Pricing TARN Using a Finite Difference Method - arXiv

1 [ ] 21 Aug 2014 Pricing TARN Using a Finite Difference MethodXiaolin Luo1, and Pavel V. Shevchenko2 Draft, this version 29 April 20131 CSIRO Mathematics, Informatics and Statistics, Sydney, Australia; e-mail: Mathematics, Informatics and Statistics, Sydney, Australia; Corresponding authorAbstractTypically options with a path dependent payoff, such as TargetAccumulationRedemption Note (TARN), are evaluated by a Monte Carlo Method . This paperdescribes a Finite difference scheme for Pricing a TARN steps in theproposed scheme involve tracking of multiple one-dimensional Finite differencesolutions, application of jump conditions at each cash flow exchange date, anda cubic spline interpolation of results after each jump.

2 Since a Finite differencescheme for TARN has significantly different features from a typical Finite dif-ference scheme for options with a path independent payoff, we give a step bystep description on the implementation of the scheme, whichis not available inthe literature. The advantages of the proposed Finite difference scheme over theMonte Carlo Method are illustrated by examples with three different knockouttypes. In the case of constant or time dependent volatility models (where MonteCarlo requires simulation at cash flow dates only), the finitedifference methodcan be faster by an order of magnitude than the Monte Carlo Method to achievethe same accuracy in price.

3 Finite difference Method can be even more efficientin comparison with Monte Carlo in the case of local volatility model where MonteCarlo requires significantly larger number of time steps. Interms of robust andaccurate estimation of Greeks, the advantage of the Finite difference Method willbe even more :Target Accumulation Redemption Note, option Pricing , Finite dif-ference, Monte Carlo11 IntroductionPath dependent options have payoffs depending on the trajectory followed by one ormore of the underlying processes. The most straightforward andeasy to implementnumerical solution for Pricing path-dependent options is based on the Monte Carlomethod.

4 In the context of Pricing path-dependent options by solving partial differ-ential equations (PDE), two additional challenges may merge due tothe presence ofpath dependency. First, the dependency may introduce new dimensions to the partialdifferential equation. Second, it may cause the resulting equation much more difficultto solve because of the lack of diffusion in the additional dimensions. For some detaileddiscussions, see Tavella and Randall (2000), Zvan et al (1998) andWilmott (2000b).The nature of the path-dependent option Pricing problem largely depends on whetherwe have a continuous or discrete sampling for the path.

5 In general ,a continuous sam-pling model of path dependency introduces additional convection terms in PDE, whilefor a discretely sampled path-dependent option the convection terms are replaced byjump conditions. There are many successful attempts in Pricing discretely sampledpath dependent options by the PDE approach Using lattice based Method ( bino-mial and trinomial trees used in Ritchken et al 1993, Hull and White 1993, Barraquandand Pudet 1996, Forsyth et al 2002), and similarly Finite volume or Finite elementmethod (Forsyth et al 1999, Zvan et al 2001). Most of these studies consider Asianor lookback options.

6 Typically, a linear interpolation is adapted in these methods inapplying the jump conditions on the auxiliary variable ( path average of the under-lying asset). The convergence study by Forsyth et al (2002) shows that it is possiblefor an algorithm based on lattice Method to be non-convergent (orconvergent to anincorrect answer) if the interpolation scheme is selected Target Accumulation Redemption Note (TARN) provides a capped sum of pay-ments over a period with the possibility of early termination (knockout) determinedby the target level imposed on the accumulated amount. A certain amount of payment( spot value minus the strike) is made on a series of cash flow dates (referred toas fixing dates) until the target level is breached.

7 The payoff function of a TARN ispath dependent in that the payment on a fixing date depends on thespot value ofthe asset as well as on the accumulated payment amount up to the fixing date. Typi-cally, commercial software solutions for Pricing a TARN are based onthe Monte Carlomethod. This paper presents a Finite difference scheme as an alternative to the MonteCarlo Method to evaluate TARN. The focuses are on the step by step implementationof the Finite difference scheme, which is not readily available in the literature, and onthe comparison of performance of the proposed scheme relative to the Monte are not aware of any Finite difference scheme published in the literature, although2a general outline of PDE approach to Pricing TARN can be found in Piterbarg (2004).

8 Without losing generality, we assume the underlying asset is the foreign exchange(FX) rate. The definitions of TARN options with three different knockout types andsome key notations are introduced in Section 2. Foreign exchange rate models aredescribed in Section 3. Finite difference scheme for TARN is presented in Section 4and numerical results for both the Finite difference and Monte Carlomethods are givenin Section 5, before concluding by Section TARN Payoff DefinitionThere are different versions of TARN products used in FX trading. For simplicity,here we consider one specific form of TARN.

9 The presented Finite difference schemecan easily be adapted to other more general forms of TARN as discussed in Denote the FX rate at timetasS(t) and other notation as follows:t0is today sdate;Kis the number of fixing dates (cash flow dates);t1, t2, .. , tKare fixing dates;Xis strike;Uis the target accrual level;S(t1), S(t2), .. , S(tK) are FX rate valuesat fixing datest1, t2, .. , tK;A(t) is accumulated amount at timet; and all amountsare per unit of notional foreign amount. On each fixing datetk, there is a cash flowpaymentfCk (S(tk) X) 1 S(tk) X,(1)where is a strategy on foreign currency ( = 1 corresponds to buy and = 1corresponds to sell), subject to the target levelUis not breached by the accumulatedamountA(tk).

10 If the target levelUis breached before or on the last fixing date, denotet Kis the first fixing date when the target is breached, min{k:A(tk) U}, k= 1,2, .. , K .(2)Otherwise, seteK=K. The actual payment on the fixing datetk t Kcan be writtenasCk(S(tk), A(tk 1)) eCk (1A(tk 1)+ Ck<U+Wk 1A(tk 1)+ Ck U),(3)andCk= 0 fortk> t K. Here,A(tk 1) is the accumulation amount immediately afterthe fixing datetk 1, andWkis the weight depending on the type of the knockout whenthe target levelUbreached. The accumulated amountA(t) is a piece-wise constantfunctionA(t) =A(tk 1),tk 1 t < tkwithA(tk) =A(tk 1) +Ck(S(tk), A(tk 1)).