Exploring the advantages and drivers of sustainable agricultural practices in Central Asia

Tadjiev, Abdusame Doctoral Thesis Exploring the advantages and drivers of sustainable agricultural practices in Central Asia Suggested Citation: Tadjiev, Abdusame (2024) : Exploring the advantages and drivers of sustainable agricultural practices in Central Asia, Universitätsund Landesbibliothek Sachsen-Anhalt, Halle (Saale), https://nbn-resolving.de/urn:nbn:de:gbv:3:4-1981185920-1211646 This Version is available at: https://hdl.handle.net/10419/319601 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/ EXPLORING THE ADVANTAGES AND DRIVERS OF SUSTAINABLE AGRICULTURAL PRACTICES IN CENTRAL ASIA Dissertation Zur Erlangung des Doktorgrades der Agrarwissenschaften (Dr.agr.) der Naturwissenschaftlichen Fakultät III Agrar‐ und Ernährungswissenschaften, Geowissenschaften und Informatik der Martin‐Luther‐Universität Halle‐Wittenberg vorgelegt von Herrn ABDUSAME TADJIEV Gutachter: Prof. Dr. Thomas Herzfeld Prof. Dr. Martin Petrick Dr. Nodir Djanibekov Tag der Verteidigung: 23.09.2024 DEDICATION This dissertation is dedicated to the cherished memory of my grandfathers, TADJIEV ABDURAHMONKHON and AKBAROV UZBEKHON, and my grandmothers, AMIROVA ROBIYAKHON and TURSUNOVA ZULFIYAKHON. ~ i ~ ACKNOWLEDGMENTS I am deeply grateful to a number of people whose unwavering support and invaluable contributions have made the completion of this Ph.D. dissertation. Their guidance, encouragement, and assistance have profoundly shaped my PhD journey and enriched the outcome of this research. First of all, I would like to thank my supervisors Prof. Dr. Thomas Herzfeld and Dr. Nodir Djanibekov who accepted me as a Ph.D. student in the framework of SUSADICA project and for their continuous guidance and support throughout the completion of this thesis. A special thanks to Dr. Nodir Djanibekov for his invaluable encouragement, steady moral support, and insightful guidance, all of which have been instrumental in enabling me to advance with this thesis in a conducive and stress-free manner. He engaged me in active participation within the SUSADICA project, thereby aiding in the enhancement of my capabilities. Moreover, Nodir was the one by my side and supportive on my initial day in Halle. Commencing my PhD at IAMO, I initially planned a field experiment in Uzbekistan, in addition, prior to my involvement in the SUSADICA project, I had participated in the AGRICHANGE project, wherein I had intended to defend my first PhD in my home country, Uzbekistan. However, COVID-19 travel restrictions prevented both my plans. Despite facing significant challenges, with support from Thomas and Nodir, I managed to adapt my PhD plans and defended my thesis remotely from IAMO, as well as remained connected with IAMO during the pandemic. I am grateful to them. Additionally, I extend thanks to Prof. Dr. Shavkat Hasanov, who is my local supervisor, and the team at the "Tashkent Institute of Irrigation and Agricultural Mechanization Engineers" National Research University (TIIAME NRU) for their support during my research visit in 2019 and their assistance in conducting the online defense during the pandemic. I gratefully acknowledge the funding made available by VolkswagenStiftung under the Doctoral Program for Sustainable Agricultural Development in Central Asia (SUSADICA, Grant Number 96264) within the funding initiative ‘‘Between Europe and the Orient – A Focus on Research and ~ ii ~ Higher Education in/on Central Asia and the Caucasus”. Furthermore, I would like to thank SUSADICA project members Dr. Nozilakhon Mukhamedova, Prof. Dr. Daniel Müller, Prof. Dr. Insa Theesfeld, Dr. Ilkhom Soliev, Zafar Kurbanov, Dr. Barchynai Kimsanova, Atabek Umirbekov, Daniela Peña Guerrero, Suray Charyyeva, Davron Niyazmetov, Hannes Knorr, Jakhongir Babadjanov and Kadyrbek Sultakeev. The insightful feedback garnered from project workshops and related events has significantly contributed to enhancing the quality and depth of this research output. My special gratitude goes to Prof. Dr. Shavkat Hasanov for his comments, moral support, encouragement, and intellectually stimulating discussions, as well as for his aid during my field trip in Uzbekistan before pandemic. His support has been instrumental in guiding me throughout the entirety of this dissertation. I also thank to Prof. Dr. Olim Murtazaev, Dr. Farhod Ahrorov, and Dr. Ibrohim Ganiev for their encouragement and support, which played a pivotal role in my decision to pursue doctoral studies. Additionally, I thank to Dr. Damir Esenaliev for his invaluable comments, enhancing the content of the third chapter of this dissertation. I would also like to acknowledge Dr. Arjola Arapi-Gjini for her expertise, which proved instrumental in refining the methodology incorporated within the fourth Chapter of this dissertation. I express my gratitude to the entire Leibniz Institute of Agricultural Development in Transition Economies (IAMO) team for their ongoing and substantial assistance throughout my research journey. Their support has extended to various facets including administrative procedures, access to literature with all possible means, IT, field-trips, conferences, accommodation, reimbursements, open access, language editing, and many more matters. I also thank to all other colleagues and friends (both friends from IAMO and Uzbekistan) whose direct and indirect contributions have significantly contributed to the successful outcome of this work. Finally, I extend my sincere gratitude to my father, Tadjiev Abduhamid, my mother, Akbarova Shohista, and my extended family members including sisters and brothers. Their unwavering ~ iii ~ love and support throughout the passage of time, whether from afar or nearby, has been immeasurable. Furthermore, my special thanks to my wife and children. Despite being far from me, their patience, love, trust and support give me moral strength. I extend my utmost gratitude to them! ~ iv ~ SUMMARY Improving soil productivity and agricultural outputs pose significant challenges in developing countries, including Central Asia. Inadequate land use during the Soviet times, coupled with the absence of a structured land management system during the period, engendered numerous issues including cropland degradation that exerts a substantial impact on agricultural productivity in the Central Asian countries. The adoption of sustainable agricultural practices stands as a crucial remedy to address these issues. Despite the comprehensive coverage within global literature regarding the benefits of sustainable agricultural practices, there persists a marked discrepancy in their adoption levels by farmers. Furthermore, there is a scarcity of empirical investigations explaining the primary drivers and the impacts associated with the adoption of sustainable agricultural practices in Central Asia. From this perspective, the overarching aim of this doctoral dissertation is to gain deeper insights into the factors that facilitate the adoption of selected sustainable agricultural practices among farmers in various settings of Central Asia. The thesis comprises five chapters, incorporating three empirical sections. The initial chapter introduces to the general problem background pertaining to the issue of sustainable agricultural development in Central Asia and the key research questions of the PhD dissertation. Empirical findings are presented in the second, third and fourth chapters. A general conclusion is given in Chapter 5. The second chapter investigates the drivers of farmers’ decision to adopt crop rotation and how its adoption impacts farmers’ cotton yields and net returns in two contrasting settings of Central Asia by applying an endogenous switching regression model to cross-sectional survey data collected from 592 cotton growers in 2019 in Kazakhstan and Uzbekistan. Cotton monoculture inherited from the former Soviet cultivation system led to the decline of soil fertility and reduced cotton yields in irrigated areas of Central Asia. Adopting a diversified crop rotation approach is a viable solution to maintain soil quality and long-term economic benefits. The chapter findings ~ v ~ highlight these two countries' differing institutional contexts surrounding cotton farming. Kazakhstani farmers' decision to adopt crop rotation is positively related to age, participation in farm training, the farmer's opinion about the quality of the irrigation canal, and the share of adopters in a village. In Uzbekistan, farmers who perceive greater land tenure security are more inclined to adopt crop rotation. In Uzbekistan, employing crop rotation leads to higher cotton yields compared to traditional crop cultivation methods. In Kazakhstan, cotton farmers experience a contrasting outcome. Employing an endogenous switching regression model on the plot-level panel data of 878 of Kyrgyzstan’s smallholders, the third chapter investigates the determinants of the decision to adopt zero tillage and its effect on smallholders’ production costs. The chapter finds that the probability of zero tillage adoption is associated with employment in agriculture, assets, agricultural shocks, fertilizer use, number of plots and average distances from the dwelling to household fields and to the main road. Furthermore, the chapter indicates that zero tillage adoption decreases land preparation costs by 23%, but increases hired labor and herbicide costs by 13% and 15%, respectively compared to the conventional tillage method. Nevertheless, zero tillage can reduce total production costs by 15%. The third chapter confirms that zero tillage can be promoted as an option for resource-scarce smallholders, especially those in remote areas with poor access to inputs and machinery services. Promoting zero tillage adoption as a laborsaving or herbicide reducing practice can create false expectations among smallholders. The fourth chapter investigates the question of how participation in informal cooperation in water management influences the intensity of the adoption of sustainable agricultural practices by using two-years of 2019 and 2022 farm survey data of Uzbekistan and employing a marginal treatment effects model. The results show that farmers who are likely to participate in informal cooperation in water management tend to benefit more from the participation in terms of higher adoption intensity of sustainable agricultural practices. Finally, the fifth chapter synthesizes the research findings, summarizes the policy implications along with research limitations and provides ideas for future research. ~ vi ~ TABLE OF CONTENTS ACKNOWLEDGMENTS ........................................................................................................... i SUMMARY ......................................................................................................................... iv TABLE OF CONTENTS........................................................................................................... vi LIST OF FIGURES ............................................................................................................... viii LIST OF TABLES ................................................................................................................... ix LIST OF TABLES IN THE APPENDIX ......................................................................................... x ABBREVIATIONS ................................................................................................................. xi 1 GENERAL INTRODUCTION ................................................................................................ 1 1.1 Challenges for sustainable agricultural development in Central Asia ................................ 1 1.2 Sustainable agricultural practices ....................................................................................... 2 1.3 Problem statement, research objectives and structure of the thesis ................................ 5 2 DETERMINANTS AND IMPACTS OF CROP ROTATION ADOPTION AMONG COTTON GROWERS IN IRRIGATED AREAS OF KAZAKHSTAN AND UZBEKISTAN ............................... 11 2.1 Reforms in the cotton sector and the adoption of crop rotation practices in Kazakhstan and Uzbekistan .............................................................................................. 11 2.2 Data and descriptive analysis ............................................................................................ 13 2.3 Methodological approach ................................................................................................. 18 2.3.1 Crop rotation adoption decision and farm outcome .............................................. 19 2.3.2 Endogenous switching regression .......................................................................... 22 2.3.3 Average treatment effects ...................................................................................... 24 2.4 Results and discussion ...................................................................................................... 26 2.4.1 Determinants of farmers’ decision to adopt crop rotation .................................... 26 2.4.2 Determinants of cotton yields and net returns of cotton growers ........................ 29 2.4.3 Cotton yield and net returns impacts of the adoption of crop rotation ................ 38 3 DOES ZERO TILLAGE SAVE OR INCREASE PRODUCTION COSTS OF SMALLHOLDERS IN KYRGYZSTAN? ........................................................................................................... 41 3.1 Smallholders’ challenges in the adoption of conservation agriculture in Kyrgyzstan ......................................................................................................................... 41 3.2 Conceptual framework ..................................................................................................... 43 3.3 Data and descriptive analysis ............................................................................................ 46 3.4 Methodological approach ................................................................................................. 52 3.4.1 Zero tillage adoption decision and production costs ............................................. 52 3.4.2 Estimation of average treatment effect on the treated ......................................... 55 3.5 Results and discussion ...................................................................................................... 56 3.5.1 Determinants of zero tillage adoption ................................................................... 56 3.5.2 Resource-saving effects of zero tillage ................................................................... 60 ~ 2 ~ Although the intensive fertilizer application has lessened the visible impact and masked the problem's full extent, irrigated croplands in Central Asia—especially in cotton-producing areas— have become focal points for land degradation. This degradation resulted in the loss of US$ 6 billion in 2001-2009, with desertification and agricultural abandonment costing US$ 1 billion each (Mirzabaev et al., 2016). Enhancing soil productivity and mitigating land degradation stand as pivotal challenges in Central Asia. Yet, the post-independence transition to market economies in these countries presented new challenges for mitigating land degradation (Pomfret, 2019). Land reforms, aimed at redistributing state-owned lands to private individuals, often lacked the necessary support mechanisms to foster sustainable land management practices (Kienzler et al., 2012). Farm fragmentation hindered the efficient use of resources and adoption of modern agricultural technologies, further contributing to land degradation and lower agricultural productivity (Lerman and Sedik, 2018). Access to knowledge, technology, financial resources, and infrastructure necessary for implementation of sustainable land management by newlyemerged agricultural producers was lacking (Hornidge et al., 2016; Kienzler et al., 2012). 1.2 Sustainable agricultural practices The adoption of sustainable agricultural practices (SAPs) present a promising solution to these challenges by improving soil fertility, capturing carbon to address climate change, and boosting both crop yields and financial returns (Manda et al., 2016). The adoption of SAPs stands as a principal strategy directed towards enhancing farm productivity, improving agricultural profitability, and reducing production expenses (e.g., Lee, 2005; Manda et al., 2016; Tadjiev et al., 2023a; Zhao et al., 2020). Scholars argue that sustainable agricultural development embodies five primary characteristics: (1) resource conservation, (2) environmental preservation, (3) technological suitability, (4) economic viability, and (5) social acceptability (FAO, 1989; Teklewold et al., 2013). Accordingly, sustainable agricultural practices are characterized by their ~ 3 ~ inherent capacity to yield positive externalities in crucial domains such as biodiversity, water conservation, soil health, landscape preservation, and climate change mitigation. This distinguishing feature sets them apart from conventional practices (Dessart et al., 2019). SAPs encompass a spectrum of farming practices that include environmental, societal, and economic aspects. These practices include crop rotation, intercropping, conservation tillage, biological methods for pest control, residue retention, improved crop varieties, animal manure, soil and water conservation (e.g., Lee, 2005; Manda et al., 2016; Teklewold et al., 2013; Zeweld et al., 2017). Insufficiency of financial, physical, and human resources stands as a predominant issue faced by rural households and individual farms within the Central Asian region (Wolfgramm et al., 2010; Djanibekov et al., 2012). Moreover, the rising expenses associated with production inputs pose significant challenges for farms in Central Asia (Djanibekov et al., 2012). Hence, it is imperative to advocate for the adoption of resource-saving practices that entail lower financial resources among farms. For example, crop rotation, zero tillage, intercropping is an approach to soil management that involves minimal input, and preferably less off-farm sources (Baker and Saxton, 2007; Tanveer et al., 2019). In various empirical chapters of my dissertation, I thus focus on studying the adoption of crop rotation, zero tillage, low-tillage, biological methods for pest control, and intercropping practices in the context of Central Asia. The utilization of these selected practices offers economic, social, and environmental advantages to farmers (e.g., Abdollahzadeh et al., 2015; Baker and Saxton, 2007; Glaze-Corcoran et al., 2020; Ogieriakhi and Woodward, 2022; Yigezu and El‐Shater, 2021; Zhao et al., 2020). The adoption of these practices provides potential for mitigating challenges of sustainable agricultural development in Central Asia (FAO, 2013; Kienzler et al., 2012; Nurbekov et al., 2016; Pender et al., 2009). Furthermore, the above-listed SAPs have been successfully tested for feasibility in Central Asia (Nurbekov et al., 2016; Pender et al., 2009). Biological pest control refers to the environmentally conscious approach of managing pests by harnessing natural adversaries (Kumari et al., 2022; Nigam and Mukerji, 2023). Biological pest ~ 4 ~ controls reduce the dependency of modern agriculture on pesticide applications and maintain high crop yields (Schneider et al., 2015). Biological control is a component of an integrated pest management strategy and the adoption of integrated pest management will be introduced to manage pest populations and crop producers’ net returns will be improved (Hoffmann and Frodsham, 1993; McNamara et al., 1991). Crop rotation is a method of cultivating crops in a systematic sequence on the same piece of land, with the goal of preserving soil fertility and ensuring that farmers maintain or increase their land-related profits (Sumner, 1982; Tanveer et al., 2019). Crop rotation, particularly with leguminous crops, has been shown to sustain and enhance farm productivity and income (FAO, 2015). Studies like Manda et al. (2016) demonstrated that maize-legume rotation, combined with improved varieties and residue retention, boosts maize yields and household income. In China, cotton-legume rotation increases cotton yields by nearly a quarter (Zhao et al., 2020). Implementing crop rotation, especially with alfalfa, sorghum, or mung beans as cover crops, within an organic-based agricultural framework in Uzbekistan enhances net present value and reduces expenses (Franz et al., 2009). Farmers who practiced crop rotation had higher welfare compared to non-adopters (Ghimire et al., 2012; Mohammad et al., 2012; Zeweld et al., 2020). Conservation agriculture means no or minimum mechanical soil disturbance, seeding or planting directly into untilled soil, and using crop residues and cover crops to protect and feed soil life and this can help to improve soil quality, and increase soil organic matter (FAO, 2023; Nurbekov et al., 2016). The study investigates the adoption of zero tillage as a form of conservation agriculture. Zero tillage, namely when crops are planted directly into a seedbed not tilled after harvesting the previous crop accumulates soil carbon and increases soil nitrogen, thus promoting soil, moisture and nutrients conservation for increasing crop productivity (Baker and Saxton, 2007; FAO, 2023; Ofstehage and Nehring, 2021). Zero tillage is also proved to be a solution to target low financial and resource capacity of smallholders in developing countries (Jaleta et al., 2016; Jaleta et al., 2019; Montt and Luu, 2020; Musafiri et al., 2022). ~ 5 ~ Intercropping is the simultaneous cultivation of two or more crop species in the same field at a given time (Stomph et al., 2020; Wang et al., 2014). Intercropping minimizes the use of chemical inputs such as pesticides and herbicides, as well as enhances soil fertility and yields (Brooker et al., 2015; Ha et al., 2023; Stomph et al., 2020). 1.3 Problem statement, research objectives and structure of the thesis As earlier mentioned, the global scholarly discourse extensively examines the issue on the adoption of sustainable agricultural practices and its impact on farm performance. To combat the land degradation, in the mid-1990s, the concept of conservation agriculture was presented by international agencies (Wolfgramm et al., 2015) and several practices have been successfully tested in Central Asia (Nurbekov et al., 2016; Pender et al., 2009; Kienzler et al., 2012). Nevertheless, the adoption of sustainable agricultural practices is still low in Central Asia and most farmers are reluctant to adopt them. The conversion to sustainable practices of crop cultivation in Central Asia, is challenged by the lack of agronomic knowledge about sustainable agricultural methods, inadequate supply of extension services, lack of seed varieties and machinery suitable for sustainable crop cultivation, as well as the absence of government incentives for adopting such practices (Kienzler et al., 2012; Nurbekov et al., 2016). Although much of the literature emphasizes the main drivers of the adoption of SAPs and their impact on farm performance (e.g., Knowler and Bradshaw, 2007; Ruzzante et al., 2021; Takahashi et al., 2020), there is a lack of empirical studies that thoroughly examine the determinants and impact of sustainable agricultural practices in the context of Central Asia. So, the issue of how the adoption of sustainable agricultural practices can be promoted among farmers in Central Asia remains an understudied research question. This dissertation addresses this existing research gap and aims to provide a more comprehensive and refined understanding of the subject matter by answering the research question “What factors determine adoption of selected SAPs and how this affects farm performance?”. By doing so, the dissertation also ~ 6 ~ contributes to the global discussion on adoption and impact of SAPs in a developing country farming system. This study encompasses three objectives aimed at comprehending the essential drivers for promoting the adoption of sustainable agricultural practices by farmers and their effects on farm performance in Central Asia: 1. to understand the main determinants of farmer’s decision to adopt SAP and its impacts on farm outcomes; 2. to investigate whether the adoption of SAP offers economic benefits to farmers by reducing production costs; 3. to explore determinants of farmers’ participation in informal cooperation in water management and its impact on the intensity of SAPs adoption. To achieve the research goal, I study commercial farms and rural households of three Central Asian countries, namely Kazakhstan, Kyrgyzstan and Uzbekistan. I utilize empirical models as methodological tools to accomplish the objectives of the study. As the sample is not truly random, but based on preselection of study regions, I follow up on the p-value warnings when specific sampling designs are ignored (Hirschauer et al., 2020). Instead of reporting the model results with p-value, I use confidence intervals (CI). As the p-value or the measure of statistical significance is not the relevant output from an analysis, it is argued that reporting the estimation results with CIs is more preferable (Imbens, 2021). CIs also provide a convenient way of summarizing the hypothesis test results for effect sizes (Greenland et al., 2016). This first objective is addressed in Chapter 2 where I used crop rotation in irrigated cotton farming systems as an example of SAP. Crop rotation, or sequentially growing cotton with leguminous crops on the same plot, can contribute to sustainability of cotton cultivation by improving soil fertility, reducing land degradation, and preventing nutrient loss in the long run (Ball et al., 2005) and affecting soil microbiology and phytotoxins (Tanveer et al., 2019). Existing ~ 7 ~ research on Central Asia’s crop rotation relies on agronomic experiments (Takata et al., 2008) or employs remote sensing tools (Conrad et al., 2017; Löw et al., 2017) rather than employing economic rigor to assess the advantages of this SAP. While the documented agronomic advantages of crop rotation in cotton cultivation schemes are well-established (Takata et al., 2008; Zhao et al., 2020), farmers are interested in monoculture for reaching short-term profit maximization goals. This raises questions about factors defining the adoption of crop rotation, and whether crop rotation improves cotton growers’ revenues and yields. Hence, I hypothesize that farmers who adopt crop rotation achieve higher cotton yields and net revenues compared to non-adopters. It is the first empirical research delving into the factors influencing the adoption of crop rotation among Central Asian farmers. Here, I examine two historically cottondominated irrigated areas in the region, namely, Turkistan province in Kazakhstan and Samarkand province in Uzbekistan by using a farm survey data collected in the framework of the AGRICHANGE 1 research project in March-April 2019. The distinctive approach, examining multiple countries instead of single-country research focusing solely on the productivity and income impacts of adopting soil-improvement practices, presents additional insights on the heterogeneity in institutional responses and policy effectiveness across contrasting national settings. This part of the study is one of the few studies globally to empirically analyze crop rotation's effect on cotton producers' performance. To better understand the impact of crop rotation on farmer’s cotton yield and cotton net returns I employ an endogenous switching regression (ESR) model and measure average treatment effects. The empirical results show implications of two countries’ contrasting institutional settings of cotton cultivation on adoption of crop rotation. Compared to conventional crop cultivation, crop rotation in Uzbekistan increases cotton yields and revenues. However, an opposite effect is observed among cotton growers in Kazakhstan. 1 Institutional Change in Land and Labour Relations of Central Asia’s Irrigated Agriculture (AGRICHANGE), www.iamo.de/en/agrichange. ~ 8 ~ The second objective is addressed in Chapter 3 where zero tillage is used as an example of an SAP offering socio-economic benefits for smallholders through lower production costs (Chatterjee and Acharya, 2021). However, there is an ongoing debate whether zero tillage only reduces smallholders’ production costs or whether it alters the production cost structure. A summary of findings from nine empirical studies on the impact of conservation tillage methods, including zero tillage, is presented in Table A3 in the Appendix. For instance, some findings suggest that zero tillage can increase monetary herbicide expenditure and total labor costs (Teklewold et al., 2013). While arguing that zero tillage reduces fuel and labor cost, Yigezu and El‐Shater (2021) found that its effect on the labor requirement and expenses are not necessarily straightforward as zero tillage can increase manual work requirements for weeding. Furthermore, while lowering female and male labor requirements, reduced and zero-tillage methods lead to higher application doses of chemical fertilizers and herbicides (Tessema et al., 2018). This empirical study contributes to this debate on whether zero tillage saves or increases production costs in smallholder settings. To achieve the research objective, I utilized plot-level panel data from the “Life in Kyrgyzstan (LiK) survey, which provides detailed longitudinal information on smallholder farmers in Kyrgyzstan. My investigation shows that zero tillage adoption increases hired labor and herbicide costs, but decreases land preparation and total production costs compared to the conventional tillage method among households in Kyrgyzstan. The third objective is addressed in Chapter 4. Farmers’ participation in informal cooperation in water management allows them to overcome water distribution disputes and to share maintenance costs. In addition, farmers’ cooperation in water management provides a platform for knowledge exchange among participants, such as about SAPs use. Participation in collective initiatives can improve soil conservation by facilitating the exchange of planting materials, information, and labor among farmers, overcoming household labor constraints, and thereby enhancing the implementation of labor-intensive soil conservation practices (Willy and HolmMüller, 2013). Additionally, community-based collective action initiatives contribute to soil ~ 9 ~ conservation through collective learning and knowledge exchange. The participation in informal cooperation is expected to enhance the information sharing among farmers, leading to an improved intensity of SAPs adoption. Over the last four decades, globally, a significant number of studies have been investigating SAPs adoption determinants including farm and farmer characteristics, institutional and behavioral factors (e g., Dessart et al., 2019; D’Emden et al., 2008; Feder et al., 1985; Ruzzante et al., 2021). Despite this voluminous literature on SAPs adoption, there is a lack of empirical research of how informal cooperation among water users affects their SAPs adoption (Willy and Holm-Müller, 2013; Xue et al., 2022). Willy and HolmMüller (2013) offer a perspective on the effects of various collaborative efforts, such as mutual support initiatives within a community, the upkeep of rural access roads, and water management, on soil conservation efforts of rural smallholders. Xue et al. (2022) investigate the impact of participation in collective action on smallholders’ decisions to adopt no-tillage technology. Several studies investigated the impact of formal cooperative membership on farmers’ SAPs adoption decisions (e.g., Wu et al., 2023; Zhang et al., 2020). In Chapter 4 of my thesis, I thus explore both the determinants and effects of informal cooperation on the intensity of SAPs adoption. For doing so, I use two waves of farm survey data of Uzbekistan collected within the framework of the AGRICHANGE and SUSADICA 2 projects in 2019 and 2022, and employ marginal treatment effects (MTEs) model. The analysis of the marginal returns associated with participation in informal cooperation contributes to the empirical knowledge on SAP adoption in developing countries. The results show that farmers who are likely to participate in informal cooperation tend to benefit more from participation in terms of intensity of SAP adoption. 2 Structured doctoral programme on Sustainable Agricultural Development in Central Asia (SUSADICA), https://www.iamo.de/en/research/research-projects/ ~ 10 ~ The final chapter of the dissertation presents conclusions and provides policy recommendations aimed at enhancing the adoption of SAPs in Central Asia. Furthermore, this concluding chapter addresses limitations, as well as proposes ideas for future research. ~ 11 ~ 2 DETERMINANTS AND IMPACTS OF CROP ROTATION ADOPTION AMONG COTTON GROWERS IN IRRIGATED AREAS OF KAZAKHSTAN AND UZBEKISTAN 2.1 Reforms in the cotton sector and the adoption of crop rotation practices in Kazakhstan and Uzbekistan Cotton farming in irrigated areas like South Kazakhstan and Uzbekistan makes significant contributions to rural livelihoods (Shtaltovna and Hornidge, 2014), and in addition to farm employment, creates jobs in the ginning and textile sectors (Baffes, 2005). Thus, cotton production is directly linked to rural incomes and employment. During the Soviet era cotton cultivation system in Central Asia, based on production plans and state regulation of procurement prices and value chain actors (Rumer, 1989), combined six-year sequences of cotton followed by three years of leguminous crops (mainly alfalfa) and fallow land (Toderich et al., 2007). As the Soviet planners demanded the fulfilment of cotton production plans, disregarding environmental consequences, crop rotation was often abandoned, and farmers relied on the intensive use of fertilizers and machinery (Rumer, 1989). The continuous practice of cotton monoculture resulted in cropland degradation. Along with this, pests and water shortages have stagnated cotton yields since the 2000s (OECD/FAO, 2022). Available sustainable agronomic practices can improve cotton yields (OECD/FAO, 2022). One example is crop rotation, an alternative to monoculture, which involves growing cotton sequentially with leguminous crops on the same plot, contributing to soil fertility, reducing land degradation, and preventing nutrient loss (Ball et al., 2005). Diversifying crop cultivation, such as with sorghum, instead of monoculture, offers a solution to the environmental challenge of soil salinity in Central Asia (Bobojonov et al., 2013). Following the dissolution of the Soviet Union, Kazakh and Uzbek governments took contrasting approaches to reform their cotton sectors (Pomfret, 2019). In Kazakhstan, an upper middleincome economy and the richest country in Central Asia thanks to its oil exports, the government rapidly reformed the cotton sector in the 1990s (Pomfret, 2019). The cotton sector in ~ 18 ~ a) b) Figure 2.2: Cotton yields (a) and net revenues (b) among crop rotation adopters and nonadopters in Kazakhstan and Uzbekistan Source: Based on the AGRICHANGE 2019 farm survey data. 2.3 Methodological approach The assessment of the technology adoption impact based on non-experimental cross-sectional data requires the correction of self-selection bias, identification of proper counterfactuals, and controlling for non-observable farm characteristics (Asfaw et al., 2012; Jaleta et al., 2016). I ~ 19 ~ explain the empirical models in the following subsections and motivate the selection of the methodology for this section. 2.3.1 Crop rotation adoption decision and farm outcome The identification of farmers’ decision to adopt crop rotation is based on the measurement of profitability and yield-increasing effects. To estimate the impact of crop rotation on farm outcomes, I followed existing literature such as Abdulai and Huffman (2014), Amadu et al. (2020), Jaleta et al. (2016), and Issahaku and Abdulai (2020), and employed a two-stage estimation approach. Farmers will adopt crop rotation (𝐶1∗) if they expect to achieve higher yields and net returns from crop rotation compared to a decision with not to adopt (𝐶0∗). Here, expected yields and net returns are not observed, but adoption decision is observed. In this perspective, adoption decision (𝐶𝑖) is treated as a dichotomous choice, namely 𝐶𝑖=1 if 𝐶1∗> 𝐶0∗ and 𝐶𝑖=0 if 𝐶0∗>𝐶1∗. Thus, farmers’ adoption decision is related to their perception of whether adoption maximizes net returns or not. Based on given latent variable model, in the first stage, determinants of adoption were analyzed by the following probit model: 𝐶𝑖∗=𝛿𝐾𝑖+𝜀𝑖 (2.1) here, 𝐶𝑖 is a dummy variable indicating whether farmer 𝑖 adopts crop rotation or not. 𝐾𝑖 is a vector of determinants of adoption decision (𝑛×𝑚). 𝛿 is a vector of parameters to be estimated 𝑚×1, 𝜀𝑖 is a vector of error term (𝑛×1) normally and independently distributed with mean 0 and variance 𝜎2. To connect the relationship between adoption of crop rotation scheme and farm outcomes, it is assumed that farmers maximize expected net returns from cotton production, and the function is expressed following Dubbert (2019), and Zheng et al. (2021): 𝑚𝑎𝑥 𝜋𝑖= 𝑃𝑖𝑄𝑖(𝑅𝑖,𝑍𝑖) − 𝐼𝑖𝑅𝑖 (2.2) ~ 20 ~ where 𝜋 is the maximum net returns of farmer 𝑖 gained from cotton production, 𝑃 is cotton price per kg, and 𝑄 is cotton yield in kg. 𝑅 represents input quantities such as fertilizer, seeds, and labor. 𝑍 represents the vector of explanatory variables, i.e. farm/farmer characteristics. 𝐼 is a vector of input prices. Net returns (𝜋𝑖) are expressed as a function of input and output prices, farm/farmer characteristics, and adoption of crop rotation scheme as follows: 𝜋𝑖=𝜋(𝑃𝑖,𝐼𝑖,𝑍𝑖,𝐶𝑖) (2.3) Applying Hoteling’s lemma to Equation (2.2) yields a reduced form of the cotton output supply function as follows: 𝑄𝑖=𝑄(𝑃𝑖,𝐼𝑖,𝑍𝑖,𝐶𝑖) (2.4) Given the challenges in measuring production costs, it is assumed that farmers aim to maximize both yield and net returns. For larger farms in Samarkand, knowledge about yields, prices, and revenues is available, but input use data is fragmented among various experts, including agronomists, machinery engineers, and irrigation experts, making measurement complex. In this study, net returns from cotton production are calculated by deducting fertilizer costs (nitrogen, phosphorus, and potassium), cotton seed costs, and labor costs (payments for land preparation and cotton cultivation) from cotton revenue (yield multiplied by cotton price). Since manual cotton-picking wages are linked to the amount of harvested crop rather than actual labor effort, this cost component is not included. Equations (2.3) and (2.4) determine net returns and cotton yield based on input and output prices, farm/farmer characteristics, and the adoption of crop rotation. In the second stage, to better understand the impact of adoption, I applied a simple model of farmers’ outcomes. Cotton yield and net returns are determined by several factors, including land, labor, and fertilizer. A Cobb-Douglas production function was used, connecting farm outputs with inputs and other factors: 𝑌𝑖=𝐹(𝐴,𝐿,𝑁) (2.5) ~ 21 ~ where 𝑌𝑖 is a vector of outcome variables of farmer 𝑖, 𝐴 stands for farm size (in this case, cotton area in ha), 𝐿 stands for labor quantity (in persons per ha), and 𝑁 stands for fertilizer use (US$ per ha). As mentioned above, family labor dominates among Kazakh cotton-growers, and thus labor quantity in persons is used in the model. Taking the logarithm of outcome variables and production inputs, I derived cotton yield (or cotton net returns) function as linearly separable (Amadu et al., 2020). Additionally, I accounted for other dummy or non-logarithmic variables. Thus, the effect of crop rotation adoption on cotton yield and net returns was modelled through a ln(𝑌) functional form related to production inputs and other factors such as farm/farmer characteristics and institutional settings as follows: 𝐿𝑛𝑌𝑖=𝛼0+𝛽𝑙𝑛𝐴+µ𝑙𝑛𝐿+𝜅𝑙𝑛𝑁+𝜓𝑍𝑖+𝜍𝐶𝑖+𝑢𝑖 (2.6) It is assumed that the outcome variable (𝑌𝑖) is associated with production inputs (𝐴, 𝐿 and 𝑁), a vector of other explanatory variables (𝑍𝑖), and rotation adoption (𝐶𝑖) take a value of 1 if a farm adopts crop rotation and 0 otherwise. 𝛼0 is a constant, 𝛽, µ, 𝜅, 𝜓 and 𝜍 are vectors of estimated parameters, and 𝑢𝑖 is an error term. The impact of the adoption of crop rotation on cotton yield and net returns is computed by the estimation of the parameter 𝜍. This approach might create biased estimates because it assumes that adoption is exogenously determined, while it is potentially endogenous (Di Falco et al., 2011). Farmers’ decision to adopt or not to adopt may be based on individual self-selection. Farmers who adopt crop rotation can have different characteristics compared to non-adopters. Furthermore, farmers can decide to adopt based on expected benefits but structurally differ in their expectations (Asfaw et al., 2012; Di Falco et al., 2011). Considering that the interviewed farmers might have self-selected into adopting crop rotation schemes, selection bias can occur because of observable and unobservable attributes affecting adoption and outcome variables at the same time. Hence, an Ordinary Least Squares (OLS) estimator might generate biased and inconsistent estimates (Di Falco et al., 2011; Dubbert, 2019). Following the arguments expressed in recent existing studies by Asfaw et al. (2012), Di ~ 22 ~ Falco et al. (2011), and Jaleta et al. (2016), I employed an endogenous switching regression (ESR) model that accounts for both endogeneity and sample selection bias. 2.3.2 Endogenous switching regression To examine the influence of crop rotation on farm outcomes, I applied the Average Treatment Effect on the Treated (ATT). The ATT estimates average differences in outcome variables between adopters who actually adopted crop rotation (observed) and those who would not have adopted it (counterfactual). Although the Propensity Score Matching (PSM) method can also calculate ATT, it does not account for unobservable factors that simultaneously influence farmers’ adoption decisions and outcome variables (Jaleta et al. 2016). For instance, Abdulai and Huffman (2014), Asfaw et al. (2012), Jaleta et al. (2016) and Issahaku and Abdulai (2020) applied the ESR model approach to analyze the impact of sustainable agricultural practices on outcome variables in the binary regime of adopters and non-adopters. Following these studies, in the second stage, the relationship between outcome variables and adoption decisions including other explanatory variables, can be formulated in two regimes with an OLS regression model. Consequently, Equation 2.6 is expressed as follows: Regime 1 (crop rotation adopters): 𝑦1𝑖 =𝑋𝑖1𝛽1+𝜔1𝑖 if 𝐶=1 (2.7a) Regime 2 (crop rotation non-adopters): 𝑦2𝑖 =𝑋𝑖2𝛽2+𝜔2𝑖 if 𝐶=0 (2.7b) where 𝑦1𝑖 and 𝑦2𝑖 are outcome variables for adopters and non-adopters. 𝑋𝑖1 and 𝑋𝑖2 are vectors of determinants of the outcome variables. 𝛽1 and 𝛽2 are vectors of parameters to be estimated. 𝜔1𝑖 and 𝜔2𝑖 are error terms. The probit model in Equation 1 supplies essential information to examine and correct the potentially resulting bias (Maddala, (1983, 223); Petrick, (2004, 151)). To test selection bias, according to Heckman (1979) the Inverse Mills Ratio (IMR) can be calculated from the results of a probit estimation as follows: ~ 23 ~ 𝜆1𝑖 = 𝜑(𝛿𝐾𝑖) 𝛷(𝛿𝐾𝑖) 𝜆2𝑖 =−𝜑(𝛿𝐾𝑖) 1−𝛷(𝛿𝐾𝑖) (2.8) where 𝜑(.) and 𝛷(.) indicate probability density function and cumulative density function of the standard normal distribution, respectively. 𝜆1𝑖 and 𝜆2𝑖 represent IMR. Equations 2.7a and 2.7b are used to correct selection bias. Thus, the outcome equations in two regimes stand for: Regime 1 (crop rotation adopters): 𝑦1𝑖 =𝑋𝑖1𝛽1+𝜎1𝜀𝜆1𝑖 +𝜂1𝑖 if 𝐶=1 (2.9a) Regime 2 (crop rotation non-adopters): 𝑦2𝑖 =𝑋𝑖2𝛽2+𝜎2𝜀𝜆2𝑖 +𝜂2𝑖 if 𝐶=0 (2.9b) where 𝜎1𝜀 and 𝜎2𝜀 are parameters to be estimated, 𝜂1𝑖 and 𝜂2𝑖 are normally distributed error terms with mean zero and constant variance. Existing studies explain that for a more robust identification, it is important to select instrumental variables (IV) that affect 𝐶𝑖 in Equation 2.1 and do not appear in explanatory variables of outcome equation. Technology adoption studies employ information sources, such as other farmers, neighbors, and relatives, as valid IVs (Asfaw et al., 2012; Di Falco et al., 2011; Manda et al., 2016). Previous findings show that adopters of maize-legume rotation or improved technologies have better access to relevant information on application and associated benefits (Manda et al., 2016). Based on these arguments, I used variables “information from other farmers and neighbors”, and “information from media, internet and radio about technologies and agronomy” as IVs for measuring the impact of crop rotation in both study regions. For Kazakhstan, I also used “village share of crop rotation adopters” as an IV. Thus, these variables were excluded from Equations 2.9a and 2.9b. I explored acceptability of instruments through a simple falsification test to determine whether the selected variables were reasonable and thus affect farmer’s adoption decision, but not outcome variables (Di Falco et al., 2011; Jaleta et al., 2016). The results of the falsification test show that selected instruments are jointly statistically significant in the adoption decision (for adoption decision χ2=7.81, p-value=0.05 for Kazakhstan; χ2=6.93, p-value=0.03 for Uzbekistan), ~ 24 ~ but statistically insignificant in the outcome equation of non-adopters (for net returns and cotton yield respectively F-stat=0.55 and 0.78, p value=0.65 and 0.51 for Kazakhstan; F-stat=0.22 and 0.18, p value=0.80 and 0.83 for Uzbekistan) (See Table A1 in Appendix A). Consequently, the selected instruments can be considered as plausible. 2.3.3 Average treatment effects The impact of crop rotation on farmers’ outcome can be tested through the comparison of expected outcomes of adopters and non-adopters in actual and counterfactual situations. For this, the Average Treatment Effect on the Treated (ATT) and the Treatment Effect on the Untreated (ATU) were computed within the ESR model. To do this, I calculated the expected outcome for adopters and non-adopters in actual and counterfactual scenarios based on Equations 2.9a and 2.9b as follows: 𝐸(𝑦1𝑖|𝑋, 𝐶𝑖=1)=𝑋1𝑖𝛽1+𝜎1𝜀𝜆1𝑖 (2.10a) 𝐸(𝑦2𝑖|𝑋,𝐶𝑖=0)=𝑋2𝑖𝛽2+𝜎2𝜀𝜆2𝑖 (2.10b) 𝐸(𝑦2𝑖|𝑋, 𝐶𝑖=1)=𝑋1𝑖𝛽2+𝜎2𝜀𝜆1𝑖 (2.10c) 𝐸(𝑦1𝑖|𝑋. 𝐶𝑖=0)=𝑋2𝑖𝛽1+𝜎1𝜀𝜆2𝑖 (2.10d) Here, Equation 2.10a is for adopters (𝐶=1) , and Equation 2.10b is for non-adopters (𝐶=0), both observed in the sample. In contrast, two other equations consider counterfactuals, such as Equation 2.10c is for adopters who would have decided not to adopt, and Equation 2.10d is for non-adopters who would have decided to adopt. The differences between Equations 2.10a and 2.10c can be formulated as Equation 2.11 which explains the comparisons of the expected outcomes (net returns in US$/ha, and cotton yield in t/ha), and allows for the calculation of the average treatment effect on the treated (ATT) as follows: 𝐴𝑇𝑇=(2.10𝑎)−(2.10𝑐)=𝐸(𝑦1𝑖|𝑋,𝐶𝑖=1)−𝐸(𝑦2𝑖|𝑋, 𝐶𝑖=1)=𝑋1𝑖 (𝛽1− 𝛽2)+ 𝜆1𝑖(𝜎1𝜀−𝜎2𝜀) (2.11) ~ 25 ~ The differences between Equations 2.10b and 2.10d can be formulated as Equation 2.12 which is the average treatment effect on the untreated (ATU): 𝐴𝑇𝑈=(2.10𝑏)−(2.10𝑑)=𝐸(𝑦2𝑖|𝑋, 𝐶𝑖=0)−𝐸(𝑦1𝑖|𝑋, 𝐶𝑖=0)=𝑋2𝑖 (𝛽1− 𝛽2)+ 𝜆2𝑖(𝜎1𝜀−𝜎2𝜀) (2.12) Thus, the heterogeneity effect is measured by utilizing Equations 2.11 and 2.12. According to Asfaw et al. (2012), Di Falco et al. (2011), and Jaleta et al. (2016) the effect of base heterogeneity (BH) for adopters can be calculated as the difference between Equations 2.10a and 2.10d, and for non-adopters as the difference between Equations 2.10c and 2.10b (see Table 2.2). Additionally, the outcomes for two groups (crop rotation adopters and non-adopters) may differ because of unobserved factors, as each group may react differently to changing conditions over time. This variation is referred to as “transitional heterogeneity” (TH), indicating that the effects of adopting crop rotation can vary across groups. Table 2.2: Expected conditional, average treatment and heterogeneity effects Subsamples Decision stage Treatment effects To adopt CR Not to adopt CR Adopters (a) 𝐸(𝑦1𝑖|𝑋, 𝐶𝑖=1) (c) 𝐸(𝑦2𝑖|𝑋, 𝐶𝑖=1) ATT Non-adopters (d) 𝐸(𝑦1𝑖|𝑋, 𝐶𝑖=0) (b) 𝐸(𝑦2𝑖|𝑋, 𝐶𝑖=0) ATU Heterogeneity effects BH1 BH2 TH Notes: (a) and (b) represent observed expected farm outcome (cotton net returns (US$/ha), and crop yield (t/ha)). (c) and (d) represent counterfactual expected farm outcome (cotton net returns (US$/ha), and crop yield (t/ha)). C = 1 if farmer 𝑖 adopted crop rotation. C = 0 if farmer 𝑖 did not adopt crop rotation. y1i = farm outcome if farmers treated with crop rotation adoption; y2i = farm outcome if farmers treated with crop rotation non-adoption. ATT = average treatment effect on treated. ATU = average treatment effect on untreated. BH1 = the effect of base heterogeneity for crop rotation adopters. BH2 = the effect of base heterogeneity for crop rotation non-adopters. TH = transitional heterogeneity (ATT-ATU). Source: Authors based on Jaleta et al. (2016). ~ 26 ~ 2.4 Results and discussion 2.4.1 Determinants of farmers’ decision to adopt crop rotation This section presents and discusses the results from the probit model. Figure 2.3 shows the estimated average marginal effect coefficients and 90% CIs of each explanatory variable. Table A2 in Appendix A provides more details to the model results. The econometric models were estimated in STATA 17 software. A cross-country comparison of the model results highlights the influence of divergent institutional settings in the cotton sectors of the two countries. Kazakhstan's approach involved distributing land to former members of state and collective farms who were actively involved in crop cultivation (Petrick et al., 2017). In contrast, Uzbekistan allocated land through auctions to residents, both rural and urban, with strong entrepreneurial skills and capital, sometimes with limited agricultural experience (Djanibekov et al., 2012). The estimation results reveal that in Kazakhstan, farmers' age positively correlates with the likelihood of adopting crop rotation, indicating that older farmers with more experience are more likely to adopt this practice. However, this is not the case for cotton growers in Uzbekistan. Furthermore, farm restructuring implied contrasting levels of land tenure security in two countries. In Kazakhstan, individuals received farmland for private use (Petrick et al., 2017). In Uzbekistan, farmland remained state-owned and could be revoked at any time, as it has been done through several farm consolidation campaigns (Djanibekov et al. 2012). Scholars continuously note that Uzbek land tenure insecurity hinders the wider adoption of sustainable agricultural practices requiring a longer lifespan for generating farm benefits (Hamidov et al., 2022; Kienzler et al., 2012). The model results confirm the effect of land tenure security for Uzbekistan, where cotton growers perceiving higher land tenure security are more likely to adopt crop rotation. This relationship is not observed for Kazakhstan, where respondents generally felt more optimistic about their land tenure security. ~ 27 ~ Another contrasting result is related to the organization of agricultural credits for cotton growers. In Kazakhstan, private ginneries provide financing through contract farming, or farmers seek commercial or subsidized credits (Petrick et al., 2017). Conversely, Uzbekistan's tightly controlled cotton sector relies on parastatal agricultural banks for short-term credits. The model results reflect this difference. In Kazakhstan, credit rationing does not impact crop rotation adoption, as farmers obtain finance and inputs through ginneries in contract farming arrangements. In Uzbekistan, farmers facing credit rationing are more likely to adopt crop rotation, which makes sense since crop rotation with legumes or alfalfa requires fewer financial resources and inputs. This aligns with Montt and Luu’s (2020) findings of a positive relationship between credit rationing and the adoption of cost-effective crop management practices. Creditconstrained farmers opt for more affordable rotation with legumes and alfalfa instead of costly high-value crops. Consequently, the results suggest that access to agricultural finance makes Uzbek cotton growers less inclined to employ crop rotation. Both study areas rely on irrigation, impacting farmers’ crop choices through variables like proximity to the main irrigation canal and farmers' perception of its condition. Inherited Soviet irrigation systems favor farmers at the canal's head, leading to decreased cotton crop rotation near canals due to the lower water requirements of legumes compared to high-value crops like rice, potatoes, vegetables, and melons. Farmers with better water access are more likely to rotate cotton with water-intensive crops. This effect is more pronounced in Uzbekistan, where limited water supply pressures cotton cultivation and crop rotation becomes a strategy. Additionally, farmers' opinions about canal conditions influence their choices, with improved conditions encouraging sustainable crop rotation and defining investment risks in multi-year land-improving practices (Hamidov et al., 2022). Farm training and knowledge delivery methods vary between Kazakhstan and Uzbekistan. In Kazakhstan, farm training participation is voluntary and based on farmers' requests (Shtaltovna and Hornidge, 2014), whereas in Uzbekistan, the government mandates cotton-focused training ~ 34 ~ Table 2.3: Second stage endogenous switching regression estimates for net returns from cotton Kazakhstan Uzbekistan Adopters Non-adopters Adopters Non-adopters Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Log of total cotton fertilizer cost (US$/ha) -0.132 (0.085) -0.275 0.011 -0.025 (0.059) -0.123 0.073 -0.330 (0.448) -1.082 0.421 0.020 (0.142) -0.214 0.254 Log of total labor in a farm (persons /ha) 0.012 (0.062) -0.091 0.116 0.024 (0.035) -0.034 0.081 0.021 (0.123) -0.186 0.228 0.083 (0.056) -0.010 0.176 Log of cotton area (ha) 0.238 (0.080) 0.103 0.371 -0.014 (0.048) -0.093 0.065 -0.012 (0.118) -0.210 0.186 0.168 (0.082) 0.033 0.304 Farmer’s age (year) 0.001 (0.007) -0.010 0.013 -0.008 (0.003) -0.012 -0.003 -0.006 (0.006) -0.015 0.004 -0.003 (0.004) -0.009 0.004 Farmer has education in agriculture (1/0) -0.070 (0.173) -0.361 0.221 -0.052 (0.098) -0.214 0.110 0.148 (0.153) -0.109 0.405 0.251 (0.067) 0.140 0.362 Farmer perceives canal condition as good (1/0) 0.242 (0.130) 0.024 0.461 0.219 (0.090) 0.070 0.369 -0.036 (0.144) -0.278 0.206 0.248 (0.097) 0.088 0.409 Credit-rationed farmer (1/0) 0.017 (0.144) -0.224 0.259 0.068 (0.079) -0.063 0.199 -0.545 (0.275) -1.005 -0.084 0.036 (0.096) -0.124 0.195 Farmer participates in farm trainings (1/0) 0.228 (0.374) -0.399 0.855 0.091 (0.263) -0.343 0.525 0.155 (0.158) -0.109 0.420 -0.020 (0.101) -0.187 0.148 Share of land with good fertility (0-1) 0.187 (0.114) -0.004 0.378 0.080 (0.075) -0.044 0.204 0.065 (0.248) -0.351 0.482 -0.016 (0.098) -0.179 0.146 Distance to the district center (km) -0.021 (0.006) -0.030 -0.011 -0.006 (0.003) -0.011 -0.001 0.014 (0.014) -0.010 0.038 -0.011 (0.006) -0.021 -0.001 Farm fields located at irrigation canal head (1/0) 0.303 (0.136) 0.074 0.532 0.098 (0.089) -0.049 0.245 0.543 (0.287) 0.061 1.025 0.043 (0.104) -0.129 0.215 Farmer perceives land tenure as secure (1/0) x x x x x x -0.721 (0.464) -1.498 0.057 0.168 (0.126) -0.040 0.376 ~ 35 ~ Table 2.3 cont. Kazakhstan Uzbekistan Adopters Non-adopters Adopters Non-adopters Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Farmer participates in contract farming (1/0) -0.243 (0.154) -0.501 0.015 0.215 (0.080) 0.083 0.348 x x x x x x Farm located in Shardara (1/0) 1.375 (0.390) 0.720 2.029 0.327 (0.157) 0.068 0.586 x x x x x x Farm located in Pastdargom (1/0) x x x x x x 0.191 (0.173) -0.100 0.481 -0.121 (0.067) -0.231 -0.010 mills1 0.204 (0.364) -0.406 0.814 x x x -1.616 -2.624 -0.609 x x x mills2 x x x -0.101 -0.737 0.535 x x x 0.183 -0.388 0.754 _cons 6.332 (0.999) 4.658 8.007 7.057 (0.324) 6.521 7.593 10.658 (3.165) 5.352 15.964 5.924 (0.713) 4.747 7.102 N 64 214 64 243 R-squared 0.423 0.163 0.419 0.163 Note: Standard error in parenthesis. Net returns from cotton is given in US$/ha (ln). Source: Based on the AGRICHANGE 2019 farm survey data. ~ 36 ~ Table 2.4: Second stage endogenous switching regression estimates for cotton yield Kazakhstan Uzbekistan Adopters Non-adopters Adopters Non-adopters Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Log of total cotton fertilizer cost (US$/ha) 0.050 (0.057) -0.046 0.147 0.104 (0.039) 0.039 0.169 0.096 (0.205) -0.247 0.440 0.159 (0.078) 0.031 0.288 Log of total labor in a farm (persons /ha) 0.049 (0.040) -0.019 0.116 0.042 (0.026) -0.001 0.085 0.051 (0.061) -0.052 0.153 -0.007 (0.030) -0.056 0.043 Log of cotton area (ha) 0.184 (0.055) 0.093 0.276 -0.009 (0.036) -0.069 0.050 -0.044 (0.061) -0.146 0.057 -0.002 (0.036) -0.061 0.057 Farmer’s age (year) 0.001 (0.004) -0.006 0.009 -0.004 (0.002) -0.008 0.000 -0.001 (0.003) -0.006 0.004 -0.002 (0.002) -0.005 0.001 Farmer has education in agriculture (1/0) -0.127 (0.111) -0.312 0.059 0.001 (0.069) -0.113 0.115 0.005 (0.074) -0.121 0.130 0.077 (0.036) 0.018 0.135 Farmer perceives canal condition as good (1/0) 0.118 (0.095) -0.041 0.278 0.120 (0.064) 0.014 0.225 -0.007 (0.079) -0.139 0.126 0.037 (0.059) -0.048 0.120 Credit-rationed farmer (1/0) -0.010 (0.97) -0.173 0.153 0.014 (0.066) -0.095 0.122 -0.279 (0.127) -0.491 -0.066 0.019 (0.048) -0.060 0.099 Farmer participates in farm trainings (1/0) 0.160 (0.239) -0.241 0.560 0.033 (0.181) -0.267 0.332 0.001 (0.077) -0.129 0.131 0.001 (0.048) -0.077 0.080 Share of land with good fertility (0-1) 0.185 (0.083) 0.045 0.324 0.096 (0.053) 0.009 0.183 0.066 (0.117) -0.131 0.262 0.062 (0.043) -0.010 0.133 Distance to the district center (km) -0.012 (0.004) -0.018 -0.005 -0.004 (0.002) -0.007 -0.001 0.012 (0.007) 0.001 0.024 -0.0004 (0.003) -0.006 0.005 Farm fields located at irrigation canal head (1/0) 0.164 (0.105) -0.012 0.339 0.055 (0.063) -0.049 0.159 0.246 (0.141) 0.009 0.483 0.046 (0.051) -0.038 0.131 Farmer perceives land tenure as secure (1/0) x x x x x x -0.333 (0.208) -0.682 0.016 0.037 (0.059) -0.061 0.135 ~ 37 ~ Table 2.4 cont. Note: Standard error in parenthesis. Cotton yield is given in kg/ha (ln). Source: Based on the AGRICHANGE 2019 farm survey data. Kazakhstan Uzbekistan Adopters Non-adopters Adopters Non-adopters Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Coeff. [90% confidence interval] Farmer participates in contract farming (1/0) -0.240 (0.114) -0.430 -0.049 0.115 (0.058) 0.020 0.211 x x x x x x Farm located in Shardara (1/0) 0.835 (0.272) 0.378 1.291 0.233 (0.095) 0.076 0.390 x x x x x x Farm located in Pastdargom (1/0) x x x x x x 0.018 (0.075) -0.107 0.143 -0.128 (0.032) -0.181 -0.075 mills1 0.156 (0.224) -0.219 0.531 x x x -0.832 (0.267) -1.279 -0.384 x x x mills2 x x x -0.065 (0.286) -0.538 0.407 x x x 0.001 (0.161) -0.265 0.266 _cons 6.697 (0.636) 5.631 7.763 7.325 (0.230) 6.945 7.706 8.630 (1.458) 6.186 11.075 7.006 (0.369) 6.396 7.616 N 64 214 64 243 R-squared 0.425 0.177 0.475 0.151 ~ 38 ~ 2.4.3 Cotton yield and net returns impacts of the adoption of crop rotation As described earlier, the impact of the adoption of crop rotation on farmers’ expected outcomes under actual and counterfactual conditions is measured by the average treatment effect on the treated (ATT) and the average treatment effect on the untreated (ATU) estimated by the ESR model. Table 2.5 presents the results from the ESR treatment effect model for Kazakhstan and Uzbekistan. The fifth column of Table 2.5 provides the treatment effects of the adoption of crop rotation. The obtained results reveal that the impact of the adoption of crop rotation on net returns and cotton yields differs between Kazakh and Uzbek farmers. The adoption of crop rotation has a negative impact on the outcomes of Kazakh farmers. This means that in Kazakhstan, the treatment effect of the adoption of crop rotation on net returns and cotton yield per ha is -0.168 and -0.139, respectively. In other words, interviewed Kazakh adopters of crop rotation would have received 15% higher net returns and 12% higher cotton yields had they not adopted crop rotation. Based on a meta-regression analysis, Ogundari and Bolarinwa (2018) show that crop rotation among other natural resource management strategies has a substantial effect on production and social measures but not on economic outcomes. Such an unexpected impact of crop rotation on the performance of cotton growers in Kazakhstan can be explained by the fact that the existing institutional environment and infrastructure in Kazakhstan after the cotton sector reform actually provide an advantage to farmers who practice “conventional cotton” monoculture. In the sample of Kazakh respondents, cotton is cultivated on 85% of farmers’ sown area (Table 2.1). This can be explained by conditions imposed by contract farming arrangements with private ginneries, favoring cotton monoculture practices to ensure a supply of raw cotton (Petrick et al., 2017). Thus, Kazakh farmers who choose sustainable crop rotation are likely to lose timely access to the ginnery’s provision of inputs like seeds, fertilizers, pesticides, and machinery services. This suggests that to promote sustainable agricultural practices in Kazakhstan’s cotton growing areas, economic incentives for both sides ~ 39 ~ of the contractual arrangement should be considered, not only for cotton growers but also for processors. In contrast, the adoption of crop rotation has a positive impact on both outcome variables of Uzbek cotton growers. The estimation results reveal that the adoption of crop rotation increases cotton yields and net revenues by 19% and 5% respectively. In other words, interviewed Uzbek farmers who actually adopted crop rotation would have obtained 19% less net returns or 5% less cotton yield had they not adopted crop rotation. These findings are confirmed by other studies. For instance, Zhao et al. (2020) found that in China crop rotation increased cotton yields on average by 20% compared with “conventional cotton” monoculture. Löw et al. (2017) found that higher cotton yields in Uzbekistan are likely to be areas with higher share of crop rotation area. For both countries, the model results show that cotton growers who actually did not adopt crop rotation would have lower net returns and cotton yields if they had adopted this practice. Although, the ATU on the outcome variable is negative, the result presents positive TH effects for net returns and cotton yields among Uzbek cotton growers indicating that net returns and cotton yields are higher among adopters of crop rotation. However, negative TH effects are observed for Kazakh cotton growers revealing that cotton yields and net returns are lower among adopters of crop rotation. Furthermore, counterfactual adopters of crop rotation have higher cotton yield and net returns than actual non-adopters in both countries (Table 2.5). The result shows that there are several important sources of heterogeneity that make adopters of crop rotation better cotton producers than non-adopters. ~ 40 ~ Table 2.5: Average expected net returns and cotton yield for adopters and non-adopters of crop rotation in Kazakhstan and Uzbekistan Category To adopt Not to adopt Treatment effect 1 2 3 4 5 6 Cotton net returns (US$/ha) (ln) Decisions in Kazakhstan [90% confidence interval] ATT (a) 6.587 (0.039) (c) 6.755 (0.025) -0.168 (0.048) -0.247 -0.088 ATU (d) 6.639 (0.031) (b) 6.723 (0.015) -0.084 (0.034) -0.140 -0.027 HE BH1= -0.052 BH2= 0.03 TH = -0.084 Decisions in Uzbekistan ATT (a) 6.591 (0.042) (c) 6.415 (0.028) 0.176 (0.051) 0.091 0.260 ATU (d) 6.263 (0.031) (b) 6.363 (0.015) -0.100 (0.034) -0.156 -0.043 HE BH1= 0.328 BH2= 0.052 TH = 0.276 Cotton yield (kg/ha) (ln) Decisions in Kazakhstan ATT (a) 7.561 (0.028) (c) 7.700 (0.017) -0.139 (0.032) -0.193 -0.084 ATU (d) 7.578 (0.021) (b) 7.668 (0.011) - 0.090 (0.024) -0.129 -0.051 HE BH1= -0.017 BH2= 0.032 TH =-0.049 Decisions in Uzbekistan ATT (a) 7.802 (0.022) (c) 7.757 (0.012) 0.045 (0.025) 0.002 0.086 ATU (d) 7.594 (0.018) (b) 7.712 (0.007) -0.118 (0.019) -0.150 -0.086 HE BH1=0.208 BH2=0.045 TH =0.163 Note: Standard errors are in parenthesis. For calculation of the percent differences of treatment effect, 100*(eATT -1) equation is used following Asfaw et al. (2012). Source: Based on the AGRICHANGE 2019 farm survey data. ~ 41 ~ 3 DOES ZERO TILLAGE SAVE OR INCREASE PRODUCTION COSTS OF SMALLHOLDERS IN KYRGYZSTAN? 3 3.1 Smallholders’ challenges in the adoption of conservation agriculture in Kyrgyzstan Kyrgyzstan is a land-locked low-income food-deficit country with a population of about 6 million, of which almost two-thirds live in rural areas (FAO, 2020). In 2021, GDP per capita was US$ 1,123 (in constant 2015 US$). Despite the progress in poverty reduction, one-fourth of the population lives below the poverty line (World Bank, 2023). Rural areas, where two-thirds of the population live in poverty, are still lagging behind these figures (FAO, 2020). Although agriculture's contribution to the country’s gross domestic product (GDP) has been steadily declining, it still plays a central role in the rural economy. In 2021, agriculture accounted for almost 15% of GDP (World Bank, 2023). As of 2019, about 20% of employment was in agriculture (World Bank, 2023). Kyrgyzstan's late-1990s land reform drove the switch from planned socialist agriculture to smallholder market-oriented agriculture (Lerman and Sedik, 2018). Through the recognition of private land ownership in 1996-1999 the government redistributed over 80% of arable land among rural families, creating a smallholder-based farming system (FAO, 2020). The majority of smallholders are characterized by intercropped and mixed crop-livestock systems with production mostly for their own consumption (Jalilova et al., 2019). In 2016, the official statistics reported about 1,150,000 rural households and peasant farms with an average size of about 0.87 ha (FAO, 2020). This includes 727,000 rural households with an average land size of about 0.1 ha, and 415,000 peasant farms with an average size of 2.2 ha (FAO, 2020). Although the smallholders have been important in food security and poverty alleviation, the fragmented nature of the farming system is prone to the problems of ‘smallness’. For instance, 3 Chapter 3 was published following open-access article: Tadjiev, A., Djanibekov, N., Herzfeld, T. (2023) Does zero tillage save or increase production costs? Evidence from smallholders in Kyrgyzstan. International Journal of Agricultural Sustainability 21 (1), 2270191. https://doi.org/10.1080/14735903.2023.2270191. This chapter builds upon that article. ~ 42 ~ in fragmented agricultural settings of Kyrgyzstan, limited physical, financial, and human resources raise concerns about the future of agricultural food production and the sustainability of arable lands (Wolfgramm et al., 2010). Among the reasons is that rural households have to cope with the increasing costs of agricultural inputs. Most public finance and agricultural subsidies do not reach rural households and are captured by large commercial farms (Lerman and Sedik, 2018). The government does not have a sufficient budget to provide adequate support to smallholders to cover field operation costs. The scarcity of agricultural machinery has been imposing high machinery service costs for land preparation among smallholders, making it 55% more expensive than in neighboring southern Kazakhstan, and has hindered agricultural productivity in Kyrgyzstan (Guadagni and Fileccia, 2009). Farmers might be facing a mix of price, risk and quantity rationing as the number of credits at affordable rates is limited (Kuhn and Bobojonov 2021). The high rates and transaction costs of commercial credits may be unacceptable for smallholders the majority of whom cannot access limited subsidized credits. The lack of access to new technologies and to knowledge of conservation tillage practices limits the wider adoption of zero tillage among smallholders in Kyrgyzstan. Kyrgyzstan’s irrigated agriculture is among the most vulnerable in Eastern Europe and Central Asia to climate change (Fay et al., 2010). A modeling study by Bobojonov and Aw-Hasan (2014) suggests that under a water shortage scenario, predicted farm incomes in the semiarid parts of Kyrgyzstan might decline by 15% harming smallholders’ profits and long-term sustainability. In light of the importance of agriculture in rural incomes and food security, the intensity and spread of land degradation and increasing pressure from water scarcity will affect agricultural productivity and threaten agricultural livelihoods. Cost-saving practices like zero tillage can be an option for smallholders that suffer from low credit access, underinvestment and are prone to water stress. In 2016, the full technical potential adoption level of conservation agriculture in Kyrgyzstan, including reduced and zerotillage and crop rotation, was estimated at 1.2 million ha of cultivated area under cereals, oil and ~ 43 ~ leguminous crops (Polo et al., 2022). The results of the financial analysis presented by Polo et al. (2022) show that conservation agriculture scores moderately with an investment return rate of 13% and a payback period of seven years. It was estimated that conservation agriculture can increase agricultural production via long-term improved soil nutrient management and water retention. For instance, raised-bed and no-tillage planting can increase wheat yield by 25–38% compared to the conventional cultivation method (Nurbekov et al., 2016). The economic value of the annual additional production due to the adoption of conservation agriculture in Kyrgyzstan was estimated at over US$ 35 million or 9% of gross agricultural value (Polo et al., 2022). However, despite these advantages, the gap between present and potential uptake has remained substantial with little change (Polo et al., 2022). 3.2 Conceptual framework Numerous studies have noted three paradigms such as “the innovation-diffusion”, “the adoption perception” and “economic constraints” to define farmers’ adoption of conservation practices (Chatterjee and Acharya, 2021; Ruzzante et al., 2021). Each paradigm assumes several factors influencing the adoption decision (Figure 3.1). For example, to illustrate adoption behavior, the economic paradigm assumes the maximization of the farmer’s profit and considers economic constraints such as access to natural resources, access to capital, investment costs and risk attitude. The innovation-diffusion paradigm assumes that access to information is the main parameter to improve adoption decisions. The adoption perception paradigm postulates that a farmer’s adoption behavior depends on perceived attributes of innovation, access to information, and individual factors such as the farmer’s experience and education, as well as institutional factors that can affect their perceptions (Ruzzante et al., 2021). I conceptualize that a household faces the decision to adopt zero tillage on a specific plot versus conventional tillage methods in crop cultivation. From this perspective, the economic paradigm stipulates that the adoption decision occurs under the farmer’s objective of profit maximization. ~ 50 ~ Table 3.1 cont. Variables 2016 2019 Full sample Mean Sd Mean Sd Mean Sd Household farm characteristics Number of household members that can work in agriculture (above 10 and under 65 years old) 4.407 1.781 4.570 1.927 4.490 1.859 Asset index 0.400 0.139 0.354 0.168 0.376 0.156 Household owns a tractor (1/0) 0.051 0.226 0.035 0.188 0.043 0.207 Number of livestock units owned by household 3.364 4.260 2.377 4.268 2.859 4.292 Household received remittances last year (1/0) 0.145 0.353 0.226 0.418 0.187 0.390 Household applied chemical fertilizers last year (1/0) 0.252 0.434 0.246 0.431 0.249 0.432 Household experienced a weather shock last year (1/0) 0.629 0.483 0.161 0.367 0.390 0.488 Household experienced an agricultural shock last year (1/0) 0.364 0.481 0.088 0.283 0.223 0.416 Plot under grains and legumes (1/0) 0.318 0.466 0.353 0.478 0.336 0.472 Plot under vegetables (1/0) 0.400 0.490 0.270 0.444 0.334 0.472 Plot under a mix of crops (grain, legumes and vegetables) (1/0) 0.073 0.260 0.022 0.146 0.047 0.211 Location characteristics Distance to main road from dwelling (km) 0.521 0.738 0.777 0.891 0.652 0.830 Distance from dwelling to plot (km) 1.470 2.840 1.174 2.536 1.319 2.693 Number of land plots owned by household 1.966 0.627 1.980 0.702 1.973 0.666 Plot size (ha) 0.694 1.340 0.794 1.898 0.745 1.649 Institutional settings Amount of credit received by household last year (US$) 210.154 775.394 339.183 1387.313 276.103 1131.974 Provinces Issyk Kul (1/0) 0.160 0.367 0.179 0.383 0.170 0.375 Djalal Abad (1/0) 0.213 0.409 0.201 0.401 0.207 0.405 Naryn (1/0) 0.073 0.260 0.048 0.213 0.060 0.237 Batken (1/0) 0.125 0.331 0.122 0.328 0.123 0.329 Osh (1/0) 0.260 0.439 0.293 0.455 0.277 0.448 Talas (1/0) 0.079 0.270 0.070 0.256 0.075 0.263 Chuy (1/0) 0.160 0.367 0.179 0.383 0.170 0.375 Note: N=1363 for 2016, N=1425 for 2019 and N=2788 for the full sample. Because of missing values of herbicide costs, the number of observations for 2016, 2019 and the full sample are 1342, 1396 and 2738, respectively. Source: Tadjiev et al. (2023a). ~ 51 ~ As a proxy for household wealth, I calculated the asset index using the principal component analysis (PCA) as suggested in Filmer and Pritchett (2001). I used binary information regarding ownership of 35 assets based on the standardized PCA scores, and the min-max normalization (feature scaling) method was used to convert the scaled data to a range (0–1). The number of total livestock units (TLU) is an additional household wealth indicator. I calculated TLU based on livestock unit coefficients 5 .. First, I multiplied each type of livestock by LU coefficients, and then summarized the result by households. The summary statistics suggest that the number of livestock units owned by a household was on average 3 in 2016 and 2 in 2019. In the study, I also considered the number of plots owned by households. Table 3.1 indicates that households have on average 2 plots in both years. Remittances and migration have been among the main income sources in rural areas of many developing countries and particularly of Kyrgyzstan where remittances affect households’ decisions in agriculture (Atamanov and Van den Berg, 2012). Following the argument by Montt and Luu (2020) that successful conservation agriculture practice requires appropriate management of external inputs such as fertilizers, I added a household’s application of fertilizers as an explanatory dummy variable in the models. Furthermore, I considered the opinion of household heads about whether their households experienced agricultural shocks over the last year such as pest infestations, crop and livestock diseases, insufficient irrigation water supply, theft of livestock, or inability to sell agricultural products as well as weather shocks such as drought, flood, heavy rain or extremely cold winter temperatures. I assumed that such agricultural and weather shocks can affect a household’s decision to adopt zero tillage practices by harming the household’s agricultural outputs and assets. 5 Total livestock units (TLU) is calculated based on livestock unit (LU) coefficients according to the following sources: (1) https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Livestock_unit_(LSU) and (2) http://adlib.everysite.co.uk/adlib/defra/content.aspx?id=000il3890w.198awldohj69f3#nix . ~ 52 ~ Smallholders often cultivate a mix of crops on a single plot. I aggregated all costs for various crop types to control their effect on zero tillage adoption decision. I generated three dummy variables which explain that the plot was cultivated (1) purely by grain and legume crops, (2) by vegetables, and (3) by a mix of grains, legumes and vegetables. 3.4 Methodological approach The following subsections explain the empirical models and motivate the selection of the estimation strategy. The assessment of the economic effects of technology adoption from nonexperimental survey data requires the correction of self-selection bias, identification of proper counterfactuals and control for non-observable farm characteristics (Asfaw et al., 2012; Jaleta et al., 2016). I based the identification of a farmer’s decision to adopt zero tillage on the measurement of profitability through its production cost reducing effects. To estimate the impact of zero tillage on production costs, I followed the existing literature such as Abdulai and Huffman (2014), Jaleta et al. (2016), Keil et al. (2020), Khonje et al. (2018), and Montt and Luu (2020) and employed a two-stage estimation approach. I assessed different models to investigate the relationships between zero tillage adoption and payments for hired labor, machinery costs for land preparation and seeding, weeding, and herbicide costs as well as total costs. The Mundlak device (Mundlak, 1978) was employed to estimate time-invariant endogeneity. Furthermore, I used the endogenous switching regression (ESR) model to account for selection bias. To estimate the association between zero tillage adoption and each production cost considered above, I used the counterfactual framework that measures average treatment effects on the treated (ATT). 3.4.1 Zero tillage adoption decision and production costs The decision to adopt zero tillage and the selection of plots under this method are made by a household head and other household members, and thus are not random. Such a self-selection problem implies a potential bias in the effect of zero-tillage adoption on production costs. In ~ 53 ~ reality, households might apply zero tillage on plots with higher production costs. As a result, the effect of zero tillage on production costs can be overestimated. As commonly done in other studies (e.g., Jaleta et al., 2016; Khonje et al., 2018; Keil et al., 2020; Montt and Luu, 2020), to correct for selection bias, I employed a two stage ESR model. In the first stage, I estimated the main determinants of zero tillage adoption. The probability of zero tillage adoption for an individual can be written as follows: 𝑃𝑟 (𝑧𝑡𝑗𝑖𝑡)=𝑓(𝑋𝑗𝑖𝑡) (3.1) where, 𝑃𝑟 (𝑧𝑡𝑗𝑖𝑡) is the probability of zero tillage adoption of 𝑖′s household in 𝑗′s plot at time 𝑡. 𝑋 is a vector of explanatory variables describing household and plot characteristics, personal characteristics, location settings, etc. I used the Mundlak approach where the means of observable time-variant variables were added to the model. The Mundlak approach is applied to panel fixed-effects in cases of variation within units over time and when time-invariant observables affect both adoption decision and outcomes (Khonje et al., 2018; Montt and Luu, 2020; Mundlak, 1978). This approach also reduces the problem of unobserved heterogeneity. The fundamental assumption of using the Mundlak approach is to consider unobserved time-invariant components by calculating and employing the mean of time-variant variables as a proxy (Montt and Luu, 2020; Mundlak, 1978). I computed the means of all time-variant variables (𝑥𝑖) and added them to a probit regression model to measure the probability of zero tillage adoption. Furthermore, I included province dummies (𝑅𝑝, here, Issyk Kul is the reference province) and a time dummy (𝑌𝑡, here, 2016 is the reference year) for all models to account for the province-level and year differences. The regional dummies allow for accounting for other cross-regional differences that can be associated with adoption decisions such as costs of machinery, labor and other inputs. Thus, from Equation (3.1), a household 𝑖′s likelihood of adopting zero tillage in their 𝑗′s plot at time t can be formulated as: ~ 54 ~ 𝑃𝑟 (𝑧𝑡𝑗𝑖𝑡 =1|𝑋𝑖,𝑅𝑝,𝑋𝑖 ,𝑌𝑡)=𝛷(𝑎𝑖+𝛽′𝑥𝑗𝑖𝑡 +𝛿′𝑥𝑖𝑡 +𝑅𝑝+𝑌𝑡) (3.2) where, 𝛽, 𝛿 and 𝛾 are the parameters to be estimated. 𝑥𝑗𝑖𝑡 contains observables at the plot level. 𝑥𝑖𝑡 contains observables at the household level. 𝑥𝑖 is the mean of time-varying variables that follow the Mundlak approach. In the second stage, an OLS model was applied under two regimes, namely, under non-adoption and adoption of zero tillage. Here, the model estimates the relationship of outcome variables for zero tillage adopters and non-adopters as follows: {𝑦1𝑗𝑖𝑡 =𝐾𝑗𝑖𝑡1𝛽1+𝑘 𝑖1𝜈1+𝑅𝑝+𝑌𝑡+𝜂1𝑗𝑖𝑡 , if 𝑍𝑇=1 𝑦0𝑗𝑖𝑡 =𝐾𝑗𝑖𝑡0𝛽0+𝑘 𝑖0𝜈0+𝑅𝑝+𝑌𝑡+𝜂0𝑗𝑖𝑡 , if 𝑍𝑇=0 (3.3) where 𝑦𝑗𝑖𝑡 is an outcome variables such as machinery costs for land preparation, machinery costs for weeding, payment for hired labor, and herbicide costs, on plot 𝑗 of 𝑖’s household at time 𝑡. 𝐾𝑗𝑖𝑡 is a set of explanatory variables that relate to the outcomes. 𝑘 𝑖 is the mean of timevarying variables. As mentioned before, 𝑅𝑝 and 𝑌𝑡 are the province and time dummies. Some households reported relatively high costs per plot and high amounts of credit. Therefore, the natural logarithm was used for these variables. However, there are some observations with “0” values. Hence, to avoid missing values, I added “+1” for these variables before transforming to the natural logarithm. The probit model supplies essential information to examine and correct the potentially resulting bias (Maddala, 1983: 223; Petrick, 2004: 151). To test selection bias, I followed Heckman (1979) and used the Inverse Mills Ratio (IMR) calculated from the results of a probit estimation as follows: 𝜆1𝑗𝑖𝑡 = 𝜑(𝛿𝑥𝑗𝑖𝑡)/𝜙(𝛿𝑥𝑗𝑖𝑡); 𝜆0𝑗𝑖𝑡 = −𝜑(𝛿𝑥𝑗𝑖𝑡)/[1−𝜙(𝛿𝑥𝑗𝑖𝑡)] (3.4) where 𝜑(.) and 𝛷(.) indicate the density and cumulative density function of the standard normal distribution, respectively. 𝜆0𝑖𝑡𝑗 and 𝜆1𝑖𝑡𝑗 represent the IMR. The calculated IMR was ~ 55 ~ added to the second stage model to correct selection bias and resulted in the following equation: {𝑦1𝑗𝑖𝑡 =𝐾𝑗𝑖𝑡1𝛽1+𝑘 𝑖1𝜈1+𝑅𝑝+𝑌𝑡+𝜆1𝑗𝑖𝑡𝜎1+ 𝑌𝑡∗𝜆1𝑗𝑖𝑡𝜏1+𝜂1𝑗𝑖𝑡 ,if 𝑍𝑇=1 𝑦0𝑗𝑖𝑡 =𝐾𝑗𝑖𝑡0𝛽0+𝑘 𝑖0𝜈0+𝑅𝑝+𝑌𝑡+𝜆0𝑗𝑖𝑡𝜎0+ 𝑌𝑡∗𝜆0𝑗𝑖𝑡𝜏0+𝜂0𝑗𝑖𝑡 ,if 𝑍𝑇=0 (3.5) Furthermore, to consider changes in the selection effect over time, the IMR was interacted with the time dummy ( 𝑌𝑡∗𝜆𝑗𝑖𝑡) following Montt and Luu (2020). Several studies emphasize the selection of valid instruments that influence adoption decisions but do not affect outcome variables. I assume households near the main road will have more convenience in using conventional tillage methods due to easy access to machinery services and, thus, thus, are less likely to adopt zero tillage than households located further away from the road. A falsification test shows that “distance to the main road” relates to zero tillage adoption decision but does not affect the outcome variables (see Table A5 in the Appendix). 3.4.2 Estimation of average treatment effect on the treated The average treatment effect was estimated within the ESR framework method to test the impact of zero tillage adoption on outcome variables. First, the expected outcomes of zero tillage adopters and non-adopters were compared in actual and counterfactual situations. The expected (actual) outcome for zero-tillage adopters can be expressed as follows: 𝐸(𝑦1𝑗𝑖𝑡|𝑧𝑒𝑟𝑜 𝑡𝑖𝑙𝑙𝑎𝑔𝑒=1)=𝐾𝑗𝑖𝑡1𝛽1+𝑘 𝑖1𝜈1+𝑅𝑝+𝑌𝑡+𝜆1𝑗𝑖𝑡𝜎1+ 𝑌𝑡∗𝜆1𝑗𝑖𝑡𝜏1 (3.6) The expected outcome for adopters had they not adopted zero tillage (counterfactual) can, thus, be expressed as follows: 𝐸(𝑦0𝑗𝑖𝑡|𝑧𝑒𝑟𝑜 𝑡𝑖𝑙𝑙𝑎𝑔𝑒=1)=𝐾𝑗𝑖𝑡1𝛽0+𝑘 𝑖1𝜈0+𝑅𝑝+𝑌𝑡+𝜆1𝑗𝑖𝑡𝜎0+ 𝑌𝑡∗𝜆1𝑗𝑖𝑡𝜏0 (3.7) Second, the differences between the actual and counterfactual expected outcomes, which explain the ATT are estimated as follows: 𝐴𝑇𝑇=𝐸(𝑦1𝑗𝑖𝑡|𝑧𝑒𝑟𝑜 𝑡𝑖𝑙𝑙𝑎𝑔𝑒=1)−𝐸(𝑦0𝑗𝑖𝑡|𝑧𝑒𝑟𝑜 𝑡𝑖𝑙𝑙𝑎𝑔𝑒=1) (3.8) ~ 56 ~ 3.5 Results and discussion 3.5.1 Determinants of zero tillage adoption This section briefly discusses the results from a probit adoption model since the primary interest of this chapter is to study the resource-saving impact of zero tillage. The average marginal effects are assessed from the probit model (Equation 3.2). The model results are given in the first column of Table 3.2. The statistical significance of the Wald test shows that all coefficients for explanatory variables are not simultaneously equal to zero. The falsification test shows a significant correlation between the instrumental variable and zero tillage adoption decision, but not with production costs (Table A5 in the Appendix). Hence, the selected instrument is plausible. The econometric models were estimated in STATA 17 software. In summary, the results show that zero tillage is favored by poorer households whose heads are employed in agriculture, have fewer plots, are located in remote areas and do not apply chemical fertilizers. More specifically, household heads with agricultural employment are more likely to use zero tillage because they are exposed to knowledge about sustainable practices. Secondly, agricultural wages are lower than in other sectors (Atamanov and Van den Berg, 2012) and thus such households are more likely to opt for zero tillage rather than apply conventional tillage. The relationship between the asset index of households and the adoption of zero tillage practices is negative. This indicates that households with more assets, i.e., wealthier households, are likely to adopt conventional agricultural practices that depend on mechanized tractor services. This result is consistent with Ngoma (2018), who found that household assets reduce the likelihood of minimum tillage adoption. Furthermore, the model results show that households with more plots are less likely to adopt zero tillage. Applying chemical fertilizer can also be related to smallholders’ wealth status, where poor smallholders have more challenges ~ 57 ~ accessing this input and often cannot afford it. The model result shows that households who apply chemical fertilizers are less likely to adopt zero tillage. Households located further away from their land plots and main roads are likely to adopt zero tillage. This is not surprising since it is expected that households located further away from their lands and road are likely to have higher costs for accessing production inputs and machinery services and, thus, likely to switch to input-saving zero tillage. This result is in line with the findings of Jaleta et al. (2016) and Tessema et al. (2018), who found a positive relationship between plot distance and minimum tillage adoption. A remote location in a rural area can be associated with lower wealth status. Households that experienced agricultural shocks are less likely to adopt zero tillage. Other studies that considered agricultural shocks, e.g., waterlogging stress by Teklewold et al. (2013), droughts and floods frequencies by Wainaina et al. (2016), did not find an association with the adoption of conservation tillage. Finally, the estimation results show that the dummy variable of “grain and legume production” is positive but statistically insignificant in relation to zero-tillage adoption. In contrast, there is a negative and statistically significant relationship between vegetable production and zero tillage adoption decision. It can be explained that some vegetable crops may not be planted using zerotillage method. The negative values of regional dummy variables show that the adoption of zero tillage among smallholders in Djalal Abad, Osh and Talas regions is less likely than among smallholders in the Issyk Kul province. At the same time, there is no significant difference in likelihood of zero-tillage adoption between smallholders in the Issyk Kul province and in its neighboring Naryn and Chuy regions and the Batken province. Various unobserved region-specific characteristics can explain these cross-regional differences in the adoption of zero-tillage. For instance, higher population density and limited availability of land in Osh and Djalal-Abad provinces (Zhunusova and ~ 58 ~ Herrmann, 2018) can reduce smallholders’ costs for hired labor in tillage operations and as a result lower the adoption rate of zero-tillage in these two provinces. Furthermore, agroecological zoning of the regions of Kyrgyzstan can account for differences in crop portfolio, production specialization and tillage methods (Jalilova et al., 2019). Chuy, Talas and Issyk Kul regions are closer agro-ecologically to each other representing the northern regions. Djalal Abad, Osh and Batken represent southern agro-ecological regions encompassing the Fergana valley. Naryn region represents the central zone with vast alpine areas of mountains and valleys suitable for winter grazing and crop cultivation. ~ 59 ~ Table 3.2: Determinants of zero tillage adoption decision Marginal effect Standard error [90% confidence interval] Age of household head (years) 0.0003 0.002 -0.002 0.003 Education level of household head (categorical, 1=illiterate…7=university) 0.002 0.005 -0.007 0.011 Female household head (1/0) -0.008 0.016 -0.033 0.018 Household head employed in agriculture (1/0) 0.048 0.025 0.007 0.089 Number of household members that can work in agriculture (above 10 and under 65 years old) -0.003 0.009 -0.018 0.011 Household head’s ethnicity (1/0) 0.022 0.017 -0.005 0.050 Asset index -0.245 0.068 -0.357 -0.132 Household owns a tractor (1/0) 0.089 0.051 -0.001 0.167 Household received remittances last year (1/0) -0.042 0.028 -0.088 0.003 Household experienced a weather shock last year (1/0) 0.034 0.022 -0.002 0.069 Household experienced an agricultural shock last year (1/0) -0.043 0.024 -0.082 -0.004 Amount of credits received by household last year (logarithm, US$) 0.002 0.003 -0.004 0.008 Number of land plots owned by household -0.021 0.010 -0.038 -0.004 Number of livestock units owned by households 0.002 0.003 -0.002 0.007 Distance from dwelling to plot (km) 0.004 0.002 0.001 0.008 Household applied chemical fertilizers last year (1/0) -0.051 0.019 -0.083 -0.020 Plot size (ha) 0.001 0.004 -0.005 0.006 Plot under grains and legumes (1/0) 0.009 0.017 -0.018 0.036 Plot under vegetables (1/0) -0.062 0.017 -0.089 -0.035 Plot under a mix of crops (grain, legumes and vegetables) (1/0) -0.026 0.034 -0.082 0.030 Djalal Abad (1/0) -0.096 0.022 -0.131 -0.061 Naryn (1/0) -0.042 0.028 -0.088 0.005 Batken (1/0) 0.012 0.022 -0.024 0.048 Osh (1/0) -0.127 0.024 -0.166 -0.088 Talas (1/0) -0.221 0.040 -0.286 -0.155 Chuy (1/0) 0.048 0.023 0.009 0.086 Survey year (2019=1) 0.133 0.015 0.108 0.158 Distance from dwelling to the main road (km) 0.019 0.007 0.008 0.031 Pseudo R2 0.168 N 2788 Source: Tadjiev et al. (2023a). ~ 66 ~ In Uzbekistan, cooperation occurs within narrow social groups already in frequent interaction, such as neighboring farmers (Radnitz et al., 2009). In this regard, informal cooperation helps participating farmers to overcome organizational problems caused by resource shortages and reduce transaction costs (Djanibekov et al., 2015). For instance, when irrigation water service provision becomes unreliable, farmers extend their reliance on mutual self-help by contributing labor or financial means to repair a commonly owned irrigation pump or canal (O’Hara, 2000). Beyond managing water resources, cooperation plays a crucial role in knowledge sharing among Uzbekistan's farming community. The primary challenges lie in the absence of extension services facilitating knowledge exchange or information sharing in agriculture (Kazbekov and Qureshi, 2011). Due to the poor extension system, farmers often seek other sources of information, such as exchanging knowledge with their peers (Kurbanov et al., 2022), for example, through the participation in informal cooperation activities, like irrigation canal cleaning and the repair of common irrigation pump. Furthermore, farmers participating in informal cooperation actively engage in social media groups, facilitating the dissemination of information and the exchange of knowledge (Tadjiev et al., 2023b). 4.2 Conceptual framework I conceptualize that farmers who participate in informal cooperation in irrigation water management are more predisposed to adopting SAPs. This assertion derives from the notion that participation in informal cooperation facilitates the dissemination of knowledge and the exchange of information related to SAPs among participants, thereby fostering higher awareness and comprehension of these practices (Olawuyi and Mushunje, 2020; Willy and Holm-Müller, 2013). Thus, I assume that participation in informal cooperation can serve as a catalyst for greater SAPs adoption among farmers, potentially leading to a broader implementation of such practices within the agricultural community. Moreover, the adoption of ~ 67 ~ new agricultural practices introduces uncertainty, wherein farmers may possess inadequate knowledge concerning the attributes of SAPs (Chavas and Nauges, 2020; Rogers, 2003). This includes understanding the suitability of new SAPs under specific soil conditions, as well as understanding how best to use the SAPs especially when combined with other production factors such as fertilizers (Chavas and Nauges, 2020). Hence, gathering information from early adopter-farmers through participation in informal cooperation in water management may reduce farmers’ resistance to the adoption of SAPs. Several factors exhibit associations with both participation in informal cooperation and the decision to adopt SAPs (figure 4.1). They can be divided into farm and farmer characteristics, farm biophysical characteristics, institutional factors, and locational settings (e g., Dessart et al., 2019; D’Emden et al., 2008; Feder et al., 1985; Ruzzante et al., 2021). The diverse array of factors exhibits varying effects on both the intensity of SAP adoption and farmers’ decision to participate in informal cooperation. Figure 4.1 shows hypothesized signs of the relationship between explanatory variables and dependent variables. Figure 4.1 Conceptual relationship between SAP adoption and informal cooperation, and explanatory variables +/‐ +/‐ +/‐ +/‐ +/‐ Intensity of SAPs adoption Participation in informal cooperation Farm and farmer characteristics (age, education,...) Farm biophysical characteristics (land size, soil fertility,.) Institutional factors (land tenure, participation in trainings, decision making autonomy) Locational settings (near/further in an irrigation canal, distance…) +/‐ ? +/‐ ~ 68 ~ I test whether participation in informal cooperation in water management increases or decreases the intensity of SAP adoption. Social learning refers to the process through which farmers acquire new knowledge, skills, attitudes, or behaviors by interacting with others in their social environment. It involves the sharing and exchange of information, experiences, and practices among farmers to improve farming methods, and enhance productivity and revenues. Through this process, participation in informal cooperation is anticipated to influence the extent of SAP adoption. This model relies on peer-to-peer learning and shared experiences to overcome barriers to adopting new agricultural technologies (Foster and Rosenzweig, 1995). On one hand, informal cooperation among farmers, facilitated by shared water management and knowledge exchange, is seen as a way to enhance SAP adoption, particularly where formal extension services are weak. In contrast, in contexts where low-level awareness regarding SAPs is high in the farming community, this very informal interaction might reinforce traditional, inputintensive cultivation practices, reducing the adoption intensity of SAPs (Bakker et al., 2021; Wagner et al., 2016). Facilitating access to agricultural technology information among farmers enhances cooperation and significantly reduces the transaction costs linked to technology adoption decisions (Ugochukwu and Phillips, 2018). During communal activities, such as hashar or water infrastructure maintenance, farmers engage in open discussions about agronomy, technology processes, and the economic implications of adopting new agricultural practices. This cooperative environment fosters a robust platform for capacity building and the exchange of critical information related to SAPs within the rural agricultural setting (Olawuyi and Mushunje, 2020). Moreover, the influence of neighborhood social dynamics plays a crucial role in fostering social learning, thereby accelerating the uptake of innovative crop cultivation methods (Willy and Holm-Müller, 2013). The interplay between information sharing, social learning, and social capital is identified as instrumental in the adoption of agricultural technologies among local farming communities (Dessart et al., 2019; Marra et al., 2003). ~ 69 ~ 4.3 Data and descriptive analysis The present study utilizes farm survey data derived from two distinct waves conducted within the framework of the AGRICHANGE and SUSADICA 7 projects from the Turkistan (Kazakhstan) and Samarkand (Uzbekistan) provinces. In the study, I focus on Uzbekistan, hence Kazakhstan subsample is excluded. The initial wave of the survey was collected in 2019, followed by a subsequent wave in 2022. In the first wave of 2019, 460 farmers actively participated from the Samarkand province. In the 2022 survey, 450 farmers were surveyed in the Samarkand province. A total of 309 farmers participated in both survey waves, thereby facilitating a longitudinal analysis of agricultural dynamics. A multistage random sampling procedure was used to select farmers for interviews. For this, I selected three districts according to their crop specialization. Pastdargom and Payarik districts in Samarkand are more specialized in cotton cultivation, while farmers in Jomboy district in Samarkand have diversified from cotton to other high-value crops such as vegetables and melons. Identified farmers answered a detailed questionnaire on individual socio-demographic data, individual behavioral perception, as well as farm, field, and location characteristics, and the adoption of SAPs. I applied several conditions to the dataset to fit the research objectives. I pooled the years 2019 and 2022, because more than 30% of farmers participated in only one wave. Additionally, variations among farm management were noted between the two years, where farm business was run by different family members in 2019 and 2022, further justifying the decision to aggregate the datasets for a comprehensive and representative analysis. Following the definition by Piñeiro et al. (2020), several farming practices were covered in the surveys, such as crop rotation, biological pest control methods, laser levelling of fields, low 7 Structured doctoral programme on Sustainable Agricultural Development in Central Asia (SUSADICA), https://www.iamo.de/en/research/research-projects/ ~ 70 ~ tillage of land, direct planting without tillage, intercropping, drip and sprinkler irrigation. For each type of practice, farmers had the option to choose one of the following three answers, thus, (1) “yes and still use it”, (2) “yes, but I stopped using it” and (3) “no, never used it”. I treated the responses “yes and still use it” as adoption, and the other two responses were aggregated into a non-adoption. By doing so, a binary variable was generated for each type of SAP. Figure 4.2 presents the adoption level of each practice. Figure 4 2: SAPs adoption level in Samarkand region in 2018 and 2021 (pooled) Source: Authors. The primary objective of this study is to scrutinize the adoption patterns of low-cost agricultural practices. Consequently, I excluded the assessment of adoption pertaining to laser leveling, drip irrigation, and sprinkler irrigation practices. The methodology involves the consolidation of the utilized SAPs by individual farmers, which is subsequently normalized by Simpson’s (Simpson, 1949) diversity index (e.g., Conrad et al., 2017; Lyson and Welsh, 1993). Therefore, I use the following formula to calculate the intensity of adoption of different SAPs relative to farm size as presented in equation 4.1: 𝑆=∑𝑎2 𝐴2 (4.1) where, 𝑆 is the intensity of SAP adoption, 𝑎 is the area under a particular SAP out of total land area, 𝐴 is a farm’s total land area. This index expresses the intensity of SAP adoption for each ~ 71 ~ farmer and ranges from zero to 2.4. An index value of 0 signifies the absence of SAP application, while a higher numerical value indicates a greater implementation of diversified SAPs across a larger share of the farmer’s land. In the analysis, it is shown that a farmer adopts maximum three SAPs out of five. The theoretically observable maximum value of S=2.4 would imply that more SAPs will be used on a farm’s total area. This index of SAP serves as the principal outcome variable in the following analysis. The survey provides information on which formal and informal form of cooperation farmers participated in regarding irrigation water management. A binary dummy variable was created, representing the treatment variable in the research, thereby categorizing farmers into two distinct groups: "participants in informal cooperation" and "non-participants in informal cooperation". I use this binary variable measuring farmers’ engagement in informal cooperation in “irrigation of fields and control of water distribution”, “repair and cleaning of irrigation and drainage canals”, and “joint maintenance, utilization, construction and repair of irrigation equipment and infrastructure”. If farmers responded that they participated in informal cooperation in irrigation water management - that is, if farmers made informal agreements with other farmers or if they participated in hashar for irrigation activities the values of the variable of interest were coded as one. The dataset of 909 observations included 51 respondents who disclosed participation in water management cooperation purely based on formal agreements. Given the relatively small number of respondents involved exclusively in formal cooperation, I excluded these cases from the estimation model to maintain focus on informal cooperation's impact. Farmers not participating in any form of cooperation, i.e. 351 responses, were assigned a value of zero. Interestingly, Table 4.1 reveals that 32 farmers were involved in both formal and informal cooperation. These individuals, alongside another 475 who were engaged exclusively in informal cooperation, were coded as one in the analysis, highlighting the predominant role of informal mechanisms in irrigation water management. Thus, I used 858 observations in the estimations, including 507 farmers participating in informal cooperation. ~ 72 ~ Table 4.1: Cross tabulation of forms of cooperation in water management Type of cooperation Informal participant non-participant Formal participant 32 51 non-participant 475 351 Total 507 402 Source: Authors. Table 4.2 provides information about the summary statistics of variables across the treatment variable used in the study. The variables are divided into “outcome variable”, which is the intensity of SAP adoption, and “explanatory variables”. Table 4.2: Descriptive statistics of variables across farmers that participated and nonparticipated in informal cooperation (pooled 2019 and 2022) Variables Description Participant (N=507) Nonparticipant (N=351) Mean diff Mean Mean Outcome variable Intensity of SAPs adoption Intensity of SAP adoption to farm size (0a farm does not use any SAPs, a higher index value denotes the adoption of at least one SAP) 0.340 (0.490) 0.386 (0.480) -0.046 Explanatory variables Age Age of farm manager (years) 44.846 (10.224) 44.934 (10.252) -0.088 Education 1 if farmer has special education in agriculture, 0 otherwise 0.444 (0.497) 0.387 (0.488) 0.056 Farm size Total available land of farm (ha) 74.198 (53.805) 79.998 (49.413) -5.800 Agronomy 1 if farmer has own knowledge on agronomy, 0 otherwise 0.462 (0.499) 0.416 (0.494) 0.046 Tractor 1 if farmer owns a tractor, 0 otherwise 0.844 (0.363) 0.823 (0.382) 0.021 Caring opinion A farmer cares about opinion of neighbors, relatives, and other farmers (index of 1-5, where 1doesn’t care and 5 =very much cares) 3.274 (0.753) 3.198 (0.768) 0.076 Training 1 if farmer participates in trainings related to SAPs, 0 otherwise 0.359 (0.480) 0.262 (0.440) 0.097*** Free decision A farmer is free to decide what crop to cultivate and where to sell harvest (index of 1-5, where 1-not free and 5=fully free) 2.398 (1.165) 1.870 (1.200) 0.528*** ~ 73 ~ Table 4.2 cont. Variables Description Participant (N=507) Nonparticipant (N=351) Mean diff Mean Mean Land tenure 1 if farmer perceives not losing land rights in the next 3 years, 0 otherwise 0.469 (0.499) 0.595 (0.492) -0.126*** Soil fertility Index of perceived soil fertility (0-low fertility … 1-high fertility) 0.669 (0.406) 0.615 (0.382) 0.054* WUA supplies water 1 if irrigation water to farm field supplied mostly by local water user association, 0 otherwise 0.712 (0.453) 0.442 (0.497) 0.270*** Canal condition 1 if farmer satisfied about condition of irrigation and drainage canals, 0 otherwise 0.921 (0.270) 0.806 (0.396) 0.115*** Plot location 1 if farm field is located at the head of the water source, 0 otherwise 0.233 (0.423) 0.182 (0.387) 0.050* Distance to house Distance to the house from farm field (km) 4.403 (5.041) 4.098 (5.014) 0.305 Distance to local market Distance to the local market from farm field (km) 12.301 (6.183) 14.610 (6.791) -2.309*** Year 1 if observations belong to 2022, 0 otherwise 0.562 (0.497) 0.330 (0.471) 0.232*** Note: Standard deviation are reported in parentheses; ***, ** and * are significant at p<0.01, p<0.05 and p<0.1 level, respectively. Source: Authors. In the questionnaire, farmers were asked about the significance they attribute to the opinions of neighbors, relatives, and farm colleagues. Respondents were required to select responses on a categorical scale ranging from 1 to 5, where "1" denoted "not at all" and "5" signified "very much". The two categorical questions, namely "how much they care about the opinion of neighbors and relatives" and "how much they care about the opinion of farm colleagues," were used to compute the average, thereby generating a new control variable called "farmers’ caring opinion of neighbors, relatives, and farm colleagues". In addition, the questionnaire also provides information on “to what extent farmers are free in crop cultivation and crop rotation to use” and “how free farmers are in deciding where to sell their main harvested crops.”. The responses to these categorical questions ranged from "1= I cannot decide myself" to "5= It is fully my decision." The control variable "freedom to decide ~ 74 ~ crop cultivation and selling" represents the arithmetic average of farmers’ responses to each of these questions. As indicated before, in the study, main outcome variable is the intensity of SAP use, calculated as in Equation 4.1. The summary statistics show that the intensity of SAP use is slightly higher for non-participant farmers than participants, but the difference is relatively small. The data presented in Table 4.2 also show significant differences in participation in trainings, free decision to crop cultivation and selling, land tenure security, soil fertility, condition of irrigation and drainage, farm field location relative to water resources and distance to the local market from farm field between participant and nonparticipant farmers. Nevertheless, a simplistic comparison of mean differences between participant and nonparticipant farmers fails to consider potential confounding factors contributing to these disparities. Consequently, I employ a state-of-the-art econometric method, namely the marginal treatment effect model, to disentangle biases stemming from self-selection into participation in informal cooperation, and, in turn, to analyze its influence on the intensity of SAP adoption. 4.4 Methodological approach 4.4.1 Methodological approach for the determinants of participation in informal cooperation Following existing literature such as Addai et al. (2023), Andresen (2018), Dubbert et al. (2023), this study employs the marginal treatment effect (MTE) approach. Thus, I assume that participation in the informal cooperation of a farm 𝑖 is a binary variable indicated by 𝐺𝑖. The assumption is that a farmer participates in informal cooperation in water management with other farmers, where they share agricultural knowledge and information that improves the SAP adoption level. Participation in informal cooperation can be expressed as a function of observable and unobservable elements in the following latent variable model: 𝐺𝑖∗=𝛽𝐺(𝑍)−𝑉𝑖 (4.2) ~ 75 ~ with 𝐺𝑖=1 if 𝐺𝑖∗≥0 and 𝐺𝑖=0 otherwise, where 𝐺𝑖 is a binary indicator that equals 1 if a farm participates in informal cooperation, and zero otherwise. 𝑍=(𝑋𝑖,𝑍 𝑖) stand for a vector of observable variables given in figure 4.1 as 𝑋𝑖 that influence the outcome equation of the intensity of SAP adoption, and an instrument for identification, 𝑍 𝑖, excluded from the outcome equation. In our case, 𝑍 𝑖, is the distance to a local market from farm field. 𝛽𝐺 is a vector of parameters to be estimated. 𝑉𝑖 is the unobserved resistance to treatment or participation in informal cooperation, i.e., the error term. The negative sign associated with the error term in the selection equation signifies the unobserved characteristics that might decrease the likelihood of an individual farmer engaging in informal cooperation. In the MTE literature, it is commonly known as the “unobserved resistance" to the treatment (Andresen, 2018; Dubbert et al., 2023). Farmers with high values of 𝑉 are less likely to participate (high resistance to participate) in informal cooperation, compared to farmers with low 𝑉 values who are more likely to participate (low resistance to participate) in informal cooperation. 4.4.2 Methodological approach for the impact of participation in informal cooperation on adoption level of sustainable agricultural practices I set out the model of the relationship between participation in informal cooperation and SAP adoption level by following an approach presented by Dubbert et al. (2023). 𝑆1𝑖 denotes the intensity of SAP adoption of farmer 𝑖 under the assumed condition where the farmer is treated, that is, participates in informal cooperation. 𝑆0𝑖 represents the intensity of SAP adoption under the assumption that the farmer 𝑖 is not treated, and does not participate in informal cooperation. This relationship between SAP adoption intensity 𝑆𝑗𝑖 and participation in informal cooperation can be modelled as follows: 𝑆𝑗𝑖 =𝛽𝑗𝑋𝑖+𝑈𝑗𝑖 𝑗=0,1 (4.3) where 𝑋𝑖 stands for a vector of observable variables as in Equation 4.2. 𝛽𝑗 vector of parameters to be estimated. 𝑈𝑗𝑖 is the error term representing unobserved characteristics that affect SAP ~ 82 ~ points. In contrast, in the participation state farmers’ ownership of a tractor decreases SAP adoption intensity by 37.3 percentage points. Scholars emphasize that adoption of SAPs, such as conservation agriculture, or any new technology require sufficient financial well-being (e.g., Knowler and Bradshaw, 2007). Farmers with their own tractor are considered wealthier and may be more likely adopt SAPs. Such farmers have also greater access to credit as they are able to use their tractors as collateral (Ruzzante et al., 2021). However, in the participation state farmers may cooperate with their peers on sharing tractors and less likely to adopt technologies such as zero tillage (Ngoma, 2018; Tadjiev et al., 2023a). Soil fertility also tends to produce differential effects on treated and untreated farmers. The estimated results demonstrate that SAP adoption intensity may decrease by 21 percentage points for farmers with better soil fertility in the non-participation state. This phenomenon may be attributed to cases where farms with fertile soil might neglect the implementation of soil management methods, such as crop rotation or intercropping. Knowler and Bradshaw (2007) showed an inverse correlation between highly productive soil and the adoption of conservation agricultural practices. However, when farmers with better soil fertility participate in informal cooperation their intensity of SAP adoption increases by 36.2 percentage points. In the untreated state, if a local WUA is primary supplier of irrigation water to farmers, intensity of SAP adoption will decrease by 68.5 percentage points. However, if farmers participate in informal cooperation their intensity of SAP adoption will increase by 64.7 percentage points when the irrigation water is mainly supplied by a local WUA. This highlights the importance of a WUA in encouraging farmers to participate in informal cooperation that can increase the intensity of SAPs adoption The coefficient of land tenure secure in Table 4.4 is positive in the non-participation state, indicating that farmers with more land tenure security will have more intensity of SAP adoption. However, the result is negative but not statistically significant in the participation state. ~ 83 ~ Moreover, in the non-participation state, the intensity of SAP adoption will decrease by 1.6 percentage points for farmers who have farm fields in distance from home dwellings. In our model, we also control year dummy (here 2019 is the reference year) variable to better understand year differences of the intensity of SAP adoption in both the non-participation and participation state. The result shows that in the non-participation state, the intensity of SAP adoption is higher by 113 percentage points in 2022 comparing to 2019, in contrast, in the participation state, the intensity of SAP adoption is lower by 83.3 percentage points in 2022 comparing to 2019. ~ 84 ~ Table 4.4: Outcome equations Variables (1) (2) Outcome (𝛽0) Outcome (𝛽1−𝛽0) Coeff. Std.err [90% confidence interval] Coeff. Std.err [90% confidence interval] Age -0.001 0.005 -0.009 0.006 0.004 0.008 -0.009 0.016 Education 0.024 0.108 -0.153 0.201 -0.004 0.169 -0.282 0.275 Farm size -0.004 0.001 -0.006 -0.002 0.008 0.002 0.004 0.011 Agronomy -0.021 0.088 -0.166 0.125 0.056 0.148 -0.188 0.299 Tractor 0.223 0.118 0.028 0.418 -0.373 0.200 -0.702 -0.044 Caring opinion -0.027 0.059 -0.124 0.070 -0.076 0.095 -0.233 0.080 Training 0.021 0.106 -0.153 0.196 0.274 0.162 0.008 0.540 Free decision -0.097 0.072 -0.215 0.021 0.080 0.112 -0.103 0.264 Land tenure 0.250 0.114 0.063 0.437 -0.222 0.208 -0.565 0.121 Soil fertility -0.210 0.116 -0.400 -0.019 0.362 0.193 0.044 0.680 WUA supplies water -0.685 0.179 -0.979 -0.390 0.647 0.315 0.128 1.167 Canal condition -0.020 0.123 -0.223 0.182 -0.100 0.303 -0.598 0.399 Plot location 0.013 0.118 -0.180 0.207 -0.038 0.180 -0.335 0.259 Distance to house -0.016 0.008 -0.030 -0.002 0.009 0.013 -0.013 0.031 Year 1.132 0.197 0.808 1.456 -0.833 0.301 -1.329 -0.338 Constant 0.232 0.249 -0.178 0.643 -0.196 0.560 -1.118 0.726 Test of observed heterogeneity, p-value 0.005 Test of essential heterogeneity, p-value 0.049 Number of observations 858 Note: Columns 1 and 2 offer the estimates of the intensity of SAPs adoption equation in the non-participation and not participation in informal cooperation states (the difference between participation and non-participation), respectively. The reported test heterogeneity shows whether the treatment effect (𝛽1−𝛽0) varies across the observed covariates (Addai et al., 2023; Andresen, 2018; Dubbert et al., 2023). Source: Authors. ~ 85 ~ 4.5.3 Average and marginal treatment effects estimates The main goal of this study is to better understand how farmers’ participation in informal cooperation in water management tends to impact the intensity of SAP adoption. This section helps in ascertaining whether farmers benefit from participation in informal cooperation and how these effects differ with regard to their unobserved characteristics. The MTE curve in Figure 4.4 illustrates the distribution of marginal returns to treatment over varying levels of unobserved resistance to treatment (referred to as 𝑈𝐺), specifically the resistance to participation in informal cooperation among farmers. It shows a downward slope, with relatively high treatment effects above 2 at the beginning of 𝑈𝐺 distribution and eventually declining to negative effects below -2 at the right end of the distribution, suggesting that the effect of participation in informal cooperation on the intensity of SAP adoption varies with levels of unobserved characteristics. In Figure 4.4, the ATE line stays at around 0.33, and the downward sloping pattern implies positive selection on unobservable patterns. This finding thus tells us that, given the unobserved characteristics, farmers who are more likely to participate in informal cooperation in water management have higher intensity of SAP adoption from participation. This pattern of unobserved heterogeneity in returns to participation is statistically significant at the 5% level (the p-values for the test of unobserved heterogeneity is given in Table 4.4) for the intensity of SAP adoption. Consequently, a lower level of unobserved resistance to participation (high propensity to participate in informal cooperation) is linked with a higher intensity of SAP adoption, but the intensity of SAP adoption tends to decrease as the unobserved resistance to participation increases. ~ 86 ~ Figure 4.4: MTE curve for the intensity of SAPs adoption Source: Authors. Table 4.5 puts forward a summary of the treatment effects in terms of SAP adoption from the participation in informal cooperation. The result of the ATE shows that participation in informal cooperation significantly increases the intensity of SAPs adoption for the average farmer. The ATE estimation for SAPs adoption is 0.328, which indicates that randomly selecting farmers from the population and having them participate in informal cooperation increases the intensity of SAP adoption by 32.8 percentage points. The findings of the ATT, which put more weight on farmers with high propensity scores for participation, suggest that participation in informal cooperation significantly – in this case, by 98.3 percentage points - increases SAPs adoption intensity for the average farmer who participates in informal cooperation. On the other hand, the ATUT estimates presents that participation in informal cooperation would decrease the intensity of SAP adoption for the average untreated farmer, but the hypothesis that ATUT equal to zero cannot be rejected. A general picture of the estimates shows that the coefficient of ATT is greater than the ATE, which is also greater than the ATUT: ATT (0.983)>ATE (0.328)>ATUT (−0.616). This ~ 87 ~ ranking of three effect measurements indicates positive selection on gains, where farmers who are probable to participate in informal cooperation tend to benefit more from participation in terms of the intensity of SAP adoption. The finding confirms the study by Willy and Holm-Müller (2013), who found that participation in collective action enhances soil conservation efforts. Table 4.5 also provides the result of an estimation of the local average treatment effect (LATE). The LATE estimate for SAP adoption is 0.664. This indicates that farmers who participate in informal cooperation due to closer location to the local market increase the intensity of SAP adoption by 66.4 percentage points. Table 4.5: Average treatment effects The intensity of SAP adoption Coeff. Std.err. [90% confidence interval] ATE 0.328 0.183 0.026 0.629 ATT 0.983 0.466 0.216 1.750 ATUT -0.616 0.514 -1.462 0.230 LATE 0.664 0.170 0.383 0.944 Test of essential heterogeneity, p-value 0.049 Source: Authors. ~ 88 ~ 5 GENERAL CONCLUSIONS AND POLICY IMPLICATIONS The present section starts by summarizing the research findings of three empirical chapters of the dissertation. Following this, it derives important policy messages based on the research findings. Finally, it presents research limitations and outlook for further research. 5.1 Synthesis of research findings Overall, the three empirical chapters of the dissertation provide a comprehensive analysis of SAPs in Central Asia and emphasize their diverse and context-specific nature, highlighting the potential of practices like crop rotation and zero tillage in improving farm outcomes, and the importance of informal cooperation in water management for expansion of sustainable agriculture. Chapter 2 delves into crop rotation in Kazakhstan and Uzbekistan, highlighting its variable impact on cotton yields and net revenues, and how factors such as farmer age, educational training, and perceptions of land tenure security play a role. Chapter 3's focus on zero tillage in Kyrgyzstan sheds light on its economic trade-offs for smallholders, balancing input cost savings against increased labor and herbicide expenses. Chapter 4’s exploration of informal cooperation in water management in Uzbekistan reveals its influence on the adoption of sustainable practices, underscored by variables like farmers’ agronomy knowledge and the quality of agricultural infrastructure. These chapters underscore the diverse and context-specific nature of adoption of sustainable agriculture in Central Asia. Chapter 2 examined factors influencing the adoption of crop rotation and its impact on cotton yields and net revenues of cotton growers by applying parametric and nonparametric empirical methods on cross-sectional farm survey data from Central Asia’s two cotton-growing areas. The results demonstrated that sample selection bias could have occurred if the impact of crop rotation was estimated without considering observable and unobservable factors in the adoption decision. Thus, to control for the selection bias issues arising from observable and ~ 89 ~ unobservable factors, the ESR model is employed that estimates differential impacts of adoption of crop rotation on continuous outcome variables like cotton yields and net revenues. The model results suggested the presence of selection bias. After controlling for the bias, the estimation results showed that crop rotation increases cotton yields by 5% and net revenue by 19% in Uzbekistan. However, an opposite (negative) impact is revealed for Kazakhstan. Such unexpected impact of crop rotation on performance of cotton growers in Kazakhstan is explained by the fact that the existing institutional environment and infrastructure in Kazakhstan after the cotton sector reform actually provide advantage to farmers who practice conventional cotton monoculture. The large role here can be assigned to contractual arrangements with private gins who favor farmers cultivating cotton each year, i.e. cotton monoculture (Petrick et al., 2017). As a result, Kazakh farmers who opt for the soil-improving crop rotation scheme are most likely to end up outside such contractual arrangements and lose timely access to external inputs like cotton seeds, fertilizer, pesticides and machinery supplied by a private ginnery through contract farming. Furthermore, the results provided insights into factors affecting farmers’ decision to adopt crop rotation as well as its impact on cotton yields and net returns. In Kazakhstan, crop rotation adoption positively associates with farmers’ age, participation in farm trainings, perception about irrigation canal condition, and village share of adopters. The results also suggested that Kazakh farmers who learned about new technologies and agronomy from peers and neighbors were less likely to use crop rotation. In Uzbekistan, the probability of adoption of crop rotation is higher among credit-rationed farmers and those who perceive land tenure as secure. Uzbek farmers in remote areas, at irrigation canal heads, and who receive agricultural information from the internet, media, and radio are less likely to adopt crop rotation. The results showed that production inputs were important determinants of cotton cultivation in both study areas. In Kazakhstan, the intensity of fertilizer use affects cotton yields of non-adopters, while size of cotton sown area is an important factor for adopters. In Uzbekistan, cotton sown area is ~ 90 ~ positively related with net returns of non-adopters, and fertilizer use positively affects cotton yields of non-adopters. In Kazakhstan, contract farming with private ginneries raises outcomes in ‘conventional cotton’ monoculture, but puts adopters of crop rotation in disadvantage. Furthermore, in both countries, crop rotation allows better use of the advantage of close location to irrigation canals. Chapter 3 measured the adoption determinants and resource-saving effects of zero tillage among smallholders in Kyrgyzstan by using parametric and nonparametric empirical methods on two waves of longitudinal data. The findings suggest that zero tillage can be an attractive option for resource-poor smallholders located in remote areas. The probability of zero tillage adoption is positively associated with household head’s employment in agriculture, and distance of household dwellings to household fields and main road. Furthermore, the probability of zero tillage adoption is negatively related to household wealth measured in asset index and number of household plots as well as fertilizer application. The findings suggest that zero tillage can generate tangible benefits to smallholders in terms of reducing input costs by 15%. At the same time, zero tillage adoption affects the structure of production costs. As expected it reduces machinery costs for land preparation and seeding by almost 23%. As a result, of substituting the machinery services with external workers, zero tillage can increase hired labor costs by 13%. Furthermore, zero tillage increases herbicide costs by 15%. Chapter 4 provides first-time estimation of marginal effects of participation in informal cooperation in water management on the adoption intensity of sustainable agricultural practices in a developing country setting. To do so, it uses two years 2019 and 2022 survey data of farmers in Uzbekistan and employed empirical methods. The MTE model was used to account for selection bias, and observable and unobservable heterogeneity among farmers. Furthermore, I investigate main determinants of farmers’ participation in informal cooperation. The findings showed that the probability of participation in informal cooperation is positively related to farmers’ knowledge on agronomy, tractor ownership, participation in SAP trainings, ~ 91 ~ decision making freedom, sources of irrigation water supply, and quality of irrigation and drainage infrastructure. Furthermore, the probability of participation in informal cooperation is negatively associated with land tenure security and distance to the local market. The findings indicate that participation in informal cooperation emerges as a favorable prospect for farmers who possess educational background in agriculture and have improved condition of irrigation and drainage infrastructure. The empirical results showed significant heterogeneity in the influence of informal cooperation on the intensity of SAP adoption. Notably, the results pertaining to observable characteristics suggest that farmers possessing own tractors tend to more likely participate in informal cooperation. However, its treatment effects reveal that the intensity of SAP adoption level becomes lower compared to nonparticipant farmers. Additionally, participation in informal cooperation exhibits a propensity to positively influence the intensity of SAP adoption, particularly among farmers endowed with larger land holdings and better soil fertility. Finally, the results on treatment effects revealed that farmers who are likely to participate in informal cooperation tend to benefit more from the participation in terms of higher intensity of SAP adoption. The three empirical chapters underscore the crucial role of SAPs in enhancing farm outcomes as exemplified in different settings of Central Asia. The analyzed practices, including crop rotation and zero tillage, demonstrate significant potential for improving crop yields and farm revenues. Furthermore, the studies highlight the pivotal role of informal cooperation in water management. This cooperation is key to enhancing the adoption intensity of SAP, reflecting the importance of community-level engagement in water resource management for fostering agricultural sustainability in the region. 5.2 Policy recommendations Empirical findings presented in this dissertation allow to derive policy recommendations related to the promotion of the adoption of sustainable agricultural practices among farmers and ~ 98 ~ Brinch, C. N., Mogstad, M., & Wiswall, M. (2017). Beyond LATE with a discrete instrument. Journal of Political Economy, 125(4), 985-1039. https://doi.org/10.1086/692712 . Brooker, R. W., Bennett, A. E., Cong, W. F., Daniell, T. J., George, T. S., Hallett, P. D., & White, P. J. (2015). Improving intercropping: a synthesis of research in agronomy, plant physiology and ecology. New Phytologist, 206(1), 107-117. https://doi.org/10.1371/journal.pone.0113984 Brück, T., Esenaliev, D., Kroeger, A., Kudebayeva, A., Mirkasimov, B., & Steiner, S. (2014). Household survey data for research on well-being and behavior in Central Asia. Journal of Comparative Economics, 42(3), 819-835. https://doi.org/10.1016/j.jce.2013.02.003. Cardenas, J. C., Rodriguez, L. A., & Johnson, N. (2011). Collective action for watershed management: field experiments in Colombia and Kenya. Environment and Development Economics, 16(3), 275-303. https://doi.org/10.1017/S1355770X10000392 Chatterjee, R., & Acharya, S. K. (2021). Dynamics of conservation agriculture: A societal perspective. Biodiversity and Conservation, 30(6), 1599–1619. https://doi.org/10.1007/s10531-021-02161-3 Chavas, J. P., & Nauges, C. (2020). Uncertainty, learning, and technology adoption in agriculture. Applied Economic Perspectives and Policy, 42(1), 42-53. https://doi.org/10.1002/aepp.13003 Conrad, C., Löw, F., & Lamers, J. P. (2017). Mapping and assessing crop diversity in the irrigated Fergana Valley, Uzbekistan. Applied Geography, 86, 102-117. https://doi.org/10.1016/j.apgeog.2017.06.016 Cornelissen, T., Dustmann, C., Raute, A., & Schönberg, U. (2018). Who benefits from universal child care? Estimating marginal returns to early child care attendance. Journal of Political Economy, 126(6), 2356-2409. https://doi.org/10.1086/699979 ~ 99 ~ Dessart, F. J., Barreiro-Hurlé, J., & Van Bavel, R. (2019). Behavioural factors affecting the adoption of sustainable farming practices: a policy-oriented review. European Review of Agricultural Economics, 46(3), 417-471. https://doi.org/10.1093/erae/jbz019 Di Falco, S., Veronesi, M., & Yesuf, M. (2011). Does adaptation to climate change provide food security? A micro‐perspective from Ethiopia. American Journal of Agricultural Economics, 93(3), 829-846. https://doi.org/10.1093/ajae/aar006 Djanibekov, N., K. Van Assche, I. Bobojonov, and J.P. Lamers. (2012). Farm restructuring and land consolidation in Uzbekistan: New farms with old barriers. Europe-Asia Studies 64 (6), 11011126. https://doi.org/10.1080/09668136.2012.691720 Djanibekov, N., Hornidge A.-K., & Ul-Hassan, M. (2012b). From joint experimentation to laissezfaire: Transdisciplinary innovation research for the institutional strengthening of a water users association in Khorezm, Uzbekistan. Journal of Agricultural Education and Extension, 18 (4), 409-423. https://doi.org/10.1080/1389224X.2012.691785 Djanibekov, N., Djanibekov, U., Sommer, R., & Petrick, M. (2015). Cooperative agricultural production to exploit individual heterogeneity under a delivery target: The case of cotton in Uzbekistan. Agricultural Systems, 141, 1-13. https://doi.org/10.1016/j.agsy.2015.09.002 Dubbert, C. (2019). Participation in contract farming and farm performance: Insights from cashew farmers in Ghana. Agricultural Economics, 50(6), 749-763. https://doi.org/10.1111/agec.12522 Dubbert, C., Abdulai, A., & Mohammed, S. (2023). Contract farming and the adoption of sustainable farm practices: Empirical evidence from cashew farmers in Ghana. Applied Economic Perspectives and Policy, 45(1), 487-509. https://doi.org/10.1002/aepp.13212 ~ 100 ~ D’Emden, F. H., Llewellyn, R. S., & Burton, M. P. (2008). Factors influencing adoption of conservation tillage in Australian cropping regions. Australian Journal of Agricultural and Resource Economics, 52(2), 169-182. https://doi.org/10.1111/j.1467-8489.2008.00409.x El-Shater, T., Mugera, A., & Yigezu, Y. A. (2020). Implications of adoption of zero tillage (ZT) on productive efficiency and production risk of wheat production. Sustainability, 12(9), 3640. https://doi.org/10.3390/su12093640. Erenstein, O., Farooq, U., Malik, R. K., & Sharif, M. (2008). On-farm impacts of zero tillage wheat in South Asia's rice–wheat systems. Field Crops Research, 105(3), 240-252. https://doi.org/10.1016/j.fcr.2007.10.010. FAO. (1989). The state of food and agriculture. FAO, Rome. https://openknowledge.fao.org/handle/20.500.14283/t0162e FAO. (2013). Conservation Agriculture in Central Asia: Status, Policy, Institutional Support, and Strategic Framework for its Promotion. FAO Sub-Regional Office for Central Asia (FAO-SEC), Ankara. FAO. (2015). Measuring Sustainability in Cotton Farming Systems Towards a Guidance Framework. FAO, Rome. https://openknowledge.fao.org/handle/20.500.14283/i4170e FAO. (2020). Smallholders and family farms in Kyrgyzstan - Country study report. FAO, Budapest. https://doi.org/10.4060/ca9826en. FAO. (2023). What is conservation agriculture? Retrieved 2023, FAO. https://www.fao.org/conservation-agriculture/overview/what-is-conservationagriculture/en/ Fay, M., Block, R., & Ebinger, J. (Eds.). (2010). Adapting to climate change in Eastern Europe and Central Asia. World Bank Publications. ~ 101 ~ Feder, G., Just, R. E., & Zilberman, D. (1985). Adoption of agricultural innovations in developing countries: A survey. Economic Development and Cultural Change, 33(2), 255-298. https://doi.org/10.1086/451461 Fischer, E., & Qaim, M. (2012). Linking smallholders to markets: determinants and impacts of farmer collective action in Kenya. World Development, 40(6), 1255-1268. https://doi.org/10.1016/j.worlddev.2011.11.018 Foster, A. D., & Rosenzweig, M. R. (1995). Learning by doing and learning from others: Human capital and technical change in agriculture. Journal of political Economy, 103(6), 1176-1209. https://doi.org/10.1086/601447 Franz, J., Bobojonov, I., & Egamberdiev, O. (2009). Assessing the economic viability of organic cotton production in Uzbekistan: a first look. Journal of Sustainable Agriculture, 34(1), 99119. https://doi.org/10.1080/10440040903396821 Filmer, D., & Pritchett, L. H. (2001). Estimating wealth effects without expenditure data—or tears: an application to educational enrollments in states of India. Demography, 38(1), 115132. https://doi.org/10.1353/dem.2001.0003. Ghimire, R., Adhikari, K. R., Chen, Z. S., Shah, S. C., & Dahal, K. R. (2012). Soil organic carbon sequestration as affected by tillage, crop residue, and nitrogen application in rice–wheat rotation system. Paddy and Water Environment, 10, 95-102. https://doi.org/10.1007/s10333-011-0268-0 Glaze-Corcoran, S., Hashemi, M., Sadeghpour, A., Jahanzad, E., Afshar, R. K., Liu, X., & Herbert, S. J. (2020). Understanding intercropping to improve agricultural resiliency and environmental sustainability. In Donald L. Sparks (Ed.), Advances in Agronomy, 162, 199256. https://doi.org/10.1016/bs.agron.2020.02.004 ~ 102 ~ Greenland, S., S. J. Senn, K. J. Rothman, J. B. Carlin, C. Poole, S. N. Goodman, & Altman D. G. (2016). Statistical tests, P values, confidence intervals, and power: A guide to misinterpretations. European Journal of Epidemiology, 31(4), 337-350. https://doi.org/10.1007/s10654-016-0149-3 Guadagni, M., & Fileccia, T. (2009). The Kyrgyz Republic: Farm Mechanization and Agricultural Productivity. FAO, Rome, Italy. http://hdl.handle.net/10986/19476. Ha, T. M., Manevska-Tasevska, G., Jäck, O., Weih, M., & Hansson, H. (2023). Farmers’ intention towards intercropping adoption: the role of socioeconomic and behavioural drivers. International Journal of Agricultural Sustainability, 21(1), 2270222. https://doi.org/10.1080/14735903.2023.2270222 Hamidov, A., Kasymov, U., Salokhiddinov, A., & Khamidov, M. (2020). How can intentionality and path dependence explain change in water-management institutions in Uzbekistan?. International Journal of the Commons , 14(1), 16-29. https://doi.org/10.5334/ijc.947 Hamidov, A., U. Kasymov, N. Allahverdiyeva, N., & Schleyer C. (2022). Governance of technological innovations in water and energy use in Uzbekistan. International Journal of Water Resources Development, 1-17. https://doi.org/10.1080/07900627.2022.2062706 Heckman, J. (1979). Sample selection as a specification bias. Econometrica, 47(1), 153-161. https://doi.org/10.2307/1912352 Heckman, J. J., & Vytlacil, E. (2005). Structural equations, treatment effects, and econometric policy evaluation 1. Econometrica, 73(3), 669-738. https://doi.org/10.1111/j.14680262.2005.00594.x Hirschauer, N., Grüner, S., Mußhoff, O., Becker, C., & Jantsch, A. (2020). Can p-values be meaningfully interpreted without random sampling? Statistics Surveys 14, 71-91. https://doi.org/10.1214/20-SS129 ~ 103 ~ Hoffmann, M. P., & Frodsham, A. (1993). Natural enemies of vegetable insect pests. Cooperative Extension, Cornell University, Ithica, 63 p. Hornidge, A. K., Shtaltovna, A., & Schetter, C. (2016). Introduction: Independence, Transformation and the Search for a Future in Agriculture. In Hornidge, A. K., Shtaltovna, A., & Schetter, C (Eds.), Agricultural Knowledge and Knowledge Systems in Post-Soviet Societies. Peter Lang. Imbens, G.W. (2021). Statistical significance, p-values, and the reporting of uncertainty. Journal of Economic Perspectives, 35(3), 157-174. https://doi.org/10.1257/jep.35.3.157 Issahaku, G., & A. Abdulai (2020). Adoption of climate-smart practices and its impact on farm performance and risk exposure among smallholder farmers in Ghana. Australian Journal of Agricultural and Resource Economics, 64(2), 396-420. https://doi.org/10.1111/14678489.12357 Jaleta, M., Kassie, M., Tesfaye, K., Teklewold, T., Jena, P. R., Marenya, P., & Erenstein, O. (2016). Resource saving and productivity enhancing impacts of crop management innovation packages in Ethiopia. Agricultural Economics, 47(5), 513-522. https://doi.org/10.1111/agec.12251 Jaleta, M., Baudron, F., Krivokapic-Skoko, B., & Erenstein, O. (2019). Agricultural mechanization and reduced tillage: Antagonism or synergy? International Journal of Agricultural Sustainability, 17(3), 219-230. https://doi.org/10.1080/14735903.2019.1613742 Jalilova, G., Knipe, D., Hardy, W. (2019). Climate-Smart Agriculture for the Kyrgyz Republic. Washington, D.C.: World Bank Group. https://coilink.org/20.500.12592/jx8rmc Jansen, H. G., Pender, J., Damon, A., Wielemaker, W., & Schipper, R. (2006). Policies for sustainable development in the hillside areas of Honduras: A quantitative livelihoods ~ 104 ~ approach. Agricultural Economics, 34(2), 141-153. https://doi.org/10.1111/j.15740864.2006.00114.x. Karimov, A. A. (2014). Factors affecting efficiency of cotton producers in rural Khorezm, Uzbekistan: Re-examining the role of knowledge indicators in technical efficiency improvement. Agricultural and Food Economics 2 (1): 1-16. https://doi.org/10.1186/s40100014-0007-0 Kazbekov, J., & Qureshi, A. S. (2011). Agricultural extension in Central Asia: Existing strategies and future needs. IWMI Working paper 145. Colombo, Sri Lanka: International Water Management Institute. Keil, A., Mitra, A., McDonald, A., & Malik, R. K. (2020). Zero-tillage wheat provides stable yield and economic benefits under diverse growing season climates in the Eastern Indo-Gangetic plains. International Journal of Agricultural Sustainability, 18(6), 567-593. https://doi.org/10.1080/14735903.2020.1794490 Khonje, M. G., Manda, J., Mkandawire, P., Tufa, A. H., & Alene, A. D. (2018). Adoption and welfare impacts of multiple agricultural technologies: evidence from eastern Zambia. Agricultural Economics, 49(5), 599-609. https://doi.org/10.1111/agec.12445. Kienzler, K. M., Lamers, J. P. A., McDonald, A., Mirzabaev, A., Ibragimov, N., Egamberdiev, O., & Akramkhanov, A. (2012). Conservation agriculture in Central Asia—What do we know and where do we go from here?. Field Crops Research, 132, 95-105. https://doi.org/10.1016/j.fcr.2011.12.008 Knowler, D., & Bradshaw, B. (2007). Farmers’ adoption of conservation agriculture: A review and synthesis of recent research. Food Policy 32, 25-48. https://doi.org/10.1016/j.foodpol.2006.01.003 ~ 105 ~ Kumari, M., Srivastava, A., Babu Sah, S., & Subhashini. (2022). Biological Control of Agricultural Insect Pests. In Ranz, R. E. R. (Eds.), Insecticides: impact and benefits of its use for humanity. IntechOpen. Kurbanov, Z., Tadjiev, A., Djanibekov, N. (2022). Adoption of sustainable agricultural practices and investments in productive assets in irrigated areas of Central Asia: Farm-survey evidence from Kazakhstan and Uzbekistan. IAMO Annual 24: 69-79. Kuhn, L., & Bobojonov, I. (2021). The role of risk rationing in rural credit demand and uptake: lessons from Kyrgyzstan. Agricultural Finance Review, 83(1), 1-20. https://doi.org/10.1108/AFR-04-2021-0039. Lee, D. R. (2005). Agricultural sustainability and technology adoption: Issues and policies for developing countries. American Journal of Agricultural Economics, 87(5), 1325-1334. https://www.jstor.org/stable/3697714 Lee, N., & Thierfelder, C. (2017). Weed control under conservation agriculture in dryland smallholder farming systems of southern Africa. A review. Agronomy for Sustainable Development, 37(5), 48. https://doi.org/10.1007/s13593-017-0453-7. Lerman, Z., & Sedik, D. (2018). Transition to smallholder agriculture in Central Asia. Journal of Agrarian Change, 18(4), 904-912. https://doi.org/10.1111/joac.12282. Liu, M. Y. (2017). Uzbek Political Thinking in the Third Decade of Independence. In: Laruelle, M. (ed.) Constructing the Uzbek State: Narratives of Post-Soviet Years. Lexington Books, pp. 119-148 Löw, F., C. Biradar, E. Fliemann, J.P. Lamers, & Conrad C. (2017). Assessing gaps in irrigated agricultural productivity through satellite Earth observations: A case study of the Fergana Valley, Central Asia. International Journal of Applied Earth Observation and Geoinformation 59: 118-134. doi: https://doi.org/10.1016/j.jag.2017.02.014 ~ 106 ~ Lyson, T. A., & Welsh, R. (1993). The Production Function, Crop Diversity, and the Debate Between Conventional and Sustainable Agriculture 1. Rural Sociology, 58(3), 424-439. https://doi.org/10.1111/j.1549-0831.1993.tb00503.x Maddala, G. S. (1983). Limited-dependent and qualitative variables in econometrics (No. 3). Cambridge University Press. Manda, J., Alene, A. D., Gardebroek, C., Kassie, M., & Tembo, G. (2016). Adoption and impacts of sustainable agricultural practices on maize yields and incomes: Evidence from rural Zambia. Journal of Agricultural Economics, 67(1), 130-153. https://doi.org/10.1111/14779552.12127 Marra, M., Pannell, D. J., & Ghadim, A. A. (2003). The economics of risk, uncertainty and learning in the adoption of new agricultural technologies: where are we on the learning curve?. Agricultural Systems, 75(2-3), 215-234. https://doi.org/10.1016/S0308-521X(02)00066-5 McNamara, K. T., Wetzstein, M. E., & Douce, G. K. (1991). Factors affecting peanut producer adoption of integrated pest management. Applied Economic Perspectives and Policy, 13(1), 129-139. https://doi.org/10.2307/1349563 Meinzen-Dick, R., Raju, K. V., & Gulati, A. (2002). What affects organization and collective action for managing resources? Evidence from canal irrigation systems in India. World Development, 30(4), 649-666. https://doi.org/10.1016/S0305-750X(01)00130-9 Mirzabaev, A., Goedecke, J., Dubovyk, O., Djanibekov, U., Le, Q. B., & Aw-Hassan, A. (2016). Economics of land degradation in Central Asia. In Nkonya, E., Mirzabaev, A., & Von Braun, J. (Eds.), Economics of land degradation and improvement–A global assessment for sustainable development (pp. 261-290). Springer, Cham. Mohammad, W., Shah, S. M., Shehzadi, S., & Shah, S. A. (2012). Effect of tillage, rotation and crop residues on wheat crop productivity, fertilizer nitrogen and water use efficiency and ~ 107 ~ soil organic carbon status in dry area (rainfed) of north-west Pakistan. Journal of Soil Science and Plant Nutrition, 12(4), 715-727. http://dx.doi.org/10.4067/S0718-95162012005000027 Montt, G., & Luu, T. (2020). Does conservation agriculture change labour requirements? Evidence of sustainable intensification in Sub-Saharan Africa. Journal of Agricultural Economics, 71(2), 556–580. https://doi.org/10.1111/1477-9552.12353 Mueller, L., Suleimenov, M., Karimov, A., Qadir, M., Saparov, A., Balgabayev, N., & Lischeid, G. (2014). Land and water resources of Central Asia, their utilisation and ecological status. In: Mueller, L., Saparov, A., Lischeid, G. (eds) Novel Measurement and Assessment Tools for Monitoring and Management of Land and Water Resources in Agricultural Landscapes of Central Asia (3-59). Environmental Science and Engineering. Springer, Cham. https://doi.org/10.1007/978-3-319-01017-5_1 Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica, 69-85. https://doi.org/10.2307/1913646. Musafiri, C. M., Kiboi, M., Macharia, J., Ng'etich, O. K., Okoti, M., Mulianga, B., & Ngetich, F. K. (2022). Does the adoption of minimum tillage improve sorghum yield among smallholders in Kenya? A counterfactual analysis. Soil and Tillage Research, 223, 105473. https://doi.org/10.1016/j.still.2022.105473. Ngoma, H. (2018). Does minimum tillage improve the livelihood outcomes of smallholder farmers in Zambia?. Food Security, 10(2), 381-396.. https://doi.org/10.1007/s12571-0180777-4. Nichols, V., Verhulst, N., Cox, R., & Govaerts, B. (2015). Weed dynamics and conservation agriculture principles: A review. Field Crops Research, 183, 56-68. https://doi.org/10.1016/j.fcr.2015.07.012 ~ 114 ~ Wang, Z. G., Jin, X., Bao, X. G., Li, X. F., Zhao, J. H., Sun, J. H., & Li, L. (2014). Intercropping enhances productivity and maintains the most soil fertility properties relative to sole cropping. PLoS ONE, 9(12), e113984. https://doi.org/10.1371/journal.pone.0113984 Wainaina, P., Tongruksawattana, S., & Qaim, M. (2016). Tradeoffs and complementarities in the adoption of improved seeds, fertilizer, and natural resource management technologies in Kenya. Agricultural Economics, 47(3), 351-362. https://doi.org/10.1111/agec.12235. Willy, D. K., & Holm-Müller, K. (2013). Social influence and collective action effects on farm level soil conservation effort in rural Kenya. Ecological Economics, 90, 94-103. https://doi.org/10.1016/j.ecolecon.2013.03.008 Wolfgramm, B., Shigaeva, J., Nekushoeva, G., Bonfoh, B., Breu, T. M., Liniger, H. P., & Maselli, D. (2010). Kyrgyz and Tajik land use in transition: Challenges, responses and opportunities. In H. Hurni & U. Wiesmann (Eds.), Global Change and Sustainable Development: A Synthesis of Regional Experiences from Research Partnerships. Perspectives of the Swiss National Centre of Competence in Research (NCCR) North-South, University of Bern, Vol. 5. Bern,: Geographica Bernensia, pp 241–254. https://boris.unibe.ch/id/eprint/5902 Wolfgramm, B., Shigaeva, J., & Dear, C. (2015). The research–action interface in sustainable land management in Kyrgyzstan and Tajikistan: challenges and recommendations. Land Degradation & Development, 26(5), 480-490. https://doi.org/10.1002/ldr.2372 World Bank, (2023). World Development Indicators. The World Bank Group. Accessed on 20 January 2023. https://datatopics.worldbank.org/world-development-indicators/ Wu, F., Guo, X., & Guo, X. (2023). Cooperative membership and new technology adoption of family farms: Evidence from China. Annals of Public and Cooperative Economics, 94(3), 719739. https://doi.org/10.1111/apce.12433 ~ 115 ~ Xue, C., Qiao, D., & Aziz, N. (2022). Influence of natural disaster shock and collective action on farmland transferees’ no-tillage technology adoption in China. Land, 11(9), 1511. https://doi.org/10.3390/land11091511 Yigezu, Y. A., Mugera, A., El-Shater, T., Aw-Hassan, A., Piggin, C., Haddad, A., Khalil, Y., & Loss, S. (2018). Enhancing adoption of agricultural technologies requiring high initial investment among smallholders. Technological Forecasting and Social Change, 134, 199–206. https://doi.org/10.1016/j.techfore.2018.06.006 Yigezu, Y. A., & El‐Shater, T. (2021). Socio‐economic impacts of zero and reduced tillage in wheat fields of the Moroccan drylands. Agricultural Economics, 52(4), 645-663. https://doi.org/10.1111/agec.12640. Zeweld, W., Van Huylenbroeck, G., Tesfay, G., & Speelman, S. (2017). Smallholder farmers' behavioural intentions towards sustainable agricultural practices. Journal of Environmental Management, 187, 71-81. https://doi.org/10.1016/j.jenvman.2016.11.014 Zeweld, W., Van Huylenbroeck, G., Tesfay, G., Azadi, H., & Speelman, S. (2020). Sustainable agricultural practices, environmental risk mitigation and livelihood improvements: Empirical evidence from Northern Ethiopia. Land Use Policy, 95, 103799. https://doi.org/10.1016/j.landusepol.2019.01.002 Zhang, S., Sun, Z., Ma, W., & Valentinov, V. (2020). The effect of cooperative membership on agricultural technology adoption in Sichuan, China. China Economic Review, 62, 101334. https://doi.org/10.1016/j.chieco.2019.101334 Zhao, J., Y. Yang, K. Zhang, J. Jeong, Z. Zeng, and H. Zang (2020). Does crop rotation yield more in China? A meta-analysis. Field Crops Research, 245, 107659. https://doi.org/10.1016/j.fcr.2019.107659 ~ 116 ~ Zheng, H., Ma, W., & Li, G. (2021). Adoption of organic soil amendments and its impact on farm performance: evidence from wheat farmers in China. Australian Journal of Agricultural and Resource Economics, 65(2), 367-390. https://doi.org/10.1111/1467-8489.12406 Zhunusova, E., & Herrmann, R. (2018). Development impacts of international migration on “sending” communities: The case of rural Kyrgyzstan. European Journal of Development Research, 30, 871-891. https://doi.org/10.1057/s41287-018-0136-5 Zinzani, A. (2016). Hydraulic bureaucracies and irrigation management transfer in Uzbekistan: The case of Samarkand Province. International Journal of Water Resources Development, 32(2), 232-246. https://doi.org/10.1080/07900627.2015.1058765 ~ 117 ~ APPENDIX Table A1: Falsification test for instrumental variables Kazakhstan Uzbekistan Joint significance test p-value Joint significance test p-value Crop rotation adoption (probit model regression) χ2 (3)=7.81 0.05 χ2 (3)= 6.93 0.03 Net returns from cotton for non-adopters F(3, 197) = 0.55 0.65 F(2, 227) =0.22 0.80 Cotton yield for nonadopters F(3, 197) = 0.78 0.51 F(2, 227) =0.18 0.83 Note: IV are Information source about new technologies and agronomy (i.e., information from other farms and community, information from media, internet and radio), and the village share of adopters of crop rotation (only for Kazakhstan for estimating the crop rotation impact). Source: Based on AGRICHANGE 2019 farm survey data. ~ 118 ~ Table A2 Probit estimation on determinants of adoption decision of crop rotation among cotton growers Kazakhstan Uzbekistan Marginal effect 90% Confidence Interval Marginal effect [90% Confidence Interval] Farmer’s age (year) [age] 0.003 (0.002) 0.001 0.006 0.001 (0.002) -0.003 0.004 Farmer has education in agriculture (1/0) [aredu] -0.091 (0.057) -0.185 0.002 -0.022 (0.044) -0.094 0.050 Farmer perceives canal condition as good (1/0) [cancon] 0.081 (0.049) 0.001 0.161 0.075 (0.051) -0.009 0.159 Credit-rationed farmer (1/0) [credrat1] -0.080 (0.056) -0.173 0.013 0.109 (0.043) 0.039 0.180 Farmer participates in farm trainings (1/0) [parttraining] 0.319 (0.051) 0.235 0.403 -0.063 (0.052) -0.149 0.022 Share of land with good fertility (0-1) [goodfert] 0.014 (0.052) -0.072 0.101 0.033 (0.051) -0.051 0.117 Distance to the district center (km) [discenter] -0.001 (0.002) -0.003 0.002 -0.009 (0.003) -0.014 -0.004 Farm fields located irrigation canal head (1/0) [locwrt] 0.052 (0.052) -0.034 0.138 -0.117 (0.057) -0.212 -0.022 Farmer receives information about new technologies and agronomy from other farms and neighbors (1/0) [infstechagr] -0.083 (0.046) -0.160 -0.007 -0.018 (0.045) -0.092 0.056 Farmer receives information about new technologies and agronomy from media, internet or radio (1/0) [media_1] 0.058 (0.058) -0.037 0.153 -0.117 (0.043) -0.188 -0.047 Village share of adopters of crop rotation (Kazakhstan) (%)[crshare_adopt_mah] 0.782 (0.396) 0.132 1.433 x x x Farmer perceives land tenure as secure (1/0) [tnr_secur] x x x 0.210 (0.042) 0.141 0.279 Farmer participates in contract farming (1/0) [contrcfrm] -0.033 (0.057) -0.127 0.062 x x x Farmer located in Shardara (1/0) -0.044 (0.086) -0.184 0.097 x x x Farmer located in Pastdargom (1/0) x x x -0.014 (0.043) -0.085 0.056 N 278 307 Pseudo R2 0.153 0.172 Wald χ2 (13)/(12) 47.66 46.10 Log likelihood -126.98 -130.08 Note: Standard error in parenthesis. Source: Based on AGRICHANGE 2019 farm survey data. ~ 119 ~ Table A3: Review of empirical studies on impact of zero-tillage adoption on crop production costs Author (s) Country Data Method Main findings Teklewold et al. (2013) Ethiopia Farm household survey Multinomial ESR Adoption of conservation tillage significantly increased pesticide application and labor demands. El‐Shater et al. (2016) Syria Farm survey of 621 wheat farmers PSM and ESR Negative relationship found between quantity of fertilizer use, quantity of labor use and zero tillage adoption. The major benefits of the adoption of ZT come from tillage and labor cost savings. Jaleta et al. (2016) Ethiopia Survey of 12 peasant associations Probit and ESR models Minimum tillage reduces total labor use, and draft power use (oxendays/ha) for land preparation. Tessema et al. (2018) Ethiopia Households survey 2 step estimation procedure, probit model, OLS Conservation tillage increases herbicide use but reduces female and male labor requirements. Keil et al. (2020) India Panel dataset from 961 farm households ESR Adoption of zero tillage leads to reduction of total variable per-unit production cost Montt and Luu (2020) Eastern and Southern Africa Longitudinal farm data of 3,617 households Multinomial logit and ERS models Conservation tillage increases labor requirements in households. However, minimum tillage saves working time during land preparation and weed control. Yigezu and El‐ Shater (2021) Morocco Survey of 995 households ESR model Zero tillage has no effect on the total amount of agricultural labor use. Erenstein et al. (2008) India & Pakistan Survey of 400 households in India and 458 households in Pakistan Descriptive analysis Resource-saving effects in diesel, tractor time and cost savings for wheat cultivation. Water savings are less pronounced than expected from onfarm trial data. Krishna & Veettil (2014) India Survey of 180 households Production function and semi-parametric technical efficiency estimation methods Wheat productivity increases by 5%, paid-out variable cost reduces by 14%, while the resource use efficiency indirectly increases by 1% ~ 120 ~ Table A4: Descriptive statistics of variables by adopters and non-adopters of zero tillage methods Variables 2016 (N=1363) 2019 (N=1425) Pooled 2016 and 2019 (N=2788) ZT plots (N=93) non-ZT plots (N=1270) mean differ ZT plots (N=290) non-ZT plots (N=1135) mean differ ZT plots (N=383) non-ZT plots (N=2405) mean differ Outcome variables Total payment for hired labor (US$/ha) 5.199 10.584 -5.385 8.891 6.199 2.692 7.994 8.514 -0.520 Machinery costs for land preparation and seeding (US$/ha) 11.272 36.898 -25.626*** 31.281 39.426 -8.146 26.422 38.091 -11.670*** Machinery costs for weeding (US$/ha) 3.714 7.395 -3.681 13.935 9.919 4.017 11.453 8.586 2.867 Herbicide costs (US$/ha) 7.940 7.213 0.727 31.644 19.497 12.147*** 25.750 12.956 12.794*** Total machinery, labor and herbicide costs (US$/ha) 28.124 62.089 -33.955*** 84.192 75.283 8.909 70.250 68.257 1.993 Household head characteristics Age of household head (years) 54.516 55.887 -1.371 56.928 56.001 0.927 56.342 55.941 0.401 Education level of household head (categorical, 1=illiterate…7=university) 4.677 4.308 0.370*** 4.245 4.210 0.035 4.350 4.262 0.088 Female household head (dummy, 1=female) 0.247 0.195 0.052 0.197 0.267 -0.070** 0.209 0.229 -0.020 Household head employment in agriculture (dummy, 1 = occupation in agriculture) 0.559 0.324 0.235*** 0.431 0.313 0.118*** 0.462 0.319 0.143*** Household head’s ethnicity (dummy, 1 = Kyrgyz) 0.893 0.779 0.114*** 0.790 0.766 0.024 0.815 0.772 0.042* ~ 121 ~ Table A4 cont. Variables 2016 (N=1363) 2019 (N=1425) Pooled 2016 and 2019 (N=2788) ZT plots (N=93) non-ZT plots (N=1270) mean differ ZT plots (N=290) non-ZT plots (N=1135) mean differ ZT plots (N=383) non-ZT plots (N=2405) mean differ Household farm characteristics Number of household members that can work in agriculture (above 10 and under 65 years old) 4.032 4.435 -0.403** 4.766 4.520 0.246* 4.587 4.475 0.113 Asset index 0.425 0.398 0.027* 0.356 0.353 0.003 0.373 0.377 -0.004 Household owns a tractor (dummy, 1=yes) 0.032 0.050 -0.018 0.052 0.030 0.022* 0.047 0.041 0.006 Number of livestock units owned by household 4.750 3.262 1.487*** 2.766 2.278 0.488* 3.247 2.798 0.450* Household received remittance last year (dummy, 1=yes) 0.065 0.151 -0.087** 0.217 0.228 -0.011 0.180 0.188 -0.007 Household applied chemical fertilizer last year (dummy, 1=applied) 0.129 0.261 -0.132*** 0.221 0.253 -0.032 0.198 0.257 -0.059** Household experienced a weather shock last year (dummy, 1=yes) 0.849 0.613 0.237*** 0.186 0.154 0.032 0.347 0.396 -0.049* Household experienced an agricultural shock last year (dummy, 1=yes) 0.204 0.376 -0.171*** 0.155 0.070 0.085*** 0.167 0.232 -0.065*** Plot under grains and legumes (dummy, 1=yes) 0.538 0.302 0.235*** 0.338 0.357 -0.019 0.386 0.328 0.058** Plot under vegetables (dummy, 1=yes) 0.043 0.426 -0.383*** 0.245 0.277 -0.032 0.196 0.356 -0.160*** Plot under a mix of crops (grain, legumes and vegetables) (dummy, 1=yes) 0.097 0.071 0.026 0.007 0.026 -0.019* 0.029 0.050 -0.021 ~ 122 ~ Table A4 cont. Variables 2016 (N=1363) 2019 (N=1425) Pooled 2016 and 2019 (N=2788) ZT plots (N=93) non-ZT plots (N=1270) mean differ ZT plots (N=290) non-ZT plots (N=1135) mean differ ZT plots (N=383) non-ZT plots (N=2405) mean differ Location characteristics Distance to main road from dwelling (km) 0.381 0.531 -0.150* 0.945 0.734 0.211*** 0.808 0.627 0.181*** Distance from dwelling to plot (km) 2.085 1.425 0.660** 1.398 1.117 0.281* 1.565 1.280 0.285* Number of land plots owned by household 2.323 1.939 0.383*** 1.776 2.033 -0.257*** 1.909 1.983 -0.075*** Plot size (ha) 1.090 0.665 0.425*** 1.007 0.739 0.268** 1.027 0.700 0.327*** Institutional settings Amount of credit received by household last year (US$) 441.183 193.236 247.9*** 624.287 266.337 357.950*** 579.826 227.735 352.091*** Provinces Issyk Kul 0.419 0.141 0.278*** 0.241 0.163 0.078*** 0.285 0.151 0.133*** Djalal Abad 0.118 0.220 -0.101** 0.155 0.212 -0.057** 0.146 0.216 -0.070*** Naryn 0.161 0.066 0.095*** 0.041 0.049 -0.008 0.070 0.058 0.012 Batken 0.172 0.121 0.051 0.217 0.098 0.119*** 0.206 0.110 0.096*** Osh 0.075 0.274 -0.199*** 0.152 0.329 -0.177*** 0.133 0.300 -0.167*** Talas 0.011 0.084 -0.073** 0.017 0.084 -0.066*** 0.016 0.084 -0.068*** Chuy 0.043 0.094 -0.051 0.176 0.065 0.111*** 0.144 0.080 0.063*** Note: ***, ** and * are significant at 1%, 5% and 10% level, respectively. Source: Based on 2016 and 2019 waves of the LiK data. ~ 123 ~ Table A5: Falsification test for instrumental variable Variables Test p-value Zero tillage adoption (probit model regression) χ2 (1)=7.68 0.006 Total payment for hired labor (US$/ha) (ln) F(1, 2376) = 0.23 0.631 Machinery costs for land preparation and seeding (US$/ha) (ln) F(1, 2376) = 0.53 0.466 Machinery costs for weeding (US$/ha) (ln) F(1, 2376) = 0.00 0.946 Herbicide costs (US$/ha) (ln) F(1, 2356) = 0.00 0.966 Total machinery, labor and herbicide costs (US$/ha) (ln) F(1, 2356) = 0.01 0.934 Note: Here, “distance to the main road” is instrument variable. Source: Based on 2016 and 2019 waves of the LiK data. ~ 130 ~ Table A6 cont. (for total cost) Total machinery, labor and herbicide costs (US$/ha) (ln) ZT plots nZT plots Coeff. [90% confidence interval] Coeff. [90% confidence interval] Age of household head (years) -0.003 -0.036 0.030 -0.009 -0.022 0.003 Education level of household head (categorical, 1=illiterate…7=university) 0.025 -0.110 0.161 -0.028 -0.086 0.031 Female household head (dummy, 1=female) -0.388 -0.864 0.089 0.169 -0.007 0.345 Household head employed in agriculture (dummy, 1 = occupation as agriculture) 1.066 0.201 1.931 0.442 0.127 0.756 Number of household members that can work in agriculture (above 10 and under 65 years old) -0.020 -0.311 0.271 0.044 -0.046 0.135 Household head’s ethnicity (dummy, 1 = Kyrgyz) -0.026 -0.617 0.565 0.174 -0.012 0.359 Assets index -0.471 -3.284 2.342 0.408 -0.512 1.328 Household owns a tractor (dummy, 1=yes) 1.059 -0.413 2.531 0.361 -0.262 0.985 Plot size, (ha) -0.013 -0.095 0.070 -0.020 -0.057 0.018 Household received remittances last year (dummy, 1=yes) -0.539 -1.375 0.298 0.115 -0.155 0.384 Household experienced a weather shock last year (dummy, 1=yes) 0.734 -0.009 1.478 0.551 0.300 0.803 Household experienced an agricultural shock last year (dummy, 1=yes) 0.239 -0.546 1.024 -0.155 -0.430 0.121 Distance from dwelling to plot (km) 0.066 -0.009 0.140 0.119 0.089 0.149 Household applied chemical fertilizers last year (dummy, 1=applied) 1.059 0.202 1.916 1.233 1.006 1.459 Number of land plots owned by household -0.363 -0.733 0.008 -0.005 -0.126 0.117 Number of livestock units owned by households -0.051 -0.146 0.044 0.004 -0.026 0.034 Amount of credits received by household last year (logarithm, US$) 0.071 -0.041 0.184 -0.021 -0.062 0.020 Plot under grains and legumes (dummy, 1=yes) 1.250 0.780 1.719 1.565 1.366 1.764 Plot under vegetables (dummy, 1=yes) 0.511 -0.349 1.371 0.346 0.131 0.561 Plot under a mix of crops (grain, legumes and vegetables) (dummy, 1=yes) 1.301 0.274 2.328 1.489 1.166 1.812 ~ 131 ~ Table A6 cont. (for total cost) Total machinery, labor and herbicide costs (US$/ha) (ln) ZT plots nZT plots Coeff. [90% confidence interval] Coeff. [90% confidence interval] Djalal-Abad -1.850 -2.889 -0.810 -1.451 -1.779 -1.122 Naryn -1.249 -2.159 -0.338 -1.181 -1.536 -0.825 Batken 0.624 -0.094 1.342 -0.974 -1.271 -0.677 Osh 0.236 -1.228 1.699 -0.740 -1.123 -0.358 Talas -2.240 -5.034 0.553 -0.975 -1.476 -0.473 Chuy -0.270 -1.290 0.750 -0.776 -1.119 -0.433 Survey year (2019=1) 1.611 -0.604 3.825 -0.022 -0.310 0.265 Mean of age of household head 0.030 -0.008 0.069 0.007 -0.007 0.021 Mean of household head employed in agriculture -0.562 -1.453 0.329 0.075 -0.280 0.430 Mean number of household members that can work in agriculture 0.188 -0.132 0.507 -0.018 -0.117 0.081 Mean of assets index 1.985 -1.849 5.818 -0.504 -1.705 0.697 Mean of household owns a tractor -2.306 -4.185 -0.426 -0.090 -0.922 0.742 Mean of household received remittances last year -0.302 -1.385 0.781 -0.199 -0.550 0.152 Mean of household experienced a weather shock last year -0.422 -1.459 0.614 -0.412 -0.756 -0.067 Mean of household experienced an agricultural shock last year -0.712 -1.614 0.191 0.353 -0.033 0.740 Mean of amount of credit received by household last year (ln) 0.028 -0.118 0.175 0.088 0.034 0.141 Mean of household applied chemical fertilizers last year 0.972 -0.332 2.276 0.657 0.283 1.032 Mean of number of livestock units owned by household 0.092 -0.017 0.201 0.019 -0.017 0.054 year (2019)*mills1 0.424 -0.745 1.593 x x x mills1 1.887 -0.380 4.153 x x x year (2019)*mills2 x x x -3.460 -4.762 -2.158 mills2 x x x 4.490 2.428 6.553 _cons -5.434 -10.869 0.002 1.883 1.222 2.544 R2 0.351 0.293 N 373 2365 Source: Based on 2016 and 2019 waves of the LiK data. ~ 132 ~ Table A7: Propensity score matching (PSM) estimation (ATT) results (impact of zero tillage adoption) Outcome variable ATT (average treatment effect on the treated) Coefficient Standard error [90% confidence Interval] Total payment for hired labor (USD $/ha) (ln) 0.170 0.071 0.053 0.287 Machinery costs for land preparation and seeding (USD $/ha) (ln) -0.292 0.159 -0.554 -0.031 Machinery costs for weeding (USD $/ha) (ln) 0.120 0.110 -0.062 0.301 Herbicide cost (USD $/ha) (ln) 0.128 0.126 -0.080 0.335 Total cost (USD $/ha) (ln) -0.311 0.167 -0.587 -0.036 Source: Based on 2016 and 2019 waves of the LiK data. Table A8: Falsification test to check validity of instrument First-Stage Second-Stage Instrument Participation in informal cooperation Intensity of Sustainable Agricultural Practices adoption of Non-participant Farmers Distance to the local market from farm field (km) - 0.012 (0.002) χ2 = 22.23 p-value = 0.000 -0.004 (0.003) F-Stat.=1.46 p-value = 0.228 Sample Size 858 351 Note: Standard errors are reported in parentheses. Source: Based on AGRICHANGE and SUSADICA farm survey data. ~ 133 ~ CURRICULUM VITAE ABDUSAME TADJIEV Doctoral researcher Department of External Environment for Agriculture and Policy Analysis Leibniz Institute of Agricultural Development in Transition Economies (IAMO) Theodor-Lieser-Str. 2, 06120 Halle (Saale), Germany Tel: +49345-2928120; Email: tadjie[email protected]e Nationality: Uzbekistan Date of birth: 01/09/1983 Present Position January 2019 – present Doctoral researcher at the Department of External Environment for Agriculture and Policy Analysis, Leibniz Institute of Agricultural Development in Transition Economies (IAMO). Dissertation title: Exploring the advantages and drivers of sustainable agricultural practices in Central Asia Professional experience September 2008 – December 2018: Assistant Professor and Researcher at Samarkand Institute of Veterinary Medicine (former Samarkand Agricultural University), Uzbekistan. Responsibilities included undergraduate classes and seminars, agricultural economics research, supervision of bachelor and master students. Academic Qualifications January 2019 - September 2024: Ph.D. in Agricultural economics, at Leibniz Institute of Agricultural Development in Transition Economies (IAMO), Halle, Germany. In the framework “SUSADICA” project, I successfully defended my second doctoral thesis entitled “Exploring the advantages and drivers of sustainable agricultural practices in Central Asia” at the faculty of “Natural Sciences III” of the Martin Luther University Halle-Wittenberg. 23rd September, 2024. January 2019 - September 2020: Ph.D. in Agricultural economics (Candidate of Sciences), at the Tashkent Institute of Irrigation and Agricultural Mechanization Engineers (TIIAME) in Uzbekistan. I successfully defended my first doctoral thesis entitled “Evaluation of land and water reforms and cooperation among farmers: The case of Samarkand province” September 2006 – July 2008: MSc. in Sectoral Economics, Samarkand Institute of Economics and Service, Uzbekistan September 2002 – July 2006: BSc. in Agricultural Economics, Samarkand Agricultural University, Uzbekistan Participation in Research projects Since 2023 – BMBF-funded project led by IAMO “UzFarmBarometer Better understanding of the adoption of sustainable agricultural practices in Uzbekistan” 2019-2023 – VolkswagenStiftung-funded project led by IAMO “Structured doctoral programme on Sustainable Agricultural Development in Central Asia (SUSADICA)” ~ 134 ~ 2015-2019 – VolkswagenStiftung-funded project led by IAMO “Institutional change in land and labour relations of Central Asia’s irrigated agriculture (AGRICHANGE)” Relevant Academic Trainings 25/06/2023 – 30/06/2023 – “Panel Data Analysis”, Barcelona School of Economics, Spain 24/10/2018 – 28/10/2018 – “Capacity Building of Young Researchers from Central Asia and Afghanistan in Water for Policy Studies”, Dushanbe office of OECE, Tajikistan 03/06/2018 – 14/06/2018 – “Regional training course on Applied Econometric analysis”, Westminster International University in Tashkent and IFPRI, Tashkent, Uzbekistan 19/02/2017 – 23/03/2017 – “Training course on Agricultural economics”, Leibniz Institute for Agricultural Development in Transition Economies (IAMO), Halle (Saale), Germany Selected publications Tadjiev, A., Djanibekov, N., Herzfeld, T. (2023) Does zero tillage save or increase production costs? Evidence from smallholders in Kyrgyzstan. International Journal of Agricultural Sustainability 21 (1) Kurbanov, Z., Tadjiev, A., Djanibekov, N. (2022) Adoption of sustainable agricultural practices and investments in productive assets in irrigated areas of Central Asia: Farm-survey evidence from Kazakhstan and Uzbekistan. IAMO Annual 24, pp. 69-79. Tadjiev, A., Kurbanov, Z., Djanibekov, N., Govind, A., Akramkhanov, A. (2023) Determinants and impact of farmers' participation in social media groups: Evidence from irrigated areas of Kazakhstan and Uzbekistan. IAMO Discussion Paper No. 201, Halle (Saale), IAMO. Tadjiev, A., Djanibekov, N., Soviadan, M. K., & Herzfeld, T. (2025) Participation in informal cooperation in water management and adoption of sustainable agricultural practices: Empirical evidence from Uzbekistan. Australian Journal of Agricultural and Resource Economics Networking Activities and Outreach Science and policy briefs Djanibekov, N., Kurbanov, Z., Tadjiev, A., Govind, A., Akramkhanov, A. (2023) Farmers' social media groups for better extension and advisory services. IAMO Policy Brief No. 46, Halle (Saale). Tadjiev, A. (2023) Impacts of crop rotation on the performance of cotton growing farmers in Central Asia. Science Brief 4, SUSADICA project. Organization of conference session 2021: "Understanding Farmers’ Decisions Towards Sustainable Agriculture in Irrigated Areas of Central Asia" at 31th Virtual International Conference of Agricultural Economists (ICAE). Selected conference presentations Tadjiev, A. (2023) Does adoption of zero tillage save or intensify production costs? Evidence from Kyrgyzstan. XVII Congress of EAAE “Agri-food systems in a changing world”, Rennes, France. ~ 135 ~ Tadjiev, A. (2023) Participation in informal cooperation and adoption of sustainable agricultural practices. SUSADICA final symposium, Tashkent, Uzbekistan. Tadjiev, A., Djanibekov, N., Sanaev, G. (2021) Adoption of sustainable agricultural practices in irrigated areas of Central Asia. XVI Online Congress of EAAE. Tadjiev, A. (2021) Impact of sustainable agricultural practices adoption on output of cotton producing farms. 31st Virtual ICAE 2021, Neu Delhi /Online, India. Tadjiev, A., Djanibekov, N., Sanaev, G. (2020) Determinants of sustainable agricultural practices in Central Asia. 6th Annual ‘Life in Kyrgyzstan’ Conference, Online, Germany. Teaching Ad hoc lectures on An econometric analysis in Stata: Impact evaluation of sustainable agricultural practices, 13–14 March 2023, IAMO, Halle (Saale), Germany. Ad hoc lectures on “Econometric analysis of farmers’ adoption decisions of sustainable agricultural practices” Part II, 7–11 April 2025, Westminster International University in Tashkent (WIUT), Uzbekistan. Ad hoc lectures on “Econometric analysis of farmers’ adoption decisions of sustainable agricultural practices”, 7–11 October 2024, Westminster International University in Tashkent (WIUT), Uzbekistan. Professional Activities and Memberships International Association of Agricultural Economists (IAAE), European Association of Agricultural Economists (EAAE) Halle (Saale), 2025 ~ 136 ~ Eidesstattliche Erklärung / Declaration under Oath Ich erkläre an Eides statt, dass ich die Arbeit selbstständig und ohne fremde Hilfe verfasst, keine anderen als die von mir angegebenen Quellen und Hilfsmittel benutzt und die den benutzten Werken wörtlich oder inhaltlich entnommenen Stellen als solche kenntlich gemacht habe. I declare under penalty of perjury that this thesis is my own work entirely and has been written without any help from other people. I used only the sources mentioned and included all the citations correctly both in word or content. ______________________ ________________________________________ Datum / Date Unterschrift des Antragstellers / Signature of the applicant