Abstract
We study the portfolio optimization problem of maximizing the outperformance probability over a random benchmark through dynamic trading with a fixed initial capital. Under a general incomplete market framework, this stochastic control problem can be formulated as a composite pure hypothesis testing problem. We analyze the connection between this pure testing problem and its randomized counterpart, and from the latter we derive a dual representation for the maximal outperformance probability. Moreover, in a complete market setting, we provide a closed-form solution to the problem of beating a leveraged exchange traded fund. For a general benchmark under an incomplete stochastic factor model, we provide the Hamilton–Jacobi–Bellman PDE characterization for the maximal outperformance probability.
References
Avellaneda, M., Zhang, S.: Path-dependence of leveraged ETF returns. SIAM J. Financ. Math. 1, 586–603 (2010)
Bayraktar, E., Song, Q., Yang, J.: On the continuity of stochastic control problems on bounded domains. Stoch. Anal. Appl. 29, 48–60 (2011)
Bayraktar, E., Huang, Y.-J., Song, Q.: Outperforming the market portfolio with a given probability. Ann. Appl. Probab. 22, 1465–1494 (2012)
Brannath, W., Schachermayer, W.: A bipolar theorem for \(L^{0}_{+}(\varOmega,\mathcal{F},\bold P)\). In: Séminaire de Probabilités, XXXIII. Lecture Notes in Math., vol. 1709, pp. 349–354. Springer, Berlin (1999)
Browne, S.: Reaching goals by a deadline: digital options and continuous time active portfolio management. Adv. Appl. Probab. 31, 551–577 (1999)
Cvitanić, J.: Minimizing expected loss of hedging in incomplete and constrained markets. SIAM J. Control Optim. 38, 1050–1066 (2000)
Cvitanić, J., Karatzas, I.: Generalized Neyman–Pearson lemma via convex duality. Bernoulli 7, 79–97 (2001)
Delbaen, F., Schachermayer, W.: A general version of the fundamental theorem of asset pricing. Math. Ann. 300, 463–520 (1994)
Delbaen, F., Schachermayer, W.: The Mathematics of Arbitrage. Springer, Berlin (2006)
El Karoui, N.E., Quenez, M.-C.: Dynamic programming and pricing of contingent claims in an incomplete market. SIAM J. Control Optim. 33, 29–66 (1995)
Föllmer, H., Leukert, P.: Quantile hedging. Finance Stoch. 3, 251–273 (1999)
Föllmer, H., Schied, A.: Convex measures of risk and trading constraints. Finance Stoch. 6, 429–447 (2002)
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 (1990)
Giga, Y., Goto, S., Ishii, H., Sato, M.-H.: Comparison principle and convexity preserving properties for singular degenerate parabolic equations on unbounded domains. Indiana Univ. Math. J. 40, 443–470 (1991)
Jarrow, R.: Understanding the risk of leveraged ETFs. Finance Res. Lett. 7, 135–139 (2010)
Jeanblanc, M., Yor, M., Chesney, M.: Mathematical Methods for Financial Markets. Springer, London (2009)
Krutchenko, R.N., Melnikov, A.V.: Quantile hedging for a jump-diffusion financial market model. In: Kohlmann, M., Tang, S. (eds.) Trends in Mathematics: Workshop of the Mathematical Finance Research Project, Konstanz, Germany, 5–7 October 2000, pp. 215–229. Birkhäuser, Basel (2001)
Lehmann, E.L., Romano, J.P.: Testing Statistical Hypotheses, 3rd edn. Springer, New York (2005)
Leung, T., Song, Q., Yang, J.: Generalized hypothesis testing and maximizing the success probability in financial markets. In: Proceedings of the International Conference on Business Intelligence and Financial Engineering (ICBIFE) (2011). Available at http://ssrn.com/abstract=2292771
Ma, J., Yong, J.: Forward-Backward Stochastic Differential Equations and Their Applications. Lecture Notes in Mathematics, vol. 1702. Springer, Berlin (1999)
Ma, J., Zhang, J.: Representation theorems for backward stochastic differential equations. Ann. Appl. Probab. 12, 1390–1418 (2002)
MathOverflow, A non-degenerate martingale. Website (version: 2011–12–24). http://mathoverflow.net/questions/84216
Romano, M., Touzi, N.: Contingent claims and market completeness in a stochastic volatility model. Math. Finance 7, 399–410 (1997)
Rudloff, B.: Convex hedging in incomplete markets. Appl. Math. Finance 14, 437–452 (2007)
Rudloff, B., Karatzas, I.: Testing composite hypotheses via convex duality. Bernoulli 16, 1224–1239 (2010)
Schied, A.: On the Neyman–Pearson problem for law-invariant risk measures and robust utility functionals. Ann. Appl. Probab. 14, 1398–1423 (2004)
Schied, A.: Optimal investments for robust utility functionals in complete market models. Math. Oper. Res. 30, 750–764 (2005)
Sekine, J.: On a robustness of quantile-hedging: complete market’s case. Asia-Pac. Financ. Mark. 6, 195–201 (1999)
Sircar, R., Zariphopoulou, T.: Bounds and asymptotic approximations for utility prices when volatility is random. SIAM J. Control Optim. 43, 1328–1353 (2005)
Spivak, G., Cvitanić, J.: Maximizing the probability of a perfect hedge. Ann. Appl. Probab. 9, 1303–1328 (1999)
Yong, J., Zhou, X.-Y.: Stochastic Controls: Hamiltonian Systems and HJB Equations. Springer, New York (1999)
Acknowledgements
The authors would like to thank the Editor and two anonymous referees for their insightful remarks, as well as Jun Sekine, Birgit Rudloff, and James Martin for their helpful discussions.
Tim Leung’s work is partially supported by NSF grant DMS-0908295. Qingshuo Song’s work is partially supported by SRG grant 7002818 and GRF grant CityU 103310 of Hong Kong.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
1.1 A.1 The role of \(\operatorname {co}(\mathcal{H})\) in V(x)
In this example, we show that the representation of V(x) in (2.8) does not hold if \(\operatorname {co}(\mathcal{H})\) is replaced by the smaller set \(\mathcal{H}\).
Example A.1
Let Ω=[0,1], and let \(\mathbb{P}\) be Lebesgue measure, i.e., \(\mathbb{P}[(a,b)] = b-a\) for a≤b. Let \(\mathcal{G} = \{G \equiv1\}\) and \(\mathcal{H} = \{H_{1}, H_{2}\}\) with
For the randomized hypothesis testing problem (2.3) with x=1, it is easy to see, e.g., from (2.8), that
along with the optimizers
In this simple example, the uniqueness follows immediately.
Now, if one switches from \(\operatorname {co}(\mathcal{H})\) to \(\mathcal{H}\) in (2.8), then a strictly larger value will result; in fact,
1.2 A.2 On the positivity of \(\hat{a}\)
Theorem 2.3 shows that the minimizer \(\hat{a}\) takes values in [0,∞) rather than (0,∞) as claimed in Proposition 3.1 and Lemma 4.3 in Cvitanić and Karatzas [7]. To illustrate this issue, we first give an example where \(\hat{a}\) takes the value zero. Then we provide a sufficient condition for \(\hat{a}>0\).
Example A.2
Let \(N_{g} := \bigcap_{G\in\mathcal{G}} \{G=0\}\) and x>0.
-
(i)
If \(\mathbb{P}[N_{g}]=1\), then \(\mathbb{E} [(G- aH)^{+}] = 0\) for all G,H,a. Thus, \(\hat{a} = 0\) is the unique minimizer of \(\{xa + \inf_{\mathcal{G} \times\mathcal{H}} \mathbb{E} [(G- a H)^{+}]\}\).
-
(ii)
If \(0<\mathbb{P}[N_{g}] <1\) and \(x > \sup_{H\in\mathcal{H}} \mathbb{E} [(H I_{N_{g}^{c}})]\), then there also exists a counterexample such that \(\hat{a} = 0 \) minimizes \(\{xa + \inf_{\mathcal{G} \times\mathcal{H}} \mathbb{E} [(G- a H)^{+}]\}\). Let us consider the following scenario. The sample space is Ω=[0,1], \(\mathbb{P}\) is Lebesgue measure on [0,1], \(\mathcal{G} = \{G\}\) with G=2I [1/2,1], and \(\mathcal{H} = \{H\}\) with H≡1. Then one can check that N g ={G=0}=[0,1/2), which in turn implies that \(G = I_{N_{g}^{c}} / \mathbb{P}(N_{g}^{c})\), and
$$\begin{aligned} xa + \inf_{\mathcal{G} \times\mathcal{H}} \mathbb{E} [(G- a H)^+] &= xa + \mathbb{E}[(G- zH)^+] \\ &= \left\{ \begin{array}{l@{\quad}l} xa & \text{if}\ a \geq\frac{1}{P[N_g^c]},\\ 1+ a (x - \mathbb{P}[N_g^c] ) & \text{if}\ 0\leq a < \frac{1}{\mathbb{P}[N_g^c]}. \end{array} \right. \end{aligned}$$(A.1)Since \(x > \sup_{H\in\mathcal{H}}\mathbb{E} [HI_{N_{g}^{c}}] = \mathbb{P}[N_{g}^{c}]\), \(\hat{a} = 0\) is the unique minimizer of (A.1).
Proposition A.3
If
then there exists \((\hat{G}, \hat{H}, \hat{a}, \hat{X}) \in \mathcal{G} \times\overline{\operatorname {co}(\mathcal{H})} \times(0,\infty) \times \mathcal{X}_{x}\) satisfying (2.5)–(2.7). In particular,
Proof
Define the function \(f_{x}(a) := xa + \inf_{\mathcal{G} \times \overline{\operatorname {co}(\mathcal{H})}} \mathbb{E} [(G- a H)^{+}]\), which is Lipschitz-continuous (see Lemma 4.1 of [7]). Since \(f_{x}(0) = \inf_{\mathcal{G}} \mathbb{E}[G]\) is in [0,∞) and lim a→∞ f x (a)=∞, there exists a finite \(\hat{a} \ge0\) that minimizes f x (a).
Now suppose that \(\hat{a} = 0\) is a minimizer of f x (a). Then it follows that f x (a)≥f x (0), ∀a>0, which leads to
In (A.3), \(\tilde{G}\) minimizes \(\mathbb{E}[G]\) over \(\mathcal{G}\), and its existence follows from convexity and closedness of \(\mathcal{G}\). Taking the limit a↘0 yields a contradiction to (A.2) because
Hence, we conclude that \(\hat{a}>0\). □
1.3 A.3 Counterexample for Remark 2.5
Let Ω={ω 1,ω 2}, P[{ω 1}]=P[{ω 2}]=1/2. Then any random variable in \(\mathcal{G}, \mathcal{H}\) or in \(\mathcal{X}_{x}\), \(\mathcal{I}_{x}\) can be represented as a point in \({\mathbb{R}}^{2}\). Let \(\mathcal{H}\) be the line segment connecting (2,4) and (6,2), and \(\mathcal{G} = \{(2,2)\}\). Given x≥0, \(\mathcal{X}_{x}\) is the convex quadrangle with four vertices (0,0),(x/3,0),(x/5,2x/5), (0,x/2) intersected with {(x 1,x 2) | 0≤x 1,x 2≤1}. For all \(H =(h_{1},h_{2})\in\mathcal{H}\) and X=(x 1,x 2), the constraint \(\mathbb{E}[HX] \leq x\) implies that \(\frac{h_{1}}{2}x_{1} + \frac{h_{2}}{2}x_{2} \leq x\). The set \(\{(x_{1},x_{2}): \frac{h_{1}}{2} x_{1} + \frac{h_{2}}{2}x_{2} \le x\}\) is the lower half-plane bounded by the line h 1 x 1+h 2 x 2=2x, which passes through (x/5,2x/5) since h 1+2h 2=5. Hence, we have
where \(\mathcal{I}_{x}=\mathcal{X}_{x}\cap\{(0,0),(0,1),(1,0),(1,1)\}\). In summary, the values are given by the following table:
x | V(x) | V 1(x) |
---|---|---|
0≤x<2 | \(\frac{3}{5}x\) | 0 |
\(2\leq x < \frac{5}{2}\) | \(\frac{3}{5}x\) | 1 |
\(\frac{5}{2}\leq x < 4\) | \(\frac{x}{3}+\frac{2}{3}\) | 1 |
x≥4 | 2 | 2 |
By inspecting the values of V 1(x) we see that its smallest concave majorant must take the value \(\frac{x}{2}\) in [0,4]. Therefore, V(x) is not the smallest concave majorant of V 1(x).
1.4 A.4 Counterexample for Remark 2.11
With reference to Theorem 2.3, we show via an example that one cannot remove the independence requirement in (C2) of Theorem 2.10 when \(\mathcal{G}\) and \(\mathcal{H}\) are not both singletons.
Example A.4
Let Ω={0,1}×[0,1], \(\mathcal{F}_{T} = \mathcal{B}(\varOmega)\). Let μ be Lebesgue measure on [0,1]. Define \(\mathbb{P}\) by
Let \(H_{0}: \{0,1\} \to\mathbb{R}\) be given by H 0(0)=1/2 and H 0(1)=3/2, and \(f:[0,1]\to \mathbb{R}\) be an arbitrarily fixed probability density function. Define the set
and the singleton \(\mathcal{G} = \{G \equiv1\}\). Let U be a uniform random variable on \((\varOmega, \mathcal{F}_{T}, \mathbb{P})\), so that \(\mathbb{P}[U\le a] = a\) for a∈[0,1].
The pure hypothesis testing problem is
Direct computation gives the success set \(\hat{A} = \{0\}\) and the value of the pure hypothesis test V 1=1/2. On the other hand, the randomized hypothesis testing problem is
We find that \(\hat{H}(\alpha, a) = H_{0}(\alpha)\) and \(\hat{X} = I_{ \{\alpha= 0\}} + 1/3 I_{\{\alpha= 1\}}\) solve this randomized hypothesis test with the optimal value V=2/3.
This shows that the values of the pure and randomized hypothesis tests are different. If one were to construct an indicator version of the randomized test as in (2.13), namely
then although this test \(\bar{X} \) still satisfies \(\mathbb{E}[ \hat{H} \bar{X}] = 1/2\), it does not solve either the pure or the randomized hypothesis test. Indeed, for \(\tilde{H} (\alpha, a) = 3 I_{a<1/3} H_{0}(\alpha) \in \mathcal{H}\), we observe the violation \(\mathbb{E}[ \tilde{H} \bar{X}] = 1> 1/2\).
1.5 A.5 A property of nondegenerate martingales
On the probability space \((\varOmega, \mathcal{F}, \mathbb{P})\) with filtration \((\mathcal{F}_{t})_{0\le t\le1}\), we denote by W a standard Brownian motion. Let Y be a \((\mathbb{P}, \mathcal{F}_{t})\)-martingale defined by
where (σ t ) is a bounded \((\mathcal{F}_{t})\)-adapted process.
Proposition A.5
Assume that c<σ t <C for some positive constants c and C. Then
for all constants b.
To prove this proposition, we use the following two facts. We first define the function \(f: \mathbb{R}^{+} \times\mathbb{R}^{+} \times\mathbb{R} \to[0,1]\) by
-
1.
By direct computation we obtain
$$\sup_{u\in\mathbb{R}} f(x,y,u) = f(x,y,0) < 1. $$ -
2.
By a scaling argument we have
$$f(\lambda x, \lambda y, u) = f\bigg(x,y, \frac{u}{\sqrt{\lambda}}\bigg) \quad\forall\lambda>0. $$
Now we are ready to present the
Proof of Proposition A.5
Since Y is a continuous process,
By Lévy’s zero–one law we have
Therefore, it is enough to show that there exists a∈(0,1) such that
Note that the martingale (Y s |Y t =u:s>t) has the same distribution as a time-changed Brownian motion starting from state u. Combining this with the estimate \(c^{2} (1-t) \le\int_{t}^{1} \sigma_{r}^{2} dr \le C^{2} (1-t)\), we have for some standard Brownian motion B that
Since f(c 2,C 2,0) is independent of t and strictly less than 1, we can simply take a=f(c 2,C 2,0). □
One may wonder whether the condition on σ in Proposition A.5 can be relaxed to σ t >0 a.s. for all t. The answer is negative, as shown by the counterexample in [22].
Rights and permissions
About this article
Cite this article
Leung, T., Song, Q. & Yang, J. Outperformance portfolio optimization via the equivalence of pure and randomized hypothesis testing. Finance Stoch 17, 839–870 (2013). https://doi.org/10.1007/s00780-013-0213-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00780-013-0213-8