678 / COMMUNICATIONS Am. J. Agr. Econ.

ple MAD for suboptimal decisions, a matter of some importance in a world where individual utility functions are but vaguely located. Consequently, while sympathetic to Ang's desire to extend the analysis in this way, we consider his suggestion both impractical and too narrowly conceived, although we are grateful for his comment that between the (estimated) optimal points, the expected utility loss may be less than the loss estimated by us. On a more specific level, we wish to disagree with Ang on two counts. First, the set of efficient, E, V plans selected with the sample MAD need not always be less efficient than the set selected with the sample variance. It is quite possible for the two estimated frontiers to intersect, sometimes repeatedly, and our simulation results also show that the sample MAD sometimes performs consistently better than the sample variance when income distributions are asymmetric. However, we would like to point out that the efficiencies dealt with in our paper concern the determination of the plans underlying these frontiers by comparison, through rival techniques, of two random variables, one distributed according to the plan on the frontier, the other distributed according to a plan above the frontier. We were not concerned with the positions of the frontiers per se, for as pointed out in our note [1], the estimation involved here "may make use of any estimation procedure considered desirable" once the "best" plan (in terms of activity levels) has been settled on. Secondly, we consider that our results are applicable for all feasible values of E, including extreme