Eliciting farmers' preferences for agricultural policy reform
How to cite
Drichoutis, Andreas; Nacka, Marina; Bazzani, Claudia; Dimitrievski, Dragi. 2025. "Eliciting farmers' preferences for agricultural policy reform." AEA RCT Registry. AEARCTR-0017024. https://doi.org/10.1257/rct.17024-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This study investigates farmers' preferences for alternative agricultural support schemes in North Macedonia using a discrete choice experiment (DCE). The research aims to inform policy design during the country’s alignment with the European Union’s Common Agricultural Policy. in the DCE, farmers face trade-offs between complex but more generous direct payment systems and simplified schemes that reallocate part of the agricultural budget toward rural development. The DCE presents respondents with 15 choice tasks, each consisting of two hypothetical new programs and the current national scheme. Programs vary along six key attributes: share of funds allocated to rural development, type of direct payments (from fully coupled to fully decoupled), reliability and administrative burden, level and scope of investment aid, environmental conditionality, and change in total annual support relative to the status quo. The design is Bayesian D-efficient, informed by a pilot with active farmers. The analysis will estimate preference structures and heterogeneity across farmers, focusing on how simplification, reliability, and the degree of coupling influence support for different policy directions. The findings will provide evidence to guide ongoing subsidy reforms and the balance between income support, investment incentives, and rural development priorities in agricultural policy.
Intervention
The intervention consists of a discrete choice experiment (DCE) in which farmers evaluate alternative agricultural policy programs. Each program is described by six attributes with multiple levels:
- Share of rural development budget: 10% (status quo), 25%, 35%
- Type of direct payments: Fully coupled (current system), partially decoupled simplified, fully decoupled
- Reliability and administrative burden: Low (often late; ≈40h/year, status quo), Medium (most payments by June; ≈20h/year), High (all payments by June; ≈8h/year)
- Investment aid (grants): Up to 50% support (current budget), up to 50% support (higher budget), up to 60% support
- Environmental conditionality: Minimum legal rules, basic cross-compliance, eco-scheme
- Change in total annual support: –20%, –15%, –10%, –5%, 0%, +5%, +10%, +15%, +20%
The full factorial design results in 3*3*3*3*3*9 = 2187 possible combinations per policy alternative, which makes a complete enumeration infeasible. To achieve statistical efficiency while keeping respondent burden reasonable, a Bayesian D-efficient design was used to select a balanced subset of 15 choice tasks per respondent.
In each task, farmers choose one option among two hypothetical new policy programs (A and B) and the current policy (status quo), which remains constant across all tasks. All participants complete the same 15 optimized choice tasks, allowing the estimation of trade-offs and preference heterogeneity across the six policy dimensions.
Experimental design
The study employs a discrete choice experiment (DCE) to elicit farmers’ preferences for alternative agricultural support programs in North Macedonia. Each respondent completes 15 choice tasks, each containing two hypothetical policy programs (A and B) and the current national policy (status quo). The six attributes and their levels are described in the intervention section.
The design is Bayesian D-efficient, developed based on priors from a pilot study with 21 farmers to improve precision and reduce cognitive burden. All respondents face the same optimized set of 15 tasks, ensuring comparability across individuals.
The sample will consist of active farmers recruited through regional agricultural offices and farmer associations. Participation occurs in controlled laboratory or field settings, where respondents complete the survey individually under supervision.
Randomization occurs at the level of choice task presentation (order randomized across respondents) and option placement (left/right randomization of Programs A and B) to avoid ordering or positional bias.
The analysis will estimate mixed logit and latent class models to capture heterogeneity in preferences and derive willingness-to-pay measures for improvements in reliability, administrative simplicity, investment aid, and rural development funding relative to total payment changes.
Design details
Cited by
No citation data available for this registration yet.