Hindawi Publishing Corporation Abstract and Applied Analysis Volume 2013, Article ID 682524, 5 pages http://dx.doi.org/10.1155/2013/682524
Research Article Homotopy Analysis Method for Boundary-Value Problem of Turbo Warrant Pricing under Stochastic Volatility
Hoi Ying Wong1 and Mei Choi Chiu2
1 Department of Statistics, The Chinese University of Hong Kong, Hong Kong 2 Department of Mathematics and Information Technology, Hong Kong Institute of Education, Hong Kong
Correspondence should be addressed to Hoi Ying Wong; [email protected]
Received 13 December 2012; Accepted 30 January 2013
Academic Editor: Bashir Ahmad
Copyright Β© 2013 H. Y. Wong and M. C. Chiu. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Turbo warrants are liquidly traded financial derivative securities in over-the-counter and exchange markets in Asia and Europe. The structure of turbo warrants is similar to barrier options, but a lookback rebate will be paid if the barrier is crossed bythe underlying asset price. Therefore, the turbo warrant price satisfies a partial differential equation (PDE) with a boundary condition that depends on another boundary-value problem (BVP) of PDE. Due to the highly complicated structure of turbo warrants, their valuation presents a challenging problem in the field of financial mathematics. This paper applies the homotopy analysis method to construct an analytic pricing formula for turbo warrants under stochastic volatility in a PDE framework.
1. Introduction properties of turbo warrants. This leads to subsequent exten- sions to turbo warrant pricing with jump-diffusion model in Turbo warrant first appears in Europe but is now available [3] and with mean reversion in [4]. Although Wong and Chan under various names in many markets including the UK, [2] approximate turbo prices under a stochastic volatility (SV) Germany, Switzerland, Italy, Australia, New Zealand, Singa- model, their solution is restricted to the assumption that the pore,SouthAfrica,Taiwan,andHongKong.Forinstance,itis mean-reverting speed of the stochastic volatility should either called turbo warrant in the Nordic Growth market, contract beclosetozeroortoinfinity.Ourgoalistoderiveananalytic for difference (CFD) in the UK, and callable bull/bear solution to turbo prices by relaxing such an assumption using contract (CBBC) in Hong Kong. According to a report by a homotopy analysis method. Hong Kong Exchanges and Clearing Limited (HKEx) in 2009 Homotopy analysis method was introduced by Ortega [1],theUKhasabigOTCmarketforCFDswhileHong and Rheinboldt [5] in 1970 and has been applied to solve Kong has a big exchange market for CBBC. Typically, HKEx many nonlinear problems since the work of Liao [6]in1992. listed 1,525 CBBCs, constituting 28% of the total number of its Zhu in [7] pioneers the use of homotopy analysis method in listed securities, and the total issued amount reached HK$704 financial mathematics and derives the first analytic formula billion or US$95 billion at the end of May 2009 [1]. Hong to American options. Zhao and Wong in [8]showthat Kong has two types of CBBC: N and R. The N-CBBC pays the approach of Zhu is also applicable to general diffusion norebatewhenapreselectedbarrieriscrossedwhereastheR- models. CBBC pays a lookback rebate. If the barrier is not breached by While general diffusion models belong to the class of the underlying asset price during the life of the contract, then complete market models, SV models are of incomplete the turbo call (put) is equivalent to the standard call (put). market models because the number of Brownian motions Therefore, turbo warrant resembles a combination of barrier driving the asset dynamics is larger than one. Park and and lookback options. Kim [9] apply homotopy analysis method to solve vanilla A comprehensive mathematical treatment of turbo war- option and barrier option prices under SV models. Leung [10] rants is given in [2], which provides several model-free extends the framework to lookback option pricing. However, 2 Abstract and Applied Analysis the application of homotopy analysis method to turbo war- where DOC(π‘, π) denotes the down-and-out call option price at rant pricing under SV models is yet to be considered. π‘,
This paper employs the homotopy analysis method to βππ π + (π ,π ,π )=πΈ [π 0 (π 0 βπΎ) 1 ] solve the PDE for the turbo warrant price under a SV model. LB π» ππ» 0 ππ» ππ» {ππ»β€π} As the price of turbo warrant under the Black-Scholes model = (0, π , (π ,πΎ),V ,π ) (3) is available, we construct a homotopy which deforms from LC ππ» min ππ» ππ» 0 the Black-Scholes solution to the desired solution under β (0, π ,π , V ,π ), theSVmodel.Wehighlightthefundamentalchallengein LC ππ» ππ» ππ» 0 turbo warrant pricing. A turbo warrant consists of a barrier option and a lookback rebate. Although the barrier option and LC(π‘,π,π,V,π0) is the floating strike lookback call at time pricing under SV models is investigated in [9], Park and Kim π‘ on π with the realized minimum π,instantaneousvolatility do not consider rebate in knockout options. Our solution V, and time to maturity π0. should cover the case of state-dependent rebate under SV. Second, Leung [10] offers an analytic pricing formula for In Proposition 1, the floating strike lookback call, (π‘,π,π,V,π ) lookback options but his formula cannot be directly used in LC 0 ,hasthepayoff calculating the lookback rebate, which is an expectation on π β {π }. π0 min π the discounted lookback option with a random starting time. 0β€πβ€π0 (4) Therefore, our homotopy analysis method has to simultane- ously solve these two problems together. Hence, its price has the representation, The reminder of this paper is organized as follows. βπ(π0βπ‘) Section 2 presents the nature of turbo warrants under SV LC (π‘, π, π, V,π0)=π πΈπ‘ [ππ β min {ππ}] . (5) 0 0β€πβ€π models and the corresponding pricing problem in a PDE 0 βππ approach. Section 3 contains the main result of this paper and πΈ [π π» 1 (π ,π ,π )] The term π‘ {ππ»β€π}LB π» ππ» 0 is called the down- solves the PDE using homotopy analysis method. Concluding and-in lookback (DIL) option in [2] as a TC holder will remark is made in Section 4. knock in the lookback option LB, shown in (3), only when the underlying asset price hits the down-side barrier π». 2. Problem Formulation In particular, if the asset price process follows the Black- Scholes (BS) model where the volatility is a constant, then the π‘β[0,π ] Let ππ‘ be the underlying asset price at time π‘.Aturbocall turbo call price at π» reduces to + (put) warrant pays the contract holder (ππ βπΎ) at maturity TC (π‘,) π = DOC (π‘,) π π if a specified barrier π»β₯πΎhas not been passed by ππ‘ at BS BS any time prior to the maturity. Denote ππ» as the first passage +[ (0,π»,πΎ,π ) LCBS 0 (6) time that the asset price crosses the barrier π»,thatis,ππ» = {π‘:π β€π»} π β€π β (0,π»,π»,π )] (π‘,) π , inf π‘ .If π» ; then the contract is void and LCBS 0 DRBS a new contract starts. The new contract is a call option on π0 ππ = minπ β€π‘β€π +π ππ‘,withthestrikepriceπΎ,andthetime where the Black-Scholes pricing formulas of DOCBS,LCBS, π» π» π» 0 βπππ» =πΈπ‘[π 1{π β€π}] to maturity π0.Moreprecisely,theturbocall(TC)attimeπ‘ and DRBS π» are classical results and canbeexpressedas can be found in [11]. However, the formulas of LCBS and DRBS also appear in some proofs of this paper. Specifically, LC (π‘,π,πΎ,π0)=ππ0(π‘, π¦, V,π0),whereπ0 is given by (19) + BS =πΈ[πβππ(π βπΎ) 1 while DR appears in (28). TCπ‘ π‘ π {ππ»>π} BS (1) βπ(π +π ) π + +π π» 0 (π 0 βπΎ) 1 ], 2.1. Turbos under Stochastic Volatility. Although the explicit ππ» {ππ»β€π} BS formula for the TC is known as in (6), its analytic pricing formula under SV model is yet to be considered. where π is the constant interest rate and πΈπ‘ represents the risk- The Heston SV model assumes the following stochastic neutral expectation given information up to time π‘.Wong differential equation (SDE) for the underlying asset price andChan[2] explain the incentive of this security design and πππ‘ π show a model-free representation that the turbo call warrant =(πβπ)ππ‘+βVπ‘ πππ‘ , (7) is sum of a down-and-out call (DOC) and an expectation of ππ‘ a nonstandard lookback option (LB) starting at ππ». where π is the constant interest rate, π is the constant dividend Proposition 1 π‘β€π yield, and Vπ‘ is the stochastic instantaneous variance of the (see [2]). At π», the model-free representa- V tion of the turbo call warrant is asset. The stochastic variance π‘ follows the SDE: V πVπ‘ =π (πβVπ‘)ππ‘+πβVπ‘ πππ‘ , (8)
βπππ» TC (π‘,) π = DOC (π‘,) π +πΈπ‘ [π 1{π β€π}LB (ππ»,ππ ,π0)] , π V π» π» where π , π,andπ are constants. In (7)and(8), ππ‘ and ππ‘ are (2) Wiener processes defined on a filtered complete probability Abstract and Applied Analysis 3
π V space (Ξ©, F, Q, Fπ‘β₯0) with πΈ[ππ‘ ππ‘ ]=ππ‘,whereQ denotes In principal, the exotic lookback option LB(π‘, π, V,π) the risk-neutral probability measure. should be solved from an alternative PDE governing lookback Let π(π‘,Μ π, V) be the turbo warrant price at time π‘ under options. Fortunately, we know from (11)thatthisexotic the stochastic volatility V.Using(1), we have lookback option is the difference between two floating strike lookback calls. Μ π (π‘, π, V) Specifically, LC(π‘,π,π,V,π)has the following homotopy
βππ + expression [10]: =πΈ[π (ππ βπΎ) 1{π >π} π» σ΅¨ LC (π‘, π, π, V,π) =πβ π(π‘,π¦,V,π,π)σ΅¨ , βπ(π +π ) π + σ΅¨π=1 (16) +π π» 0 (π 0 βπΎ) 1 |π =π,V = V]. ππ» {ππ»β€π} π‘ π‘ (9) where π¦=ln(π/π);
By Proposition 1, the turbo price can be expressed as (2). β π As the SV model assumes a continuous process for the π(π‘, π¦, V,π,π) = βππ (π‘, π¦, V,π) π . (17) π =π» π=0 underlying asset price, ππ» .Hence,wehave πΜ (π‘, π, V) The functions ππ(π‘, π¦, V,π), π = 1, 2, .,areobtainedfrom . the following iteration: =πΈ[πβππ(π βπΎ)+1 π {ππ»>π} Μ βππ π (π‘, π¦, V,π)=π(π‘,π¦,V) π (π‘, π¦, V,π), +π π» 1 (π ,π»,V ,π)|π =π,V = V], π π {ππ»β€π}LB π» ππ» 0 π‘ π‘ 1 1 (10) π(π,π¦,V)= {(βπβ V(πΌ (V) +1)2)π+ (πΌ (V) +1) π₯} , exp 8 2 where 2(πβπ) 1 LB (ππ»,π»,Vπ ,π0)=LC (0,π»,πΎ,Vπ ,π0) πΌ (V) = ,π=(πΌ (V) +1) , π» π» V 2 (11) β (0, π», π», V ,π ). π β LC ππ» 0 Μ ππ (π‘, π¦, V,π)= β« β« π (π , π, V) Mππβ1 (πβπ ,π,V,π) 0 0 Let π₯π‘ = ln(ππ‘/π»). By Itoβs lemma, we obtain the SDE for π₯π‘ as Γ πΊ (π β π , π¦,π, V)ππππ ,
Vπ‘ π π0 2 ππ₯ = (πβπβ ) + V ππ ,π₯= ( ) . π π π 1 2 π π‘ β π‘ π‘ 0 ln (12) M =ππV βππV +π (πβV) + π V , 2 π» πV ππ¦πV πV 2 πV2 Let π(π‘, π₯, V)=π(π‘,Μ π, V). Applying the Feynman-KacΛ 1 β(π¦βπ)2/2Vπ‘ β(π¦+π)2/2Vπ‘ formula to π in (10)withrespectto(12)and(8), we have πΊ (π‘, π¦, π, V)= {π +π β2πVπ‘ (L + M) π=0 β β((π¦+π+π)2/2Vπ‘)+ππ π₯ + +2π β« π ππ} , π (π, π₯, V) =(π»π βπΎ) (13) 0 (18) π (π‘, π₯, V)|π₯=0 = LB (π‘, π», V,π0), where the iteration begins with the function where π V π 1 π2 π (π‘, π¦, V,π)=π¦βπ(πβπ‘) π π(βπβ )βπβπ(πβπ‘)π(βπ+ ) L = +(πβπβ ) + V βπ, 0 π π 2 ππ‘ 2 ππ₯ 2 ππ₯ V (14) + [πβπ(πβπ‘)π(π+ ) π π2 1 π2 2(πβπ) π M =π (πβV) +ππV + π2V . πV ππ₯πV 2 πV2 βπ(πβπ‘)+(π¦(πβπ)/V) π βπ π(ππ)] , In addition, the rebate LB is related to LC through (11). (19) Using the transformation of variable LC(π‘, π, π, V,π) = π β π(π‘, log(π/π), V,π),thefunctionπ is the solution of the Β± βπ¦ + (π β πΒ± (V/2)) (πβπ‘) following BVP [10]: π = , π βV (πβπ‘) (L + M) π=0 (20) π + πβπ‘ π₯ π =π β2(πβπ)β , π (π, π₯, V,π) =π β1 π π V (15) σ΅¨ ππ (π‘, π₯, V,π)σ΅¨ σ΅¨ =0. with π(β ) being the standard normal cumulative distribution σ΅¨ ππ₯ σ΅¨π₯=0 function. 4 Abstract and Applied Analysis
3. The Homotopy Framework Consider the Taylor expansion of π(π‘, π₯, V,π)with respect to π as follows: We aim to construct a homotopy solution π(π‘, π₯, V,π) with β auxiliary parameter πβ[0,1]such that π(π‘, π₯, V,0) = π π(π‘,π₯,V,π)= βππ (π‘, π₯, V) π , (25) π0(π‘, π₯, V) and π(π‘, π₯, V, 1) = π(π‘, π₯, V), the desired pricing π=0 formula. The function π0 can be regarded as the initial guess π π ofthesolution.WeusetheBSformulaofTCin(6)as where ππ(π‘, π₯, V) = (1/π!)(π /ππ )π(π‘, π₯, V,π)|π=0.Substitut- the initial guess. Therefore, π0(π‘, π₯, V) is the solution of the ing (25)into(22) yields following PDE: Lππ (π‘, π₯, V) + Mππβ1 (π‘, π₯, V) =0, Lπ0 (π‘, π₯, V) =0 π (π, π₯, V) =0, π₯ + π π0 (π, π₯, V) =(π»π βπΎ) σ΅¨ πΎ σ΅¨ σ΅¨ πΎ ππ (π‘, π₯, V)σ΅¨ =π»[ππ (0, 0, V,π0)βππ (0, ln , V,π0)] , π (π‘, π₯, V)σ΅¨ =π»[π (0, 0, V,π )βπ (0, , V,π )] , σ΅¨π₯=0 π» 0 σ΅¨π₯=0 0 0 0 ln π» 0 (21) for π = 1, 2, . . ., (26) where L is defined in (14)andπ0 in (19). π(π‘, π₯, V,π) Consider the following PDE for : where ππ(0, π¦, V,π0) is a known function given in (19). Let π=πβπ‘.ThenthePDE(26) becomes, for π = 1, 2, ., . βπ (L + M) π=(1βπ)L (πβπ0) ππ 1 π2π V ππ π₯ + π = V π +(πβπβ ) π βππ + Mπ π(π,π₯,V,π)=(π»π βπΎ) ππ 2 ππ₯2 2 ππ₯ π πβ1 σ΅¨ πΎ π (π, π₯, V) =0 π(π‘,π₯,V,π)σ΅¨ =π»[π (0,0, V,π ,π)βπ(0, , V,π ,π)], π σ΅¨π₯=0 0 lnπ» 0 σ΅¨ (22) π (π, π₯, V)σ΅¨ =π»[π (0, 0, V,π ) π σ΅¨π₯=0 π 0 where π₯=ln(π/π»), π is defined in (17), and L and M are πΎ β defined in (14). βππ (0, ln , V,π0)] := π»π (V) . If we set π=1in (22), then we have π» (27) (L + M) π=0 This PDE can be transformed into a heat equation and π (π, π₯, V,1) =(π»ππ₯ βπΎ)+ its solution has a classical result. Specifically, the result is summarized as follows. σ΅¨ πΎ π (π‘, π₯, V,1)σ΅¨ =π»[π (0, 0, V,π0,1)βπ(0,ln , V,π0,1)] σ΅¨π₯=0 π» Theorem 2. Suppose the underlying asset price, ππ‘,followsthe SV model of (7) and (8).Then,theturbocallwarrantprice = LC (π», π», V,π0)βLC (π», πΎ, V,π0) in (1) has an analytic formula derived from the iteration: for π = 1, 2, ., . = LB (π‘, π», V,π0). π (π, π₯, V) =π»β (V) (π, π₯, V) (23) π π DRBS By comparing the systems (13)and(23), we confirm that π β + β« β« Mππβ1 (π , π₯, V) πΊ (πβπ ,π₯,π,V) ππ ππ ; π(π‘, π₯, V, 1) = π(π‘, π₯, V). 0 0 Alternatively, setting π=0in (22)yields σ΅¨ V π π σ΅¨ (π, π₯, V) = β« πΊ (πβπ ,π₯,π,V)σ΅¨ ππ ; DRBS σ΅¨ Lπ (π‘, π₯, V,0) = Lπ0 (π‘, π₯, V) 2 0 ππ σ΅¨π=0
π₯ + 1 2 π (π, π₯, V,0) =(π»π βπΎ) πΊ (π‘, π₯, π, V) = π((πβπβ(V/2))(πβπ₯)/V)β(π+((πβπβ(V/2)) /2V))π‘ β2πVπ‘ σ΅¨ πΎ π (π‘, π₯, V,0)σ΅¨ =π»[π (0, 0, V,π0,0)βπ(0,ln , V,π0,0)] 2 2 σ΅¨π₯=0 π» Γ(πβ(π₯βπ) /2Vπ‘ βπβ(π₯+π) /2Vπ‘). πΎ =π»[π (0, 0, V,π )βπ (0, , V,π )] , (28) 0 0 0 lnπ» 0 (24) An analytic pricing formula of TC under SV model is then given by π (0, π¦, V,π ) where 0 0 is obtained in (19). Therefore, β π(π‘, π₯, V,0) = π0(π‘, π₯, V).Hence,thePDE(22) is desired π (π‘, π₯, V) = π (π, π₯, V,1) = βππ (π, π₯, V) . (29) construction for the homotopy solution. π=0 Abstract and Applied Analysis 5
3.1. DOC and DIL. As the TC price paying a lookback References rebate is a sum of the DOC and DIL prices, the solution of π(π‘, π₯, V,π) from (25)and(28) is useful to determine the [1] The HKEx callable bull/bear contract (CBBC) market, Report by Research & Corporate Development Department of Hong DOC option price and DIL option price under SV. Hence, Kong Exchanges and Clearing Limited, 2009, http://www.hkex our solution also contributes to the valuation of barrier-type .com.hk/eng/stat/research/rpaper/Documents/cbbc2009.pdf. options under SV model. Note that the N-CBBC is actually a [2]H.Y.WongandC.M.Chan,βTurbowarrantsunderstochastic DOC option. volatility,β Quantitative Finance,vol.8,no.7,pp.739β751,2008. For DOC option with barrier level π»,thecorresponding [3] H. Y. Wong and K. Y. Lau, βAnalytical valuation of turbo PDE is the same as (13)exceptthattheboundarycondition π(π‘, π₯, V)| =0 warrants under double exponential jump diffusion,β Journal of is replaced by π₯=0 .Using(28)and(29), we Derivatives,pp.61β73,2008. immediately obtain that [4] H. Y. Wong and K. Y. Lau, βPath-dependent currency options β with mean reversion,β The Journal of Futures Markets,vol.28, DOC (π‘, π₯, V,π) = βDOCπ (π, π₯, V) , (30) pp. 275β293, 2008. π=0 [5]J.M.OrtegaandW.C.Rheinboldt,IterativeSolutionofNonlin- where ear Equations in Several Variables, Academic Press, New York, π β NY, USA, 1970. DOCπ (π, π₯, V) = β« β« MDOCπβ1 (π , π₯, V) [6] S. J. Liao, The proposed homotopy analysis technique for the 0 0 solution of nonlinear problems [Ph.D. thesis], Shanghai Jiao Tong ΓπΊ(πβπ ,π₯,π,V) ππ ππ , (31) University, 1992. [7] S.-P. Zhu, βAn exact and explicit solution for the valuation of (π, π₯, V) = (π‘, π₯, V,π) . American put options,β Quantitative Finance,vol.6,no.3,pp. DOC0 DOCBS 229β242, 2006. Here, we see that the DOC price under SV model from our [8]J.ZhaoandH.Y.Wong,βAclosed-formsolutiontoAmer- approach is consistent with the result by Park and Kim [9]. ican options under general diffusion processes,β Quantitative For DIL option with barrier level π»,thecorresponding Finance,vol.12,no.5,pp.725β737,2012. PDE is the same as (13) except that the terminal condition is [9] S.-H. Park and J.-H. Kim, βHomotopy analysis method for replaced by π(π, π₯, V)=0.Using(28)and(29), we obtain that option pricing under stochastic volatility,β Applied Mathematics Letters,vol.24,no.10,pp.1740β1744,2011. β [10] K. S. Leung, βAn analytic pricing formula to lookback options DIL (π‘, π₯, V,π) = βDILπ (π, π₯, V) , (32) π=0 under stochastic volatility,β Applied Mathematics Letters,vol.26, pp.145β149,2013. where [11] Y.-K. Kwok, Mathematical Models of Financial Derivatives, (π, π₯, V) =π»β (V) (π, π₯, V) Springer, Berlin, Germany, 2008. DILπ π DRBS π β + β« β« MDILπβ1 (π , π₯, V) 0 0 (33) ΓπΊ(πβπ ,π₯,π,V) ππ ππ ,
(π, π₯, V) =π»β (V) (π, π₯, V) . DIL0 0 DRBS
4. Conclusion WeuseaPDEapproachtosolvethepriceofturbowarrant underaSVmodel.ThePDEissolvedbymeansofhomotopy analysis method. The boundary condition of the PDE is simplified using the homotopy solution developed in10 [ ]. As byproducts, we offer analytic pricing formulas for DOC and DIL options under SV model. Future research can apply this solution to investigate the impact of volatility to the prices of turbo warrants empirically.
Acknowledgments H. Y. Wong acknowledges the support by GRF of Research Grant Council of Hong Kong with Project n. 403511. M. C. Chiu acknowledges the Start-up Research Grant RG44/2012- 2013R by Department of MIT, Hong Kong Institute of Education. Copyright of Abstract & Applied Analysis is the property of Hindawi Publishing Corporation and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.