Uncovering heterogeneity
Observable and non-observable factors explaining farmer preferences for label design
2026-05-26
Overview
Motivation
- Policy context
- Label rationale
- Research questions
Methods
- DCE design
- Attributes & levels
- Subjective beliefs
Results
- Main effects & WTA
- Land shortfall
- Belief heterogeneity
Outlook
- Next steps
- Policy implications
Motivation
Transition to biodiversity-friendly food systems
- EU policy push: CAP, Green Deal, Biodiversity & Farm to Fork Strategy
- Target: at least 10% of farmland as valuable habitat features
- Gap: farmers need incentives beyond existing agri-environment schemes
Key idea
Attract additional private funding for dark green measures by making farmer performance visible to consumers, thereby linking sustainability to market advantages.
Approach
Investigating label-based incentives for farmers
Can labelling increase farmer uptake of dark green agri-environmental measures?
Two potential mechanisms:
- Visibility — farmer performance made transparent via a biodiversity label
- Market returns — higher prices or better purchase conditions for labelled products
We investigate whether these mechanisms increase the attractiveness of dark green measures and what institutional design features farmers actually prefer.
Previous work
Prior research in contracts2.0
- value chain perspectives in designing new labels (Schulze et al. 2024)
- producers, labelling organisations, retailers
- ideal types of labelling schemes
- consumer preferences for hypothetical biodiversity labels
- DCE with 12.000 consumers in 6 countries
- milk and flour differently labelled, combined with a fair payment attribute
Approach here
Rationale of label
Rationale of label
Research questions and storyline
Research question 1
What are farmer preferences for different institutional features of labelling schemes?
Research question 2
How do preferences differ with respect to the amount of land missing to fulfill label criteria?
Research question 3
How do subjective probabilities towards label premia explain preference heterogeneity?
DCE Setup
Survey & sample
- DCE survey with farmers in Germany, Poland, Spain and UK
- Prior qualitative interviews to identify dark green measures per region
- Surveys measured current land allocation → identifies who qualifies already
Label requirement
Farmers must implement dark green AECM on at least 10% of their farmland.
Questionnaire flow:
- Introduction
- Land allocation to dark green measures
- Subjective probabilities on price premia (token allocation task)
- DCE — 8 choice scenarios
- Debriefing
DCE Setup
Attributes, levels & example choice card
| Attribute | Levels |
|---|---|
| Accreditation of the labelling mechanism | Government; private bodies; nature conservation |
| Public awareness for labelled products | Government campaign; retail endorsement; marketing agency |
| Advisory services | Technical; administrative; both |
| Price premium | 5–20% |
Experimental design
Questionnaire & subjective beliefs
Theoretical motivation
Farmers’ willingness to participate in a labelling scheme depends not only on its institutional design, but also on how much of a price premium they realistically expect, making subjective beliefs about market returns a key moderator of label preferences.
Token allocation task
Please distribute a total of 100 points to indicate how likely you consider different price changes to be.
- The more points you assign to an option, the more likely you consider it.
- If you think a price change is unlikely, leave “0” in that field.
- Example: If you believe a 10% price increase is most likely, assign the most points there. If a 40% increase is unrealistic, assign it 0.
Results
DCE main effects
What are farmer preferences for different institutional features of labelling schemes?
Model run by cschulze using Apollo 0.3.5 on R 4.5.0 for Windows.
Please acknowledge the use of Apollo by citing Hess & Palma (2019)
DOI 10.1016/j.jocm.2019.100170
www.ApolloChoiceModelling.com
Model name : MXL_log_DE
Model description : MXL model of German label data in wtp space with log normal distribution
Model run at : 2026-04-22 15:40:42.074042
Estimation method : bfgs
Model diagnosis : successful convergence
Min abs eigenvalue of Hessian : 4.685937
Number of individuals : 116
Number of rows in database : 1044
Number of modelled outcomes : 1044
Number of cores used : 5
Number of inter-individual draws : 2500 (sobol)
LL(start) : -1122.61
LL at equal shares, LL(0) : -1146.95
LL at observed shares, LL(C) : -1143.77
LL(final) : -864.27
Rho-squared vs equal shares : 0.2465
Adj.Rho-squared vs equal shares : 0.2308
Rho-squared vs observed shares : 0.2444
Adj.Rho-squared vs observed shares : 0.2286
AIC : 1764.54
BIC : 1853.65
Estimated parameters : 18
Time taken (hh:mm:ss) : 00:03:38.12
pre-estimation : 00:00:58.23
estimation : 00:00:49.96
post-estimation : 00:01:49.93
Iterations : 38
Unconstrained optimisation.
These outputs have had the scaling used in estimation applied to them.
Estimates:
Estimate s.e. t.rat.(0) Rob.s.e.
mu_asc -1.74801 0.3302 -5.2933 0.3934
mu_cert_ext -1.16551 0.2550 -4.5703 0.2550
mu_cert_nat -1.03893 0.2685 -3.8687 0.2875
mu_market_retail -0.07625 0.1787 -0.4266 0.1761
mu_market_special 0.15863 0.1654 0.9593 0.1731
mu_advice_tech 0.53852 0.2279 2.3629 0.2570
mu_advice_admin 0.68131 0.2336 2.9162 0.2262
mu_advice_all 0.79447 0.2541 3.1260 0.2566
mu_log_premium 2.35577 0.1406 16.7610 0.1752
sigma_asc 2.01039 0.2946 6.8230 0.2817
sigma_cert_ext 1.34063 0.2855 4.6957 0.3356
sigma_cert_nat -1.67811 0.2882 -5.8229 0.2978
sigma_market_retail -0.72203 0.2458 -2.9376 0.2367
sigma_market_special -0.13056 0.4285 -0.3047 0.1972
sigma_advice_tech -0.70971 0.3244 -2.1879 0.2935
sigma_advice_admin 0.79064 0.3391 2.3318 0.3192
sigma_advice_all 0.11420 0.3766 0.3032 0.1419
sigma_log_premium 0.90340 0.1217 7.4201 0.1258
Rob.t.rat.(0)
mu_asc -4.4436
mu_cert_ext -4.5712
mu_cert_nat -3.6138
mu_market_retail -0.4330
mu_market_special 0.9166
mu_advice_tech 2.0956
mu_advice_admin 3.0118
mu_advice_all 3.0963
mu_log_premium 13.4428
sigma_asc 7.1358
sigma_cert_ext 3.9943
sigma_cert_nat -5.6353
sigma_market_retail -3.0503
sigma_market_special -0.6621
sigma_advice_tech -2.4184
sigma_advice_admin 2.4767
sigma_advice_all 0.8046
sigma_log_premium 7.1808Results
DCE main effects
What are farmer preferences for different institutional features of labelling schemes?
Key finding
Government certification is most preferred; private certification bodies require the highest compensation. Advisory services(especially when combined) significantly increase label attractiveness.
Results
DCE main effects - WTA
What are farmer preferences for different institutional features of labelling schemes?
term mean median sd
WTA_asc_mean WTA_asc_mean 10.8612650 10.8335452 2.165888
WTA_cert_ext_mean WTA_cert_ext_mean 7.3393565 7.2690886 1.635043
WTA_cert_nat_mean WTA_cert_nat_mean 6.4760500 6.4399650 1.781002
WTA_market_retail_mean WTA_market_retail_mean 0.4925488 0.4726792 1.126163
WTA_market_special_mean WTA_market_special_mean -0.9880968 -0.9744659 1.095997
WTA_advice_tech_mean WTA_advice_tech_mean -3.3709914 -3.3509136 1.650914
WTA_advice_admin_mean WTA_advice_admin_mean -4.2948518 -4.2387034 1.478868
WTA_advice_all_mean WTA_advice_all_mean -5.0069378 -4.9360008 1.719818
p2.5 p97.5
WTA_asc_mean 6.625476 15.143693
WTA_cert_ext_mean 4.307639 10.837473
WTA_cert_nat_mean 3.151032 10.126570
WTA_market_retail_mean -1.650108 2.783566
WTA_market_special_mean -3.220300 1.157978
WTA_advice_tech_mean -6.686898 -0.173242
WTA_advice_admin_mean -7.356164 -1.508565
WTA_advice_all_mean -8.607138 -1.806447Results
DCE main effects - WTA
What are farmer preferences for different institutional features of labelling schemes?
Key finding
Farmers require a 7% price premium to accept private over government certification. Combined advisory support is valued at 5%, being the largest single driver of WTA after certification.
Results
DCE main effects - choice probabilities
What are farmer preferences for different institutional features of labelling schemes?
Predicted posterior choice probabilities
Results
DCE land interactions
How do preferences differ with respect to the amount of land missing to fulfill label criteria?
Predicted posterior choice probabilities with shortfall interactions
Results
DCE subjective probabilities interactions
How do subjective probabilities towards label premia explain preference heterogeneity?
Approach to integrate subjective probabilities
Rationale
Farmers who expect higher price premia should find labelling schemes more attractive. We test whether belief-based heterogeneity explains residual preference variation beyond land shortfall.
Results
DCE subjective probabilities interactions - mean
Predicted posterior choice probabilities with mean probability interactions
Results
DCE subjective probabilities interactions - standard deviations
Predicted posterior choice probabilities with sd probability interactions
Results
DCE subjective probabilities interactions - optimism
Predicted posterior choice probabilities with negative probability interactions
Next steps
- finalise model interactions
- repeat analysis across other countries
- pooled analysis?
- discuss findings derive adequate policy implications
References
Hybrid finance for agri-environmental climate measures