Exploring the Impact of Farmer Field Schools on the Adoption of Sustainable Agricultural Practices and Farm Production: A Case of Pakistani Citrus Growers

Abstract

In the wake of recent climate changes, extension services have become crucial drivers in disseminating information about the latest agriculture technologies and facilitating sustainable agricultural productivity. Pakistan’s traditional extension cannot yield the expected outcomes which corroborate the introduction of a participatory approach, mainly in farmers’ field schools. Using farm-level data from Punjab province, the current study examined the impact of farmers’ field schools (FFS) on adopting sustainable agriculture practices (SAPs) and citrus yield. The study employed recursive bivariate probit and propensity score matching to explore the objectives. The findings revealed that FFS participants had a higher SAPs adoption than non-participants, demonstrating the positive impact of FFS on the uptake of SAPs. Moreover, the treatment effect showed that FFS participants had a higher yield than the non-participants. The results suggest strengthening and enhancing FFS reach among citrus growers.

1. Introduction

After apples and grapes, citrus is the most grown class of fruit and a primary source of livelihood worldwide. Citrus fruits are always in demand due to their nutritional value and excellent health benefits for humans. With 40% of total fruit production, citrus is among the top valued fruit types in Pakistan. It is cultivated on 199,400 hectares, and its annual production stands at 2.46 million tons [1,2]. The citrus yield of Pakistan is around 11.60 tones/ha, far below the developed world (16.60 tones/ha) [3]. Citrus fruits are grown under specific climate conditions and are sensitive to climate change; citrus growth is severely affected by climate extremities such as extremely high and low temperatures, intensive rainfall, droughts, and relative humidity [4]. For the last two decades, Pakistan has been facing severe climate change, which has adversely impacted the productivity of cash crops, and citrus is no exception [5]. Due to its unmatched importance to the country’s foreign exchange and rural livelihood, it has been deemed imperative to elevate its production and attain maximum output. As the country’s agriculture vulnerability is well established [6,7], farmers need to adapt and incorporate sustainable practices in their current farming methods. Over the last few decades, a new course of research aimed at finding cutting-edge solutions to climatic uncertainties and promoting sustainable agriculture has mushroomed. Recent studies have shown that including climate-smart agriculture initiatives minimizes carbon footprints and improves farm production [8,9,10,11,12].

Contemporary scholarship emphasized that the shift toward sustainable agriculture practices (SAPs) promises greater wellbeing for consumers and producers. Henceforth, it should be promoted at local and national levels to eliminate environmental hazards and elevate farmers’ wellbeing. Specifically, adopting SAPs could bring long-term sustainability [13,14]. Accordingly, SAPs could help build resilience against environmental shocks, instigate economic opportunities, and boost farmers’ wellbeing [15]. Similarly, Gold [16] defined SAPs as a set of agriculture practices that sustainably bring resilience, reduce greenhouse gas (GHG) emissions, and elevate farm production.

The diffusion of agricultural technologies mainly depends on an effective extension setup, as most small farmers rely on public extension for the latest information [17]. However, the desired results have not been achieved due to Pakistan’s ineffective extension delivery system. The current extension setup often ignores marginal farmers; consequently, many smallholders remain out of the extension’s reach, and the adoption of the latest agriculture technologies remains low. Past research has demonstrated that traditional extension delivery systems were insufficient to educate farmers [18,19,20,21]. Ref. [22] suggested that three-fourths of Asian farmers had no contact with extension agents. There could be multiple factors behind poor extension delivery, such as lack of financial resources, inadequate training amid lack of planning, etc. [21]. Most extension systems often neglect the site-specific needs and problems of the farming community and ultimately fail to achieve the desired results, a concern also valid in Pakistan [19]. The weaknesses of the traditional extension delivery system justify the emergence of participatory approaches as a parallel extension, such as farmers’ field schools. The FFS approach was first introduced in Indonesia during the late 1980s. It then became popular and spread to other countries [23]. The formation of FFS has been considered a paradigm shift in conventional agriculture extension as it involves participatory training methods to assist farmers in developing analytical and critical thinking [24]. Through field experimentation and interactive learning, farmers acquire the skill to conduct experiments and solve agricultural problems independently; hence, they need fewer extension services [25]. To improve livelihoods and promote sustainable farming practices, FFS was introduced in Pakistan under the ambit of the fruit and vegetable development project. It covered around 20 districts of Punjab province [26]. So far, most studies have focused on the immediate effects of FFS on IPM adoption among cotton producers [27,28,29,30,31,32,33]. However, the recent FFS project was broadly aimed at sustainable agriculture practices and facilitating sustainable livelihood among fruit producers, specifically citrus producers [3,34,35]. So far, only a few studies have discussed the effectiveness of FFS projects among citrus producers. These were restricted mainly to descriptive analyses and lacked causal inference [36,37]. Hence, against this background, the current study could offer some valuable contributions. First, it explored the factors behind participation in the farmers’ field schools. Second, it examined the impact of FFS on adopting sustainable agriculture practices and, third, the role of FFS in enhancing farm production among citrus producers. It also applied recursive bivariate probit (RBP) and propensity score matching (PSM) to get an unbiased and robust piece of empirical evidence.

2. Empirical Framework

2.1. Conceptual Framework

Based on the empirical evidence [38] the current study followed the expected utility theory. The utility derived from participation in FFS can be expressed as UP(π) and utility from nonparticipation as UN($\pi$). Participation in the FFS program largely depended on utility derived from participation, i.e., UP($\pi$) > UN($\pi$). The expected utility from participation may be linked to a set of explanatory variables.

$$U_i\left(\pi \right)=\beta z_i+{\epsilon }_i$$

(1)

The utility gained from the participation was not observable; only the participation decision was observed. It can be expressed through a latent variable L(π) equal to 1; thus, farmer participating in FFS $L\left(\pi \right)=1 1 if UP\left(\pi \right)>UN\left(\pi \right)$and otherwise = 0.

Hence, the probability of participation can be described as follows.

$$\mathit{Pr}\left(L=1\right)=\mathit{Pr}\left[UP\left(\pi \right)>UN\left(\pi \right)\right]=\mathit{Pr}\left({\epsilon }_i>-\beta Z_i\right)=1-F\left(-\beta Z_i\right)$$

(2)

where F denotes the cumulative distribution function aimed at $\epsilon$. participation in FFS programs enhance the farmer’s skill, elevating the chances of adopting SAPs and citrus production. To link the farmer’s participation with SAPs uptake and farm yield, the subjects were deemed risk-neutral and aimed to maximize net returns $\pi$. This can be expressed as follows:

$$\max \pi =w\left(PQ\left(W,Z\right)\right)-RW$$

(3)

p and q represent the output price and expected output level, respectively. W and Z are a vector of input and households and farm characteristics. R reflects the vector of input prices while net returns are described as function variable inputs.

The households’ endowments, output prices, and technology choices are as follows:

$$\pi =\pi \left(R,d,P,Z\right)$$

(4)

With the application of Hotelling’s lemma into Equation (3), reduced specification for input demand and output supply can be expressed as follows:

$$W=W\left(R,d,P,Z\right) \mathrm{for} \mathrm{all} \mathrm{I}$$

(5)

$$Q=Q\left(R,d,P,Z\right) \mathrm{for} \mathrm{all} \mathrm{I}$$

(6)

2.2. Recursive Bivariate Model

The recursive bivariate (RBP) model uses a maximum likelihood approach to estimate the impact, specifically non-observable confounders. Our study intended to explore the effects of FFS participation on SAPs adoption, as both variables are binary and FFS participation is probably endogenous. Hence, we opted for the recursive bivariate model.

In the presence of continuous variables, probit models were identified via index function, where $Z_1$ is the observed FFS participation, $X_1$ is a vector of independent variables influencing FFS participation, ${\beta }_1$ is a vector of parameters to be estimated, and $e^1$ is the random error term.

$$Z_{1i}=X_{1i}{\beta }_{1i}+e_{1i}$$

(7)

$$Z_{1i}=\{\begin{array}{c}1 if Z_{1i}^{*}>0\\ 0 otherwise\end{array}$$

(8)

Similarly,

$$Z_{2i}=\partial Z_{1i} X_{2i}{\beta }_{2i}+e_{2i}$$

(9)

$$Z_{2i}=\{\begin{array}{c}1 if Z_{2i}^{*}>0\\ 0 otherwise\end{array}$$

(10)

where $Z_2$ is the observed adoption variable, $X_2$ is a vector of independent variables explaining the participation decisions of farmers, ${\beta }_2$ is a vector of parameters to be estimated, and $e^2$ is the random error term. $Z_{1i}$ enters Equation (3) as a dummy variable and measures the effect of FFS participation on SAPs adoption, which is captured by ∂.

There is dependence between the error terms in Equations (1) and (2). The two error terms have the following bivariate standard normal distribution with correlation ζ such that:

$$E\left[e_1\right]=\left[e_2\right]=0, var\left[e_1\right]=1 and corr\left[e_1,e_2\right]=\mathsf{\zeta }$$

(11)

2.3. Propensity Score Matching

Propensity score matching (PSM) approach was employed to compare the outcomes of FFS participants (treated) and non-participants (controlled) that were similar in observable characteristics, thus reducing the bias which could have occurred when groups were systematically different. It comprised two stages; in the first stage, we generated the propensity score for participating in FFS programs, and in the second, the average treatment on treated was calculated.

$$Pr(x_1)=Pr\left(P_1=1|Z_1\right)=E(P_1|Z_1)$$

(12)

where $P_1=\left\{0,1\right\}$ is an indicator of choosing to participate in FFS (j = 1) and $Z_1$ is the vector of pre-choice characteristics.

$$\mathrm{ATT}=E_{p\left(z_1\right)|D_1=1}\left\{E\left[\left(Y_1|D_1=1,P\left(Z_1\right)\right]-\right[\left(Y_0|D_1=1,P\left(Z_1\right)\right]\right\}$$

(13)

Multiple measures have been developed to match similar propensity scores’ participation and nonparticipation. In this study, we employed commonly used techniques, including kernel-based matching (KBM), and nearest-neighbor matching (NNM) methods, to estimate the treatment effects on treated.

3. Methodology

3.1. Study Area and Data Collection

The Punjab province was selected for the study based on its overwhelming share of 94% in overall citrus production. Amid citrus importance, Pakistan’s government has initiated multiple FFS projects to promote best agriculture practices, increase citrus yield, and enable sustainable livelihood among citrus producers [2,26,34,39,40].

In the first data collection stage, we isolated 19 districts solely responsible for local citrus production (Figure 1). In the second step, we selected Sargodha and Mandi bahauddin (Figure 2) as our study regions. In the third step, we selected two sub-districts from each district. Consequently, we selected four union councils from each sub-district, and in the fourth stage, two villages were selected from each union council. In the last step, we selected 22 farmers from each village. Farmers who attended the FFS class were considered participants, opposite those who did not participate in FFS classes, both living in the same district. The data collection was conducted between March and April 2021. The household representatives of farm households were interviewed using a structured questionnaire to investigate research objectives. A multistage random sampling technique selected 440 sampled farm households. Before beginning the study, a team of local enumerators was hired and trained on the survey questionnaire, methods, and objectives. Additionally, the questionnaire was completed on field pretesting to improve the survey’s quality and avoid missing information. The inclusive survey included household socioeconomic characteristics, citrus yield, FFS participation, and adoption of SAPs.

3.2. Variable Specification

The study utilized data of 440 farmers to explore the effects of FFS participation on sustainable agriculture practices and productivity. Participation in the FFS programs was subject to numerous socioeconomic and institutional factors. Based on the literature review [33,41,42,43] the study categorized these factors as farmers, farm level, and institutional characteristics. The description of these variables is explained in

Table 1. Descriptive statistics and definition of the variables.

Variable	Description	Mean	SD
Farm production	Citrus yield (maunds per acre)	290.437	38.011
Integrated soil fertility management (ISFM)	1 = If the farmer applies both organic and inorganic fertilizer, 0 = otherwise	0.374	0.484
Integrated pest management (IPM)	1 = If the farmer adopts all or a single IPM practices, 0 = otherwise	0.481	0.431
Drip irrigation (DI)	1 = If the farmer adopts drip irrigation, 0 = otherwise	0.281	0.473
FFS participation	1 = If the farmers have attended FFS trainings, 0 = otherwise	0.414	0.493
Age	Age of the farmer in numbers	44.756	13.436
Gender	1 = If the household head is male, 0 = otherwise	0.753	0.431
Education	=1 if farmer has received primary, =2 if household head received secondary, =3 if household head received matric education, =4 if household head received above matric	2.002	0.635
Family size	Number of family size scale,	5.790	3.441
ICT user	1 = If the farmer is ICT user, 0 = otherwise	0.482	0.500
Non-farm participation	1 = If the farmer participates in non-farm activities, 0 = otherwise	0.571	0.445
Farm owner	1 = If the farmer is the owner of the land, 0 = otherwise	0.605	0.409
Tenancy	1 = If the farmer is tenant of the land, 0 = otherwise	0.412	0.221
Livestock ownership	1 = if the farmer owns livestock, 0 = otherwise	0.562	0.496
Extension access	1 = if the farmer has access to extension services, 0 = otherwise	0.722	0.551
Organization membership	1 = If the farmer is a member to farmers organization, 0 = otherwise	0.145	0.353
Credit access	1 = if the farmer has access to credit, 0 = otherwise	0.450	0.208

. Refs. [36,43] observed that FFS training sessions in Sargodha, Mandi Bahauddin, Toba Tek Singh, and Layyah helped farmers to apply best agriculture practices such as irrigation management, balanced fertilizer, cultural practices, and improved verities. Similarly, [37] showed that FFS graduates were well equipped to apply soil management, integrated pest management, farmyard manure, and balanced fertilizer in Sargodha. [35] suggested that interactive learning and on-site training sessions vastly improved the farmer’s adaptive capacity. Based on the literature review [1,34,35,44] and local context, three sustainable agriculture practices (drip irrigation, integrated pest management, and integrated soil fertility management) were selected as response variables for this study (Table 1).

Drip irrigation is a crop irrigation system that involves controlled water supply delivery through a systematic network of pipes and tubes. It is an effective system that has been promoted through governmental and nongovernmental channels to manage stressed water resources [45] efficiently. Drip irrigation (DI) was taken as a dummy variable with 1 = if the farmer adopts drip irrigation, 0 = otherwise. The integrated soil fertility management practices (ISFM) practices came in a package that included the usage of chemical fertilizer, improved verities, and soil organic matter in a combination that ensured the prevention of soil degradation. The basic assumption in the ISFM approach was that each component contributed to soil fertility and productivity, as none of them could provide sustainable solutions individually. Hence, they should be used in a balanced way [46,47]. The ISFM were taken as 1 = if the farmer applies both organic, and inorganic fertilizer and improved verities 0 = otherwise. Integrated pest management (IPM) involves the growth of healthy crops with a minimum disturbance of the agroecological system through managing the pests in a natural mechanism consisting of pest scouting, differentiation, and timings [32,33,48]. The study operationalized IPM as 1 = if the farmer adopts all the elements of IPM practices, 0 = otherwise.

4. Results & Discussion

The study’s primary aim was to explore the effects of FFS participation on sustainable agriculture practices and farm productivity. The study applied the recursive bivariate probit (RBP) model and propensity score matching (PSM) to fulfill the research objectives. The results have been described below.

4.1. Descriptive Statistics

The descriptive statistics are shown in Table 1. They revealed that a maximum number of farmers had primary education with an average of 2.00, specifying that most could read and write. The average age for this study was 44.75. Most farmers had livestock holdings, and around 56% of the farmers accessed an agricultural advisory in the past twelve months. Moreover, 14% of the overall farmers were members of an organization.

The summary statistics of the t-test (

Table 2. Differences in characteristics between participants and non-participants of the farmers’ field school.

Variables	Non-Participant (264)	Participant (176)	t-Test
Farm production	287.669	294.346	−1.817 ***
Integrated soil fertility management (ISFM)	0.350	0.417	−3.400 ***
Integrated pest management (IPM)	0.554	0.439	−2.390 *
Drip irrigation (DI)	0.264	0.417	−3.400 **
Age	44.61	44.95	−0.261
Gender	0.754	0.752	0.050
Education	2.031	1.951	1.283
Family size	5.626	6.021	−1.185
ICT user	0.517	0.482	1.725 *
Farm ownership	0.642	0.554	1.842 *
Tenancy	0.377	0.461	−1.765 *
Credit access	0.556	0.571	−0.311
Livestock ownership	0.351	0.603	−1.257
Organization membership	0.163	0.120	1.243
Extension access	0.697	0.739	−0.955
Non-farm participation	0.552	0.598	−0.916

***, **, and * indicate significance at p ≤ 0.005, p ≤ 0.05, and p ≤ 0.1, respectively.

) reflected the differences between the FFS participants and non-participants. The coefficients signified significant differences concerning farm ownership and ICT usage. The statistics revealed that the average age of members was 38 years, while, for nonmembers, it was 40 years. This showed that younger generations were keener to participate in FFS. Moreover, in terms of gender, males were more likely to join FFS. Education was also on the higher side for participants. The average educational attainment for the participants was 2.03, while for non-participants, it was only 1.95. Around 55% of the farmers who received the extension service were in the participants category.

The membership in farming organizations also played a crucial role in stimulating the FFS association. As farmers, 16% of the participants were influenced by FOs membership. The results showed significant differences among FFS participants and non-participants regarding outcome variables. The FFS participants adopted a higher number of SAPs in contrast to non-participants. Furthermore, the participants obtained a considerably higher yield than the non-participants. Comparing the mean differences in household characteristics between the participants and non-participants indicated that participants were better off. However, it should be noted that mean difference comparisons could have resulted in bias as they did not consider other factors.

4.2. Goodness of Fit Test

We ran both the Hosmer–Lemesbow and Murphy’s score test to verify if there were any misspecifications in the RBP model [49,50]. The results are shown in

Table 3. Goodness of fit RBP model.

Groups	Murphy’s Score Test	Hosmer–Lemeshow Test
FFS Participation and IPM	chi2(9) = 1.85 Prob > 0.9936	chi2(9) = 21.43 with Prob > chi2 = 0.1187
FFS Participation and D. I	chi2(9) = 16.82 Prob > 0.5161	chi2(9) = 11.13 with Prob > chi2 = 0.7218
FFS Participation and ISFM	chi2 = 5.94 Prob > 0.7464	chi2(9) = 23.14 with Prob > chi2 = 0.5112

. The p values of both tests were insignificant, thus rejecting the null hypothesis of normality and validating the RBP model.

4.3. Recursive Bivariate Probit

The first stage estimate of the RBP model illustrated the farmers’ decisions to participate in farmers’ field schools. The estimated correlation coefficients ${\rho }_{\epsilon \mu }$ of all corresponding models (

Table 4. The RBP estimates the impact of FFS participation on the adoption of SAPs.

	Model 1		Model 2		Model 3
Coef.	Participation	IPM	Participation	DI	Participation	ISFM
Age	0.000 (0.004)	0.002 (0.004)	−0.002 (0.004)	0.002 (0.004)	−0.000 (0.004)	0.001 (0.004)
Gender	−0.027 (0.142)	0.139 (0.126)	−0.014 (0.140)	0.109 (0.132)	−0.057 (0.138)	0.270 ** (0.127)
Education	−0.095 (0.093)	0.121 (0.083)	−0.114 (0.096)	0.021 (0.093)	−0.093 (0.097)	0.010 (0.091)
Family size	0.019 (0.020)	−0.028 (0.018)	0.011 (0.017)	0.012 (0.016)	0.012 (0.017)	−0.007 (0.016)
ICT usage	−0.205 (0.130)	0.069 (0.122)	−0.213 * (0.123)	0.099 (0.117)	−0.220 * (0.127)	0.036 (0.117)
Nonfarm participation	0.104 (0.127)	−0.034 (0.113)	0.110 (0.123)	0.246 (0.117)	0.128 (0.122)	0.127 (0.113)
Farm ownership	−0.214 * (0.122)	0.190 * (0.108)	−0.226 * (0.124)	0.134 (0.116)	−0.217 * (0.124)	0.249 ** (0.115)
Tenant	0.216 * (0.123)	−0.146 (0.110)	0.233 * (0.122)	−0.197 * (0.113)	0.208 * (0.124)	−0.273 ** (0.114)
Livestock ownership	0.162 (0.268)	0.116 (0.235)	0.102 (0.309)	−0.023 (0.290)	0.121 (0.287)	−0.120 (0.255)
Credit access	0.054 (0.125)	−0.013 (0.110)	0.047 (0.120)	0.251 ** (0.114)	0.060 (0.122)	0.195 * (0.114)
Organizational membership	−0.211 (0.182)	−0.028 (0.163)	−0.178 (0.169)	0.338 ** (0.152)	−0.183 (0.172)	0.341 ** (0.152)
Extension access	−0.013 (0.144)	0.220 * (0.131)	0.010 * (0.141)	0.173 (0.136)	−0.009 (0.144)	0.158 (0.136)
FFS participation		1.617 *** (0.088)		1.639 *** (0.091)		1.558 *** (0.093)
Constant	−0.073 (0.339)	−1.172 *** (0.309)	0.095 (0.360)	−1.817 *** (0.364)	0.033 (0.364)	−1.526 *** (0.359)
${\rho }_{\epsilon \mu }$	−3.663 *** (0.991)		−2.001 *** (0.402)		−1.927 *** (0.538)
Wald test of ${\rho }_{\epsilon \mu }$ = 0	13.654 ***		24.770 ***		12.801 ***
Log-pseudolikelihood	−581.850		−549.165		−560.940
Number of observation	440		440		440

***, **, and * indicate significance at p ≤ 0.005, p ≤ 0.05, and p ≤ 0.1, respectively.

) differed significantly from zero, suggesting biases due to unobservable factors. Whereas the negative sign of ${\rho }_{\epsilon \mu }$ indicated that the farmers with a lower probability of investing in sustainable agriculture practices were more likely to participate in farmers’ field schools. Furthermore, the results of the Wald test for ${\rho }_{\epsilon \mu }$ = 0 in respective models were significantly different from zero, signifying that FFS participation as not exogenous. Hence, farmers’ decisions to invest in sustainable agriculture and FFS participation were correlated. Furthermore, the results of FFS impact on SAPs adoption are presented in the second, fourth, and sixth columns of Table 3. The estimates confirmed the positive and significant impact of FFS on the adoption of all SAPs. The other coefficients in the same columns also reflected that the adoption of SAPs as affected by multiple factors. we calculated marginal effects in

Table 5. Marginal effects of RBP model estimation on the marginal probability of adopting SAPs.

	Participation	I.P.M	Participation	D.I	Participation	I.S.F.M
Age	0.000	0.000	−0.000	0.000	−0.000	0.000
Gender	−0.010	0.041	−0.005	0.030	−0.021	0.078
Family size	0.007	−0.008	0.004	0.003	0.004	−0.002
Education	−0.036	0.036	−0.043	0.006	−0.035	0.003
ICT use	−0.077	0.020	−0.080	0.027	−0.083	0.010
Farm ownership	−0.080	0.056	−0.085	0.037	−0.082	0.072
Tenant	0.081	−0.043	0.088	−0.054	0.078	−0.079
Livestock ownership	0.061	0.034	0.038	−0.006	0.045	−0.035
Credit access	0.020	−0.003	0.017	0.070	0.022	0.056
Extension access	−0.005	0.065	0.004	0.048	−0.003	0.046
Organizational membership	−0.079	−0.008	−0.067	0.094	−0.069	0.099
Non-farm participation	0.039	−0.010	0.041	0.068	0.048	0.036
FFS participation		0.481		0.456		0.453

to give broader and more meaningful results by providing the magnitude of the impact of individual probabilities.

4.4. Determinants of FFS Participation

As the explanatory variables in all three models are the same, we decided to discuss them together (Table 4). The results revealed that mobile phone users were less likely to participate in FFS sessions. ICT plays a nonnegligible role in technology diffusion and is more convenient than physical modes of extension. Similarly, ref. [44] also found a negative association between mobile usage and FFS participation among the cocoa producers of Cameroon.

The coefficient of farm ownership was significant and negative, indicating the inverse relationship between farm ownership and FFS participation. The households with a secured land tenancy were less likely to participate in FFS training. In contrast, land tenancy was positively associated with participation. This suggested that farmers with less or no land rights were more likely to participate in farmers’ field schools, as an unsecured tenant is more resource-constrained—and thus, more likely to join in FFS due to their role in enhancing adaptive capacity. This was contradictory to [51], who reported the negative association between unsecured land tenancy and FFS among tobacco growers in Pakistan.

Access to extension services coefficient model 2 was positive and significantly different from zero. The positive relationship was logical and rational, as the extension personnel often mobilized and persuaded farmers to participate in FFS. Accordingly, [52] suggested that frequent extension visits increased farmers’ awareness regarding the importance of farming organizations. Consistent with [44], a significant and positive association between the number of extensions visits and FFS participation was found.

4.5. Determinants of SAPs Adoption

The estimates showed that participation in FFS programs significantly and positively impacted adoption of all sustainable agriculture practices (Table 4), as FFS participants were more likely to adopt IPM, DI, and ISFM by 48%, 45%, and 45%, respectively. The results emphasized the effectiveness of FFS programs in promoting sustainable agriculture. The FFS increased farmers’ knowledge and capacity to diagnose issues, identify solutions, establish strategies, and implement them, with or without outside help, thus facilitating adoption-related decisions. Accordingly, multiple studies [23,28,33,42,43] indicated the crucial role of FFS participation in adopting sustainable agriculture practices. The other coefficients in all three models showed that additional factors affected adoption decisions.

The results signified the positive role of credit access on adoption of ISFM and DI, as farmers with credit access were more likely to invest in drip irrigation and integrated soil fertility management by 7% and 5.6%, respectively. Agriculture technologies are capital intensive, whereas access to credit alleviates this financial constraint. Hence, farmers with access to credit were most likely to invest in sustainable agriculture and purchase necessary technological instruments to enhance farm efficiency. Accordingly, ref. [52] also reported credit access’s significant and positive effect in explaining farmer’s technology adoption initiatives.

Moreover, the findings revealed the importance of institutional factors in technology adoption decisions. The results signified the positive role of farmer organizations (FO) in explaining the adoption of DI and ISFM, as the members farmers were 9.4 and 9.9% more likely to invest in drip irrigation and integrated soil fertility management, respectively. Organizational membership is beneficial for disseminating information regarding introducing and executing the latest farming technologies. Social networks are crucial in minimizing the information cost of adopting and marketing agricultural commodities. Furthermore, these also aid farmers in arranging required credit. Therefore, the positive impact of FO membership is logical—and was empirically verified.

Likewise, [53] reported the significant and positive impact of farmer organizations considering agriculture technology adoption in rural Nigeria. Additionally, farmers with access to extension services were more likely to invest in SAPs. Results suggested that extension access increased the likelihood of participating in adopting IPM by 6.5%. Extension services disseminate information and help rural communities to adopt the latest technologies. Likewise, ref. [39] supported the considerable role of extension access in technology adoption decisions in rural Ghana. Similarly, [54,55] suggested the significant role of these institutional characteristics in influencing technology adoption decisions.

Findings suggested negative linkages between insecure land rights and SAPs adoption, as tenants faced resource constraints and were less likely to invest in soil management practices. Farmers with unsecured land rights were 5.4% and 7.9% less likely to adopt drip irrigation and integrated soil fertility management practices, respectively. Similarly, ref. [56] also reported an inverse relationship between land tenure security and technology adoption in rural Ghana.

4.6. Effects of FFS Participation on-Farm Production

The results for average treatment on treated (ATT) showed the effects of FFS participation (treatment group) on citrus yield, in comparison to the control group of non-participants (

Table 6. PSM Estimations for average treatment effects of FFS participation on citrus yield.

Mean Outcome			Treatment Effect
	Participants	Non-Participants	Differences	ATT
NNM	294.382	285.676	8.705	0.97 **
KBM	294.338	284.722	9.616	1.86 **

** indicate significance at p ≤ 0.05.

). The treatment effects differ from the mean differences reported in Table 2; the ATT estimates countered selection bias, as FFS participation could be systematically different from the non-participants.

Kernel-based matching (KBM) and nearest-neighbor matching (NNM) showed a significant and positive effect on citrus yield among the participants. This indicated that FFS participants were better off and attained higher yields than non-participants. Likewise, [33] supported the considerable role of FFS participation in improving wellbeing among Pakistani cotton farmers.

5. Conclusions

Though FFS was deemed an essential institutional factor that improved agricultural production, empirical evidence on the linkage between agricultural technology adoption and FFS participation has not been well established. The scarcity of comparative research (considering the benefits of FFS programs) has stalled the effective implementation of sustainable agriculture methods and livelihood. This research contributed to the literature in multiple ways: identifying the factors related to FFS participation, examining the impact of FFS on the uptake of sustainable agriculture practices, and exploring the effects of FFS on farm productivity. The study employed a recursive bivariate probit model and propensity score matching to explore the objectives. The findings suggested the significant roles of ICT usage, land tenancy status, and extension contacts in considering FFS participation. The adverse effects of ICT technologies showed that ICT technologies were more convenient in terms of technology diffusion. Hence, ICT technologies need to be embedded in the FFS training program. The study’s results strongly encourage the design and implementation of a digital farmers’ field school in order to hasten the technology adoption process. Farmers with insecure land rights faced resource constraints and were more likely to participate in farmers’ field schools due to their role in enhancing adaptive capacity. Thus, policymakers should prioritize the inclusion of land tenants in FFS training sessions.

Meanwhile, the findings suggested the positive role of frequent extension contacts in increasing FFS participation. The government needs to strengthen the extension delivery system, which will ultimately elevate technology adoption rates via farmers’ field schools. Field results indicated a significant and positive impact of FFS on the uptake of all SAPs. It also highlighted that the uptake of SAPs was significantly associated with credit access, organizational membership, extension contact, and land tenancy. The findings could aid in identifying and refining the critical entry points for the FFS program. The results signified that FFS participants had a higher yield than non-participants. The FFS impact was evident; as such, more emphasis on creating awareness of the FFS approach to all stakeholders from top to bottom, including research officers, extension agents, and policymakers, is recommended, considering the critical attributes of FFS methodology. Additionally, we observed a dire need for concentrated institutional efforts to strengthen FFS programs, facilitating sustainable agriculture and farmers’ wellbeing. The current study was limited to citrus growers; nonetheless, considering the broader range of FFS projects, general research should be conducted in the future for more comprehensive evidence of FFS relevance.

Our study was constrained by the fact that we had predominantly male respondents. Nonetheless, FFS projects help rural communities improve their way of life and reduce their vulnerability and poverty through empowerment. Future studies should consider gender equality and social inclusion as primary concerns. Additionally, it is essential to learn more about these things in order to improve social inclusion, expand the tools used to measure the effects of FFS programs, and use the FFS approach to its fullest potential in future activities, programs, and changes.