Introduction
Since the book by Thaler and Sunstein1, the concept of nudging has gained increasing attention in behavioural economics and public policy and the environmental and agricultural domains2,3, offering an alternative to traditional regulatory or financial incentives. Nudges are defined as “any aspect of the choice architecture that alters people’s behaviour in a predictable way without forbidding any options or significantly changing their economic incentives”1, p. 6. Nudges can be considered ethical insofar as they preserve individual autonomy while aiming to correct behavioural market failures4.
With the demand for green solutions on the rise, many EU projectsi and initiatives explicitly address the efficacy and efficiency of Green Nudges as a means of enhancing the adoption of Green Practices in agriculture and forestry. Efficacy refers to the achievement of the intended behavioural shift towards the desired practice by a specific nudge, while efficiency imposes further desiderata like low implementation costs and compliance with specific ethical rules. Ethics-related desiderata call for a cautious attitude towards manipulative nudges which target automatic or subconscious reactions (System 1) by the decision makers, rendering them passive and unaware receptors of the intervention. In the opposite case, nudges which make the decision maker aware of the intervention and its objectives can be considered more desirable, as they achieve their objectives through the activation of more explicit cognitive processes (System 2) leading to the conscious and presumably long-lasting adoption of the desired practices.ii This note reports on a nudge with adverse effects, arguing that informative nudges and information in general may hinder the adoption of the desired (green) practice. Furthermore, it opens the black box of the contextual and individual sources of such adverse informational effects.
The global shift toward sustainable food systems has pushed agroecology as a key lever for the agricultural industry. However, the transition towards agroecological practices often places farmers at the centre of a trade-off: on one hand, the moral call to adopt environmentally responsible practices; on the other, the economic and practical risks associated with deviating from conventional farming methods. Ethical tensions may arise when transition to agroecological practices is pushed in the light of social and political pressure, without adequately addressing the farmers’ structural constraints and value systems5,6.
Informational nudges represent a significant subset of behavioural interventions7. These nudges appeal to System 2 thinking, promoting more reflective and deliberate choices8. They differ from System 1 nudges, which often rely on affective or subconscious stimuli9. Informational nudges are often applied in domains where it is crucial to increase awareness, such as sustainability and pro-environmental behaviours10,11. The efficacy of informational nudges may depend on individual and contextual factors. Socio-demographic characteristics of farmers such as age, education, and gender have been shown to influence receptiveness to sustainability messages12. Likewise, farm characteristics like size or farming system has shown to affect how farmers interpret and act upon information13.
While specific nudging interventions have been successful in many fields, their success in agricultural contexts has not always been straightforward. As farmers base their decisions on a combination of rational, habitual, social, and emotional factors, informational nudges may lead to unintended consequences or boomerang effects7,14.
Methodology
Data collection
This study analyses data collected from a haphazard sample of farmers across eight countries in Europe—Greece, Cyprus, Italy, Spain, Portugal, Serbia, Latvia, and the Netherlands—through 16 Living Labs involved in farming activities on the following crops: apple, grapes, plum, cherry, citrus, legumes, maize, olives, onion, rice, rye, peas, soybean, triticale and winter wheat (see Appendix A in Extended Data, including the questionnaire of this study). A total of 405 observations were collected through a survey by questionnaire (ethical approval CERBSB2023-56). Informed consent for participation in the study was obtained, and individuals who indicated that they did not wish to take part in the survey were automatically screened out (see Appendix A, Q1 in Extended data). The part of the questionnaire analysed for this study included two sections.
The first section focused on socio-demographic characteristics of farmers (age, gender, education, year of experience in farming), farm structure and farming methods. Table 1 describes the sample according to these characteristics.
Table 1. Socio-demographic and farming characteristics of the sample (n=405).
| Variables | Mean (Std.
Dev.) | Min | Max |
|---|
| Male | 0.81 | 0 | 1 |
| Age (in years old) | 50 (13.14) | 24 | 84 |
| Education | | | |
Up to secondary school
degree | 0.23 | 0 | 1 |
| High school degree | 0.33 | 0 | 1 |
| University degree | 0.44 | 0 | 1 |
| Conventional farming methods | 0.63 | 0 | 1 |
| Years of activity | 20.9 (18.9) | 1 | 124 |
| Family labor in farming | 1.55 | 0 | 6 |
| Farming methods (hectares) | | | |
| Conventional farming | 28.83 (122.41) | 0 | 1,800 |
| Conventional under transition | 1.47 (12.66) | 0 | 215 |
| Organic | 17.32 (101.73) | 0 | 1,205 |
| Biodynamic | 1.87 (34.92) | 0 | 700 |
| Regenerative | 5.16 (61.89) | 0 | 1,000 |
| Agroecological | 2.38 (35.14) | 0 | 700 |
| Other | 2.42 (31.17) | 0 | 600 |
| Production destination | | | |
| Cropland | 29.32 (110.19) | 0 | 1,283 |
| Permanent crops | 8.39 (26.55) | 0 | 290 |
| Permanent grasslands | 13.11 (81.81) | 0 | 1,100 |
| Other | 3.21 (51.07) | 0 | 1,000 |
The second part of the questionnaire explored farmers' intention to reduce herbicide use and adopt AWM practices in the short and long termiii. Then a brief informational paragraph was introduced as a nudge, providing additional details about AWM (see Appendix A, Q9 in Extended data).
After reading this information, farmers were presented with the same statements on their intention to adopt AWM practices.
Econometric analysis
The first part of our analysis employs the Wilcoxon Signed-Rank Test to compare farmers’ stated intentions to adopt environmentally sustainable practices before and after the nudge intervention, both in the short-term and long-term periods. This approach allows us to assess whether the intervention led to statistically significant changes in intention levels, without making distributional assumptions about the data. In addition, we perform subgroup analyses based on farmers’ sociodemographic characteristics (age, gender, education) and farm characteristics (farming methods, family business status, and years of activity) to explore potential heterogeneity in responses to the intervention.
In the second part of our analysis, we estimate two ordered probit models in which the dependent variables are intention to reduce herbicides and adopt AWM practices in the short term, before and after the nudge.
Our general specification for the intention models is given by:
where is a latent continuous variable representing farmer i’s intentions regarding outcome j: reducing herbicide use (j = 1) or adopting AMW practices (j = 2). The vector Xi includes socio-demographic characteristics (age, gender and education), and Zi captures farm characteristics (farming methods, family business status and years of activity). The error term ei is assumed to follow a standard distribution. The observed ordinal outcome Yi,j∈{1,...,7} is linked to the latent variable via a set of threshold parameters, consistent with the ordered probit framework.
Results
Are informational nudges an effective tool for promoting herbicide reduction and adoption of AWM practices?
The results of Wilcoxon Signed-Rank Tests show a significant decline in short-term intention following the nudge (Table 2). In contrast, long-term intentions remained stable. Similarly, general future-oriented intentions showed no difference in mean values before and after the nudge, with medians remaining at 7 and no significant difference detected. These results suggest that the informational nudge had a statistically significant, negative effect on intention to apply AWM in the short run, but no observable impact on intention in the long term.
Table 2. Differences in Farmers’ Short- and Long-Term Intentions to Adopt Sustainable Practices Before and After the Nudge.
| Short term farmers’ intentions | Mean (std.dev.) |
|---|
| Before nudge: I plan to reduce the use of herbicides this year | 5.18 (2.21) |
| After nudge: I plan to use Agroecological Weed Management practices this year. | 4.91 (2.11) |
| Wilcoxon test for pre-posttest in intentions: Mdn_before=7; Mdn_after=5 W=3.109; p=0.01 |
| Long term farmers’ intentions | |
|---|
| Before nudge: I intend to reduce the use of herbicides over the next 5 years. | 5.45 (1.96) |
| After nudge: I intend to use Agroecological Weed Management practices over the next 5 years. | 5.43 (1.81) |
| Wilcoxon test for pre-posttest in intentions: Mdn_before=7; Mdn_after=6; W=0.916, p>0.10 |
| Before nudge: I will regularly try to reduce the use of herbicides in the future | 5.64 (1.89) |
After nudge: I will regularly try to introduce Agroecological Weed Management practices in the
future. | 5.64 (1.72) |
| Wilcoxon for pre-posttest in intentions Mdn_before=7; Mdn_after=7; W= 0.536, p>0.10 |
What is the influence of socio-demographic and farm characteristics on behavioural response to informational nudges?
In order to answer RQ2, Wilcoxon tests were conducted to determine whether there is a statistically significant improvement between pre- and post-nudge intentions to implement AWM practices, based on farmers' socio-economic characteristics and farm characteristics (Table 3).
Table 3. Pre- and post-nudge intentions to reduce herbicide use and adopt AWM in the light of socio-demographic and farming characteristics.
| Variables | Mean
(St. Dev.) | Wilcoxon
signed-rank test
for pre-post test
in intentions z | Obs. |
|---|
| Farming methods | | | |
| If conventional farming | | | |
| Before | 4.38 (2.211) | 1.838 | 249 |
| After | 4.09(1.990) | | |
| If sustainable farming | | | |
| Before | 6.45 (1.529) | 3.354*** | 157 |
| After | 6.21 (1.585) | | |
| Farmers’ education | | | |
| If university degree | | | |
| Before | 5.33 (2.150) | 2.774*** | 181 |
| After | 4.97 (2.130) | | |
| If college degree | | 1.204 | 132 |
| Before | 5.21 (2.255) | | |
| After | 5.07(2.066) | | |
| If secondary school diploma | | | |
| Before | 4.86 (2.282) | 1.244 | 93 |
| After | 4.55 (2.123) | | |
| Farmers’ gender | | | |
| If women | | | |
| Before | 5.87 (1.958) | 2.429** | 74 |
| After | 5.47 (2.021) | | |
| If men | | | |
| Before | 5.03 (2.245) | 2.506** | 331 |
| After | 4.77 (2.114) | | |
| Farmers’ age | | | |
| If ≤50 years-old | | | |
| Before | 5.43 (2.069) | 3.582*** | 211 |
| After | 4.99 (2.056) | | |
| If >50 years-old | | | |
| Before | 4.92 (2.343) | 0.674 | 195 |
| After | 4.82 (2.173) | | |
| Years of activity | | | |
| If ≤16 years | | | |
| Before | 5.68 (1.965) | 3.265*** | 204 |
| After | 5.27 (1.983) | | |
| If >16 years | | | |
| Before | 4.68 (2.347) | 1.066 | 202 |
| After | 4.54(2.178) | | |
| Family business | | | |
| Yes | | | |
| Before | 5.01 (2.247) | 2.226** | 343 |
| After | 4.81 (2.112) | | |
| Non | | | |
| Before | 6.11 (1.797) | 2.750*** | 63 |
| After | 5.41 (2.052) | | |
The results reveal that farmers who practice sustainable farming, have higher levels of education, are younger, and have fewer years in this economic activity reduce their intentions to adopt AWM after the nudge. These differences are only observed in short-term intentions. When looking at long-term intentions, there are no significant differences before and after the informative nudge.
Ordered probit models were applied to highlight the role of socio-demographic characteristics of farmers and farming practices in influencing intentions to reduce the use of herbicides and apply AWM practices in the short term (Table 4).
Table 4. Ordered probit model on the determinants of farmers’ intentions.
| Variables | I plan to reduce the
use of herbicides
this year | I plan to adopt
AWM practices
this year. |
|---|
| Coef. | z | Coef. | z |
|---|
| Age | 0.000 | 0.11 | 0.002 | 0.59 |
| Male | -0.428 | 2.57*** | -0.252 | 1.67* |
| Conventional farming | -1.12 | 8.58*** | -1.27 | 9.86*** |
| Family Labor | -0.532 | 2.78*** | -0.39 | 0.24 |
| College | -0.064 | 0.40 | 0.090 | 0.50 |
| University | -0.090 | 0.59 | 0.090 | 0.60 |
| Years of business activity | -0.416 | 2.89*** | -0.281 | -2.10*** |
| Log Likelihood | | -556.814 | | -637.98 |
| Pseudo R2 | | 0.103 | | 0.09 |
| Number of obs. | | 405 | | 405 |
Male farmers are significantly less likely to plan a reduction in herbicide use compared to female farmers. Similarly, those engaged in conventional farming practices are substantially less inclined to reduce herbicide use. Family businesses also seem to be less likely to reduce herbicides, while farmers’ age and education do not show significant effects. More years of business activity are negatively associated with the intention to reduce herbicides, suggesting that more experienced farmers may be less inclined to change their weed management strategies.
In contrast, when examining intention to adopt AWM practices, conventional farming again shows a strong and significant negative association. While the coefficient for male farmers remains negative, it is only marginally significant. The influence of family labor, though negative, is not statistically significant in this model. Education and age continue to show no significant effects. However, the number of years in business again presents a significant negative relationship, indicating a consistent reluctance among more experienced farmers to adopt AWM practices. The overall model fit, as measured by Pseudo R2, is slightly lower for the AWM adoption model (0.09) compared to the herbicide reduction model (0.103).
Conclusions
The contribution of this paper to the debate on the efficacy of otherwise desirable informational nudges is twofold. First, informational nudges may have an adverse effect, hindering the adoption of the desired green practice. Second, the adverse effect depends on the decision-maker’s characteristics and past farming decisions. Thus, information provision is not per se an effective means of changing behaviour in favour of more sustainable practices. In fact, these practices represent only one subset of possible actions that farmers might consider adopting and the interaction between preexisting and new green practices has been shown to reduce or reverse the effect of a green informational nudge. Whether this finding represents a local or a global shortcoming of informational nudges remains an open question. Future research should investigate whether transparent communication about both the benefits and costs of green practices can foster more deliberate, conscious adoption of sustainable methods, moving beyond the idea of behavioural change as a simple reaction to nudges, and toward a more holistic and informed decision-making process. However, the nudging failure reported here stands as strong evidence of efficacy-ethics trade-offs emerging under transparent and informationally loaded behavioural interventions.
Ethics and consent
Ethical approval was granted by the Ethics Committee of the Burgundy School of Business on June 28th 2023 (CERBSB2023-56). Informed consent was obtained electronically at the beginning of the online questionnaire.
