Abstract
In the framework of an incomplete financial market where the stock price dynamics are modeled by a continuous semimartingale (not necessarily Markovian), an explicit second-order expansion formula for the power investor’s value function—seen as a function of the underlying market price of risk process—is provided. This allows us to provide first-order approximations of the optimal primal and dual controls. Two specific calibrated numerical examples illustrating the accuracy of the method are also given.
Similar content being viewed by others
Notes
Theorem 4.5 in [13] expresses the corresponding value function as an infinite sum of weighted generalized Laguerre polynomials.
References
Altarovici, A., Muhle-Karbe, J., Soner, H.M.: Asymptotics for fixed transaction costs. Finance Stoch. 19, 363–414 (2015)
Bick, B., Kraft, H., Munk, C.: Solving constrained consumption-investment problems by simulation of artificial market strategies. Manag. Sci. 59, 485–503 (2013)
Brandt, M.W., Goyal, A., Santa-Clara, P., Stroud, J.R.: A simulation approach to dynamic portfolio choice with an application to learning about return predictability. Rev. Financ. Stud. 18, 831–873 (2005)
Campbell, J.Y.: Intertemporal asset pricing without consumption data. Am. Econ. Rev. LXXXIII, 487–512 (1993)
Campbell, J.Y., Viceira, L.M.: Consumption and portfolio decisions when expected returns are time varying. Q. J. Econ. 114, 433–495 (1999)
Cannarsa, P., Sinestrari, C.: Semiconcave Functions, Hamilton–Jacobi Equations, and Optimal Control. Progress in Nonlinear Differential Equations and their Applications, vol. 58. Birkhäuser Boston, Boston (2004)
Chacko, G., Viceira, L.M.: Dynamic consumption and portfolio choice with stochastic volatility in incomplete markets. Rev. Financ. Stud. 18, 1369–1402 (2005)
Cheridito, P., Filipović, D., Kimmel, R.L.: Market price of risk specifications for affine models: theory and evidence. J. Financ. Econ. 83, 123–170 (2007)
Cox, J.C., Huang, C.F.: Optimal consumption and portfolio policies when asset prices follow a diffusion process. J. Econ. Theory 49, 33–83 (1989)
Cvitanić, J., Goukasian, L., Zapatero, F.: Monte Carlo computation of optimal portfolios in complete markets. J. Econ. Dyn. Control 27, 971–986 (2003)
Davis, M.H.A.: Optimal hedging with basis risk. In: Kabanov, Yu., et al. (eds.) From Stochastic Calculus to Mathematical Finance. The Shiryaev Festschrift, pp. 169–187. Springer, Berlin (2006)
Detemple, J.B., Garcia, R., Rindisbacher, M.: A Monte Carlo method for optimal portfolio. J. Finance LVIII, 401–446 (2003)
Guasoni, P., Robertson, S.: Static fund separation of long-term investments. Math. Finance 25, 789–826 (2015)
Haugh, M.B., Kogan, L., Wang, J.: Evaluating portfolio policies: a duality approach. Oper. Res. 54, 405–418 (2006)
Henderson, V.: Valuation of claims on nontraded assets using utility maximization. Math. Finance 12, 351–373 (2002)
Herrmann, S., Muhle-Karbe, J., Seifried, F.T.: Hedging with small uncertainty aversion. Finance Stoch. 21, 1–64 (2017)
Hurd, T., Kuznetsov, A.: Explicit formulas for Laplace transforms of stochastic integrals. Markov Process. Relat. Fields 14, 277–290 (2008)
Kallsen, J., Muhle-Karbe, J.: Option pricing and hedging with small transaction costs. Math. Finance 25, 702–723 (2015)
Karatzas, I., Kardaras, C.: The numéraire portfolio in semimartingale financial models. Finance Stoch. 11, 447–493 (2007)
Karatzas, I., Lehoczky, J.P., Shreve, S.E.: Optimal portfolio and consumption decisions for a “small” investor on a finite horizon. SIAM J. Control Optim. 25, 1557–1586 (1987)
Karatzas, I., Lehoczky, J.P., Shreve, S.E., Xu, G.L.: Martingale and duality methods for utility maximization in an incomplete market. SIAM J. Control Optim. 29, 702–730 (1991)
Karatzas, I., Shreve, S.E.: Methods of Mathematical Finance, 1st edn. Applications of Mathematics (New York), vol. 39. Springer, New York (1998)
Kim, T.S., Omberg, E.: Dynamic nonmyopic portfolio behavior. Rev. Financ. Stud. 9, 141–161 (1996)
Kogan, L., Uppal, R.: Risk aversion and optimal policies in partial and general equilibrium economies. Working paper (2001). Available online at http://www.nber.org/papers/w8609
Kraft, H.: Optimal portfolios and Heston’s stochastic volatility model: an explicit solution for power utility. Quant. Finance 5, 303–313 (2005)
Kraft, H., Seiferling, T., Seifried, F.T.: Optimal consumption and investment with Epstein–Zin recursive utility. Finance Stoch. 21, 187–226 (2017)
Kramkov, D., Schachermayer, W.: The asymptotic elasticity of utility functions and optimal investment in incomplete markets. Ann. Appl. Probab. 9, 904–950 (1999)
Kramkov, D., Sîrbu, M.: On the two-times differentiability of the value functions in the problem of optimal investment in incomplete markets. Ann. Appl. Probab. 16, 1352–1384 (2006)
Kramkov, D., Sîrbu, M.: Sensitivity analysis of utility-based prices and risk-tolerance wealth processes. Ann. Appl. Probab. 16, 2140–2194 (2006)
Larsen, K.: A note on the existence of the power investor’s optimizer. Finance Stoch. 15, 183–190 (2011)
Larsen, K., Žitković, G.: Stability of utility-maximization in incomplete markets. Stoch. Process. Appl. 117, 1642–1662 (2007)
Larsen, L., Munk, C.: The costs of suboptimal dynamic asset allocation: general results and applications to interest rate risk, stock volatility risk, and growth/value tilts. J. Econ. Dyn. Control 36, 266–293 (2012)
Liu, J.: Portfolio selection in stochastic environments. Rev. Financ. Stud. 20, 1–39 (2007)
Monoyios, M.: Malliavin calculus method for asymptotic expansion of dual control problems. SIAM J. Financ. Math. 4, 884–915 (2013)
Mostovyi, O.: Necessary and sufficient conditions in the problem of optimal investment with intermediate consumption. Finance Stoch. 19, 135–159 (2015)
Munk, C.: Financial Asset Pricing Theory, vol. 6. Oxford University Press, London (2013)
Oliver, H.W.: The exact Peano derivative. Trans. Am. Math. Soc. 76, 444–456 (1954)
Rogers, L.C.G.: The relaxed investor and parameter uncertainty. Finance Stoch. 5, 131–154 (2001)
Szpiro, G.G.: Measuring risk aversion: an alternative approach. Rev. Econ. Stat. 68, 156–159 (1986)
Wachter, J.: Portfolio and consumption decisions under mean-reverting returns: an exact solution for complete markets. J. Financ. Quant. Anal. 37, 63–91 (2002)
Acknowledgements
The authors would like to thank the anonymous referees, the anonymous Associate Editor, the Editor Martin Schweizer, Milica Čudina, Claus Munk, Mihai Sîrbu, and Kim Weston for useful comments.
During the preparation of this work, the first author has been supported by the National Science Foundation under Grant No. DMS-1411809 (2014–2017). The second author has been supported by the National Science Foundation under grant No. DMS-1600307 (2015–2018). The third author has been supported by the National Science Foundation under Grant No. DMS-1107465 (2012–2017) and Grant No. DMS-1516165 (2015–2018). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: Details on the Kim–Omberg model
The following result summarizes the main properties in [23].
Theorem A.1
(Kim and Omberg 1996)
Let the market price of risk process be defined by (5.3), \(M:=B\), and let \(p<0\). Then there exist continuously differentiable functions \(b,c:[0,\infty)\to{\mathbb{R}}\) such that for \(t\in[0,T)\), we have
where \(\alpha_{1}:=\theta\kappa\), \(\alpha_{2} := (1+q) \beta^{2}+ \gamma^{2}\), \(\alpha_{3} := \beta^{2}+\gamma^{2}\) and \(\alpha_{4} := q \beta- \kappa\). Furthermore, the primal optimizer is given by
For \(p<0\), the above Riccati equation describing \(c\) has a “normal non-exploding solution” as defined in the Appendix of [23]. Therefore, all three functions \(a\), \(b\) and \(c\) are bounded on any finite subinterval \([0,T]\) of \([0,\infty)\).
Lemma A.2
Let \((\lambda,\lambda')\) be defined by (5.4) and (5.5) and let \(p<0\). For the basis model \(\lambda\), the primal and dual optimizers are given by
Furthermore, the processes \((\gamma^{B},\gamma^{W})\) appearing in the representation (3.6) of \(\varPhi\) are given by (5.6) and (5.7), where the functions \(C_{j}\) satisfy the ODEs
on \([0,T)\), with \((a,b,c)\) as in Theorem A.1 with \(\beta:=0\), \(\tilde{b}(t):=\kappa\theta-\gamma^{2} b(t)\) and \(\tilde{c}(t) := \kappa+\gamma^{2} c(t) \). Furthermore, for the measure \(\tilde{\mathbb{P}}^{(0)}\) defined by (2.6) and for all \(T>0\), we have
Proof
The first part follows from Theorem A.1 applied to the case \(\beta:=0\). To find the representation (3.6), we define the function
where the functions \((b,c)\) are as in Theorem A.1 and \(a\) is given by
The martingale properties of \((f(t,\hat{X}^{(0)}_{t},\lambda_{t}))\) and \((\hat{X}^{(0)}_{t}\hat{Y}^{(0)}_{t})\) as well as the proportionality property \((\hat{X}^{(0)}_{T})^{p}\propto\hat{X}^{(0)}_{T} \hat{Y} ^{(0)}_{T}\) produce
By computing the dynamics of the left-hand side, we see from Girsanov’s theorem that the two processes (see (4.13))
are independent Brownian motions under \(\tilde{\mathbb{P}}^{(0)}\). These dynamics and Itô’s lemma ensure that
is a \(\tilde{\mathbb{P}}^{(0)}\)-local martingale. Because the processes \((\lambda,\lambda')\) remain Ornstein–Uhlenbeck processes under \(\tilde{\mathbb{P}}^{(0)}\) and the functions \(C_{1},\dots,\! C_{6}\) are bounded, \(N\) is indeed a \(\tilde{\mathbb{P}}^{(0)}\)-martingale. Furthermore, thanks to the zero terminal conditions imposed on \(C_{1},\dots,C_{6}\), we see that
for \((\gamma^{B},\gamma^{W})\) defined by (5.6) and (5.7). □
Appendix B: Details on the extended affine models
The following result is from [25].
Theorem B.1
(Kraft 2005)
Let \(p<0\) and consider the model (5.8) with \(\lambda_{t}:=1\). The primal optimal control is \(\hat{\pi}^{(0)}_{t} = \frac{f(t)-1}{p-1}\), the dual optimal control corresponding to \(W\) is \(\hat{\nu}^{(0)}_{t} =f(t) \sqrt{F_{t}}\), and the deterministic function \(f\) is given by
For the model (5.10) with \(\lambda_{t}:=\lambda_{21}\), the primal optimal control is \(\hat{\pi}^{(0)}_{t} = \frac{\lambda_{21}}{1-p}\), the dual optimal controls corresponding to \((W^{(1)},W^{(3)})\) are
and the deterministic function \(f\) is given by
We then turn to the representation (3.6) of \(\varPhi:= \int _{0}^{T} \hat{\pi}^{(0)}_{t} \lambda_{t}' \sigma_{t}^{2} dt\). For the model (5.8), this is trivial because \(\int_{0}^{T} \hat{\pi}^{(0)}_{t} dt\) is deterministic (implying that \(\gamma^{B}= \gamma^{W}=0\)). The next lemma provides the representation for the model in (5.11).
Lemma B.2
Let \(p<0\), let \((\lambda,\lambda')\) be defined by (5.10) and (5.11), and define \(\hat{\pi} ^{(0)}_{t} := \frac{\lambda_{21}}{1-p}\). The integrands \((\gamma^{B}, \gamma^{W})\) appearing in the representation (3.6) of
are given by
where the functions \(C_{j}\) satisfy on \([0,T]\) the ODEs
Proof
This is similar to the proof of Lemma A.2. Because the functions \(C_{j}\) satisfy the above ODEs, we have that
is a \(\tilde{{\mathbb{P}}}^{(0)}\)-martingale (here we use the \(\tilde{\mathbb{P}}^{(0)}\)-Brownian motions defined in (4.13)). The function \(C_{0}\) is given by
with \(C_{0}(T)=0\). Then we have
Hence, we find the integrands appearing in the representation (3.6) of \(\varPhi\) to be as in the statement. □
Finally, we verify the integrability conditions (3.7) and (3.8) in Theorem 3.4, provided that \(T>0\) is small. We can define the second-order optimal controls \((\tilde{\pi}, \tilde{\nu})\) by (3.12) and (3.13). We start by considering the model (5.8). Theorem 3.1 in [17] states that the joint Laplace transform of
is finite for all \(T>0\) in some neighborhood of 0 as soon as the strict Feller condition \(2\kappa\theta> 1\) holds. This implies that both \(\varLambda\) and \(Q\) have some finite exponential positive moments (and consequently, \(\varLambda\) and \(Q\) have all moments). Because the integrand \(\gamma^{B}\) appearing in the representation (3.6) of \(\varPhi\) is zero in this model, the integrand appearing inside the exponential in (3.8) is deterministic. Consequently, (3.8) holds for all \(\varepsilon>0\).
For the model (5.11), we have the estimate
which shows that the moment requirements in (3.7) hold. Furthermore, by inserting \(\lambda'\) defined in (5.11) and \(\sigma^{2}_{t} = 1+F^{(1)}_{t}\) into the exponent in (3.8), we see that this exponent is affine in \(F^{(1)}\), \(F^{(2)}\) and \(F^{(3)}\) (with time-dependent coefficients). Consequently, the expectation in (3.8) is finite for small positive values of \(\varepsilon\).
Rights and permissions
About this article
Cite this article
Larsen, K., Mostovyi, O. & Žitković, G. An expansion in the model space in the context of utility maximization. Finance Stoch 22, 297–326 (2018). https://doi.org/10.1007/s00780-017-0353-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00780-017-0353-3
Keywords
- Continuous semimartingales
- Second-order expansion
- Incomplete markets
- Power utility
- Convex duality
- Optimal investment