Mean-variance hedging under transaction costs (Q2460042)

An incomplete financial market is considered with a non-risky assets and \(d\) risky ones. There are proportional transaction costs denoted \((\lambda_i,\mu_i,i= 1,\dots, d)\). Only self-financing strategies are used. At terminal time \(T\), the global risk to be minimized for the contingent claim is \[ E[(H- x- G_T(\theta))^2], \] where \(x\) is the initial wealth, \(\theta^i= l^i\) if \(i\)th is bought, \(m^i\) if it is sold, the terminal gain is \[ G_T(\theta)= \sum^{T-1}_{t=0} [-\langle(1+ \lambda) l_t, S_t\rangle+ \langle(1- \mu) m_t, S_t\rangle]+ \Biggl\langle \sum^{T-1}_{t= 0} l_t, S_t\Biggr\rangle- \Biggl\langle\sum^{T-1}_{t= 0} m_t, S_T\Biggr\rangle. \] A sufficient condition for arbitrage opportunity absence is given, namely ``no free lunch in \(L^2\) condition'': \(\forall C_1> 0\), \(\exists C_2\in ]0,C_1[\) such that for any strategy \(\theta\) satisfying \(\| G_T(\theta)\|_2\geq C_1\), then \(G^-_T(\theta)\geq C_2\). Then, under suitable conditions on the price processes, the set \(\{G_T(\theta)\), \(\theta\) admissible\} is closed in \(L^2\), thus there exists for every contingent claim \(H\) and every initial wealth \(x\) a global risk minimizing strategy under transaction costs.