What Drives the Use of Organic Fertilizers? Evidence from Rice Farmers in Indo-Gangetic Plains, India

: There is a growing concern about the sustainability of agriculture due to the indiscriminate use of chemical fertilizers in several parts of the world. In India, the Indo-Gangetic Plain (IGP) is a region where the externalities of excessive use of chemical fertilizers for cereal production manifest in groundwater pollution, air pollution due to emissions, and soil degradation. In this context, we study the adoption of organic fertilizers in the region and the determining factors. We use data collected from 400 rice farmers to empirically test the link between farmers’ perceptions, revenue expectations, socioeconomic factors, and the policy environment to adopt organic fertilizers. We use Tobit and Cragg’s double hurdle model to study the farmers’ expenditure and adoption of organic fertilizers, respectively. The results show that only 32% of the farmers adopted organic fertilizers in the region. Further, membership in farmer organizations, training, and education are the key variables that determine the adoption of organic fertilizers, in addition to a positive perception of the beneﬁts of their usage. The ﬁndings highlight the need for efﬁcient extension efforts in organic fertilizers and suggest policy interventions that promote collective learning through farmer groups. randomly selected four villages (Kalampura, Kachhwa, Sangohi, and Landhora from the Karnal block; Basauli, Dhobauli, Siswan, and Bharohia from Bansgaon block). In the ﬁnal stage, 50 farmers were selected randomly from each of the eight selected villages that enabled us to survey the farmers and collect data from a total of 400 rice farmers, of which only 32% of farmers adopted organic fertilizers.


Introduction
Fertilizers have played a crucial role in raising global food production and achieving food security [1]. The close causal link between global cereal production and chemical fertilizer consumption is widely acknowledged [2]. Though regional variations exist, in general, Nitrogen (N) fertilizer use is estimated to have contributed 40% to the increases in the world's per capita food production [3]. N fertilizer's direct effect on food production has driven its global consumption growth upwards of nine times from the consumption level of the 1960s [4]. However, with increasing N consumption, more unutilized N is also released to the environment through leaching, volatilization, nitrification, and denitrification [5][6][7], since crop uptake amounts to only about 30% to 50% of the total fertilizers applied to soil [8]. Overuse of chemical fertilizers leads to soil and water contamination issues and greenhouse gas emissions, thus polluting the environment [9,10]. Continuous overuse of fertilizers can negatively affect the soil quality and structure of the soil microbial community [11], resulting in the decline of soil organic matter and faster acidification of soil [12]. Thus, reducing chemical fertilizer application without threatening food security to maintain agriculture sustainability is a challenge [13]. In India, being an agriculturedependent nation, the situation warrants immediate attention considering the present level of fertilizer use, the use efficiency of nutrients, and the emissions and leaching that affect the environment [14].

Data
The study utilizes the data collected from 400 rice farmers of the IGP, India, from March to June 2020. The IGP region of India is vast, spanning from Punjab in the Northwest to West Bengal in the East [40]. Farmers there follow an input-intensive rice-wheat cropping system, and a large number of them grow Basmati rice, which fetches premium prices in the international market [41]. Yield stagnation, the decline in the groundwater table, soil degradation, and atmospheric pollution questions the sustainability of rice cultivation in IGP [42]. Nevertheless, excessive fertilizer use, especially N, continues in the region, leading to nitrate leaching and groundwater pollution [43]. Along with this, the inadequate use of organic manures increases the risk of low yield response of the crop [44]. Judicious use of chemical fertilizers, along with organic fertilizers, as per the results of a soil test, is the key to sustaining the cropping system [45]. These prevailing cropping practices and resulting sustainability and environmental concerns encouraged us to select IGP to study organic fertilizer usage. Though both rice and wheat are grown in the study area in rotation, the farmers' expectations on returns are higher from rice than wheat. This is because farmers grow wheat mainly for government procurement agencies and/or the national market, as well as rice for export markets. Due to this, we presume that farmers make better decisions on input use (including that for organic fertilizers) in rice in comparison to wheat. Hence, we select rice as the crop of our interest in this study. We used a multistage sampling technique to collect the primary data from the IGP (Figure 1). In the first stage, we randomly selected Karnal from the Upper Gangetic Plains and Gorakhpur from the Middle Gangetic Plains, among the region's districts. These districts fall under different transect zones of IGP and have varying levels of agrarian dynamism. In the second stage, we selected one block from each district (Karnal block from Karnal district and Bansgaon block from Gorakhpur district) based on the maximum area under rice cultivation. We randomly selected four villages (Kalampura, Kachhwa, Sangohi, and Landhora from the Karnal block; Basauli, Dhobauli, Siswan, and Bharohia from Bansgaon block). In the final stage, 50 farmers were selected randomly from each of the eight selected villages that enabled us to survey the farmers and collect data from a total of 400 rice farmers, of which only 32% of farmers adopted organic fertilizers.

Data
The study utilizes the data collected from 400 rice farmers of the IGP, India, March to June 2020. The IGP region of India is vast, spanning from Punjab in the N west to West Bengal in the East [40]. Farmers there follow an input-intensive rice-w cropping system, and a large number of them grow Basmati rice, which fetches prem prices in the international market [41]. Yield stagnation, the decline in the groundw table, soil degradation, and atmospheric pollution questions the sustainability of ric tivation in IGP [42]. Nevertheless, excessive fertilizer use, especially N, continues i region, leading to nitrate leaching and groundwater pollution [43]. Along with thi inadequate use of organic manures increases the risk of low yield response of the [44]. Judicious use of chemical fertilizers, along with organic fertilizers, as per the re of a soil test, is the key to sustaining the cropping system [45]. These prevailing crop practices and resulting sustainability and environmental concerns encouraged us to IGP to study organic fertilizer usage. Though both rice and wheat are grown in the area in rotation, the farmers' expectations on returns are higher from rice than wheat is because farmers grow wheat mainly for government procurement agencies and/o national market, as well as rice for export markets. Due to this, we presume that far make better decisions on input use (including that for organic fertilizers) in rice in parison to wheat. Hence, we select rice as the crop of our interest in this study. We u multistage sampling technique to collect the primary data from the IGP ( Figure 1). I first stage, we randomly selected Karnal from the Upper Gangetic Plains and Gorak from the Middle Gangetic Plains, among the region's districts. These districts fall u different transect zones of IGP and have varying levels of agrarian dynamism. In th ond stage, we selected one block from each district (Karnal block from Karnal distric Bansgaon block from Gorakhpur district) based on the maximum area under rice cu tion. We randomly selected four villages (Kalampura, Kachhwa, Sangohi, and Land from the Karnal block; Basauli, Dhobauli, Siswan, and Bharohia from Bansgaon bloc the final stage, 50 farmers were selected randomly from each of the eight selected vil that enabled us to survey the farmers and collect data from a total of 400 rice farme which only 32% of farmers adopted organic fertilizers.   The data required for the study were collected using a structured schedule that consisted of questions on the demographic characteristics, household characteristics, farm characteristics of individual farmers, and farming practices followed by them. We also included questions on farmers' perceptions of organic fertilizers and their risk preferences. Table 1 presents the descriptive statistics of the variables used in the study. While all the farmers used chemical fertilizers, only 32% used organic fertilizers in their rice fields. Although Farm Yard Manure (FYM) is the major organic fertilizer used by the farmers in the sample, some have also used biofertilizers and commercial organic fertilizer products. For analysis in this study, we consider FYM use and expenditure synonymously to organic fertilizer use and expenditure. Farmers, on average, spent five times more on chemical fertilizer in comparison to organic. We calculated a risk score for individual farmers using their responses to a set of questions specially designed for the purpose. The farmers' average risk score was 3.38, indicating a relative risk-preferring group (1 = most risk-averse, 5 = most risk-preferring) who are ready to invest in newer technologies. While 17% of the farmers have attended at least one training on organic fertilizers, 26% happened to be members of any farmer organization or cooperative.

Methodology
We analyzed the data in two stages-the fertilizer expenditure models in the first stage, and the adoption and use of organic fertilizer in the second stage. We started by estimating the fertilizer expenditure models for both chemical and organic fertilizers. Since all the farmers used chemical fertilizers, we used Ordinary Least Square (OLS) regression to estimate the determinants of chemical fertilizer expenditure. When the dependent variable was censored, that is, when the observation accumulated at the limit of the range of the variable, the OLS model was not useful. The lower limit of the range was zero in the case of variable organic fertilizer use, and observations accumulated at this limit. The Tobit model, a censored regression model, was thus used to estimate the covariates of organic fertilizer expenditure, which was appropriate since only 32% of the farmers used organic fertilizers. We used the same set of independent variables in both the models to compare and contrast the differences in fertilizer expenditure behaviours of farmers.
We specify the Tobit model as: The latent variable Y i * in Equation (1) is a stochastic variable to measure the organic fertilizer expenditure in Rupees (INR) per hectare rice area by the farmer i. X i represents the set of socioeconomic variables that can affect the adoption of organic fertilizers, and P i includes the variables to capture the farmer's perception of organic fertilizers. We assumed that the attitude towards risk can also affect the adoption of new technologies by the farmers. Hence, the variable RS i is included in the model to represent the risk score of the farmer. The final term ε i in Equation (1) represents the random disturbance term, while β 1 , β 2 , and β 3 refer to the coefficients of the parameters to be estimated. Equation (2) provides the relation between the observed expenditure on organic fertilizers Y i and the latent variable Y i *.
Next, to estimate the covariates of expenditure on chemical fertilizers by the farmers, we used OLS regression. The OLS approach was appropriate since all the farmers surveyed used chemical fertilizers. The OLS model was specified as in Equation (3), where Y i represents the farmer's expenditure on chemical fertilizers in INR per ha, and all the other variables are the same as in Equation (1). Finally, the parameters to be estimated by the model are α 1 , α 2 , and α 3 . We also performed a variance inflation factors (VIFs) test, following OLS regression, to detect multicollinearity among the independent variables. We estimated the organic fertilizer use (quintals per hectare rice area) by Cragg's double hurdle model in the second stage. Unlike the first stage, where the prime objective of the analysis was to compare the expenditure behaviours of farmers between chemical and organic fertilizers, here we estimated the covariates of adoption and use (measured in physical quantity) of organic fertilizers. Upon data examination, we found that 68% of farmers reported using no organic fertilizers (nonadopters). The Tobit model of the first stage could be applied under the assumption that there being zero observations was due to economic factors, but it cannot explain there being zero observations due to nonparticipation. Cragg's double hurdle model can be useful in this aspect. The advantage of using Cragg's double hurdle model is that it allows considering the hurdles that the farmers need to overcome to become adopters of organic fertilizers at first, and then provide the estimates of variables that determine the level of organic fertilizer use [1]. It allows separate stochastic processes for adoption and level of use decisions. These are separately captured as the selection and outcome equations in the model, as specified below: The adoption of organic fertilizer is represented by the latent variable Z i * and the level of organic fertilizer use by the latent variable Y i *. While β 1 , β 2 , and β 3 , as well as α 1 , α 2 , and α 3 , represent the parameters to be estimated, the terms u i and v i denote the errors that are assumed to be disturbed normally and independently with zero mean and a constant variance [46]. The relation between observed expenditure Y i and latent expenditure variable Y i * is as per Equation (6).
The motives behind adopting organic fertilizers and the quantity of use of organic fertilizers are complex. Therefore, we assumed the set of independent variables that affect the outcome equation and selection equation were the same, except for the set of variables on perception towards policy variables for promoting organic fertilizer use. This is because the farmers who have not adopted organic fertilizers would be handicapped in terms of formulating appropriate perceptions towards these variables.

Comparison of Adopters and Nonadopters
Among the sampled farmers, only 32% were adopters of organic fertilizers. In Table 2, we provide adopters' characteristics to test if they are different from the nonadopters in any aspect. The difference among adopters and nonadopters in the observed characteristics were tested using pairwise t-tests. We found the adopters' ages and farming experience to be significantly lower than that of the nonadopters and their education to be significantly higher. There was a significant difference in other variables such as training, membership, asset, farm size, tenure security, the holding of the soil health card, and perception towards organic fertilizers. Interestingly, there was no difference among the two groups in the chemical fertilizer expenditure and quantity used, indicating that even the adopters of organic fertilizers are not ready to use lesser quantities of chemical fertilizers.  Table 3 presents the regression results of expenditure for both chemical and organic fertilizers. It shows that the farm size and soil health card holding affected the chemical fertilizer expenditure positively and significantly. The larger farmers spent more on chemical fertilizers, suggesting that they used a higher amount of fertilizers, which may be due to their higher ability and purchasing power to buy the chemical fertilizers. However, the holding of the soil health card is supposed to reduce the chemical fertilizer use since the card provides the details on the soil fertility status of the farms and the nutrient requirements specific to the plot. As realized from the farmers' discussion, many of them could not use the information provided in the card effectively to decide on fertilization. Additionally, mostly only the large and educated farmers valued the soil health card's information. Farmers belonging to the disadvantaged section spent more on chemical fertilizers. They often depend fully on farming to earn their livelihoods, unlike their counterparts with multiple income sources. Hence, they invest more in chemical fertilizers with the expectation of extracting the maximum output from the crop. We also found that as the farmers' experience in farming increases, they reduce the chemical fertilizer expenditure. Among the other variables in the model, the farmers' perceptions that shifting to organic fertilizers will increase the pest and disease attack affected the expenditure in chemical fertilizers positively as expected. The variable's sign is negative since the variable captures the perception on a five-point scale in which higher values disagree with the statement. All the VIF values were less than 3 in the VIF test, suggesting our model specification to be devoid of multicollinearity. Interestingly, a different set of variables influences the expenditure on organic fertilizers, among which those referring to the perception of farmers towards organic fertilizers were key. Among these, the farmers who disagreed with the statement that organic fertilizer use will reduce the yield and agreed that it would provide a better price to the output spent more on the organic fertilizers. Additionally, the farmers' contentment about the policy environment variables such as the current level of extension support and subsidies received and the status of farm certification encouraged organic fertilizer expenditure. Among the farmer characteristics, age and education were the crucial ones that affected the expenditure decision. Well-educated and young farmers were likely to spend more on organic fertilizers. Membership in farmer associations/cooperatives and farm size were the other determinants that positively affected the decision to invest. As large-sized farmers have better access to the information and can invest more, it is logical that they show a higher expenditure. The results are in line with Chen et al. [1]. The significance of membership re-emphasizes the fact that technology adoption can be better if promulgated through such groups.

Determinants of Organic Fertilizer Adoption and Use by Rice Farmers
The determinants of organic fertilizer adoption and use were estimated using Cragg's double hurdle model. Table 4 presents the results and shows both the selection and outcome models. The organic fertilizer adoption decision (selection model) is significantly affected by the effects the farmers perceived with regard to organic fertilizer adoption. Their perceptions of organic fertilizer's effect on yield, output price, pest and disease attacks, and market acceptance were all associated with the adoption decision. The farmers who perceived that the yield would not be reduced and that pests and disease attacks will not be higher due to organic fertilizer being found to have a higher chance of adoption. Membership in farmer organizations and participation in organic fertilizer training also determined the adoption positively and significantly. The younger farmers showed better adoption, and the number of years of formal education also significantly affected it. Most of the variables that affected the adoption decision also significantly affected the level of organic fertilizer use (outcome model). Farmers who are young, educated, members of farmer organizations, participate in organic fertilizer training, have larger farm sizes, and have positive perceptions of organic fertilizers adopt organic fertilizers and use them at levels higher than the others. However, the perception that organic fertilizer use will increase pest and disease attacks and provide better acceptance in the market affected only the adoption and not the level of organic fertilizer use. 574.14 *** Log-likelihood −203.13 Notes: Standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

Discussion
Our analysis results provide meaningful insights into the organic fertilizer adoption and use by the farmers in the IGP region of India. All the explanatory variables included in the models were selected based on their hypothesized relation with the adoption decision and use level. We hypothesize that the farmers' expenditure on fertilizers is directly in line with their prospect revenue variables, such as the effect on yield, price, and the risk involved (in switching partially or wholly from chemical to organic fertilizer or vice versa). Farmers' perceptions of organic fertilizers were captured using these variables, and their significance in the model is in line with other studies [1,9]. The farmers who expect better yields and revenues with chemical fertilizers will continue to use them under normal circumstances unless they believe that supplementing the fertilizer mix with organic components could make a difference to their revenues. The prospect revenue variables, except risk, were found to encourage organic fertilizer use. The risk score did not affect the farmers' decisions, which is not in line with previous studies [47,48]. As noted earlier, the farmers in the IGP are those who adopted the green revolution technologies and contributed significantly to India's cereal production. The grain output from this region is generally exported or procured by the government agencies for distributing through the public distribution system; hence, they are ensured to receive a minimum price (Minimum Support Price) [49]. This may be the reason for the nonsignificance of the risk score of the farmers.
The farmers' perceptions and confidence in the existing marketing arrangement is another aspect that could play a part in decision-making. The need for contracts and their membership status in the farmer organizations/cooperatives depict this variable. Both of these variables encouraged organic fertilizer use significantly. The literature also suggests [36] that membership in farmer organizations will help the farmer to navigate the proper fertilization strategies by reducing excessive chemical fertilizers, which could prompt them to use more organic supplements. Training on organic fertilizer use is another variable that had a positive effect on its adoption. This finding complements the literature [1,50] and suggests that the efforts of the government to educate farmers through training on different types of organic fertilizers, their suitability to different soil conditions, the time lag for the effects to be visible, and the method of adoption will pay off heavily in the future by helping them to embrace sustainability in the production systems. Socioeconomic characteristics such as age, education, and the farming experience can also affect the expenditure decision since younger lots are risk-takers, and education gives them access to better information. Additionally, higher farming experience results in a better understanding of the farm nutrient needs and helps realize the significance of organic fertilizers.
Among others, the set of variables that are part of the policy environment, including subsidy, the need for certification of farms or produce, and extension support, also encourages organic fertilizer adoption. Subsidies can help increase the level of use because the adopters in the region cannot use the recommended level of organic fertilizers, which may be due to monetary constraints. Since they understand the benefits of organic fertilizers and have crossed the adoption barrier, they may be better it if they receive support in the form of subsidies. Government actions also echo this, since there is a greater emphasis on organic fertilizers in several recently implemented agricultural development schemes [33]. The future impetus will be on the Decision Support Systems (DSSs) that enable assessing the effectiveness of organic fertilizers locally or regionally [51]. Such an organic fertilizer promotional strategy utilizing a DSS will go a long way in the journey of Indian agriculture towards a sustainable future [52,53].
To summarize, in this study, we have extracted the determinants that could affect the adoption and use of organic fertilizers. Although our analysis provides several meaningful insights, there are some limitations as well. The dependency between the two hurdles in the double hurdle model is assumed based on the underlying theory and not empirically tested, which is a limitation of the model. Additionally, we have considered only one among the two crops commonly cultivated within the region, and hence there are chances of underestimating the actual quantity of organic fertilizers applied to the farms. The results should thus be interpreted cautiously, considering the variation in farmers' perceptions and the crop and soil requirements across regions. Still, our findings can be valuable to areas with similar agroecological and socioeconomic characteristics to the IGP. Further research in this area can focus on the local social networks prevalent in the region, giving a micro view on the agencies and entities to be tapped to encourage better adoption and use of organic fertilizers. Randomized control trials on organic fertilizers are another future research option that can prove the technology's worth and spread the results fast among the stakeholders. Additionally, studies utilizing time-series data on similar aspects can be of great value in the future since they can provide better insights than cross-sectional data.

Conclusions
Using data from a comprehensive survey of rice farmers in IGP, India, we studied the expenditure on organic fertilizers and the factors determining its adoption and quantity used. We used the Tobit model to study farmers' expenditure in organic fertilizers and Cragg's double hurdle model to identify the factors determining its adoption and level of use in farming. We add valuable information to the literature by linking farmers' future revenue expectations, the prevailing marketing arrangements, socioeconomic factors, policy environment, and perception of technology with technology adoption and intensity of use. The findings are relevant since the adoption of organic fertilizers in the region was found to be lower, hence policy options should be suggested to increase its usage. We found that, in general, young, educated farmers who are members of farmer organizations, attended training, and positively perceived the effect of using organic fertilizers were the adopters and showed higher levels of use. Our findings have some policy implications as well, especially concerning the future strategies of soil fertilization in the region. First, the government should encourage the farmers to join together, form groups, and make collective farming decisions suitable for the local soil properties. The uneducated, older, and untrained farmers could also experience the benefits of organic fertilizers. Additionally, it will be more comfortable and practical if training is provided to farmer groups than individual farmers. Second, since the positive perception towards organic fertilizers plays a crucial role in its adoption and level of usage, more efforts to popularize the benefits of the technology should be undertaken by the government since it takes time before the technology reaches the majority. The problem with organic fertilizers is that their benefits may not have immediate visibility. The extension system has a huge role to play here to assure the farmers of the benefits of continuing with its usage. Third, although the subsidies are being criticized for being the most common government strategy for promoting chemical fertilizer usage, an altered strategy of providing a part of these subsidies to promote organic fertilizers can be effective in improving its usage. However, it is necessary to act beyond subsidies and create an enabling environment that can form farmer groups, avail necessary training, and perceive the actual benefits before they are reaped. Finally, if generated from different locations and conditions, more research evidence can help validate the findings that can guide policymaking to benefit a broader set of farmers, nationally and internationally.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.