An extension of Arrow's result on optimality of a stop loss contract