This study has considered a fundamental relationship to examine the impact of ZT adoption on the cost associated with various operations, the overall cost of cultivation, yield, net income, and labor use. We assume that such outcomes are a linear function of the vector of explanatory variables (Xi) and the ZT adoption dummy variable (Ci). Thus, the linear regression equation can be expressed as:

Where yi∗ is the mean outcome (i.e., operation-wise cost, overall cost of cultivation, net return, yield, and labor use), and Ci is a dummy variable (0 or 1) used for ZT adoption; Ci = 1 if the ZT is adopted and Ci = 0, otherwise. The vector Xi expresses the farm-level socio-economic and environmental characteristics of the farm household. Based on these attributes, farm households decide whether to adopt ZT for maize production. Thus, the ZT adoption decision is a farmer’s selection rather than random assignment. The index function used to evaluate the adoption of ZT by a farmer can be represented as:

Where Ci∗ is a latent variable that articulates the difference between the advantages gained from adopting ZT (∩iA) and the advantage gained from not adopting the ZT (∩iN), the household should adopt ZT if Ci∗ = ∩iA-∩iN> 0. Using farm-level and household characteristics (Xi) as explanatory variables, the term Xi∗∂ estimates the difference between the advantages obtained from adopting ZT. Here, εi expresses as the error term. From Equations 1, 2, it is evident that perhaps the relationship between ZT adoption and outcome (i.e., yield, cost of cultivation, the net return, and labor use) is interdependent. Therefore, ZT adoption may help in yield, cost of cultivation, the net return, and labor use. Conversely, farmers with better yield, income, and less labor use may be better disposed toward adopting ZT. Hence, the treatment assignment is not random, with a group of adopters being systematically different. Also, the chance of selection bias increases as the unobservable factors influence error terms of both the outcome equation (τi) and the ZT adoption equation (εi). Consequently, the evaluation of the outcome equation (Equation 1) with ordinary least squares (OLS) guides toward a bias estimation.

The authors employed the Heckman two-step selection method and instrumental variable approach (IV) to overcome the above-mentioned problem. However, limitations are also present in both these approaches, like (i) the Heckman approach based on the restrictive assumption of normally distributed errors, (ii) the difficulty in identifying an instrumental variable for the IV approach, and (iii) both IV and Heckman method impose a linear functional form assumption which means that the coefficient of independent variables is same for ZT adopters and non-adopters. Using the PSM approach, one can easily overcome the limitations of the discussed parametric methods. The PSM method does not require any assumption about the functional form to specify the associations between the outcome and the anticipators of the outcome.

India cultivated maize on 9.9 million hectares and produced 31.51 million tonnes in 2020–21. Among the non-traditional maize states in India, Andhra Pradesh (AP) stands second in terms of area under maize cultivation (0.3 million ha) and production (1.78 million tonnes). The state’s agro-climatic conditions suit maize cultivation during the kharif and rabi seasons. However, farmers mainly produce paddy (rainfed) in the kharif season and choose dryland maize farming in rabi as the crop requires less irrigation than paddy. Maize production in AP accounts for 78.61% of the total production of coarse grains. A state-wise detailed analysis shows that more than 65% of the cost is incurred as variable costs in AP. A major chunk of this cost utilized labor and machinery costs during tillage (Jat et al., 2014). Hence, we found this state most suitable as our study area. In AP, out of 81 districts, Srikakulam district stands 3rd in Maize production (0.26 million tonnes in 2019–20), and more than 80% of the farmers are involved in ZT maize production during Rabi season. Like in other districts, most farmers keep a portion of the produce for their consumption and sell it in the domestic market. Hence, we purposively select this district for a face-to-face questionnaire survey (Figure 2).

According to Census (2011), the district’s farmers number was 2,266,411. Out of the total, the number of farmers who owned land and were directly involved in agriculture was 145,576. We employed Bartlett et al. (2002) sample size calculation formula to calculate an adequate sample size for the study:

Where N = Size of the population (2,266,411), n = size of the sample population; Z = confidence interval at 95% (Z = 1.96); d = error at 5% (d = 0.05); p = proportion of target population (p = 0.5); and q = 1−p (q = 0.5). We found that 384 household data were sufficient for this study. However, we surveyed 1,189 households from 20 blocks of Srikakulam district. From these 20 blocks, we selected only those villages for the questionnaire survey where ZT maize cultivation training was provided, and some farmers have already adopted the technique and have been producing maize using the technique for the last three years. Such criteria confirm that it is farmers’ choice whether to adopt ZT in maize cultivation, and it is crucial to overcome the selection bias and present the counterfactuals. Of the 1,189 households, 801 farmers adopted the ZT method, while 388 farmers followed the conventional tillage for maize cultivation.

The study uses 17 independent variables and 11 dependent variables for analysis (see Appendix 1). Table 2 shows the descriptive statistics and mean difference test for adopting and non-adopting farmers of the ZT technology. The result illustrates that the adopters and non-adopters significantly differ in cost associated with seed and seed treatment, land preparation, fertiliser application, weed management, pest management, harvest, and irrigation. Moreover, farmers adopting the ZT maize achieve 1.47 quintal acre⁻¹ more yield, INR7047 acre⁻¹ net return, and INR3939 acre⁻¹ less cost of cultivation than farmers without ZT adoption. This implies that the benefits are not limited to input cost cutting. The farmers can reap more direct returns in the form of improved crop yield. Adopting ZT maize also saves 9.71 man-days more than conventional maize growers. Descriptive statistics also specify that ZT adopters differ significantly in socio-economic and demographic attributes. The descriptive outcomes indicate that farmers following ZT in maize cultivation have a higher education level, membership in farmers’ organisations, farm size (acre), institutional credit, crop insurance, prior experience in mechanisation, and risk-taking attitude than conventional maize farmers. We also report that farmers who are relatively aged, have a higher household size, and reside relatively far from the marketplace prefer conventional practices, which might be due to their low risk-taking ability.

The probit model ascertains the factors that influence farmers’ ZT adoption. We estimate the Variance Inflation Factor (VIF) as a statistical measure to assess the presence of multicollinearity among the independent variables. According to Appendix 2, the VIF for the independent variables is below the specified threshold of 5. This suggests that the predictors exhibit no significant correlation with each other.

The limitations of ZT technology adoption are determined using a standard probit model. Table 3 shows a statistically significant inverse correlation between age and ZT maize adoption. A one-year increase in the mean age of the population is likely to decrease the chance of ZT maize adoption by 3.8%, which is in line with the findings of Ali et al. (2014). The probit analysis supports our assertion, which is drawn from the results of descriptive statistics, that younger farmers are likelier to adopt ZT technology than their elder counterparts. A potential rationale for this phenomenon could be that younger farms exhibit greater enthusiasm for adopting innovative technologies. In comparison, relatively aged farmers tend to adhere to conventional maize cultivation techniques and are reluctant to bear the risk of adopting novel technological advancements beyond their comfort zone (Marenya and Barrett, 2007; Läpple and Van Rensburg, 2011). However, our results are contrary to those of Sodjinou et al. (2015), who reported that the age of farmers positively affects their adoption of novel farming techniques.

Nevertheless, the impact of education on the acceptability of sustainable technology like ZT is of great importance, as demonstrated by a positive and statistically significant relationship between farmers’ educational levels and the adoption of ZT technology. A one-year increase in education year in the sample means an improvement in the probability of ZT adoption by 1.4%. This can be referred to the fact that farmers are inclined to adopt new technologies when they understand the new technology. At the same time, farmers who have received more formal education may exhibit an enhanced ability to comprehend the information disseminated by agricultural experts and extension agents and make informed decisions about new technologies (Ruzzante and Bilton, 2021). Also, farmers with a formal education better predict the potential ramifications of emerging technologies on both agricultural yield and financial profitability. Consequently, there is a favourable correlation between the educational level of the household head and their inclination to embrace innovative and sustainable methods in farming, such as ZT in maize farming. This observation aligns with the results reported by Duraisamy (2002), Idrisa et al. (2012), and Huffman (2001). According to Adeoti (2008) and Nonvide (2021), farmers with higher levels of education tend to have enhanced capabilities in adapting to new challenges and effectively utilising emerging technology.

The coefficient associated with membership in farmers’ organisations demonstrates a positive and statistically significant correlation with adopting ZT technology. This result aligns with the findings of Tura et al. (2010). Conley and Udry (2001, 2010) have provided evidence to support the notion that extension services and farmers’ groups are effective channels for disseminating information among agricultural producers and improving the chance of technology adoption. The results obtained in this study are consistent with the findings published in previous studies conducted by Abdulai et al. (2011), Allagbe and Biaou (2013), Barry (2016), and Seye et al. (2016). Membership in farmers’ organisations also brings economies of scale in cultivation and marketing (Trebbin and Hassler, 2012), minimises the cost of commercialisation, and improves production efficiency (Francesconi and Heerink, 2011; Bernard and Taffesse, 2012; Fischer and Qaim, 2014). It also enables the farmers to cope with the alteration in the global value chain and prevailing market inadequacies (Meinzen-Dick et al., 2004). Group/organisation membership serves as an indicator of social capital. Social networks enable producers to exchange information and participate in peer-to-peer learning. Social organisations served as an informal form of insurance during times of crisis. If members of the group support adopting any new technology, a positive attitude toward adoption is generated in farmers’ minds, which in turn helps in technology adoption. This may be why farmers with group membership are more inclined toward ZT adoption.

The results show that household size negatively affects the adoption of ZT maize. With a 1% increase in the average household size of the population, the probability of adopting the ZT maize decreases by 2.4%. This finding is consistent with the outcome reported by Dey and Singh (2023). Household size is a proxy for labor availability (Feder et al., 1985). If the household size is large, more family labor is available, and households choose labor-intensive technologies. Conventional maize production is a labor-intensive process. It requires additional labor, particularly for land preparation, weeding, and harvest. Therefore, farm families with larger household sizes choose conventional maize farming.

Meanwhile, relatively small families tend to adopt those technologies that minimise labor requirements. ZT does not require land preparation and stubble removal from the field. The sowing process is also machine-driven. Hence, the ZT technique is a less labor-dependent process. Therefore, households with fewer members tend to adopt this technology.

The study indicates a positive association between access to institutional financing and the likelihood of adopting agricultural technologies. The likelihood of ZT adoption increases by 17.1%, with the increase of institutional credit adoption by 1%. This finding is similar to the results by Tura et al. (2010) and Idrisa et al. (2012). Farmers who have obtained institutional loans are more inclined to employ agricultural technologies. Mdemu et al. (2016) and Nonvide et al. (2018) have identified a notable constraint in the use of technology due to a lack of access to formal financial services. The mentioned studies reported that financial resources could facilitate the acquisition of agricultural inputs and the implementation of innovative technologies in agriculture. The study identifies that farmers need to purchase the double-wheel marker to adopt ZT in maize production. Small and marginal farmers with limited resources find it difficult to invest money in mechanisation. However, if farmers receive financial support from formal institutions, they invest the money to purchase the required machines. This may be why farmers with access to formal credit are more inclined to adopt ZT.

The findings in Table 3 show a positive correlation between landholding size and the adoption of ZT. This implies that farmers who own larger farm sizes are more inclined to adopt ZT practices. The study findings exhibit a resemblance to Houeninvo et al. (2020), Tura et al. (2010), Ali et al. (2018), Abay et al. (2018), and Mwangi and Kariuki (2015). Furthermore, smallholder farmers may believe that using new technology may decrease their average yield, hindering their ability to attain food security and anticipated income. In contrast, relatively large landholders assign their land under the purview of both initiatives, i.e., ZT and conventional. Given the substantial scale of their farm, they possess a sense of assurance that in the event of a fall in yield resulting from using ZT, they can effectively offset any potential income losses and guarantee food security by implementing conventional maize.

A significant and positive association exists between the extension agent and ZT adoption. A 1% increase in the interaction with the extension agent improves the adoption of ZT maize by 8.3%. Extension service can be regarded as a viable alternative to formal education in promoting adopting certain practices (Feder et al., 1985; Nkamleu and Adesina, 2000). This service is found to be most effective in areas with high levels of education. Farmers learn more about the new technology by interacting with the extension agent. We observed that farmers can clear their doubts about the technology by face-to-face interaction with the extension agents. Trusted information from extension agents regarding higher production with less cultivation cost further inspires the farmers to adopt ZT in maize cultivation.

Farmers have various options available to mitigate agricultural risks. Our results indicate that adopting a specific risk management tool is always conducive to further including supplementary risk management techniques (Velandia et al., 2009). Furthermore, our findings suggest crop insurance favours adopting other risk management strategies, such as ZT, for maize production. We observed that farmers who choose to use ZT instead of conventional maize production also believe that if reduced crop output results from their lack of proficiency in implementing this new technique, crop insurance will be an extra safeguard against potential losses.

Successfully adopting new technologies in the past inspires and boosts faith in adopting smart technologies in the future (Mwombe et al., 2014). Perhaps this may be why farmers who have bought and used new technology in the past also adopt ZT maize farming. Results show that a 1% increase in the prior experience in adopting any new technology improves the chance of ZT adoption by 22.5%. Similarly, farmers who do not want to take risks follow conventional farming. However, a 1% increase in farmers’ risk-taking capacity improves the adoption of ZT maize by 12.7%.

Table 4 shows the average impact of ZT adoption on maize farmers’ cost of cultivation (at various stages of operations), yield, net income, and human resource utilization. The results further support our descriptive analysis that ZT adoption significantly decreases the cost of cultivation and improves maize yield and net income from maize production. ZT adopters are likely to achieve INR 8376 acre⁻¹ more (average based on the matching algorithms) and save 9.5 man-days acre⁻¹ than conventional farmers. Likewise, the yield of ZT farmers is 2.52 quintal ha⁻¹ higher than conventional farmers. A detailed analysis of the operation-wise cost of cultivation shows that in ZT maize farming, land preparation cost, weed and pest management cost, and harvest cost are lower by INR 2708 acre⁻¹, INR 167 acre⁻¹, and INR 649 acre⁻¹, respectively, than conventional farming. Nevertheless, seed and seed treatment costs and irrigation costs are higher by INR 108 acre⁻¹ and INR 176 acre⁻¹ for ZT adopters than non-adopters.

The outcomes shown in Table 4 are contingent upon the assumption of conditional independence and confounding. We infer from the results that when an unobserved independent variable has the potential to impact both ZT adoption and outcome variables, there is a possibility of unobserved heterogeneity arising, which might potentially modify the importance of the impact (Rosenbaum and Rubin, 1983; Becker and Caliendo, 2007). However, determining the extent of hidden bias in nonexperimental studies poses challenges due to the absence of a suitable assessment instrument. We assessed the degree to which unobserved exogenous factors impact the significance of the estimate by employing the Rosenbaum-bounds-sensitivity calculation (DiPrete and Gangl, 2004; Caliendo and Kopeinig, 2008).

Our results indicate that each ATT value is linked to a corresponding critical level of hidden bias. This number represents a significant gamma level at which one might justify the causal inference of ZT adoption. An illustration of this may be seen in the gamma value range of 2.75–2.80 for net return (in radius matching). This range suggests that if farmers possess identical baseline variables, their likelihood of ZT adoption may increase by 175–180%. However, the extent to which ZT adoption positively affects net income from maize production may be subject to scrutiny. This implies that the magnitude of the concealed bias must be sufficiently high enough to impact the results. Additionally, the study acknowledges that most exogenous factors are treated as baseline covariates that influence the treatment and dependent variables.

Unobservable variables may introduce bias into the findings identified with the PSM technique. Therefore, this study has utilized the FILM-based ESR model to address the reliability and inherent biases associated with the PSM model. Table 5 displays the average treatment effect of ZT adoption based on the ESR method for a set of dependent variables, i.e., seed and seed treatment cost, land preparation cost, fertiliser cost, irrigation cost, disease and pest management cost, cost of cultivation, yield, the net return, and farm labor required under both actual and counterfactual scenarios.

Results show that both the PSM and the ESR methods yield comparable results in evaluating the effects of ZT adoption on various outcomes. The direction and difference between ATT and ATU are statistically significant for all those variables, where ATT associated with the outcome variable in PSM are also significant. The average improvement in seed and seed treatment costs for ZT adopters is INR 52 acre⁻¹. Results show that seed and seed treatment costs are INR 2889 acre⁻¹ for conventional farmers while INR 2996 acre⁻¹ for ZT adopters. In the counterfactual scenario, if the ZT adopters decide not to continue ZT adoption, their seed and seed treatment costs will decrease by INR52 acre⁻¹. The study identifies that seed rate does not differ significantly among the ZT adopters and non-adopters, which repudiates the myth that ZT requires less seed. However, ZT adopters mostly used pre-treated seed, which may be costlier than non-treated seed used by conventional farmers. The land preparation cost is 56% less for ZT adopters (INR 2122 acre⁻¹) than non-adopters (INR 4831 acre⁻¹). Also, the finding illustrates that if the conventional maize farmers (non-adopters) adopt the ZT practice in the future, they can save INR 2482 acre⁻¹ for land preparation.

Similarly, ZT adopters can save 8.9 and 6.8% of the weed and pest management cost compared to conventional maize farmers. Considering the counterfactual scenarios, the study identifies that if the ZT non-adopters become ZT adopters, they can save INR 116 acre⁻¹ and INR 115 acre⁻¹ for weed and pest management, respectively. The harvesting cost is 11.9% less for ZT adopters (INR 4485 acre⁻¹) than non-adopters (INR 5091 acre⁻¹). Also, the finding illustrates that if the conventional maize farmers (non-adopters) adopt the ZT practice in the future, they can save INR 512 acre⁻¹ under harvesting. Unlike harvest cost, irrigation cost is relatively higher for ZT adopters (INR 1300 acre⁻¹) than conventional farmers (INR 1225 acre⁻¹). Under the counterfactual scenario, if ZT adopters become non-adopters in the future, they can save 2.7% on the irrigation cost. However, the ZT adopters (27.56 quintal acre⁻¹) achieve a 9.32% additional yield compared to non-ZT adopters (25.21 quintal acre⁻¹). Outcomes also indicate that if conventional farmers adopt ZT practice, they can gain an additional yield of 1.37 quintal acre⁻¹. Estimating the cost of cultivation suggests that it is INR 3462 acre⁻¹ less for ZT adopters than conventional maize growers. If ZT adopters discontinue following ZT practice, their cost of cultivation will increase by INR 2891 acre⁻¹. Similarly, the net return from maize cultivation is 26.1% higher for ZT adopters than conventional farmers. Considering counterfactual scenarios, our study revealed that conventional farmers must adopt ZT practice, which may significantly improve their net income from maize cultivation by 18.9%. The study implies that ZT adoption is beneficial in improving farmers’ net return in many ways. For instance, labor use can be significantly saved in ZT practice. Furthermore, if we monetise the gains, zero tillage can significantly improve the livelihood of maize farmers.