The formulation of MOTAD programming models for farm planning using subjectively elicited activity net revenue distributions

If activity net revenue distributions for farm planning under risk are subjectively assessed, the claimed advantages of MOT AD (Minimisation Of Total Absolute Deviation) programming can be reaped only if some way can be found of generating discrete vectors of net revenue deviations. Pseudo-random sampling is evaluated for this purpose and it is shown that, even with relatively large sample sizes, there is an appreciable risk that the MOTAD formulation will yield solutions some distance from the (E, V)-efficient frontier. An alternative procedure is described based on the use of linear programming to define a discrete distribution of outcomes that has means, variances and co-variances similar to those of the elicited joint normal distribution. It is shown that a MOTAD formulation based on a discrete distribution so derived can closely approximate the (E, V) frontier.