Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Farmer responses to an input subsidy and co-learning program: intensification, extensification, specialization, and diversification?

Farmer responses to an input subsidy and co-learning program: intensification, extensification,... Sustainable intensification aims to increase production and improve livelihoods of smallholder farmers in sub-Saharan Africa. Many farmers, however, are caught in a vicious cycle of low productivity and lack of incentives to invest in agricultural inputs. Moving towards sustainable intensification therefore requires support such as input subsidies and learning about new options through, for instance, co-learning approaches. Yet such support is not straightforward as agricultural developments often diverge from the envisaged pathways: extensification may occur instead of intensification and specialization instead of diversification. Understanding of farmers’ responses to incentives such as input subsidies and new knowledge is lacking. Our overarching aim was to improve this understanding, in order to better support future pathways for agricultural develop- ment in smallholder farming. Over five seasons, we compared the responses of farmers in western Kenya taking part in a novel co-learning program we developed, which included provision of an input voucher, with the responses of farmers who only received a voucher. We also assessed the differences before and during the program. We used diverse indicators that were related to the different agricultural development pathways. Farmer responses were mainly a result of the input voucher. Farmers increased maize yields (intensification) and maize area (specialization) for maize self-sufficiency. Increased farm and maize areas in combination with relatively low N application rates also pointed to extensification coupled with the risk of soil N mining. Diversification by increasing the soybean and groundnut area share was facilitated by the integrated co-learning approach, which thereby supported relatively complex farm management changes. Our results highlight the dif- ficulty of enabling yield and production increases, while also meeting environmental and economic goals. The diversity of farmer responses and constraints beyond the farm level underlined the importance of wider socio-economic developments in addition to support of sustainable intensification at farm level. Keywords Sustainable intensification · Smallholder farmers · Sub-Saharan Africa · Multi-criteria assessment · Indicators · Pathways · Yield · N use efficiency · Living income · Subsidies 1 Introduction and a changing climate require considerable changes in current farming systems (Giller 2020). Sustainable Livelihoods of smallholder farmers in sub-Saharan Africa intensification of farming is seen as a key strategy to (SSA) are under pressure. Many are caught in a poverty enhance rural livelihoods in SSA (Vanlauwe et al. 2014; trap, a vicious cycle of low productivity and lack of Jayne and Sanchez 2021). Sustainable intensification opportunities and incentives to invest in agricultural inputs aims to enhance production per unit land, nutrient, and (Tittonell and Giller 2013; Koning 2017). Additionally, labor input, while reducing environmental damage, constraints such as small farm sizes, limited market access, building resilience, and natural capital, as well as securing environmental services (e.g., Pretty et  al. 2011; The Montpellier Panel 2013). Struik and Kuyper (2017) argue * Katrien Descheemaeker that the concept of sustainable intensification can be used katrien.descheemaeker@wur.nl as a “process of inquiry and analysis” and discuss how Plant Production Systems, Wageningen University, P.O. the social and economic dimensions of sustainability can Box 430, 6700, AK, Wageningen, The Netherlands be included. Such a broad view enables identification of Central Africa Hub Office, IITA, P.O. Box 30722-00100, trade-offs that arise when agricultural systems intensify. Nairobi, Kenya Vol.:(0123456789) 1 3 40 Page 2 of 19 W. Marinus et al. Using a diverse set of indicators to describe these trade- level cereal self-sufficiency is an important indicator that fits offs, can inform decision-making by society and policy with farmers’ objectives. makers (Struik et al. 2014; Struik and Kuyper 2017). Yield-increasing inputs required for sustainable intensi- However, increasing yields through sustainable intensi- fication are beyond the reach of most smallholder farmers fication is challenging in SSA (Schut and Giller 2020) and (Vanlauwe et al. 2010) and need incentives such as input alternative pathways are often more apparent. For instance, subsidies. In the past 15 years, several fertilizer and seed extensification is currently more common than intensifica- subsidy programs were (re-)initiated by African govern- tion in many regions of SSA (Baudron et al. 2012; Ollen- ments (Jayne and Rashid 2013; Jayne et al. 2018), after their burger et al. 2016). Continued extensification is associated virtual absence during the 1990s and early 2000s (Martin with soil nutrient mining, and this trend could be reversed and Anderson 2008). In addition, social enterprises, such by strongly increasing nutrient inputs (Giller et al. 2021). as One Acre Fund (www. oneac refund. org), provide inputs However, this is constrained by widespread poverty traps though credit schemes to smallholder farmers. Increased (Tittonell and Giller 2013; Koning 2017) and the relatively input use, however, also requires new knowledge (Jayne low economic benefits of staple crop intensification in prac- et al. 2019; Jayne and Sanchez 2021). In a large-scale sub- tice (Bonilla-Cedrez et  al. 2021). Indeed, current trends sidy scheme in Malawi, the limited extension provided by show an increase in the area under maize cultivation in SSA the government was seen as a possible cause for N use effi- (van Loon et al. 2019; Santpoort 2020), which historically ciencies to remain low (Dorward et al. 2008). In addition, has been linked to an increasing population, increasing food fertilizers can be scarce and farmers may mistrust their requirements and urbanization (Smale and Jayne 2003), and quality (Michelson et al. 2021). Co-learning, an iterative hence increasing land pressure (Crowley and Carter 2000). learning framework involving farmers and researchers or Although specialization towards maize favors the production extension workers, has proven to be successful in develop- of sufficient energy, diversified cropping systems would be ing contextualized knowledge (Descheemaeker et al. 2019). more sustainable in terms of income, nutrition, crop yields, We developed an integrated co-learning approach (Mari- and risk spreading (Vanlauwe et al. 2019). Hence, identifica- nus et al. 2021), which aimed to sustainably increase farm tion of constraints and opportunities is essential to support level production by fostering increased input use through desired pathways such as diversification and intensification. the provision of a voucher, in combination with knowledge Setting sustainable intensification as an overall goal for co-creation (Fig. 1). In this paper, we apply a multi-criteria smallholder farming systems results in multiple subsidiary assessment over five seasons to analyze the outcomes of goals, e.g., increased yields, desired N use efficiencies, and food self-sufficiency at household and national level. Attain- ing all goals simultaneously is virtually impossible as trade- offs exist (Klapwijk et al. 2014; Vanlauwe and Dobermann 2020). Moreover, farmers follow their own objectives and prioritize some goals over others. Some goals also require time before they can be attained (Vanlauwe et al. 2010) and outcomes may differ between seasons, requiring assess- ment over multiple seasons, which is rarely done (Smith et  al. 2017). Measuring progress towards the multiple goals of sustainable intensification requires a multi-criteria assessment of indicators associated with the principles of sustainability. Using a framework of principles and criteria warrants transparency and a justified selection of indicators (Florin et al. 2012). According to Florin et al. (2012, p.109), “Principles are the overarching (‘universal’) attributes of a system. Criteria are the rules that govern judgement on outcomes from the system and indicators are variables that assess or measure compliance with criteria.” Criteria can Fig. 1 A farmer who took part in the integrated co-learning approach also help to decide upon benchmarks to judge whether a goal explains how she has used a new type of maize spacing to ensure increased light availability for the intercropped groundnut. Maize is reached (Schut et al. 2014). Within sustainable intensic fi a - grew more vigorously due to increased fertilizer use as part of her tion of smallholder farming systems, criteria, indicators, and intensification strategy. Moreover, by learning about maize-legume benchmarks need to address the field, farm, and household spacing options and new groundnut varieties she was able to increase level. At national level, increasing yields to a certain thresh- the area of groundnut on her farm and thereby to also diversify her cropping system. Photographed by Wytze Marinus. old is required to attain food self-sufficiency, while at farm 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 3 of 19 40 −2 a co-learning program in relation to different agricultural Busia is less densely populated with 530 people km , and development pathways. somewhat larger farms of about 1.0 ha (Jaetzold et al. 2005; Our overarching aim was to improve the understanding KNBS 2019). Both locations receive a rainfall of 1800–2000 −1 of farmer responses to input subsidies and new knowledge, mm year and a have a bi-modal rainfall pattern (Jaetzold in order to better support desired agricultural development et al. 2005), with the long-rain (LR) cropping season from pathways in smallholder farming. This materialized in the March until June and the short-rain (SR) cropping season following objectives to (1) assess the effect of co-learning from September until November. Activities started in the supported by a voucher for inputs on farmers’ decisions and SR season of 2016 and continued for five seasons until the management outcomes, by comparing it with a voucher-only SR season of 2018. Vihiga was selected as a location for its approach; (2) analyze the above effects in terms of crite- high population density, which commonly occurs in high- ria and indicators that relate to agricultural development lands areas of East Africa. Busia was selected for its com- pathways; and (3) reflect on the pathways of intensification, parably larger farm sizes than Vihiga, which could lead to extensification, specialization, and/or diversification result- more opportunities for increasing household income from ing from the co-learning and voucher program. farming. In each county, Vihiga and Busia, two sub-locations were selected and in each of these locations 11–12 farmers were 2 Methodology chosen. Farmers in one sub-location formed the co-learning group while a comparison group was formed in the other 2.1 T he integrated co‑learning approach sub-location. The sub-locations were selected to have similar farming systems, yet be sufficiently far apart to avoid spillo- We applied an integrated co-learning approach from August ver effects. All farmers in the co-learning group received a 2016 until July 2018, as described in detail by Marinus voucher and took part in the co-learning activities. Those et al. (2021). The approach combined four complementary in the comparison group received only the input voucher. elements: input vouchers, an iterative learning process, When inputs were added to the voucher based on feedback common grounds for communication, and complementary from the co-learning groups, these were added for the com- knowledge. An input voucher of US$ 100 was provided each parison group as well. A mid-season field monitoring sur - season to 47 farming households which aimed to alleviate vey included a visit by researchers to each field including resource constraints and increase input use. Inputs for maize, fields that were newly added during the program, to record groundnut, soybean, common bean, and sorghum produc- the crops cultivated and the percentage intercropping. The tion and for dairy were made available. Most inputs were farmer was asked about input use, planting dates, and other offered from the first season on, while groundnut and (short crop management practices. Field sizes were measured using duration) common bean seed and Imazapyr-treated maize a hand-held GPS before the start of the program in June seed against striga were added later during the program in 2016. Small fields with sides less than 20 m were measured response to feedback from the co-learning farmers. The feed- by hand. Yield measurements were done in two 4 × 4 m back was central to an iterative learning process in which a (16 m ) quadrats in all fields containing maize, groundnut, co-learning workshop prior to each cropping season played soybean, and/or common bean. These crops together made a pivotal role. The focus of the workshops evolved over time up about 60–70% of the total cultivated area per farm. Fresh based both on questions and on feedback from farmers dur- cob (maize) and pod (legumes) yields were measured in the ing the season as well as topics identified by the research - field, with one sub-sample per quadrat was taken to deter - ers. Discussion topics during the workshops included the mine dry weight by oven drying. Dry weights were calcu- judicious use of mineral fertilizers and the cultivation of lated back to a standardized moisture content of 14%, and −1 alternative crops such as legumes. Researchers monitored the grain yield (kg ha ) per field was calculated based on the farmers’ responses through a mid-season field survey, the average of the two quadrats. The detailed monitoring yield data collection, and an individual evaluation interview and measurement campaign during five seasons ensured a at the end of each season (see Marinus et al. 2021 for further comprehensive assessment of changes in farm management details). over time. However, the limited number of farmers per sub- location precluded a formal statistical analysis. Addition- 2.2 Research setup ally, we compared the situation during the program with a baseline study from the two seasons before the program. The The integrated co-learning approach was applied in two baseline study was held in the dry season, June 2016, before locations, Vihiga and Busia County in western Kenya. Vih- the start of the program. It used the detailed farm characteri- iga is one of the most densely populated rural areas in SSA zation survey methodology (Giller et al. 2011) to ask many −2 with 1050 people km , with small farm sizes of <0.5 ha. questions relating to the household characteristics and the 1 3 40 Page 4 of 19 W. Marinus et al. production system, including estimates of crop yields and We present the indicator values at the start and the end input use in the previous two seasons. Field sizes for all of the program in a spider web diagram, to assess possible fields in the farm were measured and farmer reported data pathways related to agricultural development. Indicators that was used to derive crop production and input use. During the were identified for intensification and extensification and program, however, crop yields were measured, and farmer- for diversification and specialization, indicated with a * in reported input use was triangulated by comparing field and Table 1, were included in the spider web diagram. In Sec- farm level application. Hence, the accuracy of the baseline tions 2.3.1–2.3.4, we describe the link for each of the indi- study and the detailed monitoring during the program differs cators with their respective pathway. Those indicators were and this needs to be considered in the comparison. scaled using a 0 to 10 score based on specific benchmarks (described in Sections 2.3.1–2.3.4), with a larger score indi- cating a more sustainable situation. Linear interpolation was 2.3 T he indicator framework: principles, criteria, applied to the indicator values to score them between 0 and and indicators 10. We used a multi-criteria assessment to analyze farmers’ 2.3.1 Productivity decisions and management outcomes of the integrated co- −1 learning program. Indicators were selected using principles Reducing yield gaps Maize grain yield (kg ha ) was and criteria (Table 1). We identified four principles of sus- measured in all maize fields, both monocropped and inter - tainable intensification of smallholder systems: productiv - cropped. A farm-level, weighted average maize grain yield ity, food self-sufficiency, environmental protection, and was calculated based on the area of each maize field. The economic viability. For each principle, one to four criteria yield benchmark (score 10) was 50% of the season-specific, and indicators were identified. The yield-related indicators water-limited yield potential in western Kenya, a yield target and food self-sufficiency focused on maize, which was the required to attain national or regional food self-sufficiency most important crop in terms of food and sale with nearly (van Ittersum et al. 2016). The average water-limited yield all households cultivating maize every season. potentials were calculated with a crop growth simulation Table 1 Indicators for agricultural development, organized according eties that were not “Local OPVs” (hybrid varieties, improved varie- to principles (in italics) and criteria. The third column identifies other ties, and improved open-pollinated varieties); PPP purchasing power principles under which an indicator may also fit. Yw water-limited parity. yield potential, AE adult equivalent; improved maize variety: all vari- Principles and criteria Indicators Related to other principle Unit Productivity Reducing yield gaps - Maize yield* % of Yw - Improved maize variety* % maize area −1 - N application rate* Environment kg N ha −1 Food production - Maize production Food self-sufficiency, economic kg produced household viability Food self-sufficiency −1 Food production - Maize self-sufficiency kg produced k g required Environmental protection Avoiding N losses and soil N mining - N use efficiency maize % −1 - N surplus maize kg N ha Ensuring diversification - Crop area: maize* Economic viability % of farm area - Crop area: legumes* Economic viability % of farm area Economic viability −1 −1 Allowing a decent living - Value of produce per crop$PPP AE day −1 - Value of produce per hectare of $PPP ha all crops combined* - Farm area* ha Spreading risk - Legume contribution to the com- % bined value of produce* Indicators for specific pathways 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 5 of 19 40 −1 model (hybrid-maize) using long-term weather data. They on an energy content of maize grain of 3500 kcal kg −1 −1 were 12.5 Mg ha and 8.0 Mg ha for the long- and the (Lukmanji et al. 2008). short-rain cropping seasons respectively (GYGA 2020). −1 The score was set to zero at a maize yield of 0 kg ha . In 2.3.3 Environmental protection addition, the water-limited yield potential of 80% was used as a benchmark for the maximum attainable yield and 15% Nitrogen use efficiency and N surplus Nitrogen (N) use was used as the low baseline found for current yields in efficiency of maize was calculated per season: the total N −1 SSA (van Ittersum et al. 2016). Using these seasonal aver- outputs in maize grain (kg N ha ) divided by the N inputs −1 age yield potentials is a simplification of what is possible in on all fields with maize (kg N ha ). N output was calculated the region, on average, as the water-limited yield potential using the farm-level weighted average maize grain yield and varies from season to season and from farm to farm. This a fixed N content in maize grain of 1.54% (Njoroge 2019). should be considered when evaluating the results against A farm level weighted average for N inputs was calculated the benchmarks. based on the mineral fertilizer used per field, as reported in Yield can be increased by using improved varieties. the monitoring survey. N use efficiency was analyzed using All varieties that were not local open-pollinated varieties the framework developed by the EU Nitrogen Expert Panel (OPVs) were classified as “improved” varieties. These (2015), with a minimum and a maximum N use efficiency include hybrid varieties and improved OPVs. The bench- of 50% and 90% respectively and a maximum N surplus of −1 mark score was 0 at no use of improved varieties and 10 if 80 kg N ha . A N use efficiency below 50% or a N surplus −1 100% of the maize area was sown with improved varieties. above 80 kg N ha indicated a high risk of N losses to the Mineral N application rates on maize were scored at 0 if no environment, while N use efficiencies above 90% indicated N fertilizer was applied and 10 if the mineral N application a high risk of soil mining. The framework also includes a −1 −1 rate on maize was 120 kg N ha or more. The above three general benchmark for a desired output of 80 kg N ha . indicators, associated with reducing the yield gap, were used We adjusted this benchmark to the N output at 50% of the as indicators for the pathway of intensification. water-limited yield potential, equivalent to 83 and 53 kg N −1 ha for the long-rain and the short-rain cropping seasons. Food production Maize is representative of the food pro- duced at farm level and in principle available for home con- Crop area of maize and legumes Assessing area per crop sumption. The total maize production at farm level (kg) was in smallholder farming is not straightforward as crops are calculated from maize yield and maize area for each season. commonly intercropped: e.g., maize is often intercropped with legumes such as common bean or soybean. Cultivated 2.3.2 Food self‑sufficiency area per crop (ha) was calculated as the sum of the areas of all fields containing that crop and was used to calculate Maize self‑sufficiency Maize self-sufficiency was considered yields. The percentage farm area per crop (%) was calculated an indicator for food production, as maize self-sufficiency using the estimated percentage intercropping and the field was reported to be an important production objective by area when comparing percentage areas of different crops. participating farmers (Marinus et  al. 2021). Maize self- When analyzing maize alone, the percentage intercrop- sufficiency may also be a prerequisite before farmers start ping was not considered as, in most common maize-legume to consider other changes in their farm towards sustainable intercropping systems used by farmers in western Kenya, intensification, e.g., diversification into legumes. Maize self- intercropping does not influence maize yield (Ojiem et al. sufficiency at household level (−) was calculated as the total 2014). The percentage farm area covered by maize was an maize production at farm level per season (kg) divided by indicator of specialization and by legumes of diversification. the maize requirements per household per season (kg). The If the percentage maize was above 75% of the farm area, the seasonal maize requirement was calculated from the annual score was 0 and if it was 25% or less it was 10. For legumes, requirement multiplied with the proportional contribution the score was 0 if no legumes were present and 10 if they of seasonal maize production to the annual production. The occupied more than 30% of the farm area. annual household requirements were calculated from the number of adult male equivalents (AMEs) per household 2.3.4 Economic viability and the energy requirements of an active male, 2500 kcal/ day (FAO/WHO/UNU 2001). The number of AMEs per Value of produce Value of produce per crop was calculated household was based on the family composition during the for maize, common bean, groundnut, and soybean based 2018SR, whereby a female was equivalent to 0.82 AME and on the total production per crop per season and the median children (0–18 years) 0.75 AME (FAO/WHO/UNU 2001). crop price for 2018. Median prices were obtained through −1 −1 The maize requirements were 260 kg AME year , based a weekly market survey after pooling the data from both 1 3 40 Page 6 of 19 W. Marinus et al. locations as there were limited differences. Value of pro- of Table 1 and subsequently analyze the different pathways for duce was expressed per adult equivalent per day based on sustainable intensification. the household composition in 2018 and season length. Input costs were not considered as these were largely covered by 3.1 Maize yield and production the voucher. The value of produce therefore paints a rela- tively optimistic picture and does not reflect farm profitabil- Median yields were about 15% of the seasonal-average water- ity. In addition, seasonal and within season price fluctuations limited yield potential before the program (Table  2) and were not considered, as this was not feasible for all crops strongly increased to almost 50% of the seasonal-average and inputs. We used the poverty line for Kenya (World Bank water-limited yield potential for most households from the first 2015) and the living income for rural Kenya (Anker and season of the program onwards. Some farms even reached 80% Anker 2017) as benchmarks. Both were corrected for infla- of the seasonal-average water-limited yield potential in some tion, using 2018 as reference year, which was the same year seasons. Those good yields were maintained during all five as for the crop prices. Both the poverty line and the living seasons of the program (Fig. 2). During the program, farmers income were expressed in $ purchasing power parity ($PPP) planted nearly all of their maize area with improved varieties per adult equivalent per day, following OECD (2011) and (96%) in both locations, while before the program this was Van de Ven et al. (2020). The value of produce per hectare only 46% in Vihiga and 63% in Busia. of all crops combined was expressed per hectare of farm land The maize production per household before the program for each season. It was scored at 0 if the value of produce was about 15% of that during the program, due to both a −1 was 0 $PPP ha . The score of 10 was assigned to the 75% yield increase and the increase in maize area (Table 2). Dur- percentile of the value of produce obtained by all farmers ing the program, the maize area remained relatively large in the short- and the long-rain cropping seasons, so it was a and some farmers even increased it over time (Supplemen- relative score based on the current production values. Value tary materials 2, Fig. 2). This trend was observed irrespec- of produce was considered an indicator for intensification. tive of the initial cultivated area of maize (Supplementary materials 2). Risk spreading Economic viability is improved if risk is spread by growing a variety of crops and not focusing solely 3.2 Maize self‑sufficiency and maize area on maize. We calculated the relative contribution of leg- umes (common bean, groundnut, and soybean) to the com- Maize self-sufficiency before the program in Vihiga was bined value of produce at farm level as an indicator for risk on average one-third of the required amount of maize per spreading. It was scored 0 if legumes did not contribute to household and in Busia this was half. During the program the value of produce and 10 if legumes contributed 50% or most households became maize self-sufficient. On aver - more to the value of produce. The degree of risk spreading age, in Vihiga, households were producing 1.62 times what was considered an indicator of diversification. they needed and in Busia 3.28 times (Fig. 3). Increases in maize area from the second season onwards resulted in an Farm area Farm area often limits the income that can be improvement in maize self-sufficiency for those households attained from farming (Marinus et al. 2022). We assessed in Vihiga which were not yet maize self-sufficient in the first the total farm area per farm based on measured field sizes season. In Busia, larger maize self-sufficiency was associ- of all fields in the farm and monitored this over time during ated with a smaller fraction of the farm area dedicated to the seasonal monitoring survey. Farm area was score 0 if the maize (Fig.  3). These relatively larger farms cultivated a farm area was 0 ha. The score of 10 was assigned to the 75% larger absolute area with maize than smaller farms of less percentile of the farm areas observed for all farmers, so it was than 0.5 ha, who tended to plant maize in most of their fields a relative score based on the current farm areas. An increase (Fig. 4). This critical area of 0.5 ha was roughly what was in farm area was considered an indicator for extensification. needed to produce twice the amount of maize required by typical households, indicating farmers’ priority to attain food self-sufficiency. Maize self-sufficiency and the good market for maize, albeit at low price, were named by farmers 3 Results as reasons to grow maize during the evaluation interviews. There were few differences between the two groups of farm- 3.3 Nitrogen application and nitrogen use ers, the co-learning and the comparison group, except for the efficiency expansion of legumes. Therefore, in the results section, no distinction is made between the two groups of farmers, except Before the program, farmers in Vihiga applied a similar where relevant differences arose. We first assess the indicators rate of mineral N fertilizer on maize as during the program 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 7 of 19 40 Table 2 Average household level indicators per location, before The crop area in % was corrected for intercropping. Crop production (averaged over two seasons), and during the program (averaged over and input use before the program were farmer estimates while field the five seasons). Indicators are grouped according to the pathways sizes were measured. Yields during the program were measured. of intensification/extensification and/or diversification/specialization. Vihiga (n = 23) Busia (n = 24) Before program During program Before program During program Intensification/extensification Farm area (ha) 0.33 0.41 0.76 1.02 Maize area (ha) 0.13 0.23 0.35 0.51 −1 Maize yield (kg ha ) 1513 4426 1260 4541 −1 Maize production (kg household ) 182 1028 367 2342 Maize variety type Local OPV 53 3 34 4 (% cultivated area) Improved 46 96 63 96 Mineral fertilizer application N 94 99 38 54 −1 on maize (kg ha ) P 29 47 19 25 Value of produce crops 48 107 43 113 −1 combined (×1000 Ksh ha ) Diversification/specialization Contribution per crop Maize 68 72 68 74 to combined value of Common bean 25 8 15 9 produce (%) Groundnut 4 8 10 9 Soybean 0 6 1 7 Total legumes 29 22 27 26 Crop area (%) Maize 32 41 36 40 Common bean 11 13 8 12 Groundnut 1 6 3 5 Soybean 0 6 0 6 Total legumes 12 25 11 23 (Table 2). The total amount of N applied on maize however A common choice was to use 60% of the voucher to buy a nearly doubled, but due to the increase in maize area, the rate 50 kg bag of DAP (di-ammonium phosphate) and a 50 kg remained similar. The N application rate in Busia increased bag of CAN (calcium ammonium nitrate), adding up to 23 by nearly 50% during the program as compared to before the kg of N which was the common maximum N use per farm program. P application rates increased in both sites during across the maize fields (Supplementary materials 3). Some the program as compared to before the program (Table 2). farmers with a larger maize area, mainly in Busia, bought There was a clear negative relationship between N small amounts of additional mineral fertilizer with their application rate and maize area in both Vihiga and Busia own money, resulting in moderate fertilizer N application −1 during the program (Fig. 5). High N application rates (> rates of around 50 kg N ha . −1 120 kg N ha ) were applied on farms with a small maize Only few farms across sites and seasons were within area (<0.2 ha) and the rates were largest in the first season the desired range of N use efficiency (white area in Fig.  6). (2016SR). Especially the farmers in Vihiga applied high Too high N use efficiencies (>90%), indicating soil min- rates, which was attributed to their extremely small culti- ing, were found for many of the farms in Busia, during vated areas. With an increased maize area from the second all five seasons, and for about half of the farms in Vihiga season onwards, the N application rates reduced. The other from the second season onwards. Too low N use efficien- −1 seasons showed a similar pattern as 2017LR. Farmers with cies (<50%) and too large N surpluses (>80 kg N ha ) a large maize area tended to distribute the fertilizers over were mainly found in Vihiga (Fig.  6), especially in the the whole area, resulting in lower application rates per first season, where large amounts of N-based fertilizers −1 hectare (40–50 kg N ha ). This relation between N appli- were applied on small maize areas (<0.2 ha). This problem cation rate and farm area seemed partly related to the size reduced from the second season onwards when the culti- of the input voucher, which limited total N use per farm. vated area of maize increased (Fig. 5). 1 3 40 Page 8 of 19 W. Marinus et al. Fig. 2 Total maize production per household in relation to the maize season. The short-dashed line indicates a maize grain yield of 80% of −1 cultivated area per household during the program for Vihiga (A) and the water-limited yield potential, 6400 kg ha for the SR and 10000 −1 Busia (B). The dotted line indicates a maize grain yield of 50% of the kg ha for the LR cropping season. The long-dashed line indicates a −1 seasonal-average water-limited yield potential, 4000 kg ha for the maize grain yield of 15% of the water-limited yield potential, 1200 kg −1 −1 −1 SR (short rains) and 6300 kg ha for the LR (long rains) cropping ha for the SR and 1900 kg ha for the LR cropping season. fraction of the farm area with legumes than larger farms, 3.4 Relative cropping area for maize and legumes but mostly in intercropping with maize. In evaluation inter- views, farmers with larger farms noted labor constraints for Before the program, the relative crop area for both maize cultivating legumes as their main reason for dedicating only and legumes was smaller than during the program (Table 2). a limited area to legumes. In Vihiga, legumes were mainly The share of maize increased by 10 to 25% and the share intercropped with maize. of legumes doubled. However, the area in common bean After increasing in the first seasons, the fraction of farm decreased, the area in groundnut increased, and soybean was area with maize decreased in the last season on the larger newly introduced to 6% of the farm area (Table 2). The frac- farms (2018SR, Supplementary materials 5). The initial tion of the farm area cropped with maize increased in the increases were realized both by replacing other crops (cas- first two seasons, whereas that with legumes increased in sava, sorghum) and by using additional land, e.g., by rent- later seasons. Co-learning farmers planted a larger fraction ing in land and using land that was previously fallow (not of their farm area with groundnut and soybean in the last two shown). Most farmers who decreased their maize area had seasons (2018LR and 2018SR) than the comparison farmers a relatively large maize area. They reported ample maize (Fig. 7), although this seemed to be at the cost of common self-sufficiency and low maize prices as main reasons for the bean. Groundnut and soybean were two focus crops of the decrease. Maize was replaced by groundnut and by leaving co-learning program, for rotational benefits and high value land fallow. of produce per hectare, with specific attention to intercrop- ping arrangements. The difference between comparison and 3.5 Value of crop produce co-learning groups was larger during the long-rain cropping season (Supplementary materials 4), which is locally seen The value of combined crop produce per hectare more as the main season for maize. Some households cultivated than doubled during the program when compared to before legumes mainly during the long rains and others mainly (Table  2). This was the result of yield increases of most during the short rains. Small farms tended to grow a larger 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 9 of 19 40 Fig. 3 Fraction of farm area under maize in relation to maize self- self-sufficient. The fraction of farm area under maize is not corrected sufficiency per season for Vihiga (A) and Busia (B). A maize self- for intercropping. SR stands for short-rain cropping season and LR sufficiency ratio of one (dashed line) means that a household is maize for long-rain cropping season. Fig. 4 Maize-cultivated area per farm in relation to farm area for Vih- intercropping. The dashed line indicates 0.5 ha of maize, above which iga (A) and Busia (B). The dotted line is a 1:1 line, indicating that all no farms cultivate only maize. SR stands for short-rain cropping sea- fields of the farm contain maize. The maize area is not corrected for son and LR for long-rain cropping season. 1 3 40 Page 10 of 19 W. Marinus et al. Fig. 5 Mineral N rate applied to maize fields in relation to the area cropped with maize per farm in 2016SR and 2017LR cropping seasons for Vihiga (A) and Busia (B). The dotted line indicates an application rate of −1 50 kg N ha (common) and −1 the dashed line 120 kg N ha (advised). SR stands for short- rain cropping season and LR for long-rain cropping season. crops. Only yields of mostly intercropped common bean important for co-learning farmers than comparison farmers in decreased during the program, because of the prolific maize the last two seasons (Fig. 8). In particular, groundnut became growth. important, contributing 14% and 8% to total value of produce Maize contributed most to the total value of produce for for co-learning farmers in Vihiga and Busia, respectively, in most households (Fig. 8), because of the large fraction of 2018LR. For comparison farmers, the value of produce of farm area on which it was grown. The contribution of maize legumes was 1% in Vihiga and 0% in Busia in 2018LR due to the total value of produce was more or less the same before to low yields and small areas with soybean. Soybean was and during the program and increased only slightly. As a mainly valued as an option to reduce striga infestation and consequence, the contribution of legumes slightly decreased. less important for its selling value. However, the share of common beans strongly decreased Only one household in Vihiga obtained a value of pro- (low yields, smaller fraction of farm area) and groundnut and duce that was equivalent to the living income in two of the soybean took over (Table 2). For some individual households, seasons (Fig. 8). In Busia, slightly more households in both legumes contributed two to three times more to the total groups obtained a living income, which was mainly related value of produce than maize, because of their larger legume to the larger farm area compared with Vihiga. The total area fraction combined with relatively good legume yields value of produce was equivalent to the poverty line for a (not shown). The expanding area of groundnut (Fig. 7) also few households per group in Vihiga and for about one-third explains why the value of produce of legumes became more of the households in Busia. 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 11 of 19 40 −1 Fig. 6 Farm level N outputs in maize grain in relation to mineral a N surplus of 80 kg N ha , which, if exceeded, indicates a risk of N −1 N inputs on maize, all in kg N ha , for Vihiga (A) and Busia (B). losses to the environment (light yellow-color); (2) a horizontal dashed The figure is based on the EU Nitrogen Expert Panel (2015) analysis line indicating a N output that is equivalent to 50% of the water-lim- −1 method. The upper and the lower diagonal lines with a y-intercept of ited yield  potential per season, 83 kg N ha for the long rains and −1 zero indicate a N use efficiency of 90% and 50% respectively. An N 53 kg N ha for the short rains. Below this output, the  maize grain use efficiency above 90% indicates a risk of soil N mining (deep yel- yield is lower than targeted (pink color). The remaining white area low color), while an N use efficiency below 50% indicates a risk of indicates the desired range of N efficiencies and output. SR stands for N losses to the environment (orange color). The cleat between these short-rain cropping season and LR for long-rain cropping season. two lines is further narrowed by (1) a dotted diagonal line indicating hectare of all crops, maize yield, and the use of improved 3.6 Indications of different agricultural varieties remained the same. N application rate even slightly development pathways decreased. The relative maize area showed a slight speciali- zation towards maize over time during the program, but at Farm area appeared to be an important characteristic for the same time the trends in relative legume area and the explaining the indicator values, especially in Busia. Based legume contribution to the value of produce pointed at diver- on Fig. 4, a cutoff point of 0.5 ha was determined to group sification and spreading of risk. Farm area slightly increased farmers with a smaller farm area (<0.5 ha), denoted “small over time, pointing towards extensification. farms”, and farmers with a larger farm area (>0.5 ha), In Busia the small farms showed a similar pattern: a large denoted “larger farms”, even though these farms are still positive change in intensification only at the start of the pro- very small. Above an area of 0.5 ha, no farmers cultivated gram and a decreasing N application rate during the program maize on all of their land, with one exception in Busia. In due to an increase in maize and total farm area (Fig. 9). The Vihiga, very few farms were larger than 0.5 ha, too few to specialization in maize (low score for maize area) was even consider as a separate category, so we excluded these from more pronounced than in Vihiga and coincided with a slight the analysis. decrease in the relative legume area. However, the contribu- In Vihiga, most intensification happened at the beginning tion of legumes to the value of produce slightly increased, of the program, while hardly any further intensification was pointing at risk spreading through diversification. Similar observed in subsequent years (Fig. 9). This was the case for to the small farms, the larger farms in Busia showed most all indicators related to intensification: value of produce per 1 3 40 Page 12 of 19 W. Marinus et al. Fig. 7 Average percentage of farm area cultivated with leg- umes crops before and during the program for the comparison (A) and co-learning (B) farmers in Vihiga and for the com- parison (C) and co-learning (D) farmers in Busia. The dashed line indicates the start of the program. Percentage areas per crop are corrected for intercrop- ping. SR stands for short-rain cropping season and LR for long-rain cropping season. Fig. 8 Value of produce for soybean, groundnut, common bean, and were ordered each season per location for their value of produce maize in $PPP per adult equivalent per day for each household for of maize. Household IDs were assigned per location. SR stands for the comparison (A) and co-learning (B) farmers in Vihiga and for the short-rain cropping season and LR for long-rain cropping season. comparison (C) and co-learning (D) farmers in Busia. Households 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 13 of 19 40 intensification at the start of the program and hardly any maize and contribution of legumes to the value of produce. further intensification, except for a slight increase in maize Our results are in line with the well-described difficulty of yield. The other indicators for intensification remained the enabling an increase in yields and agricultural production, same during the program. The farms diversified as shown while at the same time fulfilling other environmental and by both the relative maize area and the relative legume area economic goals that are important for sustainable intensi- and a large increase in the contribution of legumes to the fication of smallholder agriculture. value of produce, leading to spreading of risk. Farm area for both groups in Busia slightly increased over time, pointing 4.1 Farmers’ response to the voucher towards extensification. and integrated co‑learning Comparing the larger and the small farms in Busia showed a slightly higher degree of intensification on the The voucher seems to have resulted in changes in input use, larger farms by a larger value of produce per hectare and a yields, maize area, and farm area, independent of the co- higher maize yield during the program (Fig. 9). However, learning workshops (Table 2). Although maize yields and the differences were small. During the program, small farms input use prior to the program were based on farmer-reported were more diversified in terms of legume area and legume data, they were in line with current yields (MoALF 2015) contribution to value of produce, than larger farms. At the and input use (Sheahan et al. 2013; Valbuena et al. 2015) end of the program, however, the contribution of legumes reported in the literature for western Kenya. The measured to the total value of produce was larger for larger farms due maize yields and subsequent increased farm level production to higher yields of legumes, contributing to diversification allowed most households to achieve maize self-sufficiency for risk spreading. during the program. This is most likely due to the provision of the US$ 100 input voucher, as most farms in western Kenya only produce enough maize to feed the household for 4 Discussion half of the year (Valbuena et al. 2015). Although the voucher alleviated capital constraints for agriculture at household In this study, we used a diverse set of indicators to analyze level, co-learning helped to facilitate more complex changes five seasons of detailed farm level data, which was gath- such as diversification into (new) legumes such as soybean ered as part of a co-learning program with 47 farmers in and groundnut (Fig. 7). Although taking time, the iterative western Kenya. We also compared the outcomes during learning process facilitated learning on new intercropping the program (measured) with farmer-reported data, col- arrangements of maize and legumes and identified specific lected during a baseline study held before the program. objectives for soybean (e.g., reducing striga incidence) and −1 We compared the integrated co-learning approach (Mari- groundnut (e.g., high value of produce ha ), as described in nus et al. 2021), which included an input voucher, with more detail in Marinus et al. (2021). Co-learning can thus be a voucher-only approach. We assessed whether the inte- used to contextualize knowledge for the breadth of options grated co-learning approach and/or the input voucher-only that is needed for sustainable intensification (Descheemaeker would lead to pathways of intensification or extensification et al. 2019; Ronner et al. 2021). and pathways of diversification or specialization. We did Initially, all households in our sample, irrespective of not observe a difference between farmers only receiving a farm area, specialized in maize both in Vihiga and in Busia. voucher and those also taking part in the co-learning pro- Larger farms, however, reduced their maize area again in the gram, so we analyzed them as one group. The only excep- last season (2018SR), while small farms maintained their tion was the adoption of legumes, which were included increased maize area. Similar increases in maize area after more substantially by the co-learning farmers. Soybean was the introduction of an input voucher or subsidy have been newly introduced and groundnut substantially expanded, described before for western Kenya (Sanchez et al. 2007) which led to a more diversified cropping system. All farm- and Malawi (Holden and Lunduka 2010; Chibwana et al. ers in our sample increased maize yields (intensification) 2012), based on farmer-reported maize areas. In these stud- compared to the situation before the program, although an ies, however, an increase in maize area often resulted in a increase in farm and maize areas in combination with rela- decrease in legume area (Holden and Lunduka 2010; Chib- tively low N application rates (risk of soil N mining) also wana et al. 2012). For small farms, maintaining the large pointed to extensification and specialization. The value of maize area was associated with farmers’ objectives to be produce remained below a living income for most house- maize self-sufficient (Marinus et al. 2021). holds in our sample due to the small farm areas. This was The financial benefits of diversification into groundnut more prominent in Vihiga than in Busia. The larger farms and soybean that we observed were in line with the find- in our sample scored better in terms of diversification than ings of Franke et al. (2014), who simulated benefits of the small farms, especially related to fraction of the area in diversification with legumes for different farm types in 1 3 40 Page 14 of 19 W. Marinus et al. Fig. 9 Spider web diagrams with average indicator scores per indicator for farms with a relatively small (<0.5 ha) and a larger farm area (>0.5 ha) (data from 2015SR) in Vihiga (A) and Busia (B). For Vihiga, larger farms were left out, as there were too few. A larger score indicates a more sustainable situation. Dotted lines represent the short-rain (SR) cropping season before the start of the program, 2015SR. Dashed lines represent the first season of the program, 2016SR, while solid lines represent the last season of the program, 2018SR. The indicators maize yield, N appli- cation rate, and improved maize variety refer to maize crop level. The other indicators are at farm level. (I) An intensification indicator, (E) extensification indicator, (D) diversity indica- tor, (S) specialization indicator. The “–” sign after maize area indicates a negative relation for this specific indicator: maize area receives a higher score if the cultivated area with maize is smaller. Malawi. Diversification is important for spreading risks diggers for harvesting and shellers (Tsusaka et al. 2017), (crop failure, low prices), and for nutritional and rotational may be required to enable diversification for households benefits (Vanlauwe et al. 2019). On the small farms, leg- with a larger farm area. umes were mainly intercropped with maize resulting in limited benefits due to land constraints. However, on larger 4.2 Concurrent pathways of intensification farms, labor constraints were limiting the expansion of and extensification legumes, similar to the findings of Franke et al. (2014). This would imply that developing and promoting legume- The maize yields obtained during the program point at the specific, small-scale mechanization, such as groundnut pathway of intensification as yields were two to three times 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 15 of 19 40 greater than the yields reported by participating farmers and therefore desirable. Increasing farm area by some farm- before the program, and close to the seasonal-average bench- ers, by renting in land that was ready in active use, meant mark of 50% of the water-limited yield potential. However, that farm area decreased for other farmers. If this would as the corresponding N application rates were both above happen on a larger scale, increasing farm areas could push and below the desirable range, it appeared to be difficult others out of agriculture, requiring alternative employment to enhance N use efficiency, which is a typical challenge for those going out of agriculture (Giller 2020). pertaining to sustainable intensification (Zhang et al. 2015). Except for the first season, the N application rates were Intensified mineral fertilizer use resulted in extremely remarkably similar for the small and larger farmers in our −1 −1 high N application rates in Vihiga (>200 kg N ha in the sample at about 50 kg N ha . This was partly limited by −1 first season, ~100 kg N ha in later seasons) due to the the fixed voucher size of US$ 100. Farmers with a rela- small farm areas in our sample there, resulting in N use effi- tively large maize area, who bought additional fertilizers −1 ciencies below 50%. These farms of less than 0.2 ha were still applied N at a maximum of 50 kg N ha , despite the not able to allocate all inputs from the voucher in a useful advice in the co-learning workshops to apply more. This manner, even with an increased farm or maize area in later may be partly due to the active presence of One Acre Fund seasons. Maize yields were not negatively related to farm in the area who, as a credit provider, advises farmers to use −1 area, and in some seasons even positively related to farm this conservative rate of 50 kg N ha . One Acre Fund was area, while for a small sample, this seems to go against the already present in the program locations before the start of inverse farm size-productivity relationship (Larson et al. the program and did not change the intensity of their activi- 2014). Our finding however is in line with Desiere and Jol- ties during the program. The relatively good yields and low liffe (2018) and Gourlay et al. (2017) who also found no N fertilizer application rates are probably not sustainable negative relation between farm area and yield. Notwithstand- as they will likely result in soil N mining. Soil N mining ing higher N application rates, smaller farms did not seem is common in SSA, but usually at lower yields and lower to produce better yields than larger farms, which may be input levels than in our study (Sheahan and Barrett 2017). explained by reliance on off-farm work requiring farmers’ We diagnosed negative N balances over multiple seasons, attention (Leonardo et al. 2015) and the presence of poorer which suggest that soil mining will occur on the long term. soils (Franke et al. 2019), requiring longer term investments This may have been enabled by the increased application in soil fertility (Vanlauwe et al. 2010). of P through the mineral fertilizers. In the P-fixing soils Extensification was observed on larger farms, who of the study area, P limits mineralization and strong yield increased their farm and/or maize areas and hence distrib- responses to P can be found (Kihara and Njoroge 2013). uted N over larger areas. This was most notable in Busia, as Another reason may be that we did not account for N inputs also discussed in Marinus et al. (2023), where population from manure and N -fixation, although these were small −1 pressure is lower and fallow land is available. We hypoth- (<14 kg N ha for manure on average, with large varia- esize that the expansion in farm area was enabled by the tion in rates due to likely recall error). When good yields in voucher (Marinus et al. 2023). The preference of the farmers combination with soil N mining are continued, N and other in our sample for extensification over intensification goes nutrients (e.g., K) may become limiting (Njoroge 2019) and against one of the key objectives of sustainable intensifica- fertilizer rates will need to be adjusted. tion, namely to increase agricultural production on existing farmland (Cassman et al. 2003; Struik and Kuyper 2017). 4.3 Development pathways evaluated The preference for land expansion among African small- by multi‑criteria analysis holders to increase production, however, seems to be a gen- eral trend for crop area (Baudron et al. 2012; Ollenburger We combined indicators that farmers indicated to be et al. 2016; Jayne and Sanchez 2021; Giller et al. 2021). At important from their perspective (e.g., maize self-suffi - farm level, however, extensification may be less expensive ciency, value of produce) with indicators that are impor- than increasing input rates with the associated larger risks tant for local or national food self-sufficiency (e.g., yield of financial losses (Tittonell et al. 2007; Burke et al. 2019; and production) and environmental protection (e.g., N use Jindo et al. 2020), which can help to explain the observed efficiency, N surplus) in an integrated assessment. This farmers’ preference. The additional fields that farmers rented analysis, in combination with the discussions with the co- in were either previously fallow, or already in active use for learning farmers, identified potential constraints and trade- agriculture. Expanding into fallow land or nature areas, on offs at farm level. Achieving and even surpassing maize the one hand, can result in environmental concerns as it can self-sufficiency was a first priority for farmers, because of jeopardize current ecosystem services such as providing nat- the importance of having surplus food as a buffer for later ural habitats or erosion control. On the other hand, using fal- seasons and the reliable market for maize (Marinus et al. low land more frequently, can also be seen as intensification 2021). This priority could stimulate specialization. The 1 3 40 Page 16 of 19 W. Marinus et al. limited observed diversification goes against a common on crop area, can be subjective and thereby require trans- assumption in modeling studies that farmers are likely to parency on why they were chosen and which benchmarks diversify into other crops once they are maize self-suffi- were used (Marinus et al. 2018). cient (e.g., Hengsdijk et al. 2014; Leonardo et al. 2018). Increasing the value of produce obtained from farming was a second objective for farmers. However, despite the 5 Conclusions good yields, farm area and in Busia also, labor availabil- We analyzed whether farmer responses to a voucher and ity seemed to be overriding constraints for reaching the income benchmarks. At best, one-third of the households co-learning were indicative of different pathways for agricultural development over a period of five seasons by in our sample obtained a living income and half of the households reached the poverty line in Busia. In Vihiga, applying an indicator framework. Our overarching aim was to improve the understanding of farmer responses to input only one out of twenty-three households obtained a liv- ing income in two seasons, while at best one-fourth of subsidies and new knowledge, in order to better support desired agricultural development pathways in smallholder the sample reached the poverty line. Increasing farm area per household or extensification may thus be needed to farming. Although we focused on a limited number of farmers, 47 in total, we believe that based on the detailed increase income from farming to a living income. New employment opportunities will then be needed for those data collection over multiple seasons, some conclusions can be drawn. The novel integrated co-learning approach who choose to leave farming (Giller 2020), if no additional land is available or if an increase of agricultural land is not which we developed facilitated more complex changes in farm management, such as diversification through an desired (e.g., Godfray et al. 2010; The Montpellier Panel 2013). Following area expansion, in Busia, mechanization increase in legume area and legume contribution to the value of produce. Other responses were mainly related to could alleviate the labor constraints for cultivation of prof- itable crops such as legumes, of which the further expan- the input voucher itself. Increased input use through the voucher seemed to sion during the program seemed to be limited by labor constraints. Mechanization could thereby facilitate further increase yields and production, indicating a pathway of intensification that allowed households to achieve maize diversification into more profitable crops for economic viability. Apart from changing to more profitable crops self-sufficiency. As a result of increases in maize and farm area on larger farms, N application rates remained and increasing farm area, selling products at times of high prices can also be a way to increase income. This strategy constant, despite larger inputs. Accompanied by too low N use efficiencies, this pointed at extensification and a is not within reach of all farmers, as it depends on their short-term needs for cash. We did not consider seasonal risk of not reaching environmental protection objectives. Most small farms were only just maize self-sufficient and price fluctuations, although maize prices can differ more than a factor two between the scarce lean season and just their value of produce remained below the poverty line. Obtaining a living income was only possible on large after harvest, when maize is abundant (Burke et al. 2017). A more extensive analysis of each individual household farm areas. It should be noted, however, that we based this on the product prices of 2018. Different prices would give and price fluctuations would be required to assess this. Disaggregating the analysis per household showed a somewhat different picture, but it is clear that prices would have to increase several-fold to lift the majority that farm area limited outcomes for both small and larger farms in specific ways. For example, N use efficiencies of the farmers out of poverty. Our multi-criteria analysis highlighted the difficulty of supporting diversification as were below (for about one-third of farms in Vihiga) or above the desired range (for about half of the farms in a pathway towards sustainable intensification. Improving livelihoods requires changes that go far beyond the farm both sites), respectively, while outputs (yield) were in the desired range. Another methodological lesson learned level. Smallholder farmers in western Kenya and in many rural areas of sub-Saharan Africa are essentially part- is that assessing adoption of new crops or varieties in programs on diversification needs multi-season studies time farmers who depend on many sources of income. To increase income from farming, farm areas need to (Glover et al. 2019) as our results showed that the leg- ume area per farm differed per season and not necessar- increase, which requires off-farm employment opportu- nities for those who choose to leave farming. Whether ily according to the season when legumes were known to be most commonly cultivated. Finally, the principles, sustainable intensification of smallholder agriculture will actually happen may therefore depend on how changes criteria, and indicator framework, following Florin et al. (2012), was useful in being explicit on the underlying in farm structure—that is, capital, land and labor—are facilitated at farm level and as part of the wider socio- assumptions, i.e., criteria, on when an indicator contrib- utes to sustainability. Some of these assumptions, e.g., economic developments within a country. 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 17 of 19 40 Supplementary Information The online version contains supplemen- Zimbabwe. J Dev Stud 48:393–412. https://doi. or g/10. 1080/ 00220 tary material available at https://doi. or g/10. 1007/ s13593- 023- 00893-w .388. 2011. 587509 Bonilla-Cedrez C, Chamberlin J, Hijmans RJ (2021) Fertilizer and Authors' contributions All authors contributed to the study conception grain prices constrain food production in sub-Saharan Africa. Nat and design. Data collection and analysis were performed by Wytze Food 2:766–772. https:// doi. org/ 10. 1038/ s43016- 021- 00370-1 Marinus. The first draft of the manuscript was written by Wytze Mari- Burke M, Bergquist LF, Miguel E (2017) Selling low and buying high: nus and all authors commented on previous versions of the manuscript. an arbitrage puzzle in Kenyan Villages. https:// web. stanf ord. edu/ All authors read and approved the final manuscript.~mburke/ papers/ Burke_ stora ge_ 2013. pdf Burke WJ, Frossard E, Kabwe S, Jayne TS (2019) Understanding ferti- Funding This work was funded through the CGIAR research programs lizer adoption and effectiveness on maize in Zambia. Food Policy on MAIZE (project number A5014.09.08.01), the HumidTropics, and 86:101721. https:// doi. org/ 10. 1016/j. foodp ol. 2019. 05. 004 by the Plant Production Systems group of Wageningen University. KEG Cassman KG, Dobermann A, Walters DT, Yang H (2003) Meeting thanks the NWO-WOTRO Strategic Partnership NL-CGIAR for the cereal demand while protecting natural resources and improving financial support. environmental quality. Annu Rev Environ Resour 28:315–358. https:// doi. org/ 10. 1146/ annur ev. energy. 28. 040202. 122858 Data availability The datasets generated during and/or analyzed dur- Chibwana C, Fisher M, Shively G (2012) Cropland allocation effects ing the current study are available from the corresponding author on of agricultural input subsidies in Malawi. World Dev 40:124–133. reasonable request.https:// doi. org/ 10. 1016/j. world dev. 2011. 04. 022 Crowley EL, Carter SE (2000) Agrarian change and the changing Code availability The datasets generated during and/or analyzed dur- relationships between toil and soil in Maragoli, Western Kenya ing the current study are available from the corresponding author on (1900–1994). Hum Ecol 28:383–414. https:// doi. org/ 10. 1023/A: reasonable request.10070 05514 841 Descheemaeker KKE, Ronner E, Ollenburger M et al (2019) Which options fit best? Operationalizing the socio-ecological niche Declarations concept. Exp Agric 55:169–190. https://d oi.o rg/1 0.1 017/S 0014 47971 60004 8X Ethics approval Participating farmers were well informed about the Desiere S, Jolliffe D (2018) Land productivity and plot size: is meas- purpose of the study prior to participation and informed consent was urement error driving the inverse relationship? World Bank, obtained from all participants involved in the study. Participants were Washington, DC, USA free withdraw from the study at any moment. Data was collected and Dorward A, Chirwa E, Kelly VA et  al (2008) Evaluation of the stored according to the data management plan of the Plant Production 2006/7 Agricultural Input Subsidy Programme, Malawi. Final Systems group of Wageningen University (https://git. wur .nl/ pps/ PPS_ report. https:// eprin ts. soas. ac. uk/ 16730/1/ March Repor tRevM data_m anage ment/-/r aw/m aster/w ritin g/P PS_D ata_M anage ment.p df). ayNoT rack. pdf Ethical approval for this study was not required according to the check- EU Nitrogen Expert Panel (2015) Nitrogen Use Efficiency (NUE) list of the Social Sciences Ethics Committee of Wageningen University. - an indicator for the utilization of nitrogen in agriculture and food systems. Wageningen University, Alterra, Wageningen, the Consent to participate Verbal informed consent was obtained from all Netherlands. http://www .eunep. com/ wp- conte nt/ uploa ds/ 2017/ 03/ individual participants included in the study. Report- NUE- Indic ator- Nitro gen- Expert- Panel- 18- 12- 2015. pdf FAO/WHO/UNU (2001) Human energy requirements. Report of a joint Consent for publication Not applicable FAO/WHO/UNU expert consultation, Rome Florin MJ, van Ittersum MK, van de Ven GWJ (2012) Selecting the Open Access This article is licensed under a Creative Commons Attri- sharpest tools to explore the food-feed-fuel debate: sustainability bution 4.0 International License, which permits use, sharing, adapta- assessment of family farmers producing food, feed and fuel in tion, distribution and reproduction in any medium or format, as long Brazil. Ecol Indic 20:108–120. https:// doi. org/ 10. 1016/j. ecoli nd. as you give appropriate credit to the original author(s) and the source, 2012. 02. 016 provide a link to the Creative Commons licence, and indicate if changes Franke AC, Baijukya F, Kantengwa S et al (2019) Poor farmers - poor were made. The images or other third party material in this article are yields: socio-economic, soil fertility, and crop management indi- included in the article's Creative Commons licence, unless indicated cators affecting climbing bean productivity in northern Rwanda. otherwise in a credit line to the material. If material is not included in Exp Agric 55:14–34. https:// doi. org/ 10. 1017/ S0014 47971 60000 the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will Franke AC, van den Brand GJ, Giller KE (2014) Which farmers benefit need to obtain permission directly from the copyright holder. To view a most from sustainable intensification? An ex-ante impact assess- copy of this licence, visit http://cr eativ ecommons. or g/licen ses/ b y/4.0/ . ment of expanding grain legume production in Malawi. Eur J Agron 58:28–38. https:// doi. org/ 10. 1016/j. eja. 2014. 04. 002 Giller KE (2020) The food security conundrum of sub-Saharan Africa. Glob Food Sec 26:100431. https:// doi. org/ 10. 1016/j. gfs. 2020. References 100431 Giller KE, Delaune T, Silva JV et al (2021) The future of farming: who will produce our food? Food Secur 13:1073–1099. https://d oi.o rg/ Anker R, Anker M (2017) Living wage report Kenya - with a focus on 10. 1007/ s12571- 021- 01184-6 rural Mount Kenya Area. Global Living Wage Coalition, Series Giller KE, Tittonell P, Rufino MC et al (2011) Communicating com- 1, Report 10. p 51. http://www .iseal allia nce. or g/sites/ def ault/ files/ plexity: integrated assessment of trade-offs concerning soil fertil- Kenya_ Living_ Wage_ Bench mark_ Report. pdf ity management within African farming systems to support inno- Baudron F, Andersson JA, Corbeels M, Giller KE (2012) Failing to vation and development. Agric Syst 104:191–203. https://doi. or g/ yield? Ploughs, conservation agriculture and the problem of 10. 1016/j. agsy. 2010. 07. 002 agricultural intensification: an example from the Zambezi Valley, 1 3 40 Page 18 of 19 W. Marinus et al. Glover D, Sumberg J, Ton G et al (2019) Rethinking technological Leonardo WJ, van de Ven GWJ, Udo H et al (2015) Labour not land change in smallholder agriculture. Outlook Agric 48(3):169– constrains agricultural production and food self-sufficiency in 180. https:// doi. org/ 10. 1177/ 00307 27019 864978 maize-based smallholder farming systems in Mozambique. Food Godfray HCJ, Beddington JR, Crute IR et al (2010) Food security: Secur 7:857–874. https:// doi. org/ 10. 1007/ s12571- 015- 0480-7 the challenge of feeding 9 billion people. Science 327:812–818. Lukmanji Z, Hertzmark E, Mlingi N, Assey V (2008) Tanza- https:// doi. org/ 10. 1126/ scien ce. 11853 83 nia food composition tables. Muhimbili University of Allied Gourlay S, Kilic T, Lobell DB (2019) A new spin on an old debate: Science,Tanzania Food and Nutrition Centre and Havard School Errors in farmer-reported production and their implications for of Public Health, Dar es Salaam Tanzania inverse scale - Productivity relationship in Uganda. J Dev Econ Marinus W, Descheemaeker K, van de Ven GW et al (2023) Narrow- 141:102376. https:// doi. org/ 10. 1016/j. jdeve co. 2019. 102376 ing yield gaps does not guarantee a living income from small- GYGA (2020) Global Yield Gap Atlas - climate zone Kenya, Kaka- holder farming – an empirical study from western Kenya. Plos mega (Code=7-7-01) rainfed maize. http:// www. yield gap. org/. One 15:e0283499. https://d oi.o rg/1 0.1 371/j ourna l.p one.0 28349 9 Accessed 23 Nov 2020 Marinus W, Descheemaeker KKE, van de Ven GWJ et al (2021) “That Hengsdijk H, Franke AC, van Wijk MT, Giller KE (2014) How small is my farm” – an integrated co-learning approach for whole-farm is beautiful? Food self-sufficiency and land gap analysis of small- sustainable intensification in smallholder farming. Agric Syst holders in humid and semi-arid sub Saharan Africa, Report num- 188:103041. https:// doi. org/ 10. 1016/j. agsy. 2020. 103041 ber 562. Plant Research International, part of Wageningen, UR. Marinus W, Ronner E, van de Ven GWJ et al (2018) The devil is in the https:// edepot. wur. nl/ 331203 detail! Sustainability assessment of African smallholder farming. Holden S, Lunduka R (2010) Too poor to be efficient? Impacts of the In: Bell S, Morse S (eds) Sustainability Indicators and Indices. targeted fertilizer subsidy programme in Malawi on farm plot level Routledge, London, pp 453–476 input use, crop choice and land productivity. Noragric Rep. No Marinus W, Thuijsman ES, van Wijk MT et  al (2022) What farm 55. p 55. http:// www. umb. no/ stati sk/ norag ric/ publi catio ns/ repor size sustains a living? Exploring future options to attain a living ts/ 2010_ nor_ rep_ 55. pdf income from smallholder farming in the East African Highlands. Jaetzold R, Schmidt H, Hormetz B, Shisanya C (2005) Farm manage- Front Sustain Food Syst 5:1–15. https:// doi. org/ 10. 3389/ fsufs. ment handbook of Kenya: natural conditions and farm manage-2021. 759105 ment information. In: Subpart A1 Western Province, 2nd edn. Martin W, Anderson K (2008) Agricultural trade reform under the Ministry of Agriculture Kenya and the German agency for Techni- Doha Agenda: some key issues. Aust J Agric Resour Econ 52:1– cal Cooperation (GTZ), Nairobi 16. https:// doi. org/ 10. 1111/j. 1467- 8489. 2008. 00404.x Jayne TS, Mason NM, Burke WJ, Ariga J (2018) Review: taking stock Michelson H, Fairbairn A, Ellison B et al (2021) Misperceived quality: of Africa’s second-generation agricultural input subsidy programs. fertilizer in Tanzania. J Dev Econ 148:102579. https://doi. or g/10. Food Policy 75:1–14. https:// doi. org/ 10. 1016/j. foodp ol. 2018. 01. 1016/j. jdeve co. 2020. 102579 003 MoALF (2015) Economic Review of Agriculture 2015. Ministry of Jayne TS, Rashid S (2013) Input subsidy programs in sub-Saharan Agriculture, Livestock and Fisheries (MoALF), Nairobi, Kenya Africa: a synthesis of recent evidence. Agric Econ 44:547–562. Njoroge SK (2019) Feed the crop, not the soil! Explaining variability https:// doi. org/ 10. 1111/ agec. 12073 in maize yield responses to nutrient applications in smallholder Jayne TS, Sanchez PA (2021) Agricultural productivity must improve farms of western Kenya. PhD thesis, Wageningen University, in sub-Saharan Africa. Science (80-) 372:1045–1047. https://doi. Wageningen. https:// edepot. wur. nl/ 503185 org/ 10. 1126/ scien ce. abf54 13 OECD (2011) What are equivalence scales? OECD, p. 2. https:// Jayne TS, Snapp S, Place F, Sitko N (2019) Sustainable agricultural www. oecd. org/ els/ soc/ OECD- Note- Equiv alenc eScal es. pdf intensification in an era of rural transformation in Africa. Glob Ojiem JO, Franke AC, Vanlauwe B et al (2014) Benefits of legume– Food Sec 20:105–113. https:// doi. org/ 10. 1016/j. gfs. 2019. 01. 008 maize rotations: assessing the impact of diversity on the produc- Jindo K, Schut AGT, Langeveld JWA (2020) Sustainable intensifica- tivity of smallholders in Western Kenya. F Crop Res 168:75–85. tion in Western Kenya: who will benefit? Agric Syst 182:102831. https:// doi. org/ 10. 1016/j. fcr. 2014. 08. 004 https:// doi. org/ 10. 1016/j. agsy. 2020. 102831 Ollenburger MH, Descheemaeker K, Crane TA et al (2016) Waking Kihara J, Njoroge S (2013) Phosphorus agronomic efficiency in maize- the Sleeping Giant: agricultural intensification, extensification based cropping systems: a focus on western Kenya. F Crop Res or stagnation in Mali’s Guinea Savannah. Agric Syst 148:58–70. 150:1–8. https:// doi. org/ 10. 1016/j. fcr. 2013. 05. 025https:// doi. org/ 10. 1016/j. agsy. 2016. 07. 003 Klapwijk C, van Wijk M, Rosenstock T et al (2014) Analysis of trade- Pretty J, Toulmin C, Williams S (2011) Sustainable intensification offs in agricultural systems: current status and way forward. Curr in African agriculture. Int J Agric Sustain 9:5–24. https:// doi. Opin Environ Sustain 6:110–115. https://doi. or g/10. 1016/j. cosus t. org/ 10. 3763/ ijas. 2010. 0583 2013. 11. 012 Ronner E, Sumberg J, Glover D, et  al (2021) Basket of options: KNBS (2019) 2019 Kenya population and housing census volume 1: unpacking the concept. Outlook Agric 003072702110194. population by county and sub-county. Kenya National Bureau of https:// doi. org/ 10. 1177/ 00307 27021 10194 27 Statistics, Nairobi Sanchez P, Palm C, Sachs J et al (2007) The African Millennium Koning N (2017) Food security, agricultural policies and economic Villages. Proc Natl Acad Sci 104:16775–16780. https://doi. or g/ growth: long-term dynamics in the past, present and future. Rout-10. 1073/ pnas. 07004 23104 ledge, Abingdon, United Kindom Santpoort R (2020) The drivers of maize area expansion in sub- Larson DF, Otsuka K, Matsumoto T, Kilic T (2014) Should African Saharan Africa. How policies to boost maize production over- rural development strategies depend on smallholder farms? An look the interests of smallholder farmers. Land 9. https:// doi. exploration of the inverse-productivity hypothesis. Agric Econ org/ 10. 3390/ land9 030068 45:355–367. https:// doi. org/ 10. 1111/ agec. 12070 Schut AGT, Giller KE (2020) Sustainable intensification of agricul- Leonardo W, van de Ven GWJ, Kanellopoulos A, Giller KE (2018) Can ture in Africa. Front Agric Sci Eng. https://doi. or g/10. 15302/J- farming provide a way out of poverty for smallholder farmers in FASE- 20203 57 central Mozambique? Agric Syst 165:240–251. https://doi. or g/10. Schut M, Cunha Soares N, van de Ven G, Slingerland M (2014) 1016/j. agsy. 2018. 06. 006 Multi-actor governance of sustainable biofuels in developing 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 19 of 19 40 countries: the case of Mozambique. Energy Policy 65:631–643. Paper Series No 44. International Crops Reearch Institute for the https:// doi. org/ 10. 1016/j. enpol. 2013. 09. 007 Semi-Arid Tropics (ICRISAT), Lilongwe, Malawi. http://oar .icr is Sheahan M, Barrett CB (2017) Ten striking facts about agricultural at. org/ 10232/1/ T_ Tsusa ka_ etal_ ISEDPS_ 44. pdf input use in Sub-Saharan Africa. Food Policy 67:12–25. https:// Valbuena D, Groot JCJ, Mukalama J et al (2015) Improving rural liveli- doi. org/ 10. 1016/j. foodp ol. 2016. 09. 010 hoods as a “moving target”: trajectories of change in smallholder Sheahan M, Black R, Jayne TS (2013) Are Kenyan farmers under- farming systems of Western Kenya. Reg Environ Chang 15:1395– utilizing fertilizer? Implications for input intensification strat- 1407. https:// doi. org/ 10. 1007/ s10113- 014- 0702-0 egies and research. Food Policy 41:39–52. https:// doi. org/ 10. van de Ven GWJ, de Valença A, Marinus W et al (2020) Living income 1016/j. foodp ol. 2013. 04. 008 benchmarking of rural households in low-income countries. Food Smale M, Jayne T (2003) Maize in Eastern and Southern Africa : “ Secur. https:// doi. org/ 10. 1007/ s12571- 020- 01099-8 seeds ” of success in retrospect. EPTD DISCUSSION PAPER van Ittersum MK, van Bussel LGJ, Wolf J et al (2016) Can sub-Saharan NO. 97. International Food Policy Research Institute. Wash- Africa feed itself? Proc Natl Acad Sci 113:14964–14969. https:// ington DC, USA, p 79. https:// ebrary. ifpri. org/ digit al/ api/ colle doi. org/ 10. 1073/ pnas. 16103 59113 ction/ p1573 8coll2/ id/ 77285/ downl oad van Loon MP, Hijbeek R, Berge HFM et al (2019) Impacts of inten- Smith A, Snapp S, Chikowo R et al (2017) Measuring sustainable sifying or expanding cereal cropping in sub‐Saharan Africa on intensification in smallholder agroecosystems: a review. Glob greenhouse gas emissions and food security. Glob Chang Biol Food Sec 12:127–138. https://doi. or g/10. 1016/j. gf s.2016. 11. 002 25:3720–3730. https:// doi. org/ 10. 1111/ gcb. 14783 Struik PC, Kuyper TW (2017) Sustainable intensification in agricul- Vanlauwe B, Bationo A, Chianu J et al (2010) Integrated soil fertility ture: the richer shade of green. A review. Agron Sustain Dev 37. management operational definition and consequences for imple- https:// doi. org/ 10. 1007/ s13593- 017- 0445-7 mentation and dissemination. Outlook Agric 39:17–24 Struik PC, Kuyper TW, Brussaard L, Leeuwis C (2014) Deconstruct- Vanlauwe B, Coyne D, Gockowski J et al (2014) Sustainable intensi- ing and unpacking scientific controversies in intensification and fication and the African smallholder farmer. Curr Opin Environ sustainability: why the tensions in concepts and values? Curr Sustain 8:15–22. https:// doi. org/ 10. 1016/j. cosust. 2014. 06. 001 Opin Environ Sustain 8:80–88. https://doi. or g/10. 1016/j. cosus t. Vanlauwe B, Dobermann A (2020) Sustainable intensification of agri- 2014. 10. 002 culture in sub-Saharan Africa: first things first. Front Agric Sci The Montpellier Panel (2013) Sustainable intensification: a new Par - Eng 7:376. https:// doi. org/ 10. 15302/J- FASE- 20203 51 adigm for African Agriculture. Agriculture for Impact, London Vanlauwe B, Hungria M, Kanampiu F, Giller KE (2019) The role of leg- Tittonell P, van Wijk MT, Rufino MC et al (2007) Analysing trade- umes in the sustainable intensification of African smallholder agri- offs in resource and labour allocation by smallholder farmers culture : lessons learnt and challenges for the future. Agric Ecosyst using inverse modelling techniques: a case-study from Kaka- Environ 284:106583. https:// doi. org/ 10. 1016/j. agee. 2019. 106583 mega district, western Kenya. Agric Syst 95:76–95. https://doi. World Bank (2015) Purchasing power parities and the real size of world org/ 10. 1016/j. agsy. 2007. 04. 002 economies: a Comprehensive Report of the 2011 International Tittonell PA, Giller KE (2013) When yield gaps are poverty traps: Comparison Program. The World Bank, Washington, DC the paradigm of ecological intensification in African small- Zhang X, Davidson EA, Mauzerall DL et al (2015) Managing nitrogen holder agriculture. F Crop Res 143:76–90. https:// doi. org/ 10. for sustainable development. Nature 1–9. https://doi. or g/10. 1038/ 1016/j. fcr. 2012. 10. 007natur e15743 Tsusaka TW, Twanje GH, Msere HW et al (2017) Ex-ante assessment of adoption of small-scale post-harvest mechanization: the case Publisher's note Springer Nature remains neutral with regard to of groundnut producers in Malawi. Socioeconomics Discussion jurisdictional claims in published maps and institutional affiliations. 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Agronomy for Sustainable Development Springer Journals

Farmer responses to an input subsidy and co-learning program: intensification, extensification, specialization, and diversification?

Loading next page...
 
/lp/springer-journals/farmer-responses-to-an-input-subsidy-and-co-learning-program-tovQp42ipp

References (81)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2023
ISSN
1774-0746
eISSN
1773-0155
DOI
10.1007/s13593-023-00893-w
Publisher site
See Article on Publisher Site

Abstract

Sustainable intensification aims to increase production and improve livelihoods of smallholder farmers in sub-Saharan Africa. Many farmers, however, are caught in a vicious cycle of low productivity and lack of incentives to invest in agricultural inputs. Moving towards sustainable intensification therefore requires support such as input subsidies and learning about new options through, for instance, co-learning approaches. Yet such support is not straightforward as agricultural developments often diverge from the envisaged pathways: extensification may occur instead of intensification and specialization instead of diversification. Understanding of farmers’ responses to incentives such as input subsidies and new knowledge is lacking. Our overarching aim was to improve this understanding, in order to better support future pathways for agricultural develop- ment in smallholder farming. Over five seasons, we compared the responses of farmers in western Kenya taking part in a novel co-learning program we developed, which included provision of an input voucher, with the responses of farmers who only received a voucher. We also assessed the differences before and during the program. We used diverse indicators that were related to the different agricultural development pathways. Farmer responses were mainly a result of the input voucher. Farmers increased maize yields (intensification) and maize area (specialization) for maize self-sufficiency. Increased farm and maize areas in combination with relatively low N application rates also pointed to extensification coupled with the risk of soil N mining. Diversification by increasing the soybean and groundnut area share was facilitated by the integrated co-learning approach, which thereby supported relatively complex farm management changes. Our results highlight the dif- ficulty of enabling yield and production increases, while also meeting environmental and economic goals. The diversity of farmer responses and constraints beyond the farm level underlined the importance of wider socio-economic developments in addition to support of sustainable intensification at farm level. Keywords Sustainable intensification · Smallholder farmers · Sub-Saharan Africa · Multi-criteria assessment · Indicators · Pathways · Yield · N use efficiency · Living income · Subsidies 1 Introduction and a changing climate require considerable changes in current farming systems (Giller 2020). Sustainable Livelihoods of smallholder farmers in sub-Saharan Africa intensification of farming is seen as a key strategy to (SSA) are under pressure. Many are caught in a poverty enhance rural livelihoods in SSA (Vanlauwe et al. 2014; trap, a vicious cycle of low productivity and lack of Jayne and Sanchez 2021). Sustainable intensification opportunities and incentives to invest in agricultural inputs aims to enhance production per unit land, nutrient, and (Tittonell and Giller 2013; Koning 2017). Additionally, labor input, while reducing environmental damage, constraints such as small farm sizes, limited market access, building resilience, and natural capital, as well as securing environmental services (e.g., Pretty et  al. 2011; The Montpellier Panel 2013). Struik and Kuyper (2017) argue * Katrien Descheemaeker that the concept of sustainable intensification can be used katrien.descheemaeker@wur.nl as a “process of inquiry and analysis” and discuss how Plant Production Systems, Wageningen University, P.O. the social and economic dimensions of sustainability can Box 430, 6700, AK, Wageningen, The Netherlands be included. Such a broad view enables identification of Central Africa Hub Office, IITA, P.O. Box 30722-00100, trade-offs that arise when agricultural systems intensify. Nairobi, Kenya Vol.:(0123456789) 1 3 40 Page 2 of 19 W. Marinus et al. Using a diverse set of indicators to describe these trade- level cereal self-sufficiency is an important indicator that fits offs, can inform decision-making by society and policy with farmers’ objectives. makers (Struik et al. 2014; Struik and Kuyper 2017). Yield-increasing inputs required for sustainable intensi- However, increasing yields through sustainable intensi- fication are beyond the reach of most smallholder farmers fication is challenging in SSA (Schut and Giller 2020) and (Vanlauwe et al. 2010) and need incentives such as input alternative pathways are often more apparent. For instance, subsidies. In the past 15 years, several fertilizer and seed extensification is currently more common than intensifica- subsidy programs were (re-)initiated by African govern- tion in many regions of SSA (Baudron et al. 2012; Ollen- ments (Jayne and Rashid 2013; Jayne et al. 2018), after their burger et al. 2016). Continued extensification is associated virtual absence during the 1990s and early 2000s (Martin with soil nutrient mining, and this trend could be reversed and Anderson 2008). In addition, social enterprises, such by strongly increasing nutrient inputs (Giller et al. 2021). as One Acre Fund (www. oneac refund. org), provide inputs However, this is constrained by widespread poverty traps though credit schemes to smallholder farmers. Increased (Tittonell and Giller 2013; Koning 2017) and the relatively input use, however, also requires new knowledge (Jayne low economic benefits of staple crop intensification in prac- et al. 2019; Jayne and Sanchez 2021). In a large-scale sub- tice (Bonilla-Cedrez et  al. 2021). Indeed, current trends sidy scheme in Malawi, the limited extension provided by show an increase in the area under maize cultivation in SSA the government was seen as a possible cause for N use effi- (van Loon et al. 2019; Santpoort 2020), which historically ciencies to remain low (Dorward et al. 2008). In addition, has been linked to an increasing population, increasing food fertilizers can be scarce and farmers may mistrust their requirements and urbanization (Smale and Jayne 2003), and quality (Michelson et al. 2021). Co-learning, an iterative hence increasing land pressure (Crowley and Carter 2000). learning framework involving farmers and researchers or Although specialization towards maize favors the production extension workers, has proven to be successful in develop- of sufficient energy, diversified cropping systems would be ing contextualized knowledge (Descheemaeker et al. 2019). more sustainable in terms of income, nutrition, crop yields, We developed an integrated co-learning approach (Mari- and risk spreading (Vanlauwe et al. 2019). Hence, identifica- nus et al. 2021), which aimed to sustainably increase farm tion of constraints and opportunities is essential to support level production by fostering increased input use through desired pathways such as diversification and intensification. the provision of a voucher, in combination with knowledge Setting sustainable intensification as an overall goal for co-creation (Fig. 1). In this paper, we apply a multi-criteria smallholder farming systems results in multiple subsidiary assessment over five seasons to analyze the outcomes of goals, e.g., increased yields, desired N use efficiencies, and food self-sufficiency at household and national level. Attain- ing all goals simultaneously is virtually impossible as trade- offs exist (Klapwijk et al. 2014; Vanlauwe and Dobermann 2020). Moreover, farmers follow their own objectives and prioritize some goals over others. Some goals also require time before they can be attained (Vanlauwe et al. 2010) and outcomes may differ between seasons, requiring assess- ment over multiple seasons, which is rarely done (Smith et  al. 2017). Measuring progress towards the multiple goals of sustainable intensification requires a multi-criteria assessment of indicators associated with the principles of sustainability. Using a framework of principles and criteria warrants transparency and a justified selection of indicators (Florin et al. 2012). According to Florin et al. (2012, p.109), “Principles are the overarching (‘universal’) attributes of a system. Criteria are the rules that govern judgement on outcomes from the system and indicators are variables that assess or measure compliance with criteria.” Criteria can Fig. 1 A farmer who took part in the integrated co-learning approach also help to decide upon benchmarks to judge whether a goal explains how she has used a new type of maize spacing to ensure increased light availability for the intercropped groundnut. Maize is reached (Schut et al. 2014). Within sustainable intensic fi a - grew more vigorously due to increased fertilizer use as part of her tion of smallholder farming systems, criteria, indicators, and intensification strategy. Moreover, by learning about maize-legume benchmarks need to address the field, farm, and household spacing options and new groundnut varieties she was able to increase level. At national level, increasing yields to a certain thresh- the area of groundnut on her farm and thereby to also diversify her cropping system. Photographed by Wytze Marinus. old is required to attain food self-sufficiency, while at farm 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 3 of 19 40 −2 a co-learning program in relation to different agricultural Busia is less densely populated with 530 people km , and development pathways. somewhat larger farms of about 1.0 ha (Jaetzold et al. 2005; Our overarching aim was to improve the understanding KNBS 2019). Both locations receive a rainfall of 1800–2000 −1 of farmer responses to input subsidies and new knowledge, mm year and a have a bi-modal rainfall pattern (Jaetzold in order to better support desired agricultural development et al. 2005), with the long-rain (LR) cropping season from pathways in smallholder farming. This materialized in the March until June and the short-rain (SR) cropping season following objectives to (1) assess the effect of co-learning from September until November. Activities started in the supported by a voucher for inputs on farmers’ decisions and SR season of 2016 and continued for five seasons until the management outcomes, by comparing it with a voucher-only SR season of 2018. Vihiga was selected as a location for its approach; (2) analyze the above effects in terms of crite- high population density, which commonly occurs in high- ria and indicators that relate to agricultural development lands areas of East Africa. Busia was selected for its com- pathways; and (3) reflect on the pathways of intensification, parably larger farm sizes than Vihiga, which could lead to extensification, specialization, and/or diversification result- more opportunities for increasing household income from ing from the co-learning and voucher program. farming. In each county, Vihiga and Busia, two sub-locations were selected and in each of these locations 11–12 farmers were 2 Methodology chosen. Farmers in one sub-location formed the co-learning group while a comparison group was formed in the other 2.1 T he integrated co‑learning approach sub-location. The sub-locations were selected to have similar farming systems, yet be sufficiently far apart to avoid spillo- We applied an integrated co-learning approach from August ver effects. All farmers in the co-learning group received a 2016 until July 2018, as described in detail by Marinus voucher and took part in the co-learning activities. Those et al. (2021). The approach combined four complementary in the comparison group received only the input voucher. elements: input vouchers, an iterative learning process, When inputs were added to the voucher based on feedback common grounds for communication, and complementary from the co-learning groups, these were added for the com- knowledge. An input voucher of US$ 100 was provided each parison group as well. A mid-season field monitoring sur - season to 47 farming households which aimed to alleviate vey included a visit by researchers to each field including resource constraints and increase input use. Inputs for maize, fields that were newly added during the program, to record groundnut, soybean, common bean, and sorghum produc- the crops cultivated and the percentage intercropping. The tion and for dairy were made available. Most inputs were farmer was asked about input use, planting dates, and other offered from the first season on, while groundnut and (short crop management practices. Field sizes were measured using duration) common bean seed and Imazapyr-treated maize a hand-held GPS before the start of the program in June seed against striga were added later during the program in 2016. Small fields with sides less than 20 m were measured response to feedback from the co-learning farmers. The feed- by hand. Yield measurements were done in two 4 × 4 m back was central to an iterative learning process in which a (16 m ) quadrats in all fields containing maize, groundnut, co-learning workshop prior to each cropping season played soybean, and/or common bean. These crops together made a pivotal role. The focus of the workshops evolved over time up about 60–70% of the total cultivated area per farm. Fresh based both on questions and on feedback from farmers dur- cob (maize) and pod (legumes) yields were measured in the ing the season as well as topics identified by the research - field, with one sub-sample per quadrat was taken to deter - ers. Discussion topics during the workshops included the mine dry weight by oven drying. Dry weights were calcu- judicious use of mineral fertilizers and the cultivation of lated back to a standardized moisture content of 14%, and −1 alternative crops such as legumes. Researchers monitored the grain yield (kg ha ) per field was calculated based on the farmers’ responses through a mid-season field survey, the average of the two quadrats. The detailed monitoring yield data collection, and an individual evaluation interview and measurement campaign during five seasons ensured a at the end of each season (see Marinus et al. 2021 for further comprehensive assessment of changes in farm management details). over time. However, the limited number of farmers per sub- location precluded a formal statistical analysis. Addition- 2.2 Research setup ally, we compared the situation during the program with a baseline study from the two seasons before the program. The The integrated co-learning approach was applied in two baseline study was held in the dry season, June 2016, before locations, Vihiga and Busia County in western Kenya. Vih- the start of the program. It used the detailed farm characteri- iga is one of the most densely populated rural areas in SSA zation survey methodology (Giller et al. 2011) to ask many −2 with 1050 people km , with small farm sizes of <0.5 ha. questions relating to the household characteristics and the 1 3 40 Page 4 of 19 W. Marinus et al. production system, including estimates of crop yields and We present the indicator values at the start and the end input use in the previous two seasons. Field sizes for all of the program in a spider web diagram, to assess possible fields in the farm were measured and farmer reported data pathways related to agricultural development. Indicators that was used to derive crop production and input use. During the were identified for intensification and extensification and program, however, crop yields were measured, and farmer- for diversification and specialization, indicated with a * in reported input use was triangulated by comparing field and Table 1, were included in the spider web diagram. In Sec- farm level application. Hence, the accuracy of the baseline tions 2.3.1–2.3.4, we describe the link for each of the indi- study and the detailed monitoring during the program differs cators with their respective pathway. Those indicators were and this needs to be considered in the comparison. scaled using a 0 to 10 score based on specific benchmarks (described in Sections 2.3.1–2.3.4), with a larger score indi- cating a more sustainable situation. Linear interpolation was 2.3 T he indicator framework: principles, criteria, applied to the indicator values to score them between 0 and and indicators 10. We used a multi-criteria assessment to analyze farmers’ 2.3.1 Productivity decisions and management outcomes of the integrated co- −1 learning program. Indicators were selected using principles Reducing yield gaps Maize grain yield (kg ha ) was and criteria (Table 1). We identified four principles of sus- measured in all maize fields, both monocropped and inter - tainable intensification of smallholder systems: productiv - cropped. A farm-level, weighted average maize grain yield ity, food self-sufficiency, environmental protection, and was calculated based on the area of each maize field. The economic viability. For each principle, one to four criteria yield benchmark (score 10) was 50% of the season-specific, and indicators were identified. The yield-related indicators water-limited yield potential in western Kenya, a yield target and food self-sufficiency focused on maize, which was the required to attain national or regional food self-sufficiency most important crop in terms of food and sale with nearly (van Ittersum et al. 2016). The average water-limited yield all households cultivating maize every season. potentials were calculated with a crop growth simulation Table 1 Indicators for agricultural development, organized according eties that were not “Local OPVs” (hybrid varieties, improved varie- to principles (in italics) and criteria. The third column identifies other ties, and improved open-pollinated varieties); PPP purchasing power principles under which an indicator may also fit. Yw water-limited parity. yield potential, AE adult equivalent; improved maize variety: all vari- Principles and criteria Indicators Related to other principle Unit Productivity Reducing yield gaps - Maize yield* % of Yw - Improved maize variety* % maize area −1 - N application rate* Environment kg N ha −1 Food production - Maize production Food self-sufficiency, economic kg produced household viability Food self-sufficiency −1 Food production - Maize self-sufficiency kg produced k g required Environmental protection Avoiding N losses and soil N mining - N use efficiency maize % −1 - N surplus maize kg N ha Ensuring diversification - Crop area: maize* Economic viability % of farm area - Crop area: legumes* Economic viability % of farm area Economic viability −1 −1 Allowing a decent living - Value of produce per crop$PPP AE day −1 - Value of produce per hectare of $PPP ha all crops combined* - Farm area* ha Spreading risk - Legume contribution to the com- % bined value of produce* Indicators for specific pathways 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 5 of 19 40 −1 model (hybrid-maize) using long-term weather data. They on an energy content of maize grain of 3500 kcal kg −1 −1 were 12.5 Mg ha and 8.0 Mg ha for the long- and the (Lukmanji et al. 2008). short-rain cropping seasons respectively (GYGA 2020). −1 The score was set to zero at a maize yield of 0 kg ha . In 2.3.3 Environmental protection addition, the water-limited yield potential of 80% was used as a benchmark for the maximum attainable yield and 15% Nitrogen use efficiency and N surplus Nitrogen (N) use was used as the low baseline found for current yields in efficiency of maize was calculated per season: the total N −1 SSA (van Ittersum et al. 2016). Using these seasonal aver- outputs in maize grain (kg N ha ) divided by the N inputs −1 age yield potentials is a simplification of what is possible in on all fields with maize (kg N ha ). N output was calculated the region, on average, as the water-limited yield potential using the farm-level weighted average maize grain yield and varies from season to season and from farm to farm. This a fixed N content in maize grain of 1.54% (Njoroge 2019). should be considered when evaluating the results against A farm level weighted average for N inputs was calculated the benchmarks. based on the mineral fertilizer used per field, as reported in Yield can be increased by using improved varieties. the monitoring survey. N use efficiency was analyzed using All varieties that were not local open-pollinated varieties the framework developed by the EU Nitrogen Expert Panel (OPVs) were classified as “improved” varieties. These (2015), with a minimum and a maximum N use efficiency include hybrid varieties and improved OPVs. The bench- of 50% and 90% respectively and a maximum N surplus of −1 mark score was 0 at no use of improved varieties and 10 if 80 kg N ha . A N use efficiency below 50% or a N surplus −1 100% of the maize area was sown with improved varieties. above 80 kg N ha indicated a high risk of N losses to the Mineral N application rates on maize were scored at 0 if no environment, while N use efficiencies above 90% indicated N fertilizer was applied and 10 if the mineral N application a high risk of soil mining. The framework also includes a −1 −1 rate on maize was 120 kg N ha or more. The above three general benchmark for a desired output of 80 kg N ha . indicators, associated with reducing the yield gap, were used We adjusted this benchmark to the N output at 50% of the as indicators for the pathway of intensification. water-limited yield potential, equivalent to 83 and 53 kg N −1 ha for the long-rain and the short-rain cropping seasons. Food production Maize is representative of the food pro- duced at farm level and in principle available for home con- Crop area of maize and legumes Assessing area per crop sumption. The total maize production at farm level (kg) was in smallholder farming is not straightforward as crops are calculated from maize yield and maize area for each season. commonly intercropped: e.g., maize is often intercropped with legumes such as common bean or soybean. Cultivated 2.3.2 Food self‑sufficiency area per crop (ha) was calculated as the sum of the areas of all fields containing that crop and was used to calculate Maize self‑sufficiency Maize self-sufficiency was considered yields. The percentage farm area per crop (%) was calculated an indicator for food production, as maize self-sufficiency using the estimated percentage intercropping and the field was reported to be an important production objective by area when comparing percentage areas of different crops. participating farmers (Marinus et  al. 2021). Maize self- When analyzing maize alone, the percentage intercrop- sufficiency may also be a prerequisite before farmers start ping was not considered as, in most common maize-legume to consider other changes in their farm towards sustainable intercropping systems used by farmers in western Kenya, intensification, e.g., diversification into legumes. Maize self- intercropping does not influence maize yield (Ojiem et al. sufficiency at household level (−) was calculated as the total 2014). The percentage farm area covered by maize was an maize production at farm level per season (kg) divided by indicator of specialization and by legumes of diversification. the maize requirements per household per season (kg). The If the percentage maize was above 75% of the farm area, the seasonal maize requirement was calculated from the annual score was 0 and if it was 25% or less it was 10. For legumes, requirement multiplied with the proportional contribution the score was 0 if no legumes were present and 10 if they of seasonal maize production to the annual production. The occupied more than 30% of the farm area. annual household requirements were calculated from the number of adult male equivalents (AMEs) per household 2.3.4 Economic viability and the energy requirements of an active male, 2500 kcal/ day (FAO/WHO/UNU 2001). The number of AMEs per Value of produce Value of produce per crop was calculated household was based on the family composition during the for maize, common bean, groundnut, and soybean based 2018SR, whereby a female was equivalent to 0.82 AME and on the total production per crop per season and the median children (0–18 years) 0.75 AME (FAO/WHO/UNU 2001). crop price for 2018. Median prices were obtained through −1 −1 The maize requirements were 260 kg AME year , based a weekly market survey after pooling the data from both 1 3 40 Page 6 of 19 W. Marinus et al. locations as there were limited differences. Value of pro- of Table 1 and subsequently analyze the different pathways for duce was expressed per adult equivalent per day based on sustainable intensification. the household composition in 2018 and season length. Input costs were not considered as these were largely covered by 3.1 Maize yield and production the voucher. The value of produce therefore paints a rela- tively optimistic picture and does not reflect farm profitabil- Median yields were about 15% of the seasonal-average water- ity. In addition, seasonal and within season price fluctuations limited yield potential before the program (Table  2) and were not considered, as this was not feasible for all crops strongly increased to almost 50% of the seasonal-average and inputs. We used the poverty line for Kenya (World Bank water-limited yield potential for most households from the first 2015) and the living income for rural Kenya (Anker and season of the program onwards. Some farms even reached 80% Anker 2017) as benchmarks. Both were corrected for infla- of the seasonal-average water-limited yield potential in some tion, using 2018 as reference year, which was the same year seasons. Those good yields were maintained during all five as for the crop prices. Both the poverty line and the living seasons of the program (Fig. 2). During the program, farmers income were expressed in $ purchasing power parity ($PPP) planted nearly all of their maize area with improved varieties per adult equivalent per day, following OECD (2011) and (96%) in both locations, while before the program this was Van de Ven et al. (2020). The value of produce per hectare only 46% in Vihiga and 63% in Busia. of all crops combined was expressed per hectare of farm land The maize production per household before the program for each season. It was scored at 0 if the value of produce was about 15% of that during the program, due to both a −1 was 0 $PPP ha . The score of 10 was assigned to the 75% yield increase and the increase in maize area (Table 2). Dur- percentile of the value of produce obtained by all farmers ing the program, the maize area remained relatively large in the short- and the long-rain cropping seasons, so it was a and some farmers even increased it over time (Supplemen- relative score based on the current production values. Value tary materials 2, Fig. 2). This trend was observed irrespec- of produce was considered an indicator for intensification. tive of the initial cultivated area of maize (Supplementary materials 2). Risk spreading Economic viability is improved if risk is spread by growing a variety of crops and not focusing solely 3.2 Maize self‑sufficiency and maize area on maize. We calculated the relative contribution of leg- umes (common bean, groundnut, and soybean) to the com- Maize self-sufficiency before the program in Vihiga was bined value of produce at farm level as an indicator for risk on average one-third of the required amount of maize per spreading. It was scored 0 if legumes did not contribute to household and in Busia this was half. During the program the value of produce and 10 if legumes contributed 50% or most households became maize self-sufficient. On aver - more to the value of produce. The degree of risk spreading age, in Vihiga, households were producing 1.62 times what was considered an indicator of diversification. they needed and in Busia 3.28 times (Fig. 3). Increases in maize area from the second season onwards resulted in an Farm area Farm area often limits the income that can be improvement in maize self-sufficiency for those households attained from farming (Marinus et al. 2022). We assessed in Vihiga which were not yet maize self-sufficient in the first the total farm area per farm based on measured field sizes season. In Busia, larger maize self-sufficiency was associ- of all fields in the farm and monitored this over time during ated with a smaller fraction of the farm area dedicated to the seasonal monitoring survey. Farm area was score 0 if the maize (Fig.  3). These relatively larger farms cultivated a farm area was 0 ha. The score of 10 was assigned to the 75% larger absolute area with maize than smaller farms of less percentile of the farm areas observed for all farmers, so it was than 0.5 ha, who tended to plant maize in most of their fields a relative score based on the current farm areas. An increase (Fig. 4). This critical area of 0.5 ha was roughly what was in farm area was considered an indicator for extensification. needed to produce twice the amount of maize required by typical households, indicating farmers’ priority to attain food self-sufficiency. Maize self-sufficiency and the good market for maize, albeit at low price, were named by farmers 3 Results as reasons to grow maize during the evaluation interviews. There were few differences between the two groups of farm- 3.3 Nitrogen application and nitrogen use ers, the co-learning and the comparison group, except for the efficiency expansion of legumes. Therefore, in the results section, no distinction is made between the two groups of farmers, except Before the program, farmers in Vihiga applied a similar where relevant differences arose. We first assess the indicators rate of mineral N fertilizer on maize as during the program 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 7 of 19 40 Table 2 Average household level indicators per location, before The crop area in % was corrected for intercropping. Crop production (averaged over two seasons), and during the program (averaged over and input use before the program were farmer estimates while field the five seasons). Indicators are grouped according to the pathways sizes were measured. Yields during the program were measured. of intensification/extensification and/or diversification/specialization. Vihiga (n = 23) Busia (n = 24) Before program During program Before program During program Intensification/extensification Farm area (ha) 0.33 0.41 0.76 1.02 Maize area (ha) 0.13 0.23 0.35 0.51 −1 Maize yield (kg ha ) 1513 4426 1260 4541 −1 Maize production (kg household ) 182 1028 367 2342 Maize variety type Local OPV 53 3 34 4 (% cultivated area) Improved 46 96 63 96 Mineral fertilizer application N 94 99 38 54 −1 on maize (kg ha ) P 29 47 19 25 Value of produce crops 48 107 43 113 −1 combined (×1000 Ksh ha ) Diversification/specialization Contribution per crop Maize 68 72 68 74 to combined value of Common bean 25 8 15 9 produce (%) Groundnut 4 8 10 9 Soybean 0 6 1 7 Total legumes 29 22 27 26 Crop area (%) Maize 32 41 36 40 Common bean 11 13 8 12 Groundnut 1 6 3 5 Soybean 0 6 0 6 Total legumes 12 25 11 23 (Table 2). The total amount of N applied on maize however A common choice was to use 60% of the voucher to buy a nearly doubled, but due to the increase in maize area, the rate 50 kg bag of DAP (di-ammonium phosphate) and a 50 kg remained similar. The N application rate in Busia increased bag of CAN (calcium ammonium nitrate), adding up to 23 by nearly 50% during the program as compared to before the kg of N which was the common maximum N use per farm program. P application rates increased in both sites during across the maize fields (Supplementary materials 3). Some the program as compared to before the program (Table 2). farmers with a larger maize area, mainly in Busia, bought There was a clear negative relationship between N small amounts of additional mineral fertilizer with their application rate and maize area in both Vihiga and Busia own money, resulting in moderate fertilizer N application −1 during the program (Fig. 5). High N application rates (> rates of around 50 kg N ha . −1 120 kg N ha ) were applied on farms with a small maize Only few farms across sites and seasons were within area (<0.2 ha) and the rates were largest in the first season the desired range of N use efficiency (white area in Fig.  6). (2016SR). Especially the farmers in Vihiga applied high Too high N use efficiencies (>90%), indicating soil min- rates, which was attributed to their extremely small culti- ing, were found for many of the farms in Busia, during vated areas. With an increased maize area from the second all five seasons, and for about half of the farms in Vihiga season onwards, the N application rates reduced. The other from the second season onwards. Too low N use efficien- −1 seasons showed a similar pattern as 2017LR. Farmers with cies (<50%) and too large N surpluses (>80 kg N ha ) a large maize area tended to distribute the fertilizers over were mainly found in Vihiga (Fig.  6), especially in the the whole area, resulting in lower application rates per first season, where large amounts of N-based fertilizers −1 hectare (40–50 kg N ha ). This relation between N appli- were applied on small maize areas (<0.2 ha). This problem cation rate and farm area seemed partly related to the size reduced from the second season onwards when the culti- of the input voucher, which limited total N use per farm. vated area of maize increased (Fig. 5). 1 3 40 Page 8 of 19 W. Marinus et al. Fig. 2 Total maize production per household in relation to the maize season. The short-dashed line indicates a maize grain yield of 80% of −1 cultivated area per household during the program for Vihiga (A) and the water-limited yield potential, 6400 kg ha for the SR and 10000 −1 Busia (B). The dotted line indicates a maize grain yield of 50% of the kg ha for the LR cropping season. The long-dashed line indicates a −1 seasonal-average water-limited yield potential, 4000 kg ha for the maize grain yield of 15% of the water-limited yield potential, 1200 kg −1 −1 −1 SR (short rains) and 6300 kg ha for the LR (long rains) cropping ha for the SR and 1900 kg ha for the LR cropping season. fraction of the farm area with legumes than larger farms, 3.4 Relative cropping area for maize and legumes but mostly in intercropping with maize. In evaluation inter- views, farmers with larger farms noted labor constraints for Before the program, the relative crop area for both maize cultivating legumes as their main reason for dedicating only and legumes was smaller than during the program (Table 2). a limited area to legumes. In Vihiga, legumes were mainly The share of maize increased by 10 to 25% and the share intercropped with maize. of legumes doubled. However, the area in common bean After increasing in the first seasons, the fraction of farm decreased, the area in groundnut increased, and soybean was area with maize decreased in the last season on the larger newly introduced to 6% of the farm area (Table 2). The frac- farms (2018SR, Supplementary materials 5). The initial tion of the farm area cropped with maize increased in the increases were realized both by replacing other crops (cas- first two seasons, whereas that with legumes increased in sava, sorghum) and by using additional land, e.g., by rent- later seasons. Co-learning farmers planted a larger fraction ing in land and using land that was previously fallow (not of their farm area with groundnut and soybean in the last two shown). Most farmers who decreased their maize area had seasons (2018LR and 2018SR) than the comparison farmers a relatively large maize area. They reported ample maize (Fig. 7), although this seemed to be at the cost of common self-sufficiency and low maize prices as main reasons for the bean. Groundnut and soybean were two focus crops of the decrease. Maize was replaced by groundnut and by leaving co-learning program, for rotational benefits and high value land fallow. of produce per hectare, with specific attention to intercrop- ping arrangements. The difference between comparison and 3.5 Value of crop produce co-learning groups was larger during the long-rain cropping season (Supplementary materials 4), which is locally seen The value of combined crop produce per hectare more as the main season for maize. Some households cultivated than doubled during the program when compared to before legumes mainly during the long rains and others mainly (Table  2). This was the result of yield increases of most during the short rains. Small farms tended to grow a larger 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 9 of 19 40 Fig. 3 Fraction of farm area under maize in relation to maize self- self-sufficient. The fraction of farm area under maize is not corrected sufficiency per season for Vihiga (A) and Busia (B). A maize self- for intercropping. SR stands for short-rain cropping season and LR sufficiency ratio of one (dashed line) means that a household is maize for long-rain cropping season. Fig. 4 Maize-cultivated area per farm in relation to farm area for Vih- intercropping. The dashed line indicates 0.5 ha of maize, above which iga (A) and Busia (B). The dotted line is a 1:1 line, indicating that all no farms cultivate only maize. SR stands for short-rain cropping sea- fields of the farm contain maize. The maize area is not corrected for son and LR for long-rain cropping season. 1 3 40 Page 10 of 19 W. Marinus et al. Fig. 5 Mineral N rate applied to maize fields in relation to the area cropped with maize per farm in 2016SR and 2017LR cropping seasons for Vihiga (A) and Busia (B). The dotted line indicates an application rate of −1 50 kg N ha (common) and −1 the dashed line 120 kg N ha (advised). SR stands for short- rain cropping season and LR for long-rain cropping season. crops. Only yields of mostly intercropped common bean important for co-learning farmers than comparison farmers in decreased during the program, because of the prolific maize the last two seasons (Fig. 8). In particular, groundnut became growth. important, contributing 14% and 8% to total value of produce Maize contributed most to the total value of produce for for co-learning farmers in Vihiga and Busia, respectively, in most households (Fig. 8), because of the large fraction of 2018LR. For comparison farmers, the value of produce of farm area on which it was grown. The contribution of maize legumes was 1% in Vihiga and 0% in Busia in 2018LR due to the total value of produce was more or less the same before to low yields and small areas with soybean. Soybean was and during the program and increased only slightly. As a mainly valued as an option to reduce striga infestation and consequence, the contribution of legumes slightly decreased. less important for its selling value. However, the share of common beans strongly decreased Only one household in Vihiga obtained a value of pro- (low yields, smaller fraction of farm area) and groundnut and duce that was equivalent to the living income in two of the soybean took over (Table 2). For some individual households, seasons (Fig. 8). In Busia, slightly more households in both legumes contributed two to three times more to the total groups obtained a living income, which was mainly related value of produce than maize, because of their larger legume to the larger farm area compared with Vihiga. The total area fraction combined with relatively good legume yields value of produce was equivalent to the poverty line for a (not shown). The expanding area of groundnut (Fig. 7) also few households per group in Vihiga and for about one-third explains why the value of produce of legumes became more of the households in Busia. 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 11 of 19 40 −1 Fig. 6 Farm level N outputs in maize grain in relation to mineral a N surplus of 80 kg N ha , which, if exceeded, indicates a risk of N −1 N inputs on maize, all in kg N ha , for Vihiga (A) and Busia (B). losses to the environment (light yellow-color); (2) a horizontal dashed The figure is based on the EU Nitrogen Expert Panel (2015) analysis line indicating a N output that is equivalent to 50% of the water-lim- −1 method. The upper and the lower diagonal lines with a y-intercept of ited yield  potential per season, 83 kg N ha for the long rains and −1 zero indicate a N use efficiency of 90% and 50% respectively. An N 53 kg N ha for the short rains. Below this output, the  maize grain use efficiency above 90% indicates a risk of soil N mining (deep yel- yield is lower than targeted (pink color). The remaining white area low color), while an N use efficiency below 50% indicates a risk of indicates the desired range of N efficiencies and output. SR stands for N losses to the environment (orange color). The cleat between these short-rain cropping season and LR for long-rain cropping season. two lines is further narrowed by (1) a dotted diagonal line indicating hectare of all crops, maize yield, and the use of improved 3.6 Indications of different agricultural varieties remained the same. N application rate even slightly development pathways decreased. The relative maize area showed a slight speciali- zation towards maize over time during the program, but at Farm area appeared to be an important characteristic for the same time the trends in relative legume area and the explaining the indicator values, especially in Busia. Based legume contribution to the value of produce pointed at diver- on Fig. 4, a cutoff point of 0.5 ha was determined to group sification and spreading of risk. Farm area slightly increased farmers with a smaller farm area (<0.5 ha), denoted “small over time, pointing towards extensification. farms”, and farmers with a larger farm area (>0.5 ha), In Busia the small farms showed a similar pattern: a large denoted “larger farms”, even though these farms are still positive change in intensification only at the start of the pro- very small. Above an area of 0.5 ha, no farmers cultivated gram and a decreasing N application rate during the program maize on all of their land, with one exception in Busia. In due to an increase in maize and total farm area (Fig. 9). The Vihiga, very few farms were larger than 0.5 ha, too few to specialization in maize (low score for maize area) was even consider as a separate category, so we excluded these from more pronounced than in Vihiga and coincided with a slight the analysis. decrease in the relative legume area. However, the contribu- In Vihiga, most intensification happened at the beginning tion of legumes to the value of produce slightly increased, of the program, while hardly any further intensification was pointing at risk spreading through diversification. Similar observed in subsequent years (Fig. 9). This was the case for to the small farms, the larger farms in Busia showed most all indicators related to intensification: value of produce per 1 3 40 Page 12 of 19 W. Marinus et al. Fig. 7 Average percentage of farm area cultivated with leg- umes crops before and during the program for the comparison (A) and co-learning (B) farmers in Vihiga and for the com- parison (C) and co-learning (D) farmers in Busia. The dashed line indicates the start of the program. Percentage areas per crop are corrected for intercrop- ping. SR stands for short-rain cropping season and LR for long-rain cropping season. Fig. 8 Value of produce for soybean, groundnut, common bean, and were ordered each season per location for their value of produce maize in $PPP per adult equivalent per day for each household for of maize. Household IDs were assigned per location. SR stands for the comparison (A) and co-learning (B) farmers in Vihiga and for the short-rain cropping season and LR for long-rain cropping season. comparison (C) and co-learning (D) farmers in Busia. Households 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 13 of 19 40 intensification at the start of the program and hardly any maize and contribution of legumes to the value of produce. further intensification, except for a slight increase in maize Our results are in line with the well-described difficulty of yield. The other indicators for intensification remained the enabling an increase in yields and agricultural production, same during the program. The farms diversified as shown while at the same time fulfilling other environmental and by both the relative maize area and the relative legume area economic goals that are important for sustainable intensi- and a large increase in the contribution of legumes to the fication of smallholder agriculture. value of produce, leading to spreading of risk. Farm area for both groups in Busia slightly increased over time, pointing 4.1 Farmers’ response to the voucher towards extensification. and integrated co‑learning Comparing the larger and the small farms in Busia showed a slightly higher degree of intensification on the The voucher seems to have resulted in changes in input use, larger farms by a larger value of produce per hectare and a yields, maize area, and farm area, independent of the co- higher maize yield during the program (Fig. 9). However, learning workshops (Table 2). Although maize yields and the differences were small. During the program, small farms input use prior to the program were based on farmer-reported were more diversified in terms of legume area and legume data, they were in line with current yields (MoALF 2015) contribution to value of produce, than larger farms. At the and input use (Sheahan et al. 2013; Valbuena et al. 2015) end of the program, however, the contribution of legumes reported in the literature for western Kenya. The measured to the total value of produce was larger for larger farms due maize yields and subsequent increased farm level production to higher yields of legumes, contributing to diversification allowed most households to achieve maize self-sufficiency for risk spreading. during the program. This is most likely due to the provision of the US$ 100 input voucher, as most farms in western Kenya only produce enough maize to feed the household for 4 Discussion half of the year (Valbuena et al. 2015). Although the voucher alleviated capital constraints for agriculture at household In this study, we used a diverse set of indicators to analyze level, co-learning helped to facilitate more complex changes five seasons of detailed farm level data, which was gath- such as diversification into (new) legumes such as soybean ered as part of a co-learning program with 47 farmers in and groundnut (Fig. 7). Although taking time, the iterative western Kenya. We also compared the outcomes during learning process facilitated learning on new intercropping the program (measured) with farmer-reported data, col- arrangements of maize and legumes and identified specific lected during a baseline study held before the program. objectives for soybean (e.g., reducing striga incidence) and −1 We compared the integrated co-learning approach (Mari- groundnut (e.g., high value of produce ha ), as described in nus et al. 2021), which included an input voucher, with more detail in Marinus et al. (2021). Co-learning can thus be a voucher-only approach. We assessed whether the inte- used to contextualize knowledge for the breadth of options grated co-learning approach and/or the input voucher-only that is needed for sustainable intensification (Descheemaeker would lead to pathways of intensification or extensification et al. 2019; Ronner et al. 2021). and pathways of diversification or specialization. We did Initially, all households in our sample, irrespective of not observe a difference between farmers only receiving a farm area, specialized in maize both in Vihiga and in Busia. voucher and those also taking part in the co-learning pro- Larger farms, however, reduced their maize area again in the gram, so we analyzed them as one group. The only excep- last season (2018SR), while small farms maintained their tion was the adoption of legumes, which were included increased maize area. Similar increases in maize area after more substantially by the co-learning farmers. Soybean was the introduction of an input voucher or subsidy have been newly introduced and groundnut substantially expanded, described before for western Kenya (Sanchez et al. 2007) which led to a more diversified cropping system. All farm- and Malawi (Holden and Lunduka 2010; Chibwana et al. ers in our sample increased maize yields (intensification) 2012), based on farmer-reported maize areas. In these stud- compared to the situation before the program, although an ies, however, an increase in maize area often resulted in a increase in farm and maize areas in combination with rela- decrease in legume area (Holden and Lunduka 2010; Chib- tively low N application rates (risk of soil N mining) also wana et al. 2012). For small farms, maintaining the large pointed to extensification and specialization. The value of maize area was associated with farmers’ objectives to be produce remained below a living income for most house- maize self-sufficient (Marinus et al. 2021). holds in our sample due to the small farm areas. This was The financial benefits of diversification into groundnut more prominent in Vihiga than in Busia. The larger farms and soybean that we observed were in line with the find- in our sample scored better in terms of diversification than ings of Franke et al. (2014), who simulated benefits of the small farms, especially related to fraction of the area in diversification with legumes for different farm types in 1 3 40 Page 14 of 19 W. Marinus et al. Fig. 9 Spider web diagrams with average indicator scores per indicator for farms with a relatively small (<0.5 ha) and a larger farm area (>0.5 ha) (data from 2015SR) in Vihiga (A) and Busia (B). For Vihiga, larger farms were left out, as there were too few. A larger score indicates a more sustainable situation. Dotted lines represent the short-rain (SR) cropping season before the start of the program, 2015SR. Dashed lines represent the first season of the program, 2016SR, while solid lines represent the last season of the program, 2018SR. The indicators maize yield, N appli- cation rate, and improved maize variety refer to maize crop level. The other indicators are at farm level. (I) An intensification indicator, (E) extensification indicator, (D) diversity indica- tor, (S) specialization indicator. The “–” sign after maize area indicates a negative relation for this specific indicator: maize area receives a higher score if the cultivated area with maize is smaller. Malawi. Diversification is important for spreading risks diggers for harvesting and shellers (Tsusaka et al. 2017), (crop failure, low prices), and for nutritional and rotational may be required to enable diversification for households benefits (Vanlauwe et al. 2019). On the small farms, leg- with a larger farm area. umes were mainly intercropped with maize resulting in limited benefits due to land constraints. However, on larger 4.2 Concurrent pathways of intensification farms, labor constraints were limiting the expansion of and extensification legumes, similar to the findings of Franke et al. (2014). This would imply that developing and promoting legume- The maize yields obtained during the program point at the specific, small-scale mechanization, such as groundnut pathway of intensification as yields were two to three times 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 15 of 19 40 greater than the yields reported by participating farmers and therefore desirable. Increasing farm area by some farm- before the program, and close to the seasonal-average bench- ers, by renting in land that was ready in active use, meant mark of 50% of the water-limited yield potential. However, that farm area decreased for other farmers. If this would as the corresponding N application rates were both above happen on a larger scale, increasing farm areas could push and below the desirable range, it appeared to be difficult others out of agriculture, requiring alternative employment to enhance N use efficiency, which is a typical challenge for those going out of agriculture (Giller 2020). pertaining to sustainable intensification (Zhang et al. 2015). Except for the first season, the N application rates were Intensified mineral fertilizer use resulted in extremely remarkably similar for the small and larger farmers in our −1 −1 high N application rates in Vihiga (>200 kg N ha in the sample at about 50 kg N ha . This was partly limited by −1 first season, ~100 kg N ha in later seasons) due to the the fixed voucher size of US$ 100. Farmers with a rela- small farm areas in our sample there, resulting in N use effi- tively large maize area, who bought additional fertilizers −1 ciencies below 50%. These farms of less than 0.2 ha were still applied N at a maximum of 50 kg N ha , despite the not able to allocate all inputs from the voucher in a useful advice in the co-learning workshops to apply more. This manner, even with an increased farm or maize area in later may be partly due to the active presence of One Acre Fund seasons. Maize yields were not negatively related to farm in the area who, as a credit provider, advises farmers to use −1 area, and in some seasons even positively related to farm this conservative rate of 50 kg N ha . One Acre Fund was area, while for a small sample, this seems to go against the already present in the program locations before the start of inverse farm size-productivity relationship (Larson et al. the program and did not change the intensity of their activi- 2014). Our finding however is in line with Desiere and Jol- ties during the program. The relatively good yields and low liffe (2018) and Gourlay et al. (2017) who also found no N fertilizer application rates are probably not sustainable negative relation between farm area and yield. Notwithstand- as they will likely result in soil N mining. Soil N mining ing higher N application rates, smaller farms did not seem is common in SSA, but usually at lower yields and lower to produce better yields than larger farms, which may be input levels than in our study (Sheahan and Barrett 2017). explained by reliance on off-farm work requiring farmers’ We diagnosed negative N balances over multiple seasons, attention (Leonardo et al. 2015) and the presence of poorer which suggest that soil mining will occur on the long term. soils (Franke et al. 2019), requiring longer term investments This may have been enabled by the increased application in soil fertility (Vanlauwe et al. 2010). of P through the mineral fertilizers. In the P-fixing soils Extensification was observed on larger farms, who of the study area, P limits mineralization and strong yield increased their farm and/or maize areas and hence distrib- responses to P can be found (Kihara and Njoroge 2013). uted N over larger areas. This was most notable in Busia, as Another reason may be that we did not account for N inputs also discussed in Marinus et al. (2023), where population from manure and N -fixation, although these were small −1 pressure is lower and fallow land is available. We hypoth- (<14 kg N ha for manure on average, with large varia- esize that the expansion in farm area was enabled by the tion in rates due to likely recall error). When good yields in voucher (Marinus et al. 2023). The preference of the farmers combination with soil N mining are continued, N and other in our sample for extensification over intensification goes nutrients (e.g., K) may become limiting (Njoroge 2019) and against one of the key objectives of sustainable intensifica- fertilizer rates will need to be adjusted. tion, namely to increase agricultural production on existing farmland (Cassman et al. 2003; Struik and Kuyper 2017). 4.3 Development pathways evaluated The preference for land expansion among African small- by multi‑criteria analysis holders to increase production, however, seems to be a gen- eral trend for crop area (Baudron et al. 2012; Ollenburger We combined indicators that farmers indicated to be et al. 2016; Jayne and Sanchez 2021; Giller et al. 2021). At important from their perspective (e.g., maize self-suffi - farm level, however, extensification may be less expensive ciency, value of produce) with indicators that are impor- than increasing input rates with the associated larger risks tant for local or national food self-sufficiency (e.g., yield of financial losses (Tittonell et al. 2007; Burke et al. 2019; and production) and environmental protection (e.g., N use Jindo et al. 2020), which can help to explain the observed efficiency, N surplus) in an integrated assessment. This farmers’ preference. The additional fields that farmers rented analysis, in combination with the discussions with the co- in were either previously fallow, or already in active use for learning farmers, identified potential constraints and trade- agriculture. Expanding into fallow land or nature areas, on offs at farm level. Achieving and even surpassing maize the one hand, can result in environmental concerns as it can self-sufficiency was a first priority for farmers, because of jeopardize current ecosystem services such as providing nat- the importance of having surplus food as a buffer for later ural habitats or erosion control. On the other hand, using fal- seasons and the reliable market for maize (Marinus et al. low land more frequently, can also be seen as intensification 2021). This priority could stimulate specialization. The 1 3 40 Page 16 of 19 W. Marinus et al. limited observed diversification goes against a common on crop area, can be subjective and thereby require trans- assumption in modeling studies that farmers are likely to parency on why they were chosen and which benchmarks diversify into other crops once they are maize self-suffi- were used (Marinus et al. 2018). cient (e.g., Hengsdijk et al. 2014; Leonardo et al. 2018). Increasing the value of produce obtained from farming was a second objective for farmers. However, despite the 5 Conclusions good yields, farm area and in Busia also, labor availabil- We analyzed whether farmer responses to a voucher and ity seemed to be overriding constraints for reaching the income benchmarks. At best, one-third of the households co-learning were indicative of different pathways for agricultural development over a period of five seasons by in our sample obtained a living income and half of the households reached the poverty line in Busia. In Vihiga, applying an indicator framework. Our overarching aim was to improve the understanding of farmer responses to input only one out of twenty-three households obtained a liv- ing income in two seasons, while at best one-fourth of subsidies and new knowledge, in order to better support desired agricultural development pathways in smallholder the sample reached the poverty line. Increasing farm area per household or extensification may thus be needed to farming. Although we focused on a limited number of farmers, 47 in total, we believe that based on the detailed increase income from farming to a living income. New employment opportunities will then be needed for those data collection over multiple seasons, some conclusions can be drawn. The novel integrated co-learning approach who choose to leave farming (Giller 2020), if no additional land is available or if an increase of agricultural land is not which we developed facilitated more complex changes in farm management, such as diversification through an desired (e.g., Godfray et al. 2010; The Montpellier Panel 2013). Following area expansion, in Busia, mechanization increase in legume area and legume contribution to the value of produce. Other responses were mainly related to could alleviate the labor constraints for cultivation of prof- itable crops such as legumes, of which the further expan- the input voucher itself. Increased input use through the voucher seemed to sion during the program seemed to be limited by labor constraints. Mechanization could thereby facilitate further increase yields and production, indicating a pathway of intensification that allowed households to achieve maize diversification into more profitable crops for economic viability. Apart from changing to more profitable crops self-sufficiency. As a result of increases in maize and farm area on larger farms, N application rates remained and increasing farm area, selling products at times of high prices can also be a way to increase income. This strategy constant, despite larger inputs. Accompanied by too low N use efficiencies, this pointed at extensification and a is not within reach of all farmers, as it depends on their short-term needs for cash. We did not consider seasonal risk of not reaching environmental protection objectives. Most small farms were only just maize self-sufficient and price fluctuations, although maize prices can differ more than a factor two between the scarce lean season and just their value of produce remained below the poverty line. Obtaining a living income was only possible on large after harvest, when maize is abundant (Burke et al. 2017). A more extensive analysis of each individual household farm areas. It should be noted, however, that we based this on the product prices of 2018. Different prices would give and price fluctuations would be required to assess this. Disaggregating the analysis per household showed a somewhat different picture, but it is clear that prices would have to increase several-fold to lift the majority that farm area limited outcomes for both small and larger farms in specific ways. For example, N use efficiencies of the farmers out of poverty. Our multi-criteria analysis highlighted the difficulty of supporting diversification as were below (for about one-third of farms in Vihiga) or above the desired range (for about half of the farms in a pathway towards sustainable intensification. Improving livelihoods requires changes that go far beyond the farm both sites), respectively, while outputs (yield) were in the desired range. Another methodological lesson learned level. Smallholder farmers in western Kenya and in many rural areas of sub-Saharan Africa are essentially part- is that assessing adoption of new crops or varieties in programs on diversification needs multi-season studies time farmers who depend on many sources of income. To increase income from farming, farm areas need to (Glover et al. 2019) as our results showed that the leg- ume area per farm differed per season and not necessar- increase, which requires off-farm employment opportu- nities for those who choose to leave farming. Whether ily according to the season when legumes were known to be most commonly cultivated. Finally, the principles, sustainable intensification of smallholder agriculture will actually happen may therefore depend on how changes criteria, and indicator framework, following Florin et al. (2012), was useful in being explicit on the underlying in farm structure—that is, capital, land and labor—are facilitated at farm level and as part of the wider socio- assumptions, i.e., criteria, on when an indicator contrib- utes to sustainability. Some of these assumptions, e.g., economic developments within a country. 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 17 of 19 40 Supplementary Information The online version contains supplemen- Zimbabwe. J Dev Stud 48:393–412. https://doi. or g/10. 1080/ 00220 tary material available at https://doi. or g/10. 1007/ s13593- 023- 00893-w .388. 2011. 587509 Bonilla-Cedrez C, Chamberlin J, Hijmans RJ (2021) Fertilizer and Authors' contributions All authors contributed to the study conception grain prices constrain food production in sub-Saharan Africa. Nat and design. Data collection and analysis were performed by Wytze Food 2:766–772. https:// doi. org/ 10. 1038/ s43016- 021- 00370-1 Marinus. The first draft of the manuscript was written by Wytze Mari- Burke M, Bergquist LF, Miguel E (2017) Selling low and buying high: nus and all authors commented on previous versions of the manuscript. an arbitrage puzzle in Kenyan Villages. https:// web. stanf ord. edu/ All authors read and approved the final manuscript.~mburke/ papers/ Burke_ stora ge_ 2013. pdf Burke WJ, Frossard E, Kabwe S, Jayne TS (2019) Understanding ferti- Funding This work was funded through the CGIAR research programs lizer adoption and effectiveness on maize in Zambia. Food Policy on MAIZE (project number A5014.09.08.01), the HumidTropics, and 86:101721. https:// doi. org/ 10. 1016/j. foodp ol. 2019. 05. 004 by the Plant Production Systems group of Wageningen University. KEG Cassman KG, Dobermann A, Walters DT, Yang H (2003) Meeting thanks the NWO-WOTRO Strategic Partnership NL-CGIAR for the cereal demand while protecting natural resources and improving financial support. environmental quality. Annu Rev Environ Resour 28:315–358. https:// doi. org/ 10. 1146/ annur ev. energy. 28. 040202. 122858 Data availability The datasets generated during and/or analyzed dur- Chibwana C, Fisher M, Shively G (2012) Cropland allocation effects ing the current study are available from the corresponding author on of agricultural input subsidies in Malawi. World Dev 40:124–133. reasonable request.https:// doi. org/ 10. 1016/j. world dev. 2011. 04. 022 Crowley EL, Carter SE (2000) Agrarian change and the changing Code availability The datasets generated during and/or analyzed dur- relationships between toil and soil in Maragoli, Western Kenya ing the current study are available from the corresponding author on (1900–1994). Hum Ecol 28:383–414. https:// doi. org/ 10. 1023/A: reasonable request.10070 05514 841 Descheemaeker KKE, Ronner E, Ollenburger M et al (2019) Which options fit best? Operationalizing the socio-ecological niche Declarations concept. Exp Agric 55:169–190. https://d oi.o rg/1 0.1 017/S 0014 47971 60004 8X Ethics approval Participating farmers were well informed about the Desiere S, Jolliffe D (2018) Land productivity and plot size: is meas- purpose of the study prior to participation and informed consent was urement error driving the inverse relationship? World Bank, obtained from all participants involved in the study. Participants were Washington, DC, USA free withdraw from the study at any moment. Data was collected and Dorward A, Chirwa E, Kelly VA et  al (2008) Evaluation of the stored according to the data management plan of the Plant Production 2006/7 Agricultural Input Subsidy Programme, Malawi. Final Systems group of Wageningen University (https://git. wur .nl/ pps/ PPS_ report. https:// eprin ts. soas. ac. uk/ 16730/1/ March Repor tRevM data_m anage ment/-/r aw/m aster/w ritin g/P PS_D ata_M anage ment.p df). ayNoT rack. pdf Ethical approval for this study was not required according to the check- EU Nitrogen Expert Panel (2015) Nitrogen Use Efficiency (NUE) list of the Social Sciences Ethics Committee of Wageningen University. - an indicator for the utilization of nitrogen in agriculture and food systems. Wageningen University, Alterra, Wageningen, the Consent to participate Verbal informed consent was obtained from all Netherlands. http://www .eunep. com/ wp- conte nt/ uploa ds/ 2017/ 03/ individual participants included in the study. Report- NUE- Indic ator- Nitro gen- Expert- Panel- 18- 12- 2015. pdf FAO/WHO/UNU (2001) Human energy requirements. Report of a joint Consent for publication Not applicable FAO/WHO/UNU expert consultation, Rome Florin MJ, van Ittersum MK, van de Ven GWJ (2012) Selecting the Open Access This article is licensed under a Creative Commons Attri- sharpest tools to explore the food-feed-fuel debate: sustainability bution 4.0 International License, which permits use, sharing, adapta- assessment of family farmers producing food, feed and fuel in tion, distribution and reproduction in any medium or format, as long Brazil. Ecol Indic 20:108–120. https:// doi. org/ 10. 1016/j. ecoli nd. as you give appropriate credit to the original author(s) and the source, 2012. 02. 016 provide a link to the Creative Commons licence, and indicate if changes Franke AC, Baijukya F, Kantengwa S et al (2019) Poor farmers - poor were made. The images or other third party material in this article are yields: socio-economic, soil fertility, and crop management indi- included in the article's Creative Commons licence, unless indicated cators affecting climbing bean productivity in northern Rwanda. otherwise in a credit line to the material. If material is not included in Exp Agric 55:14–34. https:// doi. org/ 10. 1017/ S0014 47971 60000 the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will Franke AC, van den Brand GJ, Giller KE (2014) Which farmers benefit need to obtain permission directly from the copyright holder. To view a most from sustainable intensification? An ex-ante impact assess- copy of this licence, visit http://cr eativ ecommons. or g/licen ses/ b y/4.0/ . ment of expanding grain legume production in Malawi. Eur J Agron 58:28–38. https:// doi. org/ 10. 1016/j. eja. 2014. 04. 002 Giller KE (2020) The food security conundrum of sub-Saharan Africa. Glob Food Sec 26:100431. https:// doi. org/ 10. 1016/j. gfs. 2020. References 100431 Giller KE, Delaune T, Silva JV et al (2021) The future of farming: who will produce our food? Food Secur 13:1073–1099. https://d oi.o rg/ Anker R, Anker M (2017) Living wage report Kenya - with a focus on 10. 1007/ s12571- 021- 01184-6 rural Mount Kenya Area. Global Living Wage Coalition, Series Giller KE, Tittonell P, Rufino MC et al (2011) Communicating com- 1, Report 10. p 51. http://www .iseal allia nce. or g/sites/ def ault/ files/ plexity: integrated assessment of trade-offs concerning soil fertil- Kenya_ Living_ Wage_ Bench mark_ Report. pdf ity management within African farming systems to support inno- Baudron F, Andersson JA, Corbeels M, Giller KE (2012) Failing to vation and development. Agric Syst 104:191–203. https://doi. or g/ yield? Ploughs, conservation agriculture and the problem of 10. 1016/j. agsy. 2010. 07. 002 agricultural intensification: an example from the Zambezi Valley, 1 3 40 Page 18 of 19 W. Marinus et al. Glover D, Sumberg J, Ton G et al (2019) Rethinking technological Leonardo WJ, van de Ven GWJ, Udo H et al (2015) Labour not land change in smallholder agriculture. Outlook Agric 48(3):169– constrains agricultural production and food self-sufficiency in 180. https:// doi. org/ 10. 1177/ 00307 27019 864978 maize-based smallholder farming systems in Mozambique. Food Godfray HCJ, Beddington JR, Crute IR et al (2010) Food security: Secur 7:857–874. https:// doi. org/ 10. 1007/ s12571- 015- 0480-7 the challenge of feeding 9 billion people. Science 327:812–818. Lukmanji Z, Hertzmark E, Mlingi N, Assey V (2008) Tanza- https:// doi. org/ 10. 1126/ scien ce. 11853 83 nia food composition tables. Muhimbili University of Allied Gourlay S, Kilic T, Lobell DB (2019) A new spin on an old debate: Science,Tanzania Food and Nutrition Centre and Havard School Errors in farmer-reported production and their implications for of Public Health, Dar es Salaam Tanzania inverse scale - Productivity relationship in Uganda. J Dev Econ Marinus W, Descheemaeker K, van de Ven GW et al (2023) Narrow- 141:102376. https:// doi. org/ 10. 1016/j. jdeve co. 2019. 102376 ing yield gaps does not guarantee a living income from small- GYGA (2020) Global Yield Gap Atlas - climate zone Kenya, Kaka- holder farming – an empirical study from western Kenya. Plos mega (Code=7-7-01) rainfed maize. http:// www. yield gap. org/. One 15:e0283499. https://d oi.o rg/1 0.1 371/j ourna l.p one.0 28349 9 Accessed 23 Nov 2020 Marinus W, Descheemaeker KKE, van de Ven GWJ et al (2021) “That Hengsdijk H, Franke AC, van Wijk MT, Giller KE (2014) How small is my farm” – an integrated co-learning approach for whole-farm is beautiful? Food self-sufficiency and land gap analysis of small- sustainable intensification in smallholder farming. Agric Syst holders in humid and semi-arid sub Saharan Africa, Report num- 188:103041. https:// doi. org/ 10. 1016/j. agsy. 2020. 103041 ber 562. Plant Research International, part of Wageningen, UR. Marinus W, Ronner E, van de Ven GWJ et al (2018) The devil is in the https:// edepot. wur. nl/ 331203 detail! Sustainability assessment of African smallholder farming. Holden S, Lunduka R (2010) Too poor to be efficient? Impacts of the In: Bell S, Morse S (eds) Sustainability Indicators and Indices. targeted fertilizer subsidy programme in Malawi on farm plot level Routledge, London, pp 453–476 input use, crop choice and land productivity. Noragric Rep. No Marinus W, Thuijsman ES, van Wijk MT et  al (2022) What farm 55. p 55. http:// www. umb. no/ stati sk/ norag ric/ publi catio ns/ repor size sustains a living? Exploring future options to attain a living ts/ 2010_ nor_ rep_ 55. pdf income from smallholder farming in the East African Highlands. Jaetzold R, Schmidt H, Hormetz B, Shisanya C (2005) Farm manage- Front Sustain Food Syst 5:1–15. https:// doi. org/ 10. 3389/ fsufs. ment handbook of Kenya: natural conditions and farm manage-2021. 759105 ment information. In: Subpart A1 Western Province, 2nd edn. Martin W, Anderson K (2008) Agricultural trade reform under the Ministry of Agriculture Kenya and the German agency for Techni- Doha Agenda: some key issues. Aust J Agric Resour Econ 52:1– cal Cooperation (GTZ), Nairobi 16. https:// doi. org/ 10. 1111/j. 1467- 8489. 2008. 00404.x Jayne TS, Mason NM, Burke WJ, Ariga J (2018) Review: taking stock Michelson H, Fairbairn A, Ellison B et al (2021) Misperceived quality: of Africa’s second-generation agricultural input subsidy programs. fertilizer in Tanzania. J Dev Econ 148:102579. https://doi. or g/10. Food Policy 75:1–14. https:// doi. org/ 10. 1016/j. foodp ol. 2018. 01. 1016/j. jdeve co. 2020. 102579 003 MoALF (2015) Economic Review of Agriculture 2015. Ministry of Jayne TS, Rashid S (2013) Input subsidy programs in sub-Saharan Agriculture, Livestock and Fisheries (MoALF), Nairobi, Kenya Africa: a synthesis of recent evidence. Agric Econ 44:547–562. Njoroge SK (2019) Feed the crop, not the soil! Explaining variability https:// doi. org/ 10. 1111/ agec. 12073 in maize yield responses to nutrient applications in smallholder Jayne TS, Sanchez PA (2021) Agricultural productivity must improve farms of western Kenya. PhD thesis, Wageningen University, in sub-Saharan Africa. Science (80-) 372:1045–1047. https://doi. Wageningen. https:// edepot. wur. nl/ 503185 org/ 10. 1126/ scien ce. abf54 13 OECD (2011) What are equivalence scales? OECD, p. 2. https:// Jayne TS, Snapp S, Place F, Sitko N (2019) Sustainable agricultural www. oecd. org/ els/ soc/ OECD- Note- Equiv alenc eScal es. pdf intensification in an era of rural transformation in Africa. Glob Ojiem JO, Franke AC, Vanlauwe B et al (2014) Benefits of legume– Food Sec 20:105–113. https:// doi. org/ 10. 1016/j. gfs. 2019. 01. 008 maize rotations: assessing the impact of diversity on the produc- Jindo K, Schut AGT, Langeveld JWA (2020) Sustainable intensifica- tivity of smallholders in Western Kenya. F Crop Res 168:75–85. tion in Western Kenya: who will benefit? Agric Syst 182:102831. https:// doi. org/ 10. 1016/j. fcr. 2014. 08. 004 https:// doi. org/ 10. 1016/j. agsy. 2020. 102831 Ollenburger MH, Descheemaeker K, Crane TA et al (2016) Waking Kihara J, Njoroge S (2013) Phosphorus agronomic efficiency in maize- the Sleeping Giant: agricultural intensification, extensification based cropping systems: a focus on western Kenya. F Crop Res or stagnation in Mali’s Guinea Savannah. Agric Syst 148:58–70. 150:1–8. https:// doi. org/ 10. 1016/j. fcr. 2013. 05. 025https:// doi. org/ 10. 1016/j. agsy. 2016. 07. 003 Klapwijk C, van Wijk M, Rosenstock T et al (2014) Analysis of trade- Pretty J, Toulmin C, Williams S (2011) Sustainable intensification offs in agricultural systems: current status and way forward. Curr in African agriculture. Int J Agric Sustain 9:5–24. https:// doi. Opin Environ Sustain 6:110–115. https://doi. or g/10. 1016/j. cosus t. org/ 10. 3763/ ijas. 2010. 0583 2013. 11. 012 Ronner E, Sumberg J, Glover D, et  al (2021) Basket of options: KNBS (2019) 2019 Kenya population and housing census volume 1: unpacking the concept. Outlook Agric 003072702110194. population by county and sub-county. Kenya National Bureau of https:// doi. org/ 10. 1177/ 00307 27021 10194 27 Statistics, Nairobi Sanchez P, Palm C, Sachs J et al (2007) The African Millennium Koning N (2017) Food security, agricultural policies and economic Villages. Proc Natl Acad Sci 104:16775–16780. https://doi. or g/ growth: long-term dynamics in the past, present and future. Rout-10. 1073/ pnas. 07004 23104 ledge, Abingdon, United Kindom Santpoort R (2020) The drivers of maize area expansion in sub- Larson DF, Otsuka K, Matsumoto T, Kilic T (2014) Should African Saharan Africa. How policies to boost maize production over- rural development strategies depend on smallholder farms? An look the interests of smallholder farmers. Land 9. https:// doi. exploration of the inverse-productivity hypothesis. Agric Econ org/ 10. 3390/ land9 030068 45:355–367. https:// doi. org/ 10. 1111/ agec. 12070 Schut AGT, Giller KE (2020) Sustainable intensification of agricul- Leonardo W, van de Ven GWJ, Kanellopoulos A, Giller KE (2018) Can ture in Africa. Front Agric Sci Eng. https://doi. or g/10. 15302/J- farming provide a way out of poverty for smallholder farmers in FASE- 20203 57 central Mozambique? Agric Syst 165:240–251. https://doi. or g/10. Schut M, Cunha Soares N, van de Ven G, Slingerland M (2014) 1016/j. agsy. 2018. 06. 006 Multi-actor governance of sustainable biofuels in developing 1 3 Farmer responses to an input subsidy and co‑learning program: intensification,… Page 19 of 19 40 countries: the case of Mozambique. Energy Policy 65:631–643. Paper Series No 44. International Crops Reearch Institute for the https:// doi. org/ 10. 1016/j. enpol. 2013. 09. 007 Semi-Arid Tropics (ICRISAT), Lilongwe, Malawi. http://oar .icr is Sheahan M, Barrett CB (2017) Ten striking facts about agricultural at. org/ 10232/1/ T_ Tsusa ka_ etal_ ISEDPS_ 44. pdf input use in Sub-Saharan Africa. Food Policy 67:12–25. https:// Valbuena D, Groot JCJ, Mukalama J et al (2015) Improving rural liveli- doi. org/ 10. 1016/j. foodp ol. 2016. 09. 010 hoods as a “moving target”: trajectories of change in smallholder Sheahan M, Black R, Jayne TS (2013) Are Kenyan farmers under- farming systems of Western Kenya. Reg Environ Chang 15:1395– utilizing fertilizer? Implications for input intensification strat- 1407. https:// doi. org/ 10. 1007/ s10113- 014- 0702-0 egies and research. Food Policy 41:39–52. https:// doi. org/ 10. van de Ven GWJ, de Valença A, Marinus W et al (2020) Living income 1016/j. foodp ol. 2013. 04. 008 benchmarking of rural households in low-income countries. Food Smale M, Jayne T (2003) Maize in Eastern and Southern Africa : “ Secur. https:// doi. org/ 10. 1007/ s12571- 020- 01099-8 seeds ” of success in retrospect. EPTD DISCUSSION PAPER van Ittersum MK, van Bussel LGJ, Wolf J et al (2016) Can sub-Saharan NO. 97. International Food Policy Research Institute. Wash- Africa feed itself? Proc Natl Acad Sci 113:14964–14969. https:// ington DC, USA, p 79. https:// ebrary. ifpri. org/ digit al/ api/ colle doi. org/ 10. 1073/ pnas. 16103 59113 ction/ p1573 8coll2/ id/ 77285/ downl oad van Loon MP, Hijbeek R, Berge HFM et al (2019) Impacts of inten- Smith A, Snapp S, Chikowo R et al (2017) Measuring sustainable sifying or expanding cereal cropping in sub‐Saharan Africa on intensification in smallholder agroecosystems: a review. Glob greenhouse gas emissions and food security. Glob Chang Biol Food Sec 12:127–138. https://doi. or g/10. 1016/j. gf s.2016. 11. 002 25:3720–3730. https:// doi. org/ 10. 1111/ gcb. 14783 Struik PC, Kuyper TW (2017) Sustainable intensification in agricul- Vanlauwe B, Bationo A, Chianu J et al (2010) Integrated soil fertility ture: the richer shade of green. A review. Agron Sustain Dev 37. management operational definition and consequences for imple- https:// doi. org/ 10. 1007/ s13593- 017- 0445-7 mentation and dissemination. Outlook Agric 39:17–24 Struik PC, Kuyper TW, Brussaard L, Leeuwis C (2014) Deconstruct- Vanlauwe B, Coyne D, Gockowski J et al (2014) Sustainable intensi- ing and unpacking scientific controversies in intensification and fication and the African smallholder farmer. Curr Opin Environ sustainability: why the tensions in concepts and values? Curr Sustain 8:15–22. https:// doi. org/ 10. 1016/j. cosust. 2014. 06. 001 Opin Environ Sustain 8:80–88. https://doi. or g/10. 1016/j. cosus t. Vanlauwe B, Dobermann A (2020) Sustainable intensification of agri- 2014. 10. 002 culture in sub-Saharan Africa: first things first. Front Agric Sci The Montpellier Panel (2013) Sustainable intensification: a new Par - Eng 7:376. https:// doi. org/ 10. 15302/J- FASE- 20203 51 adigm for African Agriculture. Agriculture for Impact, London Vanlauwe B, Hungria M, Kanampiu F, Giller KE (2019) The role of leg- Tittonell P, van Wijk MT, Rufino MC et al (2007) Analysing trade- umes in the sustainable intensification of African smallholder agri- offs in resource and labour allocation by smallholder farmers culture : lessons learnt and challenges for the future. Agric Ecosyst using inverse modelling techniques: a case-study from Kaka- Environ 284:106583. https:// doi. org/ 10. 1016/j. agee. 2019. 106583 mega district, western Kenya. Agric Syst 95:76–95. https://doi. World Bank (2015) Purchasing power parities and the real size of world org/ 10. 1016/j. agsy. 2007. 04. 002 economies: a Comprehensive Report of the 2011 International Tittonell PA, Giller KE (2013) When yield gaps are poverty traps: Comparison Program. The World Bank, Washington, DC the paradigm of ecological intensification in African small- Zhang X, Davidson EA, Mauzerall DL et al (2015) Managing nitrogen holder agriculture. F Crop Res 143:76–90. https:// doi. org/ 10. for sustainable development. Nature 1–9. https://doi. or g/10. 1038/ 1016/j. fcr. 2012. 10. 007natur e15743 Tsusaka TW, Twanje GH, Msere HW et al (2017) Ex-ante assessment of adoption of small-scale post-harvest mechanization: the case Publisher's note Springer Nature remains neutral with regard to of groundnut producers in Malawi. Socioeconomics Discussion jurisdictional claims in published maps and institutional affiliations. 1 3

Journal

Agronomy for Sustainable DevelopmentSpringer Journals

Published: Jun 1, 2023

Keywords: Sustainable intensification; Smallholder farmers; Sub-Saharan Africa; Multi-criteria assessment; Indicators; Pathways; Yield; N use efficiency; Living income; Subsidies

There are no references for this article.