Generalized stochastic target problems for pricing and partial hedging under loss constraints—application in optimal book liquidation

Bruno Bouchard; Ngoc-Minh Dang

doi:10.1007/s00780-012-0198-8

Finance and Stochastics

January 2013, Volume 17, Issue 1, pp 31–72 | Cite as

Generalized stochastic target problems for pricing and partial hedging under loss constraints—application in optimal book liquidation

Authors
Authors and affiliations

Bruno Bouchard
Ngoc-Minh Dang

Article

First Online: 26 September 2012

Received: 02 August 2010

Accepted: 14 March 2012

386 Downloads
12 Citations

Abstract

We consider a singular version with state constraints of the stochastic target problems studied in Soner and Touzi (SIAM J. Control Optim. 41:404–424, 2002; J. Eur. Math. Soc. 4:201–236, 2002) and more recently Bouchard et al. (SIAM J. Control Optim. 48:3123–3150, 2009), among others. This provides a general framework for the pricing of contingent claims under risk constraints. Our extended version perfectly fits the market models with proportional transaction costs and the order book liquidation issues. Our main result is a direct PDE characterization of the associated pricing function. As an example application, we discuss the valuation of VWAP-guaranteed-type book liquidation contracts, for a general class of risk functions.

Keywords

Stochastic target problems State constraints Pricing under risk constraint Book liquidation

Mathematics Subject Classification (2010)

49L25 60J60 91G80

JEL Classification

G11 G13 C61

This is a preview of subscription content, to check access.

Appendix: Proof of Theorem 2.8

For the reader’s convenience, we provide here the main elements of the proof of Theorem 2.8. It follows line by line the arguments in [10], which in turn are based on [24]. One difference is that we formulate our result in terms of a family of stopping times $\{\theta^{\phi},\;\phi\in \mathcal {A}\} \subset\mathcal {T}_{[t,T]}$ and not in terms of a fixed one θ. This does not change anything, except at the level of notation. Another one is that the assumption A1 (pasting property) in [24] is not “formally” satisfied by elements of $\mathcal{L}$. This is however easily handled.

Let t∈[0,T] be fixed. We must show that $V(t)=\bar{V}(t)$, where

$$\bar{V}(t):= \Big\{ z \in\mathbb{R}^{d+1}: \exists\phi\in\mathcal {A} \mbox{ s.t. } \big( Z^\phi_{t,z}(\tau\wedge{\theta^{\phi}}) \in O \bigoplus^{\tau,{\theta ^{\phi}}} V\ \mbox{for\ all}\ \tau\in\mathcal{T}_{[t,T]}\big)\Big\}. $$

In the following, we work on the d-dimensional canonical space endowed with the Wiener measure ℙ. The filtration is given by the augmentation of the raw filtration generated by W. The canonical path is denoted by ω=(ω _t)_t∈[0,T]. We use the notation ω ^s:=(ω _t∧s)_t∈[0,T] and δ _s ω:=(ω _t∨s−ω _s)_t∈[0,T], s∈[0,T].

We split the proof in several lemmas.

Lemma A.1

$V(t) \subset\bar{V}(t)$.

Proof

Fix z∈V(t) and $\phi\in \mathcal {A}$ such that $Z^{\phi}_{t,z}(s) \in O(s)$ for all s∈[t,T] ℙ-a.s. By the flow property, we have $Z^{\phi}_{{\theta^{\phi}},\xi}(s \vee {\theta^{\phi}}) \in O(s \vee{\theta^{\phi}})$ for all s∈[t,T] ℙ-a.s., where $\xi:=Z^{\phi}_{t,z}({\theta^{\phi}})$. Hence

Open image in new window

It follows that for ℙ-a.e. ω∈Ω,

Open image in new window

where $\tilde{\phi}_{\omega}$ is the map $\varOmega\ni\tilde{\omega}\mapsto\phi(\omega^{\theta(\omega)},\delta _{\theta(\omega)}\tilde{\omega})$, with ϕ viewed as a map on Ω. Using the right-continuity of $Z^{\tilde{\phi}_{\omega}}_{{\theta^{\phi }}(\omega),\xi(\omega)}$ and the Standing Assumption 3, we then deduce that for ℙ-a.e. ω∈Ω, $Z^{\tilde{\phi}_{\omega}}_{{\theta^{\phi}}(\omega),\xi(\omega)}(s \vee {\theta^{\phi}}(\omega) )(\tilde{\omega}) \in O(s\vee{\theta^{\phi }}(\omega) ) $ for all s∈[θ ^ϕ(ω),T] for ℙ-a.e. $\tilde{\omega}\in\varOmega$. This shows that $\xi=Z^{\phi}_{t,z}({\theta ^{\phi}})\in V({\theta^{\phi}})$ ℙ-a.s. Since we already know that $Z^{\phi}_{t,z}(\tau) \in O(\tau)$ for all $\tau\in\mathcal{T}_{[t,T]}$, this shows that $z\in\bar{V}(t)$. □

It remains to prove the opposite inclusion. In the following we equip $\mathcal{A}=\mathcal{U}\times\mathcal{L}$ with the distance

$$d\big((\nu,L),(\nu',L')\big):=\mathbb {E}\bigg[{\int_{0}^{T}\big|\nu_{s}-\nu '_{s}\big|^{2} \,ds + \sup_{s\le T }\big|L_{s}-L'_{s}\big|^{2}}\bigg ]^{\frac{1}{2}} $$

and denote by $\mathcal{B}_{\mathcal{A}}$ the induced Borel σ-field.

Lemma A.2

$\bar{V}(t) \subset V(t)$.

Proof

We now fix $z \in\bar{V}(t)$ and $\phi=(\nu,L)\in\mathcal {A}$ such that for all $\tau\in\mathcal{T}_{[t,T]}$,

Open image in new window

(A.1)

(1) We fix an arbitrary $\tau\in\mathcal{T}_{[t,T]}$ and first work on the event {θ ^ϕ<τ}. On this set, we have $Z^{\phi}_{t,z}({\theta^{\phi}}) \in V({\theta^{\phi}}) $ and therefore

$$\big({\theta^{\phi}},Z^\phi_{t,z}({\theta^{\phi}})\big) \in\mathcal {D}:= \big\{ (t,z) \in[0,T]\times \mathbb {R}^{d+1}: z \in V(t)\big\}. $$

Let $\mathcal{B}_{\mathcal{D}}$ denote the collection of Borel subsets of $\mathcal{D}$. Applying Lemma A.3 below to the measure induced by $({\theta ^{\phi}},Z^{\phi}_{t,z}({\theta^{\phi}}))$ on [0,T]×ℝ^d+1, we can construct a measurable map $\bar{\phi}:(\mathcal{D},\mathcal{B}_{\mathcal{D}}) \to(\mathcal {A},\mathcal{B}_{\mathcal{A}})$ such that

$$Z^{\bar{\phi}({\theta^{\phi}},Z^\phi_{t,z}({\theta^{\phi}}))}_{{\theta ^{\phi}},Z^\phi_{t,z}({\theta^{\phi}})}(\vartheta) \in O(\vartheta) \quad \mbox{for\ all}\ \vartheta\in\mathcal{T}_{[{\theta^{\phi}},T]}. $$

Since $\mathcal{A}$ is a separable metric space, we can then find a progressively measurable process ϕ ₁=(ν ¹,L ¹) such that $\phi_{1}=\bar{\phi}({\theta^{\phi}},Z^{\phi}_{t,z}({\theta^{\phi}}))$ on 〚θ ^ϕ,T〛 dt×dℙ-a.e.; see Lemma 2.1 in [24]. We then define $\hat{\phi}=(\hat{\nu},\hat{L})$ by

Open image in new window

Note that ν ¹, and therefore $\hat{\nu}$, need not be square-integrable, but that both are at least ℙ-a.s. square-integrable and take values in U. Similarly, $\hat{L}_{T}$ need not belong to L ²(Ω), but $\hat{L}$ is continuous and nondecreasing. Since the dynamics of $Z^{\phi_{1}}_{\theta^{\phi},Z^{\phi}_{t,z}({\theta^{\phi}})}$ only depends on the increments of L ¹ on [θ ^ϕ,T], the above defined control satisfies

$$Z^{\hat{\phi}}_{{\theta^{\phi}},Z^\phi_{t,z}({\theta^{\phi}})}(\tau) \in O(\tau) \quad\mbox{on}\ \{ {\theta^{\phi}} < \tau\}. $$

(2) Let $\hat{\phi}$ be defined as above and note that by (A.1), we also have

$$Z^{\hat{\phi}}_{t,z}(\tau)=Z^\phi_{t,z}(\tau) \in O \bigoplus^{\tau ,{\theta^{\phi}}} V=O(\tau) \quad \mathrm{on}\ \{ \tau\le{\theta ^{\phi}} \}. $$

(3) Combining the two above steps implies that

$$Z^{\hat{\phi}}_{t,z}(\tau) \in O (\tau). $$

Since $\tau\in\mathcal{T}_{[t,T]}$ is arbitrary, our Standing Assumption 2 implies that $Z^{\hat{\phi}}_{t,z}(s)\in O(s)$ for all s∈[t,T] ℙ-a.s. Although the control $\hat{\phi}$ might not satisfy the integrability condition imposed on elements of $\mathcal {A}$, we know from step (1) above that $\hat{\nu}$ is at least ℙ-a.s. square-integrable and that $\hat{L}$ is continuous and nondecreasing. Our Standing Assumption 4 thus allows one to conclude that $\mathcal {A}_{t,z}\ne\emptyset$, i.e., z∈V(t). □

It remains to prove the following result which was used in the previous proof.

Lemma A.3

For any probability measure μ on [0,T]×ℝ^d+1, there exists a Borel-measurable function $\bar{\phi}:(\mathcal{D},\mathcal{B}_{\mathcal{D}}) \to(\mathcal {A},\mathcal{B}_{\mathcal{A}})$ such that

$$\bar{\phi}(t,z) \in \mathcal {A}_{t,z} \quad\mbox{\textit{for}}\ \mu\mbox{-\textit{a.e.}}\ (t,z) \in\mathcal{D}. $$

Proof

Set $B:=\{ (t,z,\phi) \in[0,T]\times \mathbb {R}^{d +1} \times\mathcal {A} : \phi\in\mathcal{A}_{t,z} \}$. It follows from our Lipschitz-continuity assumptions that the map

$$(t,z,\phi) \in[0,T]\times \mathbb {R}^{d+1 } \times\mathcal{A} \to Z^\phi_{t,z}(r) \in L^{2}(\varOmega) $$

is continuous, and therefore Borel-measurable, for any r≤T. Then for any bounded continuous function f, the map $\psi^{r}_{f} : (t,z,\phi) \in[0,T]\times \mathbb {R}^{d+1 } \times\mathcal{A} \to \mathbb {E}[f(Z^{\phi}_{t,z}(r))]$ is Borel-measurable. Since O(r) is a Borel set, the map 1 _O(r) is the limit of a sequence of bounded continuous functions (f ⁿ)_n. Therefore $\psi^{r}_{\mathbf {1}_{O(r)}}=\lim_{n \to\infty}\psi^{r}_{f^{n}}$ is a Borel function. This implies that

Open image in new window

is a Borel set. Appealing to the Standing Assumption 3 and the right-continuity of $Z^{\phi}_{t,z}$, we then deduce that B=⋂_{r≤T, r∈ℚ} B ^r. This shows that B is a Borel set and therefore an analytic subset of $[0,T]\times \mathbb {R}^{d+1 } \times\mathcal{A}$; see [4].

Applying the Jankov–von Neumann theorem (see [4] Proposition 7.49), we then deduce that there exists an analytically measurable function $\bar{\phi}: \mathcal {D} \to\mathcal{A}$ such that

$$\bar{\phi}(t,z) \in\mathcal{A}_{t,z} \quad\mbox{for\ all}\ (t,z) \in \mathcal{D}. $$

Since an analytically measurable map is also universally measurable, the required result follows from Lemma 7.27 in [4]. □

References

1.
Almgren, R.: Optimal trading in a dynamic market. Technical report (2009). http://www.math.nyu.edu/financial_mathematics/content/02_financial/2009-2.pdf
2.
Almgren, R., Harts, B.: A dynamic algorithm for smart order routing. Technical report, white paper streambase (2008). http://www.streambase.com/wp-content/uploads/downloads/StreamBase_White_Paper_Smart_Order_Routing_low.pdf
3.
Almgren, R.F., Chriss, N.: Optimal execution of portfolio transactions. J. Risk 3, 5–39 (2000) Google Scholar
4.
Bertsekas, D.P., Shreve, S.E.: Stochastic Optimal Control: The Discrete Time Case. Mathematics in Science and Engineering. Academic Press, San Diego (1978) zbMATHGoogle Scholar
5.
Bertsimas, D., Lo, A.W., Hummel, P.: Optimal control of execution costs for portfolios. Comput. Sci. Eng. 1, 40–53 (1999) CrossRefGoogle Scholar
6.
Bouchard, B., Elie, R., Touzi, N.: Stochastic target problems with controlled loss. SIAM J. Control Optim. 48, 3123–3150 (2009) MathSciNetCrossRefGoogle Scholar
7.
Bouchard, B., Dang, N.M., Lehalle, C.A.: Optimal control of trading algorithms: a general impulse control approach. SIAM J. Financ. Math. 2, 404–438 (2011) MathSciNetzbMATHCrossRefGoogle Scholar
8.
Bouchard, B., Nutz, M.: 2011, Weak dynamic programming for generalized state constraints. Preprint. http://arxiv.org/abs/1105.0745
9.
Bouchard, B., Touzi, N.: Explicit solution of the multivariate super-replication problem under transaction costs. Ann. Appl. Probab. 10, 685–708 (2000) MathSciNetzbMATHGoogle Scholar
10.
Bouchard, B., Vu, T.N.: The American version of the geometric dynamic programming principle: application to the pricing of American options under constraints. Appl. Math. Optim. 61, 235–265 (2010) MathSciNetzbMATHCrossRefGoogle Scholar
11.
Crandall, M., Ishii, H., Lions, P.L.: User’s guide to viscosity solutions of second order partial differential equations. Am. Math. Soc. 27, 1–67 (1992) MathSciNetzbMATHCrossRefGoogle Scholar
12.
Cheridito, P., Soner, M., Touzi, N.: The multi-dimensional super-replication problem under gamma constraints. Ann. Inst. Henri Poincaré, Sér. C: Anal. Non-Linéaire 22, 633–666 (2005) MathSciNetzbMATHCrossRefGoogle Scholar
13.
Cvitanić, J., Pham, H., Touzi, N.: Super-replication in stochastic volatility models with portfolio constraints. J. Appl. Probab. 36, 523–545 (1999) MathSciNetzbMATHCrossRefGoogle Scholar
14.
Cvitanić, J., Pham, H., Touzi, N.: A closed-form solution to the problem of super-replication under transaction costs. Finance Stoch. 3, 35–54 (1999) zbMATHCrossRefGoogle Scholar
15.
Dupuis, P., Ishii, H.: SDEs with oblique reflection on nonsmooth domains. Ann. Probab. 21, 554–580 (1993) MathSciNetzbMATHCrossRefGoogle Scholar
16.
Föllmer, H., Leukert, P.: Quantile hedging. Finance Stoch. 3, 251–273 (1999) MathSciNetzbMATHCrossRefGoogle Scholar
17.
Föllmer, H., Leukert, P.: Efficient hedging: cost versus shortfall risk. Finance Stoch. 4, 117–146 (2000) MathSciNetzbMATHCrossRefGoogle Scholar
18.
Kabanov, Y.M.: Hedging and liquidation under transaction costs in currency markets. Finance Stoch. 3, 237–248 (1999) MathSciNetzbMATHCrossRefGoogle Scholar
19.
Pagès, G., Laruelle, S., Lehalle, C.A.: Optimal split of orders across liquidity pools: a stochastic algorithm approach. SIAM J. Financ. Math. 2, 1042–1076 (2011) zbMATHCrossRefGoogle Scholar
20.
Soner, H.M.: Optimal control with state-space constraint I. SIAM J. Control Optim. 24, 552–561 (1986) MathSciNetzbMATHCrossRefGoogle Scholar
21.
Soner, H.M.: Optimal control with state-space constraint II. SIAM J. Control Optim. 24, 1110–1122 (1986) MathSciNetzbMATHCrossRefGoogle Scholar
22.
Soner, H.M., Touzi, N.: Super-replication under gamma constraints. SIAM J. Control Optim. 39, 73–96 (2000) MathSciNetzbMATHCrossRefGoogle Scholar
23.
Soner, H.M., Touzi, N.: Stochastic target problems, dynamic programming and viscosity solutions. SIAM J. Control Optim. 41, 404–424 (2002) MathSciNetzbMATHCrossRefGoogle Scholar
24.
Soner, H.M., Touzi, N.: Dynamic programming for stochastic target problems and geometric flows. J. Eur. Math. Soc. 4, 201–236 (2002) MathSciNetzbMATHCrossRefGoogle Scholar
25.
Soner, H.M., Touzi, N.: The problem of super-replication under constraints. In: Carmona, R.A., Çinlar, E., Ekeland, I., Jouini, E., Scheinkman, J., Touzi, N. (eds.) Paris–Princeton Lectures on Mathematical Finance. Lecture Notes in Mathematics, vol. 1814, pp. 133–172. Springer, Berlin (2002) CrossRefGoogle Scholar
26.
Touzi, N.: Direct characterization of the value of super-replication under stochastic volatility and portfolio constraints. Stoch. Process. Appl. 88, 305–328 (2000) MathSciNetzbMATHCrossRefGoogle Scholar

Copyright information

Authors and Affiliations

Bruno Bouchard
- 1
- 2
Email author
Ngoc-Minh Dang
- 1
- 3

1.CEREMADEUniversité Paris DauphineParis cedex 16France
2.CRESTENSAE-ParisTechMalakoffFrance
3.CA CheuvreuxCourbevoieFrance

Generalized stochastic target problems for pricing and partial hedging under loss constraints—application in optimal book liquidation

Abstract

Keywords

Mathematics Subject Classification (2010)

JEL Classification

References

Copyright information

Authors and Affiliations

Personalised recommendations

Cite article