optimal surrender policy for variable annuity guarantees
TRANSCRIPT
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Optimal Surrender Policy for VariableAnnuity Guarantees
Anne MacKay
University of Waterloo
January 31, 2013
Joint work withDr. Carole Bernard, University of Waterloo
Max Muehlbeyer, Ulm University
Research funded by the Hickman Scholarship of the Society of Actuariesand NSERC
Anne MacKay – University of Waterloo Optimal Surrender for VA 1/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Outline
1 Introduction
2 Optimal surrender boundary: GMAB case
3 Optimal exercise boundary: Path-dependent case
4 Conclusion
Anne MacKay – University of Waterloo Optimal Surrender for VA 2/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Variable annuities and financial options
� Variable annuities: similar to mutual funds with additionalguarantees
� VA guarantees paid for via a fixed fee rate throughout the lifeof the contract (not paid upfront)
� Decreases return on fund� Impacts value of option
� VA contract can be surrendered (Knoller, Kraut, andSchoenmaekers (2011))
� Financial needs� Higher costs of opportunity� Moneyness of the option
Anne MacKay – University of Waterloo Optimal Surrender for VA 3/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Continuous fee and the surrender option
� Simple payoff at maturity:
max(G ,FT ) = FT + (G − FT )+
� Option paid by continuous fee set as percentage of fund:� Fee is low when option value is high� Incentive to surrender when fund value is high
� Surrender region: surrender benefit higher than payoffexpected if contract is kept
Anne MacKay – University of Waterloo Optimal Surrender for VA 4/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Setting
Index value St :
dSt
St= r dt + σdWt
Account value Ft based on index:
dFt
Ft= (r − c)dt + σdWt
⇒ Ft |Fs ∼ LN (log(Ft) + (r − c − σ2
2)(t − s), σ2(t − s))
Anne MacKay – University of Waterloo Optimal Surrender for VA 5/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Contract
� Accumulation benefit:
max(FT ,G ), G = F0egT , g < r
� Surrender benefit:e−κ(T−t)Ft
Anne MacKay – University of Waterloo Optimal Surrender for VA 6/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Integral representation for surrender option
� Similarities between surrender option and American option
� Can use techniques developed for American options (Kim(1990), Kim and Yu (1996), Carr, Jarrow, and Myneni (1992),Wu and Fu (2003))
� Integral can be obtained in different ways:� Finite number of surrender times (as in Kim (1990))� No-arbitrage arguments (as in Kim and Yu (1996))
Anne MacKay – University of Waterloo Optimal Surrender for VA 7/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Trading strategy for surrender option
� Confirm that optimal strategy is a threshold strategy
� Hold the VA below the surrender boundary Bt
� When Ft crosses Bt from below, sell the VA and invest theproceeds
� When Ft crosses Bt from above, use the investment to buythe VA
� Payoff of portfolio is payoff of VA
Anne MacKay – University of Waterloo Optimal Surrender for VA 8/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Gain from surrender
Proposition
The benefit associated with the surrender option between[t, t + dt] for an infinitesimal time step dt is given by
e−κ(T−t)(c − κ)Ftdt
Anne MacKay – University of Waterloo Optimal Surrender for VA 9/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Proof of Proposition
� Suppose surrender at t.
� Policyholder receives
e−κ(T−t)Ft = e−κ(T−t)e−ctSt = e−κT e−(c−κ)tSt
� To buy the VA at time t + dt, policyholder only needs
e−κ(T−(t+dt))Ft+dt = e−κT e−(c−κ)(t+dt)St+dt
� Investment made at t becomes
e−κT e−(c−κ)(t+dt)St+dt + e−κT e−(c−κ)tSterdt(1− e−(c−κ)dt)
= e−κ(T−(t+dt))Ft+dt + e−κ(T−t)(c − κ)Ftdt + o(dt)
Anne MacKay – University of Waterloo Optimal Surrender for VA 10/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Price of VA with surrender
Price of VA contract
V (t,Ft) = E [e−rt max(G ,FT )|Ft ]︸ ︷︷ ︸Maturity benefit
+
e−κT (c − κ)Ftect
∫ T
te−(c−κ)uΦ
(d1(Ft ,Bu, u, t)
)du︸ ︷︷ ︸
Surrender option
� Maturity benefit: Similar to vanilla option under Black-Scholes
� Surrender option:∫ T
te−r(u−t)
∫ ∞Bu
e−κ(T−u)(c − κ)xfFu(x |Ft)dxdu.
Anne MacKay – University of Waterloo Optimal Surrender for VA 11/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Optimal exercise boundary condition
� At maturity BT = GT and along the surrender boundary,V (Ft , t) = e−κ(T−t)Ft = Bt .
� Work backwards to solve for Bt
Bt =v(Ft , t) + e(Ft , t)
=e−c(T−t)Bteκ(T−t)Φ(d1(Bte
κ(T−t),GT ,T , t))
+ e−r(T−t)GTΦ(d2(Bteκ(T−t),GT ,T , t))
+ (c − κ)Bte(c−κ)t
∫ T
t
e−(c−κ)uΦ(d1(Bteκ(T−t),Bu, u, t))du
Anne MacKay – University of Waterloo Optimal Surrender for VA 12/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Numerical example
Assumptions:
� T = 15
� G = F0 = 100
� κ = 0, unless otherwise indicated
� c = 0.91% (fair fee for maturity benefit)
� r = 0.03
� σ = 0.2, unless otherwise indicated
Anne MacKay – University of Waterloo Optimal Surrender for VA 13/22
0 5 10 15
100
120
140
160
180
200
Time t (Years)
Und
erly
ing
fund
pric
e F
(t)
Optimal exercise boundary, sensitivity analysis: sigma
sigma=0.15sigma=0.20sigma=0.25sigma=0.30
0 5 10 15
100
120
140
160
180
200
Time t (Years)
Und
erly
ing
fund
pric
e F
(t)
Optimal exercise boundary, sensitivity analysis: kappa
k=0k=c/4k=c/3k=c/2
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Geometric average
� Consider the payoff max(GT ,YT ), where YT is the geometricaverage defined as
Yt = exp
1
t
t∫0
lnFsds
� The conditional distribution of Yu|(Yt ,Ft) for u > t is again
log-normal with mean and variance given by
Mgt =
t
ulnYt +
u − t
ulnFt +
r − c − σ2
2
2u(u − t)2
V gt =
σ2
3u2(u − t)3
Anne MacKay – University of Waterloo Optimal Surrender for VA 16/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Pricing formula
Theorem
Let V g (Yt ,Ft , t) denote the price at time t of the VA with guarantee GT
and a surrender benefit equal to e−κ(T−t)Yt . Then V g (Yt ,Ft , t) can bedecomposed into a European part vg (Yt ,Ft , t) and an early exercisepremium eg (Yt ,Ft , t)
V g (Yt ,Ft , t) = vg (Yt ,Ft , t) + eg (Yt ,Ft , t),
where
vg (Yt ,Ft , t) = e−r(T−t)eMgt +
Vgt2 Φ(
− ln(GT )+Mgt +V g
t√V gt
)+ e−r(T−t)GTΦ
(ln(GT )−Mg
t√V gt
),
eg (Yt ,Ft , t) = e−κT ertT∫t
eu(κ−r)eV̂u,t
2 Ytut F
u−t2u
t E [k(u,Fu, t)] du
Anne MacKay – University of Waterloo Optimal Surrender for VA 17/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Particularities of the path-dependent case
� Optimal surrender behaviour depends on account value Ft andgeometric average Yt .⇒ Optimal surrender surface
� To solve for optimal surrender surface, need to consider manyvalues Ft at each time t.
� To simplify calculations, assume that Bt(Ft) has the form
Bt(Ft) = max(GT e−r(T−t), at + btFt)
Anne MacKay – University of Waterloo Optimal Surrender for VA 18/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Numerical example
Additional assumptions:
� T = 10
� Payoff: max(YT ,GT )
� GT = F0e0.025T
� c = 0.0197
Anne MacKay – University of Waterloo Optimal Surrender for VA 19/22
0 1 2 3 4 5 6 7 8 9 10
6080
100
120140
160180
20090
100
110
120
130
140
150
160
timeF
t
Opt
imal
Bou
ndar
y
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Conclusion
� Integral representation for the surrender option
� Can retrieve optimal surrender boundary
� Can be used for path independent and path dependent payoffs
Future work:
� Use for other types of fee structures
� Consider flexible premium (as in Chi and Lin (2013))
Anne MacKay – University of Waterloo Optimal Surrender for VA 21/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
References
Carr, P., R. Jarrow, and R. Myneni (1992): “Alternative characterizations of American put options,”Mathematical Finance, 2(2), 87–106.
Chi, Y., and S. X. Lin (2013): “Are Flexible Premium Variable Annuities Underpriced?,” working paperavailable at SSRN.
Kim, I. J. (1990): “The Analytic Valuation of American Options,” The Review of Financial Studies, 3(4), 547–572.
Kim, I. J., and G. G. Yu (1996): “An alternative approach to the valuation of American options andapplications,” Review of Derivatives Research, 1(1), 61–85.
Knoller, C., G. Kraut, and P. Schoenmaekers (2011): “On the Propensity to Surrender a VariableAnnuity Contract,” Discussion paper, Working Paper, Ludwig Maximilian Universitaet Munich.
Wu, R., and M. C. Fu (2003): “Optimal exercise policies and simulation-based valuation for American-Asianoptions,” Operations Research, 51(1), 52–66.
Anne MacKay – University of Waterloo Optimal Surrender for VA 22/22
Introduction Optimal surrender boundary: GMAB case Optimal exercise boundary: Path-dependent case Conclusion References
Thank you for your attention!
Anne MacKay – University of Waterloo Optimal Surrender for VA 23/22