Abstract
We analyze mean-variance-optimal dynamic hedging strategies in oil futures for oil producers and consumers. In a model for the oil spot and futures market with Gaussian convenience yield curves and a stochastic market price of risk, we find analytical solutions for the optimal trading strategies. An implementation of our strategies in an out-of-sample test on market data shows that the hedging strategies improve long-term return-risk profiles of both the producer and the consumer.
Similar content being viewed by others
References
Bertus, M., Godbey, J., Hilliard, J.: Minimum variance cross hedging under mean-reverting spreads, stochastic convenience yields, and jumps: application to the airline industry. J. Futures Mark. 29, 736–756 (2009)
Caldentey, R., Haugh, M.: Optimal control and hedging of operations in the presence of financial markets. Math. Oper. Res. 31, 285–304 (2006)
Carmona, R., Ludkovski, M.: Spot convenience yield models for the energy markets. In: Yin, G., Zhang, Q. (eds.) Mathematics of Finance, vol. 351, pp. 65–80. AMS, Providence (2004)
Chen, S.-S., Lee, C.-f., Shrestha, K.: Futures hedge ratios: a review. Q. Rev. Econ. Finance 43, 433–465 (2003)
Černý, A., Kallsen, J.: On the structure of general mean-variance hedging strategies. Ann. Probab. 35, 1479–1531 (2007)
Černý, A., Kallsen, J.: Mean-variance hedging and optimal investment in Heston’s model with correlation. Math. Finance 18, 473–492 (2008)
Delbaen, F., Schachermayer, W.: The variance-optimal martingale measure for continuous processes. Bernoulli 2, 81–105 (1996). Amendments and corrections. Bernoullli 2, 379–380 (1996)
Delbaen, F., Schachermayer, W.: Attainable claims with p-th moments. Ann. Inst. Henri Poincaré 32, 743–763 (1996)
Duffie, D., Pan, J., Singleton, K.: Transform analysis and asset pricing for affine jump-diffusions. Econometrica 68, 1343–1376 (2000)
Fontana, C., Schweizer, M.: Simplified mean-variance portfolio optimisation. Math. Financ. Econ. 6, 125–152 (2012)
Föllmer, H., Schweizer, M.: Hedging of contingent claims under incomplete information. In: Davis, M.H.A., Elliott, R.J. (eds.) Applied Stochastic Analysis. Stochastics Monographs, vol. 5, pp. 389–414. Gordon and Breach, London (1991)
Geman, H.: Mean reversion versus random walk in oil and natural gas prices. In: Fu, M.C., Jarrow, R.A., Yen, J.-Y., Elliott, R.J. (eds.) Advances in Mathematical Finance, pp. 219–228. Birkhäuser, Boston (2007)
Gouriéroux, C., Laurent, J.-P., Pham, H.: Mean-variance hedging and numéraire. Math. Finance 8, 179–200 (1998)
Lien, D., Tse, Y.: Some recent developments in futures hedging. J. Econ. Surv. 16, 357–383 (2002)
Miltersen, K., Schwartz, E.: Pricing of options on commodity futures with stochastic term structures of convenience yields and interest rates. J. Financ. Quant. Anal. 33, 33–59 (1998)
Nascimento, J., Powell, W.: An optimal solution to a general dynamic jet fuel hedging problem (2008). Available at http://www.castlelab.princeton.edu/Papers/NascimentoPowell-JetFuelHedging.pdf
Pham, H.: Continuous-Time Stochastic Control and Optimization with Financial Applications. Springer, Berlin (2009)
Pham, H., Rheinländer, T., Schweizer, M.: Mean-variance hedging for continuous processes: new results and examples. Finance Stoch. 2, 173–198 (1998)
Platen, E.: On the role of the growth optimal portfolio in finance. Aust. Econ. Pap. 44, 365–388 (2005)
Rheinländer, T., Schweizer, M.: On L 2-projections on a space of stochastic integrals. Ann. Probab. 25, 1810–1831 (1997)
Schweizer, M.: On the minimal martingale measure and the Föllmer–Schweizer decomposition. Stoch. Anal. Appl. 13, 573–599 (1995)
Schweizer, M.: Approximation pricing and the variance-optimal martingale measure. Ann. Probab. 24, 206–236 (1996)
Schweizer, M.: A guided tour through quadratic hedging approaches. In: Jouini, E., Cvitanić, J., Musiela, M. (eds.) Option Pricing, Interest Rates and Risk Management, pp. 538–574. Cambridge University Press, Cambridge (2001)
Sin, C.: Complications with stochastic volatility models. Adv. Appl. Probab. 30, 256–268 (1998)
Acknowledgements
We should like to thank the two anonymous referees and the AE for their constructive criticism and many suggestions that improved the presentation of the paper. We should also like to thank Wolfgang Runggaldier for his valuable feedback on the estimation techniques.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
All proofs are given in this appendix. Sections A.1 and A.2 review some key results from the literature on affine models and on quadratic hedging that we use in this paper. Section A.3 contains the proofs of our main results.
1.1 A.1 Proof of Theorem 2.4
Assumption (2.7) implies that
with A(τ)=a′(τ) and B(τ)=b′(τ), and thus by (2.8)
for all T>t. Comparing with (2.4) we obtain σ ϵ (t,T)=B(T−t)η, and then (2.5) and (2.11) yield
By separating this into a sum of deterministic terms and linear terms in X t , we obtain
Since B(0)=b′(0)=1, we obtain B(T−t)=e −β(T−t) with β=κ+γ 1 η and
with ϵ 0=A(0) and ϵ 1=η(σρ−β 1). Integrating A and B then yields (2.12) and (2.13). Finally (2.14) follows by applying Itô’s formula to (2.2).
1.2 A.2 General results on the mean-variance hedging problem
In this appendix, we collect some technical background on the material in Sect. 3.1 and deduce the proof of Theorem 3.3 from the literature on the general mean-variance hedging problem. For a continuous semimartingale F on [0,T] with canonical decomposition F=F(0)+M+A, we write \(\mathbf {F}\in\mathcal {S}^{2}(P)\) if
Recall the set \(\mathcal{A}\) of admissible strategies in Definition 3.1, the set \(\mathcal{M}\) of measures, and the notation G(π)=∫π⋅d F. Various authors have worked with a different set Θ of admissible strategies. Define
and note that the set \(\mathcal{A}\) corresponds to the set \(\bar{\varTheta}\) in Černý and Kallsen [5].
Theorem A.1
-
(a)
We have
-
(b)
\(\varTheta\subset\mathcal{A}\), and \(G_{T}(\mathcal{A})\) is the closure of G T (Θ) in L 2(P).
Part (a) is Theorem 2.8 in Černý and Kallsen [6], where the inclusion ‘‘⊇’’ follows from Theorems 1.2 and 2.2 in Delbaen and Schachermayer [8], and part (b) is Corollary 2.9 part 1 in Černý and Kallsen [5]. In particular, (a) says that the set of admissible strategies \(\mathcal{A}\) coincides with the set of strategies used in Gouriéroux et al. [13]. Part (b) in particular says that \(G_{T}(\mathcal{A})\) is closed in L 2(P).
Proof of Theorem 3.3
(a) This result is obtained as a special case of Theorem 4.10 in Černý and Kallsen [5] as follows. Let L denote the opportunity process, \(\tilde{a}\) the adjustment process, and Q ∗ the variance-optimal signed martingale measure in the sense of Definitions 3.3, 3.8 and 3.12 in [5]. Since F is continuous, Q ∗ is equal to the variance-optimal martingale measure \(\tilde{P}\) by Theorem 1.3 in [7]. By (3.16) in [5], we have
By Theorem 3.25, equation (3.33) in [5], we have \(\tilde {a}\mathcal{E}( - \int\tilde{a}\cdot d\mathbf{F}) \in\mathcal{A}\), so the process \(\mathcal{E}( - \int\tilde{a}\cdot d\mathbf{F})\) is a \(\tilde{P}\)-martingale by Theorem A.1(a). So the process \(\tilde{Z}\) defined in (3.8) satisfies
Applying Itô’s formula in the last equation and comparing with (3.9), we obtain \(\tilde{a}_{t} = - \frac{\tilde{\zeta }_{t}}{\tilde{Z}_{t}}\). Furthermore, V(λ) is the mean-value process in (4.2) of [5]. Since F and V(λ) are continuous, the predictable covariation does not depend on the probability measure, and therefore \(\tilde{\xi}\) is the pure hedge coefficient in Definition 4.6 of [5] by equation (4.8) of [5]. So (3.11) is equivalent to equation (4.14) in [5] and thus the assertion follows from Theorem 4.10 in [5].
(b)–(d) The minimal value in (3.6) is given by
with \(R = E [ ( \tilde{Z}_{T} \int_{0}^{T} \frac {1}{\tilde{Z}_{s}} \,dL_{s} )^{2} ]\). This follows from rewriting (5.3) in [13] under \(\tilde{P}\) and P, and using the relation (4.13) in [20] between the Galtchouk–Kunita–Watanabe decompositions of the discounted cashflows under the measures \(\tilde{P}\) and \(\tilde{R}\), respectively. Next for each m∈ℝ define
The same proof as for Proposition 6.6.5 in [17] shows that
and if λ m is a maximizer in (A.3), the process π(λ m ) in (3.11) is an optimal control for B(m) in (A.2). Using (A.1), straightforward calculations yield the maximizer \(\lambda_{m} = \frac{\tilde{Z}_{0} m - \mathcal{X}_{0} - \tilde {E}[H_{T}]}{\tilde{Z}_{0} - 1} \) and the value
This yields (c). For given v≥R, (d) follows from setting B(m)=v and solving the quadratic equation for its largest root. Finally, by definition of U(a) and B(m) we have
and by (A.4) the maximum in (A.5) is attained at \(m^{*} = \frac{\tilde{Z}_{0} - 1}{2a} + \mathcal{X}_{0} + \tilde {E}[H_{T}]\). Moreover the optimal solution to (3.3) is now given by the optimal control to B(m ∗), which by the above observation is \(\pi (\lambda_{m^{*}})\). Combining the formulas for λ m and m ∗ yields (b). □
1.3 A.3 Proofs of Theorems 3.5–3.8
We resume the setup and definitions in Theorem 3.5. We start with
Lemma A.2
Let x(⋅),y(⋅),z(⋅) be differentiable functions which satisfy x(0)=y(0)=z(0)=0. The process
for t∈[0,T] is a P-local martingale if and only if x(⋅),y(⋅),z(⋅) are a finite solution to the ODE system
on [0,T]. In this case the process \(\hat{Z}:= \frac{Y }{Z }\) satisfies \(\hat{Z}_{T} = Z_{T}\) and
where \(\hat{W}_{t} = W_{t} + \int_{0}^{t} \hat{\theta}_{u} du\) and \(\psi_{t} = ( \hat{\theta}_{t}^{0} , \, \hat{\theta}_{t}^{1} - ( \eta y(T-t) + 2 \eta z(T-t) X_{t} ), \, 0 )\).
Proof
Recall from (2.11) that \(\hat{\theta}_{t} = ( \beta _{0} + \gamma_{0} X_{t}, \beta_{1} + \gamma_{1} X_{t}, 0 )\). From (2.8) we have
Applying Itô’s formula to (A.6),
Using κ=β−γ 1 η and writing the drift in (A.11) as a quadratic function in X t with deterministic coefficients, we find that the drift vanishes (that is, Y is a P-local martingale) if and only if (A.7), (A.9) hold true. Finally, we note that Itô’s formula, (A.11) and \(dZ_{t} = - Z_{t} \hat{\theta}_{t} \cdot dW_{t}\) imply
which gives (A.10). □
The solution of the Ricatti equation system (A.7)–(A.9) can be expressed in closed form.
Lemma A.3
For constant coefficients a,b,c,f,h,k∈ℝ with a,c>0, define \(d=\sqrt{b^{2} - 4ac}\) and \(g=\frac{d+b}{d-b}\). Then the ODE system
has the solution
with
The above formulas are to be understood as their analytic continuation if d=0. The solution exists on the open interval [0,T max) with
For \(b \in(-\sqrt{4ac},\sqrt{4ac}]\), the function \(b \mapsto\frac {1}{\sqrt{b^{2} - 4ac}} \log\frac{b+\sqrt{b^{2} - 4ac}}{b-\sqrt{b^{2} - 4ac}}\) is to be understood as its analytic continuation out of the domain \(( \sqrt{4ac}, \infty)\).
Proof
The solution formulas are verified by lengthy but straightforward computations. The time horizon \(T_{\rm max}\) is determined by the smallest positive zero of the function τ↦1+ge −dτ. □
Proof of Theorem 3.5
Define Y as in (A.6) with (A.7), (A.9). Since z(T−t)>0 for all t<T, there exists a constant c>0 such that
for t<T and thus \(Y_{t} \geq c Z_{t}^{2}\) for all t∈[0,T]. Since Y is a continuous process, the stopping times
satisfy τ n ↗T for n→∞, and the processes \(Y^{\tau_{n}}\) and \(Z^{\tau_{n}}\) are bounded P-martingales by Lemma A.2 and \(Z_{t} \leq\sqrt{\frac{1}{c} Y_{t}}\). Applying Doob’s inequality to \(Z^{\tau_{n}}\), we find
for some constant c 2>0. Letting n→∞ and applying monotone convergence in the last inequality, we obtain \(E [ \sup_{0 \leq t \leq T} Z_{t}^{2} ] \leq\frac{c_{2}}{c} Y_{0} < \infty\). So \(\hat{P}\in\mathcal {M}\). □
For the proof of Theorem 3.6 we need the following result.
Lemma A.4
Let W be a d-dimensional Brownian motion on some filtered probability space \((\varOmega,\mathcal{F},\mathbb{F},P)\), and a an ℝd-valued and b,σ,ν ℝd×d-valued deterministic functions. Let V be an ℝd-valued and S an ℝ-valued adapted process satisfying S 0>0 and
Then S is a martingale.
Proof
The proof follows the ideas in Sin [24]. S is a positive local martingale and hence a supermartingale, so it suffices to show that E[S T ]=S 0 for each T>0. Define the stopping times
Since ν⋅V is a locally bounded process, we have τ n ↗∞ P-a.s. for n→∞. Moreover, the stopped process \(S^{\tau_{n}} \) is a martingale by Novikov’s condition. Hence we can define a probability measure P n≈P by \(\frac{dP^{n}}{dP} = \frac {S_{T}^{\tau_{n}}}{S_{0}^{\tau_{n}}}\). Then the process
is a d-dimensional P n-Brownian motion by Girsanov’s theorem, and V satisfies
Now define a process \(\hat{V} \) by \(\hat{V}_{0} = V_{0}\) and
and a sequence of stopping times \(\hat{\tau}_{n}\) by
Then the distribution of τ n under P n is the same as the distribution of \(\hat{\tau}_{n}\) under P. Moreover, \(\hat{\tau}_{n} \nearrow\infty\) P-a.s. for n→∞ since \(\nu_{t} \cdot\hat{V}_{t}\) is locally bounded. Monotone convergence therefore yields
This finishes the proof. □
Applying Lemma A.4 to V=X immediately yields
Corollary A.5
Let x(⋅),y(⋅),z(⋅) be a solution to (A.7), (A.9) which satisfies x(0)=y(0)=z(0)=0. Then the process \(\hat{Z}= \frac{Y }{Z }\) in Lemma A.2 is a \(\hat{P}\)-martingale.
Proof of Theorem 3.6
(a) We start by noting that \(\hat{P}\) is a signed Θ-martingale measure in the sense of Sect. 1 of [22]. Indeed, since \(\hat{P}\in \mathcal{M}\), the process G(π) is a \(\hat{P}\)-martingale for each π∈Θ by Theorem A.1; hence ZG(π) is a P-martingale, and thus \(E \big[ \frac{d\hat{P}}{dP} G_{T}(\pi) \big] = 0\). To show that \(\hat{P}\) is the variance-optimal measure, by Lemma 1 (c) in [22] it now suffices to show that
where M 0∈[1,∞) and J T is in the L 2(P)-closure of G T (Θ), that is, in \(G_{T}(\mathcal{A})\) by Theorem A.1 b). To prove this we proceed in three steps.
Step 1. Let \(\mathbb{G}= (\mathcal{G}_{t})_{t \in[0,T]}\) be the filtration \(\mathcal{G}_{t} := \sigma( (\hat{W}_{s}^{0},\hat{W}_{s}^{1}) \, |\, s \leq t )\) generated by the 2-dimensional \(\hat{P}\)-Brownian motion \((\hat{W}^{0},\hat{W}^{1})\) and define the \(\mathbb{G}\)-stopping times
Since X is continuous, we have τ k ↗T a.s. for k→∞. Next define the processes
Since Z T and \(Z_{\tau_{k}}\) are \(\mathcal{G}_{T}\)-measurable and \(\mathcal{F} _{t} = \mathcal{G}_{t} \vee\sigma(W_{s}^{2} \, |\, s \leq t )\) with \(W^{2} = \hat{W}^{2}\) independent of \(\mathbb{G}\), we obtain \(M_{t} = \hat{E}[ Z_{T} \, |\, \mathcal{G}_{t} ]\) and \(M_{t}^{(k)} = \hat{E}[ Z_{\tau_{k}} \, |\, \mathcal{G}_{t} ]\), and hence by Itô’s representation theorem
for some predictable processes h=(h 0,h 1,0) and h (k)=g(h (k,0),h (k,1),0). Setting \(J_{T} = \int _{0}^{T} h_{s} \cdot d \hat{W}_{s}\), we obtain (A.12) with \(M_{0} = E [ Z_{T}^{2} ] \geq E [ Z_{T} ]^{2} = 1\).
Step 2. It remains to show that J T is in \(G_{T }(\mathcal{A})\). To this end recall that Z is a square-integrable P-martingale by Theorem 3.5, so dominated convergence and Doob’s inequality imply that
for k→∞, and therefore
Since \(G_{T}(\mathcal{A})\) is closed in L 2(P), it thus suffices to show that \(\int_{0}^{T} h_{s}^{(k)} \cdot d \hat{W}_{s} \in G_{T }(\mathcal {A})\) for each k. To verify this, first note that the nonsingularity of the volatility matrix of F allows us to write
for a suitable predictable and F-integrable process ζ (k). By Theorem A.1(a), the assertion now follows once we show that G(ζ (k)) is a Q-martingale for each \(Q \in \mathcal{M}\).
Step 3. Fix k∈ℕ and \(Q \in\mathcal{M}\). Clearly G(ζ (k)) is a Q-local martingale. To show the martingale property under Q, we start by computing
where \(\hat{\theta}_{t}^{(k)} = \hat{\theta}_{t} I_{ \{ t < \tau_{k} \} }\) is a process bounded by some constant c k depending on k and the model parameters, \(B_{t} = e^{ \int_{0}^{t} \| \hat{\theta}_{s}^{(k)} \|^{2} \,ds }\), and \(N = \mathcal{E}( - \int\hat{\theta}^{(k)} \cdot d\hat{W}) \) is a \(\hat{P} \)-martingale by Novikov’s condition. Hence
It follows that sup t∈[0,T]|G t (ζ (k))|∈L 2(P) by Theorem 3.5 and Doob’s inequality. Hence
by the Cauchy–Schwarz inequality, and so G(ζ (k)) is a Q-martingale.
(b) By (a) we have \(\tilde{P}= \hat{P}\) and thus \(\frac{d\tilde{P}}{dP} = Z_{T} = \hat{Z}_{T}\). Since \(\hat{Z} \) is a \(\hat{P}\)-martingale by Corollary A.5, it follows that
Equation (3.14) now follows from Lemma A.2. For (3.15), note that (2.14) and \(d\hat{W}_{t} = dW_{t} + \hat{\theta}_{t}\, dt\) imply
and
Plugging this into (3.14) yields (3.15). Finally \(( \tilde{\zeta} ^{i}, \zeta^{h} ) \in\mathcal{A}\) follows from using the uniqueness of the VOMM and the representations (3.9) and (3.15). □
Proof of Theorem 3.7
We give the proof under the assumption ϕ>β>0, which is satisfied for the parameter estimates we find in our calibration procedure. The result can be easily extended to general parameter values of ϕ and β.
(a) By definition of the spot-futures spread in (2.1) with T 1=T 1(t), we have
From (3.16) we then compute
Fix u∈[0,T] and T j . We claim that
for deterministic functions m j (τ),n 1(τ),n 2(τ) such that m j (0)=n 2(0)=0 and n 1(0)=1. Indeed, Itô’s formula for \(M_{t}^{j}(u) := F (t,T_{j} ) e^{m_{j}(u-t) + n_{1}(u-t) Y_{t} + n_{2}(u-t) X_{t}}\) and using
from (2.14), (2.8) and (2.15), gives, writing n 1=n 1(u−t), n 1=n 1(u−t) and m j =m j (u−t),
Hence the drift of M j(u) is zero if m j ,n 1,n 2 satisfy the ODE system
with m j (0)=n 2(0)=0 and n 1(0)=1, and lengthy but straightforward calculations show that the solution to this system is given by
where \(\alpha= \frac{\nu(\gamma_{0} c_{0} + \gamma_{1} c_{1})}{\phi- \beta}\) and
In this case, M j(u) is a \(\tilde{P}\)-local martingale, and since the diffusion coefficient is of the form \(M_{t}^{j}(u) c(t)\) with a (deterministic) bounded function c(t), the process M j(u) is a \(\tilde{P} \)-martingale by Novikov’s condition. Now (A.15) follows from \(M_{u}^{j}(u) = F (u, T_{j} ) e^{Y_{u}}\). Together with (A.14) we obtain (3.17).
(b) Plugging \(M_{t}^{j}(u) = F (t,T_{j} ) q_{j}(t,u)\) into (A.17) and using (A.16), we obtain
Moreover by (3.17) we have
Applying Itô’s formula and using that V(λ) and M j(u) are \(\tilde{P}\)-martingales gives
Plugging in (A.21) here, we obtain
with a \(\tilde{P}\)-local martingale L orthogonal to F. Plugging (A.13) into the last equation yields (3.18). □
Proof of Theorem 3.8
The structure of the proof is analogous to the proof of Theorem 3.7(a), so we only give a sketch. As in (A.14) we obtain
so (3.21) follows once we show that for all t≤u,
with deterministic functions p,ℓ j ,s j satisfying p(0)=1 and s j (0)=ℓ j (0)=0. To this end, we apply Itô’s formula to the RHS of (A.22), use (2.14), (2.8) and (2.15), and as in the proof of Theorem 3.7(a), we find that the RHS of (A.22) is a P-local martingale, and then indeed a martingale, if and only if the functions p,s j ,ℓ j fulfill a system of ODEs. This system can be solved explicitly, and lengthy but straightforward computations yield
where \(\alpha= \frac{\eta\gamma_{1}}{\beta\kappa} - \frac{\sigma }{\kappa} ( \rho_{0} \gamma_{0} + \rho_{1} \gamma_{1} )\) and
To verify (3.22), similarly as above we compute
Hence the assertion follows once we show
for all u,v, and it suffices to establish (A.26) for u≥v by symmetry of the function q ij (u,v) in u and v. So let u≥v. We note that by (A.22) we have
and thus (A.26) follows once we prove for all t∈[0,v]
with deterministic functions w,w ij ,m ij satisfying the equations m ij (0)=0, w(0)=1+p(u−v), and w ij (0)=s i (u−v). To this end, we proceed as above. We apply Itô’s formula to the RHS of (A.27), use (2.14), (2.8) and (2.15), and as in the proof of Theorem 3.7(a), we find that the RHS of (A.27) is a P-local martingale, and then indeed a martingale, if and only if the functions w,w ij ,m ij fulfil a system of ODEs. This system can be solved explicitly, and lengthy but straightforward computations yield that
where p(⋅) and s i (⋅) are defined in (A.23), (A.24), \(\alpha= \frac{\eta\gamma_{1}}{\beta\kappa} - \frac {\sigma }{\kappa} ( \rho_{0} \gamma_{0} + \rho_{1} \gamma_{1} )\), and
This finishes the proof. □
Rights and permissions
About this article
Cite this article
Wang, L., Wissel, J. Mean-variance hedging with oil futures. Finance Stoch 17, 641–683 (2013). https://doi.org/10.1007/s00780-013-0203-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00780-013-0203-x