Agricultural policy informed by farmers’ adaptation experience to climate change in Veneto, Italy

Abstract

There is an increasing call for agricultural water management to adapt to climate change, yet efforts in this direction often consider only the policy dimension, or planned adaptation perspective. However, it is crucial to include an assessment of farmers’ autonomous adaptation into the design and evaluation of rural policy measures. Amongst others, this helps avoid doubling efforts and ensure the effectiveness of the policies proposed. Moreover, farmers are the primary receivers of climate-proofing agricultural policies. Hence, to maximise a policy’s success, it is fundamental to include farmers in its design phases. Farmers autonomously react to changes and policies should build on ongoing efforts. This work, carried out in the Veneto Region of Italy, proved the advantages of approaching adaptation as a continuum between autonomous and planned, rather than addressing the two separately. We first collected farmers’ perceptions of and adaptations to change through an online questionnaire. We then identified the major determinants of their choice to adapt through a multinomial probit model. We analysed farmers’ expectations of effectiveness of five different adaptation options for water conservation, via an ad-hoc online decision support system tool, mDSSweb. Our work provided policy makers with information on how different typologies of farmers are (not) adapting their practices to climate change. We clearly identified which groups of farmers the policies should target first and with what type of support. Both policy makers and farmers reacted positively to our approach and expressed interest in up-scaling it to become more inclusive.

Appendix: The MNP model

Suppose that a farmer i chooses alternative k and therefore, the difference between the latent variable and the $J-1$ is defined as follows:

$$\begin{aligned} v_{ijk}=y_{ij}-y{ik}=x_{i}(\beta _{j}-\beta _{k})+\epsilon _{ij}-\epsilon _{ik}=x_{i}\theta _{j}+u_{ij} \end{aligned}$$

(3)

Where $j'=j$ if $j < k$ and $j'=j-1$ if $j >k$ such that $j'=1,\ldots ,J-1$. The variance and covariance of the error terms $u_{ij}$ are defined as $Var(u_{ij})=Var(\epsilon _{ij}-\epsilon _{ik})=2$ and $Cov(u_{ij},u_{im})=1$ for $j'\ne m'$. The probability that alternative $k$ is chosen is given by the following equation:

$$\begin{aligned} Pr(i \,{\mathrm{chooses }}\, k)= Pr(v_{ilk} \le 0,\ldots ,v_{i,J-1,k} \le 0) = Pr(u_{il} \le -x_{i}\sigma _{l},\ldots ,u_{i,J-1} \le -x_{i}\sigma _{J-1}) \end{aligned}$$

(4)