Review Reports - One Cow, Two Gains: Welfare Improvement and Chemical Fertilizer Reduction from an Incentive-Based Cattle Support Program Among Poor Smallholders in Guangxi, China

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study conducts a thorough assessment of the Yijiang Daibu incentive-based cattle support program in Guangxi, China. The research effectively illustrates that engagement in small-scale cattle rearing results in a dual benefit: a notable rise in per capita consumption expenditure and a considerable decrease in the use of chemical fertilizers. The authors effectively demonstrate that the main mechanism involves substituting cattle manure for chemical fertilizers. The subject is relevant to the field of Sustainability, and the application of a two-wave panel dataset alongside advanced econometric methods (CEM, Fixed-Effects, and IV) represents a notable advantage. The manuscript exhibits a coherent structure and is articulated with clarity.

Major Comments

The essence of this study lies in its policy significance. The discourse regarding the program's long-term sustainability is presently inadequate. The Yijiang Daibu program is predicated on incentives and dependent on subsidies. The authors must expressly address the fundamental inquiry regarding the fate of the "two gains" following the cessation or termination of the subsidy. Does the first income increase generate sufficient self-sustaining capital for farmers to persist in cattle rearing and manure utilization? A specific section or an extensively elaborated portion of the Discussion must evaluate the program's exit strategy and the prospects for enduring behavioral change. This is a crucial factor for any manuscript in a sustainability journal.
The reduction in chemical fertilizer use represents a significant environmental contribution. This variable is obtained from self-reported household surveys. The authors acknowledge this limitation in the Conclusion; however, it requires a more comprehensive discussion in the Methodology and Discussion sections. Self-reported input data is prone to recall bias and social desirability bias. The authors are required to:
• Specify the survey questions and methodologies employed to reduce potential measurement error.
• Analyze the possible extent and orientation of the bias affecting the results.
Conduct a sensitivity analysis or compare self-reported data with available administrative or regional fertilizer sales data to validate this important finding externally.
The study identifies the replacement of chemical fertilizer with cattle manure as the main mechanism. The analysis requires further depth to substantiate this claim. Could the authors present descriptive statistics or qualitative evidence regarding the rate of substitution? For example, what is the reduction in chemical fertilizer (in kg/ha) per unit of cattle manure produced or utilized? A more detailed analysis of the manure's nutrient content (e.g., N, P, K) and its effects on crop yield or soil health, if data allows, would greatly improve the environmental relevance and mechanistic understanding of the paper.
The study exhibits significant contextual specificity, concentrating on the Yijiang Daibu program in Guangxi, a region characterized by distinct resource endowments, including hilly terrain and fragmented land. The micro-level evidence presented is valuable; however, the authors should allocate additional space in the Discussion to clearly define the boundary conditions for the generalizability of their findings. What elements of the program's success can be applied to other regions in China or developing countries, and which are uniquely tied to the specific policy design and local context of Guangxi?

Minor Comments

Title Clarification: The title, "One Cow, Two Gains," is catchy but could be slightly misleading if the program involves more than a single cow per household. Please clarify in the Introduction whether the program strictly limits the number of cattle or if "One Cow" is a symbolic representation of small-scale rearing.
Table 1 Presentation: In Table 1, the definition for "Consumption" is "Per capita consumption expenditure (USD/year)." Please ensure all monetary values in the paper are consistently presented (e.g., all in USD or all in RMB with a clear conversion rate provided). The note at the bottom of the table mentions the 2020 exchange rate, which is helpful, but consistency is key.
Equation Presentation: Equations (3) and (4) are very similar. Please ensure the distinction between the two is clearly explained in the accompanying text to avoid confusion for readers unfamiliar with the fixed-effects and first-difference models.
Conclusion Section: The Conclusion section (Section 6) is well-written but could benefit from a more explicit summary of the policy recommendations derived from the findings, moving beyond a restatement of the results. What are the three most important takeaways for policymakers in China?

Author Response

Response to Reviewer 1 Comments

We sincerely appreciate the reviewer’s positive assessment and encouraging comments regarding the relevance of our topic, the rigor of our empirical strategy, and the clarity of our manuscript. We are delighted that the "dual benefit" finding—income growth coupled with fertilizer reduction—resonates with the reviewer as a meaningful contribution to the field of Sustainability.

To further enhance the quality and robustness of the study as recognized by the reviewer, we have made the following refinements in this revised version. We have highlighted all changes in the revised manuscript with a yellow background.

Major Comments:

The essence of this study lies in its policy significance. The discourse regarding the program's long-term sustainability is presently inadequate. The Yijiang Daibu program is predicated on incentives and dependent on subsidies. The authors must expressly address the fundamental inquiry regarding the fate of the "two gains" following the cessation or termination of the subsidy. Does the first income increase generate sufficient self-sustaining capital for farmers to persist in cattle rearing and manure utilization? A specific section or an extensively elaborated portion of the Discussion must evaluate the program's exit strategy and the prospects for enduring behavioral change. This is a crucial factor for any manuscript in a sustainability journal.

We are very grateful for this insightful comment, which touches upon the core mission of Sustainability. We fully agree that addressing the fate of the "two gains" after the subsidy ceases is crucial for evaluating the policy's long-term value. In response to your suggestion, we have significantly expanded the Discussion section. Specifically, in the penultimate paragraph (second to last paragraph) of Section 5, we have added a dedicated discussion on the program's exit strategy and behavioral persistence.

We argue that the Yijiang Daibu policy promotes self-sustaining change rather than dependency. Economically, unlike simple cash transfers, cattle act as productive biological assets; the income generated from the initial subsidized cycle provides the necessary capital for reinvestment, allowing farmers to cross the threshold of self-sufficiency. Environmentally, the substitution of manure for chemical fertilizer is driven by rational cost minimization. Once farmers master the technique and realize the cost savings, the economic incentive to continue using manure persists independently of the subsidy. This ensures that the behavioral shift toward greener production is enduring.

The reduction in chemical fertilizer use represents a significant environmental contribution. This variable is obtained from self-reported household surveys. The authors acknowledge this limitation in the Conclusion; however, it requires a more comprehensive discussion in the Methodology and Discussion sections. Self-reported input data is prone to recall bias and social desirability bias. The authors are required to:

Specify the survey questions and methodologies employed to reduce potential measurement error.
Analyze the possible extent and orientation of the bias affecting the results.

Conduct a sensitivity analysis or compare self-reported data with available administrative or regional fertilizer sales data to validate this important finding externally.

We sincerely appreciate the reviewer's rigorous attention to data validity. We fully acknowledge the challenges associated with self-reported micro-data and have taken multiple steps to address them. We have revised the manuscript to provide comprehensive details on our data quality control and external validation:

Strict Survey Protocol (Section 3.1): We have clarified that our data was collected through one-on-one, face-to-face interviews, not self-administered questionnaires. As detailed in the revised Section 3.1, we employed a "cross-calculation verification" technique during interviews. By simultaneously recording total cost, unit price, and physical quantity, enumerators verified logical consistency on-site to minimize measurement error and recall bias.

External Validation with Macro Data (Section 5): To validate our findings externally, we analyzed the aggregate fertilizer usage trends in the study region using official data from the Guangxi Statistical Yearbook. The statistics reveal a distinct downward trend: the aggregate chemical fertilizer usage (pure content) across the six prefecture-level cities covered by our study decreased from 1.42 million tons in 2017 to 1.32 million tons in 2020, representing a total reduction of approximately 7.5%. This quantitative alignment between our micro-survey data and macro-administrative statistics strongly supports the validity of our conclusion that the policy effectively reduced fertilizer use.

The study identifies the replacement of chemical fertilizer with cattle manure as the main mechanism. The analysis requires further depth to substantiate this claim. Could the authors present descriptive statistics or qualitative evidence regarding the rate of substitution? For example, what is the reduction in chemical fertilizer (in kg/ha) per unit of cattle manure produced or utilized? A more detailed analysis of the manure's nutrient content (e.g., N, P, K) and its effects on crop yield or soil health, if data allows, would greatly improve the environmental relevance and mechanistic understanding of the paper.

Thank you for this insightful suggestion. While our study relies on household socio-economic surveys and thus lacks laboratory-tested data on specific N-P-K content or precise physical weights to calculate a technical substitution rate (kg/kg), we have substantiated this mechanism through rigorous empirical evidence and literature support. In Section 4.2, we expanded our analysis to show that participation in the program significantly increases the probability and quantity of organic fertilizer application (Table 4). When viewed alongside the baseline reduction in chemical fertilizer (Table 2), this simultaneous increase in organic inputs provides robust evidence of a substitution effect rather than a mere reduction in input intensity.

Regarding environmental relevance and nutrient content, we have revised the Discussion (Section 5) to incorporate authoritative agronomic literature (e.g.,Jie Wei et al., 2025). These studies, conducted on similar sloping lands, confirm that the N-P-K supply from cattle manure is sufficient to maintain soil fertility and improve soil physicochemical properties. This literature provides the necessary agronomic theoretical basis to justify the substitution observed in our micro-level data.

The study exhibits significant contextual specificity, concentrating on the Yijiang Daibu program in Guangxi, a region characterized by distinct resource endowments, including hilly terrain and fragmented land. The micro-level evidence presented is valuable; however, the authors should allocate additional space in the Discussion to clearly define the boundary conditions for the generalizability of their findings. What elements of the program's success can be applied to other regions in China or developing countries, and which are uniquely tied to the specific policy design and local context of Guangxi?

We appreciate this thoughtful comment regarding the external validity of our study. We agree that distinguishing between context-specific factors and generalizable policy lessons is essential.

We have added a new paragraph titled "Boundary conditions and generalizability" in the Discussion section. We clarify that the agronomic success relies on Guangxi’s specific endowment (hilly terrain, fragmented land, and abundant crop residues like sugarcane), making the technical model most suitable for similar mountainous developing regions rather than flat, industrialized agricultural zones. We highlight that the economic mechanism—the incentive-compatible "award in lieu of subsidy" design—is widely applicable. This policy logic, which ties transfers to productive asset accumulation to prevent dependency, offers a valuable lesson for poverty alleviation programs in other developing countries, regardless of geography.

Minor Comments:

Title Clarification: The title, "One Cow, Two Gains," is catchy but could be slightly misleading if the program involves more than a single cow per household. Please clarify in the Introduction whether the program strictly limits the number of cattle or if "One Cow" is a symbolic representation of small-scale rearing.

We thank the reviewer for identifying this crucial point of clarification. We have chosen to retain the "One Cow, Two Gains" prefix as it succinctly captures the study’s core focus on the marginal impact of acquiring productive assets for the poor. As the reviewer correctly surmised, "One Cow" is a symbolic representation of the low entry barrier and the small-scale nature of the beneficiaries, rather than a strict policy cap limiting households to a single animal.

To prevent misunderstanding and ensure precision, we have made two adjustments. First, we clarified in the second paragraph of the Introduction that "One Cow" symbolizes the entry threshold rather than a limit. Second, we refined the subtitle to "...Welfare Improvement and Chemical Fertilizer Reduction..." to more accurately reflect the specific variables measured in our study. These changes ensure the title remains catchy while being academically rigorous and not misleading.

Table 1 Presentation: In Table 1, the definition for "Consumption" is "Per capita consumption expenditure (USD/year)." Please ensure all monetary values in the paper are consistently presented (e.g., all in USD or all in RMB with a clear conversion rate provided). The note at the bottom of the table mentions the 2020 exchange rate, which is helpful, but consistency is key.

We appreciate the reviewer’s attention to detail regarding unit consistency. We have carefully reviewed the entire manuscript and confirmed that all monetary variables (including income, consumption, and input costs) presented in Table 1, the descriptive text, and all subsequent regression results are consistently reported in USD.

We have retained the footnote specifying the exchange rate used (1 USD ≈ 6.89 RMB, based on the 2020 average) in Table 1 and have ensured this conversion standard is applied uniformly throughout the paper to facilitate readability for international audiences.

Equation Presentation: Equations (3) and (4) are very similar. Please ensure the distinction between the two is clearly explained in the accompanying text to avoid confusion for readers unfamiliar with the fixed-effects and first-difference models.

We thank the reviewer for this helpful comment regarding the presentation of the empirical model. We agree that the distinction between the algebraic derivation and the final notation should be made clearer for a broader audience.

We have revised the text connecting Equations (3) and (4) on Page 6. We now explicitly clarify that Equation (3) illustrates the mechanism of first-differencing (showing how the time-invariant term Z_i is eliminated), while Equation (4) presents the same model using standard difference notation (Δ) for compactness and ease of interpretation.

Conclusion Section: The Conclusion section (Section 6) is well-written but could benefit from a more explicit summary of the policy recommendations derived from the findings, moving beyond a restatement of the results. What are the three most important takeaways for policymakers in China?

We appreciate the reviewer’s suggestion to sharpen the policy implications. We agree that the Conclusion should offer actionable advice beyond summarizing the findings.

We have rewritten the second paragraph of the Conclusion (Section 6) to explicitly outline three key takeaways for policymakers in China.

Thank you again for your helpful comments and suggestions.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Generally speaking, the paper has done a research regarding the Income Growth and Reduced Chemical Fertilizer Use of small household in China, and it has revealed some new findings and conclusions. The paper has used some good research methods to revealed the impact mechanism, it is a good empirical paper, while, I still have some comments for the author to improve the paper:

Thetitle of the paper should be re-organized, it can add a subtitle to present the research samples.
I find the abstract have present the mainly research contents, the methods and the policy implication, this should be admired.
I suggest the author try to construct a theoretical framework that can construct the causal relationship of the “yijiang daibu”policy on household income and also the chemical use, the impact mechanisms that have been introduced in the abstract should be presented here.
Has the authors try to use another method in the robust check section, I think except for the CEM, the author also need to try the PSM methods instead.
Actually, the authors have done great regression results in the baseline model, and also tried some measures in the robust check models, these should be admired, while as I have mentioned, the PSM should also be considered.
The Iv-FE can be good methods to solve the problems of endogenous, while the author should explained more clearly how they choose the Iv, why the Iv is a good variable and it is reasonable at the situation, if other better IV?
Thecontents in the conclusion and the discussion are repeated, Actually, the authors can provide a lot of effective policy implications according to their empirical research and the regression results, while, the policy implication in the conclusion is too simple and not enough, this should be added. The authors have presented some research limitations of the paper, this should be admired.

Author Response

Response to Reviewer 2 Comments

Thank you for all your comments and suggestions. We have revised the paper according to the specific comments in the main text. Below we provide itemized response to the few more comprehensive comments and suggestions.

Comments:

The title of the paper should be re-organized, it can add a subtitle to present the research samples.

We thank the reviewer for this constructive suggestion regarding the title's structure and precision. We fully agree that explicitly presenting the research sample is crucial for readers to quickly grasp the study’s scope.

In response, we have carefully refined the title structure to "Main Title: Subtitle." Specifically, the revised title is: One Cow, Two Gains: Welfare Improvement and Chemical Fertilizer Reduction among Poor Smallholders in Guangxi, China. This modification not only specifies the key outcome variables (Welfare and Fertilizer) but also, as per your suggestion, clearly identifies the research sample (Poor Smallholders) and the geographic location (Guangxi, China). We believe this finalized structure maximizes clarity by precisely defining both the study's impact and its context.

I find the abstract have present the mainly research contents, the methods and the policy implication, this should be admired.

We sincerely appreciate the reviewer’s positive assessment of our abstract. We made a concerted effort to ensure that the research content, methodology, and policy implications were presented clearly and comprehensively. We are very encouraged by your kind recognition of this work.

I suggest the author try to construct a theoretical framework that can construct the causal relationship of the “yijiang daibu” policy on household income and also the chemical use, the impact mechanisms that have been introduced in the abstract should be presented here.

We sincerely appreciate this insightful suggestion. We agree that constructing a clear theoretical framework is essential to elucidate the causal pathways connecting the Yijiang Daibu policy to household income and chemical fertilizer use before presenting the empirical results. Therefore, we have added a new subsection, "2.4 Theoretical Framework and Research Hypotheses," at the end of Section 2.

In this new section, we ground our analysis in household production theory and the Permanent Income Hypothesis. Specifically, we explicitly articulate two primary mechanisms: the resource substitution effect (where manure acts as a shadow-priced input replacing commercial fertilizer) and the income effect (derived from asset accumulation and cost savings). Furthermore, as suggested, we have integrated the mechanisms introduced in the abstract—credit access, cooperative membership, and internet usage—into this framework, presenting them as key environmental conditions that moderate the policy's impact.

Has the authors try to use another method in the robust check section, I think except for the CEM, the author also need to try the PSM methods instead. Actually, the authors have done great regression results in the baseline model, and also tried some measures in the robust check models, these should be admired, while as I have mentioned, the PSM should also be considered.

We thank the reviewer for this constructive suggestion using Propensity Score Matching (PSM) to further validate our findings. Accordingly, we have conducted a robustness check using the PSM k-nearest neighbor matching algorithm (k=4). The results, now presented in Table 5 of Section 4.2, are highly consistent with our baseline estimates. This confirms that the positive impact on household consumption and the negative impact on chemical fertilizer intensity are robust across different matching specifications.

While the PSM results support our conclusions, we retain Coarsened Exact Matching (CEM) as our primary identification strategy. Following King and Nielsen (2019), we favor CEM because it avoids the need to specify a functional form for the propensity score (reducing model dependence) and guarantees 'monotonic imbalance bounding' ex ante. Thus, the newly added PSM results serve as strong supplementary evidence that our findings are not driven by the choice of algorithm.

The Iv-FE can be good methods to solve the problems of endogenous, while the author should explained more clearly how they choose the Iv, why the Iv is a good variable and it is reasonable at the situation, if other better IV?

We thank the reviewer for this insightful comment. In the revised Section 4.3, we have clarified that our choice of IV—an indicator for whether neighbors or close relatives engage in cattle rearing—follows the standard approach in recent development literature (Ma and Abdulai, 2016; Lasdun Violet et al, 2025). This selection relies on the mechanism of peer effects and social learning, which are particularly strong in rural China. As confirmed by recent evidence in the Journal of Development Economics (Lasdun Violet et al, 2025), households are significantly more likely to adopt agricultural practices when they observe neighbors doing so due to reduced information asymmetry, ensuring strong relevance for our instrument.

Regarding the validity and reasonableness of the IV, we argue that it satisfies the exclusion restriction. It is plausible to assume that the decision of neighbors to rear cattle does not directly alter the focal household’s budget constraints or agricultural production environment (e.g., soil quality). Neighbors’ behavior affects the focal household’s consumption welfare and chemical fertilizer intensity only through the channel of influencing the household’s own adoption decision. By controlling for household characteristics and fixed effects, we minimize the risk of omitted variable bias, ensuring the IV captures peer influence rather than common shocks.

Finally, we confirm the robustness of this IV through statistical tests. The first-stage Kleibergen-Paap rk Wald F-statistic reported in Table 6 significantly exceeds the threshold of 10, rejecting the concern of weak identification. While we considered alternative instruments such as geographic variables (e.g., terrain or distance), these are typically time-invariant and thus absorbed by the fixed effects in our panel model, or they may directly influence agricultural productivity. Therefore, the peer-based IV remains the most robust and valid identification strategy for this specific study context.

The contents in the conclusion and the discussion are repeated, Actually, the authors can provide a lot of effective policy implications according to their empirical research and the regression results, while, the policy implication in the conclusion is too simple and not enough, this should be added. The authors have presented some research limitations of the paper, this should be admired.

We sincerely thank the reviewer for pointing out the repetition between the Discussion and Conclusion sections, and for the constructive suggestion to strengthen the policy implications. We fully agree that our empirical results offer more concrete guidance than previously stated. In response, we have:

Restructured Section 5 (Discussion) to focus strictly on comparing our findings with existing literature and interpreting the underlying mechanisms, removing summary statements that belong in the conclusion.

Rewritten Section 6 (Conclusion and Policy Implications) to avoid repetition. Most importantly, we have significantly expanded the Policy Implications subsections. Instead of general suggestions, we now provide specific recommendations derived directly from our empirical findings, such as targeting peer effects for technology promotion, subsidizing small-scale integrated farming to reduce fertilizer use, and designing welfare-enhancing livestock programs. These changes highlight the practical value of our research.

Thank you again for your helpful comments and suggestions.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

This paper deserves publication, but some aspects need improvement. I have attached my suggestions. Here I list my two main concerns:

1) I think this paper has a point of methodological weakness because not all the households in the treatment group have been exposed to the treatment. In my opinion, this exposes your final results to criticism. My suggestion is to explain, as well as you can, why you consider those who have not been exposed to the treatment to be part of a treatment group. This is a crucial point. The CME applied is an ex post statistical strategy that helps reduce bias, but the methodological bias remains.

2) Looking at the test of the endogeneity, the coefficients of fertilizer usage became positive; the more cattle rearing, the more fertilizer usage. This is probably due to the fact that initially you omitted from your model the 'peer effect', learning from what the neighbors do. In my opinion, you should discuss this effect in depth. In this case, the program has not taken into consideration the social mechanism of contagion. When you keep the mechanism of contagion under control, your basic relationship (cattle rearing and fertilizer usage) becomes positive and strong. I am a sociologist, and these mechanisms are often overlooked by policymakers. I suggest discussing this result in depth, without trying to hide these very interesting research results.

Comments for author File: Comments.pdf

Author Response

Response to Reviewer 3 Comments

Comments:

I think this paper has a point of methodological weakness because not all the households in the treatment group have been exposed to the treatment. In my opinion, this exposes your final results to criticism. My suggestion is to explain, as well as you can, why you consider those who have not been exposed to the treatment to be part of a treatment group. This is a crucial point. The CME applied is an ex post statistical strategy that helps reduce bias, but the methodological bias remains.

We gratefully acknowledge the reviewer’s concern regarding the exposure of the treatment group and potential selection bias. We would like to clarify that non-exposure within the treatment group is institutionally impossible in this study due to the specific "Build First, Subsidize Later" mechanism of the Yijiang Daibu policy. As detailed in the revised Section 2.1, subsidies are granted only after households have physically acquired cattle, completed construction, and passed government verification. Therefore, our treatment variable (T=1) strictly captures actual adoption behavior rather than mere policy eligibility, ensuring that every household in the treatment group was fully exposed to the intervention.

Regarding the methodological concern of bias arising from voluntary participation (self-selection), we have addressed this by combining Coarsened Exact Matching (CEM) with Household Fixed Effects (FE). CEM balances the treatment and control groups on key observable characteristics (e.g., labor resources, land size) prior to estimation, while fixed effects control for time-invariant unobserved heterogeneity. By defining the treatment based on verified adoption and rigorously matching households, we aim to isolate the causal Average Treatment Effect on the Treated (ATT) and minimize the methodological bias mentioned. We have updated Section 3.1 to make these definitions and strategies more explicit.

Looking at the test of the endogeneity, the coefficients of fertilizer usage became positive; the more cattle rearing, the more fertilizer usage. This is probably due to the fact that initially you omitted from your model the 'peer effect', learning from what the neighbors do. In my opinion, you should discuss this effect in depth. In this case, the program has not taken into consideration the social mechanism of contagion. When you keep the mechanism of contagion under control, your basic relationship (cattle rearing and fertilizer usage) becomes positive and strong. I am a sociologist, and these mechanisms are often overlooked by policymakers. I suggest discussing this result in depth, without trying to hide these very interesting research results.

We offer our sincerest apologies for a significant typographical error in the previous version of the manuscript, which led to this confusion. The reviewer is absolutely correct to point out the anomaly in the coefficient.

Due to a formatting oversight in the previous draft, the negative sign was accidentally omitted from the coefficient for fertilizer usage in the IV-FE model (Table 6). The correct coefficient is, in fact, -0.309 (negative), not positive. We have corrected this error in the revised manuscript. We are deeply grateful to the reviewer for scrutinizing the results so carefully; without your observation, this critical error might have gone unnoticed.

Although the strictly positive relationship the reviewer observed was due to a typo, your sociological insight regarding “peer effects” and “social contagion” remains highly relevant and valuable for interpreting the true negative result. With the correct coefficient (-0.309), which is larger in magnitude than the baseline estimate (-0.158), the "social mechanism" effect suggests a story of positive social learning. It implies that households influenced by their neighbors (the compliers in the IV strategy) are engaging in deep social learning. They represent a group that adopts cattle rearing through social networks, where they likely acquire not just the motivation to raise cattle, but also the technical knowledge on how to effectively substitute manure for chemical fertilizers.

Therefore, strictly following your suggestion, we have expanded the discussion in Section 4.3 and Section 5 to analyze this "social learning" mechanism. We argue that the "contagion" here facilitates the diffusion of green production practices, explaining why the fertilizer reduction effect is even stronger when social networks are accounted for.

Response to specific annotations within the provided PDF file

Thank you for the detailed annotations in the PDF provided. We have carefully reviewed each note and found that they primarily reinforce the two major concerns addressed above regarding treatment definition and peer effects.

Accordingly, we have revised Section 3.1 to clarify that treatment is defined by verified participation to ensure full exposure. Furthermore, we have significantly expanded Section 4.3 and the Discussion (Section 5) to incorporate the "social contagion" and peer effects mechanism as suggested. These sections now explicitly interpret the results through the lens of social learning to explain the enhanced fertilizer reduction effects.