Water Quality, Policy Diffusion Effects and Farmers’ Behavior

The nitrogen cycle is one of the most disrupted geo-chemical cycles on earth. Human activity, mainly through intensive farming, releases nitrogen by-products such as nitrates and ammonium into the environment where they have wide ranging impacts on human health, biodiversity and climate change. One of the earliest and most ambitious regulations of nitrogen use in the world is the EU Nitrate Directive, which not only sets limitations on the amount and timing of nitrogen application but also makes the adoption of modern nitrogen management tools mandatory in an effort to enhance nitrogen use efficiency. We leverage the geographical and temporal variation in the implementation of the Nitrate Directive in France to estimate its causal effects on water quality, biodiversity and farmers’ practices, productivity and profits in a Difference In Difference (DID) framework. We modify the DID estimator to account for the existence of diffusion effects along river streams, leveraging recent developments in the analysis of Randomized Controlled Trials over a network of interrelated units. This is a methodological extension that can be of interest for similar applications. We find that the EU Nitrate Directive reduced the concentration of nitrates in surface water by 1.23 milligrams per liter: a decrease of 8%. We find a clear dose-response relationship, with higher impacts where more of the upstream area is covered by the Directive. We also find that other biochemical indicators, as well as biodiversity, as measured by the number of fish and fish species, also improved as a result of the Directive. We also find that the Directive managed to improve farmers’ nitrogen use efficiency and productivity and did not decrease their profits. These findings are consistent with the Porter hypothesis. Finally, we show that not accounting for diffusion effects biases downwards the estimate of the effect of the Directive obtained with a classical DID estimator and the more recent geographic discontinuity estimator.