Introduction

Limited understanding of the role of extension in rolling out innovations and technologies that have been tested and validated by research and extension services has been cited as a major constraint to adoption in the global south, particularly in Africa, since the 1980s (Odhong 2017; Pan et al. 2018). To avert this, participatory approaches including action research (PAR)—which hinges on the development of human capital, enhances questioning the knowledge, opens opportunities for novelty, affirms that experience can be a basis for knowing and that experimental learning can lead to a legitimate form of knowledge that influences practice—has been key mechanisms for community and rural development (Baum et al. 2006). PAR aims to captivate action through innovativeness, fostering (re)search and (un)learning cycle and perpetuating the quest for improvement. Over the past 3 decades, a ‘mother-baby’ PAR approach has been widely used to promote integrated soil fertility management (ISFM) components including improved varieties, inorganic fertilizer, organic manure, soil erosion control, and water conservation measures on smallholder farms in Africa and Asia (Rusike et al. 1998; Snapp et al. 2019). As a research tool, mother-baby PAR has been used in variety selection (Abebe 2005; Buah et al. 2013) and the technical evaluation of the performance of ISFM (Reddy et al. 2015). Researcher-farmer co-designed and farmer-managed ‘mother’ trials are established in a centralised, easily accessible location where the performance of ISFM technologies and agronomic practices are tested and demonstrated to the wider community. From the mother trial, farmers select a subset of technologies and agronomic practices and establish a ‘baby’ trial on their farms where they compare the ISFM performance against their local practices (Snapp 2002). The PAR effects are expected to reach the wider community through existing farmer-farmer networks. It is timely to explore if smallholder farmers engaged in PAR adopt ISFM from mother-baby trial plots and demonstration plots to practice on their farms.

This study is based on a PAR use case by the Africa RISING (Research in Sustainable Intensification for the Next Generation) project that has been testing and promoting ISFM components in east and southern Africa since 2013 using a mother-baby approach. Through the project, researchers, extension agents, and smallholder farmers formed a co-learning coalition. The researchers provided small packets of inorganic fertilizers and new crop varieties for demonstration. They integrated with complimentary locally available organic sources such as manure and soil and water conservation measures. After observation and learning on mother-baby PAR trials and out-scaling demonstration plots, farmers provided feedback through field days and subsequent training sessions. The Africa RISING project partnered with NAFAKA (meaning ‘grains’ in Swahili) and, as of 2018, ISFM and associated technologies along the value chain were out scaled to over 50,000 households covering 58,000 hectares of farmland in Tanzania (IITA 2018; USAID 2021).

Much as the learning effects of extension services have been widely studied (Emmanuel et al. 2020; Pan et al. 2018), few studies have been conducted to evaluate the influence of PAR approaches (Calkins and Thao 2004) especially concerning ISFM components. To date, mother-baby PAR evaluation has been in terms of design assessments of farmer evaluation and learning from trials (Snapp 2002; Snapp et al. 2019) and tests of statistical rigour for the incomplete block PAR trial designs (Gary et al. 2002). A recent PAR study shows a positive effect of participatory input demonstration on subsequent input purchase decisions (Sseguya et al. 2021). Within the PAR framework also, farmers’ uptake of tested ISFM components is assumed to be supported by advances in social capital including knowledge systems, gender roles, information access (Baum et al. 2006). As the public extension to farmer, ratio continues to shrink (1:4000 while recommended is 1:400), and with increasing attention to participatory approaches such as PAR, the farmer-to-farmer cooperative engagements through local networks are becoming key in knowledge dissemination (Qamar 2005). (Kangmennaang et al. 2017; Snapp et al. 2019; van Vugt et al. 2017) Moreover, within the community where PAR is implemented, some farmers are not directly engaged but learn through information sharing with peers and existing social capital (Njuki et al. 2008). In addition, in predominantly patrilineal systems, the contribution of both men and women to decision-making is deemed an essential gendered social capital that facilitates the effective contribution of the two genders (Gladwin et al. 2015; Njuki et al. 2008).

Soil fertility decline and nutrient mining are considered a major hindrance to agricultural development in Africa and to reverse decades of nutrient mining, increasing fertilizer input from 8 to 50 kg NPK ha−1 was set as a continental-wide policy goal in 2006 to be realised by 2015 (AU 2006). Although some countries made noteworthy progress, Tanzania’s fertilizer use has been critically low, averaging 3 to 9 kg NPK ha−1 between 1990 and 2009, against the recommendation of 100 kg N plus 40 kg P ha−1 required for maize production (IFDC 2012). Fertilizer adoption in northern Tanzania was less than 5% in 2009 (IFDC 2012) with some areas still as low as 3% in 2013 (Kihara et al. 2015). As such, increasing the uptake of fertilizers is important as a component of ISFM to address soil nutrient limitations (Sanginga and Woomer 2009). Farmers also rely on organic manure to improve soil fertility and crop production. In northern Tanzania, farmers apply on average, 3.5 t ha−1 farmyard manure (FYM) supplying 56.7 kg N, 17.5 kg P, and 45.5 kg K (Kihara et al. 2015). Still, few farmers apply or incorporate crop residues essential for the maintenance of soil organic matter and active soil carbon (Chamberlin et al. 2021; Massawe and Mruttu 2006). Interestingly, and as part of ISFM, farmers in northern Tanzania have widely adopted the agroecological suited improved crop varieties (Kihara et al. 2015). New varieties have been bred to enhance pest and disease resistance, climate smartness, and yield (Mabaya et al. 2019).

The aim of the study is therefore to explore factors influencing or constraining uptake of ISFM components including fertilizer, organic manure, residue recycling, and soil and water conservation (SWC) strategies by farmers who have been engaged or exposed to participatory action research. We hypothesize that farmers, with resources, information access, and in whose households, women are empowered and those who have experimented on trial plots have adopted ISFM components and are practicing on their other plots. They are either early adopters of new components such as fertilizers and new varieties or, with increased knowledge, are among the majority adopters of existing components such as manure, soil and water conservation measures, and residue incorporation.

Methodology

Study Site, Design and Sampling

We used engagement differentiated community strata to explore how PAR influenced uptake of ISFM practices from a mother-baby PAR use-case implemented by Africa RISING in Babati, Northern Tanzania. Farmers were engaged in an integrated inquiry, action, and reflection technology development process (Fig. 1). First, the agricultural survey for the 2007/08 season and the project baseline characterisation of 2012 identified productivity constraints including soil fertility limitations, low extension outreach and low technology uptake (Charles et al. 2016; Hillbur 2013; Kihara et al. 2015; URT 2012). Based on the baseline, community awareness and engagement campaigns were carried out and, in 2013, PAR mother trials were set up that have been hubs for researching several combinations of fertilizers, varieties, and other agronomic practices in a centrally and easily accessible location that was selected at village-level meetings.

Fig. 1
figure 1

Actor engagement and knowledge development phases for the Africa RISING project in Babati District, Northern Tanzania. NB: Green lower line represents farmer knowledge and existing platforms on which the PAR trials and demos (continuous black lines), the private–public partnership platform and the scaling program (dotted lines) are established

Full size image

The trials generated evidence of positive crop yield responses to fertilizer application with appropriate agronomic practices (Kihara et al. 2020). In 2015, the best-bet technologies were selected for field testing and validation in baby trials. Over the project timeline, 328 mother-baby trials were conducted across the study site, with a range of treatments added stepwise to meet the evolving research needs using the ISFM framework (Vanlauwe et al. 2010). Farmers were also exposed to the ISFM at different degrees including participation in technology demonstrations, farmer field days, implementation of ISFM baby trials, and farmer evaluation of technologies. The researchers trained extension agents on conducting participatory field days, who were then responsible for conducting them in their respective sub-areas and exposed the technologies to the wider community. Other activities implemented to promote ISFM included linking farmers to input suppliers (improved seeds) and output markets through collective action. In 2017, a scaling activity by Africa RISING -NAFAKA was rolled out to reach farmers beyond communities where validation was done. Outreach was supported by agro-advisory channels including the MWANGA mobile platform (a free farmer messaging service developed by the project), farmer to farmer knowledge transfer, and radio and TV programs.

In 2014, Africa RISING facilitated the establishment of Babati Integrated Agricultural Research for Development (abbreviated from Swahili as JUMBA) platform. The JUMBA platform is a public–private partnership (PPP) with farmer representatives, policy makers, government extension, seed and fertilizer companies, agro-dealers and researchers from international agricultural organisations, universities and national Selian Agricultural Research Institute (SARI) (Kihara et al. 2020; Ngulu et al. 2017). Farmers avail existing networks, local knowledge and land use plans on which PAR is developed while researchers provide technologies and approaches for addressing farming constraints and train extension agents. After training, the extension agents under Babati District Council become the key agents responsible for the training of hosting farmers and scaling. Seed and fertilizer companies (Meru Agro Seed, AMINATA Quality Seed, and Minjingu Mines & Fertilizer Ltd) supported trials and opened markets in the area for fertilizers and modern improved seeds bred for nutrition, tolerance to drought, and general climatic adaptability (Kihara et al. 2020; Kinyua et al. 2021). The lessons from PAR engagement by different actors are integrated and used for formulating strategic district agricultural development plans during JUMBA committee meetings attended by District Assembly and (Ngulu et al. 2017).

To explore behavioural impacts 9 years since PAR project initiation, a quasi-experimental design with counterfactual non-participating villages that were set up could not be used because the selection of participants was not strategically targeted. Moreover, the evolution of interventions over time, which is a key feature of PAR, makes the interpretation of results from trials more challenging (Leykum et al. 2009). The project was initiated with a biophysical characterization in 2013 to understand production constraints, on-farm nutrient response trials from 2014 to 2016, scaling up of good agricultural practices (GAP) in 2017/18, and crop spacing pattern improvements in on-farm trials from 2018. Matching the baseline of 2013 with the PAR participant list, we found that only 4% of farmers that were randomly sampled following the biophysical characterization (Kihara et al. 2015) participated in on-farm trials and out-scaling demonstrations.

Considering the design challenges, we adopted a cross-sectional survey drawing households using multi-stage purposive and stratified random sampling involving seven villages (Fig. 1). Purposely chosen, the villages include Long and Seloto where biophysical characterization was done in 2013, Gallapo, Riroda, and Sabilo where PAR was subsequently implemented plus Endanoga and Hallu which were reached during out-scaling activity by the AR-NAFAKA partnership. As indicated in Fig. 2, the villages are located along the elevation gradient from 1180 to 2170 m. Four of them are on the periphery of Babati District while others are close to Lake Balangida. The households were randomly sampled proportional to the village population from three strata: ‘fully engaged’ (2.4% of the population) who managed either mother and baby trials and received improved seeds and fertilizer plus intensive training; 10.1% ‘partially engaged’ PAR farmers who received improved inputs and brief training during out-scaling; and ‘non-participants’ within the PAR villages who were not directly involved (87.7%). Therefore, within the experimental and out-scaling villages, unbalanced stratified sampling was used to group farmers into these three groups.

Fig. 2
figure 2

Distribution of sampled farming households in Babati District, Northern Tanzania

Full size image

Considering that the study aims to evaluate multiple technologies with multiple covariates and outcomes, the sample size was adjusted for design effects (Rutterford et al. 2015) as:

$$n=\left[{z}^{2}* \frac{p*q}{{e}^{2}}\right]*d*r$$

where n = sample size, z = standard normal distribution linked to 95% confidence interval (1.96), p = expected prevalence (as a fraction of 1), q = 1-p (expected non prevalence), e = relative desired precision, d = design effect, r = non-response rate of 5%. From the previous baseline, Africa RISING and out-scaling outreach, and anecdotal expert observations, we assume that some ISFM and SWC measures are widely adopted (such as intercropping over 80%) while others are adopted by a few (such as inorganic fertilizers less than 20%). Therefore, we set the expected prevalence, p, to be an average of 50% which yields the largest sample size and is thus ideal to capture the adoption and impact of a wide range of technologies. The precision, e, is set at 5% considering that the expected usage of some technologies is as low as 20% or less. The design effect, d, account for heterogeneity between the strata arising from cluster sampling of baseline farmers and differences in usage of ISFM technologies among Africa RISING, out-scaling, and ‘other’ farmers. A sample size of 607 farmers was determined which represents 4.9% of the household population in the region and 9.9% of target villages.

Empirical Model Parameterization and Data Analysis

As farmers make decisions to increase and sustain food production, most operate at sub-optimal levels, due to several inefficiencies including cognitive abilities, time and labour constraints, and imperfect information. Thus, their decision is bounded rationally with a long-run goal of optimizing utility from their farmlands while their short-run objective is a pseudo-random search for alternatives (Wall 1989). With limited information, farmers examine alternatives one at a time and in some cases simultaneously. Moreover, the search for innovative technology is done only when information about better alternatives becomes available. First, we assume that farmers implement ISFM technologies independently and apply a binary logistic model for the dichotomous data. Unlike the linear probability model (LPM), the binary logit model is robust as the estimated probabilities fall within the [0, 1] range (Wooldridge 2012).

Then, adapting the ISFM framework by Vanlauwe et al (2010) to the adoption levels among farmers of Babati, from the most widely adopted to the least, we derived four-component combinations: Improved Maize Variety (IMV), IMV + Organic Matter inputs (manure and residues, OM), IMV + OM + Soil and Water Conservation, and IMV + OM + SWC + Inorganic Fertilizer. Whilst the ISFM component combinations would be treated as ordinal (Mponela et al. 2016; Otieno et al. 2021), the components were captured at the household level and hence do not represent combined usage and they could have been used in different plots. Rather, they indicate the diversity of technologies that a household uses and are unordered multinomial choices. Hence, we use the multinomial logit model (MNL) (Greene 2012).

The underlying random utility model and parameterisation of binary logistic and multinomial logit models are similar (Greene 2012). The main difference is that the binary logistic model compares the adopter vs non-adopter while for the multinomial logit model, the comparison is made among pairs of the polytomous component combinations. We make the IMV, which is widely used, a base against which component combinations are compared with. The conditional probability for using each ISFM component/combination is: \({p}_{i}=\frac{1}{1+ {e}^{-z}}\),

$$\begin{array}{cc}\mathrm{the}\;\mathrm{probability}\;\mathrm{for}\;\mathrm{not}\;\mathrm{adopting}\;\mathrm{is}:&(1-p_i)=1-\frac1{1+e^{-z}}\end{array}$$
$$\begin{array}{cc}\mathrm{and}\;\mathrm{odds}\;\mathrm{of}\;\mathrm{ISFM}\;\mathrm{use}\;\mathrm{against}\;\mathrm{not}\;\mathrm{adopting}\;\mathrm{is}:&\frac{p_i}{1-p_i}=\frac{1+e^{-z}}{1-e^z}=e^z,\end{array}$$

which is > 1 for the likelihood to use ISFM component/combination and < 1 for the likelihood of not to use.

Hence, the household’s decision to adopt ISFM component/combination (z) is empirically estimated as:

$$\begin{array}{cc}y_i^\ast=\beta_0+\beta_ix_i+\mu,&z_i=\left\{\begin{array}{c}\begin{array}{ccc}1&if&\;y\overset\ast i>0\end{array}\\\begin{array}{ccc}0\;&if&\;other\end{array}\end{array}\right.\end{array}$$

Where xi represents explanatory factors, βi are coefficients, and µ is the error term which is assumed to be normally distributed with zero mean–variance.

Explanatory Factors

The choice of explanatory variables was based on practical knowledge, theory, and statistics. The three key variables of interest that were directly influenced by the Africa RISING activity and could influence ISFM adoption include 1) PAR (training and inputs), 2) local networking through groups and peer to peer information sharing and 3) women empowerment. The other variables included information sources, extension network, age, education level of head and highest level amongst the household members, labour availability (man equivalents) and labour dependency ratio (number of dependents/number of workers), land size and plot fragments, and income from livestock and non-farm activities which have been found to influence farmers uptake of land management practices (Glover et al. 2019; Pattanayak et al. 2003). The best-fitting model for each of the ISFM components was parameterised using stepwise regression and models with the lowest Akaike Information Criterion were considered to capture sufficient variation (Vrieze 2012; Zellner 1999). In this way, we got different sets of variables that were associated with the adoption of each of the ISFM components.

Detailed data were collected on how farmers learnt about ISFM practices: inorganic fertilizers, organic manures, crop residue management, and soil and water conservation as well as on crop types, varieties, seed, crop marketing, and post-harvest handling. For each of these farming practices, 74% of farmers indicated that they acquired knowledge from extension agents including government extension (61%), Africa RISING (53%), other non-governmental organisations (24%), and agro-dealers (32%). Forty-four per cent of farmers acquired ISFM knowledge through social networks such as farmer cooperatives (8%), farmer groups, and relative/neighbour/friend (36%). Eleven per cent acquired knowledge through media such as radio/newspaper/TV (11%) and book-guide/leaflets/manual (2.4%) and 2.2% from agricultural curricula during formal education. In terms of diversity of sources as expressed by the total number explored, around 20% did not explore any of these 9 knowledge sources, 25% accessed two of the sources, 20% 3 sources, 15% 4 sources and the remaining 20% explored 5–8 sources.

The mean age of the household head was 48 ± 13 years. Although more than 80% of the household heads just completed primary education, a third of the households had members who completed secondary education, and 13% with college certificates. Farmers own and cultivate mostly below 5 ha (average 1.6 ± 2.0 ha) consisting of 2.5 fragmented plots. They own on average 5.2 ± 16 tropical livestock units and are reported to earn from non-farm activities and livestock sales 159 ± 283 and 106 ± 183 US$ per annum, respectively.

The women’s bargaining power was estimated based on the husband vs wife’s contribution to decision making: 1 if only the wife (adjusted by -0.25 if consult husband), 0 if the husband (adjusted by 0.25 if consult wife), and 0.5 if both. These were aggregated within five indicators: farm input (seed, fertilizer, manure), food security (buy food with farm proceeds), resources (livestock), market (sell produce), and income (employment, casual, business, remittances). In general, slightly above 50% of women contributed to joint decision-making while more than 10% make unilateral decisions over income and farm inputs. Still, in over 30% of married households, men make unilateral decisions in overall decision domains.

Resource-endowment cluster analysis was conducted from ownership of transport, farm, and communication implements. The correlations among input variables were generally low (< ± 0.4) except between cart and plough (0.658) and their sampling adequacy averaged 0.79. First, the principal component analysis with varimax rotation in SPSS was run and the component score coefficients for the 5 principal components that are representative of the original 11 variables, not linearly related and accounted for 69% variation were used in subsequent k-mean cluster analysis. Implement-based cluster analyses classified farmers into poor, moderate, and high resource endowed (Table 1).

Table 1 Farming household resource endowment clusters for Babati in Northern Tanzania
Full size table

Results and Discussion

ISFM Usage in Babati, Northern Tanzania

In the 2020–2021 season, farmers in Babati grew 2–4 crops including maize, beans, pigeon peas, sunflower, and others. On average 98% of farmers used a total of 17 improved maize varieties and almost all purchased certified seeds. Among these, a substantial proportion (70%) of farmers used old varieties (released before 2013), 12% mixed new and old while 18% planted new varieties. Seven ‘old’ varieties grown (arranged in descending (%)) include DK8031(23), SC627(20), SC403(16), Pannar691(12), Pioneer253(10), H614D(4), SC513(2), Mixed(0.8) and Situka(0.5). There has been a variety shift since 2013 (Kihara et al. 2015) with DK8031 increasing by 109% while SC627 dropped by 19% and H614D by 4.5%. A proportion of the market share (in %) was also taken-up by 10 newly released varieties including SC719(8.5), Meru513(7.7), DK777(5.3), Lubango(4.0), Meru515(3.8), SC419(2.7), Aminata(1.6), Meru623(0.8), Zamseed(0.64) and Sygenta634(0.16).

Wide usage of improved maize varieties has also been reported in an earlier study (Kihara et al. 2015) and is associated with the source of information, social capital, and farmers' confidence in government support (Beyene and Kassie 2015). For other crops, local varieties were mainly used (beans 97%, pigeon peas 85%, sunflower 96%). Using the modified ISFM framework, we found that of the 98% of farmers that planted Improved Maize Varieties in Babati, 56% combined IMV with Organic Matter inputs (manure and residues), 25% with OM + Soil and Water Conservation, and 9% IMV + OM + SWC + Inorganic Fertilizer.

On average, the study has found that 15% of farmers are purchasing and applying fertilizers, a large shift from the less than 3% observed in 2013 (Kihara et al. 2015). The proportion of farmers using fertilizers is 10% among non-participants, 16.5% among partially engaged, and 17.8% among those that received intensive training plus inputs, with each category applying an average of 23 ± 15, 27 ± 20, 40 ± 30 kg ha−1, respectively. These levels though indicating an increase are still lower than the targets set at the continental level (AU 2006) and much lower than the recommended input levels (IFDC 2012). Along the topological sequence, 9.8% of farmers in lower midlands, 9.9% in upper midlands, and 42.9% in highlands applied fertilizer in amounts averaging (± sd) of 41 ± 24, 35 ± 36, 23 ± 10 kg ha−1, respectively. Although the Africa RISING PAR activity occurred mostly in midlands and uplands, a higher proportion of farmers purchasing fertilizer in highlands is because of high crop yield potential with reduced production risk (i.e., due to high rainfall).

For five years (2016 – 2021), one-third of the farmers did not apply manure despite the still relatively low proportion of farmers using chemical fertilizers. The proportion of farmers applying the manure did not differ among the PAR engagement groups: 68% non-experimenting farmers, 69% fully engaged, and 67% partially engaged farmers, with application amounts ranging from 3.6 to 4.4, 3.6 t ha−1 year−1. The farmers not applying stand to lose since a yield increase of 0.8 t ha−1 has been associated with manure application (Kihara et al. 2015). There were notable interannual differences as 42% applied during the 2020–21 season, 61% in 19–20, 52% in 18–20, and 35% in 2016–2017. An economic study of manure usage in Morogoro revealed that farmers sometimes leave the manure to build up in the Kraal and apply it in subsequent years (Kangondo 2015). Farm-farm manure transfer is considerable with 18% of farmers getting manure from other households. Almost all those who applied used animal manure while a few, a small proportion applied compost.

A small proportion (16%) of the farmers incorporate residues: 28% among mother-baby farmers, 17% among out-scaling whilst 9% of non-participants. The incorporation of residues is important for the sustainability of farming systems through soil microbiological support, nutrient cycling/supply, improvements of soil physical properties including soil water retention (Mng’ong’o et al. 2021; Wasonga et al. 2010). However, most farmers (76%) take out and feed livestock away from the cropping field while a sizeable proportion that does not take out (30%) residues feed them to livestock on site (i.e., in the field). On average, 25% took out residues used as fuelwood, 9% for construction of fences, and 11% gave to other farmers, a small proportion used leftovers after feeding livestock for compost.

Sixty-two per cent (62%) of farmers established measures for controlling soil loss and conserving soil moisture. Terracing was used by 45%, contour ploughing 49%, earth bunds 41%, restricting animals in crop fields 35%, trees and bananas 32%, grass strips 12%, mulching 12%, furrow 12%, and floodwater harvesting 8%. For those that practised before and stopped, dis-adoption for terrace was 7%, grass strips 3.2%, contour ploughing 6%, trees & bananas 4.4% and earth bunds 6%. The main reasons for dis-adopting SWC strategies included labour, land, and occurrence of pests and diseases, key triggers of SWC dis-adoption even when environmental awareness is raised (Bravo-ureta et al. 2006).

ISFM Adoption

The binary logistic and multivariate logit results (Tables 2, 3 and 4) show that increased human capital in terms of learning through technical training and input demonstration significantly influenced fertilizer use, residue incorporation, and soil and water conservation. The odds for farmers to purchase and apply fertilizer was 1.84 folds in favour of PAR participants who received inputs and agronomic advice and 3.51 for farmers who received additional expert training in the ‘Mother baby’ trials. Training plus demonstration had a considerable influence on the decision to adopt the 4-component combination of improved maize varieties plus OM, SWC, and fertilizer with 3.44 odds compared to those that used IMV only (Table 3). These findings show that PAR significantly influenced fertilizer choice: independently or in combination with other components and confirm the significance of PAR in promoting fertilizer use (Kpaka et al. 2021). These results further support the findings by (Sseguya et al. 2021) that technology demonstration has a positive influence on farmers’ farm input expenditures. Our findings reveal a much high propensity when demonstrations are supplemented by agronomic training. Specifically, PAR spurred fertilizer application especially among medium and high resource farmers by 3.3 and 6.6 odds respectively (Table 4).

Table 2 Determinants of adoption of fertilizer, residue, manure, soil and water conservation, and new improved maize varieties in Northern Tanzania (reporting odds from the binary logistic model)
Full size table
Table 3 Determinants of progressing on the ISFM ladder in Northern Tanzania (reporting odds from MNL)
Full size table
Table 4 Resource-cluster odds for adoption of fertilizer, residue, manure, soil and water conservation, and improved maize varieties use in Northern Tanzania (reporting odds from the binary logistic model)
Full size table

In most parts of Africa with acidic soils, the use of Urea and ammonium sulphate induced further acidification (Graham and Haynes 2005), lowering productivity (Lungu and Dynoodt 2008) and farmers developed a belief that chemical fertilizers “kill the soil” (Africa RISING 2018). Our results reaffirm PAR as an effective strategy to demystify and challenge existing knowledge and open opportunity for innovation. In Tanzania, PAR trials of diammonium phosphate (DAP) fertilizer and rock phosphates highlighted that farm productivity can significantly be enhanced with a yield advantage of over the 0.7 t ha−1 against the yield obtained under farmer practice of 3.1 t ha−1 (Kihara et al. 2020). Further evidence from long term experimentation in Brazil, South Africa, Zambia, and Rothamsted establishes that acidification becomes detrimental if excessive amounts of ammonium fertilizers of above 80 kg ha−1 are used (Fageria et al. 2010; Graham and Haynes 2005; Lungu and Dynoodt 2008), even when treated with liming (Macdonald et al. 2020). Although not the focus of the current study, the input demonstration offered in PAR was based on research-informed input levels of 50 kg N ha−1 and 20 kg P ha−1 that ensure profitability and can be sustainable (Africa RISING 2018; Kihara et al. 2020).

Soil and water conservation practices were enhanced by the provision of inputs plus training by 1.64 odds, especially among the high resource farmers. However, knowledge acquired through training and input demonstration was associated with a reduction in the likelihood to incorporate residues among resource-poor farmers by 91% (Table 4). These findings are indicative that for PAR to contribute to closing the nutrient gap and increase farm productivity, the program should consider both short-term productivity gains and long-term sustainability outcomes. The PAR’s negative influence on organic inputs such as residue incorporation has sustainability and profitability implications. The soils in the region have average organic carbon of 1.78 ± 0.75% (Kihara et al. 2015), which is low for soil structural stability of 2.0% but sufficient for crop fertilization response (Tamene et al. 2019). As established by Chamberlin et al (2021), crop response from the applied inorganic fertilizer is increased in presence of active carbon mostly from freshly applied organic matter, which may be affected if crop residues are not incorporated.

The bonding social capital through local networks and the bridging social capital in terms of the variety of sources of information and intensity of extension contacts are key for farmers to embrace fertilizer input, residue incorporation, and SWC (Table 3). Being informed through local networks such as cooperatives farmer groups, relatives and neighbours are associated with an increase in the odds of residue incorporation and soil and water conservation by more than 100%. With the decentralised planning through JUMBA platform, community engagement in research and technology scaling through the existing NAFAKA network supported by Mwanga mobile agro-advisory messaging, the community groups became central to agricultural information exchange and these findings on the bonding social capital effects in this study support the notion that cooperative membership increases the chances for farmers to learn and take up ISFM (Njuki et al. 2008; Otieno et al. 2021). Due to increased interaction, trust supported by evidence from research, peer-to-peer learning through social networks fosters technology adoption (Beyene and Kassie 2015) and has been found to complement extension services (Wang et al. 2020). In few cases, however, groups tend to be for programs that reduce ISFM adoption, e.g. in southern Africa (Mponela et al. 2016). Access to a variety of information sources and having more extension contacts supports the household’s decision to adopt fertilizer and residue incorporation by 80% and 75%, respectively.

The resource-cluster specific analysis (Table 4) revealed that local networks support SWC, especially among medium resource farmers but significantly reduce the odds to try new varieties among resource-poor farmers. However, having access to diverse information sources supports testing of new varieties among these resource-poor farmers and supports fertilizer application decisions especially among medium and high resource farmers. An earlier study found that diversity in information sources accelerates farmers’ adoption of improved varieties in Tanzania (Beyene and Kassie 2015).

The increased gendered social capital as indicated by women’s empowerment in decision-making has a multi-dimensional influence on ISFM adoption. The effects also vary within the gender domains. A point increase in women’s bargaining power on purchase of farm inputs increases the odds in favour of fertilization and manuring by 2.6 and 13.5 odds, respectively, and for high resource farms to implement SWC measures by 2 odds. The MNL further establishes that if women are empowered to decide on farm inputs, the propensity to complement IVM with the OM, SWC, and fertilizer increases. Similar findings of gendered social capital on fertilisation were made by Njuki et al. (2008), who found that when husbands value and consider the contribution of women to decision making, the household’s likelihood to purchase and use inorganic fertilizer increases. The positive influence on manuring however is in sharp contrast to earlier studies that found women in disfavour of manure because its labour requirements are not in sync with household chores (Mponela et al. 2020).

Our results show that the increase in women’s input bargaining power is associated with a 65% decrease in the propensity for a household to incorporate residues, especially among medium resource farmers. Increased women’s bargaining power on livestock resources increases the chances to incorporate residues by 7.8 odds, especially among medium resource farmers but reduces the odds for medium resource farmers to apply manure by 98%. In the area, residues are managed and transported to households using carts to feed livestock and spread on the farm after harvest mostly by men to retain productivity. Considering that the task is labour intensive and requires extra effort after harvest, households that make joint decisions, with increased women's bargaining power, are more likely to decide on a time to incorporate before free-range animals feed on residues left on the farm (Acosta et al. 2020).

If women have increased contribution to income usage decisions, there is an associated reduction in the likelihood for farmers to early adopt new varieties and implement SWC measures. We see similar effects for polytomous decisions. Income empowerment reduces the odds for farmers to complement IMV with the OM, SWC, and fertilizer. Similar observations were made by Otieno et al. (2021) in central Kenya that if women are in decision-making positions in houses disfavour adoption of animal manure + anti-erosion + inorganic fertilizer + crop rotation + legume intercrop as ISFM technological set. In most patrilineal systems, monotonous decisions did not give room to farm system change. Gender constancy is challenged, and women’s voices are increasingly being recognised. As women become empowered, they influence the decisions to suit their roles, responsibilities, and aspirations leading to system transformation.

Farmer and farm attributes enabling ISFM use include labour availability (enabling new variety testing), higher education level of household members (enhancing decision to apply manure as well as to combine IMV plus OM and SWC, plus IF), age of head (enabling SWC), farm size (supporting residue incorporation and SWC), having plot fragments in various locations (open opportunity for fertilizer application, manure application, and SWC). In Kenya, Otieno et al (2021) found a contrasting effect of age, as household decision-makers age, they become more risk-averse to combining animal manure, SWC, inorganic fertilizer, crop rotation, and legume intercrop. As expected, higher labour dependency reduces the odds for households to put in place measures for SWC by around 20% and quite counter-intuitive, labour availability also reduces the propensity to implement SWC. Dependency further reduces the odds for farmers to combine IMV with organic inputs and SWC as well as the addition of fertilizers to the ISFM basket. A similar study in southern Africa found that labour constraints affect the adoption of inorganic fertilizer, FYM and residues as a component combination (Mponela et al. 2016). The increasing number of children and old household members that the adult worker must look after has been found to put enormous pressure on them, affecting the adoption of labour demanding technologies such as SWC in east and southern Africa (Mponela et al. 2020). Apart from addressing gender and youth inclusivity, the ISFM technologies must reduce drudgery as youth dependency ratios are still high in sub-Saharan Africa (UN 2020).

Income from livestock sales supports farmers to apply fertilizers, manure and establishing SWC measures but has an inverse association with residue incorporation. The component combination results, also reveal that livestock income positively influences farmers’ decision to combine IMV plus OM and SWC. In contrast, income from non-farm operations supports residue incorporation but reduces the adoption of SWC. Contrasting effects of non-farm income have been reported. In southern Africa, non-farm income supported component combination whilst in Kenya affected ISFM adoption (Mponela et al. 2016; Otieno et al. 2021). Household diversifying into non-farm and integrating with livestock as major livelihood strategies forming a multi-enterprise entity. For the financial requiring component such as fertilizer, the leveraged income from one of the livestock livelihood portfolios and address constraints in the cropping portfolio by investing in fertilizer (Okoboi and Barungi 2012), which is essential to replenish the extracted nutrients and boost crop productivity. The inverse relationship between livestock income and residue incorporation is indicative that farmers who generate higher financial benefits from livestock have a corresponding higher tendency to utilise the residues as fodder. They instead invest in SWC while their counterparts, those that generate less income from livestock, tend to incorporate the residues and have less likelihood to implement SWC measures.

There are significant altitudinal differences that reflect the potential for ISFM technology use regarding the topo-sequence. Farmers on higher elevations have a higher likelihood to apply fertilizer and implement SWC measures while those on lower elevations have a significantly higher propensity to incorporate residues and test new varieties. On component combination, we see those farmers at higher altitudes are less likely to add OM to IVM than those at lower altitudes but have a higher likelihood to adopt the 4-component combination of IMV plus OM, SWC, and fertilizer. The higher elevation site (at 2150–2450 masl) receives 1200 mm rainfall/annum and it is highly productive. As a result, there is less variability and a higher probability of recovering fertilizer expenditure, supported by SWC to avoid soil erosion and nutrient drain. On the other hand, those on lower elevation sites (1200–1500 masl) which receive 750–900 mm annual rainfall face drought risks and adapt by incorporating residues and testing new drought-resistant varieties.

Our study further contributes to the literature on economies of scale and risk-averse investment behaviours. Although some studies found an inverse productivity and farm size relationship (Matchaya 2007) which could explain the diversified use of ISFM technologies on small plots (Mponela et al. 2016), several studies revealed that relatively larger farm sizes provide space for farmers to experiment and adopt multiple technologies (Croppenstedt et al. 2003; Dorward 1999; Just and Zilberman 1983; Mukasa 2018). Our results show that parcel diversity increases the likelihood for farmers to adopt ISFM components. This corresponds with Mponela et al. (2016) who also found that fragmentation increases the chances for farmers to adopt ISFM. In their review, Just and Zilberman (1983) found that, in developing countries, high inter-farm diversity resulting from fragmentation induces farmers to explore and implement different strategies to address multiple farm constraints. Moreover, these farmers are less risk-averse and spread investments in farms thereby reducing the risk of total crop investment failure. Fragmented plots also accord farmers room for experimentation and implementation of several technologies on some of the portions.

Conclusions

The study has established that farmers’ usage of ISFM components is influenced not only by human capital from training and demonstration but also by gender social capital in terms of women empowerment, bonding social capital through local networks, and labour. With the increasing shortage of extension agents, strategies that transform knowledge systems and foster innovative minds among rural communities are a feasible pathway to rural development. We find that the mother-baby PAR approach had priming effects resulting in improvements in agricultural practices among smallholders of the highlands of Northern Tanzania. Training plus demonstration had a considerable influence on the decision to adopt the 4-component combination of improved maize varieties plus OM, SWC, and fertilizer compared to those that used IMV only. Beyond the priming effect, PAR as a social learning innovation that is still in the testing and development phase faces continuity problems associated with the project cycle. However, PAR is not implemented using the linear technology development and dissemination pathway of ‘researcher-extension-farmer’, it brings actors together in an integrated social learning approach. This strategy is an efficient approach in decentralised systems for consolidating joint efforts amongst active stakeholders in response to the contextual needs of the farming community but faces scaling challenges if the lessons are to be used in different environments.

Moreover, PAR engagement is information and resource-intensive, hence, the need to balance multi-actor needs and knowledge. The study has established that the ISFM use depends on the complementary role of the local network. As farmers acquire technical knowledge, the community increasingly rely on local knowledge systems. Integration of solutions and matching with needs was achieved by the public–private partnerships that have been platforms for resource mobilisation and effective engagement and the scaling strategies for reaching farmers outside demo villages. Within these family farm communities, the role of gender gendered social capital i.e., women’s empowerment in decision-making, influences the adoption of fertilization, manuring and SWC measures. The study further establishes that if women are empowered to decide on farm inputs, the propensity to complement IVM with the OM, SWC, and fertilizer increases. In the resource-constrained, with low extension and technical support, and men-dominated patrilineal farming systems of Africa, the study places PAR, social networks, and gender inclusivity as key approaches for improving smallholder’ ISFM.