside the expectation in (4) corresponds to a call option on acoupon paying bond. The coupon payments at the time instants T + k are kpx.
• Jamshidian [3] showed that if the short rate follows a one-factor process, thenthe option price on a coupon paying bond equals the price of a portfolio ofoptions on the individual zero-coupon bonds (Brigo and Mercurio [2], p. 68):
CBO(t, T, τττ , ccc,K) =n∑
i=1
ci ZBO(t, T, Ti,Ki) (5)
Guaranteed Annuity Options 11
Applying the Hull-White Short Rate Model
• We now specify the term structure of interest rates via the following short ratedynamics (Hull-White one-factor model):
drt = κ(θ(t)− rt
)dt + σ dW (t) (t ≥ 0)
• In the Hull-White context, bond options can be calculated explicitly, see forinstance Brigo and Mercurio [2], p. 65.
• Example: European call option with strike K, maturity S written on a zero-bondmaturing at time T > S:
ZBC(t, S, T,K) = P (t, T ) Φ(h)−K P (t, S) Φ(h− σp) (6)
Guaranteed Annuity Options 12
with
σp =σ
κ
√1− exp{−2κ(S − t)}
2κ
(1− e−κ(S−t)
), h =
1σp
log( P (t, T )P (t, S)K
)+
σp
2
• Combining (4), (5) and (6), one obtains an explicit formula for the price of theGAO:
πt(X) = S(t)∑J
k=1 kpxZBO(t, T, T + k,Kk)P (t, T )
T−tpx−(T−t)
g(7)
Guaranteed Annuity Options 13
Discussion
• the market value for the GAO is proportional to S(t)
• the simple pricing formula (7) relies on strong assumptions such as
- equity returns are independent of interest rates
- term structure of interest rates is given by a one-factor Gaussian short ratemodel
• Formula (7) can be generalized by assuming that S(T ) and P (T, T + k) arejointly lognormally distributed. In that case, a closed-form solution for the valueof the GAO can still be derived.
Guaranteed Annuity Options 14
The Need for Multi-Factor Models
• Example: Hull-White two-factor model:
drt = κ(θ(t) + u(t)− rt
)dt + σ1 dW1(t)
du(t) = −b u(t) dt + σ2 dW2(t)
with u(0) = 0 and dW1(t)dW2(t) = ρdt.
• Note: the value of a swaption depends on the joint distribution of the forwardrates (F (t;T0, T1), F (t;T1, T2), . . . , F (t;Tn−1, Tn)). The payoff can thus not beadditively separated as in the case of e.g. a cap
• Later we will show that the payoff of GAOs can be (statically) replicated by aportfolio of receiver swaptions
Guaranteed Annuity Options 15
• Correlation among the forward rates has an impact on the contract value
• Multi-factor models allow for more general correlation patterns than one-factormodels
• Thus: simple one-factor models usually give reasonable prices for instruments,but good hedging schemes will assume many factors
Guaranteed Annuity Options 16
Interest Rate Swaps (IRS)
• Definition (Interest rate swap): contract that exchanges fixed payments forfloating payments, starting at a future time instant
• Tenor structure:
reset dates: Tα, Tα+1, . . . , Tβ−1
payment dates: Tα+1, . . . , Tβ−1, Tβ
• fixed-leg payments: N τi K (N : notional amount, τi: year-fraction from Ti−1 to Ti)
• floating-leg payments: N τi F (Ti−1;Ti−1, Ti)
Guaranteed Annuity Options 17
Value of a (payer) IRS
V (t) = EQ
[β∑
i=α+1
P (t, Ti) τi
(F (Ti−1, Ti)−K
)∣∣∣Ft
]= . . .
=β∑
i=α+1
(P (t, Ti−1)− (1 + τi K)P (t, Ti)
)
• Forward swap rate: value of the fixed-leg rate K that makes the present valueof the contract equal to zero:
Sα,β(t) =P (t, Tα)− P (t, Tβ)∑β
i=α+1 τi P (t, Ti)
Guaranteed Annuity Options 18
Swap options, Swaptions
• Definition: A European payer swap option where the holder has the right to payfixed and receive floating, is an option on the swap rate Sα,β(t). A Europeanreceiver swap option where the holder has the right to pay floating and receivefix, is an option on the swap rate Sα,β(t).
• The swaption maturity often coincides with the first reset date of the underlyingIRS
• Example: receiver swaption provides payments of the form(K − Sα,β(Tα)
)+.
If K = 7% and Sα,β(Tα) = 5%, it is optimal to exercise the option and receivefixed payments of Sα,β(Tα) + (K − Sα,β(Tα))+ = K
• By entering a receiver swaption, the holder protects itself against the risk thatinterest rates will have fallen when the swaption matures.
Guaranteed Annuity Options 19
IRS and Swaptions in a Nutshell
Type Discounted payoff at Tα Price (time-t value)
Payer IRS
β∑i=α+1
P (Tα, Ti) τi(F (Ti−1, Ti)− K
) β∑i=α+1
(P (t, Ti−1)− (1 + τi K)P (t, Ti)
)=(
Sα,β(Tα)− K)Aα,β(Tα)
PayerSwaption
( β∑i=α+1
P (Tα, Ti) τi(F (Ti−1, Ti)− K
))+Aα,β(t) EQA
[(Sα,β(Tα)− K
)+∣∣Ft
]=(
Sα,β(Tα)− K)+
Aα,β(Tα)
Receiver IRS
β∑i=α+1
P (Tα, Ti) τi(K − F (Ti−1, Ti)
) β∑i=α+1
(−P (t, Ti−1) + (1 + τi K)P (t, Ti)
)=(
K − Sα,β(Tα))Aα,β(Tα)
ReceiverSwaption
( β∑i=α+1
P (Tα, Ti) τi(K − F (Ti−1, Ti)
))+Aα,β(t) EQA
[(K − Sα,β(Tα)
)+∣∣Ft
]=(
K − Sα,β(Tα))+
Aα,β(Tα)
Guaranteed Annuity Options 20
Hedging the Interest Rate Risk of a GAO
• The quantity Aα,β(t) is given by∑β
i=α+1 τi P (t, Ti) and defines the change ofmeasure from the spot martingale measure Q to the measure QA:
dQA
dQ=
Aα,β(Tβ)/Aα,β(0)B(Tβ)/B(0)
=Aα,β(Tβ)
Aα,β(0)B(Tβ).
• Recall that the GAO gives the right to obtain a series of cash payments npxg atdifferent dates T1, T2, . . . . Hence, the interest rate exposure in a GAO is similarto that under a swaption.
• Pelsser [4] advocates the usage of long-dated receiver swaptions for dealing withthe interest rate risk under a GAO (static replicating portfolio approach)
• Price of the GAO ≡ value of a portfolio of long-dated receiver swaptions
Guaranteed Annuity Options 21
Static Replicating Portfolio
• Recall that in the case of GAOs the expresssion
(a(T )
g− 1)+
=
(J∑
k=1
kpx
gP (T, T + k)− 1
)+
gives the right to receive a series of cash payments (kpx/g) for a price of 1
• Cash flows from GAO are gradually decreasing over time (due to the decreasingsurvival probabilities), whereas cash flows associated with an N -year swap areconstant over time
• Idea: Combine positions in receiver swap contracts all starting at time T , butwith different maturities T + k
Guaranteed Annuity Options 22
Construction of the Hedge Portfolio
• Aim: determine the amount to be invested in each swap
• Let ω be the limiting age of the mortality table (e.g. ω = 120)
• At time T + (ω − x):
- cash flow to be replicated: (ω−xpx/g)- cash flow of a swap with fixed leg Kω−x and length ω − x : 1 + Kω−x
- amount Hω−x to be invested at time t:
Hω−x := ω−xpx
g(1 + Kω−x)(8)
Guaranteed Annuity Options 23
• Note: equation (8) can be rewritten as
Hω−xKω−x = ω−xpx
g−Hω−x (9)
• At time T + (ω − x)− 1:
- cash flow to be replicated: (ω−x−1px/g)- cash flow from swap with fixed leg Kω−x and length ω − x : Kω−x
- cash flow from swap with fixed leg Kω−x−1 and length ω−x−1 : 1+Kω−x−1
- amount Hω−x−1 to be invested at time t:
Hω−x−1 :=
(ω−x−1px − ω−xpx
)/g + Hω−x
1 + Kω−x−1(10)
• Observe that Hω−xKω−x + Hω−x−1(1 + Kω−x−1) = ω−x−1px/g
Guaranteed Annuity Options 24
I This yields a recursive relation for the amounts to be invested in swaps withtenor length n:
Hn =
(npx − n+1px
)/g + Hn+1
1 + Kn
Guaranteed Annuity Options 25
Price of the GAO
• With the portfolio of swaps
ω−x∑n=1
HnV swap(T,Kn)
all cash flows of the GAO can be replicated
• Value of the GAO:(ω−x∑n=1
HnV swap(T,Kn)
)+
≤ω−x∑n=1
Hn (V swap(T,Kn))+ =ω−x∑n=1
HnV swapt(T,Kn)
Guaranteed Annuity Options 26
Discussion of the Static Replicating PortfolioApproach
• Pros and cons:
+ no need for dynamic hedging (no further buying and selling until maturity)
+ based on the “right type of interest rate options”
+ swap market is more liquid than bond market
+ cheaper and better protection than reserving (reserving at 99%-level may beinsufficient)
− hedge against the interest rate risk only (hedging mortality risk by sellingmore life insurance?)
− in a period of rising stock returns, insurer must keep purchasing swaptions
Guaranteed Annuity Options 27
B. Valuation and Risk Measurement
Pricing Derivative Securities
• Consider an economy of d + 1 assets (S0, S1, . . . , Sd)
• Trading strategy: H = (H0,H1, . . . ,Hd) with Hi(t) denoting the number ofunits held of the ith asset at time t
• Value of the portfolio at time t:
V (t;H) =d∑
i=0
Hi(t) Si(t)
Guaranteed Annuity Options 28
• The strategy H is self-financing if
V (t)− V (0) =d∑
i=0
∫ t
0
Hi(u) dSi(u)
• When pricing a derivative, the drift parameters µi in the dynamics of Si donot appear: one does not need to know anything about an investor’s attitudetowards risk
• Rationale: risk preferences are irrelevant because contingent claims can bereplicated by trading in the underlying assets
• Price of the derivative: minimal investment to implement the trading strategy
Guaranteed Annuity Options 29
The Role of the Measures P and Q
Real-world measure P Risk-neutral measure Q
asset returns vary by asset class the rate of return on any risky assetis the same as the risk-free rate
the measure P describes theempirical dynamics of asset prices
in EQ[ · ]-expectation, the riskyassets behave like the money marketaccount
Guaranteed Annuity Options 30
Risk Measurement
Risk measurement poses the question:
How does the portfolio value V change in response to changes in the underlyingrisk factors?
• Z = (Z1, Z2, . . . , Zm)′ vector of risk factors
• ∆Z = Z(t + ∆t)− Z(t) : change in Z over ∆t
• Portfolio loss:L = V (t;Z)− V (t + ∆t;Z(t) + ∆Z)
Guaranteed Annuity Options 31
Calculation of the loss distribution function FL
If we were to determine the loss distribution function FL by means of Monte Carlosimulation, we would have to proceed as follows:
• For each of n replications,
(i) generate a scenario under P, i.e. a vector of risk factor changes ∆Zi
(ii) re-value the portfolio V (t + ∆t;Z(t) + ∆Zi) at time t + ∆t under Q, giventhe outcome of ∆Zi
(iii) compute the loss Li = V (t;Z)− V (t + ∆t;Z(t) + ∆Zi)
• Estimate P[L ≤ x] using
1n
n∑i=1
1{Li≤x}
Guaranteed Annuity Options 32
Estimation of a conditional expectation
The bottleneck in the above recipe is the portfolio revaluation of step (ii). Thismeans computing a conditional expectation (or an estimate thereof)
Guaranteed Annuity Options 33
References
[1] Boyle, P. and Hardy, M. (2003). Guaranteed Annuity Options. ASTINBulletin, Vol. 33, No. 2, 125-152.
[2] Brigo, D. and Mercurio, F. (2001). Interest Rate Models. Theory andPractice. Springer, Berlin.
[3] Jamshidian, F. (1989). An exact bond option formula. Journal of Finance,44, 205-209.
[4] Pelsser, A. (2003). Pricing and Hedging Guaranteed Annuity Options viaStatic Option Replication. Insurance: Mathematics and Economics. Vol. 33,283-296.
[5] Wilkie, A. D., Waters, H. R., and Yang, S. (2003). Reserving, Pricing andHedging for Policies with Guaranteed Annuity Options. British ActuarialJournal, Vol. 9, No II, 263-425.
Guaranteed Annuity Options 34