Partial stochastic analysis with the Aglink-Cosimo model : a methodological overview

Aglink-Cosimo is a recursive-dynamic partial equilibrium model developed and maintained by the Organisation for Economic Co-operation and Development (OECD) and the Food and Agriculture Organization of the United Nations (FAO) Secretariats as a collaborative effort. The model is primarily used to prepare the OECD-FAO Agricultural Outlook, a yearly publication aiming to provide baseline projections for the main global agricultural commodities over the medium term. These deterministic projections are enhanced by a partial stochastic analysis tool, which allows the analysis of specific market uncertainties. This is done by producing counterfactual scenarios to the baseline originating from varying yields and macroeconomic variables stochastically. The aim of this report is to propose and evaluate different methods of analysing stochastically important yields and macroeconomic uncertainty drivers. In the first stage, we identify and evaluate the best parametric method to extract unexplained variability, which we consider to be uncertainty in the macroeconomic and yield drivers. In the second stage, we test parametric and non-parametric methods side by side to simulate 10 years of potentially different macroeconomic and yield environments. The results can be summarised as follows. For yields, we find that a parametric cubic trend method performs best in the first stage and a non-parametric hierarchical copula (Clayton) method is more appropriate in the second stage. For macroeconomic variables, a vector autoregressive model performs best in the first stage, while a non-parametric hierarchical copula (Frank) method is more appropriate in the second stage.