The effects of a private-sector-driven smallholder support programme on productivity, market participation and food and nutrition security: Evidence of a nucleus-outgrower scheme from Zambia

Sakketa, Tekalign Gutu; Herrmann, Raoul; Nkonde, Chewe; Lukonde, Mwelwa; Brüntrup, Michael Working Paper The effects of a private-sector-driven smallholder support programme on productivity, market participation and food and nutrition security: Evidence of a nucleus-outgrower scheme from Zambia IDOS Discussion Paper, No. 19/2022 Provided in Cooperation with: German Institute of Development and Sustainability (IDOS), Bonn Suggested Citation: Sakketa, Tekalign Gutu; Herrmann, Raoul; Nkonde, Chewe; Lukonde, Mwelwa; Brüntrup, Michael (2022) : The effects of a private-sector-driven smallholder support programme on productivity, market participation and food and nutrition security: Evidence of a nucleus-outgrower scheme from Zambia, IDOS Discussion Paper, No. 19/2022, ISBN 978-3-96021-198-3, German Institute of Development and Sustainability (IDOS), Bonn, https://doi.org/10.23661/idp19.2022 This Version is available at: https://hdl.handle.net/10419/267728 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ The Effects of a Private-Sector-Driven Smallholder Support Programme on Productivity, Market Participation and Food and Nutrition Security Evidence of a Nucleus-Outgrower Scheme from Zambia Tekalign Gutu Sakketa Raoul Herrmann Chewe Nkonde Mwelwa Lukonde Michael Brüntrup IDOS DISCUSSION PAPER 19/2022 The effects of a private-sector-driven smallholder support programme on productivity, market participation and food and nutrition security Evidence of a nucleus-outgrower scheme from Zambia Tekalign Gutu Sakketa Raoul Herrmann Chewe Nkonde Mwelwa Lukonde Michael Brüntrup Bonn 2022 Dr Tekalign Gutu Sakketa is a researcher in the “Transformation of Economic and Social Systems” programme at the German Institute of Development and Sustainability (IDOS). Email: [email protected] Dr Raoul Herrmann is an associate researcher in the “Transformation of Economic and Social Systems” programme at the German Institute of Development and Sustainability (IDOS). Email: [email protected] Dr Chewe Nkonde is a lecturer and researcher in the Department of Agricultural Economics and Extension, University of Zambia. Email: [email protected] Mwelwa Lukonde is a lecturer and researcher in the Department of Food Science and Nutrition, University of Zambia Email: lukonde.mwelw[email protected] Dr Michael Brüntrup is a senior researcher in the “Transformation of Economic and Social Systems” programme at the German Institute of Development and Sustainability (IDOS). Email: [email protected] Published with financial support from the Federal Ministry for Economic Cooperation and Development (BMZ). Suggested citation: Sakketa, T. G., Herrmann, R., Nkonde, C., Lukonde, M., & Brüntrup, M. (2022). The effects of a private-sectordriven smallholder support programme on productivity, market participation and food and nutrition security: Evidence of a nucleus-outgrower scheme from Zambia (IDOS Discussion Paper 19/2022). Bonn: German Institute of Development and Sustainability (IDOS). https://doi.org/10.23661/idp19.2022 Disclaimer: The views expressed in this paper are those of the author(s) and do not necessarily reflect the views or policies of the German Institute of Development and Sustainability (IDOS). Except otherwise noted, this publication is licensed under Creative Commons Attribution (CC BY 4.0). You are free to copy, communicate and adapt this work, as long as you attribute the German Institute of Development and Sustainability (IDOS) gGmbH and the author(s). IDOS Discussion Paper / German Institute of Development and Sustainability (IDOS) gGmbH ISSN 2751-4439 (Print) ISSN 2751-4447 (Online) ISBN 978-3-96021-198-3 (Print) DOI: https://doi.org/10.23661/idp19.2022 © German Institute of Development and Sustainability (IDOS) gGmbH Tulpenfeld 6, 53113 Bonn Email: [email protected] http://www.idos-research.de Printed on eco-friendly, certified paper. IDOS Discussion Paper 19/2022 III Acknowledgements This discussion paper was funded by the Federal Ministry for Economic Cooperation and Development (BMZ) through two different IDOS research projects: “One World – No Hunger (SEWOH)” and “Social Cohesion in Africa”. The authors are very grateful to Francesco Burchi, Tilman Altenburg and Srinivasa Reddy Srigiri for their critical and constructive comments and suggestions. The authors also benefitted from comments provided by experts from Amatheon Agri Zambia Limited. Furthermore, the support provided by the University of Zambia for the survey work conducted in 2018 and 2021 is greatly appreciated. Moreover, the authors would like to express their gratitude to their interviewees and survey respondents for taking the time to share their valuable insights. The views and opinions expressed in this paper, as well as any errors and omissions, are those of the authors. Bonn, December 2022 IDOS Discussion Paper 19/2022 IV Abstract Nucleus-outgrower schemes (NOSs) are supposed to be a particularly effective private-sector mechanism to support smallholder farmers and contribute towards mitigating the problematic aspects of pure large-scale agricultural investments. This discussion paper uses panel household survey data collected in two rounds in Zambia to analyse some agro-ecological and socio-economic impacts of the outgrower programme of one of the largest agricultural investments in Zambia: Amatheon Agri Zambia (AAZ) Limited. The descriptive results show that the type of participation in the programme varies across participants and components, with most participating in trainings. Econometric results suggest the following key findings. First, although the overall impact of the AAZ outgrower programme on the uptake of conservation agriculture practices is robust and promising, impacts on the adoption of other agricultural technologies is less obvious and the effect depends on the type of support provided. Second, the programme has had a significant impact on maize productivity promoted in the initial phase but not on the other crops – mainly oilseeds – promoted later. Third, the initially less productive farmers seem to benefit slightly more than already better performing ones. Fourth, although the impact on overall household security was insignificant, there is some suggestive evidence (although the effect is weak) that the programme has a positive effect on improving women’s uptake of micronutrients. Finally, our findings show that the three components of the programme (trainings, seed loans and output purchases) have different effects on the adoption of sustainable agricultural practices and productivity, and to some extent on food security. Overall, the results suggest that NOSs, with all their risks, can play a role in the adoption of sustainable agricultural practices, improving farm-level agricultural technologies, providing input credit, and thereby improving productivity and smallholder livelihoods. However, this is not automatically the case, as it crucially depends on the design and management of the project; the availability of good policies and institutions governing the rules of operation; the types of crops promoted; the duration of the project; and the political commitment of host countries, among others. IDOS Discussion Paper 19/2022 V Contents Acknowledgements III Abstract IV Abbreviations VIII Executive summary 1 1 Introduction 6 2 Amatheon Agri Zambia and the outgrower programme 9 3 Framework: Potential economic effects of the AAZ outgrower programme 11 4 Data and empirical approach 15 4.1 Sample design and data collection 15 4.2 Outcome variables of interest 22 4.3 Empirical approach 25 5 Results 27 5.1 Descriptive results 27 5.1.1 Household characteristics 27 5.1.2 Primary (intermediate) outcomes of interest 30 5.1.3 Secondary outcomes of interest (final outcomes) 35 5.2 Empirical results 40 5.2.1 Effects on primary outcomes 40 5.2.2 Effects on secondary outcomes 42 5.2.3 Individual treatment components: Intervention types matter? 47 5.2.4 Heterogeneity impact of the programme: Quantile DID 50 5.2.5 Robustness checks 51 6 Conclusions and policy implications 53 References 57 Appendices 62 Appendix A 62 Appendix B 65 IDOS Discussion Paper 19/2022 VI Tables Table 1: Difference in characteristics of attrited vs panel households (HHs) at the time of the baseline 16 Table 2: Summary of the distribution of participants by districts, balanced panel 21 Table 3: Outcome indicators 24 Table 4: Mean difference between outgrower participants and non-participants 29 Table 5: Adoption of improved agricultural technologies by treatment categories and survey round 32 Table 6: Adoption of sustainable land management practices by treatment indicator 34 Table 7: Adoption of tillage power sources and tillage methods by participation status 35 Table 8: Mean differences for secondary outcomes of interest between the baseline and the follow-up 38 Table 9: Mean differences for food security indicators by participation status and survey round 39 Table 10: Effect of AAZ participation on the adoption of technology components, PSM-DID estimates 41 Table 11: The effects of AAZ participation on the adoption of SLM practices, PSM-DID estimates 42 Table 12: Effect of AAZ participation on productivity, profits and crop diversification, PSM-DID estimates 44 Table 13: Impact of AAZ participation on crop commercialisation, PSM-DID estimates 45 Table 14: Effect of AAZ programme participation on household and women’s food and nutrition security, PSM-DID estimates 46 Table 15: Effect of AAZ participation on the production of other crops, PSM-DID 47 Table 16: Impact of AAZ interventions on selected outcomes of interest, RE estimates 49 Table 17: The quantile treatment effect on the treated (QTT) estimates of the effect of AAZ participation on productivity 51 Table 18: The effects of AAZ programme participation on outcomes of interest, using robust treatment indicator 52 Figures Figure 1: AAZ impact framework 14 Figure 2: Timeline of project intervention and survey implementation 20 Figure 3: Study site – Mumbwa and Chibombo districts 21 IDOS Discussion Paper 19/2022 VII Tables in Appendix A Table A1: Definition of variables used in the regression analyses 62 Table A2: AAZ programme participation dynamics, unbalanced panel HHs 63 Tables in Appendix B Table R1-B: The effects of AAZ programme participation on individual technology adoption: Improved seed and fertiliser use, semiparametric DID estimates 65 Table R2-B: The effect of AAZ participation on the adoption of SLM practices, semiparametric DID estimates 65 Table R3-B: Effects of AAZ participation on productivity, profits and crop diversification, semiparametric DID estimates 66 Table R4-B: AAZ interventions on commercialisation, semiparametric DID estimates 66 Table R5-B: Impact of SLM practices and technology adoption on agricultural productivity: Test of mechanisms, RE estimates 67 Table R6-B: Impact of AAZ components on production of oil crops, RE estimates 67 Table R13B1: Effects of AAZ participation on the adoption of SLM practices, PSM estimates 68 Table R13B2: Effects of AAZ participation on technology adoption, using k-nearest neighbours matching 68 Table R13B3: Effects of AAZ participation on agricultural productivity, using k-nearest neighbours matching 69 Table R13B4: Effects of AAZ participation on commercialisation, using k-nearest neighbours matching 69 Table R13B5: Effects of AAZ participation on food and nutrition security and off-farm employment, using k-nearest neighbours matching 70 Table R13B6: Effects of AAZ participation on the adoption of other crops, using k-nearest neighbours matching 70 IDOS Discussion Paper 19/2022 6 1 Introduction There has been growing foreign commercial interest to invest in agricultural land in developing countries since the mid-2000s (Anseeuw et al., 2013; Deininger et al., 2011). Among other factors, this trend was sparked by the biofuel boom of the mid-2000s (Brüntrup, Anders, Herrmann, Schmitz, & Kaup, 2010; Hirschl et al., 2014) and the 2007/08 food price crisis (Anseeuw et al., 2013; Baumgartner, von Braun, Abebaw, & Müller, 2015). Many countries in sub-Saharan Africa (SSA), where productive agricultural land has been often perceived as abundant and low cost (Deininger et al., 2011; Schoneveld, 2014), have experienced an upsurge in large-scale foreign farmland acquisitions. Zambia has been among the major recipients in SSA. Similar to many other countries in the region, the Zambian government has attempted to attract foreign commercial investments in agriculture by facilitating farmland acquisitions, for instance through the farm block programme targeting 1 million hectares (ha) in all of Zambia’s 10 provinces (Matenga & Hichaambwa, 2017). Large-scale foreign farmland acquisitions have been highly controversial in policy and scholarly debates (Borras & Franco, 2012; Deininger et al., 2011; White, Borras Jr, Hall, Scoones, & Wolford, 2012), raising questions of potentials and risks. On the one hand, it has been often argued that foreign commercial investments in agriculture more generally are likely to create rural employment, transfer of technology and knowledge, and modernise agricultural value chains (Deininger et al., 2011; Deininger & Xia, 2016; Robertson & Pinstrup-Andersen, 2010; Songwe & Deininger, 2009). On the other hand, critics have pointed out that large-scale land acquisitions specifically may displace local populations and increase resource conflicts, jeopardising an inclusive rural development strategy (e.g. Cotula, Vermeulen, Leonard, & Keeley, 2009; De Schutter, 2011; White et al., 2012). The empirical evidence on the effects for SSA, generated with both qualitative (Bluwstein et al., 2018; Bruna, 2019; Engström & Hajdu, 2019; Hall, Scoones, & Tsikata, 2017; Nolte & Subakanya, 2016; Nolte & Väth, 2015; Sulle, 2017) and quantitative methods (Bottazzi, Crespo, Bangura, & Rist, 2018; Deininger & Xia, 2016; Herrmann, 2017; Herrmann & Grote, 2015; Khadjavi, Sipangule, & Thiele, 2020; Lay, Nolte, & Sipangule, 2020; Osabuohien, Efobi, Herrmann, & Gitau, 2019), so far suggests that potential effects for local communities are very context-specific and influenced by how the investment is implemented. By integrating outgrower schemes in their supply chains, large-scale commercial investments can provide market opportunities for small-scale farmers and address production constraints regarding access to national and international output markets, high-quality inputs, information, credits and/or technology (Barrett et al., 2012; Biggeri, Burchi, Ciani, & Herrmann, 2018; Deininger et al., 2011; Deininger & Xia, 2016; Herrmann & Grote, 2015). Others have been less optimistic and stress the potential negative effects from the exploitation of farmers due to monopsonic market structures (Little & Watts, 1994; Sivramkrishna & Jyotishi, 2008) or negative distributional effects, for example when landand resource-poor farmers are excluded (Goldsmith, 1985; Key & Runsten, 1999; White et al., 2012). To attenuate critics, but also for other reasons such as efficiency gains, cost reduction, search for political allies or simply lack of sufficient land, some large-scale investments have engaged smallholder farmers as suppliers (outgrower or more loose cooperation agreements). Such arrangements that aim at increasing the market participation of smallholder farmers and often try to increase their productivity are often seen as win-win-win solutions for investors, smallholder farmers and rural development. The impacts of such programmes, however, are complex and still not well understood, not least because the external and internal conditions of cooperation vary from case to case. Moreover, the empirical evidence on the impact of outgrower models for staple crops in SSA is still IDOS Discussion Paper 19/2022 7 relatively limited and offers mixed results (Bellemare & Novak, 2017; Herrmann, Jumbe, Brüntrup, & Osabuohien, 2018; Maertens & Vande Velde, 2017; Negash & Swinnen, 2013; Ragasa, Lambrecht, & Kufoalor, 2018). This discussion paper attempts to contribute towards filling this knowledge gap by analysing the ex-post effects of a foreign (private) commercial investment in farmland in Zambia – the Amatheon Agri Zambia (AAZ) “outgrower” programme – on various welfare outcomes. Specifically, the paper compares households participating in at least one of the AAZ outgrower activities with non-participants in terms of adoption of agricultural technology (seed and fertiliser), CA practices, crop productivity, food and nutrition security (of women of reproductive age), and commercialisation. To expand its trading volume and as part of its corporate social responsibility, AAZ has developed an outgrower programme with farmers, first in Mumbwa district in 2013/14 and then expanded to Chibombo district in 2017. The first phase of AAZ’s outgrower programme (season 2013/14 to season 2015/16) was comprised of three components: i) free trainings on conservation agriculture (CA) and sustainable intensification, input use, post-harvest crop handling, marketing and business skills; ii) input support in the form of seed loans, specifically improved maize and soybean seeds; iii) purchase of maize and soybeans without prior contracts with farmers. Establishing trading depots (about 30) to sell inputs and buy grains was also part of the market support. The second phase (season 2016/17 to season 2018/19) of the programme expanded its activities in phase one (both in terms of geographic coverage and crop focus) to include additional crops for sunflower, cowpeas and groundnuts as part of its input support programme as well as the purchase of these promoted crops. Consequently, the purchase of promoted crops – initially limited to maize and soybeans – was expanded to include sunflower and legumes, supported through input loans. During the second phase, the programme also incorporated different public–private and non-governmental organisation (NGO)–private partnership arrangements. For instance, the programme received support from the United States Agency for International Development (USAID), the German Investment Corporation (DEG) and Musika (a Zambian non-profit company involved in the promotion of agricultural markets) to expand the construction of depots in Chibombo district. After 2019 (the third phase), AAZ re-designed its outgrower programme to focus on high-value export crops such as quinoa and chia with a contract-farming approach in Mumbwa district, while scaling-down or stopping some of its other activities, specifically the provision of input support and the purchase of cereals. In contrast to other contract-farming schemes, the first and second phases of AAZ outgrower programmes, which are the focus of this paper, did not involve contractual obligations on either AAZ or farmers at the time of the surveys. In fact, although AAZ’s outgrower programme can best be described as a “nucleus-outgrower” scheme (NOS), we prefer to use the term “outgrower” programme to remain consistent with the company’s term.2 In this paper, all farmers who participated in one or more of the components of the programme (training, input support and purchase of grain from farmers) are subsumed under the term “outgrowers”. This discussion paper makes a number of contributions to scholarly and policy debates. First, it contributes to a growing literature on the socio-economic effects of the different business models for large-scale land-based agricultural investments (e.g. Brüntrup et al., 2018; Nolte & Ostermeier, 2017) by shedding light on the potentials and limitations of smallholder outgrower 2 “The nucleus is a large farm unit, in this case AAZ, which guarantees a certain minimum provision of raw material for a large-scale processing plant or other downstream aggregation use, while the other part of the raw material is procured from smallholder farmers who are linked through contractual arrangements to the nucleus” (Brüntrup et al., 2018, p. 1). IDOS Discussion Paper 19/2022 8 programmes to complement an investor’s large-scale commercial farming operation. Second, the paper contributes to the emerging literature on the socio-economic effects of outgrower farming in staple crop sub-sectors of SSA (Maertens & Vande Velde, 2017; Ragasa et al., 2018), specifically to the NOSs (Brüntrup et al., 2018; Herrmann & Grote, 2015). Third, the paper aims to contribute towards the debate on the sustainable intensification of small-scale agricultural systems in SSA through CA (Giller et al., 2015; Rodenburg, Büchi, & Haggar, 2020) and the potential role of the private-sector as a driver for CA adoption (Westengen, Nyanga, Chibamba, Guillen-Royo, & Banik, 2018). Zambia is one of the countries in which CA was promoted early in SSA, yet in spite of more than two decades of promotion, adoption rates in Zambia and other SSA countries have been limited, and its potential for smallholder agriculture is controversially debated (Corbeels, Naudin, Whitbread, Kühne, & Letourmy, 2020; Rodenburg et al., 2020). Lastly, the paper will add to the literature on the role of private-sector outreach programmes in improving the adoption of smallholder farm-level agricultural technologies, productivity and market participation in the developing world. The analysis of this paper relies on a panel dataset from two waves of household surveys covering about 800 farming households conducted in 2018 and in 2021 with farmers who had participated in the first two phases of the outgrower programme and non-participants. To complement the quantitative analyses, qualitative interviews (focus group discussions) were carried out in the six villages – three from each district – where AAZ operates its investment as well as implemented its outgrower programme. The paper uses both descriptive as well as various econometric techniques, such as propensity score matching difference-in-differences (PSM-DID) and semiparametric DID (to account for differences in relevant observable characteristics) in comparing households participating in the AAZ programme with nonparticipant households both before (in an early phase) and after the programme intervention. Furthermore, we ran various robustness checks to examine whether our results are sensitive to various definitions of outgrower participation, estimation techniques and changes in programme focus. Moreover, a quantile treatment effect in DID has been carried out to examine the heterogeneous effects of programme participation at different levels of outcomes of interest. The discussion paper is structured as follows: Section 2 presents background information on AAZ’s commercial investment and the AAZ outgrower programme. Section 3 discusses the theory of change. Section 4 contains all information on the sampling strategy, the methodology for our analysis and a description of the AAZ outgrower programme based on the sample information. Thereafter, Section 5 provides the results of the descriptive and econometric analyses of the outcome comparisons between programme participants and non-participants as of the 2016/17 agricultural season. Section 6 provides conclusions and policy implications relevant for programme design and further research. IDOS Discussion Paper 19/2022 9 2 Amatheon Agri Zambia and the outgrower programme Amatheon Agri Holding N.V. is a German agribusiness company founded in 2011. It operates in SSA, where it established a subsidiary in Zambia in 2012, Amatheon Agri Zambia (Amatheon Agri, 2013), and later subsidiaries in Uganda and Zimbabwe. Amatheon Agri’s headquarters is located in Berlin, Germany. In Zambia, it acquired a 99-year land lease of initially 32,000 ha in Mumbwa district (Amatheon Agri, 2013), which was increased to 38,760 ha in 2014 (Amatheon Agri, 2015). According to the Land Matrix website, the privatised and titled farm block was originally designated in 1973-1974 and previously used for large-scale planting by various owners (Land Matrix Global Observatory – LMGO, 2019). AAZ implemented a system of largescale commercial farming in 2013, focussing on maize, soybean and wheat on both irrigated and rainfed fields. At the end of 2014, the company reported 1,430 ha under cropping as well as approval for an expansion of 4,000 ha (Amatheon Agri, 2015). By 2016, AAZ reported to have 2,988 ha under cropping (Amatheon Agri, 2017). AAZ increasingly expanded its own crop portfolio towards high-value and horticulture products (Amatheon Agri, 2018, 2019), eventually focussing on “natural healthy foods” such as chia seeds and quinoa for export (Amatheon Agri, 2019; personal communication with AAZ). In addition to crop farming, AAZ operates a cattleranching component. Alongside its large-scale farming operations, AAZ started trading with farmers in its project area in 2013 by purchasing crops, selling inputs and offering extension services (Amatheon Agri, 2014). The smallholder engagement started with AAZ initiating a small extension project for surrounding farmers with a Zambian NGO called Musika. In addition, it established a retail/farm shop for farmers to buy maize and soybean inputs, receive advice on input use, and to sell maize and soybean outputs to AAZ (Amatheon Agri, 2015, p. 27). By integrating the trading component into its general business operations (Amatheon Agri, 2014, 2016), AAZ aimed at increasing its trading volume while “achieving significant social impact”, fostering entrepreneurship, stimulating local productivity and diversifying sources of income (Amatheon Agri, 2016, p. 35). AAZ referred to itself as an “anchor investor in rural areas to uplift neighbouring communities economically, socially and environmentally” (Amatheon Agri, 2016, p. 32), with its commercial farm to serve as a “hub” for surrounding communities by providing infrastructure and market access (Amatheon Agri, 2016). Its outgrower programme has undergone significant changes since its start. The extension component aimed at increasing the productivity of small-scale farmers through CA and businessskills training (Amatheon Agri, 2014, 2015). According to the Food and Agriculture Organization (Food and Agriculture Organization [FAO], 2014a), CA is an approach for resource-efficient agricultural production involving three main components: minimum tillage, mulching and crop rotation (CR) with legumes. Training was provided free of charge and also covered the issues of lime, fertiliser and herbicide application, post-harvest handling, marketing and accounting (“farming as a business”). Training was implemented in collaboration with two NGOs: the Conservation Farming Unit (CFU) and World Vision. CFU has implemented CA training in Zambia since the 1990s. Training is based on a lead-farmer-extension or train-the-trainer model, that is, experienced farmers were selected as farmer coordinators (FCs) and trained by CFU and World Vision, after which they trained 60100 farmers in self-organised groups. FCs were not employed by AAZ but received inputs as incentives (personal communication with AAZ staff). By 2016, AAZ reported to have a network of 8,000 farmers who participated in training (Amatheon Agri, 2017, p. 9). IDOS Discussion Paper 19/2022 10 As part of the trading component, AAZ established rural trading depots in close proximity to communities for farmers to buy inputs and sell their crops. Although maize and soybeans comprised the majority of crops traded during the initial phase of the programme, in 2017, the company started promoting cowpeas, groundnuts and sunflower (Amatheon Agri, 2017, p. 17). AAZ had a partnership with the UN World Food Programme to operate as the buyer of smallholder cowpeas to supply them to the World Food Programme for their national school feeding programme (Amatheon Agri, 2017, p. 9), but the partnership had stopped before the follow-up survey was conducted. At some point, AAZ also operated a livestock purchase and service component (Amatheon Agri, 2017). AAZ also negotiated an input financing scheme with Zanaco Bank (Amatheon Agri, 2015, p. 18), with AAZ acting as co-guarantor for the loans in a tripartite agreement (personal communication with AAZ staff). In the loan agreement, farmers were to pay 50 per cent of the loan up front and the remaining amount plus interest after harvest, with loan recipients receiving a crop purchase guarantee from AAZ (personal communication with AAZ staff). About 140 farmers overall participated in the initial loan trial (Amatheon Agri, 2016), expanding to 285 farmers in 2017/18, with AAZ guaranteeing purchasing the harvest (personal communication with AAZ staff). In 2017, AAZ established a partnership with USAID to expand the outgrower programme to Chibombo district and “to support an additional 6,000 farmers with access to inputs, credit, trainings and markets” (Amatheon Agri, 2018, p. 9; personal communication). Along with its expansion, AAZ additionally focussed more on cowpeas, sunflower and groundnuts (Amatheon Agri, 2019, p. 1). To improve seed access for these crops, a seed bank was established (Amatheon Agri, 2018, p. 9; 2019, p. 9). By 2018, the outgrower programme had reached around 15,000 smallholder farmers in Mumbwa and Chibombo districts (Amatheon Agri, 2019, p. 9). Around 2019, AAZ restructured the programme by introducing quinoa to farmers in the vicinity of its farm in Mumbwa, with 1,000 smallholder farmers joining the trial and more than 100 farmers being officially certified by an internationally recognised certification body as organic (quinoa) farmers (Amatheon Agri, 2021). Although AAZ did not have purchase agreements with farmers in the other trading components – with neither side obliged to purchase or sell – AAZ acts as a “guaranteed offtaker” in the quinoa project with guaranteed prices beforehand (Amatheon Agri, 2021; personal communication). In order to develop the quinoa contract farming and with USAID’s funding ending, the company decided to reduce other programme activities, especially in Chibombo, with plans to expand later once the quinoa scheme has been fully established (personal communication). This overview suggests that the Amatheon outgrower programme has been changing over time and is still developing, capturing different projects that focus on trading with and supporting farmers in Mumbwa and Chibombo. During the first years, when operating mainly in Mumbwa (the first phase of the outgrower programme), the focus was on maize and soybean production, mainly involving training and trading activities, with a small credit component. With its expansion to Chibombo (the second phase), AAZ had increasingly integrated higher-value crops, using loan schemes to support access to the seeds for these crops. IDOS Discussion Paper 19/2022 11 3 Framework: Potential economic effects of the AAZ outgrower programme Along its large-scale farming operations, Amatheon uses its commercial farms as a nucleus to support and trade with smallholder farmers. In addition, it uses its newly constructed processing and storage facility to collect produce grown by farmers. Unlike a typical outgrower scheme, which provides inputs to farmers and guarantees the purchase of the entire or part of their harvest, AAZ outgrower programmes envisage incorporating smallholder farmers into the rural value chain, build outgrower networks, foster entrepreneurship, stimulate local productivity and diversify sources of income through three core interventions (channels): technical assistance, provision of input support (seed and fertiliser) and the purchase of the harvested crops (Figure 1). Although some of the interventions may need more years to establish and produce long-term impacts, AAZ interventions affect outcomes through various channels. The first channel is technical assistance. Of the various technical assistance activities, training is the component in which most farmers participate. Although training covers a range of agricultural and non-agricultural topics, including the adequate use of agricultural technologies (e.g. fertiliser and herbicides), the project focusses on CA and related agricultural extension services. CA is a sustainable agricultural-intensification approach and is often defined by its three main practices: (1) minimum mechanical soil disturbance (minimum tillage); (2) retaining sufficient crop residues to permanently cover the soil, for example mulching (residue retention – RR)3; and (3) CR with nitrogen-fixing crops (FAO, 2014b; Kassam, Friedrich, Shaxson, & Pretty, 2009). Although trainings on CA and improved technologies (such as seeds or fertiliser) may enhance farmers’ know-how and knowledge, it might not be enough for adopting CA or any other technologies if there are constraining trade-offs or farmers lack access to capital (Arslan, McCarthy, Lipper, Asfaw, & Cattaneo, 2014; Giller et al., 2015). CA practices may improve yields by increasing soil fertility, reducing erosion, or conserving and improving soil moisture (Arslan et al., 2014; Giller et al., 2015; Ngwira, Thierfelder, Eash, & Lambert, 2013) as well as reducing labour demand (e.g. by stopping weed growth) (Giller et al., 2015). Although some studies have confirmed the positive yield and/or income effects of adopting CA (Ngoma, Mason, & Sitko, 2015; Tambo & Mockshell, 2018), estimated effects vary considerably (Corbeels et al., 2020; Rodenburg et al., 2020). Since experimentation with, and adoption of, new practices and technologies can be a rather longer-term process, it might be more realistic to observe the differences in intermediate outcomes in the short term, such as the adoption of promoted practices and technologies rather than yield effects, which may be expected only in the medium term (Giller, Witter, Corbeels, & Tittonell, 2009). The second channel is the provision of input support (improved seed and fertiliser) in the form of loans and agricultural input credit. The advantage of outgrower farming may not only come from enabling smallholders to access commercial value chains, but also by addressing input market failures, thereby increasing access to inputs, and eventually crop yields and incomes (Barrett et al., 2012). Input market access can be improved, for example, when contracts are used as collateral, when credit schemes are part of the agreements (e.g. tri-partite arrangements with commercial banks) or when earnings are sufficient to purchase inputs (Govereh & Jayne, 2003; Grosh, 1994; Herrmann, 2017). Another advantage for farmers’ investments and input use is the potential risk-reduction effect for farmers of having a guaranteed market. In the case of the AAZ outgrower programme, input supply components may improve input access via 3 It is often stated that at least 30 per cent of the soil has to be covered (Giller et al., 2009). IDOS Discussion Paper 19/2022 12 increased community-level input availability due to the input depots or the loan programme, and because of improved grain and legume market opportunities that may raise the profitability of using external inputs. Having access to the input market might also address technology adoption constraints, for example constraints to access legume seeds, sell legumes and access fertiliser and/or herbicides. CA projects therefore often involve the additional promotion of external inputs such as fertiliser, herbicides and/or improved seeds as part of a full CA package (Arslan et al., 2014; Giller et al., 2009). However, poorer farmers may lack the resources to purchase inputs. Although the programme’s loan component may address financial constraints for resource-poor farmers, the risks associated with taking up credits may prevent them from taking them up, and creditors may also favour larger, more solvent commercial farmers rather than smallholders. The third channel is the output purchases (with or without prior guarantee). Through AAZ’s grain purchases, farmers may improve the sales of crops already part of their portfolio (e.g. maize and soybean) or of the new crops purchased by AAZ (groundnuts, sunflower and/or cowpeas), thereby raising or stabilising revenues and incomes. Especially for crops with markets and buyers that are based in urban centres, proximity to AAZ’s input and output aggregation depots located close to communities help farmers overcome barriers to market access by reducing the distance to markets, which helps to decrease transaction costs. Positive price effects for farmers, however, are likely to be stronger for high-value crops and limited for staple crops, such as maize or soybeans, which are commonly traded on local markets. The fact that AAZ did not have purchase agreements at the time of the survey may allow farmers to sell to the buyer who offers the highest price. But the lack of purchase guarantees can also expose them to more risks, especially when adopting crops that were newly introduced by AAZ to the region. For instance, if contracts are reliable, more stable effects can be observed. A drop in maize prices in 2017, for example, which led AAZ to reduce crop purchases, could have made farmers worse off if they had initially increased production because of AAZ’s presence but could not find alternative markets available. Yet, since the major crops of the first phase of the AAZ outgrower programme in 2016/17 (maize and soybean) are also commonly traded locally, such market risks could be less relevant, as farmers would be more likely to find other buyers. Although the impacts remain to be seen, the introduction of a pre-fixed price off-take contract (with quinoa and chia for export) in the third phase of the programme can be a positive adaptation strategy. Overall, the benefits from participating in the AAZ outgrower programme are likely to be greater when households have access to more or all three components (training, inputs and crop sale). Training is likely to be accessible to most farmers, as it is free of charge and only involves the opportunity costs of time. However, production effects might be lower than when combined with other components. Access to inputs, which could be the most important factor for improving yields, may be limited to better-off households, at least at an initial stage, when the credit system is not yet fully developed. Linking farmers to the export markets, particularly through outgrower schemes, is expected to increase the bargaining power of farmers. The purchase of produce at the farm gate creates market access and decreases transaction costs, and hence can increase income and reduce poverty and food insecurity. Through increased income, production diversification and/or commercialisation, as well as improved access to input and output markets, the project might then improve other dimensions of households’ wellbeing, such as the food and nutrition security of household members and their health. The generated changes in IDOS Discussion Paper 19/2022 13 production and marketing systems as well as interactions between different approaches can be beneficial to society at large, for example by increasing social cohesion. However, how effectively these programmes work depends on several factors: project design and management; availability of good policies and institutions that govern interactions among stakeholders, including clarifying the rules of operation; the types of crops promoted (cereals or oilseeds; staple or cash crops); the duration of the project; the types of support provided; the political commitment by host countries, among others. Figure 1 summarises the potential socioeconomic channels and effects of the AAZ programme. Figure 1: AAZ impact framework Source: Authors IDOS Discussion Paper 19/2022 15 4 Data and empirical approach 4.1 Sample design and data collection In order to assess the impact of interventions, two rounds of survey data were collected in two districts: Mumbwa and Chibombo (see Figure 2 for a timeline of treatment and survey implementation). A baseline survey was conducted in 2018 focussing on the 2016/17 agricultural season. The follow-up survey took place in 2021, with the focus on the 2019/20 main agricultural season. In addition, qualitative interviews (focus group discussions) were carried out in the two districts, which is where AAZ operates its investment as well as implemented its outgrower programme. During the data collection, tablet-based face-to-face interviews with household heads or their spouses were used, using a structured questionnaire. One of the challenges during the data collection was the difficulty in locating the same households interviewed during the baseline. As a result, the attrition rate was remarkably high, since about 206 (about 25 per cent) of the 793 baseline sample households could not be re-interviewed during the follow-up survey. Hence, we ended up with a balanced panel of 590 households that were re-interviewed during the follow-up. The main reasons reported for this high attrition rate were migration (household moved away – about 127 households), non-contact (56) and death (10 households), among others. Although migration was high among treated households (also stated during the focus group discussion), it was evenly distributed between both districts and not necessarily concentrated in Mumbwa district, where AAZ operates its investment. Since such a high attrition rate could bias our estimates, we checked the patterns of attrition in this. Results show that there were no systematic differences between the attrited and balanced panel households in most of the socio-demographic and economic characteristics, except for the size of the land cultivation for maize, soya and agricultural assets as well as kinship ties with a chief or headman (see Table 1). This was a desirable outcome, as this lends credence to our analysis. Furthermore, 100 additional households engaged in quinoa production in 2021 were surveyed during the follow-up, although we excluded them from the main analysis. IDOS Discussion Paper 19/2022 22 4.2 Outcome variables of interest Outcome variables Since the objectives of the outgrower programme were to provide learning opportunities for farmers to diversify their crop portfolios, increase productivity, reduce post-harvest losses and increase average annual household incomes, hence overall food security, we identified two broad categories of outcomes, namely primary and secondary outcomes. Our primary outcomes of interest are adoption of CA/SLM practices and the use of improved agricultural technologies. In line with the literature (e.g. Kunzekweguta, Rich, & Lyne, 2017; Ngwira et al., 2013), we measured CA/SLM uptake by focussing on three main practices: minimum tillage (MT), residue retention (RR) and crop rotation (CR). Drawing on Zambia’s Rural Agricultural Livelihood Survey – which asked farmers whether they had prepared planting basins, performed zero tillage or ripping on a given plot – we defined MT as a dummy variable taking a value of 1 when at least one plot was treated with planting basins (when using hand hoes), zero tillage or ripping (when using mechanisation) (e.g. Ng’ombe, Kalinda, & Tembo, 2017). CFU additionally recommends till during the dry season, which is argued to reduce labour requirements (Arslan et al., 2014). We therefore also collected information on the timing of the main tillage method. RR was measured by asking farmers for each plot about the crop/field residue use during the 2016/17 and 2019/20 agricultural seasons, that is, one year before the reference seasons. Accordingly, farmers who left crop residues on the field (e.g. as mulch) and did not plough it back into the soil nor let livestock graze it were defined as RR adopters. To measure CR with legumes or other nitrogen-fixing crops, farmers were asked whether legumes were planted on a given plot, and in which combination they rotate crops on a given plot (e.g. by rotating cereals with legumes/nitrogen-fixing crops). CR is also a binary outcome. CFU also recommends erosion control practices, such as contour farming (contour bunds, grass hedges, contour ploughing, etc.), as part of the CA package, which we also asked about (Arslan et al., 2014). As to the adoption of agricultural technologies, we focussed on those mainly promoted through the outgrower programme such as improved seeds for maize, soybean, groundnuts, sunflower, cowpeas; and chemical fertiliser use, herbicides, pesticides and weedicides.6 For each technology, a dummy variable taking a value of 1 indicates if the household has used or applied the respective technology during the agricultural reference period; and 0 otherwise. In addition, technology adoption is also measured in terms of use intensity. As side information, we also documented the effect of the programme on land use and acquisition (methods of land acquisition and tenure status) as another result in the primary outcome. Our secondary outcome variables are productivity, profits (crop revenue), household food and nutrition security indicators, all of which are considered to be directly related to household welfare outcomes, as well as commercialisation indices. Productivity is measured as maize yield in kg/ha, soya yield in kg/ha or overall cropland productivity. Since AAZ was engaged in the construction of community trading depots to purchase crops from farmers, such intervention 6 We used two different definitions of improved seeds. The first defined the sources of the seeds, which was improved if the seeds came from outside the community (not own harvest/ recycled seed, no other farmer, nor friends/relatives). The second directly asked for the name of the seed, or named them unimproved if farmers referred to them as local or recycled hybrids. IDOS Discussion Paper 19/2022 23 would potentially affect farmers’ marketing performance. To capture such an effect, we computed commercialisation indices for maize, soya and the overall crop commercialisation. In particular, the construction of depots can make it easier for outgrowers to reach markets, which may in turn affect farm gate prices. Profits from crop production are measured as total crop revenue minus the cost of fertiliser and seeds.7 Subsistence production was valued by market prices. Whenever a household did not report crop sales during the 2016/17 and 2019/20 seasons, median prices in the village, ward or district were used. Food security indicators used in this impact evaluation include the household food insecurity access score, the availability of adequate micronutrients in the household for women of reproductive age and the minimum dietary diversity score for women of reproductive age. As a robustness check, we also considered household per capita consumption expenditure – including the values of home production, purchased commodities and gifts – as another indicator of household welfare. We have excluded observations that had clearly inconsistent production and sales data, and we have trimmed income figures at the 1 per cent and 99 per cent levels to account for unrealistic outliers. Finally, we constructed three commercialisation indices to proxy households’ market participation: maize, soybean and the overall crop commercialisation index. All three indices are computed by dividing kg sold by kg harvested. Hence, it also measures the intensity of market participation. For instance, the maize commercialisation index is computed as maize revenue (the price of maize per kg multiplied by kg sold) divided by the maize value of production (price of maize per kg multiplied by kg of maize harvested). A similar procedure is followed to construct the soybean and overall crop commercialisation index. By construction, the indices are continuous variables. Table 3 summarises the various outcomes of interest considered in our study. 7 The cost of fertiliser is computed using the average price of urea and nitrogen-phosphate plus sulphur per kilogram in each district. IDOS Discussion Paper 19/2022 24 Table 3: Outcome indicators Outcome indicator Description Primary outcomes (outcome related to CA and technology adoption) Minimum tillage (MT) =1 if a household is using MT on at least one plot in the 2016/17 and/or 2018/19 agricultural season Retaining crop residues (RR) =1 if using RR on at least one plot in 2016/17 and/or 2018/19 Crop rotation (CR) =1 if using CR on at least one plot in 2016/17 and 2018/19 At least two CA measures =1 if using at least two CA measures in 2016/17, dummy All three CA measures =1 if using all CA measures in 2016/17 and 2018/19 Improved seeds Seed is not local nor recycled hybrid, dummy External seeds =1 if seeds are not from own harvest/recycled seed, no other farmer, nor friends/relatives in 2016/17 and 2018/19 agricultural season HH produced grounduts =1 if HH cultivated groundnuts on at least one plot in 2016/17 and 2018/19 HH produced sunflower =1 if cultivated sunflower on at least one plot in 2016/17 and 2018/19 HH produced cowpeas =1 if cultivated cowpeas on at least one plot in 2016/17 and 2018/19 HH acquired herbicides =1 if used herbicides in 2016/17 and 2018/19 on at least one plot HH acquired fertiliser =1 if used fertiliser in 2016/17 and 2018/19 on at least one plot, dummy Use of fertiliser (kg/ha) Amount of fertiliser used per ha in kg in 2016/17 and 2018/19 Secondary outcomes Outcomes related to productivity Maize yield in kg/ha Maize yield in kg per ha (in log form), a continuous variable Soybean yield in kg/ha Maize yield in kg per ha (in log form), a continuous variable Crop value in ZMW/ha Total value of all crops (cereals, vegetables and fruits) in ZMW per ha Net productivity per ha Value of crop production minus variable costs (seed and fertiliser), continuous variable Crop diversificaiton Total number of crops grown by a household, discrete variable Outcomes related to food and nutrition outcomes HFIS Household food insecurity score (0=not food-insecure, 21=often) MDDW Minimum dietary diversity score for women of reproductive age Micro_adequate Dummy=1 if women’s diets in households are adequate in micronutrients Outcomes related to commercialisation indicators Maize commercialisation index Maize revenue (the price of maize per kg multiplied by kg sold) divided by maize value of production (price of maize per kg multiplied by kg of maize harvested), hence continuous variable Soybean commercialisation index Computed the same way as maize commercialisation, a continuous variable Crop commercialisation index Computed the same way as maize commercialisation, a continuous variable Source: Authors We have also carefully constructed various additional covariates that are indicators of sociodemographic and economic indicators and which need to be considered in the outcome estimation framework. Table 1 shows these selected additional covariates used in the regression analyses, while Table A1 in Appendix A presents the full list of variables and their description. IDOS Discussion Paper 19/2022 25 4.3 Empirical approach AAZ activities (training, awareness-creation, construction of depots, input loans) are expected to ease several constraints limiting the adoption of profitable agricultural technologies, sustainable agricultural practices and marketing strategies and/or investment decisions faced by farmers. For instance, training by AAZ increases awareness as well as the adoption of CA and improved agricultural technologies, which can be considered as an investment decision to enhance agricultural productivity if adopted. In order to evaluate the impact of the programme, we employed a DID estimation strategy, in which households in the outgrower programme and non-outgrower farmers are compared before and after the programme intervention. Here, we focussed on evaluating the impact of AAZ programme participation on the adoption of CA, agricultural technologies, productivity and household food security outcomes. The decision to adopt new agricultural practices or improved agricultural technologies can be understood as a dichotomous outcome whereby household 𝑖𝑖 participates in one of the AAZ activities. This could then inspire behavioural changes for the household to adopt CA practices and/or agricultural technologies (mainly seed and fertiliser). This in turn increases productivity or food security when a household anticipates a higher than expected utility due to participation in the AAZ programme as compared to if it had not. The main specification is given as follows: 𝑌𝑌𝑖𝑖𝑖𝑖 =𝛽𝛽1𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑖𝑖+𝛽𝛽2𝑇𝑇𝑇𝑇𝑤𝑤𝑤𝑤𝑇𝑇𝑖𝑖𝑖𝑖 +𝛽𝛽3(𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑖𝑖𝑖𝑖 ∗𝑇𝑇𝑇𝑇𝑤𝑤𝑤𝑤𝑇𝑇𝑖𝑖𝑖𝑖) + 𝑋𝑋′𝑖𝑖𝑖𝑖𝛼𝛼+𝑍𝑍𝑖𝑖𝜂𝜂+𝑢𝑢𝑖𝑖𝑖𝑖 (1) where 𝑌𝑌𝑖𝑖𝑖𝑖 are the outcome variables of interest (CA practices, technology adoption, productivity, food security indicators) for household 𝑖𝑖 at the time of each survey t; 𝑊𝑊𝑤𝑤𝑤𝑤𝑤𝑤𝑖𝑖 is a binary variable that denotes the survey round/year and takes the value “1” if the survey year is 2021, and 0 otherwise; 𝑇𝑇𝑇𝑇𝑤𝑤𝑤𝑤𝑇𝑇𝑖𝑖𝑖𝑖 is a dummy variable that equals 1 for households that participated in one of the AAZ activities at the time of each survey 𝑇𝑇 and 0 otherwise; 𝑋𝑋′𝑖𝑖𝑖𝑖 is a vector representing all time-varying household and community characteristics observed at the time of survey t; 𝑍𝑍𝑖𝑖 are household and community time-invariant factors; 𝑢𝑢𝑖𝑖𝑖𝑖 is a normal stochastic term. The main parameter of interest, 𝛽𝛽3, captures the average treatment effect on the treated (ATT), that is, participation effect for the AAZ outgrower programme participants. The main identification challenge in our settings is that participation in AAZ is not random, since farmers self-select themselves for the programme, hence participation in the programme is endogenous (Khandker, Koolwal, & Samad, 2010). In other words, since AAZ participation is not randomised, the estimation strategy needs to consider that AAZ participants and nonparticipants were likely to be systematically different at the baseline. It should be noted that the interpretation of these effects as causal depends on the identifying assumption: the parallel trend assumption. The assumption states that households with and without AAZ participation during the programme operation would have had the same time trend in the impacts on the outcome variables without the AAZ interventions. Although our difference-in-differences (DID) setting could allow us to remove time-invariant characteristics driving participation, still time-variant characteristics could be the main source of endogeneity, which biases our estimates. To address also this potential endogeneity problem and to exploit the panel structure of our datasets, we employed a propensity score matching (PSM) in combination with DID (PSM-DID) estimators to measure the true effects of AAZ participation (treatment) on the various outcome variables of interest discussed earlier (Abadie & Imbens, 2011; Villa, 2016). One of the main advantages of this approach is that it can consistently estimate the effect of programme participation on our outcome variables of interest, given that our outcome model is specified correctly (Furno & Caracciolo, 2020). PSM helps to IDOS Discussion Paper 19/2022 26 reduce potential bias resulting from self-selection into AAZ activities. Using panel data, this method estimates the ATT. Analytically: 𝐴𝐴𝑇𝑇𝑇𝑇𝐷𝐷𝑖𝑖𝐷𝐷−𝑃𝑃𝑃𝑃𝑃𝑃 = 1 𝑁𝑁𝑇𝑇𝑇𝑇 ∑𝑖𝑖∈𝑇𝑇𝑇𝑇∩𝑃𝑃 ��𝑌𝑌𝑖𝑖,𝑖𝑖+1 𝑝𝑝−𝑌𝑌𝑖𝑖,𝑖𝑖 𝑛𝑛𝑝𝑝�−∑𝑗𝑗∈𝑇𝑇𝑛𝑛𝑇𝑇∩𝑃𝑃 𝑊𝑊𝑖𝑖𝑗𝑗(𝑌𝑌 𝑗𝑗,𝑖𝑖+1 𝑝𝑝−𝑌𝑌 𝑗𝑗,𝑖𝑖 𝑛𝑛𝑝𝑝)� (2) Where Tp (Tnp) represents a treated (non-treated) group, Wij is the nearest neighbour matching weights and S is the area of common support for the covariates. The PSM creates statistically comparable groups based on observable characteristics before performing the DID estimator. This approach has three main advantages (Gebel & Voßemer, 2014). First, it is more robust in minimising misspecification errors. Second, the method ensures a more suitable weighting of covariates. Third, a classical linear regression would extrapolate outside of the area of common support, making comparisons of non-comparable households. A probit model with the following specification is estimated to predict AAZ participation: 𝑇𝑇𝑇𝑇𝑤𝑤𝑤𝑤𝑇𝑇𝑖𝑖𝑖𝑖 =𝜇𝜇𝑖𝑖+𝑋𝑋′𝑖𝑖𝑖𝑖𝛽𝛽+𝑢𝑢𝑖𝑖𝑖𝑖 (3) where X denotes vectors of household time-variant characteristics assumed to influence household i’s participation in the programme, 𝜇𝜇𝑖𝑖 controls for all household and community timeinvariant fixed effects such as initial socio-economic conditions of communities and relationship to the FC that are possibly determining participation. The choice of covariates is based on considerations of the relevant literature and the availability of data that explain participation and outcomes: household socio-demographic and economic characteristics (age, gender and highest level of education in the household, household size, farm size, livestock ownership household assets, housing conditions); institutional and other accessand community-related factors believed to affect participation in AAZ activities); and institutional and access-related factors (credit access, non-farm income source and distance to district capital, extension access, membership in organisation). Based on this estimation, a propensity score is calculated, which is then used to match real AAZ participants in the treatment group with their most similar counterfactual from the comparison group (i.e. the matched controls) via one-to-one nearestneighbour matching without replacement. Since matching results can also be sensitive to variable selection and choice of the matching algorithm, we also carried out some robustness checks. We also conducted additional robustness checks using a semiparametric DID, “a reweighting technique that addresses the imbalance of characteristics between treated and untreated groups” (Abadie, Drukker, Herr, & Imbens, 2004), and by restricting the sample of households that had benefited from the programme in both waves. In the latter case, households are considered as treated if they meet at least one of the following criteria: received AAZ training/advice in both waves; sold to AAZ in wave one and two; or received or acquired a loan or input purchase in both waves. This means that if household i did not participate in the first wave, it is not considered to be part of the treated group. This estimation can also provide some insights about the medium/long-term impacts of the programme. Finally, we conducted additional robustness checks by analysing the impacts of programme participation on outcomes of interest by survey round, that is, separately for the first round and the follow-up using PSM techniques. Heterogeneous treatment effects Although the average treatment effect is interesting in determining the effect of programme participation on outcomes of interest, it fails to unravel the heterogeneous and/or distributional IDOS Discussion Paper 19/2022 27 effects of the treatment (programme participation) at different levels of outcomes. In other words, it helps to address the question: What is the distribution of productivity across the treated household groups if the treatment had not been offered? In addition, policy-makers may be more interested in knowing the effects of programme participation, say on the productivity of farmers at the tail end of the productivity distribution. Thus, as a final analysis, we employed the quantile DID treatment effect on the treated (QTT) framework following Callaway and Li (2019) to estimate the effects of programme participation on quantiles of productivity, mainly maize productivity (kg/ha), soybean productivity (kg/ha), net productivity (ZMW/ha) as well as commercialisation (soybean), as they are the only continuous variable to run quantile regressions. 5 Results 5.1 Descriptive results In presenting the main descriptive results, we used the general treatment indicator (T), that is, participation in one of the three treatment categories. As stated earlier, an individual is considered a participant if: i) they received advice/training from AAZ prior to the survey (free trainings on cultivation, handling of high-value crops, on conservation farming and business skills), ii) if they received input support in the form of seed or purchased seed in cash, iii) if they sold any of their crops to AAZ under AAZ’s guaranteed off-take of the harvested crops, or iv) a combination of the three categories. In our regression analyses, however, we also provided results of the individual treatment (programme) components to gain additional insight. For instance, the effects of training are likely to differ from access to loan input or other aspects of the programme interventions. As a result, analysing each component enables us to uncover the potential benefits of each programme activity. In addition, given the low number of households that participated in the input loans and guaranteed off-take of the harvested crop components of the programme, the overall potential effect of the programme could be obscured by the low level of participation in some potentially more rewarding components. Expanding smaller but more rewarding components in the future may increase the overall average effects. Moreover, the results of these various treatment arms would serve as robustness checks. 5.1.1 Household characteristics Table 4 reports the summary statistics of household characteristics for AAZ programme participants and non-participants from the balanced panel of 582 households, including tests for covariate balancing between the different treatment groups to check the validity of our empirical strategy presented in Section 5.2. Columns 5 to 8 present the summary statistics for households in the different treatment groups at the baseline, and column 6 presents summary statistics for households in the control group at the baseline. Finally, columns 7 to 9 show the mean difference between households in the control group (C) and the various treatment arms (T1-T4). As stated earlier, our baseline data is not ideal since data was collected two years after the programme had started (however, although the programme had started in 2013/14, many sampled treated households only joined in 2015, the year the baseline refers to). For instance, about 57 per cent (339 households out of the 587 outgrower participants) and 32 per cent (192 households out of the 228) of the outgrower participants had already received training and acquired/purchased inputs from AAZ, respectively. Interestingly, however, there are no significant differences on a series of socio-demographic and economic indicators (age and sex IDOS Discussion Paper 19/2022 28 of head, years of education, household size) between the AAZ participants and non-participants, except for a few variables such as kinship ties (head relationship to village chief), cropland and household size. We found that most AAZ participants have closer relationships to the village chief or FCs tend to have more cropland and have a larger family size compared to nonparticipants. For instance, more than 50 per cent of programme participants are related to village headmen, but only around one-third of non-participants are related to the chiefs or headmen. Such differences might affect programme participation, specifically access to loans. Although participants’ mean area of land owned at the baseline (about 5.42 ha) is slightly larger than for non-participants (4.42 ha), the cropped area is almost equal at around 4 ha. Similarly, households in the AAZ programme have higher agricultural asset scores and durable consumer goods (e.g. TV, radio, mobile phone) indices than households in the non-outgrower programme group at the baseline. IDOS Discussion Paper 19/2022 29 Table 4: Mean difference between outgrower participants and non-participants Note: All data is from the 2017/18 survey dataset. Columns 7 to 9 show the mean difference between treatment arm one and control group [T1-C], treatment arm two and the control group [T2-C], and treatment arm three and control group [T3-C]. HH is household; AAZ=Amatheon Agri Zambia; pca=principal component analysis; C=non-AAZ participants. Source: Authors Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Overall T: either T1, T2 or T3 T1: training T2 : input support T3: crop purchase C T1-C T2-C T3-C Mean Mea n Mean difference Female-headed household, 1=Yes 0.17 0.16 0.17 0.15 0.16 0.19 -0.02 -0.04 -0.03 Dummy=1 if married heads 0.83 0.83 0.84 0.84 0.86 0.81 0.03 0.03 0.05 Age of HH head, years 44.61 44.89 44.82 45.38 43.14 44.05 0.77 1.33 -0.91 HH size 7.11 7.31 7.23 7.65 7.65 6.72 0.51 0.93 0.93 Head’s education, years 7.41 7.40 7.45 7.82 7.49 7.42 0.03 0.4 0.07 Maximum education for any HH member, years 9.20 9.23 9.27 9.57 9.33 9.14 0.13 0.43 0.19 Agricultural index (pca), incl. tractor/oxen 0.10 0.15 0.12 0.60 0.24 0.02 0.1 0.58 0.22 Agricultural index, only tractor/oxen/plough (pca) 0.06 0.08 0.06 0.30 0.12 0.03 0.03 0.27 0.09 Agricultural index (pca), w/o tractor/oxen 0.08 0.12 0.10 0.50 0.22 0 0.1 0.5 0.22 HH wealth index (pca), incl. motorbike/cars 0.08 0.15 0.15 0.42 0.30 -0.05 0.2 0.47 0.35 HH wealth index (pca), w/o motorbike/cars 0.09 0.17 0.16 0.39 0.25 -0.07 0.23 0.46 0.32 Dummy=1 if main tillage done before the rains on any plot 0.09 0.12 0.13 0.13 0.19 0.02 0.11 0.11 0.17 Total cultivated land, ha 3.75 3.96 3.93 4.44 4.36 3.34 0.59 1.1 1.02 Cost of crop inputs per hectare seed+fertiliser, ZMW 983.65 988.40 984.20 1112.00 1084.38 974.17 10.03 137.83 110.21 Dummy=1 if HH rents land 0.03 0.03 0.04 0.03 0.03 0.02 0.02 0.01 0.01 Per capita farm size, ha 0.76 0.79 0.80 0.84 0.79 0.7 0.1 0.14 0.09 Land access per capita, ha 0.76 0.79 0.80 0.85 0.79 0.71 0.09 0.14 0.08 Dummy=1 if HH accessed loan from sources other than from AAZ 0.28 0.30 0.29 0.24 0.34 0.24 0.05 0 0.1 Dummy=1 if HH irrigated any field 0.00 0.00 0.00 0.01 0.00 0 0 0.01 0 Dummy=1 if HH experienced a shock in past five years 0.73 0.74 0.73 0.75 0.72 0.72 0.01 0.03 0 Chief related to head or spouse of HH, 1=Yes 0.09 0.10 0.09 0.11 0.07 0.07 0.02 0.04 0 Headman related to head or spouse of HH, 1=Yes 0.47 0.52 0.53 0.45 0.52 0.36 0.17 0.09 0.16 Farmer coordinator related to head or spouse of HH, 1=Yes 0.38 0.49 0.51 0.51 0.49 0.17 -0.02 -0.04 -0.03 Chief or headman related to head or spouse of HH, 1=Yes 0.48 0.54 0.55 0.47 0.54 0.36 0.03 0.03 0.05 No. of HHs 587 386 339 192 138 196 IDOS Discussion Paper 19/2022 30 5.1.2 Primary (intermediate) outcomes of interest By comparing the means at the baseline (2016/17 agricultural season) and the follow-up (2019/20 agricultural season) for each group separately, we present below the main descriptive results for the different primary and secondary outcome variables of interest we discussed earlier. Adoption of agricultural technologies One of the mechanisms through which AAZ interventions are expected to improve productivity, household income or food security is through the promotion and adoption of improved farming techniques. Conceptually, the presence of a large-scale farm (LSF) operation could facilitate improved technology adoption among farmers in nearby communities through learning effects and cost effects (Liverpool-Tasie et al., 2020). Essentially, the cost effects refer to the transaction cost reductions for the smallholder farmers that come with purchasing inputs from LSFs. The learning effect results from training or extension messaging from medium and largescale farms that increases the productivity of smallholder farms (Liverpool-Tasie et al., 2020). As we show later in our regression analysis, the evidence does indeed suggest that AAZ interventions impact productivity and food security via these mechanisms. Transaction cost reductions result from the pooling of purchases between LSFs and neighbouring small farms in cases where the two grow similar crops, given that LSFs enjoy economies of scale in transport (Deininger & Xia, 2016). If the location of an LSF is accompanied with investments in public infrastructure such as roads, transport costs may decrease; however, this is not always the case, as investments may locate to areas with already developed infrastructures (Lay et al., 2020). There are also transaction cost reductions that may arise because of the location of input suppliers closer to the community due to the presence of an LSF. From the learning aspect, the location of LSFs near smaller farms creates an enabling environment for smallholder farmers’ access to better extension services provided by LSFs. This result has been demonstrated for Tanzania, where Wineman et al. (2021) associate the presence of large farms in an area with improved extension access, an increased likelihood of the cultivation of cropland and increased usage of improved seed for the neighbouring farms. As mentioned earlier, AAZ has a range of interventions, including providing input support to farmers in the form of seed loans and improving access to improved seeds so that they can buy inputs. In addition, free training on crop cultivation, handling and post-harvest loss could improve the adoption of these technologies. Hence, the interventions could have a positive effect on the adoption of agricultural technologies. In Table 5, the trends in the adoption of technologies among farmers in the participating and non-participating groups are presented. For the key crops promoted during the first two phases of the outgrower programme (soybeans, groundnuts, cowpeas and sunflower), results show a significantly higher adoption of groundnuts, sunflower and cowpeas among the outgrower participants during the first wave of data collection. Yet, the follow-up results show a statistically significant decline in the share of hybrid or open pollinated varieties of the same crops used by households, suggesting that many initial adopters did not continue after the end of the programme. Among non-treated, sunflower and cowpea adoption was low, both at the baseline and the follow-up. There are significant increases in the fertiliser application rates and share of households using fertiliser on their soybean fields in both the treatment and control groups. Yet, a significant share of respondents in both groups reported applying fertiliser on any fields they cultivated, about 82 IDOS Discussion Paper 19/2022 31 per cent for participating groups and 68 per cent for non-participating groups; although we observed a decline in the share of households applying fertilisers. However, for farmers applying fertiliser, the change in the soybean fertiliser application rate is higher among AAZ programme participants than non-participants. The opposite was observed for maize production, with fertiliser-use intensity declining by almost 96 kg in the treatment group and about 75 kg in the control group. An important side note here is that maize fertiliser also comes from the government’s subsidy programme, and the result may reflect the programme implementation or a shift of fertiliser towards soybeans. Although not statistically significant, the use of nonrecycled seed declined for both groups. A similar trend is observed for seeds acquired from friends/family or other farmers. This suggests that sources of seeds other than family or friends became more important for farmers. Table 8: Mean differences for secondary outcomes of interest between the baseline and the follow-up Note: After the first wave, most of the activities of AAZ shifted to the promotion of cash crops such as quinoa. Net productivity per ZMW/ha is the difference between the value of crop production per ha minus seed costs and fertiliser costs. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors A B C C-B D E F F-E Treatment (outgrower participants) Control (non-participants) Variables Overall Baseline Follow-up Difference Overall Baseline Follow-up Difference Participation in crop production Dummy=1 if HH produces maize, 0 otherwise 0.98 0.98 0.97 -0.01 0.95 0.97 0.91 -0.06** Dummy=1 if HH produces soya beans, 0 otherwise 0.83 0.81 0.85 0.04 0.76 0.77 0.74 -0.04 Dummy=1 if HH produces cowpeas 0.07 0.1 0.04 -0.06*** 0.02 0.02 0.02 0 Dummy=1 if HH produces groundnuts 0.46 0.45 0.47 0.02 0.44 0.43 0.45 0.03 Dummy=1 if HH produces sunflower 0.13 0.11 0.14 0.03 0.1 0.08 0.13 0.04 Secondary outcomes Maize hectares planted 2.13 2.07 2.19 0.12 1.95 1.89 2.03 0.14 Maize yield, kg/ha planted 2,063.3 2,326 1,829 -496.87*** 2,014 2,246 1,692 -554.78*** Maize price per kg (farm gate), ZMW 1.76 1.18 2.30 1.13*** 1.61 1.13 2.35 1.21*** Value of maize sales, ZMW 5,145 3,686 6,969 3,282.32*** 4,968 3,692 7,869 4,176.28*** Soybean hectares planted 1.52 1.34 1.67 0.33*** 1.43 1.34 1.55 0.21 Soybean yield, kg/ha planted 1,053 1,591 597 -993.7*** 1,089 1,491 543 -947.5*** Soya price per kg (farm gate), ZMW 3.68 3.05 4.38 1.33*** 3.5 2.9 4.7 1.72*** Value of soybean sales, ZMW 6,356 5,783.27 7,002.03 1,218.76** 5,608 5,527 5,757 229.6 Value of fruit/vegetable sales, ZMW 2,133 10,709 535 -10,173.76*** 6,338 30,335 431 -29,903.77*** Crop total production value, ZMW 20,890 15,964.02 25,619.72 9,655.71*** 17,879 13,902 24,135 10,233.13*** Crop land productivity (in ZMW/ha) 4,388 3,671 5,076 1,405.5*** 4,093 3,731 4,663 931.75*** Value of crop sales, ZMW 15,765 12,197 19,191 6,994.7*** 12,820 9,879 17,445 7,566.28*** Crop and fruit sales value, ZMW 16,065 12,627 19,286 6,658.76*** 13,250 10,621 17,154 6,532.75*** Total production value (crops, fruits/vegetables), ZMW 21,229 16,693 25,478 8,784.55*** 18,470 15,405 23,020 7,614.91*** Net productivity per ZMW/ha 3,561 2,682 4,404 1,721.98*** 3,227 2,757 3,968 1,211.3*** Commercialisation Maize commercialisation index 1 0.46 0.65 0.2*** 0 0 1 0.19*** Soybean commercialisation index 0.82 0.90 0.73 -0.16*** 0.85 0.91 0.75 -0.16*** Crop commercialisation index 0.72 0.66 0.77 0.11*** 0.67 0.61 0.77 0.15*** No. of HHs 822 386 436 347 196 151 IDOS Discussion Paper 19/2022 39 Household food insecurity and women’s nutritional outcomes We also examined if there are mean differences in food insecurity status between the baseline and the follow-up for the participant and non-participant groups. As discussed earlier, food and nutrition security is measured in terms of the share of the household food insecurity access score and the minimum dietary diversity score (DDS) for women of reproductive age. As indicated in Table 9, on average, the food insecurity access score for households in the general treatment group was about 5.67 at the baseline. This figure had increased at the follow-up (5.75). Conversely, the average household food access insecurity score for farmers who had never participated in any of the three components was about 5.67 at the baseline and 5.19 at the follow-up. The differences for the two groups over the two study periods are not statistically significant. As to the number of households with adequate micronutrients for women of reproductive age, we found a slight decrease at the follow-up compared with the baseline for both treatment and control groups, more so in the control group. However, the difference between the baseline and the follow-up is statistically significant only for the control group. Table 9: Mean differences for food security indicators by participation status and survey round Outgrower participants Non-outgrower participants A B C C-B E F G G-F Variables N Overall N Baseline N Followup Difference N Overall N Baseline N Followup Difference HHs with adequate micronutrients for women of reproductive age 611 36% 290 39% 321 33% -6% 244 31% 145 35% 99 24% -11* HH food access insecurity score 822 5.71 386 5.67 436 5.75 0.09 347 5.46 196 5.67 151 5.19 -0.48 Minimum DDS for women of reproductive age 611 4.19 290 4.26 321 4.13 -0.12 244 4.11 145 4.16 99 4.03 -0.13 Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors In summary, the simple means comparisons suggest that participants in the AAZ programme performed on average slightly better than non-participants in terms of technology adoption, CA practices, productivity, and food and nutrition security, except for the household food access insecurity score. However, since mean comparisons presented earlier do not take into account the socio-economic and institutional differences of households (e.g. differences in terms of access to relevant resources, assets, transport costs and other household characteristics) as well as communities, it is important to account for these factors, as done in the next section, to draw intuitive conclusions. In addition, mean comparisons do not help to quantify the relative effect of the programme on participants compared to non-participants and why. Now we present the regression results. IDOS Discussion Paper 19/2022 40 5.2 Empirical results In the following section, we discuss the regression results for the main outcome variables of interest grouped along the themes discussed earlier: technology adoption, sustainable land management practices, land productivity (for maize and soybeans, net and aggregate), commercialisation, and household food and nutrition security. In most of our analyses, we used propensity score matching difference-in-differences (PSM-DID) discussed earlier to estimate the average treatment effect on the treated (ATT). The main discussions are based on the general treatment indicator discussed earlier – that is, participation in one of the three activities/programmes of AAZ trainings offered, input loan provision and purchase of cereals (off-take of the harvested crops) – by comparing change over the time of the outcomes of interest across the treatment groups. We adjusted for differences between the treatment and control groups on the observable characteristics at the baseline that are correlated to the propensity score. 5.2.1 Effects on primary outcomes Effect on the adoption of improved technologies Table 10 reports the estimation results from the main specification (1) using PSM-DID for the adoption of improved agricultural technologies, mainly improved seed (OPV, hybrid seed varieties) for maize and soybeans and inorganic fertiliser use on maize and soybean fields. Overall, our analysis shows that participating in at least one of the AAZ programmes does not significantly affect the adoption of improved seeds compared to non-participation, but it does affect the use of inorganic fertilisers. In terms of household fertiliser use on maize fields, the programme has an adverse and statistically significant effect; however, it has a positive and statistically significant effect on household fertiliser use on soybean fields. For instance, participation in at least one of the programmes decreased households’ use of chemical fertilisers on maize fields by 20 percentage points (a relative decrease of 28 per cent compared to the control group), but raised households’ use of chemical fertilisers on soybean fields by 3 percentage points. Interestingly, even if the programme led to the increased use of fertilisers on soybean fields, it did not result in significant changes with regard to the adoption of OPV soybean seeds. The same is true with fertiliser application on maize fields, where no impact on OPV maize seeds was found. To determine whether the impact of each intervention on the adoption of improved seeds is also insignificant, we examined further the effects of the three types of intervention (training, seed loans, crop sales to AAZ) on the adoption of these technologies. By doing so, we found that input support in the form of seed loans increased the use of non-recycled seeds, particularly OPV maize seeds, whereas the purchase of cereals from farmers (treatment T3) increased the adoption of improved soybean seeds. The provision of training (on conservation farming, business skills, production and processing) had no effects on the adoption of these technologies (see Table R1-B in Appendix B for the detailed results). Since AAZ interventions had already begun at the time of the baseline survey and some project activities had changed over the course of the project, evaluating the impact of programme participation separately for the baseline (in its early phase) and the follow-up (four years after the start of the operation) is useful. For that, we examined the effects of outgrower participation separately at the baseline and the follow-up using the PSM estimation techniques. Our estimation results suggest that the programme had greater effects on the adoption of improved IDOS Discussion Paper 19/2022 41 maize seeds and fertiliser application on soybean fields at the time of the baseline survey than at the time of the follow-up survey (see Tables R13B1-R13B2, Appendix B). In its early operation, AAZ, with the support of USAID, had engaged in the construction of community trading depots, which are accessible to farmers. This may bias our ATT estimates. In order to reduce the bias, we controlled for proximity to community trading depots (proxied by household distance to the nearest depot centre) and AAZ farm blocks (measured by the distance between AAZ farm blocks and location of household residence). We found that households’ distance to the nearest depot or AAZ farm block is negatively and statistically significantly associated with most of the indicators of agricultural technologies (except for the adoption of OPV soybeans) (Table 10). This means that households closer to trading depots were more likely to adopt these improved technologies compared to those located farther away. For instance, a 1 per cent increase in household distance to the nearest depots reduces the likelihood of using improved seeds and/or fertiliser applications on fields by about 4 per cent to 6 per cent, depending on the type of improved seeds. Table 10: Effect of AAZ participation on the adoption of technology components, PSM-DID estimates (1) Adopted OPV/hybrid soybean seed (2) Adopted nonrecycled seed, not from relatives (3) Adopted nonrecycled seed (4) Purchased seeds in community (5) Adopted OPV/hybrid maize seed (6) Applied fertiliser on any maize field (7) Applied fertiliser on any soya field (8) Log (fertiliser on maize field, kg/a) (9) Log (fertiliser on soybeans field, kg/a) ATT_PSM-DID -0.0370 -0.0529 -0.0627 0.0260 -0.0673 -0.196*** 0.0249* -1.226*** 0.121* (0.0516) (0.0531) (0.0514) (0.0631) (0.0543) (0.0521) (0.0177) (0.276) (0.0715) Observations 1,061 1,154 1,154 1,008 1,129 1,162 1,162 1,139 1,061 R-squared 0.104 0.004 0.008 0.045 0.006 0.041 0.009 0.040 0.016 Log (distance to nearest depot) 0.000 (0.098) -0.279** (0.105) -0.273* (0.106) -0.246** (0.094) -0.269* (0.110) -0.222 (0.116) -0.072 (0.175) - - Log (distance to AAZ) -0.016 (0.011) -0.008 (0.012) -0.011 (0.012) 0.007 (0.011) -0.018 (0.013) -0.038** (0.014) 0.017 (0.019) - - Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. If we did not control for the presence of depots and distance to AAZ farm blocks, the use of nonrecycled seed turns positive and statistically significant, further supporting the positive effects of depots and AAZ farm blocks. We also found that herbicide and insecticide application was significantly reduced. ATT_PSM_DID denotes the average treatment effect on the treated estimated using Kernel PSM-DID. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors Effect on the adoption of sustainable land management practices As discussed earlier, AAZ – along with the Conservation Farming Unit (CFU) – has promoted the adoption of SLM practices in both districts. Irrespective of the estimation techniques used (conventional PSM-DID, FE or semiparametric difference-in-differences) and in line with the descriptive results, we found a strong and consistent significant effect of AAZ programme participation on the adoption of various SLM practices, except for the adoption of minimum soil IDOS Discussion Paper 19/2022 42 disturbance (MSD) and crop rotation (CR) (Table 11). For instance, our estimates suggest that participation in AAZ increases the adoption of full-suite CA practices (MSD + CR cereals to legumes or cereals to fallow + residue retention (RR)) by 8 percentage points, compared to nonparticipants. Considering the overall mean of the treatment/sample group is 0.049, households in the programme are 63 per cent more likely to adopt SLM practices than those not in the programme. Moreover, proximity to community trading depots and AAZ farm blocks increases the adoption of SLM/CA practices. In addition, although the programme had a positive effect on most of the SLM practices considered in this discussion paper, there were cases in which the programme had also negatively affected the adoption of some of the SLM practices, such as MSD in combination with CR or RR. These results, however, should be interpreted with caution and not attributed as causal effects, as they might be driven by unobservable time-variant factors. It would also be interesting to explore further how the construction of trading depots by AAZ (to sell inputs and buy grains) and proximity to AAZ farm blocks have contributed to the adoption of SLM practices. Table 11: The effects of AAZ participation on the adoption of SLM practices, PSM-DID estimates (1) (2) (3) (4) (5) (6) Variables Adopted CR and RR practices Adopted MSD and RR practises Adopted MSD and CR practises Adopted full suite CA practises (MSD+CR+RESID) Adopted MSD+CR or MSD+RR Main tillage done before the rains on any plot ATT_PSM-DID 0.0699 0.0799*** -0.103** 0.0779*** -0.0978** 0.644** (0.0490) (0.0308) (0.0454) (0.0287) (0.0460) (0.207) Log (distance to nearest depot) -0.00396 -0.00944 -0.0688*** -0.00829 -0.0699*** -0.003 (0.0243) (0.0165) (0.0229) (0.0159) (0.0232) (0.122) Log (distance to AAZ) -0.00103 -0.00354* -0.0118*** -0.00336* -0.0119*** -0.028 (0.00279) (0.00190) (0.00263) (0.00182) (0.00267) (0.015) Observations 1,064 1,064 1,064 1,070 1,070 R-squared 0.041 0.057 0.061 0.050 0.057 Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. ATT_PSM-DID denotes the average treatment effect on the treated estimated using PSM-DID. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors 5.2.2 Effects on secondary outcomes Effect on productivity, returns (net and gross) and crop diversification In the previous sections, we showed that households participating in the AAZ programme are more likely to use improved seeds, apply chemical fertilisers (though for soybean production) and adopt SLM practices than non-participating households. If so, one can expect positive effects of these technologies and improved SLM practices on productivity, returns as well as crop diversification. The next question then is to explore whether the adoption of improved seeds, adjustments to fertiliser use and use of improved SLM practices led to improved yields, productivity (net as well as gross) and crop diversification. In Table 12, we examine the impact of AAZ participation on maize productivity (measured in kg/ha and monetary values (ZMW/ha)), Soybean productivity (in kg/ha and ZMW/ha), aggregate crop productivity (sum of cereals, fruits and vegetables in ZMW/ha), net productivity (in ZMW/ha) and crop diversification (number of crops grown). Descriptive analysis suggests mixed results in the sense that participating IDOS Discussion Paper 19/2022 43 households have higher farm productivity (net returns per hectare) when monetary values are used instead of yield harvested and grow more crops than non-participating households. To make it comparable with crop net returns, we also used monetary values per ha for the analysis of maize and soybean productivity. The regression results suggest no statistically significant causal effect of AAZ interventions on various indicators of productivity and returns (despite a positive trend in net productivity), except for maize productivity. Interestingly, participating households would have reduced the maize yield/ha by 12 percentage points had they not participated in the programme, a relative loss of 7 per cent (Table R3). Earlier we showed that the programme increased the use of improved maize seed and decreased fertiliser use on maize fields. The weak or lack of a significant effect of the programme on productivity could be due to various interrelated factors. First, although AAZ interventions may not have a direct significant effect on improving farmers’ productivity or net returns, they do encourage the adoption of CA and improved technologies, as shown below (Table R3-B, Appendix B). We found that the adoption of these technologies and SLM practices have a strong positive impact on productivity. It could also be possible that adoption of improved agricultural technologies and CA practices may not improve yields in the short run. Second, the focus and activities of the AAZ programme had changed between the baseline and follow-up surveys from maize and soybeans (as well as groundnuts, sunflower and cowpeas) to cash crops such as quinoa and chia.8 Third, estimation techniques based on pooled panel data for the baseline and the follow-up might obscure the heterogeneous effects of the programme on productivity. To verify if this was the case, we re-ran our analysis separately for the two survey periods, that is, at the time of the first wave and the second wave. To minimise the bias influencing programme participation due to observable covariates, we used the PSM approach. Our estimation results suggest that there is no statistically significant effect for most of the productivity indicators. However, we observed that the magnitude of our estimates are larger at the follow-up period. This could be suggestive evidence of some longer-term effects of the programme. In addition, we found that AAZ programme participation has a positive and more significant effect (at 10 per cent) on aggregate productivity (crop, fruit and vegetables) at the time of the follow-up than at the baseline period. The results are presented in Table R13B3 in Appendix B. Fourth, given the nature and design of the programme, spillovers/contamination with non-participating households is highly likely, which could underestimate the true effect of the programme. Finally, the attrition rate was high among participating households. As to the effect of the interventions on crop diversification – one important strategy to diversify the risk of crop failure – our estimates suggest that participation in one of the AAZ components increases crop diversification, although we did not find a statistically significant difference. Separately analysing by survey rounds suggests, however, that programme participation has a strong and statistically significant positive effect on crop diversification at the baseline survey, about 32 percentage points (Table R13B3). It is also interesting to note that exposure to AAZ farm blocks (being closer to AAZ farm blocks) is positively associated with soybeans and aggregate productivity, but proximity to trading depots is not statistically significant. Examining the adoption of agricultural technologies and SLM practices as potential mechanisms, we found that both the adoption of agricultural technologies (seed and fertiliser) and SLM practices had a positive effect on soybean productivity, maize productivity (improved 8 Focus group discussion participants indicated that most of the AAZ activities decreased since 2017, including late delivery of inputs and late payment for the crops purchased, and significant decrease in purchase of maize by AAZ in 2017. IDOS Discussion Paper 19/2022 44 technologies only), aggregate productivity (total value of crop and fruits per ha) and crop diversification (see Table R5-B, Appendix B). In addition, the adoption of improved technologies seems to have a stronger effect on productivity than SLM practices. All in all, the findings suggest that the programme affects productivity and nutrition security, especially that of women, by driving the adoption of SLM practices and improved agricultural technologies, such as the use of improved seed and fertiliser. Table 12: Effect of AAZ participation on productivity, profits and crop diversification, PSM-DID estimates (1) Log maize yield in kg/ha (2) Log soybean yield in kg/ha (3) Log of crop and fruits/vegetables value in ZMW per ha (4) Log of crop land productivity in ZMW/ha (5) Log of net productivity per ZMW/ha (6) No. of crops grown ATT_ PSM-DID 0.127** -0.0954 0.0971 -0.101 0.00272 0.0855 (0.0613) (0.127) (0.168) (0.104) (0.133) (0.123) Observations 1,044 876 1,035 1,020 990 1,064 Log (distance to nearest depot) -0.0131 5.84e-05 -0.0749 -0.0329 0.0507 0.0112 (0.0242) (0.0628) (0.0666) (0.0497) (0.0634) (0.0606) Log (distance to AAZ) 0.00165 -0.0199*** -0.0131* -0.0140** -0.00976 0.00336 (0.00277) (0.00711) (0.00757) (0.00562) (0.00717) (0.00697) R-squared 0.009 0.0263 0.040 0.034 0.054 0.005 Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. ATT_PSM-DID denotes the average treatment effect on the treated estimated using PSM-DID. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors Effect on commercialisation (cereal market participation) Since one of the main project activities of AAZ is to support farmers market participation and trade with farmers through direct purchases using constructed rural trading depots as well as by providing free training on business skills, we also examined whether the programme has improved commercialisation patterns (maize, soybeans and aggregate crop commercialisation). Table 13 presents the results of this econometric analysis. We found that participation in AAZ programmes decreased farmers’ commercialisation of maize, soybeans and overall crop commercialisation, though this effect is statistically significant only for the latter outcome. If one looks at the aggregate measure of crop commercialisation patterns, it seems that participating households tend to have lower market participation levels than non-participating households, by about 6.5 percentage points. This is not surprising, given that AAZ’s grain purchases from smallholders (through its guaranteed off-take of the harvested crops) were intense only during the initial phase of its “outgrower” programme. Later, the focus of AAZ had shifted to high-value crops such as quinoa purchases. This is further substantiated by the fact that programme participation has a positive and statistically significant effect on the crop commercialisation index during the baseline but not during the follow-up (see Table R13B4, Appendix B). In fact, we observed a decreasing trend of commercialisation patterns during the follow-up survey. It is also interesting to note that households that live farther away from depots showed lower market IDOS Discussion Paper 19/2022 45 participation levels, more so regarding soybeans sales (Table R4-B). For instance, a 1 per cent increase in households’ distance to the nearest depots is associated with a 2 percentage point reduction in sales of soybeans. Table 13: Impact of AAZ participation on crop commercialisation, PSM-DID estimates (1) Maize commercialisation index (2) Soybean commercialisation index (3) Crop commercialisation index ATTPSM-DID -0.0680 -0.000183 -0.0649** (0.0435) (0.0278) (0.0279) Log (distance to nearest depot) -0.000642 -0.0246* -0.00398 (0.0180) (0.0147) (0.0121) Log (distance to AAZ) 0.0150 -0.0167 -0.00795 (0.0431) (0.0359) (0.0314) Observations 925 798 1,073 R-squared 0.119 0.180 0.093 Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. ATT_PSM-DID denotes the average treatment effect on the treated estimated using PSM-DID. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors Effect on household food and nutrition security Hunger and malnutrition are still major challenges in SSA, and smallholders are among many of the food-insecure households (Sibhatu, Arslan, & Zucchini, 2022). The situation is no different in Zambia. Thus, it is important to examine the impact of the programme on household food and nutrition security. One would expect that agricultural training, input support in the form of seed and extension services, and output market interventions (such as a guaranteed off-take of the harvested crops) help to alleviate the food and nutrition insecurity of smallholder farmers and improve household dietary diversity through increased productivity, diversification of crop portfolios and market participation. In this regard, the study investigated whether programme participation has any effect on household food and nutrition security and/or dietary diversity measured using the household food insecurity access score and the minimum DDS for women of reproductive age. We observed that participation in AAZ programmes (training, input loans/purchase, selling grains to AAZ) did not significantly reduce household food insecurity conditions and/or improve women’s dietary diversity in participating households. Interestingly, however, we found suggestive evidence of the programme’s effects on improving the likelihood of women meeting their micronutrient adequacy in beneficiary households. For instance, our estimates suggest that participating households had a 17 percentage point (about 48 per cent) higher likelihood of having adequate diets for women in the household compared to nonparticipants if the programme had not been offered (Table 14). This suggests that the programme has a positive effect on improving women’s nutrition in the household. A detailed analysis by survey round and intervention type did not alter our main conclusions (see Table R13B5, Appendix B). However, the use of quantile regression analysis suggests that the programme had stronger effects on those who were extremely food-insecure, that is, the programme benefited mostly those households with the highest food insecurity score (results not reported here). IDOS Discussion Paper 19/2022 46 The introduction of new nutritious crops (or crop diversification) and extension services – including training and/or participation in off-farm employment – are some of the potential mechanisms through which the programme could contribute to such improvements. For instance, we examined the effect of programme participation on households’ participation in offfarm employment, which is the most common livelihood diversification strategy among poor households in developing countries. In this regard, our analysis suggests that there is an increasing upward trend in off-farm employment for both groups, but the programme increases participation in off-farm employment of participating households by about 8 percentage points, and this effect is statistically significant (a relative increase of 50 per cent compared to sample average) (see Table R13B5, Appendix B).9 In addition, we observed that the programme increased productivity through the adoption of technologies and CA practices. Increased productivity would contribute to such improved nutritional outcomes. Table 14: Effect of AAZ programme participation on household and women’s food and nutrition security, PSM-DID estimates (1) Household food insecurity score (2) Minimum DDS for women of reproductive age (3) Dummy=1 if women’s diets in households are adequate in micronutrients ATTPSM-DID 0.184 0.197 0.171** (0.565) (0.157) (0.0676) Observations 1,162 828 828 R-squared 0.010 0.013 0.030 Note: Training has a positive and significant effect on the DDS of women, suggesting the importance of the training/education component of the intervention in improving women’s nutritional outcomes. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors Effect on the production of oilseed crops As explained above, with the programme’s expansion from Mumbwa to Chibombo, the programme’s focus increasingly shifted around 2017 towards supporting the adoption of sunflower, cowpeas and groundnuts through the provision of seeds and by purchasing the crops from farmers. This component, however, was abandoned when the programme switched towards quinoa and chia. Table 15 shows the estimation results for the PSM-DID, that is, after the follow-up survey. The results do not suggest significant effects on crop adoption after the programme has ended. 9 However, further analysis is required if the increase in off-farm employment is due to the loss of land as a result of the expansion of AAZ, or if it is because of the search for alternative livelihoods. IDOS Discussion Paper 19/2022 47 Table 15: Effect of AAZ participation on the production of other crops, PSM-DID (1) Produced legumes (2) Produced sunflower (3) Produced cowpeas (4) Produced groundnuts ATTPSM-DID 0.0699 0.0620* -0.0759*** -0.00916 (0.0490) (0.0407) (0.0268) (0.0634) Observations 1,064 1,070 1,070 1,064 R-squared 0.315 0.004 0.027 0.001 Note: A household is treated if it received advice/training from AAZ or acquired an AAZ input loan/ input purchase or sold grain to AAZ. ATT_PSM-DID denotes the average treatment effect on the treated estimated using PSM-DID. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors 5.2.3 Individual treatment components: Intervention types matter? So far, our analyses focussed on whether households benefiting from (or participating in) any of the AAZ interventions – such as advice/training (T1), input support (in the form of loan or access) (T2), or a guaranteed off-take of the harvested crops (purchase of cereals/grains (T3)) – are different from non-beneficiary households. Nonetheless, it is plausible to assume that the effects of each of the interventions on those outcomes differ. For instance, the effect of training on productivity could be different from that of input support or output market guarantee. As a result, the general treatment indicator (T) we used earlier could potentially conceal (or underestimate) the effects of each of the treatments on the various outcomes of interest discussed earlier. To explore this, we examined the effects of each of the three treatments (T1, T2, T3) on the relevant outcome variables of interest. Results from the various specifications of equation (1) are presented in Table 16. In order to save space, we focussed on selected technologies, SLM practices, productivity and food security. Our results suggest the following: In terms of the impact on the adoption of agricultural technologies (specifically the adoption of improved maize seed varieties and fertiliser use), input support in the form of seed loans and output purchases were more effective (positive and statistically significant effects) than training (Table 17, columns 1-3). As to the effect on the adoption of various combinations of SLM practices, all three types of intervention had positive and significant effects (Table 16, columns 5-8). As to the impacts on land productivity (both on aggregate and individual crops), we found no clear evidence of impacts on maize or soybean productivity, but input loans and grain purchases had positive effects on the aggregate productivity of beneficiary households (Table 17, columns 9-11). Interestingly, only training had a consistently positive and statistically significant effect in improving the dietary diversity of women and the micronutrient adequacy of women’s diets in beneficiary households (Table 17, columns 13-14). Output market interventions seem to reduce household food insecurity, though statistically insignificant (Table 17, column 12). In addition, farmers who received input support or sold grain to AAZ diversified their crop production portfolios more than those who did not receive this support (not reported here). In sum, the results suggest that AAZ interventions affect outcomes of interest differently: Whereas training on CA and business skills appears to be more effective in improving women’s nutrition outcomes as well as enhancing SLM practices, input support and output market interventions were more effective in improving technology adoption, enhancing SLM practices and productivity. In other words, despite consistently positive effects, we found little evidence to IDOS Discussion Paper 19/2022 54 well understood, not least because the external and internal conditions of cooperation vary from case to case. For instance, the absence of institutional and/or proper contracts between farmers and the company for product delivery (inputs and outputs) and extension services could undermine the effectiveness of such interventions. In our case, qualitative findings suggest that product delivery (of harvested crops) contracts are given without price guarantee mechanisms for all the crops grown by farmers. This discussion paper attempts to contribute to filling this knowledge gap by analysing the expost effects of a foreign commercial investment in farmland in Zambia – the AAZ “outgrower” programme – on various household outcomes: the adoption of CA, improved agricultural technologies, crop productivity, food and nutrition security, and market participation. AAZ had made several adjustments during its operation. In the initial phase (phase 1), the programme focussed mainly on maize and soybeans as part of its strategy to vertically integrate grain producers into the food value chain. During phase 2, the programme expanded to Chibombo district and included the supply of seeds for cowpeas, groundnuts and sunflower, as well as the purchase of these promoted crops. In phase 3 (after 2019), AAZ re-focussed to quinoa and chia. As a result, input support for maize and oilseeds as well as the purchase of these crops as part of the company’s outgrower programme dropped significantly after 2019. Although the programme activities changed over the course of the project, free training remained the focus of the programme. This paper focusses on the impacts of the first and second phases of the AAZ outgrower programme. It uses two rounds of household survey data: a baseline survey in 2018 focussing on the 2016/17 main agricultural season, and the follow-up survey, which took place in 2021, with the focus on the 2019/20 main agricultural season. The paper uses both descriptive analyses as well as various econometric techniques that compare households participating in the AAZ programme with non-participant households before (in an early phase) and after the programme intervention. Descriptive results illustrate that AAZ programme participants performed, on average, slightly better than non-participants in terms of technology adoption (mainly uptake of promoted crops, except cowpeas/groundnuts), uptake of CA practices, productivity, market participation (mainly maize marketing) and nutritional outcomes of women. We found that programme participants scored higher in adopting at least one or two CA practices, in the number of agricultural technologies adopted, crop diversification, the dietary diversity of women in the household as well as market participation (specifically maize). Econometric results demonstrate that, although the overall impact of the AAZ outgrower programme on the uptake of CA practices is robust and promising, impacts on other aspects of technology adoption – specifically improved seed varieties – depend on the types of crops promoted. For instance, we found that AAZ outgrower participants are more likely to use at least one or a combination of CA practices, but less likely to apply fertiliser on maize fields (used less per hectare). We also observed that the focus on cereals and oilseeds in an early phase of the programme yielded more impact. Besides, the duration of the intervention matters and should not be overlooked in interventions that necessitate gaining experience and learning. As to the AAZ programme components, we found that types of intervention (or support given) matter to enhance technology adoption. For instance, training was effective in enhancing the uptake of CA practices and more so during the second round survey; however, input support in the form of seeds and guaranteed off-take of harvested crops were more effective in increasing the adoption of agricultural technologies. IDOS Discussion Paper 19/2022 55 As to the impacts of outgrower programme participation on crop productivity, we found weaker effects during the first phase, except on maize productivity. In addition, the impacts on crop productivity were more significant during its early phase than in later phases, that is before the programme shifted its focus towards commercial crops. One explanation could lie in the way the programme increased both productivity and market participation through the purchase of selected crops during the early phase of its programme operation than in later phases. As to the programme’s integrated value chain strategy, the overall effects of the programme on crop commercialisation suggest that AAZ’s strategy of integrating smallholder farmers into their primary production of staple crops seems to be ineffective, although there were some positive effects detected during the early phase of the programme operation.10 Furthermore, our results suggest that the effects of the programme on productivity and commercialisation vary by project focus (whether the focus was on cereals or oilseeds) and duration (in an early phase or after some years). Moreover, our results indicate that the programme could increase household productivity (and nutritional outcomes) by enhancing the adoption of agricultural technologies and uptake of CA practices.11 Of the various interventions, training was found to be the most useful tool to enhance the adoption of sustainable land management practices that to have longterm impacts on productivity. However, seed loans and output purchases from farmers seem the most promising in terms of improving crop productivity. In this regard, the positive impact of the construction of trading depots during the early phase of the project or purchase of staple crops would need further support in order to have a sustainable impact on the livelihoods of smallholder farmers. That, however, contradicts the reorientation of the project towards cash crops. This also implies that such interventions would play an important role in enhancing productivity and closing productivity gaps. Moreover, AAZ programme participation has heterogeneous effects on participants: Programme participation benefited more those with low productivity levels than those with higher productivity. Further consultation with Amatheon is necessary to find out whether supporting the least productive is financially beneficial and if there are fewer benefits to working with larger farmers. Recent emerging empirical evidence suggests that large-scale agricultural intervention programmes that integrate smallholders in their value chain have the potential to help alleviate hunger and malnutrition among smallholder farmers by improving productivity and market participation simultaneously. Our paper indicates that, although the effects are not very strong, there is some suggestive evidence of the programme’s effect on improving the nutritional outcomes of women in participating households. However, further research would be necessary to confirm the pathways through which these effects occur exactly, for example nutritional education programmes, production diversification or local employment effects. Our analysis of the inter-household distributional impacts of participation in the programme suggests that the programme had the largest effects on those who were extremely food-insecure. As highlighted in the results section, the introduction of new nutritious crops (or crop diversification during phase two), nutrition-related extension services, off-farm employment (better rural-urban 10 The qualitative interview suggests that a change or shift in project focus (from cereals to oilseeds); a deterioration or malfunctioning of the constructed trading depots during the follow-up period; a continued change in the design of the project; and a lack of an institutional/proper contract between farmers and the company are some of the factors that might have contributed to such a low effect or none at all. Since the lack of a proper contract between farmers and AAZ is mentioned as one reason for the constraints limiting commercialisation, exploring the effects of the recent shift to quinoa contract farming would be interesting. 11 We analysed potential pathways through which AAZ interventions affect productivity and nutritional outcomes. Two such pathways we considered are adoption of improved agricultural technologies and the uptake of CA practices. IDOS Discussion Paper 19/2022 56 linkages), the improved productivity of low productive groups and increased joint decisionmaking in the household regarding the allocation of land for crops grown are some of the potential mechanisms (qualitative interviews pointed to these factors). Analysis of AAZ programme components also suggests that complementing input and output market interventions with nutrition education in the training package could yield significant impacts on improving household food security in general, and women’s consumption of micronutrients in particular. General remarks: Our analyses confirm that large-scale agricultural investments that integrate smallholders, with all their risks, can offer opportunities for the farmers or communities that they are operating in. They would be appropriate for enhancing the production of important crops and improving sustainability via enhancing the adoption of sustainable agricultural practices and providing input credit, as well as establishing processing facilities in rural areas where such infrastructure is often lacking, thus improving the welfare of households. However, for their benefits to outweigh their drawbacks, the appropriate policies and institutions (for instance, in the areas of land policies, contract enforcement, the monitoring and evaluation of project implementation, clarifying the rules) that shape such investments in a development-friendly way are vital. In addition, a strong political commitment from the hosting government is needed, not only to protect the vulnerable farmers, but also to attract and promote more developmentfriendly, foreign-based, large-scale agricultural investments. Moreover, support should be provided to farmers and local communities where such investments are operating to negotiate for win-win outcomes. Finally, it is important to acknowledge some of the limitations of this paper. The first is related to the caveat of the analysis: the missing plausible test of the parallel trends assumption in using DID. The results might be due to diverging trends in the outcome variables of interest between the AAZ outgrower programme participants and non-participants prior to AAZ intervention. As such, our results should be interpreted with caution and may not necessarily imply causality. Second, although our employed methods enable us to control for various household and community characteristics and remove time-invariant effects, it should be noted that unobserved time-variant factors might bias the estimates. Third, although we included exposure to depots and proximity to AAZ farm blocks, spillover effects from the programme or contamination to nonparticipating households are highly likely. This would, in turn, underestimate the true effects of the programme. Future research that analyses the extent of such spillovers would be important to understand the complete benefits of such programmes. Fourth, although a high attrition rate is expected in panel surveys such as this, it is higher in our dataset (we failed to track about 20 per cent of the baseline participants) than in the standard panel surveys. As highlighted in the data section, we conducted attrition-bias tests and carefully included variables that explain the attrition rate in our regression analyses to reduce the potential biases. Finally, future interventions interested in scaling out of such intervention models should take into account the external validity of this paper. And future studies considering wider geographic coverage (ours is only two districts) and agro-ecological conditions would give a clearer picture of the effects of private outreach programmes (or nucleus-outgrower schemes) on the welfare of smallholder farmers. In this regard, an interesting hypothesis to be tested by widening the scope of future studies would be if interventions of nucleus large farms would likely be better targeting the needs of smallholders (they may be better able to understand them than pure large processors or traders). IDOS Discussion Paper 19/2022 57 References Abadie, A., Drukker, D., Herr, J. L., & Imbens, G. W. (2004). Implementing matching estimators for average treatment effects in Stata. The Stata Journal, 4(3), 290-311. Abadie, A., & Imbens, G. W. (2011). Bias-corrected matching estimators for average treatment effects. Journal of Business & Economic Statistics, 29(1), 1-11. https://doi.org/10.1198/jbes.2009.07333 Amatheon Agri. (2013). Annual overview 2012. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2014). Annual overview 2013. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2015). Annual overview 2014. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2016). Annual overview 2015. Retrieved from http://www.amatheon-agri.com/publicmedia/Amatheon-Agri-2015-Annual-Overview.pdf Amatheon Agri. (2017). Annual overview 2016. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2018). Annual overview 2017. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2019). Annual report 2018. Retrieved from https://amatheon-agri.com/ Amatheon Agri. (2021, 13 January). First outgrower quinoa harvest kicks off. Retrieved from http://www.amatheon-agri.com/news/first-outgrower-quinoa-harvest-kicks-off Anseeuw, W., Boche, M. Breu, T., Giger, M., Lay, J., Messerli, P., & Nolte, K. (2013). Transnational land deals for agriculture in the Global South. Analytical report based on the land matrix database. Retrieved from https://www.semanticscholar.org/paper/Transnational-land-deals-for-agriculture-inthe-%3A-Anseeuw-Boche/e41239449b50b476777457a6ad23d7d082dc1bc8 Arslan, A., McCarthy, N., Lipper, L., Asfaw, S., & Cattaneo, A. (2014). Adoption and intensity of adoption of conservation farming practices in Zambia. Agriculture, Ecosystems & Environment, 187, 72-86. https://doi.org/10.1016/j.agee.2013.08.017 Barrett, C. B., Bachke, M. E., Bellemare, M. F., Michelson, H. C., Narayanan, S., & Walker, T. F. (2012). Smallholder participation in contract farming: Comparative evidence from five countries. World Development, 40(4), 715-730. http://dx.doi.org/10.1016/j.worlddev.2011.09.006 Baumgartner, P., von Braun, J., Abebaw, D., & Müller, M. (2015). Impacts of large-scale land investments on income, prices, and employment: Empirical analyses in Ethiopia. World Development, 72, 175190. https://doi.org/10.1016/j.worlddev.2015.02.017 Bellemare, M. F., & Novak, L. (2017). Contract farming and food security. American Journal of Agricultural Economics, 99(2), 357-378. https://doi.org/10.1093/ajae/aaw053 Biggeri, M., Burchi, F., Ciani, F., & Herrmann, R. (2018). Linking small-scale farmers to the durum wheat value chain in Ethiopia: Assessing the effects on production and wellbeing. Food Policy, 79, 77-91. https://doi.org/10.1016/j.foodpol.2018.06.001 Bluwstein, J., Lund, J. F., Askew, K., Stein, H., Noe, C., Odgaard, R., …Engström, L. (2018). Between dependence and deprivation: The interlocking nature of land alienation in Tanzania. Journal of Agrarian Change, 18(4), 806-830. https://doi.org/10.1111/joac.12271 Borras, S. M. J., & Franco, J. C. (2012). Global land grabbing and trajectories of agrarian change: A preliminary analysis. Journal of Agrarian Change, 12(1), 34-59. Bottazzi, P., Crespo, D., Bangura, L. O., & Rist, S. (2018). Evaluating the livelihood impacts of a largescale agricultural investment: Lessons from the case of a biofuel production company in northern Sierra Leone. Land Use Policy, 73, 128-137. Bruna, N. (2019). Land of plenty, land of misery: Synergetic resource grabbing in Mozambique. Land, 8(8), 113. https://doi.org/10.3390/land8080113 IDOS Discussion Paper 19/2022 58 Brüntrup, M., Anders, D., Herrmann, R., Schmitz, J., & Kaup, F. (2010). New paths for biofuels in Africa. Retrieved from http://www.rural21.com/fileadmin/_migrated/content_uploads/rural_2010_6_2528_01.pdf Brüntrup, M., Schwarz, F., Absmayr, T., Dylla, J., Eckhard, F., Remke, K., & Sternisko, K. (2018). Nucleusoutgrower schemes as an alternative to traditional smallholder agriculture in Tanzania – strengths, weaknesses and policy requirements. Food Security, 10(4), 807-826. https://doi.org/10.1007/s12571-0180797-0 Callaway, B., & Li, T. (2019). Quantile treatment effects in difference in differences models with panel data. Quantitative Economics, 10(4), 1579-1618. Corbeels, M., Naudin, K., Whitbread, A. M., Kühne, R., & Letourmy, P. (2020). Limits of conservation agriculture to overcome low crop yields in sub-Saharan Africa. Nature Food, 1(7), 447-454. https://doi.org/10.1038/s43016-020-0114-x Cotula, L., Vermeulen, S., Leonard, R., & Keeley, J. (2009). Land grab or development opportunity? Agricultural investment and international land deals in Africa. Retrieved from https://www.ifad.org/documents/10180/a2ea06a0-d0b2-4e99-a9b4-1f23e0491afb De Schutter, O. (2011). How not to think of land-grabbing: Three critiques of large-scale investments in farmland. The Journal of Peasant Studies, 38(2), 249-279. https://doi.org/10.1080/03066150.2011.559008 Deininger, K., Byerlee, D., Lindsay, J., Norton, A., Selod, H., & Stickler, M. (2011). Rising global interest in farmland. Can it yield sustainable and equitable benefits? Retrieved from https://siteresources.worldbank.org/DEC/Resources/Rising-Global-Interest-in-Farmland.pdf Deininger, K., & Xia, F. (2016). Quantifying spillover effects from large land-based investment: The case of Mozambique. World Development, 87, 227-241. http://dx.doi.org/10.1016/j.worlddev.2016.06.016 Engström, L., & Hajdu, F. (2019). Conjuring “win-world” – resilient development narratives in a large-scale agro-investment in Tanzania. The Journal of Development Studies, 55(6), 1201-1220. https://doi.org/10.1080/00220388.2018.1438599 FAO (Food and Agriculture Organization). (2014a). What is conservation agriculture? Retrived from http://www.fao.org/ag/ca/1a.html FAO (2014b). CA adoption worldwide. FAO-CA website. Retrived from https://www.fao.org/conservationagriculture/en/ Furno, M., & Caracciolo, F. (2020). Multi-valued double robust quantile treatment effect. Empir Econ, 58, 2545-2571. https://doi.org/10.1007/s00181-018-1584-7 Gebel, M., & Voßemer, J. (2014). The impact of employment transitions on health in Germany. A difference-in-differences propensity score matching approach. Social Science and Medicine, 108, 128-136. https://doi.org/10.1016/j.socscimed.2014.02.039 Giller, K. E., Andersson, J. A., Corbeels, M., Kirkegaard, J., Mortensen, D., Erenstein, O., & Vanlauwe, B. (2015). Beyond conservation agriculture. Frontiers in Plant Science, 6(870). https://doi.org/10.3389/fpls.2015.00870 Giller, K. E., Witter, E., Corbeels, M., & Tittonell, P. (2009). Conservation agriculture and smallholder farming in Africa: The heretics’ view. Field Crops Research, 114(1), 23-34. https://doi.org/10.1016/j.fcr.2009.06.017 Goldsmith, A. (1985). The private sector and rural development: Can agribusiness help the small farmer? World Development, 13(10-11), 1125-1138. http://dx.doi.org/10.1016/0305-750X(85)90031-2 Govereh, J., & Jayne, T. S. (2003). Cash cropping and food crop productivity: Synergies or trade-offs? Agricultural Economics, 28(1), 39-50. https://doi.org/10.1111/j.1574-0862.2003.tb00133.x Grosh, B. (1994). Contract farming in Africa: An application of the new institutional economics. Journal of African Economies, 3(2), 231-261. https://doi.org/10.1093/oxfordjournals.jae.a036805 IDOS Discussion Paper 19/2022 59 Haggblade, S., & Tembo, G. (2003) Conservation farming in Zambia (EPTD Discussion Paper 108). Washington, DC: International Food Policy Research Institute. Hall, R., Scoones, I., & Tsikata, D. (2017). Plantations, outgrowers and commercial farming in Africa: Agricultural commercialisation and implications for agrarian change. The Journal of Peasant Studies, 44(3), 515-537. https://doi.org/10.1080/03066150.2016.1263187 Herrmann, R. (2017). Large-scale agricultural investments and smallholder welfare: A comparison of wage labor and outgrower channels in Tanzania. World Development, 90, 294-310. http://dx.doi.org/10.1016/j.worlddev.2016.10.007 Herrmann, R., & Grote, U. (2015). Large-scale agro-industrial investments and rural poverty: Evidence from sugarcane in Malawi. Journal of African Economies, 24(5), 645-676. https://doi.org/10.1093/jae/ejv015 Herrmann, R., Jumbe, C., Brüntrup, M., & Osabuohien, E. (2018). Competition between biofuel feedstock and food production: Empirical evidence from sugarcane outgrower settings in Malawi. Biomass and Bioenergy, 114, 100-111. https://doi.org/10.1016/j.biombioe.2017.09.002 Hirschl, B., Dietz, K., Backhouse, M., Dunkelberg, E., Herrmann, R., & Vogelpohl, T. (2014). Biokraftstoffe zwischen Sackgasse und Energiewende: Sozial-ökologische und transnationale Perspektiven. Munich: oekom Verlag. Kassam, A., Friedrich, T., Shaxson, F., & Pretty, J. (2009). The spread of conservation agriculture: Justification, sustainability and uptake. International Journal of Agricultural Sustainability, 7(4), 292320. https://doi.org/10.3763/ijas.2009.0477 Key, N., & Runsten, D. (1999). Contract farming, smallholders, and rural development in Latin America: The organization of agroprocessing firms and the scale of outgrower production. World Development, 27(2), 381-401. http://dx.doi.org/10.1016/S0305-750X(98)00144-2 Khadjavi, M., Sipangule, K., & Thiele, R. (2020). Social capital and large-scale agricultural investments: An experimental investigation. The Economic Journal, 131(633), 420-449. https://doi.org/10.1093/ej/ueaa050 Khandker, S. R., Koolwal, G. B., & Samad, H. A. (2009). Handbook on impact evaluation: Quantitative methods and practices. Washington, DC: World Bank Publications. Kunzekweguta, M., Rich, K. M., & Lyne, M. C. (2017). Factors affecting adoption and intensity of conservation agriculture techniques applied by smallholders in Masvingo district, Zimbabwe. Agrekon, 56(4), 330-346. https://doi.org/10.1080/03031853.2017.1371616 Land Matrix Global Observatory – LMGO (2019). Retrieved from https://landmatrix.org/ Lay, J., Nolte, K., & Sipangule, K. (2020). Large-scale farms in Zambia: Locational patterns and spillovers to smallholder agriculture. World Development, 140, 105277. https://doi.org/10.1016/j.worlddev.2020.105277 Little, P. D., & Watts, M. J. (1994). Living under contract: Contract farming and agrarian transformation in sub-Saharan Africa. Madison, WI: The University of Wisconsin Press. Liverpool-Tasie, L.S.O., et al. (2020). A scoping review of market links between value chain actors and small-scale producers in developing regions. Nat Sustain 3, 799-808. https://doi.org/10.1038/s41893020-00621-2 Maertens, M., & Vande Velde, K. (2017). Contract-farming in staple food chains: The case of rice in Benin. World Development, 95, 73-87. https://doi.org/10.1016/j.worlddev.2017.02.011 Matenga, C. R., & Hichaambwa, M. (2017). Impacts of land and agricultural commercialisation on local livelihoods in Zambia: Evidence from three models. The Journal of Peasant Studies, 44(3), 574-593. https://doi.org/10.1080/03066150.2016.1276449 Negash, M., & Swinnen, J. F. M. (2013). Biofuels and food security: Micro-evidence from Ethiopia. Energy Policy, 61, 963-976. https://doi.org/10.1016/j.enpol.2013.06.031 IDOS Discussion Paper 19/2022 60 Ng’ombe, J. N., Kalinda, T. H., & Tembo, G. (2017). Does adoption of conservation farming practices result in increased crop revenue? Evidence from Zambia. Agrekon, 56(2), 205-221. https://doi.org/10.1080/03031853.2017.1312467 Ngoma, H., Mason, N. M., & Sitko, N. J. (2015). Does minimum tillage with planting basins or ripping raise maize yields? Meso-panel data evidence from Zambia. Agriculture, Ecosystems & Environment, 212, 21-29. https://doi.org/10.1016/j.agee.2015.06.021 Ngwira, A. R., Thierfelder, C., Eash, N., & Lambert, D. M. (2013). Risk and maize-based cropping systems for smallholder Malawi farmers using conservation agriculture technologies. Experimental Agriculture, 49(4), 483-503. https://doi.org/10.1017/S0014479713000306 Nolte, K., & Ostermeier, M. (2017). Labour market effects of large-scale agricultural investment: Conceptual considerations and estimated employment effects. World Development, 98, 430-446. http://dx.doi.org/10.1016/j.worlddev.2017.05.012 Nolte, K., & Subakanya, M. (2016). Relationship between large-scale agricultural investors and local communities: Lessons from two investments in Zambia. Retrieved from http://www.iapri.org.zm/images/PolicyBriefs/ps_79.pdf Nolte, K., & Väth, S. J. (2015). Interplay of land governance and large-scale agricultural investment: Evidence from Ghana and Kenya. The Journal of Modern African Studies, 53(1), 69-92. https://doi.org/10.1017/S0022278X14000688 Osabuohien, E. S., Efobi, U. R., Herrmann, R. T., & Gitau, C. M. W. (2019). Female labor outcomes and large-scale agricultural land investments: Macro-micro evidencefrom Tanzania. Land Use Policy, 82, 716-728. https://doi.org/10.1016/j.landusepol.2019.01.005 Ragasa, C., Lambrecht, I., & Kufoalor, D. S. (2018). Limitations of contract farming as a pro-poor strategy: The case of maize outgrower schemes in Upper West Ghana. World Development, 102, 30-56. https://doi.org/10.1016/j.worlddev.2017.09.008 Robertson, B., & Pinstrup-Andersen, P. (2010). Global land acquisition: Neo-colonialism or development opportunity? Food Security, 2(3), 271-283. https://doi.org/10.1007/s12571-010-0068-1 Rodenburg, J., Büchi, L., & Haggar, J. (2020). Adoption by adaptation: Moving from conservation agriculture to conservation practices. International Journal of Agricultural Sustainability, 19(5-6), 437455. https://doi.org/10.1080/14735903.2020.1785734 Schoneveld, G. C. (2014). The geographic and sectoral patterns of large-scale farmland investments in sub-Saharan Africa. Food Policy, 48, 34-50. http://dx.doi.org/10.1016/j.foodpol.2014.03.007 Sibhatu, K. T., Arslan, A., & Zucchini, E. (2022). The effect of agricultural programs on dietary diversity and food security: Insights from the smallholder productivity promotion program in Zambia. https://doi.org/10.1016/j.foodpol.2022.102268 Sipangule, K., & Lay, J. (2015). The impact of foreign large-scale land acquisitions on smallholder productivity: Evidence from Zambia (AGRODEP Working Paper 0011). Retrieved from https://www.ifpri.org/publication/impact-foreign-large-scale-land-acquisitions-smallholderproductivity-evidence-zambia Sitko, N. J., & Jayne, T. S. (2014). Structural transformation or elite land capture? The growth of “emergent” farmers in Zambia. Food Policy, 48, 194-202. http://dx.doi.org/10.1016/j.foodpol.2014.05.006 Sivramkrishna, S., & Jyotishi, A. (2008). Monopsonistic exploitation in contract farming: Articulating a strategy for grower cooperation. Journal of International Development, 20(3), 280-296. https://doi.org/10.1002/jid.1411 Smalley, R. (2013). Plantations, contract farming and commercial farming areas in Africa: A comparative review. Brighton: Future Agricultures Consortium. Songwe, V., & Deininger, K. (2009). Foreign investment in agricultural production: Opportunities and challenges (Agricultural and Rural Development Notes, No. 45.) Retrieved from https://openknowledge.worldbank.org/handle/10986/9501 IDOS Discussion Paper 19/2022 61 Sulle, E. (2017). Social differentiation and the politics of land: Sugar cane outgrowing in Kilombero, Tanzania. Journal of Southern African Studies, 43(3), 517-533. https://doi.org/10.1080/03057070.2016.1215171 Tambo, J. A., & Mockshell, J. (2018). Differential impacts of conservation agriculture technology options on household income in sub-Saharan Africa. Ecological Economics, 151, 95-105. https://doi.org/10.1016/j.ecolecon.2018.05.005 Villa, J. M. (2016), Diff: Simplifying the estimation of difference-in-differences treatment effects. The Stata Journal, 16(1), 52-71. Westengen, O. T., Nyanga, P., Chibamba, D., Guillen-Royo, M., & Banik, D. (2018). A climate for commerce: The political agronomy of conservation agriculture in Zambia. Agriculture and Human Values, 35(1), 255-268. https://doi.org/10.1007/s10460-017-9820-x White, B., Borras Jr, S. M., Hall, R., Scoones, I., & Wolford, W. (2012). The new enclosures: Critical perspectives on corporate land deals. The Journal of Peasant Studies, 39(3-4), 619-647. https://doi.org/10.1080/03066150.2012.691879 Wineman, A., et al. (2021). Characteristics and spillover effects of medium-scale farms in Tanzania. Eur J Dev Res, 33, 1877-1898. https://doi.org/10.1057/s41287-020-00323-7 Zulu-Mbata, O., & Chapoto, A. (2016). Conservation agriculture: Gendered impacts on households' livelihoods. Zambia Social Science Journal, 6(2), 7. IDOS Discussion Paper 19/2022 62 Appendices∗ Appendix A Table A1: Definition of variables used in the regression analyses Variables Description HH characteristics HH age Age of the HH head (years) Female-headed HH =1 if HH head is female HH education (head) Education of the HH head (years of formal education) HH size No. of HH members (size) Edu. max (within HH) Maximum education for any HH member (years of formal education) Head is married =1 if HH head is married Total cultivated crop, ha Total cultivated land for crop production (ha) Cultivated maize, ha Total cultivated land for maize production (ha) Cultivated soya, ha Total cultivated land for soybeans (ha) Wealth indices Agricultural asset index (pca), incl. tractor/oxen Agricultural asset index computed using pca, including tractor/oxen Agricultural asset index, only tractor/oxen/plough (pca) Agricultural asset index, only tractor/oxen plough (using pca) Agricultural asset index (pca), w/o tractor/oxen Agricultural asset index, without tractor/oxen HH wealth index (pca), incl. motorbike/cars HH wealth index (pca), including motorbikes or cars HH wealth index (pca), w/o motorbike/cars HH wealth index (pca), without motorbikes or cars HH experienced a shock =1 if a HH experienced a shock in the past five years HH irrigated a field =1 if a HH irrigated a field HH accessed a loan other than from AAZ =1 if a HH has accessed a loan other than from AAZ Land variables Per capita farm size Per capita farm size (ha) HH rented at least one field =1 if a HH rented land for at least one field Kinship ties Head or spouse is related to the chief =1 if a HH or spouse is related to the chief Head or spouse is related to the headman =1 if a HH or spouse is related to the headman Head or spouse is related to the farmer coordinator =1 if a HH or spouse is related to the farmer coordinator Head or spouse is related to the chief or headman =1 if a HH or spouse is related to the chief or headman ∗ Authors are the source for all tables in Appendices A and B, unless otherwise indicated. IDOS Discussion Paper 19/2022 63 Table A2: AAZ programme participation dynamics, unbalanced panel HHs All districts Mumbwa Chibombo Variables N % N % N % Households with a member that has received training/advice from AAZ (baseline), N=239 239 100% Share of HHs that received training from AAZ in 2014 239 31% Share of HHs that received training from AAZ in 2015 239 32% Share of HHs that received training from AAZ in 2016 239 41% Share of HHs that received training from AAZ in 2017 239 11% Share of HHs that received training from AAZ in 2018 239 1% No. of years HH has been trained by AAZ 239 1.17 Households with a member that has received training or advice from AAZ (follow-up) 395 100% 212 100% 183 100% Share of HHs that received training or advice from AAZ in 2016 244 16% 148 24% 96 4% Share of HHs that received training or advice from AAZ in 2017 244 24% 148 26% 96 20% Share of HHs that received training or advice from AAZ in 2018 244 59% 148 50% 96 73% Share of HHs that received training or advice from AAZ in 2019 244 29% 148 34% 96 21% Share of HHs that received training or advice from AAZ in 2020 244 14% 148 23% 96 0% Share of HHs that received training or advice from AAZ in 2021 244 8% 148 14% 96 0% Average no. of years HH has been trained/advised by AAZ 244 1.5 148 1.7 96 1.18 Household bought inputs from AAZ (baseline), N=75 75 100% 66 100% 9 100% Purchased inputs from AAZ in 2013/14, baseline 75 8% 66 9% 9 0% Purchased inputs from AAZ in 2014/15, baseline 75 21% 66 24% 9 0% Purchased inputs from AAZ in 2015/16, baseline 75 52% 66 58% 9 11% Purchased inputs from AAZ in 2016/17, baseline 75 28% 66 29% 9 22% Purchased inputs from AAZ in 2017/18, baseline 75 23% 66 15% 9 78% No. of years purchased inputs from AAZ, 2014-18 75 1.32 66 1.35 9 1.11 Household bought inputs from AAZ (follow-up), N=87 87 100% 64 100% 23 100% Dummy=1 if purchased AAZ inputs in 2016/17 87 51% 64 59% 23 26% Dummy=1 if purchased AAZ inputs in 2017/18 87 24% 64 17% 23 43% Dummy=1 if purchased AAZ inputs in 2018/19 87 15% 64 9% 23 30% Dummy=1 if purchased AAZ inputs in 2019/20 87 15% 64 17% 23 9% Dummy=1 if purchased AAZ inputs in 2020/21 87 17% 64 22% 23 4% Number of years HH has purchased inputs from AAZ 81 1.31 59 1.36 22 1.18 IDOS Discussion Paper 19/2022 70 Table R13B5: Effects of AAZ participation on food and nutrition security and off-farm employment, using k-nearest neighbours matching (1) (2) (3) (4) (5) HH food insecurity score Minimum dietary diversity score for women of reproductive age Dummy=1 if women’s diet in HH is micronutrient adequate Number of HHs participating in off-farm employment 1 if HH member works for pay between 1 May to 30 April ATT_ PSM1 0.005 (0.406) 0.089 (0.109) 0.040 (0.050) 0.149** (0.045) 0.081*** (0.021) Observations 579 433 433 579 579 ATT_ PSM2 0.322 (0.490) 0.093 (0.135) 0.080* (0.053) 0.028 (0.048) 0.016 (0.036) Observations 570 415 415 570 570 Note: ATT_PSM1 and ATT_PSM2 refers to ATT estimated using PSM based on wave 1 and wave 2, respectively. Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001 Table R13B6: Effects of AAZ participation on the adoption of other crops, using k-nearest neighbours matching (1) (2) (3) (4) (5) Produced legumes Produced sunflower Produced cowpeas Produced groundnuts Used OPV/hybrid groundnut seed ATT_ PSM1 0.050* (0.030) 0.027 (0.026) 0.086*** (0.023) 0.025 (0.044) 0.104* (0.052) constant 0.836*** (0.024) 0.082*** (0.022) 0.015 (0.018) 0.431*** (0.036) 0.111** (0.043) Observations 579 579 579 579 253 ATT_ PSM2 -0.000 (0.008) 0.003 (0.033) 0.018 (0.018) 0.016 (0.049) 0.009 (0.022) constant 0.007 (0.007) 0.135*** (0.029) 0.021 (0.016) 0.453*** (0.042) 0.016 (0.019) Observations 570 570 570 567 264 Note: ATT_PSM1 and ATT_PSM2 refers to ATT estimated using PSM based on wave 1 and wave 2, respectively. Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001