Introduction
Fertilizers play a crucial role in addressing food security, increasing agricultural productivity, and reducing rural poverty (McArthur & McCord, Reference McArthur and McCord2017; Stewart & Roberts, Reference Stewart and Roberts2012). To promote fertilizer use and enhance crop yields, many developing countries have implemented fertilizer support programmes (Bruinsma, Reference Bruinsma2003; Pingali, Reference Pingali2012; M. Sutton et al., Reference Sutton, Bleeker, Howard and Erisman2013). Fertilizer subsidies constitute a significant portion of agricultural budgets in developing countries.Footnote 1 Given the significant economic costs associated with these subsidies, it is essential to promote best nutrient management practices to improve fertilizer efficiency and maximize agricultural benefits (Kishore et al., Reference Kishore, Alvi and Krupnik2021; Timsina et al., Reference Timsina, Dutta, Devkota, Chakraborty, Neupane, Bishta, Amgain, Singh, Islam and Majumdar2021).
One of the key pillars of agricultural sustainability is the judicious use of inputs in crop production (Cui et al., Reference Cui, Zhang, Chen, Zhang, Ma and Nature2018; Zhang et al., Reference Zhang, Davidson, Mauzerall, Searchinger, Dumas and Shen2015). However, farmers in developing countries often apply fertilizers based on availability and affordability rather than scientifically recommended doses and plant nutrient demands (Bora, Reference Bora2022; Kishore et al., Reference Kishore, Alvi and Krupnik2021). Research indicates that Nepalese farmers do not consistently practice a balanced application of nitrogen (N), phosphorus (P), and potassium (K) for cereal crops (N. R. Pandit, Gaihre, Choudhary, et al., Reference Pandit, Gaihre, Choudhary, Subedi, Thapa, Maharjan, Khadka, Vista and Rusinamhodzi2022) or vegetables such as cauliflower (N. R. Pandit, Gaihre, Gautam, et al., Reference Pandit, Gaihre, Gautam, Maharjan, Vista and Choudhary2022). Inefficient nutrient application—whether through imbalanced quantities, improper placement, or incorrect timing—reduces plant nutrient uptake and lowers crop productivity and profitability (Johnston & Bruulsema, Reference Johnston and Bruulsema2014).
Due to the government’s heavy subsidies (65–70%) on urea (46% nitrogen content) compared to 25–30% subsidies for diammonium phosphate (DAP, 18% nitrogen content) and 30–32% subsidies for muriate of potash (MOP, no nitrogen content) (MOAD, 2022), farmers are more likely to apply excessive urea fertilizers leading to overuse of nitrogen. Several studies indicated that excessive nitrogen use has significant environmental consequences (Zhen et al., Reference Zhen, Zoebisch, Chen and Feng2006; Sheahan and Barrett, Reference Sheahan and Barrett2017; Wu et al., Reference Wu, Xi, Tang, Luo, Gu, Lam, Vitousek and Chen2018; Ren et al., Reference Ren, Xu, Li, Zheng, Liu, Reis, Lu, Zhang, Gao and Gu2021), threatens agricultural sustainability (Lu & Tian, Reference Lu and Tian2017; M. A. Sutton et al., Reference Sutton, Simpson, Levy, Smith, Reis, van Oijen and de Vries2008), and poses risks to human health (Erisman et al., Reference Erisman, Galloway, Seitzinger, Bleeker, Dise, Roxana Petrescu, Leach and de Vries2013; Gourevitch et al., Reference Gourevitch, Keeler and Ricketts2018; Wang & Lu, Reference Wang and Lu2020; Wangunci & Zhao, Reference Wangunci and Zhao2019). Overuse often results in diminishing returns; plants cannot absorb the excess nitrogen, leading to wasted inputs and reduced cost-effectiveness. Over time, excessive nitrogen use can degrade soil health, leading to lower productivity and the need for more inputs to maintain yields (Zhang et al., Reference Zhang, Davidson, Mauzerall, Searchinger, Dumas and Shen2015).
Nepal’s agricultural policy emphasizes the importance of efficient fertilizer use to drive agricultural growth and productivity (ADS, 2015; APP, 1995). The Agriculture Development Strategy (ADS)—Nepal’s guiding policy document—identifies poor knowledge of fertilizer best practices as a key constraint to productivity and calls for the promotion of efficient fertilizer management. A widely recognized approach to optimizing fertilizer use is the 4R nutrient stewardship framework, which advocates applying the right source of nutrients, at the right rate, at the right time, and in the right place. Developed by the International Plant Nutrition Institute (IPNI) with global partners, the 4Rs are now widely adopted as a best-practice standard for optimizing fertilizer use efficiency and minimizing environmental impacts (Johnston & Bruulsema, Reference Johnston and Bruulsema2014; Bruulsema et al. Reference Bruulsema, Fixen and Sulewski2016; Jones, Reference Jones2021; Mikkelsen et al., Reference Mikkelsen, Schwab and Randall2009; Santos, Reference Santos2011; Stewart et al., Reference Stewart, Sawyer and Alley2009) and can narrow the existing yield gap in agriculture (Mueller et al., Reference Mueller, Gerber, Johnston, Ray, Ramankutty and Foley2012).
In Nepal, limited research has been conducted on assessing fertilizer application practices among farmers and the nitrogen use efficiency of crops (Dhakal et al., Reference Dhakal, Baral, Pokhrel, Pandit, Gaihre and Vista2021; N. R. Pandit, Choudhary, et al., Reference Pandit, Choudhary, Maharjan, Dhakal, Vista and Gaihre2022; Timilsina et al., Reference Timilsina, Khanal and Vista2023). To the best of our knowledge, we are not aware of any studies that assess the determinants of 4R soil nutrient management adoption and excessive use of N fertilizers in Nepal. This study seeks to address this gap by examining the factors influencing the determinants of 4R soil nutrient management adoption and excessive N use focusing on rice, maize, and cauliflower among Nepalese farmers.Footnote 2
Material and methods
Data
We conducted a cross-sectional Beneficiary-Based Survey (BBS) across 14 districts in Nepal, following the methodologies outlined in the Sampling Guide for Beneficiary-Based Surveys for Selected Feed the Future Agricultural Annual Monitoring Indicators (Stukel & Friedman, Reference Stukel and Friedman2016). The survey was part of the Nepal Seed and Fertilizer (NSAF) project, implemented by the International Maize and Wheat Improvement Center (CIMMYT).
The project was implemented between 2016 and 2024, during which annual surveys were conducted with beneficiary farmers. However, only the 2021 and 2022 survey rounds included questions related to the use of 4Rs nutrient management. These two rounds were used in this study, but these are not the baseline and endline surveys. As part of the project, selected farmers received training on fertilizer best management practices. About 926 farmers were selected for survey in 2021 (426) and 2022 (500) of which 63% were from Terai and 37% from hilly region. We focus only on rice, maize, and cauliflower, identified as Nepal’s most significant crops (Table 1).
Table 1. Crops targeted, study district, and sample size
The survey employed a two-stage cluster sampling design to ensure representativeness. In the first stage, farmer groups served by the project at the village or community level—hereafter referred to as clusters—were identified across four provinces (Sudurpaschim, Gandaki, Lumbini, and Bagmati), spanning both hill (sub-tropical) and Terai (tropical) ecological regions. Roughly 100 such groups cultivating six key crops (rice, maize, lentil, onion, cauliflower, and tomato) were listed. From this pool, 30 to 32 sampling units were randomly chosen according to crop suitability within each province and ecological region.
The second stage focused on household selection. For each sampled group, a list of project beneficiaries was compiled, from which 20 to 30 households were randomly drawn using a random number generator. Trained enumerators then conducted face-to-face interviews with the selected households using Open Data Kit (ODK) software. A carefully designed and pretested questionnaire captured information on socio-economic and agricultural characteristics, food consumption, access to credit, and 4R nutrient management practices.Footnote 3
Empirical model
Multivariate probit (MVP) model to assess factors influencing 4R fertilizer usage
To analyse the determinants of 4R fertilizer use, we employed a multivariate probit (MVP) model. The 4R nutrient management practices—right source, right time, right dose, and right placement—as promoted by the project, are outlined in Table 2. Based on these guidelines, we created four binary dependent variables, each representing adherence to one of the 4R principles.
Table 2. 4R nutrient management practices promoted by the Nepal Seed and Fertilizer project
Source: Nepal Agriculture Research Council; 30 katha = 1 hectare, DAP = diammonium phosphate, MOP = muriate of potash.
The best nutrient management practice, recommended by the Nepal Agricultural Research Council, defines right fertilizer application techniques. Farmers following the guidelines in Table 2 were considered to be using the ‘right’ technique. Conversely, deviations from these recommendations were categorized as ‘wrong’ techniques. For instance, if farmers applied only urea and DAP but excluded MOP, the crop lacked the essential potassium nutrient, resulting in unbalanced nutrition—constituting an incorrect practice. For maize cultivation, urea should be applied at the 6-leaf stage with a second top dressing at the 10-leaf stage. Moreover, if urea was applied only as basal dose during planting, it was classified as a ‘wrong’ technique.Footnote 4 The modelling approach enables understanding of the factors driving or hindering the adoption of proper 4R practices and informs targeted interventions for improved nutrient management.
Some farmers may apply fertilizers at the right rate and in the right place but may fail to apply them at the right time or may use only N while neglecting P and K. These decisions are often influenced by factors such as the timely availability of fertilizers and the farmers’ knowledge of proper fertilizer use. Consequently, the adoption decisions across the four components of the 4R nutrient management framework are likely to be correlated.
While univariate logit or probit models can be used to analyse the adoption of each individual ‘R’, these methods may yield biased and inefficient estimates due to potential interdependencies among the adoption decisions. To address this issue, we employed the MVP model, which accounts for the correlations among multiple decisions and provides more accurate estimates with correct standard errors. The MVP model accommodates error correlations across equations and controls for unobserved heterogeneity (Cappellari & Jenkins, Reference Cappellari and Jenkins2003). The MPV model is specified as follows:
$${T_{ih}} = {\alpha _i} + {\beta _i}{\rm{\;}}{R_{ih}} + {\rm{\;}}{\delta _i}{\rm{\;}}{M_{ih}} + {\rm{\;}}{\mu _i}{H_{ih}} + {\theta _i}{A_{ih}} + {\varepsilon _{ih}}$$
where
${T_{ih}}$
is a binary indicator of adoption for the 4R practice i by household h (where i = 1,2,3,4 for each of the 4Rs),
$\;{R_{ih}}$
is a binary indicator for rice growers,
$\;{M_{ih}}$
is a binary indicator for maize growers (cauliflower growers serve as the reference category,
$)\;{H_{ih}}$
are the vector of socio-economic characteristics,
${A_{ih}}$
are the vector of agricultural characteristics, and
${\varepsilon _{ih}}$
is the error term capturing unobserved factors.
The agricultural characteristics included in the model are total farmed area, livestock index, access to irrigation facilities, engagement in contract farming, participation in fertilizer demonstration programmes, and time to reach nearest cooperative and agrovets. The socio-economic characteristics incorporated into the model are gender of the household head, age of the household head, education level, family size, migration status, ethnicity, access to credit, smartphone ownership, household income, and affiliation with cooperatives. These variables were selected based on insights from prior studies (Mariano, Villano and Fleming, Reference Mariano, Villano and Fleming2012; Ghimire, Huang and Shrestha, Reference Ghimire, Huang and Shrestha2015; Kumar et al., Reference Kumar, Takeshima, Thapa, Adhikari, Saroj, Karkee and Joshi2020; Aryal, Rahut, et al., Reference Aryal, Rahut, Thapa and Simtowe2021; Aryal, Sapkota, et al., Reference Aryal, Sapkota, Krupnik, Rahut, Jat and Stirling2021).
In equation 1,
${\alpha _i}$
,
${\beta _i},{\delta _i}\;,{\mu _i},{\theta _i}$
are the model parameters representing the effects of explanatory variables on the adoption decision. The error terms (
${\varepsilon _1},{\varepsilon _2},{\varepsilon _3},{\varepsilon _4})$
are assumed to follow a multivariate normal distribution with a zero conditional mean and variance normalized to 1, that is,
${\varepsilon _1},{\varepsilon _2},{\varepsilon _3},{\varepsilon _4}\mathop \to \limits^{\quad MVN\quad } \left( {0,\omega } \right)$
. The variance–covariance matrix
$\left( \omega \right)$
is given by:
$$\omega = \left[ {\matrix{ 1 & {{\rho _{12}}} & {{\rho _{13}}} & {{\rho _{14}}} \cr {{\rho _{21}}} & 1 & {{\rho _{23}}} & {{\rho _{24}}} \cr {{\rho _{31}}} & {{\rho _{32}}} & 1 & {{\rho _{34}}} \cr {{\rho _{41}}} & {{\rho _{42}}} & {{\rho _{43}}} & 1 \cr } } \right]$$
where rho
$\left( {{\rho _{ij}}} \right)$
is the correlation coefficient between the error terms of 4R adoption decision i and j. For the MVP model to be valid, the off-diagonal correlation terms (
${\rho _{ij}}$
) should be non-zero, indicating interdependencies between the adoption decisions. We tested this using a likelihood ratio (LR) test, which evaluates whether the covariances across the equations are statistically significant. The null hypothesis assumes no correlation between the error terms (
${\rho _{21}}$
=
${\rho _{31}}$
=
${\rho _{41}}$
=
${\rho _{32}}$
=
${\rho _{42}}$
=
${\rho _{43}}$
= 0). Rejecting this hypothesis supports the use of the MVP model, confirming that the farmers’ choices regarding the adoption of different 4R practices are interrelated.
Assessing factors influencing excessive nitrogen use through a binary probit model
We employed a binary probit model to examine the factors driving excessive nitrogen use in rice, maize, and cauliflower cultivation. To construct the variable representing excessive nitrogen application, we calculated the total nitrogen from urea and DAP, using their standard nutrient compositions of 46% N and 18% N, respectively. We then measured the deviation from the recommended nitrogen application rate for each household. Households applying nitrogen at a rate exceeding a 10% deviation from the recommended dose were classified as overusers. This 10% threshold was chosen under the assumption that such an excess could negatively impact crop growth and development.
While we recognize that fertilizer recommendations can vary based on soil fertility conditions, the lack of plot-level data and information on prior soil fertility constrained our ability to adjust for these factors. Given that N requirements differ by crop, we developed separate binary indicators for overapplication in rice, maize, and cauliflower. For instance, the variable ‘overuse of N for rice’ equals 1 if the nitrogen application exceeds the 10% deviation threshold and 0 otherwise. To account for crop-specific factors, we estimated a separate binary probit model for each crop using the following specification:
$${N_{ch}} = {\alpha _c} + {\beta _c}{\rm{\;}}{H_{ch}} + {\theta _c}{A_{ch}} + {\gamma _c}{E_{ch}} + {\delta _c}{Y_{ch}} + {\varepsilon _{ch}}$$
where
${N_{ch}}$
is a binary indicator (1= overuse of nitrogen, 0 = otherwise) for crop c in household h. Crop type includes rice (c = 1), maize (c = 2), and cauliflower (c = 3).
$\;{H_{ch}}$
represents the household socio-economic characteristics (e.g., family size, household income, age, and education level),
${A_{ch}}$
captures agricultural characteristics (e.g., total cultivated land, livestock index, and involvement in contract farming),
${E_{ch}}$
is a dummy variable for the ecological zone,
${Y_{ch}}$
is a dummy variable for the surveyed year, and
${\varepsilon _{ch}}$
is the error term, assumed to follow a normal distribution. Parameters
${\alpha _c},{\beta _c},{\delta _c},{\gamma _c}$
, and
${\theta _c}\;$
are estimated for each crop. The analysis was conducted using Stata 17 statistical software. This modelling approach allows for a nuanced understanding of the socio-economic, agricultural, and ecological factors contributing to excessive nitrogen use, tailored to each crop type.
Results
This section presents descriptive and empirical results, organized into distinct subsections for clarity.
Descriptive results
Table 3 summarizes the key statistics (mean and standard deviation) of the variables used in the analysis, categorized by crop type. The final column provides detailed descriptions of each variable. For binary (dummy) variables coded as 1 (Yes) or 0 (No), the mean represents the proportion of farmers exhibiting a specific characteristic or condition, which can be interpreted as a percentage.
Table 3. Descriptive statistics (mean) of the variables used in the analysis
Notes: Standard deviations are in parenthesis; D indicates dummy variable (1/0).
Fertilizer application varied notably across crop types. All sampled rice farmers (100%) applied urea, while 97% used DAP. In contrast, 59% of cauliflower farmers used MOP. Specific adherence to the ‘4R’ principles of nutrient management also varied by crop. About 56% of cauliflower farmers applied the right fertilizer source. Only 12% of maize farmers applied nitrogenous fertilizers at the recommended dose. About 10% of farmers applied fertilizers at the right time, and 30% ensured placement in the right location for rice cultivation. About 8% of the sampled rice farmers, 40% of the maize farmers, and 81% of the cauliflower farmers have applied excessive nitrogenous fertilizers.
The socio-economic profiles of farmers varied based on the crops they cultivated. Among three crops, higher proportion of cauliflower growers were female-headed (32%) and had a greater average number of years of education (5.59 years). A notable proportion of maize growing HHs belonged to the Dalit caste (16%) and were affiliated with cooperatives (97%). A larger share of rice growing households (HHs) had access to credit facilities (22%).
Agricultural practices also differed significantly among crop growers. A larger proportion (65%) of maize growing HHs cultivated hybrid varieties, maintained higher livestock holdings (as indicated by the livestock index), and participated in contract farming (7%) in comparison to rice and cauliflower growing households. A greater share of rice farmers had access to irrigation facilities (29%) and cultivated larger-than-average farm area (0.56 hectares). These descriptive results highlight the diversity in fertilizer use, socio-economic conditions, and agricultural practices among smallholder farmers, which are likely to influence nitrogen management decisions across different crops.
Empirical results
Factors influencing the adoption of 4R nutrient management technology
The suitability of the MVP model was first evaluated. Pairwise correlation coefficients among the residuals of the MVP model revealed statistically significant correlations for most 4R practices (Table 4), indicating interdependence among farmers’ decisions regarding different 4R practices. The LR test further confirmed this, rejecting the null hypothesis of no correlation among the error terms (χ2 (df = 6) = 481.20, p < 0.01). These results validate the use of the MVP model, suggesting that farmers view the 4R practices as complimentary rather than substitutive (Table 4).
Table 4. Pairwise correlation coefficients across 4R’s fertilizer application decisions
Note: Likelihood ratio test of rho21 = rho31 = rho41 = rho32 = rho42 = rho43 = 0: chi2(6) = 481.20 Prob > chi2 = 0.0000.
Table 5 presents the empirical results of the MVP model. The model is statistically significant based on the Wald chi-square test (p = 0.00), and key factors influencing 4R adoption were identified. Female-headed HHs exhibit a higher probability of adopting the right source and right placement of fertilizers. Households with higher education have higher probability of adopting right source, right time, or right placement of fertilizers. The age of the household head is positively and significantly associated with the adoption of 4R practices. Adoption behaviours differ by ethnic group. HHs from Dalit communities (historically marginalized groups) are less likely to apply fertilizers at the right rate compared to households from other ethnic backgrounds.Footnote 5
Table 5. Factors associated with the adoption of 4Rs use of fertilizer (results from multivariate probit model)
Robust standard errors are in parentheses.
*** p < 0.01, ** p < 0.05, * p < 0.1.
A negative relationship was found between farm size and 4R adoption. Households with smaller landholdings are more inclined to use 4R practices. Households with credit access are significantly more likely to apply fertilizers at the right rate. Cooperative members are more likely to apply N, P, and K fertilizers in the right place. Contract farming shows mixed effects. While farmers engaged in contract farming are more likely to adopt the right fertilizer source and placement, they are less likely to adhere to the correct nitrogen application rate.
Access to agricultural inputs and information varies with distance from cooperatives and agrovets. Farmers located closer to these resources are more likely to adopt 4R practices. Information and Communications Technology (ICT) access, particularly smartphone ownership, positively influences the adoption of the right rate of nitrogen applications. Households cultivating hybrid crop varieties are more likely to use the appropriate fertilizer types. Similarly, access to irrigation facilities enhances 4R adoption, especially regarding the right fertilizer type and placement.
Factors influencing the excessive use of nitrogenous fertilizers
Table 6 presents empirical results from the probit model assessing the determinants of excessive nitrogenous fertilizer use across rice, maize, and cauliflower farmers. To enhance interpretability, we report marginal effects rather than odds ratios, allowing for a clearer understanding of the impact of each factor on the probability of nitrogen overuse. All three crop-specific models are statistically significant.
Table 6. Factors influencing excessive nitrogen use in Nepal: probit model marginal effects (10% above recommended rate)
Robust standard errors are in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.
Land size is negatively correlated with excessive nitrogen application across all three crops. The livestock index is positively associated with the probability of excessive use among rice and maize farmers. Those growing hybrid maize varieties are more likely to apply excessive nitrogen. In contrast, rice and cauliflower farmers cultivating hybrid varieties are less likely to overapply nitrogen.
Farmers with irrigation access, particularly rice and maize growers, have a higher likelihood of overusing nitrogenous fertilizers. Rice and maize farmers who have access to credit exhibit a higher probability of excessive nitrogen application.
For cauliflower farmers, proximity to agricultural cooperatives reduces the likelihood of excessive nitrogen use. Rice and maize farmers in the Terai region are less likely to overapply nitrogenous fertilizers compared to those in the hilly region. Maize farmers in rural areas have an approximately 18% lower probability of excessive nitrogen use compared to their urban counterparts.
To address concerns about the threshold’s arbitrariness, we conducted a sensitivity analysis using a more conservative threshold of 20% above the recommended N dose (Supplementary Material Table S1). The results demonstrated that the signs and statistical significance of most coefficients remained consistent, reinforcing the robustness of our findings. This suggests that the observed patterns of excessive nitrogen use are stable across varying threshold definitions.
Discussion
Our study highlights the suboptimal adoption of 4R nutrient management practices among smallholder farmers in Nepal. Fertilizer use varies significantly by crop and nutrient type, with 100% of rice farmers applying urea compared to 80% of maize and 88% of cauliflower farmers. Despite urea’s widespread use—Nepal’s most popular fertilizer—only 25% of rice and maize farmers used the right fertilizer source, and over 90% failed to apply fertilizers at the right rate.Footnote 6 Notably, only 4% of maize farmers applied fertilizers in the right place, and just 2% of all farmers applied the correct nitrogen dose. These inefficiencies suggest that improving fertilizer management by adopting the 4Rs could substantially reduce yield gaps corroborated with earlier studies in similar agroecological zones in Nepal (Pandit et al., Reference Pandit, Adhikari, Vista and Choudhary2025).
The observation that farmers apply fertilizers at the right place but do not consistently apply the right rate, at the right time, or use the right source offers several critical insights into the fertilizer management practices and highlights areas for targeted intervention. The correct placement of fertilizer is often easier to observe and adopt because farmers can directly see the effect of placement on crop growth (e.g., root zone application leading to better plant vigour). However, right rate, time, and source require a deeper understanding of nutrient dynamics, soil fertility, and crop nutrient requirements, which may not be as intuitive. When farmers understand best practices, resource constraints may force them to make suboptimal decisions.
Nepal’s reliance on fertilizer imports exacerbates the issue. The government spent 9,071 million NPR on fertilizer subsidies in 2020, with 74% allocated to urea, 25% to DAP, and 1% to MOP (Thapa et al., Reference Thapa, Gaihre and Choudhary2025). Yet, chronic fertilizer shortage persist—only 60% of national demand is met (Gautam et al., Reference Gautam, Gaihre, Acharya, Dongol and Choudhary2022; World Bank, Reference World Bank2016). Famers often cannot access recommended fertilizer doses during the critical growing season. A study found that just 25% of farmers could purchase their full fertilizer needs (Kyle, Resnick and Karkee Reference Kyle, Resnick and Karkee2017), and many are willing to pay significantly above market prices to secure supplies (Thapa et al., Reference Thapa, Gaihre and Choudhary2025). These systemic issues hinder the adoption of 4R practices, leading to inefficient fertilizer use despite the high economic cost.
Our MVP model revealed significant socio-economic and agricultural factors influencing 4R adoption. Older, more experienced farmers show a higher likelihood of adopting 4R practices, possibly due to their accumulated knowledge and understanding of the long-term benefits of balanced fertilizer use. Dalit households are less likely to apply the right fertilizer dose, reflecting systemic inequalities. Dalits, often smallholder farmers with lower incomes and education levels, face structural barriers to technology adoption (Kumar et al., Reference Kumar, Takeshima, Thapa, Adhikari, Saroj, Karkee and Joshi2020). Our data show household heads from Dalits have lower education levels (4.3 years) compared with Brahmin/Chhetry (6.1) and Janajati (4.7) household heads (Table S2). Further, Dalits also have higher poverty rates compared to Brahmin/Chhetri and Janajati households, limiting their access to resources and information.
Education significantly increased the adoption rate of 4R practices, suggesting that more educated farmers have a better understanding of the benefits of efficient fertilizer use. Our findings align with previous studies that identify education as a strong predictor of agricultural technology and fertilizer adoption (Omamo et al., Reference Omamo, Williams, Obare and Ndiwa2002; Takeshima et al., Reference Takeshima, Adhikari, Kaphle and Shivakoti2016). Smallholders are more inclined to adopt 4R practices, likely due to the need to maximize productivity on limited land. However, small landholders are also more prone to overapply nitrogen, aiming to boost yields.
Credit access facilitates fertilizer purchases, increasing the likelihood of correct application rates. Cooperative membership also plays a crucial role, improving access to fertilizers and information. Farmers closer to cooperatives are more likely to adopt 4R practices, underscoring the importance of local institutions in promoting sustainable agriculture. Smartphone ownership enhances 4R adoption, particularly for the correct nitrogen rate. ICT tools provide timely information and guidance, as demonstrated in other contexts where mobile-based interventions improved agricultural practices and yields (Casaburi et al., Reference Casaburi, Kremer, Mullainathan and Ramrattan2014; Giulivi et al., Reference Giulivi, Harou, Gautam and Guereña2023).
Farmers growing hybrid varieties or with irrigation access are more likely to adopt specific 4R practices, especially regarding fertilizer type and placement. Hybrid crops often require more inputs, while irrigation increases nutrient availability, encouraging higher fertilizer use. However, this can also lead to overapplication, particularly of nitrogen, as farmers aim to maximize yield potential.
The probit model revealed that several socio-economic and agricultural factors contribute to the excessive use of nitrogenous fertilizers. Smallholders tend to overapply nitrogen, aiming to boost yields on limited land. Livestock ownership and credit access increase the likelihood of nitrogen overuse in rice and maize, potentially due to higher input affordability. Regional disparities influence fertilizer practices: farmers in the Terai region apply less excessive nitrogen than those in the hills, while rural maize farmers are 18% less likely to overuse nitrogen compared to urban farmers. Proximity to cooperatives reduces nitrogen overuse, particularly among cauliflower farmers, by improving access to information on proper fertilizer use.
Hybrid crops often have higher nutrient requirements due to their enhanced yield potential, and this has important implications for defining the ‘right dose’ in our analysis. Considering this, applying a uniform recommended dose across both hybrid and landrace varieties could indeed lead to a misclassification of what constitutes excessive fertilizer use. Extension services should consider promoting variety-specific fertilizer guidelines that account for the higher demands of hybrid crops, helping farmers optimize input use while avoiding true overapplication.
Conclusion and policy implications
This study examined the adoption of 4R soil nutrient management practices among smallholder farmers in Nepal, focusing on rice, maize, and cauliflower cultivation. Using data from two survey rounds (2021 and 2022), the findings reveal that the overall adoption of 4R practices remains low, particularly regarding the right rate and timing of nitrogen application. While cauliflower farmers demonstrated relatively higher adherence to 4R practices, the majority of rice and maize farmers continue to apply fertilizers inefficiently, resulting in potential yield gaps and reduced fertilizer efficiency.
Our analysis highlights that socio-economic and demographic factors significantly influence 4R adoption. Female-headed households and older farmers are more likely to adopt specific 4R practices, suggesting that experience and gender dynamics play a role in nutrient management decisions. Marginalized groups, such as Dalits, face systemic barriers to adopting optimal fertilizer practices, reflecting broader social and economic disparities. Education is a key driver of 4R adoption, though its impact varies across specific practices. More educated farmers are better positioned to understand and implement sustainable nutrient management strategies. Farmers with credit access are more capable of purchasing the appropriate quantity and variety of fertilizers, increasing the likelihood of applying fertilizers at the right rate. Smartphone ownership enhances nitrogen management, enabling farmers to access real-time agricultural information and best practices. Proximity to cooperatives and engagement in contract farming positively influence the adoption of 4R practices by improving input accessibility and promoting knowledge-sharing. Farmers cultivating hybrid crops or with access to irrigation are more likely to adopt proper fertilizer types and application methods. However, these farmers also face a higher risk of excessive nitrogen use, underscoring the need for targeted extension services and fertilizer-use guidelines.
To promote efficient and balanced fertilizer use in Nepal, several policy interventions are recommended. Farmer education and extension services need to be strengthened. Training programmes and field demonstrations could be implemented to raise awareness about 4R nutrient management and its long-term benefits. Government could improve fertilizer accessibility and affordability. The current subsidy structure could be reformed by increasing support for DAP and MOP to encourage balanced nutrient use beyond urea. Government could provide affordable credit options to enable smallholders to invest in the recommended quantity and variety of fertilizers. The ICT tools need to be leveraged such as smartphone applications and SMS-based advisory services—to disseminate information on best practices. Cooperatives need to be strengthened as key actors in fertilizer distribution and knowledge-sharing, ensuring that smallholders have access to inputs and training. Government could implement inclusive policies that provide tailored support for disadvantaged groups, such as Dalit farmers, to bridge adoption gaps and promote equity.
Nepal’s fertilizer challenges—marked by heavy reliance on imports, inefficient use, and widespread subsidies—mirror those faced by many developing countries, particularly in South Asia. The insights from this study can inform fertilizer management strategies in similar agricultural settings, supporting global efforts towards sustainable agriculture. By promoting 4R nutrient management, Nepal can reduce fertilizer dependency, minimize environmental harm, and enhance agricultural productivity. Ultimately, these efforts align with the broader objectives of the United Nations Sustainable Development Goals (SDGs), particularly those related to food security, climate action, and responsible resource use. Future research could explore the impact of fertilizer supply chain dynamics and government distribution mechanisms on farmers’ fertilizer use patterns, as this remains a critical issue for ensuring sustainable nutrient management.
The study possesses some limitations. The study primarily relies on self-reported survey data, which can introduce recall bias, especially in estimating fertilizer application rates and timings. The absence of plot-level data and detailed information on soil fertility limits the ability to tailor nutrient recommendations or fully capture site-specific management practices. The study focuses on chemical fertilizer use but excludes the quantification of organic inputs (e.g., farmyard manure, green manures). This omission may lead to an underestimation of total nutrient inputs, particularly for farmers who integrate organic fertilizers into their management strategies. The 10% deviation from recommended nitrogen rates used to define excessive application, though practical, is somewhat arbitrary and may not accurately reflect the ecological or agronomic thresholds for overuse across different soil types and crop conditions. With data collected over just two survey rounds, the study captures a snapshot rather than long-term trends. Seasonal variability or yearly fluctuations in fertilizer availability and prices could influence farmers’ practices, limiting the generalizability of the findings. The findings are specific to Nepalese smallholder farmers and may not be directly applicable to regions with different agroecological conditions, market dynamics, or policy frameworks. Broader implications should be drawn with caution.
Our sampling strategy could still introduce minor biases: farmers who are more connected to local cooperatives or agricultural extension services may have been more likely to be included in the sampling frame, potentially leading to slightly higher adoption rates of certain best practices. The use of community registries may have unintentionally excluded informal farmers or those in more remote locations, who may exhibit different fertilizer use behaviours. Future studies could adopt larger, more randomized sampling frameworks or use longitudinal data to further minimize selection bias. Farmers selected in the survey cultivate multiple crops. We acknowledge the non-independence of observations from the same farmers growing rice, maize, and cauliflower. Our survey design was not able to focus exclusively on farmers growing an individual crop.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0014479725100203
Open access
