Résumé
Agri-environmental schemes (AES) are a central component of the European environmental policy, but few of these schemes have been carefully evaluated and doubts are often expressed about their effectiveness. We use original data collected from winegrowers who participated in an AES targeting non-point source pollution from herbicides in 2011 and 2012 in the South region of France. Using the variation in the implementation of the scheme across time and space and a matching approach, we show that the quantity of herbicides used by participants in the scheme in 2011 ranges from 38 to 53 \% below what they would have used without the scheme and this percentage is between 42 and 50 \% in 2012. Further, our results suggest that least demanding AES options are effective in avoiding pollution peaks when weed pressure is high, whereas more demanding AES options guarantee an overall reduction in herbicide use, even during relatively easy farming years in which less weed pressure is experienced.