Asymptotics for fixed transaction costs (Q2339123)

The purpose of this paper is to study the effect of small transaction costs on the trading, instead of proportional costs as it is usually assumed. The market presents a safe asset \(S^0_t= e^{rt}\) and \(d\) risky assets, the prices of which being driven by a \(d\)-dimensional standard Brownian motion. Fixed transaction costs mean that, at trading stopping time \(\tau_i\), the position \((x,y)\in\mathbb{R}\times\mathbb{R}^d\) goes to \((x- \sum^d_{j=1} m^j_i-\lambda,y+m)\), \(y,m_i\in\mathbb{R}^d\), where \(m_i\) is the trade at time \(\tau_i\) and \(\lambda\) is the fixed transaction cost. With a consumption rate \(c\), a strategy is defined as triplet \(\nu= (c, (\tau_i, i\geq 1),\, (m_i, i\geq 1))\). Such a strategy is said to be admissible if the derived positions \((X^{\nu,x}_t, Y^{\nu,x}_t)\) remain at all times in the solvency region: either \(x+ \sum^d_{j=1} my^j- \lambda\geq 0\) or \(\min_{i=1,\dots,d}\{x, y^i\}\geq 0\). The utility function \(U^\gamma:c\mapsto{1\over 1-\gamma} c^{1-\gamma}\), \(0<\gamma< 1\), or \(\log c\), leads to the optimization problem \((c,\tau,m)\mapsto E[\int^\infty_0 e^{-\beta t} U^\gamma(c_t)\,dt]\) on the set of admissible strategies. After heuristic considerations, the main results are proved using, among other methods, the general methodology developed by \textit{G. Barles} and \textit{B. Perthame} [RAIRO, Modélisation Math. Anal. Numér. 21, 557--579 (1987; Zbl 0629.49017)] and \textit{L. C. Evans} [Proc. R. Soc. Edinb., Sect. A, Math. 111, No. 3--4, 359--375 (1989; Zbl 0679.35001)] (viscosity solutions). Firstly an explicit expansion of the value function with respect to \(\lambda\) is given. Secondly, an almost optimal policy is defined according to a no-trade region. Unlike for proportional costs where there is an infinite number of small `local time type' trades, fixed transaction costs lead to finitely many bulk trades over finite time intervals.