Downside risk minimization via a large deviations approach (Q417076)

The author deals with the problem of minimizing the downside risk associated with an investor's total wealth in a certain incomplete market model. More precisely, consider a market model with one riskless asset, \(m\) risky assets and \(n\) factors, the prices of the assets at time \(t\) being \(S^0(t),S^1(t),\dots, S^m(t)\). It is assumed that the bond price is governed by the ordinary differential equation \[ dS^0(t) = r(X_t)S^0(t)\,dt,\quad S^0(0)= s^0. \] The other security prices and the factors are assumed to satisfy the stochastic diffential equations \[ dS^i(t)= S^i(t) \Biggl[\alpha^i(X_t)\,dt+ \sum^{n+m}_{k=1} \sigma^i_k(X_t)\,dW^k_t\Biggr],\quad S^i(0)= s^i \] and \[ dX_t=\beta(X_t)\,dt+ \lambda(X_t)\,dW_t,\quad X(0)= x, \] where \(W^t= (W^k_t)\) \((k= 1,\dots,n+ m)\) is a standard \((m+n)\)-dimensional Brownian motion. The total wealth of the investor is given by \[ V_t= N^0_t S^0_t+\cdots+ N^m_t S^m_t. \] Here, \(N^i_t\) is the number of shares of security \(i\) and \(h^i_t= N^i_t S^i_t/V_t\) is the proportion of the portfolio invested in the \(i\)th security. It is assumed that the strategy \(h_t= (h^1_t,\dots, h^m_t)\) is self-financing (noting that \(h^0_t= 1-(h^1_t+\cdots+ h^m_t)\)). Let \({\mathcal H}(T)\) denote the set of all strategies \(h(t)\) \((0\leq t\leq T)\) which are \({\mathcal G}_t\)-progressive measurable such that \[ \int^T_0|h(s)|^2 ds<\infty\quad\text{almost surely}, \] where \({\mathcal G}_t= \sigma(S(u), X(u), u\leq t)\). For a given constant \(\chi\) the author studies the asymptotics of minimization of a downside risk \[ J(\kappa) := \liminf_{T\to\infty} T^{-1} \text{inf}_{h\in{\mathcal H}(T)}\log P(T^{-1}\log(V_T(h)/S^0_T)\leq \kappa).\tag{\(*\)} \] By considering the dual problem of (\(*\)), i.e., \[ \widehat\kappa(\gamma):= \liminf_{T\to\infty} T^{-1} \text{inf}_{h\in{\mathcal A}(T)} J(v,x;h,T)\tag{\(**\)} \] (where \(J(v,x;h,T):= \log E(\exp(\gamma\log(V_T(h)/S^0_T))]\) and \({\mathcal A}(T)\) is the set of all admissible strategies), the author obtains a strategy \((\widehat h_t(\kappa,T))\) such that \[ J(\kappa)= \lim_{T\to\infty} T^{-1}\log P(T^{-1}\log(V_T(\widehat h(\kappa, T)/S^0_T))\leq \kappa) \] which finally gives that \[ J(\kappa)= -\sup_{\gamma< 0} (\gamma\kappa- \widehat\kappa(\gamma)). \]