Portfolio optimization for an investor with a benchmark
Carl Lindberg (AP2 - The Second Swedish National Pension Fund, Sweden)
Joint work with Ralf Korn

Wednesday June 4, 11:30-12:00 | session 4.2 | Portfolio Optimization | room CD

The most common equity mandate in the financial industry is to try to outperform an externally given benchmark with known weights. The standard quantitative approach to do this is to optimize the portfolio over short time horizons consecutively, using one-period models. However, it is not clear that this approach actually yields good performance in the long run. We provide a theoretical justification to this methodology by verifying that applying the one-period benchmark-relative mean-variance portfolio, i. e. the industry standard optimal portfolio, continuously is in fact the solution to a specific continuous time portfolio optimization problem: a maximum expected utility problem for an investor who is compared against a benchmark, and evaluates her performance based on exponential utility at a deterministic future date.

Market viability and martingale measures under partial information
Claudio Fontana (University of Evry, France)
Joint work with Bernt Oksendal and Agnès Sulem

Wednesday June 4, 12:00-12:30 | session 4.2 | Portfolio Optimization | room CD

We consider a general jump-diffusion model (with possibly infinite activity jumps) and study the viability of the financial market, defined as the ability to solve a portfolio optimization problem. Our main goal consists in characterizing the notion of market viability in terms of martingale measures, in a sense to be made precise in the following, studying under which conditions the (normalized) marginal utility of terminal wealth gives rise to a martingale measure. We refrain from a-priori imposing no-arbitrage restrictions on the model and, moreover, we suppose that market participants have only access to a partial information flow.
In order to solve portfolio optimization problems under partial information, we employ necessary and sufficient maximum principles for stochastic control problems under partial information. This approach allows us to characterize the optimal solution via an associated BSDE, which in turn requires a good control on the integrability properties of the processes involved. Since such integrability conditions are not satisfied in general, a localization procedure is needed.
Our main contributions can be outlined as follows:
1) we show that the financial market under partial information is locally viable, in the sense that a portfolio optimization problem admits a solution up to a stopping time, if and only if there exists a partial information equivalent martingale measure (PIEMM) up to a stopping time. Moreover, the density of such PIEMM is given by the (normalized) marginal utility of the optimal terminal wealth, thus recovering the classical result of financial economics;
2) we prove that, if the financial market under partial information is globally viable, in the sense that it is locally viable for a sequence of stopping times, then there exists a partial information local martingale deflator (PILMD). Furthermore, such a PILMD can be constructed by aggregating the densities of all local PIEMMs if and only if the locally optimal portfolios satisfy a consistency condition. In the special case of bounded coefficients, this PILMD is actually the density process of a PIEMM on the global time horizon;
3) by means of a simple example we show that, even for regular utility functions and continuous processes with good integrability properties but unbounded coefficients, a PIEMM may fail to exist globally but, nevertheless, a PILMD exists.

Optimal Portfolios under Affine Models with Markov Switching
Daniela Neykova (Technical University Munich, Germany)
Joint work with Marcos Escobar and Rudi Zagst

Wednesday June 4, 12:30-13:00 | session 4.2 | Portfolio Optimization | room CD

We consider a stochastic volatility financial model where the asset price process and the volatility process depend on an observable Markov chain. The processes for the asset price and for the stochastic volatility exhibit an affine structure. We are faced with a finite time investment horizon and derive optimal dynamic investment strategies that maximize the investor’s expected utility from terminal wealth. To this aim we apply Merton’s approach, i.e. we solve the HJB equations, which in our case correspond to a system of coupled non-linear PDEs. Exploiting the affine structure of the model, we derive simple expressions for the solution in the case with no leverage, i.e. no correlation between the Brownian motions driving the asset price and the stochastic volatility. In the presence of leverage we propose a separable ansatz, which allows us to reduce this case to the one with no leverage. General verification results are also proved. The results are illustrated with the example of the Markov modulated Heston model.