Optimal Portfolio Positioning on Multiple Assets Under Ambiguity


This paper determines the optimal financial portfolio in the multidimensional setting when the investor exhibits ambiguity aversion. We consider the Maccheroni et al. (Econometrica 74(6):1447–1498, 2006) framework which includes both the Gilboa and Schmeidler’s (J Math Econ 18(2):141–153, 1989) multiple priors preferences and the (American Econ Rev 91:60–66, 2001) multiplier preferences. We determine the optimal portfolio profile under ambiguity when the investors can invest on various risky assets. We investigate in particular the CRRA case while introducing an ambiguity index based on the relative entropy criterion. Such result extends Ben Ameur and Prigent (Econ Model 34:89–97, 2013) when there is only one risky asset. Indeed, we show for example how the ambiguity on the correlations between the risky assets crucially modify the optimal payoff. Such results have important practical applications in structured portfolio management when investing on multiple financial indices and basket options.

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

Fig. 1
Fig. 2
Fig. 3
Fig. 4


  1. 1.

    See Campbell and Viceira (2002) and Prigent (2007) for an overview of such results.

  2. 2.

    Recall that the relative entropy \(R(\mathbb {P},\mathbb {P}_{0})\) of the probability distribution \(\mathbb {P}\) with respect to the probability distribution \(\mathbb {P}_{0}\) is defined by:

    $$\begin{aligned} R(\mathbb {P},\mathbb {P}_{0})=E_{\mathbb {P}_{0}}\left[ \left( \frac{d\mathbb {P }}{d\mathbb {P}_{0}}\right) Log\left( \frac{d\mathbb {P}}{d\mathbb {P}_{0}} \right) \right] . \end{aligned}$$

    Note in particular that \(R(\mathbb {P}_{0},\mathbb {P}_{0})=0\). The relative entropy has been introduced in both the probability and information theories to measure the difference between the probability distributions. It corresponds to the Kullback-Leibler divergence. It quantifies the information gain when we revise beliefs from the prior probability distribution to the posterior probability distribution.

  3. 3.

    This corresponds for example to the distribution of a geometric Brownian motion \((S_{t})_{t}\) in a continuous-time framework, satisfying

    $$\begin{aligned} S_{t}=S_{0}exp\left[ (\mu -1/2\sigma ^{2})t+\sigma W_{t}\right] , \end{aligned}$$

    with \(d=(\mu -1/2\sigma ^{2})\) and where W denotes the standard Brownian motion.

  4. 4.

    Minimax Theorem (Sion 1958): (Saddle-point ) Let C and K be two closed convex sets in two topological vector spaces X and Y respectively. Let further F(xy) : \( C\times K\rightarrow R\) be a function which is quasiconcave in x and quasiconvex in y. If F is upper (or lower) semi continuous in x and lower semi continuous in y, while K is compact, then the function F(xy) possesses a saddle-value on \(C\times K\) and

    $$\begin{aligned} \sup _{x\in C}\inf _{y\in K}F(x,y)=\inf _{y\in D}\sup _{x\in K}F(x,y). \end{aligned}$$
  5. 5.

    Function g is the Radon-Nikodym density \(d\mathbb {Q}/d\mathbb {P}\).

  6. 6.

    We examine this case because it is the most common in practice. Other cases can also be considered if the Log return of the risky asset is no longer Gaussian.

  7. 7.

    See e.g. Prigent (2007).


  1. Allais, M. (1953). Le comportement de l’homme rationnel devant le risque: Critique des postulats et axiomes de l’école amé ricaine. Econometrica, 21, 503–546.

    Google Scholar 

  2. Ben Ameur, H., & Prigent, J.-L. (2013). Optimal portfolio positioning under ambiguity. Economic Modelling, 34, 89–97.

    Google Scholar 

  3. Bertrand, P., Lesne, J.-P., & Prigent, J.-L. (2001). Gestion de portefeuille avec garantie : l’allocation optimale en actifs dérivés. Finance, 22, 7–35.

    Google Scholar 

  4. Biagini, S., & Pinar, M. C. (2017). The robust Merton problem of an ambiguity averse investors. Mathematics and Financial Economics, 11, 1–24.

    Google Scholar 

  5. Breeden, D., & Litzenberger, R. (1978). Prices of state contingent claims implicit in options prices. Journal of Business, 51, 621–651.

    Google Scholar 

  6. Brennan, M. J., & Solanki, R. (1981). Optimal portfolio insurance. Journal of Financial and Quantitative Analysis, 16(3), 279–300.

    Google Scholar 

  7. Buhlmann, H., Delbaen, F., Embrechts, P., & Shiryaev, A. (1996). No-arbitrage, change of measure and conditional Esscher transforms. CWI Quaterly, 9, 291–317.

    Google Scholar 

  8. Calafiore, G. C. (2007). Ambiguous risk measures and optimal robust portfolios. SIAM Journal of Optimization, 18, 853–877.

    Google Scholar 

  9. Campbell, J. Y., & Viceira, L. M. (2002). Strategic asset allocation: Portfolio choice for long-term investors. Clarendon lecture in economics. New York: Oxford University Press.

    Google Scholar 

  10. Carr, P., & Madan, D. (2001). Optimal positioning in derivative securities. Quantitative Finance, 1, 19–37.

    Google Scholar 

  11. Chen, Z., & Epstein, L. (2002). Ambiguity, risk, and asset returns in continuous time. Econometrica, 70, 1403–1443.

    Google Scholar 

  12. Chen, X., Deelstra, G., Dhaene, J., & Vanmaele, M. (2008). Static super-replicating strategies for a class of exotic options. Insurance: Mathematics and Economics, 42(3), 1067–1085.

    Google Scholar 

  13. Cox, J., & Huang, C.-F. (1989). Optimal consumption and portfolio policies when assets prices follow a diffusion process. Journal of Economic Theory, 49, 33–83.

    Google Scholar 

  14. Crouzeix, J. P., & Lindberg, P. O. (1986). Additively decomposed quasi-convex functions. Mathematical Programming, 35, 42–57.

    Google Scholar 

  15. Cvitanic, J., & Karatzas, I. (1996). Contingent claim valuation and hedging with constrained portfolio. In M. Davis, D. Duffie, W. Fleming, S. Shreve. (Eds.), IMA volume in mathematics (Vol. 65, pp. 13–33). Springer.

  16. De Giovanni, D., Ortobelli, S., & Svetlozar Rachev, S. (2008). Delta hedging strategies comparison. European Journal of Operational Research, 185(3), 1615–1631.

    Google Scholar 

  17. Debreu, G., & Koopmans, T. C. (1982). Additively decomposed quasiconcave functions. Mathematical Programming, 24, 1–38.

    Google Scholar 

  18. Ekeland, I., & Turnbull, T. (1983). Infinite-dimensional optimization and convexity., Chicago lectures in mathematics Chicago: The University of Chicago Press.

    Google Scholar 

  19. El Ghaoui, L., Oks, M., & Oustry, F. (2003). Worst-case value-at-risk and robust portfolio optimization: A conic programming approach. Operations Research, 51(4), 543–556.

    Google Scholar 

  20. El Karoui, N., Jeanblanc, M., & Lacoste, V. (2005). Optimal portfolio management with American capital guarantee. Journal of Economic Dynamics and Control, 29, 449–468.

    Google Scholar 

  21. Ellsberg, D. (1961). Risk, ambiguity, and the Savage axioms. Quarterly Journal of Economics, 75(4), 643–669.

    Google Scholar 

  22. Epstein, L. G. (1999). A definition of uncertainty aversion. Review of Economic Studies, 66, 579–608.

    Google Scholar 

  23. Epstein, G., & Ji, S. (2013). Ambiguous volatility and asset pricing in continuous time. The Review of Financial Studies, 26, 1740–1786.

    Google Scholar 

  24. Fabozzi, F. J., Huang, D., & Zhou, G. (2010). Robust portfolios: Contributions from operations research and finance. Annals of Operations Research, 176(1), 191–220.

    Google Scholar 

  25. Fei, W. (2007). Optimal consumption and portfolio choice with ambiguity and anticipation. Information Sciences, 177, 5178–5190.

    Google Scholar 

  26. Fouque, J. P., Chiu, M. C., & Wong, H. Y. (2016). Portfolio optimization with ambiguous correlation and stochastic volatilities. SIAM Journal on Control and Optimization, 54, 2309–2338.

    Google Scholar 

  27. Gilboa, I., & Schmeidler, D. (1989). Maxmin expected utility with a non-unique prior. Journal of Mathematical Economics, 18(2), 141–153.

    Google Scholar 

  28. Hansen, L., & Sargent, T. (2000). Wanting robustness in macroeconomics. In B. M. Friedman & M. Woodford (Eds.), Handbook of monetary economics. Amsterdam: North-Holland.

    Google Scholar 

  29. Hansen, L. P., & Sargent, T. J. (2001). Robust control and model uncertainty. American Economic Review, 91, 60–66.

    Google Scholar 

  30. Hansen, L., & Sargent, T. (2011). Robust control and model uncertainty. American Economic Review, 91, 60–66.

    Google Scholar 

  31. Ismail, A., & Pham, H. (2016). Robust Markowitz mean-variance portfolio selection under ambiguous volatility and correlation. arXiv:1610.06805.

  32. Jacod, J., & Shiryaev, A. (2002). Limit Theorems for Stochastic Processes. Berlin and Heidelberg: Springer-Verlag.

    Google Scholar 

  33. Knight, F. H. (1921). Risk, uncertainty, and profit. Boston: Houghton Mifflin Co.

    Google Scholar 

  34. Koziol, C., Proelssz, J., & Schweizer, D. (2011). Are institutional investors ambiguity averse? Evidence from portfolio holdings in alternative investments. International Journal of Theoretical and Applied Finance, 14, 465–484.

    Google Scholar 

  35. Leland, H. (1980). Who should buy portfolio insurance? Journal of Finance, 35, 581–594.

    Google Scholar 

  36. Leland, H. E., & Rubinstein, M. (1976). The evolution of portfolio insurance. In D. L. Luskin (Ed.), Portfolio insurance: A guide to dynamic hedging. New York: Wiley.

    Google Scholar 

  37. Liu, H. (2011). Dynamic portfolio choice under ambiguity and regime switching mean returns. Journal of Economic Dynamics and Control, 35, 623–640.

    Google Scholar 

  38. Maccheroni, F., Marinacci, M., & Rustichini, A. (2006). Ambiguity aversion, robustness, and the variational representation of preferences. Econometrica, 74(6), 1447–1498.

    Google Scholar 

  39. Maenhout, P. J. (2004). Robust portfolio rules and asset pricing. Review of Financial Studies, 17, 951–983.

    Google Scholar 

  40. Markowitz, H. (1952). Portfolio selection. Journal of Finance, 7(1), 77–91.

    Google Scholar 

  41. Merton, R. C. (1971). Optimum consumption and portfolio rules in a continuous time model. Journal of Economic Theory, 3, 373–413.

    Google Scholar 

  42. Pellizzari, P. (2005). Static hedging of multivariate derivatives by simulation. European Journal of Operational Research, 166, 507–519.

    Google Scholar 

  43. Peng, S. (2006). G -Expectation, G -Brownian Motion and Related Stochastic Calculus of Ito’s type, The Abel Symposium 2005, Abel Symposia 2, Edit. Benth et. al., 541C567, Springer-Verlag, 2006.

  44. Pflug, G. C., Pichler, A., & Wozabal, D. (2012). The 1/N investment strategy is optimal under high model ambiguity. Journal of Banking and Finance, 36, 410–417.

    Google Scholar 

  45. Pflug, G. C., & Wozabal, D. (2007). Ambiguity in portfolio selection. Quantitative Finance, 7, 435–442.

    Google Scholar 

  46. Prigent, J.-L. (2007). Portfolio optimization and performance analysis. Boca Raton, FL: Chapman & Hall.

    Google Scholar 

  47. Sion, M. (1958). On general minimax theorems. Pacific Journal of Mathematics, 8, 171–176.

    Google Scholar 

  48. Tebaldi, C. (2005). Hedging using simulation: A least squares approach. Journal of Economic Dynamics and Control, 29(8), 1287–1312.

    Google Scholar 

  49. Tobin, J. (1958). Liquidity preference as behavior toward risk. Review of Economic Studies, 25, 65–86.

    Google Scholar 

  50. Tütüncü, R. H., & Koenig, M. (2004). Robust asset allocation. Annals of Operations Research, 132, 157–187.

    Google Scholar 

  51. von Neumann, J., & Morgenstern, O. (1947). Theory of games and economic behavior. Princeton: Princeton University Press.

    Google Scholar 

  52. Wozabal, D. (2012). A framework for optimization under ambiguity. Annals of Operations Research, 193, 21–47.

    Google Scholar 

Download references

Author information



Corresponding author

Correspondence to J. L. Prigent.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



Proof of Proposition 13

In what follows, we use the following notations (see Jacod and Shiryaev 2002). For any two semimartingales X and Y, \(\left[ X,Y\right] \) denotes the quadratic variation of the processes X and Y. The process \( \left\langle X,Y\right\rangle \) denotes the predictable compensator of these processes. Recall that we have:

$$\begin{aligned} \left[ X,Y\right] _{t}=\left\langle X,Y\right\rangle _{t}+\sum _{s\le t}\Delta X_{s}\Delta Y_{s}, \end{aligned}$$

where \(\Delta X_{t}\) and \(\Delta Y_{t}\) denote the jumps of the processes and \(\left\langle X,Y\right\rangle _{t}\) is defined from the following condition: The respective martingales parts \(M^{X}\) and \(M^{Y}\) of X and Y are such that \(\left( M_{t}^{X}M_{t}^{Y}-\left\langle X,Y\right\rangle _{t}\right) _{t}\) is a (local)martingale. We have also: (integration by part formula)

$$\begin{aligned} d\left( XY\right) =XdY+YdX+d\left[ X,Y\right] . \end{aligned}$$

The process \(\mathcal {E}(X)\) denotes the Dade-Doléans stochastic exponential, defined from the stochastic differential equation (SDE):

$$\begin{aligned} d\mathcal {E}(X)=\mathcal {E}(X)dX. \end{aligned}$$

Note that, for continuous semimartingales X, we have \(\left[ X,Y\right] _{t}=\left\langle X,X\right\rangle _{t}\) thus we get:

$$\begin{aligned} \mathcal {E}(X_{t})=\mathcal {E}(X_{0})\exp \left[ X_{t}-\frac{1}{2}\left\langle X,X\right\rangle _{t}\right] . \end{aligned}$$

The result is established by using the Girsanov’s theorem. We parallel the approach of Buhlmann et al. (1996). Each of the d basic assets \(S_{i,t}\) must satisfy the following condition: when they are discounted by the nominal money market account C, they must be martingales with respect to the risk-neutral probability \(\mathbb {Q}\). This is equivalent to the fact that, when they are multiplied by the Radon-Nikodym density \(\eta \) and divided by C, they must be martingales with respect to the historical probability \(\mathbb {P}\). Note that we have:

$$\begin{aligned} S_{i,t}\eta _{t}/C_{t}=\mathcal {E}\left[ \left( \mu _{i}-r\right) t+\sigma _{i}W_{i,t}\right] \mathcal {E}\left[ -\sum _{j=1}^{d}\lambda _{j}W_{j,t} \right] , \end{aligned}$$

which is also equal to (Yor’s formula):

$$\begin{aligned} \mathcal {E}\left[ \left( \mu _{i}-r\right) t+\sigma _{i}W_{i,t}-\sum _{j=1}^{d}\lambda _{j}W_{j,t}-\sigma _{i}\left( \sum _{j=1}^{d}\lambda _{j}\rho _{i,j}\right) t\right] . \end{aligned}$$

Note that we have:

$$\begin{aligned} \left\langle \sigma _{i}W_{i,t},\sum _{j=1}^{d}\lambda _{j}W_{j,t}\right\rangle =\sigma _{i}\left( \sum _{j=1}^{d}\lambda _{j}\rho _{i,j}\right) t. \end{aligned}$$

The fact that the processes \(\left( S_{i,t}\eta _{t}/C_{t}\right) \) are martingales with respect to \(\mathbb {P}\) is equivalent to the following property: Their bounded variation components are equal to 0. This later condition implies the d following equalities: for all \(i=1,\ldots ,d\),

$$\begin{aligned} \left( \mu _{i}-r\right) -\sigma _{i}\left( \sum _{j=1}^{d}\lambda _{j}\rho _{i,j}\right) =0, \end{aligned}$$

which leads to Eq. (19).

Proof of Corollary 14

We have:

Proof of Proposition 16

In what follows, we prove that, for the fundamental example, the optimal portfolio value satisfies:

$$\begin{aligned} V_{t}^{*}= & {} P_{t}+C_{t}, \\&\text {with} \\ P_{t}= & {} e^{-r(T-t)}pV_{0}\text { (the floor),} \\ C_{t}= & {} \psi (t)\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{m_{i}^{*}}\text { (the cushion),} \end{aligned}$$

- Indeed, as seen in (23), the portfolio value \(V_{T}^{*}\) at maturity T is given by:

$$\begin{aligned} V_{T}^{*}= & {} pV_{0}+\widetilde{\psi }(T)\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{\left( \frac{1}{\phi }\frac{\lambda _{i}}{\sigma _{i}}\right) } \\&\text {with} \\ \widetilde{\psi }(t)= & {} c\exp \left[ 1/2\frac{1}{\phi }\left( \sum _{i=1,j=1}^{d}\lambda _{i}\lambda _{j}\rho _{i,j}\right) t\right] \prod \limits _{i=1}^{d}\exp \left( -\frac{1}{\phi }\frac{\lambda _{i}}{\sigma _{i}} (\mu _{i}-1/2\sigma _{i}^{2})t\right) \end{aligned}$$

- From the martingale property, the portfolio value \(V_{t}\) at any time t satisfies:

$$\begin{aligned} V_{t}^{*}=e^{-r(T-t)}\mathbb {E}\left[ V_{T}^{*}\left| \mathcal {F }_{t}\right. \right] . \end{aligned}$$

Thus, we have:

$$\begin{aligned} V_{t}^{*}=e^{-r(T-t)}pV_{0}+\widetilde{\psi }(T)\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{\left( \frac{1}{\phi }\frac{ \lambda _{i}}{\sigma _{i}}\right) }\mathbb {E}\left[ \prod \limits _{i=1}^{d}\left( S_{i,T}/S_{i,t}\right) ^{\left( \frac{1}{\phi }\frac{ \lambda _{i}}{\sigma _{i}}\right) }\left| \mathcal {F}_{t}\right. \right] . \end{aligned}$$

It means that the portfolio value \(V_{t}^{*}\) is a function of the basic assets \(S_{i,t}\) at time t, given by:

$$\begin{aligned} V_{t}^{*}= & {} e^{-r(T-t)}pV_{0}+\psi (t)\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{\left( \frac{1}{\phi }\frac{\lambda _{i}}{\sigma _{i}}\right) } \\&\text {with} \\ \psi (t)= & {} \widetilde{\psi }(t)\mathbb {E}\left[ \prod \limits _{i=1}^{d}\left( S_{i,T}/S_{i,t}\right) ^{\left( \frac{1}{\phi }\frac{ 1}{\gamma }\frac{\lambda _{i}}{\sigma _{i}}\right) }\left| \mathcal {F} _{t}\right. \right] . \end{aligned}$$

from which we deduce the result.

- We examine now the term \(\mathbb {E}\left[ \prod \limits _{i=1}^{d}\left( S_{i,T}/S_{i,t}\right) ^{\left( \frac{1}{\phi }\frac{\lambda _{i}}{\sigma _{i}}\right) }\left| \mathcal {F}_{t}\right. \right] \):

We have:

$$\begin{aligned} \left( S_{i,T}/S_{i,t}\right) ^{m_{i}^{*}}= & {} \exp \left[ m_{i}^{*}\left( (\mu _{i}-1/2\sigma _{i}^{2})(T-t)+\sigma _{i}\left( W_{i,T}-W_{i,t}\right) \right) \right] \\= & {} \exp \left[ m_{i}^{*}(\mu _{i}-1/2\sigma _{i}^{2})(T-t)+1/2\left( m_{i}^{*}\right) ^{2}\sigma _{i}^{2}(T-t)\right] \\&\times \exp \left[ -1/2\left( m_{i}^{*}\right) ^{2}\sigma _{i}^{2}(T-t)+m_{i}^{*}\sigma _{i}\left( W_{i,T}-W_{i,t}\right) \right] . \end{aligned}$$


$$\begin{aligned}&\mathbb {E}\left[ \prod \limits _{i=1}^{d}\left( S_{i,T}/S_{i,t}\right) ^{m_{i}^{*}}\left| \mathcal {F}_{t}\right. \right] \\&=\exp \left[ \sum _{i=1}^{d}m_{i}^{*}(\mu _{i}-1/2\sigma _{i}^{2})(T-t)+1/2\left( m_{i}^{*}\right) ^{2}\sigma _{i}^{2}(T-t)\right] . \end{aligned}$$

Proof of Proposition 17

In what follows, we prove that, the Delta hedging strategies are given by:

$$\begin{aligned} \Delta _{i,t}^{*}=\frac{e_{i,t}^{*}}{S_{i,t}}=\frac{m_{i}^{*}C_{t}}{S_{i,t}}=\left( \frac{1}{\phi }\frac{\lambda _{i}}{\sigma _{i}} \right) \frac{\psi (t)}{S_{i,t}}\left( S_{i,t}/S_{i,0}\right) ^{\left( \frac{ 1}{\phi }\frac{\lambda _{i}}{\sigma _{i}}\right) }\prod \limits _{j=1,j\ne i}^{d}\left( S_{j,t}/S_{i,0}\right) ^{\left( \frac{1}{\phi }\frac{\lambda _{j}}{\sigma _{j}}\right) }. \end{aligned}$$

To determine the optimal portfolio shares, first we determine the SDE satisfied by the optimal portfolio value. We have:

$$\begin{aligned} dV_{t}^{*}=dP_{t}+\psi (t)d\left( \prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{m_{i}^{*}}\right) +\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{m_{i}^{*}}\psi ^{\prime }(t)dt \end{aligned}$$

We have to identify the factors that multiply the terms \(dS_{i,t}/S_{i,t}\). For this purpose, we note that:

$$\begin{aligned} d\left( \prod \limits _{i=1}^{d}S_{i,t}^{m_{i}^{*}}\right) =\sum _{i=1}^{d}\left( m_{i}^{*}\left[ \prod \limits _{j=1,j\ne i}^{d}S_{j,t}^{m_{i}^{*}}\right] S_{i,t}^{m_{i}^{*}-1}dS_{i,t}\right) +\vartheta _{t}, \end{aligned}$$

where \(\vartheta \) is a bounded variation process. Therefore we have also:

$$\begin{aligned} d\left( \prod \limits _{i=1}^{d}S_{i,t}^{m_{i}^{*}}\right) =\sum _{i=1}^{d}\left( m_{i}^{*}\left[ \prod \limits _{i=1}^{d}S_{i,t}^{m_{i}^{*}}\right] \frac{dS_{i,t}}{S_{i,t}} \right) +\vartheta _{t}. \end{aligned}$$

Since the portfolio value satisfies:

$$\begin{aligned} V_{t}^{*}-P_{t}=\psi (t)\prod \limits _{i=1}^{d}\left( S_{i,t}/S_{i,0}\right) ^{m_{i}^{*}}, \end{aligned}$$

we deduce that:

$$\begin{aligned} dV_{t}^{*}=\sum _{i=1}^{d}m_{i}^{*}\left[ V_{t}^{*}-P_{t}\right] \frac{dS_{i,t}}{S_{i,t}}+\varkappa _{t}, \end{aligned}$$

where \(\varkappa \) is a bounded variation process. Finally, by identifying the factors of \(\frac{dS_{i,t}}{S_{i,t}}\), we get:

$$\begin{aligned} e_{i,t}^{*}=m_{i}^{*}\left[ V_{t}^{*}-P_{t}\right] . \end{aligned}$$

The Delta hedging strategies are immediately deduced from the equality \( \Delta _{i,t}^{*}=\frac{e_{i,t}^{*}}{S_{i,t}}\).

Proof of Lemma 26

The Radon-Nikodym density \(\frac{d\mathbb {P}}{d\mathbb {P}_{0}}\) is determined as follows:

We denote by \(f_{S_{T}}\) the probability distribution function (pdf) of asset \(S_{T}=\left( S_{T}^{(1)},S_{T}^{(2)}\right) \) with respect to \( \mathbb {P}\) and by g the ratio \(f_{S_{T}}/f_{0,S_{T}}\).


$$\begin{aligned} \alpha _{T}^{(1)}= & {} \left( \mu ^{(1)}-0.5\sigma ^{(1)2}\right) T; \alpha _{T}^{(2)}=\left( \mu ^{(2)}-0.5\sigma ^{(2)2}\right) T, \\ \widetilde{\Sigma }= & {} \left[ \begin{array}{cc} \sigma ^{(1)2}T &{} \sigma ^{(1)}\sigma ^{(2)}\rho T \\ \sigma ^{(1)}\sigma ^{(2)}\rho T &{} \sigma ^{(2)2}T \end{array} \right] . \end{aligned}$$

We have:

$$\begin{aligned}&f_{S_{T}}(s^{(1)},s^{(2)})=\frac{1}{2\pi }\frac{1}{s^{(1)}s^{(2)}\sqrt{\det \left( \widetilde{\Sigma }\right) }}\\&\exp \left[ -\frac{1}{2}\left[ \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)}.\Sigma ^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)} \end{array} \right) \right] \right] \\&f_{0,S_{T}}(s^{(1)},s^{(2)})=\frac{1}{2\pi }\frac{1}{s^{(1)}s^{(2)}\sqrt{ \det \left( \widetilde{\Sigma }_{0}\right) }}\\&\exp \left[ -\frac{1}{2}\left[ \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)}.\Sigma _{0}^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)} \end{array} \right) \right] \right] \\&g(s^{(1)},s^{(2)})=\frac{\sqrt{\det \left( \widetilde{\Sigma }_{0}\right) }}{ \sqrt{\det \left( \widetilde{\Sigma }\right) }}\\&\qquad \times \exp \left[ -\frac{1}{2}\left[ \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)}.\Sigma ^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)} \end{array} \right) - \\ \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)}.\Sigma _{0}^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)} \end{array} \right) \end{array} \right] \right] \end{aligned}$$

Thus, we get:

$$\begin{aligned}&\log \left[ g(s^{(1)},s^{(2)})\right] =\ln \left( \frac{\sqrt{\det \left( \widetilde{\Sigma }_{0}\right) }}{\sqrt{\det \left( \widetilde{\Sigma } \right) }}\right) \\&-\frac{1}{2}\left[ \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)}.\Sigma ^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{T}^{(2)} \end{array} \right) - \\ \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)},\ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)}.\Sigma _{0}^{-1}\left( \begin{array}{c} \ln \left[ s^{(1)}\right] -\alpha _{0,T}^{(1)} \\ \ln \left[ s^{(2)}\right] -\alpha _{0,T}^{(2)} \end{array} \right) \end{array} \right] . \end{aligned}$$

Therefore, we deduce:

$$\begin{aligned} \delta _{T}= & {} \int _{0}^{+\infty }\int _{0}^{+\infty }g(s^{(1)},s^{(2)})\log \left[ g(s^{(1)},s^{(2)})\right] f_{0,S_{T}}(s^{(1)},s^{(2)})ds^{(1)}ds^{(2)}\\= & {} \ln \left( \frac{\sigma _{0}^{(1)}\sigma _{0}^{(2)}\sqrt{1-\rho _{0}^{2}}}{ \sigma ^{(1)}\sigma ^{(2)}\sqrt{1-\rho ^{2}}}\right) -1\\&+\frac{1}{2\left( 1-\rho _{0}^{2}\right) }\left( \begin{array}{c} \frac{\sigma ^{(1)2}}{\sigma _{0}^{(1)2}}+\frac{\sigma ^{(2)2}}{\sigma _{0}^{(2)2}}+\left( \frac{\alpha _{T}^{(1)}-\alpha _{0,T}^{(1)}}{\sigma _{0}^{(1)}\sqrt{T}}\right) ^{2}+\left( \frac{\alpha _{T}^{(2)}-\alpha _{0,T}^{(2)}}{\sigma _{0}^{(2)}\sqrt{T}}\right) ^{2} \\ -2\frac{\rho _{0}}{\sigma _{0}^{(1)}\sigma _{0}^{(2)}T}\left[ \sigma ^{(1)}\sigma ^{(2)}\rho T+\left( \alpha _{T}^{(1)}-\alpha _{0,T}^{(1)}\right) \left( \alpha _{T}^{(2)}-\alpha _{0,T}^{(2)}\right) \right] \end{array} \right) . \end{aligned}$$

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Ben Ameur, H., Boujelbène, M., Prigent, J.L. et al. Optimal Portfolio Positioning on Multiple Assets Under Ambiguity. Comput Econ 56, 21–57 (2020). https://doi.org/10.1007/s10614-019-09894-y

Download citation


  • Portfolio optimization
  • Ambiguity
  • Multiple assets
  • Structured portfolio

JEL Classification

  • C61
  • G11
  • L10