Factors Influencing the Adoption of Agroecological Vegetable Cropping Systems by Smallholder Farmers in Tanzania

Abstract

Vegetable production is vital to smallholder farmers, who often struggle to overcome pests, diseases, and extreme weather. Agroecological cropping systems offer sustainable solutions to these issues but their adoption rates in Tanzania remain low. This study examines the factors influencing smallholder farmers’ adoption of selected agroecological cropping systems for vegetable production in Tanzania, which remains underexplored. Using a multistage sampling technique, cross-sectional data were gathered from 525 crucifer and traditional African vegetable farming households within the Arusha and Kilimanjaro regions. Multivariate probit regression analysis, which accounts for the simultaneous adoption of multiple systems, revealed several significant variables influencing adoption. The number of training sessions attended and access to market information positively influenced adoption (p < 0.01), while gross income from vegetable production also had a positive influence (p < 0.05). Conversely, the age of the household head and the region where the farm was located showed negative effects on adoption (p < 0.05). These findings highlight the need for targeted extension services and training sessions focusing on the benefits, methods, and management techniques of agroecological cropping systems. Gender-sensitive policies and interventions should also be developed to address the factors influencing the adoption of agroecological cropping systems.

1. Introduction

Vegetable production in sub-Saharan Africa (SSA) significantly enhances nutrition and food security, improves household income, creates employment opportunities, and generates foreign exchange [1,2]. Tanzania stands out as a leading vegetable producer within SSA, ranking among the top 20 globally [3]. The sector has experienced rapid growth with an average annual rate of 11% in recent years, making it the fastest-growing subsector within Tanzania’s agricultural sector [4,5]. Smallholder farmers, who operate on less than two hectares of land, are the backbone of vegetable production in Tanzania [4,6]. They grow a variety of crops that include cruciferous vegetables, such as cabbage, kale, and cauliflowers, along with traditional African vegetables like amaranth and African nightshade for their nutritional benefits and health advantages [1,3,4,7].

Despite the significant contributions of vegetable production to the nutrition, food security, and economic development of Tanzania, smallholder farmers struggle to overcome various challenges that limit productivity and profitability [1,3,4,7]. Pests and diseases are rampant, frequently devastating crops and causing economic losses by increasing the cost of production [1]. Soil degradation intensifies these challenges since declining soil fertility reduces productivity and prevents farmers from fully realizing their crop production potential [2,8]. Additionally, smallholder farmers, who often rely on rain-fed agriculture, are particularly vulnerable to extreme weather conditions, such as droughts and floods, which are becoming more frequent and severe due to climate change [9,10,11]. These challenges highlight the urgent need for sustainable agricultural practices that can enhance resilience and productivity [12].

The agroecological approach to vegetable production offers a promising solution, potentially mitigating these challenges faced by smallholder vegetable farmers while promoting environmental sustainability and improving livelihoods [13]. Agroecology is defined as the application of ecological and social concepts and principles to the design and management of sustainable agroecosystems [13,14,15]. Following agroecological guidelines, a variety of practices have been developed to improve the ecological functioning of cropping systems, such as intercropping, crop rotation, cover cropping, and agroforestry [16,17]. These emerging systems are commonly referred to as agroecological cropping systems [18] and have been shown in numerous studies to offer significant benefits to Tanzania’s vegetable production.

For instance, ref. [19] found that vegetable farmers in Morogoro, Tanzania, who implemented agroecological practices like intercropping incurred lower costs of production and achieved greater gross margins than those using conventional practices, such as industrial fertilizers and pesticides. Similarly, ref. [20] found that the use of appropriate intercropping combinations for crop diversification by farmers in Dodoma and Tabora, Tanzania, resulted in enhanced crop yields and improved agroecosystem resilience. Additionally, ref. [21] found that smallholder vegetable farmers in the Uluguru Mountains, Tanzania, who adopted agroecological practices, including crop rotation and intercropping, experienced higher quality produce with fewer disease incidents. The study by [22] in Morogoro, Tanzania, also found that the implementation of agroecological practices, including cover cropping, crop rotation, and intercropping, had a positive effect on food security and the well-being of smallholder farmers. Despite the potential benefits, the adoption of these cropping systems in Tanzania is still generally low [17,23]. Understanding the factors driving the adoption of these cropping systems is essential, as their benefits can only be fully attained when smallholder farmers adopt them [24].

Studies carried out in Latin America and Asia on the factors influencing the adoption of agroecological cropping systems have highlighted the roles of information access, education level, and access to agricultural training in the adoption of these systems [25,26]. In SSA countries such as Kenya and Uganda, factors such as gender, literacy levels, household size, and credit access have been identified as critical determinants [27,28,29]. Nonetheless, the literature concludes that no single set of factors universally explains adoption; instead, adoption varies with the households considered, the technology, and the specific location [25,30]. Literature on the factors influencing the adoption of agroecological cropping systems in Tanzania is sparse [31,32]. The few existing studies, such as those by [17,33], indicate that factors such as gender dynamics, credit access, distance to a market, and household income play a significant role in the uptake of these systems. However, most of this research is heavily focused on widely consumed cereal crops, with only a few studies examining the factors influencing the adoption of agroecological cropping systems specifically for vegetable production, which plays a critical role in food security and household income diversification [1,3,34,35,36]. Furthermore, while previous studies have explored some of these cropping systems, such as intercropping and crop rotation, they often analyze them in isolation using univariate models [27,34,37]. This approach overlooks the adoption of multiple cropping systems and their interaction effects, which are crucial since some systems need to be implemented concurrently or sequentially at different production stages to be effective [25,27].

This study therefore seeks to address these research gaps by examining the factors influencing the concurrent adoption of five agroecological cropping systems among smallholder vegetable farmers in Tanzania. The cropping systems considered include mixed cropping, strip cropping, row intercropping, border cropping, and crop rotation (Table S1). These systems were selected because, despite being indigenous to the research areas and actively promoted, their adoption levels remain low [38]. By focusing on multiple cropping systems and their simultaneous adoption, this study seeks to overcome the limitations of previous univariate analyses and provides a more comprehensive understanding of the adoption dynamics in vegetable farming.

2. Materials and Methods

2.1. The Study Area Description

The study was conducted in the Arusha and Kilimanjaro administration regions, two of the five regions located in northern Tanzania, as shown in Figure 1. Figure 1A provides a broader view of Tanzania, highlighting the selected study regions, while Figure 1B offers a more detailed view of the sampling points within Arusha and Kilimanjaro. Arusha (3.3869° S, 36.6829° E) covers an area of approximately 37,576 km² and consists of seven districts, with an estimated population of 1.7 million people [39,40]. The region experiences average daily temperatures of between 18 °C and 22 °C, with rainfall ranging from 1800 mm annually on Mount Meru to 508 mm in the semi-arid plains [40]. The elevation of Arusha ranges from 900 to 1600 m above sea level. The region’s key economic activities revolve around agriculture, particularly the cultivation of coffee, as well as various vegetables, such as cabbages, broccoli, spinach, and amaranth [40,41,42]. In addition to agriculture, the region engages in mineral extraction, mining resources such as magnesite and meerschaum, as well as deposits of salt, mica, saltpeter, and ochre [40].

The Kilimanjaro region (3.1508° S, 37.5842° E) spans about 13,250 km² and is also divided into seven districts, with a population of about 1.6 million people [39,43]. The region experiences average annual rainfall ranging from 500 mm in the lowland areas to over 2000 mm in the mountainous regions, with temperatures varying between 15 °C and 30 °C [43]. The elevation of the Kilimanjaro region ranges from 800 to 5895 m above sea level. The fertile volcanic soils and favorable climatic conditions found on the slopes of Mount Kilimanjaro create ideal conditions for agricultural production, supporting roughly three-quarters of the region’s population [44]. The most common crops grown in the region include cabbage, peas, tomatoes, maize, and beans [44,45].

The Arusha and Kilimanjaro regions were chosen for this study due to their significance in vegetable cultivation, which plays a crucial role in local food systems and livelihoods [42,45]. The regions’ diverse climates, population density, and agricultural intensity provide a rich context for understanding the dynamics of agroecological cropping systems.

2.2. Sample Size

The sample size, representing the number of household farmers, was determined using the sampling methodology outlined by [46], as follows:

$$n_o={\displaystyle \frac{z^{2}pq}{e^2}}$$

(1)

where

n_o = sample size, number of households;

z = standard value at a given confidence level (α = 0.05);

p = the estimated portion of the population exhibiting the attribute in question (assuming maximum variability);

q = the complement of p, calculated as 1 − p;

e = the desired precision level.

This study utilized a confidence level of 95% and a precision level of ±5% [39]. An assumed variability of p = 0.5, representing approximately 50% of the population was used. This assumption was made because of the large size of the study population and the lack of information regarding its variability [39]. The sample size was determined as follows:

$$n=\frac{{\left(1.96\right)}^2\left(0.5\right)\left(0.5\right)}{{\left(0.05\right)}^2}=385\mathrm{farmers}$$

(2)

To account for possible data collection errors or missing information, the initially targeted sample size was increased from 385 to 550 respondents [39]. However, the actual number of respondents during the survey was 525.

2.3. Sampling Procedure

The study used a multistage sampling method to identify households involved in crucifer and traditional African vegetable farming in Tanzania [39]. The first stage involved the purposive selection of two regions, Arusha and Kilimanjaro. In the second stage, key informant interviews were conducted with district agricultural officers and staff from the AGROVEG project (which focuses on promoting intensified agroecological-based cropping systems to enhance food security, environmental safety, and the income of smallholder producers of crucifers and traditional African vegetables in East Africa) to purposively select eight wards in Meru district (Ambureni, Akheri, Kikwe, Maroroni, Ngarenanyuki, Nkoanrua, Seela Sing’isi, and Knoanekoli) and eight wards in Moshi district (East Kahe, East Kibosho, Kahe, Kirima, North Mwika, Njia Panda, West Makuyuni, and West Marangu) in the Arusha and Kilimanjaro regions, respectively. In the third stage, villages were randomly selected within each ward based on the estimated population of crucifer and traditional African vegetable farmers [39]. In the last stage, 525 households engaged in crucifer and traditional African vegetable farming were randomly chosen from the selected villages. This sampling procedure and dataset are the same as those used in our previous publication [39].

2.4. Data Collection

Data were collected using structured and pretested questionnaires designed in CSPro version 7.6 (Census and Survey Processing System), a software tool used for data entry, validation, and management and administered through computer-assisted personal interviews (CAPI). The questionnaires were delivered by a team of eight enumerators who were well trained in CS entry and had a good understanding of the local language [39]. Developed after a comprehensive review of the existing literature, the questionnaire sought to gather information on the factors influencing the farmers’ adoption of agroecological vegetable cropping systems. It covered aspects such as the type of agroecological cropping system adopted, household demographics, socioeconomic factors, farm-specific factors, institutional factors, and geographical factors. The main respondent targeted for this study was the primary decision maker in vegetable production within the household [39], and data collection occurred in April 2022.

2.5. Theoretical Framework

Several theories can explain the factors that influence the adoption of agroecological cropping systems, such as the theory of planned behavior, the diffusion of innovations theory, and the random utility theory. However, since the theory of planned behavior focuses primarily on psychological factors, such as attitudes, subjective norms, and perceived behavioral control, its predictive power may be limited when considering broader economic, institutional, and geographical factors [47,48]. This limitation may reduce its usefulness in complex situations, such as those involving the adoption of multiple cropping systems by smallholder farmers. Similarly, the diffusion of innovations theory focuses on the adoption of individual innovations over time, rather than the simultaneous adoption of multiple innovations, such as cropping systems [49,50].

This study is therefore based on the random utility theory (RUT), which is grounded in the hypothesis that everyone is a rational decision maker, maximizing utility relative to their choices [51,52]. According to the RUT, when faced with numerous agroecological cropping systems to choose from, the decision-making process of each smallholder farmer will be informed by the likely utility that they will gain from adopting the cropping system [52]. The smallholder vegetable farmers’ decision to adopt agroecological cropping systems is influenced by a range of internal and external factors [53]. However, they have incomplete information on the various agroecological cropping systems available, which implies that uncertainty in the choice of agroecological cropping systems must be taken into consideration [52]. The RUT models decision making under uncertainty, allowing the simultaneous adoption of multiple cropping systems [54].

2.6. Model Specification

This study examines the factors that influence the adoption of agroecological cropping systems in vegetable production. The cropping systems analyzed include mixed cropping, strip cropping, row intercropping, border cropping, and crop rotation. A smallholder farmer was classified as an adopter if they had utilized a cropping system for at least one cropping season before the time of the interview [55]. Building on previous research [27,28,30,35,55,56], we acknowledge that smallholder farmers adopt these systems either as substitutes or complements to enhance their output and address production challenges, such as pests, diseases, and weeds. Ignoring the interrelationships between these cropping systems and analyzing these multivariate relationships as separate adoption equations using univariate techniques, such as probit or logit analysis for discrete choice dependent variables, would lead to biased estimates [28,57].

The multivariate probit (MVP) model accounts for the correlation between error terms by jointly modeling the effects of a set of explanatory variables on each of the agroecological cropping systems and estimating multiple binary probit models [57]. This regression model enables the identification of relationships between the adoption of different agroecological cropping systems, while also accounting for potential correlations between unobserved disturbances [34]. The MVP regression approach was therefore employed to jointly analyze the factors influencing the probability of adopting each agroecological cropping system. Following the random utility framework, the MVP can be represented through two systems of equations. The first system involves a set of equations with latent (unobservable) dependent variables, which are expressed as a linear function of a set of observed households ($h$) and plot ($p$) characteristics $X_{hp}$, along with multivariate normally distributed stochastic terms ${\epsilon }_{hp}$ [28,55]. Each equation in this system can be formulated as:

$${Y^{*}}_{hpj}=X_{hp}{\beta }_j+{\epsilon }_{hp}$$

where j = M, S, R, B, C, ${Y^{*}}_{hpj}$ represents the latent dependent variables, reflecting the level of expected utility derived from adopting mixed cropping (M), strip cropping (S), row intercropping (R), border cropping (B), and crop rotation (C), and $\beta$ denotes a vector of the parameters to be estimated. Household ($h)$ may opt to adopt a specific combination of cropping systems (j) if the expected utility from adoption exceeds that of non-adoption [28].

The second system, which outlines the observable dichotomous choice of households, is expressed as:

$$Y_{hpj}=\left\{\begin{array}{c}1if{Y^{*}}_{hpj}>0\\ 0ifotherwise\end{array}\right.$$

where $Y_{hpj}$ is the adoption of the jth agroecological cropping system by the hth household on plot p.

The error terms are assumed to jointly follow a multivariate normal distribution pattern with a zero conditional mean and a variance normalized to a unit where $\epsilon ~MVN\left(0,\Omega \right)$ [30,58]. The symmetric covariance matrix Ω is illustrated as follows:

$$\Omega =\left[\begin{array}{cccc}1& \rho MS& \dots & \rho MC\\ \rho SM& 1& \dots & \rho SC\\ ⋮& ⋮& \ddots & ⋮\\ \rho CM& \dots & \rho CB& 1\end{array}\right]$$

where $\rho$ is the pairwise correlation coefficient of the error terms between any two of the estimated adoption equations in the model [57]. The off-diagonal elements in the covariance matrix capture the unobserved correlation between the stochastic components of the different agroecological cropping systems adopted (e.g., $\rho MS$, $\rho CM$) [57].

The explanatory variables that are likely to affect the adoption of agroecological cropping systems in this study ($X_{hp}$) were derived from reviewed theoretical and empirical literature on the adoption of cropping systems [17,28,29,35,55]. These variables are presented in Table S2. The levels of significance of all the variables in this study were evaluated at p < 0.1, p < 0.05, and p < 0.01 to assess the robustness of the identified relationships.

3. Results and Discussion

3.1. Descriptive Statistics

A total of 525 households participated in this study, and the distribution of farmers surveyed in each region was as follows: 294 farmers (56%) in Arusha and 231 farmers (44%) in the Kilimanjaro region.

Table 1. Descriptive statistics of the dependent variables used in the econometric model.

Variables:	Description of Variables	Mean	Std. Dev.
Mixed cropping	Dummy = 1 if HH has adopted mixed cropping, 0 if otherwise	0.067	0.250
Strip cropping	Dummy = 1 if HH has adopted strip cropping, 0 if otherwise	0.036	0.187
Row intercrop	Dummy = 1 if HH has adopted row cropping, 0 if otherwise	0.029	0.167
Border cropping	Dummy = 1 if HH has adopted border cropping, 0 if otherwise	0.034	0.182
Crop rotation	Dummy = 1 if HH has adopted crop rotation, 0 if otherwise	0.608	0.489

presents summary statistics of the dependent variables used in the analysis of the adoption of agroecological cropping systems. Crop rotation (60.8%) was the most adopted cropping system, while the adoption rates for mixed cropping, strip cropping, border cropping, and row intercropping were 6.7%, 3.6%, 3.4%, and 2.9%, respectively. This shows that most of the farmers had adopted crop rotation, but the rate of adoption for all the other cropping systems was low. Similarly, [35] found that 66.7% of the farmers sampled from the southern highlands of Tanzania had adopted crop rotation. This indicates that smallholder vegetable farmers recognize and value the benefits of crop rotation, including improved soil fertility and pest and disease management [59]. The low adoption of the other agroecological cropping systems may indicate the existence of potential barriers to their adoption [60]. The observed disparity in adoption rates highlights the need for targeted agricultural extension services and educational programs to promote the benefits and feasibility of these less-adopted cropping systems.

Table 2. Descriptive statistics of the independent variables used in the econometric model.

	Combined (N = 525)		Adopters (N = 355)		Non-Adopters (N = 170)		p Value
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
(1) Socioeconomic Factors
Age	50.042	12.174	49.986	11.750	50.159	13.049	0.879 a
Education	8.023	3.367	8.085	3.330	7.894	3.449	0.545 a
Ownership:
Male	0.491	0.500	0.499	0.501	0.476	0.501
Female	0.152	0.360	0.132	0.339	0.194	0.397
Both	0.223	0.417	0.228	0.420	0.212	0.410
Rented	0.133	0.340	0.141	0.348	0.118	0.323	0.309 b
Management:
Male	0.381	0.486	0.375	0.485	0.394	0.490
Female	0.309	0.462	0.307	0.462	0.312	0.465
Both	0.310	0.463	0.318	0.466	0.294	0.457	0.844 b
Gross income	1,275,984.40	5,061,009.00	1,455,155.00	5,705,668.00	901,835.00	3,319,389.00	0.242 a
Nutrition security index	4.665	2.784	5.048	2.836	3.865	2.495	0.000 a ***
(2) Farm-Specific Factors
Farm Workers	10.779	17.741	11.606	18.907	9.053	14.922	0.123 a
Cost of insecticides	49,236.581	135,017.840	50,999.190	156,204.500	45,555.830	73,439.220	0.666 a
(3) Institutional Factors
Credit access	0.086	0.280	0.104	0.306	0.047	0.212	0.029 b **
Training	0.463	0.966	0.575	1.026	0.229	0.777	0.000 a ***
Market and information platform	0.251	0.434	0.296	0.457	0.159	0.367	0.001 b ***
Group membership	0.147	0.354	0.169	0.375	0.100	0.301	0.036 b **
(4) Geographical Factors
Distance to input market	88.983	90.398	92.899	94.239	80.806	81.457	0.152 a
Distance to output market	95.550	86.903	100.355	89.474	85.518	80.607	0.067 a
Region	0.440	0.497	0.417	0.494	0.488	0.501	0.123 b

Notes: N refers to number of observations; ^a 2-sample independent t-test; ^b Pearson chi-squared test; the asterisks *** and ** represent 1% and 5% significance levels, respectively.

presents the means of the explanatory variables for the adopters and non-adopters of the cropping systems, along with the mean differences, pooled means, standard deviations, and p-values from Pearson’s chi-square and t-tests. The proxy indicator for the nutrition security status of the households in this study is the household dietary diversity score (HDDS). The score is based on households’ consumption of 12 food groups over a 24 h recall period, with the target level of diversity determined as the average dietary diversity of the top 33% of households with the highest diversity [61]. It was found that nutrition-secure households consume more than seven food groups per day. Our results, however, indicate that, on average, the households consume approximately five food groups per day and are therefore food-insecure. This finding aligns with broader national assessments, which indicate that over 40% of households in Tanzania experience stressed or crisis-level food insecurity, limiting their access to diverse and nutritious diets [62]. Similarly, the study by [63] revealed that the majority of the households sampled in Lushoto, northern Tanzania, were food-insecure, with the dietary diversity score falling to as low as 1. On the other hand, the results indicate that households that adopted the cropping systems were significantly (p < 0.01) more nutrition-secure (5.048) than non-adopters (3.865). This implies that households with access to a more diverse and nutritious diet, and therefore better nutrition security, may be more willing to invest in cropping systems that promote long-term food quality, such as agroecological cropping systems [64,65].

As displayed in Table 2, the study found that only 8.6% of all the farmers had received production and marketing credit. This aligns with findings in Tanzania’s livestock sector where approximately 5% of smallholder farmers have access to credit facilities [66,67], suggesting that credit access is generally limited across both crop and livestock enterprises. The results also showed that adopters of the cropping systems had received significantly (p < 0.05) more production and marketing credit (10.4%) than non-adopters (4.7%). Credit facilities play a crucial role in the adoption of agroecological cropping systems by enabling smallholder farmers to hire labor, and limited access to these financial resources can impede their ability to invest in new or improved farming practices [65,68]. Furthermore, the study revealed that, on average, the smallholder farmers had participated in approximately 0.463 training sessions. This indicates either that many farmers did not attend any training sessions or that only a small portion of the farmers attended one or more sessions. However, the findings also showed that the farmers who adopted the cropping systems had attended significantly (p < 0.01) more training sessions (0.575) than the non-adopters (0.229). This implies that the lower participation in training by non-adopters could be a significant barrier to the adoption of agroecological cropping systems since farmers may lack awareness of the benefits, methods, and management techniques associated with these systems [34].

Approximately 25.1% of all the smallholder farmers utilized market and information platforms. These results are like those of [69], who found that 20% of the farmers surveyed in the Iringa region, Tanzania, used digital platforms on their mobile phones to access agricultural information. The findings also show that the usage of market and information platforms was significantly (p < 0.01) higher for adopters (29.6%) than non-adopters (15.9%). This indicates that these platforms allow smallholder farmers to access market information and therefore adopt better-paying cropping systems [70]. Our study also found that around 14.7% of the smallholder farmers were members of production and marketing groups, while adopters had significantly (p < 0.05) higher participation in these groups (16.9%) than non-adopters (10.0%). This finding underscores the importance of group membership in rural areas as an essential source of social capital, which can enhance the adoption of agricultural technologies like agroecological cropping systems [34].

This study also explored various other socioeconomic, farm-specific, and geographic factors hypothesized to influence the adoption of agroecological cropping systems. The socioeconomic variables included age, education, gender of the plot owner and manager, and gross income from vegetable production, as they are often linked to farmers’ capacity and decision-making processes regarding agroecological cropping systems [17,23,35,71,72,73]. Farm-specific factors, including the number of farm workers and the cost of insecticides, were analyzed to identify their potential impact on adoption, as they reflect labor availability and the expenses associated with crop management [30,74,75,76]. The geographic factors considered included the distance to input and output markets and the region where the farm was located since these factors may influence access to resources, market information, better output prices, and support networks for smallholder farmers [27,30]. Although these additional variables did not display any statistically significant differences between adopters and non-adopters, they provide a valuable context for understanding the diverse conditions in which smallholder farmers operate.

3.2. Complementarity and Tradeoff Among Agroecological Cropping Systems

The results showcasing the complementarity (positive correlation) and substitutability (negative correlation) of the agroecological cropping systems are provided in

Table 3. Complementarity and substitutability of agroecological cropping systems: correlation coefficient of error term matrix.

	Mixed Cropping	Strip Cropping	Row Intercropping	Border Cropping	Crop Rotation
Mixed cropping	1
Strip cropping	−0.298 (0.256)	1
Row intercropping	−0.439 *** (0.125)	0.359 * (0.195)	1
Border cropping	0.191 (0.192)	−0.395 *** (0.121)	−0.192 (0.175)	1
Crop rotation	−0.185 ** (0.090)	−0.057 (0.094)	0.211 * (0.113)	0.029 (0.099)	1

Notes: Robust standard errors are in parenthesis; the likelihood ratio test of regression interdependence χ²(10) = 19.719 **; the asterisks ***, **, and * represent 1%, 5%, and 10% significance levels, respectively.

[34]. The likelihood ratio test [χ²(10) = 19.719] was significant at p < 0.05, leading to the rejection of the null hypothesis of independence among the agroecological cropping systems [28]. This implies that using a univariate probit or logit regression estimation model would have yielded biased or inefficient estimates [27,77]. The findings indicate that both row intercropping and crop rotation are negatively correlated with mixed cropping. Similarly, border cropping was found to be negatively correlated with strip cropping. The observed tradeoff in the adoption of these cropping systems may be because incorporating multiple crops in various arrangements, such as row intercropping and mixed cropping, significantly increases the complexity of farm management, deterring farmers from adopting them simultaneously [78]. Conversely, the results revealed that row intercropping was positively correlated with strip cropping, and crop rotation was positively correlated with row intercropping. This could be due to the synergistic benefits that farmers acquire from adopting these cropping systems concurrently [79].

3.3. Factors Influencing the Adoption of Agroecological Cropping Systems

The maximum likelihood estimates of the MVP model of the adoption of agroecological cropping systems are presented in

Table 4. Adoption of agroecological cropping systems: multivariate probit (MVP) model results.

	Mixed Cropping		Strip Cropping		Row Intercropping		Border Cropping		Crop Rotation
	Coeff.	Std. Err.	Coeff.	Std. Err.	Coeff.	Std. Err.	Coeff.	Std. Err.	Coeff.	Std. Err.
Socioeconomic factors
Age	0.002 (0.000)	0.008	−0.024 ** (−0.001)	0.011	−0.004 (−0.000)	0.010	−0.012 (−0.001)	0.008	−0.001 (−0.000)	0.005
Education	−0.010 (−0.001)	0.033	−0.069 (−0.004)	0.060	−0.018 (−0.001)	0.031	−0.015 (−0.001)	0.027	0.004 (0.001)	0.017
Ownership:
Female	−0.345 (−0.036)	0.352	−0.597 * (−0.026)	0.351	−0.236 (−0.009)	0.357	−0.322 (−0.022)	0.434	−0.360 * (−0.117)	0.201
Both	0.099 (0.012)	0.230	−1.206 ** (−0.059)	0.476	−4.260 *** (-)	0.514	0.006 (−0.001)	0.274	−0.121 (−0.034)	0.163
Rented/Leased	0.379 (0.045)	0.262	−0.118 (−0.007)	0.410	−0.423 (−0.027)	0.466	−0.703 * (−0.043)	0.376	0.129 (0.034)	0.179
Management:
Female	0.052 (0.008)	0.277	0.672 * (0.032)	0.360	0.258 (0.009)	0.340	−0.330 (−0.020)	0.330	0.236 (0.076)	0.179
Both	0.237 (0.028)	0.222	0.908 *** (0.044)	0.295	0.490 * (0.029)	0.287	−0.209 (−0.013)	0.272	−0.000 (−0.006)	0.162
Gross Income (log)	−0.015 (−0.003)	0.061	0.288 ** (0.015)	0.120	−0.055 (−0.004)	0.077	0.140 ** (0.009)	0.071	0.100 ** (0.033)	0.043
Nutrition security	0.067 * (0.007)	0.034	−0.070 (−0.004)	0.050	−0.026 (−0.002)	0.054	−0.010 (−0.000)	0.039	0.101 *** (0.034)	0.024
Farm-specific factors
Farm workers	−0.005 (−0.001)	0.006	−0.035 ** (−0.002)	0.016	−0.052 *** (−0.004)	0.019	−0.038 *** (−0.002)	0.014	0.002 (0.001)	0.004
Cost of insecticides (log)	−0.081 *** (−0.009)	0.021	−0.076 ** (−0.004)	0.035	0.013 (0.000)	0.031	0.044 (0.003)	0.036	0.010 (0.003)	0.016
Institutional factors
Credit access	−0.320 (−0.041)	0.389	0.175 (0.008)	0.385	0.691 * (0.050)	0.362	0.406 (0.027)	0.370	0.376 (0.116)	0.260
Training	0.007 (0.001)	0.096	−0.268 (−0.013)	0.189	−0.009 (−0.000)	0.137	−0.193 (−0.012)	0.145	0.248 *** (0.083)	0.084
Market and information platform	−0.147 (−0.012)	0.227	1.194 *** (0.062)	0.285	0.617 *** (0.049)	0.226	−0.028 (−0.004)	0.257	0.102 (0.042)	0.152
Group membership	−0.030 (−0.004)	0.224	−0.129 (−0.005)	0.408	−0.656 (−0.041)	0.484	0.164 (0.005)	0.426	−0.049 (0.000)	0.192
Geographical factors
Distance to input market (log)	−0.055 (−0.006)	0.130	0.581 *** (0.027)	0.175	0.266 (0.013)	0.184	−0.167 (−0.009)	0.102	0.156 ** (0.057)	0.075
Distance to output market (log)	−0.040 (−0.006)	0.122	−0.380 *** (−0.020)	0.106	0.091 (0.008)	0.167	−0.037 (−0.002)	0.063	−0.008 (−0.005)	0.063
Region	0.213 (0.022)	0.190	−0.028 (−0.003)	0.323	−0.128 (−0.015)	0.202	−0.082 (−0.005)	0.233	−0.326 ** (−0.109)	0.132
Constant	−0.809	0.978	−4.284 ***	1.295	−2.238 **	1.064	−1.927 **	0.780	−2.146 ***	0.633

Notes: The asterisks ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; figures in parentheses are average marginal effects; number of observations = 525; log likelihood = −588.424; Wald χ²(90) = 526.450; prob > χ² = 0.0000.

. The Wald chi-square statistic for the overall significance of the model [χ²(90) = 526.450] was significant at p < 0.01, justifying the use of the MVP model for analysis [35]. The results show that the age of the household heads has a significant and negative effect on the adoption of strip cropping (p < 0.05), with each additional year reducing the probability of adopting strip cropping by 0.1%. This implies that older farmers may be less inclined to adopt strip cropping than younger farmers due to a decline in energy, which is essential for implementing energy-intensive agroecological cropping systems [17,23]. These results are consistent with those of [17], who reported that the age of the household heads had a statistically significant negative effect on the implementation of agroecological practices among smallholder farmers in Singida district, Tanzania. This finding underscores the importance of developing targeted interventions to address age-related barriers to the adoption of strip cropping.

Male-owned plots were used as the reference category when analyzing the gender of the owner. The findings indicated a significant and negative effect on the adoption of strip cropping and crop rotation (p < 0.1) for farms that were owned by female household members. Studies have shown that women landowners in rural Tanzania often face limited access to secure land tenure [80,81]. This can therefore constrain their ability to implement agroecological cropping systems such as crop rotation, which often involves longer-term investment in the land while yielding delayed returns [82,83]. Additionally, female landowners usually hold smaller, less ideal plots, particularly in rural areas where land is inherited or subdivided [81]. This makes it difficult to implement strip cropping, which usually requires larger, flatter plots to enhance crop interactions and maximize yield [84,85]. Compared with male-owned plots, plots that were jointly owned were significantly less likely to adopt strip cropping and row intercropping at p< 0.05 and p < 0.01, respectively. This could indicate that the complexity of decision making in joint ownership arrangements acts as a barrier to adopting certain cropping systems [86]. Smallholder farmers who rented or leased land were significantly less likely (p < 0.1) to adopt border cropping compared to those who owned land, particularly male owners, who were often the household heads. This implies that farmers who do not own land may be less inclined to adopt cropping systems such as border cropping, which often involves growing perennial crops and requires longer-term commitment and investment in the land [87,88]. These findings align with those of [17], who identified land ownership as a key factor motivating farmers in Tanzania to adopt agroecological cropping systems that yield benefits over the long term.

Our study also considered the gender of the plot manager with male-managed plots serving as the reference category. The findings indicated that female-managed plots were significantly (p < 0.1) more likely to adopt strip cropping than male-managed plots. This result contrasts with earlier findings on female landownership, suggesting that while female landowners are constrained in adopting certain cropping systems due to limited land tenure security and smaller plot sizes, female farm managers, who are responsible for day-to-day land use and improvement decisions, are less constrained by ownership issues [71,89]. Consequently, they may prioritize agroecological cropping systems like strip cropping, which enhance long-term sustainability and food security for the household [85,90]. These findings support [89] the assertion that rights to land ownership and management do not always overlap; therefore, the concepts of land ownership and management should not be used interchangeably. Plots jointly managed were significantly more likely to adopt strip cropping and row intercropping at p < 0.01 and p < 0.1, respectively, than those that were exclusively managed by male household members. This suggests that joint farm management fosters the collaborative decision making, resource sharing, and shared responsibilities that support the adoption of these cropping systems [91,92]. These results align with the findings of [73], who noted that in Uganda and Tanzania, intercropping was most frequently adopted on farms that were jointly managed. Similarly, [86] observed that in Uganda, the most prevalent decision-making pattern for adopting intercropping involved joint agreement by spouses.

Gross income from vegetable production was found to have a positive and significant effect (p < 0.05) on the adoption of strip cropping, border cropping, and crop rotation, suggesting that higher income allows farmers to invest in the inputs and technology required for the adoption of these cropping systems. This is consistent with the findings of [17], which showed that the income of smallholder farmers in Tanzania had a positive and significant impact on the implementation of agroecological practices. The study also revealed that the household food security status has a positive and significant effect on the adoption of mixed cropping (p < 0.1) and crop rotation (p < 0.01). This implies that smallholder vegetable farmers with higher dietary diversity scores are more likely to adopt cropping systems such as mixed cropping and crop rotation, probably because they experience greater stability and resilience in their food supply, enabling them to allocate resources toward adopting practices that enhance long-term sustainability [64,65]. However, these findings contrast with those of [30], who found that households in Nigeria facing lower levels of food security were more likely to adopt mixed cropping systems.

The study revealed a significant negative relationship between the number of people working on a farm during a growing season and the adoption of strip cropping (p < 0.05), row intercropping (p < 0.01), and border cropping (p < 0.01). This may be because these cropping systems require careful planning and coordination; consequently, farms with more workers may face challenges in managing their increased complexity, leading to lower adoption rates. This finding is similar to that of [74], who noted that in the Mbeya and Songwe regions of Tanzania, a marginal increase in the number of people working on a farm was associated with a decrease in the probability of adopting crop residue as a climate-smart agricultural practice. However, there is a slight discrepancy with our results, as [74] did not find a significant relationship between the number of farm workers and the adoption of intercropping or crop rotation.

The cost of insecticides was significantly negatively associated with the adoption of mixed cropping and strip cropping at p < 0.01 and p < 0.05, respectively. This indicates that higher costs of insecticides are linked with a lower likelihood of adopting these agroecological cropping systems. A plausible reason for this is that the smallholder farmers perceive these cropping systems as alternatives to conventional pest management strategies. When farmers therefore incur higher costs for insecticides, they might be less likely to adopt mixed cropping and strip cropping because they view these cropping systems as substitutes rather than complements of chemical pest control methods [93].

Access to credit was found to have a positive and significant (p < 0.1) effect on the adoption of row intercropping, highlighting the importance of financial resources in implementing this cropping system, which may involve higher management and equipment costs compared to monocropping systems [94,95]. These results confirm the results of [96], who observed that access to credit was positively related to the adoption of intercropping in Malawi, Tanzania, and Uganda. Additionally, the number of training sessions that each household had attended had a positive and significant (p < 0.01) impact on the adoption of crop rotation. Given that crop rotation was the most widely adopted cropping system under study, this finding underscores the effectiveness of training in encouraging the adoption of agroecological cropping systems. Similarly, [97] found in their study of farmers in eastern and southern Africa that attending training sessions positively influenced the likelihood of Tanzanian farmers adopting crop rotation.

Our study found that access to market and information platforms had a positive and significant (p < 0.01) impact on the adoption of strip cropping and row intercropping. These findings indicate that farmers who are better connected to markets and have access to relevant information on best practices are more likely to adopt these cropping systems. These findings align with the results of [98], who found that using radio and mobile SMS messages to provide smallholder farmers in Tanzania with information on practices such as intercropping and the use of organic fertilizer increased the adoption of these sustainable agricultural practices.

The distance from the farm to the input market (in walking minutes) positively and significantly impacted the adoption of strip cropping (p < 0.01) and crop rotation (p < 0.05). This suggests that farmers that are farther from input markets are more likely to adopt these cropping systems that reduce their dependence on external inputs, such as fertilizers and insecticides, which are often purchased from these markets [30]. On the other hand, the distance to the output market was found to have a significant negative effect (p < 0.01) on the adoption of strip cropping. This suggests that farmers that are farther from output markets are less likely to adopt strip cropping as it may require additional inputs and equipment that are more feasible with reliable market access [99]. The results indicate that farmers located in the Moshi district of the Kilimanjaro region had a 10.9% lower probability of adopting crop rotation than those located in the Meru district of the Arusha region. This disparity may be attributable to soil fertility conditions, as the volcanic soils in Moshi are highly fertile, potentially diminishing the perceived need for crop rotation to maintain soil health [100,101]. In contrast, the need to preserve soil fertility might be more acute in Meru, making crop rotation a more appealing practice there [59]. These findings underscore the importance of localized approaches to promoting the adoption of agroecological cropping systems.

4. Conclusions and Policy Implications

The adoption of agroecological cropping systems would lower production costs, enhance yields, and improve the livelihoods of smallholder farmers. Therefore, this study aimed to investigate the factors that influence the adoption of agroecological cropping systems for vegetable production by smallholder farmers in Tanzania. The correlation results revealed significant complementarities and tradeoffs among the agroecological cropping systems studied, demonstrating that analyzing them together reduces any potential biases that could have occurred if the cropping systems had been analyzed separately.

The multivariate probit results indicated that female plot ownership was linked to lower adoption rates of various cropping systems compared to male ownership, while female management was associated with higher adoption rates than male management. Therefore, policies and programs tailored to women, such as the creation of women-led farmer groups, could provide essential resources to female landowners and managers, helping to overcome the unique barriers they face. The positive impact of gross income and credit access on the adoption of agroecological cropping systems emphasizes the importance of financial resources. Governments and agricultural stakeholders should improve farmers’ access to finances and stable markets to boost their incomes and support the adoption of agroecological practices. Additionally, the positive influence of training and access to market and information platforms underscores the importance of capacity building and knowledge sharing in the adoption decisions of smallholder farmers. Policies and programs should be developed that increase the frequency and coverage of training programs and allow smallholder farmers to connect with larger markets and buyers of their produce. Geographic factors such as proximity to input and output markets and regional conditions also significantly influence the adoption of agroecological cropping systems. These findings highlight the importance of tailored, localized approaches in promoting agroecological practices.

While this study offers useful insights into the factors influencing the adoption of agroecological cropping systems, we recognize that there may be additional variables beyond the scope of this research that also play a role in adoption decisions. We recommend that future studies explore the impact of social and cultural norms, particularly those related to gender dynamics, on adoption. Furthermore, it is crucial to note that this study’s findings are restricted to the study area, and broader generalizations should be made cautiously, as adoption drivers are context-specific. We suggest that future research should explore the dynamics of the adoption of agroecological cropping systems across diverse regions and cultural settings. This broader exploration will lead to a deeper understanding of adoption patterns and support the development of targeted interventions and policies that promote the adoption of agroecological cropping systems.