RECURSIVE STOCHASTIC-PROGRAMMING APPLIED TO DAIRY-COW REPLACEMENT

Several models have been used to determine optimal replacement policies for dairy herds in grassland and non-grassland situations. The dynamic programming models reviewed do not model individual cows but model a series of states which are compared at each stage to determine an optimal replacement policy. As a consequence, it is not possible to reconcile the predicted net present values of future income streams with actual expected cash flows from cows in the herd. A recursive stochastic programming model to rank cows based on future profitability was developed to overcome these problems. This approach should allow the farmer to realize the expected net present values of income streams and, thus, should be applicable to dairy herds in general. The model was compared with the payment production index for the culling of cows. The recursive stochastic model gave an increase, per 1% culled within herd, in the net present value per cow of NZ$1.80 which out-performed culling on the payment production index by NZ$1.10. © 1990.