Most of ex-ante impact assessment policy models have been based on a parametric approach. We develop a novel non-parametric approach, called Inverse DEA. We use non parametric efficiency analysis for determining the farm’s technology and behaviour. Then, we compare the parametric approach and the Inverse DEA models to a known data generating process. We use a bio-economic model as a data generating process reflecting a real world situation where often non-linear relationships exist. Results suggest that traditional parametric approaches are biased and inconsistent. The Inverse DEA model under variable return to scale preserving technical efficiency scores outperforms any other specifications. However such non-parametric approach is by nature sensitive to noise which hampers its accuracy when it prevails. The use of panel data is preferable.