Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process

On Black and Scholes market Investor buys a European call option. At each moment of time till the maturity he is allowed to resell the option for the quoted market price. In Kukush et al. (2006) On reselling of European option, Theory Stoch. Process., 12(28), 75-87, a similar problem was investigate...

Full description

Saved in:
Bibliographic Details
Date:2008
Main Authors: Pupashenko, M., Kukush, A.
Format: Article
Language:English
Published: Інститут математики НАН України 2008
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/4573
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process / M. Pupashenko, A. Kukush // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 3-4. — С. 114-128. — Бібліогр.: 6 назв.— англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860083114473160704
author Pupashenko, M.
Kukush, A.
author_facet Pupashenko, M.
Kukush, A.
citation_txt Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process / M. Pupashenko, A. Kukush // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 3-4. — С. 114-128. — Бібліогр.: 6 назв.— англ.
collection DSpace DC
description On Black and Scholes market Investor buys a European call option. At each moment of time till the maturity he is allowed to resell the option for the quoted market price. In Kukush et al. (2006) On reselling of European option, Theory Stoch. Process., 12(28), 75-87, a similar problem was investigated for another model of the market price. We propose a more realistic model based on Cox-Ingersoll-Ross process. Discrete approximation for this model is investigated, which is arbitrage–free. For this discrete model, a formula for penultimate optimal stopping domains is derived.
first_indexed 2025-12-07T17:17:50Z
format Article
fulltext Theory of Stochastic Processes Vol.14 (30), no.3-4, 2008, pp.114-128 MYKHAILO PUPASHENKO AND ALEXANDER KUKUSH RESELLING OF EUROPEAN OPTION IF THE IMPLIED VOLATILITY VARIES AS COX-INGERSOLL-ROSS PROCESS On Black and Scholes market Investor buys a European call option. At each moment of time till the maturity he is allowed to resell the option for the quoted market price. In Kukush et al. (2006) On reselling of European option, Theory Stoch. Process., 12(28), 75-87, a similar problem was investigated for another model of the market price. We propose a more realistic model based on Cox-Ingersoll-Ross process. Discrete approximation for this model is investigated, which is arbitrage–free. For this discrete model, a formula for penultimate optimal stopping domains is derived. 1. Introduction In this paper we consider the European call option. For this type of option Investor is not entitled to exercise the option before the time T and should wait until the maturity. However it is known that on real financial markets he has an opportunity to resell the option before the maturity. Thus we investigate the reselling problem. In this paper we treat the following model. On the Black-Scholes security market with an interest rate r, at the moment t0 = 0 Investor buys the European call option with the strike price K and the maturity T on the stock with initial value S0, at the price C0 computed by the Black-Scholes formula. At any moment t ∈ (0, T ) he can resell the option for a certain market price Cm t , which may differ from the ”fair” price Ct. The paper is organized as follows. In Section 2 we propose a model for the market price in terms of implied volatility, where the latter follows Cox- Ingersoll-Ross process. In Section 3 we describe optimal stopping domains in terms of implied volatility. Sections 4-6 focuse on discrete approximation Invited lecture. 2000 Mathematics Subject Classifications: 62P05, 65C50, 91B28. Key words and phrases. Arbitrage, Cox-Ingersoll-Ross process, European option re- selling, implied volatility, optimal stopping domain, option market price. 114 RESELLING OF EUROPEAN OPTION 115 of the proposed model and some properties of it are derived. The formula for penultimate stopping domain is derived in Section 7. Section 8 contains the main result that the proposed discrete model is arbitrage–free, and Section 9 concludes. 2. Model for option market price Consider the classical Black and Scholes market in continuous time [5]:{ St = S0e (μ−σ2 2 )t+σWt , t ≥ 0, Bt = B0e rt, t ≥ 0. (1) Here μ, σ, and r are positive parameters, S0 and B0 are positive and nonran- dom, Wt is Wiener process on the filtered probability space (Ω,F , (Ft)t≥0, P ). Consider a European call option with maturity T and pay-off function g(ST ) = (ST −K)+ = max{ST −K, 0}. We suppose that the Investor buys the option at fair price C0 = E∗ S0 e−rTg(ST ) =: f(S0, T ; σ, r), (2) Here E∗ is expectation w.r.t. the martingale measure P ∗, and E∗ S0 denotes the expectation (w.r.t. P ∗) provided S0 is the value of the stock price at t = 0. It is well known that under P ∗, μ = r holds. Now, suppose that the Investor can resell the option at any moment t ∈ [0, T ] for a certain market price Cm t . Naturally, we assume that Cm 0 = C0, Cm T = g(ST ). (3) The problem of the optimal reselling of the option is an optimization problem: Ψ(τ) := Ee−rτCm τ → max (4) in the class of all (Markov) stopping times τ ∈ [0, T ]. The maximizing time is called an optimal reselling time and we denote it by τopt. The ”fair” market price at moment t ∈ [0, T ] equals Ct = f(St, T − t; σ, r) := E∗[e−r(T−t)g(ST )|St]. (5) Corollary 2.3 from [2] states the following: Corollary 1. If an option price coincides with the Black-Scholes price, then: a) τopt = 0 if μ < r, b) τopt = T if μ > r, c) any stopping time is optimal if μ = r. 116 M. PUPASHENKO AND A. KUKUSH If Cm t = Ct for all t ∈ [0, T ] then Corollary 1 holds true, and the problem (4) has no practical sense. Therefore we choose a stochastic model for the market price. At any moment t ∈ [0, T ] an implied volatility σt is defined as a solution to the equation f(St, T − t; σt, r) = Cm t , σt > 0. (6) Under natural assumptions, see [2], the equation (6) has unique solution. Note that Cm 0 = C0 implies σ0 = σ. We model σt as a stochastic volatility. The model for σt as geometric Brownian motion presented in [2] is not appropriate in practical sense, be- cause of σt can considerably deviate from σ, as t grows. Instead we propose the model based on Cox-Ingersoll-Ross process, see [1]: dσ2 t = −α(σ2 t − σ2)dt+ β √ σ2 t dW ′ t , σ0 = σ, (7) where α, β > 0 and β2 ≤ 2ασ2, W ′ t is a Wiener process on (Ω,F , (F̃t)t≥0, P ), with a new filtration (F̃t)t≥0. Under imposed restrictions on α, β, and σ, the process (7) is well defined for all t ≥ 0 [1]. Also we assume the following: Wiener processes Wt and W ′ t are jointly Gaussian and positively correlated. (8) This condition can be understood as follows. If St grows, then so does σt, which makes Cm t go beyond the ”fair” price Ct = f(St, T − t; σ, r). This corresponds to Investor’s aim to hold an option if the stock price is growing. On the other hand, when the stock price drops, the Investor is willing to get rid of an option, which makes Cm t go below its ”fair” price. Let Gt = Ft ∨ F̃t. In the reselling model (1), (6), and (7), the optimal stopping time τopt is defined as a maximum point of Ψ(τ) in a class of all stopping times w.r.t. the filtration Gt. 3. Stopping sets Similarly to Section 5.1 from [2] we have τopt = inf{t ∈ [0, T ] : (St, C m t ) ∈ Gt}, (9) where nonrandom stopping sets are given by Gt = {(s, c)|s ≥ 0, c = ft(s, c)}, (10) the function ft(s, c) is a reward function, ft(s, c) := sup τ∈[t,T ] E[e−r(τ−t)Cm τ |St = s, Cm t = c], (11) RESELLING OF EUROPEAN OPTION 117 and the upper bound is taken over all Gt - stopping times valued in [t, T ]. Since ft is jointly continuous, Gt is a closed subset of [0,∞) × [0,∞). By definition, ft(0, c) = 0. It is helpful to rewrite (9)-(11) in terms of St and σt: τopt = inf{t ∈ [0, T ] : (St, σt) ∈ Ht}, (12) Ht := {(s, d)|s ≥ 0, d = ht(s, d)}, (13) ht(s, d) := sup τ∈[t,T ] E[e−r(τ−t)Cm τ |St = s, σt = d]. (14) Relations (9)-(11), as well as (12)-(14), are based on the next observa- tion. The problem (4) is a problem of optimal realization of an American type option on two correlated assets St and Cm t with pay-off function g(St, C m t ) = Cm t . (15) 4. Discrete approximation In order to construct ε−optimal strategies of Investor, see [3], we deal with discrete approximations. Divide [0, T ] into n parts, and let Δ = T/n. We approximate the model (1) by the discrete model, which is a famous approximation of the Black-Scholes market by the Cox-Ross-Rubinstein market: ⎧⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎩ Sn(tj+1) = Sn(tj)e σ √ Δδj , Btj = Btj−1 erΔ, j = 1, n, tj = j n T, where δj = δnj = ±1 - i.i.d. with distribution: P (δj = 1) = eμΔ−e−σ √ Δ eσ √ Δ−e−σ √ Δ =: pn, P (δj = −1) = 1 − pn =: qn, (16) Consider the reselling problem for the binomial market (S(tj), Btj , j = 0, n). The fair value of the European option with the pay-off from Section 2 equals [4]: C0n = E∗ S0 e−rTg(ST ) =: fn(S0, T ; σ, r). Here E∗ corresponds to the martingale measure P ∗ n , for which instead of (16) we have P ∗(δj = 1) = erΔ − e−σ √ Δ eσ √ Δ − e−σ √ Δ =: p∗n. The fair value at the moment t = k n T equals Ctn = E∗[e−r(T−t)g(ST )|St] =: fn(St, T − t; σ, r). Introduce a market option price Cm tn with Cm 0n = C0n, C m Tn = CTn = g(ST ). 118 M. PUPASHENKO AND A. KUKUSH For t ∈ { k n T, k = 0, n− 1}, the implied volatility σnt is defined as a solution to the equation Cm tn = fn(St, T − t; σnt, r). (17) We model the implied volatility as a stochastic volatility in such a way, that the process σn(t), which is a linear interpolant of {σnt, t = 0, T n , . . . , T}, converges to {σt, t ∈ [0, T ]} from (7) in the sense of the first two moments. The σn(t) is presented in more detail in Section 5. 5. Approximation of implied volatility In this section we derive the discrete approximation of the implied volatility described by model (7). Return back to (7). Let zt = σ2 t , t ≥ 0; z0 = σ2. (18) Then by (7) we have dzt = −α(zt − σ2)dt+ β √ ztdW ′ t , z0 = σ2. (19) The stochastic differential equation (19) describes the Cox-Ingersoll- Ross process. Its first two moments and covariance function are as follows, see [6]: Ezt = σ2, (20) V arzt = σ2β2 2α (1 − e−2αt), (21) Cov(zt, zs) = σ2β2 2α · e−α(s+t)(e2α(s∧t) − 1), (22) where s ∧ t := min(s, t). Based on (19) we propose an approximation scheme for zt. From (19) we have zt+h − zt = −α ∫ t+h t (zt − σ2)dt+ β ∫ t+h t √ ztdW ′ t . Then, zt+h − zt ≈ −α(zt − σ2)h+ β √ hztγth, γth ∼ N(0, 1). (23) We need an approximation with Bernoulli variables. Therefore instead of γth we use εth which equals ±1 with equal probabilities. Then, zt+h ≈ zt(1 − αh) + αhσ2 + β √ hztεth. (24) RESELLING OF EUROPEAN OPTION 119 Now we use the relation (24) for a uniform partition of [0, T ] with step Δ = T n . We write zj = zn( j n T ), j = 0, n. Here is the approximation scheme: zj+1 = zj(1 − αΔ) + αΔσ2 + β √ zjΔεnj, (25) where εnj = ±1 with equal probabilities, and {εnj, j = 0, n} is i.i.d. se- quence. Then we set zn(t) = zj , t ∈ [ j n T, j+1 n T ), j = 0, n− 1; zn(T ) = zn. Next we find the upper and lower bounds for all zj , j = 0, n. Lemma 1. For β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) < 1 α we have the lower bound for zj: zj > Z(L) := ( −β√Δ + √ β2Δ + 4α2σ2Δ2 2αΔ )2 , for all j = 0, n. (26) Proof. We prove by induction. For all Δ > 0, the base of induction z0 > Z(L) holds true, indeed z0 = σ2 = ( −β√Δ + β √ Δ + √ 4α2σ2Δ2 2αΔ )2 > Z(L). For fixed j ≤ n− 1 we assume that zj > Z(L) and want to prove that zj+1 > Z(L). For Δ > 0 we have that Z(L) from (26) satisfies Z(L)(1 − αΔ) + αΔσ2 − β √ Z(L)Δ = Z(L). (27) Now we will prove the next inequality for β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) : zj(1 − αΔ) + αΔσ2 − β √ Δzj > Z(L)(1 − αΔ) + αΔσ2 − β √ ΔZ(L). (28) We can rewrite (28) as follows: (zj − Z(L))(1 − αΔ) > β √ Δ( √ zj − √ Z(L)). Since zj > Z(L) and Δ < 1 α we have √ zj + √ Z(L) > β √ Δ (1 − αΔ) , √ zj + √ Z(L) > 2 √ Z(L), and we can prove that:√ Z(L) = −β√Δ+ √ β2Δ+4α2σ2Δ2 2αΔ ≥ β √ Δ 2(1−αΔ) ⇔ ⇔ (1 − αΔ) √ β2Δ + 4α2σ2Δ2 ≥ β √ Δ ⇔ ⇔ 2α(2ασ2 − β2) + αΔ(β2 − 8ασ2) + 4α4σ2Δ2 ≥ 0 ⇐ ⇐ 2α(2ασ2 − β2) + αΔ(β2 − 8ασ2) ≥ 0 ⇔ 0 < Δ ≤ 4ασ2−2β2 α(8ασ2−β2) . 120 M. PUPASHENKO AND A. KUKUSH Then by (27) and (28) we have for β2 < 2ασ2 and Δ < 4ασ2−2β2 α(8ασ2−β2) next relations: zj+1 ≥ zj(1 − αΔ) + αΔσ2 − β √ Δzj > > Z(L)(1 − αΔ) + αΔσ2 − β √ ΔZ(L) = Z(L). Lemma 1 is proven. Note that from Lemma 1, for β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) , it follows immediately that zj > 0 for all j = 0, n, and relation (25) is well defined. But it is easy to show that for the positivity of zj, the condition β2 < 2ασ2 can be subsided. Lemma 2. For Δ < 1 α we have the upper bound for zj: zj < Z(U) := ( β √ Δ + √ β2Δ + 4α2σ2Δ2 2αΔ )2 , for all j = 0, n. (29) Proof. We prove by induction. For all Δ > 0, the base of induction z0 < Z(U) holds true, indeed Z(U) = ( β √ Δ + √ β2Δ + 4α2σ2Δ2 2αΔ )2 > σ2 = z0. For fixed j ≤ n− 1 we assume that zj < Z(U) and want to prove that zj+1 < Z(U). For Δ > 0 we have that Z(U) from (29) satisfies Z(U)(1 − αΔ) + αΔσ2 + β √ ΔZ(U) = Z(U). (30) Now, we have the next inequalities for Δ < 1 α : zj+1 ≤ zj(1 − αΔ) + αΔσ2 + β √ Δzj < < Z(U)(1 − αΔ) + αΔσ2 + β √ ΔZ(U). (31) Then (30) and (31) imply zj+1 ≤ zj(1 − αΔ) + αΔσ2 + β √ Δzj < < Z(U)(1 − αΔ) + αΔσ2 + β √ ΔZ(U) = Z(U). Lemma 2 is proven. In the next two lemmas we prove the convergence zn(t) → zt, as n→ ∞ (32) RESELLING OF EUROPEAN OPTION 121 in the sense of convergence for the first two moments of finite-dimensional distributions. Denote uj = zj − σ2. From (25) we get uj+1 = uj(1 − αΔ) + β √ Δ(uj + σ2)εnj, u0 = 0. (33) Lemma 3. Convergence (32) holds true in the sense of the first moments, that is lim n→∞ Ezn(t) = Ezt. Proof. First we consider t = T . From (33) we have Euj+1 = Euj(1 − αΔ), Eu0 = 0, therefore Euj = 0, j = 0, n. (34) Then lim n→∞ Ezn = lim n→∞ E(un + σ2) = σ2 = EzT . Next, verify (32) for all t ∈ (0, T ). We have zn(t) = uλn+σ2, λn = [ t Δ ] ; λ→ t T , as n→ ∞. Then by (33) we have Euλn = 0, (35) and lim n→∞ Ezλn = lim n→∞ E(uλn + σ2) = σ2 = Ezt. Lemma 3 is proven. Lemma 4. Convergence (32) holds true in the sense of the second moments, that is lim n→∞ Cov(zn(s), zn(t)) = Cov(zs, zt). Proof. Similarly to Lemma 1, first we consider t = T . From (33) we get V ar(uj+1) = V ar(uj(1 − αΔ)) + β2V ar(εnj · √ uj + σ2)Δ. We have V ar(εnj √ uj + σ2) = V ar √ uj + σ2V arεnj + (E √ uj + σ2)2V arεnj + +(Eεnj) 2V ar √ uj + σ2 = V ar √ uj + σ2 + (E √ uj + σ2)2 = = E(uj + σ2) = σ2, and V ar(uj+1) = V ar(uj(1 − αΔ)) + β2σ2Δ, V ar(u0) = 0. 122 M. PUPASHENKO AND A. KUKUSH Then by induction we obtain the variance of uj: V ar(uj) = σ2β2Δ j−1∑ i=0 (1 − αΔ)2i = σ2β21 − (1 − αΔ)2j 2α− α2Δ , j = 1, n; V ar(u0) = 0, (36) moreover V ar(un) = σ2β21 − (1 − αΔ)2n 2α− α2Δ . (37) Now, (1 − αΔ)2n = (1 − αΔ) 1 αΔ ·2nαΔ = (1 − αΔ) 1 αΔ ·2αT → e−2αT , as n→ ∞. Then from (37) we have lim n→∞ V ar(zn) = lim n→∞ V ar(un) = σ2β2 2α (1 − e−2αT ) = V ar(zT ). Next, similarly to Lemma 1, verify (32) for all t ∈ (0, T ). We have zn(t) = uλn + σ2, λn = [ t Δ ] ; λ→ t T , as n→ ∞. By (36) we get V ar(uλn) = σ2β21 − (1 − αΔ)2λn 2α− α2Δ . (38) Next, (1 − αΔ)2λn = (1 − αΔ) 1 αΔ ·2αnλΔ = (1 − αΔ) 1 αΔ ·2αTλ → e−2αt, as n→ ∞. Then from (38) we have the desirable convergence of variances: lim n→∞ V ar(zλn) = lim n→∞ V ar(uλn) = σ2β2 2α (1 − e−2αt) = V ar(zt). Next we have to verify the convergence of the covariance functions. First we show that the covariance functions of uj , ui and zj , zi are equal. We assume j > i (the case j = i was considered earlier). We have Cov(zi, zj) = Cov(ui, uj), Cov(ui, uj) = E [ ui(uj−1(1 − αΔ) + β √ (uj−1 + σ2)Δεnj−1) ] = = (1 − αΔ)E(uiuj−1) + β √ ΔE( √ uj−1 + σ2εnj−1ui) = = (1 − αΔ)Cov(ui, uj−1) + β √ ΔEεnj−1 ·E( √ uj−1 + σ2ui) = = (1 − αΔ)Cov(ui, uj−1). RESELLING OF EUROPEAN OPTION 123 Then by induction for j ≥ i, using (36) we have: Cov(zi, zj) = (1−αΔ)j−iV ar(zi) = σ2β2(1−αΔ)i+j (1 − αΔ)−2i − 1 2α− α2Δ . (39) Similarly to the convergence of variances, we have the convergence of co- variance functions: Cov(uλ1n, un) → Cov(ut, uT ), Cov(uλ2n, uλ1n) → Cov(us, ut), as n→ ∞, where for all s, t such that 0 < s < t < T , λ1n = [ t Δ ], λ2n = [ s Δ ] and λ1 → t T , λ2 → s T , as n→ ∞. Lemma 4 is proven. 6. Correlation Now due to condition (8) we want to derive a correlation between εnj and δnj . Let EWtW ′ t = 2ρ1t, 0 < ρ1 < 1 2 . (40) Here 2ρ1 is the correlation coefficient between the two processes. Now, we introduce a correlation between εnj and δnj due to the table of joint probabilities: δn�εn -1 1 -1 (1 − pn) − 1−ρ 2 1−ρ 2 1 pn − ρ 2 ρ 2 Then Eεnj = 0, Eδnj = pn− (1− pn) = 2pn− 1, and Eδnj → 0, as n→ ∞. Next we find the covariance between εnj and δnj. E(εnj · δnj) = ( ρ 2 + (1 − pn) − 1 − ρ 2 ) − ( pn − ρ 2 + 1 − ρ 2 ) = 2ρ− 2pn, and we investigate its convergence: Cov(εnj, δnj) = E(εnj · δnj) − E(εnj)E(δnj) → 2ρ− 1, as n→ ∞. Then we set ρ = ρ1 + 1 2 which corresponds to (40). Moreover from (40) we have that 1 2 < ρ < 1. 124 M. PUPASHENKO AND A. KUKUSH 7. Structure of stopping domain in discrete time For fixed Δ, we have two sequences Sj = Sn(tj), j = 0, n, n = T Δ , (41) σj = σn(tj), j = 0, n, σ0 = σ, (42) where σ is historical volatility, which describes the behavior of the stock price St in continuous time. In discrete time we have τopt = τopt,n = min{tj : (σj , sj) ∈ Γj}, (43) where Γj ⊂ [0,∞) × [0,∞). Consider European call option with pay-off function from Section 2. If at the moment tj the relation eσ √ Δ(n−j) · Sj ≤ K holds true, then Sn(T ) ≤ K, and the market option price equals zero at moments t ≥ tj . Therefore we have the next relation: [0,∞) × [0, Ke−σ √ Δ(n−j)] ⊂ Γj , j ≤ n− 1. Due to Section 3 we can write the stopping sets as follows: Γj = { (v, s) ⊂ [0,∞) × [0,∞) : Cm tj ∣∣∣ Sj=s,σj=v ≥ (44) ≥ sup tj+1≤τ≤T E [ e−r(τ−tj) · Cm τ ∣∣Sj = s, σj = v ]} ∪ ∪ ( [0,∞) × [0, Ke−σ √ Δ(n−j)] ) , j ≤ n− 1; Γn = [0,∞) × [0,∞). We can simplify this formula for the case j = n− 1. Introduce a function p∗n(v) = erΔ − e−v √ Δ ev √ Δ − e−v √ Δ . (45) First, we find the LHS of inequality in (44): Cm tj ∣∣∣ Sj=s,σj=v = fn(s,Δ(n− j), v) = E∗ σj=v [ e−rΔ(n−j)g(Sn) ∣∣Sj = s ] = = e−rΔ(n−j) n−j∑ l=0 ( n− j l ) g(sev √ Δle−v √ Δ(n−j−l))(p∗n(v)) l(1 − p∗n(v)) n−j−l, and for j = n− 1 we get: Cm tn−1 ∣∣ Sn−1=s,σn−1=v = e−rΔ [ g(sev √ Δ)p∗n(v) + g(se−v √ Δ)(1 − p∗n(v)) ] . (46) RESELLING OF EUROPEAN OPTION 125 Then, we can rewrite the RHS of inequality in (44) for j = n− 1: E [ e−rΔCm T ∣∣Sn−1 = s, σn−1 = v ] = e−rΔE [ g(Sn) ∣∣Sn−1 = s, σn−1 = v ] = = e−rΔ [ g(seσ √ Δ)pn + g(se−σ √ Δ)(1 − pn) ] , (47) where pn is defined in (16). Then from relations (44), (46), and (47) we can write the formula for Γn−1: Γn−1 = { (v, s) ⊂ [0,∞) × [0,∞) : g(sev √ Δ)p∗n(v) + +g(se−v √ Δ)(1 − p∗n(v)) ≥ g(seσ √ Δ)pn + g(se−σ √ Δ)(1 − pn) } ∪ ∪ ( [0,∞) × [0, Ke−σ √ Δ] ) . (48) Now, based on (48) we plot some stopping domains (all plots are in the coordinate plane (v, s)). Parameters of our model are the following: K = 5, n = 100, α = 2, σ = 1, β = √ 2, ρ = 0.7, r = 0.1, and μ takes one of three values: 0.05, 0.1, or 0.2. Stopping sets Γn−1 for three relations between μ and r on the plane (v, s): 8. Absence of arbitrage In this section we consider Markov stopping times with discrete values {t0, t1, . . . , tn} = {0,Δ, . . . , nΔ}. We start with a standard definition of arbitrage. Definition 1. In a model (16), (17), (18), and (25) a stopping time τ provides an arbitrage possibility if: a) P (e−rτCm τ ≥ C0) = 1, 126 M. PUPASHENKO AND A. KUKUSH b) P (e−rτCm τ > C0) > 0. Lemma 5. For β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) we have the next inequality: σj+1∗(σj) < σj < σ∗ j (σj+1), for all j = 0, n− 1, (49) where σj+1∗(σj) := √ σ2 j (1 − αΔ) + αΔσ2 − βσj √ Δ, σ∗ j+1(σj) := √ σ2 j (1 − αΔ) + αΔσ2 + βσj √ Δ. Proof. First we prove that σj+1∗(σj) < σj . We have to verify that: σ2 j (1 − αΔ) + αΔσ2 − βσj √ Δ < σ2 j ⇔ αΔσ2 j + βσj √ Δ − αΔσ2 > 0. It is easy to solve the latter inequality: σj > −β√Δ + √ β2Δ + 4α2σ2Δ2 2αΔ , but this holds true for β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) due to Lemma 1. Next, we prove σ∗ j+1(σj) > σj . We must verify that: σ2 j (1 − αΔ) + αΔσ2 + βσj √ Δ > σ2 j ⇔ αΔσ2 j − βσj √ Δ − αΔσ2 < 0. It is easy to solve the last inequality: 0 < σj < β √ Δ + √ β2Δ + 4α2σ2Δ2 2αΔ , which holds true for Δ ≤ 4ασ2−2β2 α(8ασ2−β2) < 1 α due to Lemma 2. Lemma 5 is proven. Theorem 1. For β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) there is no arbitrage pos- sibility in the discrete model (16), (17), (18), and (25). Proof. We construct a martingale measure P ∗∗ = P ∗∗ n under which e−rtCm t is a martingale w.r.t. the filtration Gt (if such a measure exists then there is no arbitrage possibility), which means: E∗∗ ( e−rΔ(j+1)Cm Δ(j+1) ∣∣∣Gj) = e−rΔjCm Δj ⇔ ⇔ E∗∗ ( e−rΔ(j+1)Cm Δ(j+1) ∣∣∣Sj , σj) = e−rΔjCm Δj ⇔ ⇔ E∗∗ ( fn(Sj+1, T − Δ(j + 1); σj+1, r) ∣∣∣Sj, σj) = = fn(Sj, T − Δj; σj , r). (50) RESELLING OF EUROPEAN OPTION 127 Further in this section we write f j(Sj, σj) = fn(Sj , T − Δj; σj , r). We choose a measure P ∗∗ such that εnj, j = 0, 1, ..., n have the same distribution as under P ∗ n , and εnj and δnj are independent with P ∗∗(εnj = 1) = h(Sj), P ∗∗(εnj = −1) = 1 − h(Sj). Then by (25) and (49) we have E∗∗ ( f j+1(Sj+1, σj+1) ∣∣∣Sj, σj) = = E∗∗ ( f j+1(Sj+1, σ ∗ j+1(σj))h(Sj) + f j+1(Sj+1, σj+1∗(σj))(1 − h(Sj)) ∣∣∣Sj, σj) = = f j(Sj , σ ∗ j+1(σj))h(Sj) + f j(Sj , σj+1∗(σj))(1 − h(Sj)). Now we can rewrite (50): f j(Sj, σ ∗ j+1(σj))h(Sj) + f j(Sj, σj+1∗(σj))(1 − h(Sj)) = f j(Sj , σj). And we can choose h(Sj) as follows: h(Sj) = f j(Sj, σj) − f j(Sj, σj+1∗(σj)) f j(Sj, σ∗ j+1(σj)) − f j(Sj, σj+1∗(σj)) , (51) and 0 < h(Sj) < 1 for β2 < 2ασ2 and Δ ≤ 4ασ2−2β2 α(8ασ2−β2) by the strict mono- tonicity of f j in the second argument and by Lemma 5. Thus P ∗∗ is equivalent to P , and P ∗∗ is a martingale measure. Theorem 1 is proven. 9. Conclusion We considered the reselling problem for European option and proposed a stochastic model for the market option price. For this model, we con- structed the discrete approximation and proved the absence of arbitrage in it. Optimal strategy of the Investor in this discrete model is described by nonrandom stopping sets in the phase space of possible implied volatility and stock price. We derived the formula for penultimate stopping domain. The structure of this domain is illustrated by plots for some numerical values of parameters. References 1. Cox, J., Ingersoll, J., and Ross, S. (1985) A theory of the term structure of interest rates, Econometrica, 53, 385-407. 2. Kukush, A.G., Mishura, Yu.S., and Shevchenko, G.M. (2006) On reselling of European optoin, Theory Stoch. Process., 12(28), 75-87. 128 M. PUPASHENKO AND A. KUKUSH 3. Kukush, A.G., and Silvestrov, D.S. (1999) Optimal stopping strategies for American type options with discrete and continuous time, Theory Stoch. Process., 5(21), 71-79. 4. Shiryaev, A.N., Kabanov, Y.M., Kramkov, O.D., and Mel’nikov, A.V., (1994) Towards the theory of pricing of option of both European and Amer- ican types. I: Discrete time, Theory Probab. Appl., 39, no.1, 23-79. 5. Shiryaev, A.N., Kabanov, Y.M., Kramkov, O.D., and Mel’nikov, A.V., (1994) Towards the theory of pricing of option of both European and Amer- ican types. II: Continuous time, Theory Probab. Appl., 39, no.1, 80-129. 6. Swishchuk, A. (2004) Modeling of variance and volatility swaps for financial markets with stochastic volatility, WILMOTT Magazine September, Issue 2, 64-72. Department of Probability Theory and Mathematical Statistics, National Taras Shevchenko University of Kyiv, Volodymyrska st. 64, 01033 Kyiv, Ukraine. E-mail: myhailo.pupashenko@gmail.com Department of Mathematical Analysis, National Taras Shevchenko University of Kyiv, Volodymyrska st. 64, 01033 Kyiv, Ukraine. E-mail: alexander kukush@univ.kiev.ua
id nasplib_isofts_kiev_ua-123456789-4573
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0321-3900
language English
last_indexed 2025-12-07T17:17:50Z
publishDate 2008
publisher Інститут математики НАН України
record_format dspace
spelling Pupashenko, M.
Kukush, A.
2009-12-07T15:37:58Z
2009-12-07T15:37:58Z
2008
Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process / M. Pupashenko, A. Kukush // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 3-4. — С. 114-128. — Бібліогр.: 6 назв.— англ.
0321-3900
https://nasplib.isofts.kiev.ua/handle/123456789/4573
On Black and Scholes market Investor buys a European call option. At each moment of time till the maturity he is allowed to resell the option for the quoted market price. In Kukush et al. (2006) On reselling of European option, Theory Stoch. Process., 12(28), 75-87, a similar problem was investigated for another model of the market price. We propose a more realistic model based on Cox-Ingersoll-Ross process. Discrete approximation for this model is investigated, which is arbitrage–free. For this discrete model, a formula for penultimate optimal stopping domains is derived.
en
Інститут математики НАН України
Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
Article
published earlier
spellingShingle Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
Pupashenko, M.
Kukush, A.
title Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
title_full Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
title_fullStr Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
title_full_unstemmed Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
title_short Reselling of European option if the implied volatility varies as Cox-Ingersoll-Ross process
title_sort reselling of european option if the implied volatility varies as cox-ingersoll-ross process
url https://nasplib.isofts.kiev.ua/handle/123456789/4573
work_keys_str_mv AT pupashenkom resellingofeuropeanoptioniftheimpliedvolatilityvariesascoxingersollrossprocess
AT kukusha resellingofeuropeanoptioniftheimpliedvolatilityvariesascoxingersollrossprocess