Abstract
This work presents a study about dynamics of territorial gaps in agriculture between EU NUTS 2 and 3 from three perspectives: agricultural production model, rural development and productivity. Our choice of methodology is driven by two considerations. First of all, different performances in agriculture of EU areas may lead to multimodal distributions of indicators, and this implies that no significant average behaviour a parametric estimates refers to can be identified. Secondly, even if one wants to highlight multimodality within distributions, techniques like univariate density estimates can not catch intra-distribution dynamics which lead to the formation of "poles" of units showing a "similar conduct". Due to this reasons, we adopt a statistical approach which aims to point out the main features of a whole distribution, rather than to focus on a single parameter estimate. The non parametrical statistical technique of stochastic kernel (Quah, 1999) allows to detect the underlying law of motion of a distribution, as well as the influence of a "conditioning factor" on it following a regression-like rationale.