A Multi-objective Bi-level Optimisation model for Agricultural Policy in Scotland
Ververidis, Konstantinos Athanasiou
Agricultural policy analysis can be visualised as a multiple objective hierarchical optimisation problem whereby sequential non-cooperative interactions between the policy makers and the farmers take place. The objectives and choices of policy makers will almost always diverge from the objectives and choices of farmers. Policy makers exercise authority over some, but not all, of the variables in the total system whereas other variables affecting their multiple goals are under the direct control of myriad farmers who operate according to their own utility maximising motives. In order to advance their own objectives, the policy makers unilaterally and pre-emptively set the policy measures to influence the farmers. The farmers execute their decisions after, and in view of, the policies and make their production decisions that observe their goals best. Ultimately, the payoffs to both the policy makers and the farmers depend not only on the actions of the former, but also on the reactions of the latter. Such problems are difficult to solve due to their intrinsic nonconvexity and multiple objectives. This thesis shows how multi-objective genetic algorithms (MOGA) in conjunction with mathematical programming (MP) can be used for solving this type of problems. A MP model is developed to capture the production choices of farmers. The model is based on positive mathematical programming and its objective function parameters are estimated using the method of generalised maximum entropy. The model is nested in and controlled by a MOGA which captures the process of multi-objective optimisation of policy decisions. The approach is illustrated using a case study taken from the Scottish agricultural systems, where several socio-economic and environmental objectives for policy making are considered. Four types of policy instruments are examined: the current single payment scheme, a multi-payment scheme based on land use, an input taxation and a regulatory scheme. For a selection of scenarios alternative Pareto-optimal solutions are discovered and tradeoffs between the policy objectives are presented along with their associated production patterns. The performance of the modelling tool developed suggests that it is well suited to dealing with real-world policy issues. It offers considerable possibilities for exploring tradeoffs between non-commensurable and conflicting objectives relevant to sustainable development of Scottish agriculture.