1 2 Do behavioral nudges endure? Effect of reminders on sustained demand for quality seed of sweet potato Julius J Okello, David R Just, Arjan Verschoor, Mingcong Xie, Chalmers Mulwa, Sylvester Ojwang, Janet Mwende Mutiso, Sam Namanda, Reuben T Ssali, Bernard Yada, Srinivasulu Rajendran, Hugo Campos November 2024 3 Do behavioral nudges endure? Effect of reminders on sustained demand for quality seed of sweet potato. © International Potato Center 2024 DOI: 10.4160/cip.2024.11.001 CIP publications contribute important development information to the public arena. Readers are encouraged to quote or reproduce material from them in their own publications. As copyright holder CIP requests acknowledgement and a copy of the publication where the citation or material appears. Please send a copy to the Communications Department at the address below. International Potato Center P.O. Box 1558, Lima 12, Peru cip@cgiar.org • www.cipotato.org Citation: Okello, J.; Just, D.; Verschoor, A.; Xie, M.; Mulwa, C.; Ojwang, S.; Mwende, J.; Namanda, S.; Ssali, R.; Yada, B.; Rajendran, S.; Campos, H. 2024. Do behavioral nudges endure? Effect of reminders on sustained demand for quality seed of sweet potato. International Potato Center. 53 p. DOI: 10.4160/cip.2024.11.001 Design and Layout: Communications Department November 2024 CIP also thanks all donors and organizations that globally support its work through their contributions to the CGIAR Trust Fund: www.cgiar.org/funders © 2024. This publication is copyrighted by the International Potato Center (CIP). It is licensed for use under the Creative Commons Attribution 4.0 International License http://www.cipotato.org/ http://www.cgiar.org/funders 4 Contents 1 Abstract ...................................................................................................................................................... 5 2 Introduction................................................................................................................................................ 6 3 Context ....................................................................................................................................................... 7 4 Study methods ......................................................................................................................................... 10 4.1 Ethical approval ...................................................................................................................................... 10 4.2 Study site and seed source ..................................................................................................................... 10 4.3 Experimental treatments ........................................................................................................................ 10 4.4 Quality seed ............................................................................................................................................ 11 5 Empirical strategy ..................................................................................................................................... 13 6 Results ...................................................................................................................................................... 15 7 Discussion and conclusions ....................................................................................................................... 27 8 Acknowledgments .................................................................................................................................... 29 9 References ................................................................................................................................................ 30 10 Appendix .................................................................................................................................................. 33 5 1 Abstract Delivering quality seed of to farmers is critical to the development of smallholder crop sector in developing countries. Smallholder farmers continue to use poor quality recycled seed obtained from own or social network. Behavioral nudges are currently being considered as a low-cost approach of stimulating demand for agricultural innovations including quality seed. However, past studies continue to cast doubt on the longevity of nudges. That is, there is doubt on whether nudges can be used to stimulate sustained demand for agricultural innovations. In this study, we examine whether nudges may be used to influence sustained adoption of quality seed of improved sweetpotato varieties (QS-ISV) in an informal seed system. The seed used in the study were inspected and certified as disease and pest-free. We focus on the use of text message reminders designed to increase the repurchase of QS-ISV introduced into informal farmer network-based seed systems. The study site is a sweetpotato growing district in Uganda where yield is severely depressed due to sweetpotato virus diseases and pests, especially sweetpotato weevils. The study used a randomized controlled trial in which quality seed was displayed for sale in all study villages with treatment group receiving text message reminders, while the control group only getting exposure to QS-ISV. Seed was presented at salespoints for three consecutive seasons with reminders being sent for two consecutive seasons focusing on seed purchasers in the first two seasons. We find that reminders initially increased the likelihood of seed repurchase and subsequently, when reminders are repeated, reduce this likelihood. This suggests that nudges may be used to encourage the swifter integration of formal and informal elements in seed systems, but that their overuse may be counterproductive. 6 2 Introduction The use of improved agricultural inputs (innovations) including quality seed is fundamental to agricultural transformation through productivity enhancement. While there have been significant investments in formalizing the seed systems to enhance access to quality seed by producers, a large proportion of such farmers in developing countries still rely on informal seed systems for seed acquisition (McGuire and Sperling, 2016). This is especially true for vegetatively propagated crops (VPCs) such as sweetpotato, cassava, bananas, yam, and potato, often referred to as root, tubers and bananas (RTB) crops. The nature of the seed material for these crops (i.e., stem, vines, and suckers) is conducive for farmer-to-farmer exchange of planting material and for farmers to recycle seed, resulting in little incentive for the participation by private seed entrepreneurs. Consequently, seed systems for RTB crops remain largely informal, with cultural norms and social relations playing a big role in governing seed acquisition (Hodgkin et al, 2007; Scholey and Padmanabhan, 2016). Varieties disseminated using such informal channels are usually uninspected and uncertified, with some not even being formally registered and released. Under this system, farmers plant the same material repeatedly over many seasons, and often years. This results in a build-up of pests and diseases, and rapid seed degeneration in seed quality and the associated reduction in crop growth vigor and yield (Ogero et al., 2023; McGuire and Sperling, 2016). While new improved varieties of the RTB crops have been developed that are resistant to pests and diseases, dissemination of these varieties within existing informal seed systems is a challenge due to the aforementioned nature of seed. This contributes to the observed low variety turnover (Gatto et al, 2021). It is, for example, estimated that adoption among RTB crops reaching ceiling at 40% in sub-Saharan Africa (McEwan et al., 2021). Moreover, the injection of new seed varieties in these settings has typically been limited to pathways such as the provision of emergency seed aid (Sperling and McGuire, 2010; Namanda et al, 2011; Sperling et al, 2020). As a result of the informal nature of RTB seed systems and related challenges of introducing new varieties, landraces continue to dominate the cropping systems of most RTB crops (Zawedde et al, 2014). Low genetic potential of these varieties, especially around agronomic traits such as yield gain and resistance to biotic and abiotic stressors, constrain productivity and limit their potential to contribute to the key development goal of combating food insecurity (Okello et al, 2023; Bayiyana et al., 2024). Moreover, consumers of these crops often favor landraces due to idiosyncratic quality and culinary characteristics that are superior or preferred to those of improved varieties (Mulwa et al., 2023; Gatto et al, 2021; Thiele et al, 2021). Beyond the informal seed system structure for RTB crops, and quality-related preferences for landraces, other demand and supply factors contribute to low diffusion of improved varieties and seed recycling among RTB crops. On the demand side, unfamiliarity with new technology and the associated perceived risk of its poor performance form a strong disincentive against the adoption of new varieties, and has been shown to reduce the adoption of improved RTB crop varieties (Jogo et al, 2021; Almekinders et al, 2019). High prices of quality seed, whether real or perceived, also contribute to seed recycling among the farmers (Bayiyana et al, 2024). In addition, lack of authentic signals for quality of VPC seeds limits differentiability of certified and uncertified seed, thereby reducing trust. Seed is a credence good whose quality cannot be directly observed (McEwan et al, 2021; Spielman et al, 2021), and signals serve as a strong incentive to differentiate between seed bundles. For instance, seed for cereal crops is easily packaged and labelled to signal quality, which is challenging to achieve, for instance, with sweetpotato vines and cassava cuttings. A rational farmer may therefore be unwilling to spend more to purchase seed that is not clearly differentiated as being of high quality, compared to other cheaper options. Development literature highlight high incidence of market failure for such seed (Loch and Boyce, 2003; McEwan et al, 2021). 7 On the supply side, key impediments to farmer replacement of seed of VPCs include unavailability and the high transaction and transportation costs of accessing quality seed (Gibson et al, 2011; McEwan et al, 2021). Community seed multipliers, which are often the only source of quality seed, tend to be few and geographically dispersed, hence tend to be located too far from the majority of the farmers. In additional seed of VPCs tend to be characterized by high bulkiness and perishability. Both characteristics increase the transaction costs of acquiring quality seed. Moreover, high perishability results make such seed subject to high temporal asset- specificity which exacerbate the transaction costs (Okello and Swinton, 2007) Development literature has long identified innovation/technology uptake as a social process (Rogers, 1962). Rogers (1995), for instance, argued that the diffusion of agricultural innovations is influenced by the cognitive factors relating to knowledge and information availability. Leeuwis and Aarts (2021) argue that adoption of agricultural innovations is a cognitive process that is driven by a host of interdependent sociological processes. Okello et al (2019) argue that farmers decision to adopt agricultural innovations is motivated by psychosocial and mental/emotional factors. Building on this, there is emerging interest in the role behavioral nudges can play in influencing farmer decision making process relating to technology use. Nudges are low-cost approaches that are designed to alter decision-making process in a predictable way without forbidding any options or changing their economic incentives (Thaler and Sustein, 2008 pg 6; Chai et al, 2023). They work by altering the choice environment and have been applied extensively in nutrition interventions in developed countries. However, application of nudges in agriculture is relatively new but growing. Recent applications include Balew et al (2023), Okello et al (2023), Rola-Rubzen et al (2023). In this paper, we report the results of an intervention that combined elements of formal and informal seed systems to assess the effect of behavioral nudges on sustainable demand for quality seed of VPCs. Specifically, the paper examines the impact of mobile phone-based text message reminders on purchase and repurchase of quality sweetpotato seed (vines), using a large field experiment in Uganda. Uganda is an interesting case to study because it is the secondary center of diversity of sweetpotato varieties globally. Zawedde et al (2014) found more than 1,300 varieties of sweetpotato in Uganda. The bulk of these varieties are landraces which, as earlier noted, are particularly popular among farmers (Mwanga et al, 2021). It is estimated that farmers maintain, on average, four varieties in their gardens, the majority of them landraces (Okello et al, 2022). 3 Context This study was implemented jointly Production Department of the Ministry of Agriculture Animal Industry and Fisheries (MAAIF), that is, the local government of Uganda. It therefore also serves to provide baseline information to the district on the implementation of the Parish Development Model (PDM). The PDM is a grassroots bottom-up development strategy implemented by the Uganda government and was launched the same year this study commenced. Information generated on the behavior of sweetpotato farmers towards quality seed of improved crop varieties therefore serve as baseline knowledge and information regarding how farmers respond to efforts to promote use of improved seed varieties and should contribute to the design of future extension programming efforts. More specifically, the findings of this study contribute to Pillar 1 of the PDM that focuses on production, storage, processing and marketing of agricultural commodities, in our case focusing on sweetpotato. In Uganda, sweetpotato is grown both as a commercial and as a subsistence/food security crop, with many households having a strong cultural affiliation with the crop because it forms a key part of local food systems (Echodu et al, 2019). In the Teso sub-region and Eastern Uganda, where the study was conducted, sweetpotato 8 is grown in the first season mainly for fresh root consumption and sale (Bayiyana et al, 2024), and in the second season for making dried sweetpotato chips and flakes, called amukeke1 and inginyo2, respectively (Echodu et al, 2019). Amukeke is used for making popular cultural food by the same name. Two types of sweetpotato seed systems exist concurrently in Uganda. The first is based on exchange within a social network while the second is based on commercial/market purchases (Rachkara et al, 2017; Sperling et al, 2023). These are however both informal, as seed flowing through these seed systems are usually uninspected, uncertified and are often composed of unregistered and unreleased varieties. The dominant seed system is social network-based system. This is not just for case sweetpotato, but also for other vegetatively propagated crops. The system thrives on farmer-to-farmer seed exchange. It serves family, friends, neighbors and other farmers within the community. Seed, in this system, is largely exchanged for free, or borrowed, with reciprocation culturally expected. Quality is determined through observation of the crop while in the field during the growth stage and/or visual observation of the vines (McEwan et al, 2021). Seed quality assessment is based on farmers’ indigenous knowledge of disease and pest manifestations on growing plants and/or leaves and stems. Farmers collect/harvest the seed directly from neighbors’ gardens, hence select what they perceive to be satisfactory in quality. The level of keenness put in selecting the best in terms of quality depends on availability, with quality largely disregarded the scarcer (less available) the seed is. The informal commercial/market sweetpotato seed systems are nascent and small, and strongest in the northern region of Uganda where drought desiccates all the sweetpotato planting material (Sperling and Almekinders, 2023). This system serves both local and distant farmers. Seed is produced off-season on wetlands and either sold to neighbors or traded in local roadside markets. A limited amount is traded across the border into Southern Sudan. Non-governmental organizations (NGOs) that supply seed aid play a major role in this system and deliver seed to distant communities. Non-NGO distant trade is facilitated by seed transportation using public bus and minibus services (Rachkara et al, 2017). The seed traded is of unknown quality. Seed quality assurance is “informal” based on trust and/or maintaining good reputation (Sperling and Almekinders, 2023). An important component of market/commercial sweetpotato seed systems revolves around community seed production by decentralized vine multipliers (DVMs). The decentralized seed multiplication is mostly piloted by research and development projects as a transitional formal model to bridge access to disease-free planting material of improved varieties. This system gained prominence with the introduction of biofortified crops and is mainly project-based (Gibson, 2013). It combines elements of informal and formal systems. The formal components encompass several aspects: training of seed multipliers on seed production and quality maintenance mechanisms. Starter material is sourced from credible certified private sector sources and/or from a public laboratory run by the National Agricultural Research Organization (NARO). Multipliers’ operation and seed plots are routinely inspected, tested the dominant diseases and certificates of quality assurance issued (Gibson, 2013). Alternatively, seed can be visually inspected usually by agricultural staff and declared to be clean relative to seed from local seed network. Seed from this second source is therefore referred to as quality declared seed (QDS), because it is inspected according to set guideline to minimize, but not eliminate, pest and disease infestation (Mukasa et al, 2016). Due to high cost of seed testing and certification, DVMs mainly produce QDS. It is sold predominantly to NGOs at high and sometimes subsidized prices. Some DVMs sell, usually leftovers, to local 1 To make amukeke, farmers peel sweetpotato roots and chip into large chunks which are then sun-dried. The process therefore needs strong sun over several days, conditions which are more prevalent after the second season harvest. 2 See below for a description of inginyo. 9 farmers/neighbors at lower prices than offered by NGOs. With the exception of a few cases, only registered and released varieties pass through this system. Sweetpotato “seeds” found in most communities will often have circulated in the community, or been recycled by farmers, for many years. Such seed accumulate high loads of diseases and pests that affect their yield. The most devasting of these pests and diseases are sweetpotato virus diseases (SPVD) and the sweetpotato weevil, respectively. The SPVD-infected seeds produce thin elongated unmarketable roots used to make non-traded dry flakes known as inginyo (Okello et al, 2023). Weevil infected seeds produce roots with dark spots and holes in the flesh, making the roots inedible. The use of poor-quality seed therefore has a hefty yield penalty (Low et al, 2020). These diseases and pests are blamed for the low average yield of 4 tons/ha for sweetpotato in smallholder farms in Uganda is as compared to more than 15 tons/ha obtained from the use of quality seed under the same farming conditions (Namanda et al., 2019). The penalty therefore is in terms of depressed production, which can also be a matter of shape, size or number of roots produced. Yield penalties of up to 86% have been reported for smallholder farmers who use poor quality sweetpotato seed (Van Vught and Franke, 2018). Mugisa et al (2023) estimate that sweetpotato weevil damage to roots can cause losses ranging from 60% to 100% Given the poor quality of seed circulating in the communities, a major seed system development question is how to best place quality seed in the hands of smallholder farmers. Domiguez (2023) identifies three broad seed delivery pathways for the case of vegetatively propagated seed. They are formal, informal and intermediate. The first (formal) pathway encompasses seed parastatals, private seed delivery, seed aid programs, project-based seed delivery, public-private seed programs, and multi-sector seed partnerships. The informal seed delivery pathways in turn are farmer to farmer dissemination also known as social seed network, local traders and informal seed multipliers. Between these two polar ends of the seed delivery pathway is sandwiched the intermediate pathway consisting of seed producer cooperatives/associations, local seed enterprises, decentralized seed multipliers, and village-based advisors. Dominguez et al (2023) discuss these pathways in detail. In this study, the approach taken combined elements of the three pathways. It utilized the seed quality assurance system typically used in the formal seed system by way of inspection and certification of seed sources. Two sources were the focus of the study namely, the early generation seed production in government-inspected and certified private sector operation. The certified seed from this source was a Class 1 (C1) seed. The second was the seed certification of community based decentralized seed multipliers that are registered under seed producer association/cooperative, thus falling under the intermediate pathway. Lastly, seed was delivered to locations in villages making it easily accessible to farmers thus reducing the transport and transaction costs. This approach of delivering seed to communities is commonly used by development agencies/NGOs as part of the seed aid. Unlike the typical NGO approach in which seed is provided to farmers at no cost, the study provided seed at a subsidized cost. The approach further used model farmers as seed sellers thus mimicking the informal seed trader pathway. 10 4 Study methods 4.1 Ethical approval This study was conducted in accordance with the ethical research guidelines laid down in the Declaration of Helsinki. It was implemented jointly by the International Potato Center, Cornell University, the Norwich Institute for Sustainable Development, and the Ugandan Ministry of Agriculture, Animal Industry and Fisheries (MAAIF) under the Cornell University Institutional Review Board (IRB) ethics approval ID# 2110010648. 4.2 Study site and seed source This study was implemented in Amuria district which in eastern region of Uganda. The district borders the northern region and has trade links with markets for sweetpotato seed in the northern districts. Farmers in the district use both social network and market sources of seed. However, purchase of seed is very limited, mainly because of availability and cost. In fact, most of the seed sold in the local market is uninspected and uncertified. At the time of the study, the nearest source of quality seed produced by trained decentralized seed multipliers was 70km away, with no direct public bus service, hence greatly restricting access to quality seed. Trained seed multipliers that were actively producing quality seed at the time of the study were based in Serere district (70km), Bukedea district (112km) and Kamuli district (223km). Quality seed could also be obtained from a trained seed multiplier based in Gulu district (258km). Notably, these sources are quite far and out of reach of smallholder farmers who typically deal in small amounts of seed making per unit cost of transporting seed from distant sources quite high. Further, high perishability of sweetpotato seed makes it difficult to maintain the integrity of the seed for extended period, with about 5 days as the most it can take between harvest and planting. This, as discussed earlier, increases the transaction costs of delivering seed to smallholder farmers. The high combined (transport and transaction) cost of quality seed makes them unfordable to smallholder farmers. Indeed, Bayiyana et al (2024) find that smallholder farmers grow local varieties that are readily available within the community, that is, they use seed found within their social networks. Their study further found that a liquidity constraint – that is, lack of money – is a major factor contributing to the low use of market sources of seed. 4.3 Experimental treatments The interventions implemented in this study covered the whole of Amuria district, comprising two counties, fifteen sub-counties and 91 parishes/wards. The district had never had a sweetpotato crop improvement program/project prior to this study. The interventions were implemented in 120 villages randomly selected from the 91 parishes/wards proportional to size of each parish/ward. The 120 villages were randomized into treatment (n=64) and control (n=56) groups. The imbalance was due to care being taken not to assign, of any two villages that are less than 5km apart, one to the treatment group and the other to the control group, to avoid contamination (McCann et al, 2018). The spatial distribution of the treatment and control villages is shown in Figure 1. Lastly, in each of the 120 villages, 10 households were randomly selected from a list of households that had grown sweetpotato in the year preceding the study, and 11 their purchase behavior tracked over time. Thus, a total of 1,200 households, which prior power calculations deemed sufficient to detect purchase behavior due to treatment, participated in the study.3 Figure 1. Layout of treatment and control villages in the district. To understand how behavioral nudges influence sustainable demand for quality seed, this study deployed a number of interventions that embody elements both of formal and of informal sweetpotato seed systems. The formal system aspects relate to use of registered/released varieties, seed plot inspection, seed testing, certification and maintenance of identity of the variety as it flows through the system from the lab to the farmer’s field. In the absence of the interventions, sweetpotato seed flow to and among farmers is informal, and the material is usually of unknown quality. Thus, the interventions deployed attempted to formalize this informal system and then examine how behavioral nudges can sustainably promote demand for quality seed. The nudge deployed was in the form of text message reminders sent to past seed purchasers one week before the delivery of quality seed to salespoints. The reminders were in the two vernacular languages used in the study area. One message was sent at a time each day of the week for 7 days, then repeated during the second week, during which the seed was on display at the salespoints. They reminded past purchasers that the seed was available at the salespoint, and of seed quality. The reminder text messages were thus designed to nudge repurchase of quality seed. Reminders were deployed from the second growing season to the end of the interventions. 4.4 Quality seed We used quality seed of four sweetpotato varieties. The varieties were chosen jointly by breeders and agronomists to represent the most popular varieties in the country, and that are known to perform well under conditions like those in the study district. The selected varieties included two released landraces (Ejumula and Tanzania) and two newly bred and released varieties (Joweria/NASPOT 13 and New Dimbuka/NAROSPOT 1). 3 Power calculation was based on a 9% adoption rate – implying a standard deviation of approximately 0.29, a within-village correlation coefficient of 0.10 and a 20-25% desired increase in adoption. With these assumptions, an 𝑛 = 1000 (100 villages, 10 households per village) is sufficient to detect an increase in adoption of 20% with 97% probability (and 15% with 80% probability). Twenty extra villages were added to allow for potential attrition, yielding 120 villages. 12 Figure 2 summarizes the characteristics of the four varieties used. Ejumula and Tanzania are both widely grown varieties in the study district. However, due to repeated planting, locally available seed of these varieties is infested with SPVD and sweetpotato weevil, which greatly reduces their yield, therefore reducing (in the second season) the volume of roots available for making amukeke. At the time of the study, farmers were not growing Ejumula and Tanzania as the most preferred variety because of poor yield performance and lack of access to quality seed (Bayiyana et al, 2024). Quality seed of the four varieties were produced from a starter material sourced from a certified/accredited early generation seed producer. This seed producer is a private sector firm specialized in production of Class 1 seed that is usually bulked before sale to farmers. The seed was bulked by trained and certified seed multipliers into Class 2 seed, under strict monitoring. The multipliers had been inspected by a government seed inspector and their operations certified. One week prior to harvest, the seed plot of each variety was inspected for viruses and sweetpotato weevil, and samples tested for virus load by scientists at the International Potato Center. Only plots that tested negative for viruses were harvested for use in the study. In addition to the pathogen testing, fields and samples were also closely observed for symptoms of major pests of sweetpotato, especially the sweetpotato weevil. Figure 2. Varieties of sweetpotato used in the study To relax the seed access constraint, seed distribution outlets (also referred to as salespoints) were constructed in each study village (Figure 3). The salespoints were in easily accessible locations and were hosted by salespersons recruited from among model/progressive farmers in each village with the help of government extension officers. The criteria for recruitment included good reputation, easy to approach, living within the village, and literacy (see Appendix 3). To avoid contamination, salespoints were not placed in or near locations where they can be easily seen by non-villagers, such as major roads, health centers, religious centers (churches/mosques), and schools. From the first rain season of 2022, and for the next four season, 12 bags of seed each with 30-centimeter-long cuttings, were delivered to the salespoints for sale to co-villagers by the salesperson. Figure 3 shows two salespersons receiving seed for sale. Due to high perishability, seed were delivered in two rounds separated by one week, to reduce losses. Salespersons captured data on transactions made in each village and season. Data on sales were captured in a booklet with a form for individual purchasers. Both the book with sales data and the money from seed sales were collected at the end of the season. To reward the salespersons for their effort, each was given a pair of 13 rubber boots and new gardening hoe after the first season and one full bag of quality seed for free in subsequent seasons. The hoe and boots were valued at UGX 50,000 while the bag of quality seed was valued at UGX 10,000. Figure 3. Village delivery and display for sale 5 Empirical strategy We examine the impact of text message reminders on farmers in the treatment group who previously purchased vines, comprising 194 participants in 2022 and 235 in 2023. A linear probability model (LPM) is estimated as follows: 𝑌𝑐𝑠ℎ = 𝛽0 + 𝛽1𝑇𝑒𝑥𝑡𝑐𝑠ℎ + 𝛾′𝑋𝑐𝑠ℎ + 𝑣𝑐 + 𝜀𝑐𝑠ℎ (1) In this model, 𝑌𝑐𝑠ℎ represents the purchase decision of household ℎ in village 𝑠 , subcounty 𝑐 during the second season of each year, while 𝑇𝑒𝑥𝑡𝑐𝑠ℎ is a binary variable indicating whether text message reminders were sent to the household. The term 𝑣𝑐 denotes a vector of subcounty fixed effects, which are included to account for unobservable characteristics at this very level. In certain specifications, the model includes 𝑋𝑐𝑠ℎ, a vector of household-level controls, such as past vine purchase frequency, gender, age, educational level, household size, and the acreage of land available for cultivation. The error term is given by 𝜀𝑐𝑠ℎ. Our primary interest, 𝛽1, captures the effect of the text message intervention. Results are presented with both robust standard errors and standard errors clustered at the village level, with symmetric tests conducted according to standard practice. To account for potential selection bias in the data, a two-stage Heckman selection model is employed. In the first stage, a probit regression is estimated to model the probability that a household provides a phone number for future contact. The selection equation is specified as: 𝑃(𝑃ℎ𝑜𝑛𝑒𝑁𝑢𝑚𝑏𝑒𝑟𝐿𝑖𝑠𝑡𝑒𝑑𝑐𝑠ℎ = 1 | 𝑋𝑐𝑠ℎ ′) = Φ(𝛼0 + α𝑋𝑐𝑠ℎ ′) (2) where 𝑃ℎ𝑜𝑛𝑒𝑁𝑢𝑚𝑏𝑒𝑟𝐿𝑖𝑠𝑡𝑒𝑑𝑐𝑠ℎ is a binary variable indicating whether household ℎ in village 𝑠 subcounty c, provided a phone number. The vector 𝑋𝑐𝑠ℎ ′ includes household-level covariates such as respondent sex, age, 14 education level, household size and the acreage of own cultivable land. The term Φ(𝛼0 + α𝑋𝑐𝑠ℎ ′) represents the cumulative distribution function of the standard normal distribution. From this probit model, the inverse Mills ratio is calculated, which corrects for selection bias and is subsequently incorporated into the second-stage model. In the second stage, another linear probability model (LPM) is specified to examine the household’s purchase decision: 𝑌𝑐𝑠ℎ = 𝛽0 + 𝛽1𝑇𝑒𝑥𝑡𝑐𝑠ℎ + 𝛽2𝐼𝑛𝑣𝑒𝑟𝑠𝑒𝑀𝑖𝑙𝑙𝑠𝑅𝑎𝑡𝑖𝑜𝑐𝑠ℎ + 𝛾′𝑋𝑐𝑠ℎ + 𝑣𝑐 + 𝜀𝑐𝑠ℎ (3) This equation is structurally identical to the earlier model (1) but includes the inverse Mills ratio to adjust for the selection effect, following the method introduced by Heckman (1979). Our primary focus remains on the coefficient 𝛽1, which estimates the effect of the text message reminders on the purchase decision. Standard errors are adjusted for clustering at the village level, in addition to the use of robust standard errors, to account for correlation and ensure the robustness of the statistical inference. Another variable in the dataset records whether participants could recall the text message reminder. The analysis categorizes participants into four groups: those who received and recalled the text, those who received but did not recall the text, those who did not receive but recalled the text, and those who did not receive nor recall the text. Due to the limited number of households responding to both relevant questions, the analysis includes 94 and 158 farmers in 2022 and 2023 respectively. To examine these groups, a similar model is specified: 𝑌𝑐𝑠ℎ = 𝛽0 + 𝛽1𝑆𝑒𝑛𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ + 𝛽2𝑆𝑒𝑛𝑡𝑁𝑜𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ + 𝛽3𝑁𝑜𝑡𝑆𝑒𝑛𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ + 𝛾′𝑋𝑐𝑠ℎ + 𝑣𝑐 + 𝜀𝑐𝑠ℎ (4) where 𝑆𝑒𝑛𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ, 𝑆𝑒𝑛𝑡𝑁𝑜𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ and 𝑁𝑜𝑡𝑆𝑒𝑛𝑡𝑅𝑒𝑐𝑎𝑙𝑙𝑒𝑑𝑐𝑠ℎ are indicators for whether household ℎ in village 𝑠 , subcounty 𝑐 falls into the respective group, with the group “text not sent and not recalled” serving as the reference category. All other variables and the presentation of results remain consistent with the previous model. To assess whether text message reminders significantly impact the total sales of the village, two additional linear regressions are estimated as follows: 𝑇𝑜𝑡𝑎𝑙𝑆𝑎𝑙𝑒𝑠𝑠 = 𝛽0 + 𝛽1𝐺𝑟𝑜𝑢𝑝ℎ𝑖𝑔ℎ + 𝛽2𝐺𝑟𝑜𝑢𝑝𝑙𝑜𝑤 + 𝛽3𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 + 𝛽4𝑃𝑟𝑒𝑣𝑖𝑜𝑢𝑠𝑆𝑎𝑙𝑒𝑠𝑠 + 𝜀𝑠 (5) 𝑇𝑜𝑡𝑎𝑙𝑆𝑎𝑙𝑒𝑠𝑠 = 𝛽0 + 𝛽1𝑁𝑢𝑚𝑃𝑒𝑜𝑝𝑙𝑒𝑇𝑒𝑥𝑡 + 𝛽2𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 + 𝛽3𝑃𝑟𝑒𝑣𝑖𝑜𝑢𝑠𝑆𝑎𝑙𝑒𝑠𝑠 + 𝜀𝑠 (6) In equation (5), 𝐺𝑟𝑜𝑢𝑝ℎ𝑖𝑔ℎ indicates whether more than two-thirds of the 10 participants in village s were sent text messages, while 𝐺𝑟𝑜𝑢𝑝𝑙𝑜𝑤 shows whether the percentage is less than one-third. Both 𝛽1 and 𝛽2 are of research interest in this model. In equation (6), 𝑁𝑢𝑚𝑃𝑒𝑜𝑝𝑙𝑒𝑇𝑒𝑥𝑡 denotes the total number of farmers in village s who were sent text message reminders over the two years. The coefficient 𝛽1  is the main parameter of interest in this model. In both regressions, 𝑇𝑜𝑡𝑎𝑙𝑆𝑎𝑙𝑒𝑠𝑠 refers to the sum of total sales (in thousands) in village s starting from the second season in 2022 to the first season in 2024. 𝑃𝑟𝑒𝑣𝑖𝑜𝑢𝑠𝑆𝑎𝑙𝑒𝑠𝑠accounts for sales (in thousands) during the first season of 2022, and 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 indicates whether village s is part of the treatment group. The error term is denoted by 𝜀𝑠, and symmetric statistical tests are also conducted. 15 6 Results Table 1 displays summary statistics for the control and treatment groups respectively based on surveys. There is little statistical difference between the two groups suggesting effective randomization. Table 2 provides results of the treatment within the initial season of implementation. While estimates of the treatment effects are of a moderate size, they are quite fragile and noisy. The treatment is only marginally significant within the first season, and only when using robust standard errors. That said, the significance of the treatment effect is unaffected by the inclusion of control variables, supporting the notion of effective randomization. Table 1: Baseline Randomization Balance Check between Treatment Group and Control Group Control Treatment Difference Variables N Mean N Mean Mean t-value Purchased 525 0.446 608 0.442 0.003 0.111 Years of Schooling 525 5.977 608 5.855 0.122 0.546 Respondent Sex (1=Male) 525 0.613 608 0.582 0.031 1.064 Respondent Age (years) 525 41.65 608 41.96 -0.306 -0.316 Household Size (count) 525 7.208 608 7.174 0.033 0.171 Own Cultivable Land (acres) 525 3.303 608 3.341 -0.038 -0.249 Distance to Sales Point (minutes) 379 19.480 459 18.296 1.184 0.826 Frequency of Vine Purchase Every Season 525 0.101 608 0.107 -0.006 -0.327 Every Other Season 525 0.109 608 0.138 -0.030 -1.505 Once in a While 525 0.310 608 0.332 -0.022 -0.781 Never 525 0.480 608 0.423 0.057 1.935* Note: Column 6 is the t-value of the difference between the mean value of the control and treatment groups. Asterisks indicate the following: ***=p<0.01, **=p<0.05, and *=p<0.1 16 Table 2. Effect of behavioral nudges on the purchase of quality seed of improved sweetpotato varieties. (1) (2) (3) (4) (5) (6) Variables Purchased Purchased Purchased Purchased Purchased Purchased Treatment 0.065* 0.070* 0.068* 0.065 0.070 0.068 (1.68) (1.85) (1.82) (1.44) (1.57) (1.55) Respondent Sex -0.026 -0.030 -0.026 -0.030 (-0.81) (-0.95) (-0.84) (-0.99) Respondent Age -0.000 -0.000 -0.000 -0.000 (-0.25) (-0.20) (-0.22) (-0.17) Years of Schooling 0.017*** 0.017*** 0.017*** 0.017*** (3.83) (3.78) (3.62) (3.57) Household Size 0.008* 0.007 0.008* 0.007 (1.82) (1.61) (1.69) (1.51) Own Cultivable Land 0.016*** 0.016*** 0.016*** 0.016*** (2.96) (2.97) (2.93) (2.94) Constant 0.349*** 0.166* 0.134 0.349*** 0.166 0.134 (4.94) (1.87) (1.50) (3.79) (1.49) (1.19) Observations 1,133 1,133 1,133 1,133 1,133 1,133 Adjusted R-squared 0.017 0.039 0.039 0.017 0.039 0.039 Number of Counties 15 15 15 15 15 15 Frequency of Vine Purchase No No Yes No No Yes Sub-county FE Yes Yes Yes Yes Yes Yes Note: Columns (1) to (3) show the results from the baseline regression with robust standard errors and columns (4) to (6) are results with village-level clustered standard error. 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡, representing whether the household receives the treatment, is interpreted as the average treatment effect of behavioral nudges. All models include sub- county-level fixed effects. T-statistics are shown in parentheses. Asterisks indicate the following: ***=p<0.01, **=p<0.05, and *=p<0.1. 17 Table 3. Effect of text message reminders on the purchase of quality vines of improved sweet potato varieties 2022 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchased Repurchased Repurchased Repurchased Repurchased Repurchased Text Sent 0.042 0.038* 0.027 0.042* 0.038** 0.027 (0.027) (0.023) (0.024) (0.022) (0.019) (0.019) Respondent Sex -0.034 -0.034 (0.028) (0.030) Respondent Age 0.002* 0.002** (0.001) (0.001) Years of Schooling 0.009* 0.009** (0.005) (0.004) Household Size 0.004 0.004 (0.003) (0.003) Own Cultivable Land -0.009** -0.009** (0.004) (0.004) Constant -0.035 0.103 0.097 -0.035* 0.103 0.097 (0.023) (0.143) (0.164) (0.020) (0.147) (0.179) Frequency of Vine Purchase No Yes Yes No Yes Yes Observations 194 194 194 194 194 194 Adjusted R-squared 0.045 0.106 0.123 0.045 0.106 0.123 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes 18 Table 3 provides an analysis of the impact of text message reminders on the purchase of quality seed by previous buyers in the treatment group, examining data from both 2022 and 2023. In 2022, the likelihood of households purchasing improved vines increased by 3.8% and 4.2% in two of the models when text message reminders were sent. However, this result shows some fragility, as models (3) and (6) do not yield statistically significant outcomes. These two models incorporate a set of control variables, revealing that socioeconomic factors, particularly age, years of schooling, and the acreage of cultivable land, is associated with the farmer’s decision to purchase improved seed. The lack of significance when controlling for these factors suggests the potential that the reception of text reminders is affected by wealth, which is also correlated with repurchase. In contrast, the results from 2023 tell a different story. In models (3) and (6) for 2023, after accounting for socioeconomic factors, previous purchase frequency, and subcounty fixed effects, the likelihood of previous buyers in the treatment group purchasing quality seed in the second season decreased by 18.1%. Other models for 2023 did not yield statistically significant results. This indicates that the effect of text message reminders is either not statistically significant or, in some cases, even negatively significant in 2023. The lack of positive statistical significance suggests that the impact of text messages either diminishes over time or may be negatively autocorrelated over seasons. By the following year, sending more and more reminders are no longer effective in increasing the likelihood of households purchasing quality seed. The findings indicate that text messages might be effective for repurchase, but not for several seasons in a row. Their influence does not sustain over longer periods. However, the variability and noise in these results suggests that further research is necessary to clarify these findings and better understand the dynamics. To correct for potential sample selection bias in the above results, a probit regression was employed as the first stage of the Heckman selection model, providing the inverse Mills ratio for the subsequent stage of analysis. Table 4 presents the results of this probit regression, which estimates the probability of individuals listing their phone numbers for the years 2022 and 2023. The results demonstrate that age and years of schooling were significant determinants of phone number listing in both years. As reflected by the negative coefficients for age, older individuals were less likely to list their phone numbers, while those with more years of education were more likely to provide their contact information. Furthermore, the coefficient of household size showed statistical significance in 2022, and the acreage of own cultivable land was a significant predictor in 2023. However, neither variable demonstrated statistically significant effects consistently across both years. 19 Table 3 continued. Effect of text message reminders on the purchase of quality vines of improved sweet potato varieties 2023 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchased Repurchased Repurchased Repurchased Repurchased Repurchased Text Sent -0.121 -0.141 -0.181** -0.121 -0.141 -0.181** (0.087) (0.085) (0.084) (0.090) (0.086) (0.083) Respondent Sex -0.059 -0.059 (0.041) (0.042) Respondent Age -0.002* -0.002* (0.001) (0.001) Years of Schooling 0.004 0.004 (0.006) (0.006) Household Size -0.013** -0.013** (0.006) (0.005) Own Cultivable Land 0.016* 0.016* (0.008) (0.009) Constant 0.217** 0.130 0.307** 0.217** 0.130 0.307** (0.103) (0.093) (0.131) (0.104) (0.091) (0.128) Frequency of Vine Purchase No Yes Yes No Yes Yes Observations 235 235 235 235 235 235 Adjusted R-squared -0.017 0.015 0.035 -0.017 0.015 0.035 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes Note: * indicates a confidence level of 90%, ** indicates 95%, and *** indicates 99%. 20 Table 4. Summary of probit regression results - first stage of Heckman selection model 2022 2023 VARIABLES Phone Number Listed Phone Number Listed Respondent Sex 0.061 0.074 (0.116) (0.128) Respondent Age -0.015*** -0.017*** (0.003) (0.004) Years of Schooling 0.127*** 0.098*** (0.019) (0.021) Household Size 0.029* 0.029 (0.016) (0.018) Own Cultivable Land 0.012 0.077*** (0.021) (0.027) Constant 0.744*** 0.826*** (0.232) (0.263) Observations 953 857 Pseudo R-squared 0.161 0.153 Log Likelihood -351.07 -290.93 Note: * indicates a confidence level of 90%, ** indicates 95%, and *** indicates 99%. 21 The second stage of the Heckman selection model, as shown in Table 5, assesses the effect of text message reminders on households' seed repurchase decisions in the second season while accounting for selection bias through the inclusion of inverse Mills ratio. Similarly, the analysis covers data from both 2022 and 2023, focusing on households that previously purchased seed. In 2022, the coefficient for the text message variable is positive, but only reaches statistical significance in models (4) and (5) when using clustered standard errors at the parish- village level. In these two models, the probability of repurchasing vines increases by 4.1% or 3.8% when the text message reminder is sent. This suggests that sending text message reminders had a small but positive effect on individual’s repurchasing decision, though this result is somewhat sensitive to model specifications. Other variables, such as respondent age, years of schooling, and own cultivable land, also show statistical significance in some models. Notably, younger individuals and those with more land were less likely to repurchase seed, whereas better educated farmers were more inclined to do so. In contrast, the results for 2023 tell a different story. The coefficient for text message reminders is negative and statistically significant in models (3), (5), and (6), indicating that the reminders had a substantially negative effect on repurchase behavior in that year. Households that received text message reminders had a 14.8% or 19.0% lower probability of repurchasing quality seed in the year 2023. This suggests that repeated reminders may have become less effective or even counterproductive over time. Additional factors, such as respondent age, household size and own cultivable land also appear to significantly influence repurchase decisions in 2023. Although the inverse Mills ratio does not achieve statistical significance in either year, its inclusion helps adjust for potential selection bias. The combined results from both 2022 and 2023 reflect a notable shift in the effectiveness of text message reminders, confirming our previous conclusion that their impact may diminish or even reverse over time. To compare vine purchasing decisions across different groups of farmers, 94 participants were divided into four categories: those who received and recalled a text message, those who received but did not recall the text, those who did not receive but recalled the text message (probably from a neighbor), and those who neither received nor recalled the text message. Table 6 presents a comparative analysis of seed purchasing behavior among these groups, with a focus on the influence of text message recall and receipt. In 2022, models (1), (2), (4), and (5) show significant results for the group that received and recalled the text messages. Depending on the specific model, participants in this group were 18.7% or 15.3% more likely to purchase quality seed compared to those who did not receive and could not recall the text messages. However, this finding is not entirely robust, as models (3) and (6) did not yield statistically significant results. In these latter models, household size emerged as a significant factor influencing farmers' purchasing decisions, suggesting that particular household characteristics may also play a critical role in this context. In contrast, the 2023 analysis reveals that there were no statistically significant differences in seed purchasing behavior between those who did receive and could recall the text messages and the base group. This pattern mirrors the findings in Table 3, indicating that while sending recallable messages can initially increase the likelihood of purchasing seed, this positive effect fades away over time. Moreover, the variability and inconsistency in the results also underscores the need for further research to confirm these preliminary results. 22 Table 5. Summary of ordinary least squares regression results - second stage of Heckman selection model 2022 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchase d Repurchase d Repurchase d Repurchase d Repurchase d Repurchased Text Sent 0.041 0.038 0.026 0.041* 0.038** 0.026 (0.029) (0.026) (0.029) (0.022) (0.018) (0.019) Respondent Sex -0.032 -0.032 (0.031) (0.029) Respondent Age 0.002* 0.002** (0.001) (0.001) Years of Schooling 0.009 0.009** (0.006) (0.004) Household Size 0.004 0.004 (0.004) (0.003) Own Cultivable Land -0.009* -0.009** (0.005) (0.004) Inverse Mills Ratio -0.388 -0.327 -0.375 -0.388 -0.327 -0.375 (0.432) (0.401) (0.420) (0.452) (0.380) (0.386) Constant 0.095 0.208 0.207 0.095 0.208 0.207 (0.138) (0.224) (0.248) (0.144) (0.199) (0.227) Frequency of Vine Purhcase No Yes Yes No Yes Yes Observations 194 194 194 194 194 194 Adjusted R-squared 0.045 0.104 0.123 0.045 0.104 0.123 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes 23 Table 5 continued. Summary of ordinary least squares regression results - second stage of Heckman selection model 2023 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchased Repurchased Repurchased Repurchased Repurchased Repurchased Text Sent -0.129 -0.148 -0.190** -0.129 -0.148* -0.190** (0.093) (0.091) (0.092) (0.091) (0.087) (0.084) Respondent Sex -0.058 -0.058 (0.043) (0.042) Respondent Age -0.002 -0.002* (0.001) (0.001) Years of Schooling 0.003 0.003 (0.006) (0.006) Household Size -0.014** -0.014** (0.006) (0.006) Own Cultivable Land 0.017* 0.017* (0.009) (0.009) Inverse Mills Ratio 0.691 0.612 0.684 0.691 0.612 0.684 (0.459) (0.470) (0.470) (0.426) (0.448) (0.439) Constant -0.008 -0.064 0.097 -0.008 -0.064 0.097 (0.196) (0.195) (0.215) (0.173) (0.181) (0.192) Frequency of Vine Purchase No Yes Yes No Yes Yes Observations 235 235 235 235 235 235 Adjusted R-squared -0.006 0.023 0.046 -0.006 0.023 0.046 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes Note: * indicates a confidence level of 90%, ** indicates 95%, and *** indicates 99%. 24 Table 6. Comparative analysis of vine purchasing behavior among farmers based on text message recall and receipt 2022 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchased Repurchased Repurchased Repurchased Repurchased Repurchased Text Sent & Recalled 0.187* 0.153** 0.132 0.187* 0.153** 0.132 (0.099) (0.091) (0.092) (0.094) (0.071) (0.085) Text Sent & Not Recalled 0.136 0.092 0.098 0.136 0.092 0.098 (0.090) (0.078) (0.099) (0.090) (0.075) (0.091) Text Not Sent & Recalled 0.122 0.115 0.074 0.122 0.115 0.074 (0.084) (0.074) (0.094) (0.082) (0.070) (0.106) Respondent Sex -0.069 -0.069 (0.061) (0.063) Respondent Age 0.003 0.003 (0.002) (0.002) Years of Schooling 0.009 0.009 (0.010) (0.006) Household Size 0.014* 0.014* (0.008) (0.008) Own Cultivable Land -0.019 -0.019 (0.012) (0.013) Constant -0.162* 0.192 0.114 -0.162* 0.192 0.114 (0.092) (0.264) (0.304) (0.088) (0.267) (0.304) Frequency of Vine Purchase No Yes Yes No Yes Yes Observations 94 94 94 94 94 94 Adjusted R-squared 0.054 0.092 0.116 0.054 0.092 0.116 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes 25 Table 6 continued. Comparative analysis of vine purchasing behavior among farmers based on text message recall and receipt 2023 (1) (2) (3) (4) (5) (6) Robust SE Parish-Village Clustered SE VARIABLES Repurchased Repurchased Repurchased Repurchased Repurchased Repurchased Text Sent & Recalled -0.039 -0.082 -0.106 -0.039 -0.082 -0.106 (0.063) (0.069) (0.073) (0.065) (0.069) (0.074) Text Sent & Not Recalled 0.035 -0.026 -0.060 0.035 -0.026 -0.060 (0.092) (0.098) (0.098) (0.092) (0.104) (0.097) Text Not Sent & Recalled 0.193 0.165 0.150 0.193 0.165 0.150 (0.127) (0.129) (0.128) (0.138) (0.143) (0.136) Respondent Sex -0.022 -0.022 (0.066) (0.063) Respondent Age -0.005** -0.005*** (0.002) (0.002) Years of Schooling -0.001 -0.001 (0.008) (0.009) Household Size -0.008 -0.008 (0.007) (0.007) Own Cultivable Land 0.014 0.014 (0.014) (0.015) Constant 0.214* 0.101 0.296** 0.214 0.101 0.296** (0.111) (0.096) (0.127) (0.148) (0.128) (0.127) Frequency of Vine Purchase No Yes Yes No Yes Yes Observations 158 158 158 158 158 158 Adjusted R-squared -0.015 0.007 0.012 -0.015 0.007 0.012 Subcounty Fixed Effects Yes Yes Yes Yes Yes Yes 26 Table 7 examines the impact of text message reminders on total seed sales at the village level over four seasons from 2022 to 2024, accounting for treatment effects and initial sales in the first season of 2022. In columns (1) and (2), villages are categorized into three groups based on the proportion of households receiving text reminders: more than two-thirds, between one- third and two-thirds, and less than one-third. The results indicate no statistically significant differences in total quality seed sales among these groups, suggesting that the proportion of households receiving reminders does not significantly influence overall sales of a particular village. Columns (3) and (4) shift the analysis to the number of individuals in each village who received text reminders, but similarly, this variable also fails to yield a statistically significant impact on sales. This analysis necessarily suffers from a low number of observations in what appears to be a very noisy treatment. Overall, the findings in this table suggest that the effect of sending text message reminders on total vine sales at the village level is small relative to the random noise in the system and does not persist over the examined period. Table 7. Effect of text message reminders on total vine sales at the village level (1) (2) (3) (4) Total Sales Total Sales Total Sales Total Sales VARIABLES (in ‘000) (in ‘000) (in ‘000) (in ‘000) More than 2/3 receiving text 24.947 24.905 (41.821) (41.267) Less than 1/3 receiving text -0.898 -14.834 (13.641) (15.262) Number of people receiving text -2.056 3.231 (2.755) (4.307) Treatment -24.848* -28.123 (12.825) (17.699) Previous Sales (in thousands) 0.188 0.151 0.164 0.143 (0.120) (0.120) (0.120) (0.119) Constant 23.905 51.496** 29.490** 37.433** (14.751) (20.364) (13.820) (14.601) Observations 106 106 106 106 Adjusted R-squared -0.002 0.025 0.010 0.024 Note: * indicates a confidence level of 90%, ** indicates 95%, and *** indicates 99%. 27 7 Discussion and conclusions This study examined the effect of a bundle of interventions including behavioral nudges on sustained use of quality seed of sweetpotato varieties. Seed systems of sweetpotato, a vegetatively propagated crop, are largely informal (McGuire and Sperling, 2016). Farmers rely on poor quality seed obtained from their own harvest or from a network of family members, neighbors and friends. Such seed is often infected with pests and diseases which greatly reduces yield, affecting household food and income security (Okello et al, 2023). Nudges in the form of mobile phone-based text reminders have been shown to be effective in inducing behavior change among smallholder farmers (Fay- Rubzen, 2023). Our study finds some evidence that this could indeed be the case for smallholder sweetpotato farmers also. We find that after resolving other key constraints that smallholder farmers face, namely access to seed and quality assurance, text message reminders increased the likelihood that farmers will repurchase quality seed. In particular, our findings indicate that exposure to text message reminders increases the chances of farmers repurchasing quality seed by 4%. Moreover, zooming into the nature of exposure by examining recall of text messages received also indicates that nudges increase the likelihood of purchase of quality seed among farmers who both received the nudges and could recall them. These findings suggest that nudges in the form of text reminders can influence farmers’ behavior towards innovation use in agriculture. The initially positive effect of reminders suggest that they can be usefully deployed in the promotion of quality seed of vegetatively propagated crops. However, the results are not robust to all model specifications indicating fragility. We, for instance, find a null effect of text reminders for other model specifications as well as a reduction in the likelihood of purchase of quality seed when reminders are repeated. Some elements of our findings corroborate those of Zachmann et al (2023). They found that using nudges to promote the use of practices that reduce pesticide applications by grapevine farmers resulted in null effect. We also find that continued deployment of text messages reminders either diminishes seed purchase over time or may be negatively autocorrelated with the seed purchase over seasons. The diminishing effect of nudges on behavior has been reported in other fields. Harnischmaker et al (2023) for instance find a drastic decline in nudges aimed at keeping recommended distances to contain the spread of COVID-19 in retail settings one year after the intervention commenced. Sasaki et al (2023), on the other hand, report a negative effect gain-framed repeated nudges aimed at encouraging contact avoidance to curtail the spread of COVID-19 in Japan, thus corroborating our finding that under certain contexts repeated exposure to nudges can be counterproductive. The fragility of the findings needs to be put in context. First, as a vegetatively propagated crop, farmers mostly plant vines/material harvested from own gardens (McGuire and Sperling, 2016) and resort to using external sources when own vines show visible signs of pest/disease infection and significant reduction in yield (Gibson et al, 2011). Both are unlikely to occur when farmers use quality seed. It is estimated that it takes more than three seasons of repeated planting of sweetpotato seed for the signs of seed degeneration (disease/pest infection and depressed yield) to be noticeable (Ogero et al, 2023). Better performance of the quality seed (by being disease free and high yield) may therefore have discouraged farmers from purchasing quality seed. In other instances, farmers who initially purchased seed in the previous season may have opted to buy less than previously in the second season because they already have good material and are simply supplementing (topping up) what they have. The latter behavior has been documented for several of the vegetatively propagated crops (Jacobson et al, 2019; Ogero et al, 2019; Navarette et al, 2022). A second important contextual factor relates to the typical use of mobile phones and literacy rates of the farmers in the study area. The majority of rural farmers keep their phones turned off at certain times when not in use to 28 preserve battery power. In other instances, phones can be off for extended periods of time during electricity blackouts (Houngbonon et al, 2021). These result in intermittence of use of phones that could have resulted in some farmers missing to receive/see text messages. Further, some less literate rural farmers use mobile phones for voice (calling and receiving) calls only and do not know how to navigate the text message platforms (Okello, 2013). Our findings also highlight the socio-demographic factors notable wealth, education, age, household size and endowment with land driving the likelihood of repurchase of seed. The finding that land ownership and wealth influence seed repurchase decisions is interesting for two reasons. First, it corroborates the findings of Bayiyana et al (2023) that liquidity constraint is a major factor influencing demand for quality seed of sweetpotato varieties. Second, this finding has a major gender implication. Women tend to dominate sweetpotato seed acquisition and other field operations (Bayiyana et al, 2023). At the same time, women also tend to have less access to land (Mudege et al, 2018) and wealth (Gilligan et al, 2023). Hence, the current finding that wealth and access to land influence seed repurchases suggest that women farmers are likely to be excluded in the use quality unless supported. Notwithstanding the fragility of findings relating to impacts on repurchases, this study finds evidence that nudges in the form of mobile phone-based text message reminders initially positively and next negatively influence the continued purchase of quality seed by smallholder sweetpotato farmers, which suggests that reminders should be judiciously used and, in particular, that their repetition may be counterproductive. They suggest that nudges can influence positively sustainable use of quality seed in an informal seed system when other key constraints are resolved, provided that they are not overused. In this study two of these constraints were resolved – namely, seed access and quality assurance. The fragility of the results suggests the need for further studies to better understand the effect of text message reminder nudges on the use of quality seed. 29 8 Acknowledgments This research was funded by the Bill and Melinda Gates Foundation (BMGF), through its investment [OPP1213329] awarded to the International Potato Center (SweetGAINS) and the CGIAR Initiative on Market Intelligence and Seed Equal as well as Program for Seed Systems Innovation for Vegetatively Propagated Crops in Africa (PROSSIVA). We also acknowledge research assistance support from Mr. Edwin Sserenkuma, Ms. Harriet Ayoko, and Mr. John Robert Otukei, and logistical support from the Kuju Sub-County Agricultural Officer Mr. Dennis Odeke. 30 9 References Almekinders, C. J., Walsh, S., Jacobsen, K. S., Andrade-Piedra, J. L., McEwan, M. A., de Haan, S., ... & Staver, C. (2019). Why interventions in the seed systems of roots, tubers and bananas crops do not reach their full potential. Food Security, 11, 23-42. Audi, P., Nagarajan, L., Jones, R., & Ibrahim, M. S. (2009). Pigeonpea seed supply and diversity: A case study of local seed markets in makueni district, eastern Kenya. In Seed Trade in Rural Markets (pp. 95-123). Routledge. Balew, S., Bulte, E., Abro, Z., & Kassie, M. (2022). Incentivizing and nudging farmers to spread information: Experimental evidence from Ethiopia. American Journal of Agricultural Economics. 105(3), 994 – 1010. https://doi.org/10.1111/ajae.12346 Chai, Y., Pannell, D. J., & Pardey, P. G. (2023). Nudging farmers to reduce water pollution from nitrogen fertilizer. Food Policy, 120, 102525. Echodu, R., Edema, H., Wokorach, G., Zawedde, C., Otim, G., Luambano, N., ... & Asiimwe, T. (2019). Farmers' practices and their knowledge of biotic constraints to sweetpotato production in East Africa. Physiological and molecular plant pathology, 105, 3-16. Gasura, E., & Mukasa, S. B. (2010). Prevalence and implications of sweetpotato recovery from sweet potato virus disease in Uganda. African Crop Science Journal, 18(4). Gatto, M., de Haan, S., Laborte, A., Bonierbale, M., Labarta, R., & Hareau, G. (2021). Trends in varietal diversity of main staple crops in Asia and Africa and implications for sustainable food systems. Frontiers in Sustainable Food Systems, 5, 626714. Gibson, R. W., Namanda, S., & Sindi, K. (2011). Sweetpotato seed systems in East Africa. In African Crop Science Conference Proceedings. Available at: https://cgspace.cgiar.org/server/api/core/bitstreams/10879672-4a13- 4c81-9efd-f089c89dce57/content Jogo, W., Bocher, T., & Grant, F. (2021). Factors influencing farmers’ dis-adoption and retention decisions for biofortified crops: the case of orange-fleshed sweetpotato in Mozambique. Agrekon, 60(4), 445-459. Harnischmacher, J., Merkl, L. M., & Germelmann, C. C. (2023, May). Nudging Physical Distance During COVID-19: Short-Term and Long-Term Wear-Out Effects of Nudges in a Retail Setting. In Academy of Marketing Science Annual Conference (pp. 197-207). Cham: Springer Nature Switzerland. Hodgkin, Toby, R. Rana, J. Tuxill, Didier Balma, A. Subedi, I. Mar, Deborah Karamura et al. "Seed systems and crop genetic diversity in agroecosystems." In Managing biodiversity in agricultural ecosystems, pp. 77-116. Columbia University Press, 2007. Houngbonon, G. V., Le Quentrec, E., & Rubrichi, S. (2021). Access to electricity and digital inclusion: evidence from mobile call detail records. Humanities and Social Sciences Communications, 8(1), 1-11. Low, J. W., Ortiz, R., Vandamme, E., Andrade, M., Biazin, B., & Grüneberg, W. J. (2020). Nutrient-dense orange- fleshed sweetpotato: advances in drought-tolerance breeding and understanding of management practices for sustainable next-generation cropping systems in sub-Saharan Africa. Frontiers in Sustainable Food Systems, 4, 50. McEwan, M. et al. "“Breaking through the 40% adoption ceiling: Mind the seed system gaps.” A perspective on seed systems research for development in One CGIAR." Outlook on Agriculture 50.1 (2021): 5-12. https://doi.org/10.1111/ajae.12346 https://cgspace.cgiar.org/server/api/core/bitstreams/10879672-4a13-4c81-9efd-f089c89dce57/content https://cgspace.cgiar.org/server/api/core/bitstreams/10879672-4a13-4c81-9efd-f089c89dce57/content 31 McEwan, M., Spielman, D. J., Okello, J. J., Hareau, G., Bartle, B., Mbiri, D., ... & Maredia, M. K. (2021). Exploring the regulatory space for improving availability, access and quality of vegetatively propagated crop seed: potato in Kenya. Available at: https://cgspace.cgiar.org/server/api/core/bitstreams/25a09177-9454-4092-9937- 608c9595a039/content McGuire, S., & Sperling, L. (2016). Seed systems smallholder farmers use. Food security, 8, 179-195. Mugisa, I., Karungi, J., Musana, P., Odama, R., Anyanga, M. O., Edema, R., ... & Yada, B. (2023). Heterotic gains, transgressive segregation and fitness cost of sweetpotato weevil resistance expression in a partial diallel cross of sweetpotato. Euphytica, 219(10), 110. Mukasa, S.B., Namanda, S., Musoke, C., Magezi, S., Kyalo, G., Mwanga, R.O.M.,Ssemakula, G., Ball, A-M., Kasharu, A., Kyamanywa, S., Tumuboine, E., 2016. Technical guidelines for inspection and certification of sweetpotato planting material in Uganda: recommended practices. ISSN 2518-248X. Mwanga, R. O., Mayanja, S., Swanckaert, J., Nakitto, M., Zum Felde, T., Grüneberg, W., ... & Muzhingi, T. (2021). Development of a food product profile for boiled and steamed sweetpotato in Uganda for effective breeding. International journal of food science & technology, 56(3), 1385-1398. Namanda, S., Gibson, R., & Sindi, K. (2011). Sweetpotato seed systems in Uganda, Tanzania, and Rwanda. Journal of Sustainable Agriculture, 35(8), 870-884. Namanda, S.; Mwanga, R.; Mukasa, S.; Talengera, D.; Musoke, C.; Kyalo, G.; Low, J.; Ssemakula, G.; Magezi, S.; Ball, A.M. 2019. Sweetpotato virus pathogen-tested planting material of susceptible varieties results in root yield increase in Uganda. Crop Protection. ISSN: 0261-2194. 124. 1-5 p. https://hdl.handle.net/10568/102370, DOI : https://doi.org/10.1016/ j.cropro.2019.104851 Navarrete, I., López, V., Borja, R., Oyarzún, P., Garrett, K. A., Almekinders, C. J., ... & Andrade-Piedra, J. L. (2022). Variety and on-farm seed management practices affect potato seed degeneration in the tropical highlands of Ecuador. Agricultural Systems, 198, 103387. Okello, J., Shikuku, K. M., Lagerkvist, C. J., Rommel, J., Jogo, W., Ojwang, S., ... & Elungat, J. (2023). Social incentives as nudges for agricultural knowledge diffusion and willingness to pay for certified seeds: Experimental evidence from Uganda. Food Policy, 120, 102506. Okello, J. J., Swanckaert, J., Martin-Collado, D., Santos, B., Yada, B., Mwanga, R. O., ... & Campos, H. (2022). Market intelligence and incentive-based trait ranking for plant breeding: a sweetpotato pilot in Uganda. Frontiers in Plant Science, 13, 808597. Okello, J., Zhou, Y., Barker, I., & Schulte-Geldermann, E. (2019). Motivations and mental models associated with smallholder farmers’ adoption of improved agricultural technology: Evidence from use of quality seed potato in Kenya. The European Journal of Development Research, 31, 271-292. Okello, J. J. (2013). Use of information and communication tools and services by rural grain traders. Technology, Sustainability, and Rural Development in Africa, 152. Okello, J. J., & Swinton, S. M. (2007). Compliance with international food safety standards in Kenya's green bean industry: Comparison of a small-and a large-scale farm producing for export. Applied Economic Perspectives and Policy, 29(2), 269-285. Ogero, K., Okuku, H. S., Wanjala, B., McEwan, M., Almekinders, C., Kreuze, J., ... & van der Vlugt, R. (2023). Degeneration of cleaned-up, virus-tested sweetpotato seed vines in Tanzania. Crop Protection, 169, 106261. https://cgspace.cgiar.org/server/api/core/bitstreams/25a09177-9454-4092-9937-608c9595a039/content https://cgspace.cgiar.org/server/api/core/bitstreams/25a09177-9454-4092-9937-608c9595a039/content https://hdl.handle.net/10568/102370 https://doi.org/10.1016/%20j.cropro.2019.104851 32 Ogero, K. O., Kreuze, J. F., McEwan, M. A., Luambano, N. D., Bachwenkizi, H., Garrett, K. A., ... & Van Der Vlugt, R. A. A. (2019). Efficiency of insect-proof net tunnels in reducing virus-related seed degeneration in sweet potato. Plant Pathology, 68(8), 1472-1480. Rachkara, P., Phillips, D. P., Kalule, S. W., & Gibson, R. W. (2017). Innovative and beneficial informal sweetpotato seed private enterprise in northern Uganda. Food Security, 9, 595-610. Rola-Rubzen, M. F., Sarmiento, J. M., Villano, R. A., Murray-Prior, R., Khan, M. F. U., Alam, M. M., ... & Das, K. K. (2023). Nudging Via Text Reminders: Can Behavioural Economics Increase the Adoption of Conservation Agriculture-based Sustainable Intensification Technologies in South Asia? Available at: https://ageconsearch.umn.edu/record/335953?ln=en&v=pdf Rogers EM (1995) Diffusion of innovations, 4th edn. Free Press, New York Rogers EM (1962) Diffusion of innovations, 1st edn. Free Press, New York Sasaki, S., Kurokawa, H., & Ohtake, F. (2021). Effective but fragile? Responses to repeated nudge-based messages for preventing the spread of COVID-19 infection. The Japanese Economic Review, 72(3), 371-408. Schöley, M., & Padmanabhan, M. (2017). Formal and informal relations to rice seed systems in Kerala, India: Agrobiodiversity as a gendered social-ecological artifact. Agriculture and Human Values, 34(4), 969-982. Sperling, L., & Almekinders, C. J. (2023). Informal commercial seed systems: Leave, suppress or support them?. Sustainability, 15(18), 14008. Sperling, L., Louwaars, N., de Ponti, O., Smale, M., Baributsa, D., & van Etten, J. (2020). COVID-19 and seed security response now and beyond. Food Policy, 97, 102000. Spielman, D. J., Gatto, M., Wossen, T., McEwan, M., Abdoulaye, T., Maredia, M. K., & Hareau, G. (2021). Regulatory options to improve seed systems for vegetatively propagated crops in developing countries (Vol. 2029). Intl Food Policy Res Inst. Thaler, R., & Sunstein, C. (2008). Nudge: Improving decisions about health, wealth and happiness. In Amsterdam Law Forum; HeinOnline: Online (p. 89). Thiele, G., Dufour, D., Vernier, P., Mwanga, R. O., Parker, M. L., Schulte Geldermann, E., ... & Hershey, C. (2021). A review of varietal change in roots, tubers and bananas: consumer preferences and other drivers of adoption and implications for breeding. International Journal of Food Science & Technology, 56(3), 1076-1092. Van Vugt, D., & Franke, A. C. (2018). Exploring the yield gap of orange-fleshed sweet potato varieties on smallholder farmers’ fields in Malawi. Field crops research, 221, 245-256. Zachmann, L., McCallum, C., & Finger, R. (2023). Nudging farmers towards low-pesticide practices: Evidence from a randomized experiment in viticulture. Journal of the Agricultural and Applied Economics Association, 2(3), 497- 514. Zawedde, B. M., Harris, C., Alajo, A., Hancock, J., & Grumet, R. (2014). Factors influencing diversity of farmers’ varieties of sweet potato in Uganda: implications for conservation. Economic botany, 68, 337-349. https://ageconsearch.umn.edu/record/335953?ln=en&v=pdf 33 10 Appendix 10.1 Appendix 1: Poster used to activate feelings of nostalgia and loss aversion among treatment farmers 34 10.2 Appendix 2: Text message reminders sent to purchasers of quality vines of ISV at the beginning of Season 2 English language 1. Did you know that vines from local sources often contain many pests and diseases? 2. Did you know that planting quality vines found at your village sales stand leads to higher yields? 3. Hurry! Get quality sweetpotato vines at your village sales stand. 4. You get more, bigger and better roots when you use quality vines in your village sales stand local vines. 5. Farmers who plant quality vines sold in your village sales stand worry less about sweetpotato pests and diseases. 6. Farmers who are planting quality vines from your village sales stand are getting more, bigger and better roots. 7. Planting quality vines sold at your village sales stand gives you more sweetpotato to cook and sell, and more amukeke. Ateso language 1. Kijeni lem ijo ebe, akwii nu acok nu itocununete amisirin wok etapit duc ajaut keda ikur ka adekasinei nu acok. 2. Kijeni lem ijo ebe, aira akwii nu itojokaritai nu ejaasi agwelanaret na ocalo kon einakini acok aimini ejok. 3. Kiwonyuni !, kodum akwii nu acok nu itojokaritai kane egwelanarere kocaalo kon. 4. Icoki bino nywal apol, adito kede abeco ka ipito oboke icok akome yot ame otye acato icalo ni. 5. Akoriok luiraete akwii nuitojokaritai nukwana egwelario ocalokus, mam iyalongongoete kanu ikur ka adekasinei nu acok 6. Akoriok lu iraete akwii nu itojokaritai nu ejaasi agwelanaret na ocalo kon edumunete acok nu ejokuka, koburok ido kominitos ejok. 7. Aira akwii nu itojokaritai nu ejaasi agwelanaret na ocalo kon ijaikini ijo acok nu ipu nu nyaman ka agwelar da. Ka amukeke da na epol. Langi language 1. Onwongo ingoe ni oboke icok me kin paco kom gi pe yot; otye kede kudi apol acamo gi kede two pol? 2. Onwongo ingoe ni pito oboke icok akome yot mio icok nyak adwong? 3. Bunye! Nwong oboke icok akome yot ame two pe iye me apita ame otye acato icalo ni 4. Ikunyu icok apol, adito kede abeco ka ipito oboke icok akome yot ame ber kato pito oboke me kin paco 1. Icoki bino nywal apol, adito kede abeco ka ipito oboke icok akome yot ame otye acato icalo ni. 5. Oput ame pito oboke icok akomgi yot pe paro ni kudi acamo onyo two amako icok gi 6. Opur ame pito oboke icok akome yot gin kunyu icok apol, adito kede abeco 7. Pito oboke icok akome yot mii inwongo icok apol ame konyi me ateda kede acata, dang mii otere adwong 35 10.3 Annex 3: Criteria used for selecting salespersons 36 SEED SALES RECORD DATA FORM 37 STUDY QUESTIONNAIRE: USING BEHAVIORAL INTERVENTIONS TO PROMOTE ADOPTION OF IMPROVED SWEETPOTATO VARIETIES Informed Consent Form We are from [International Potato Centre (CIP)] and Amuria District Agricultural Office Sector. We are conducting a survey of sweetpotato farmers to understand their sources of agricultural information and sweetpotato production practices. You, as a sweetpotato farmer in Amuria district, have been randomly selected to participate in this study. Please note that your participation in this research is voluntary and your personal information (name and contacts) will be kept confidential. Your responses will be reported after being aggregated with those of others. You can choose to refuse to answer any questions and you are free to withdraw from further participation in this interview at any time. In case you decline or withdraw, this will not result in any negative consequences against you. We would, however, appreciate your participation and completion of the interview and your honest answers to the questions. If after the study you have any questions or complaint about this research, you can contact: Dr. Julius Okello from the International Potato Center (Phone: xxxx) or Mr. Moses Okim the DAO-Amuria District (Phone: xxx) The survey should take about 1hour 20 minutes to complete. Please note that you indicate consent to voluntarily participate in this survey by agreeing to continue with the interview. Can we continue with the interview? (1. Yes ,0. No) (If No, terminate the interview) 38 RESPONDENT IDENTIFICATION Details of demographic and socio-economic characteristics of respondent. Farmer ID Interview Date (DD/MM/YYYY) Interview Start Time (HH:MM) Enumerator Supervisor/Team leader District Sub-county (codes) Parish (codes) Village (codes) Id1. Name of respondent ………………………………………. Id2. Respondent’s other names …………………………………….. Id3. Telephone number (or that of family member or nearest neighbor) __________________ [if new] Id4. Sex of respondent 1. Male; 0. Female Id5. Age of Respondent (years) ____ Id6. Did you take part in the interviews in November -December last year (2021) 1=Yes 0=No Id7. If No to Q Id6, Are you representing your household member interviewed that time? 1=Yes 0=No 1d8. If Yes to QId7 What is the name of the person you are representing_________ Id9. Can you read and write Ateso or English language? 0- No 1- Yes, only in Ateso 2- Yes, in English 3- Yes, both Ateso and English Id10. Marital Status of Respondent (1. Single, 2. Married and live together with spouse, 3. Married but the spouse is away, 4. Separated or divorced, 5. Widowed) Id11. Relationship of respondent to household head (1. Household head; 2. Spouse; 3. Son/daughter; 4. Parent; 5. Brother/sister; 6. Son/daughter in-law; 7. Grandchild; 8. Other relative; 9. Hired worker; 10. Friend; 11. Non-relative) Id12. What is the gender of the Head of your HH? [if not respondent/spouse] 1- Male; 0- Female Id13. Years of schooling of respondent Id14. Primary occupation of respondent (1. Farming (crops and livestock); 2. Salaried employment; 3. Self- employed off-farm; 4. Casual laborer on-farm; 5. Casual laborer off- farm; 6. Student; 88. Other (Specify); 96. None) 1d15. Name of household head if not the respondent. 39 PART A: RESPONDENT AND HOUSEHOLD CHARACTERISTICS a1 How many people live in your household? ………… a2. Do you have salaried income? 1. Yes ,0. No a3. Does your spouse or other dependent member of your family have salaried income? 1. Yes ,0. No a4. Are you a member of any farmer/social organization (group, association, coop, SACCO, chama – merry-go- round, etc.) 1. Yes ,0. No a5. If Yes to Qa4, which of the following services have you ever received from the organization between April 2022 and now? 1=sweetpotato production practices 2=sweetpotato marketing 3=credit for sweetpotato 4=sweetpotato value addition 88=Other (specify for sweetpotato)______ 99. N/A 40 PART A2: EXPERIMENTATION AND INNOVATIVENESS a6. Please indicate your level of agreement with these statements 1= Not at all like me……………... 7= Very much like me 1. I don’t have adequate resources to try out new crop varieties 2. Before deciding to use a new crop variety, I would be able to try it out properly 3. In general, I am among the first in my circle of friends to buy a new seed variety when it appears 4. If I like a variety I rarely switch from it just to try something different 5. If I’m to purchase vines, I will feel it is safer to buy varieties I am familiar with 6. I generally consider changes to be a negative thing. I like to do the same old things rather than try new and different ones 7. I would rather try a new variety I am not very sure of than stick with a local variety I am used to 8. Often, I feel a bit uncomfortable even about changes that may potentially improve my life 9. When someone pressures me to change something, I tend to resist it even if I think the change may ultimately benefit me 10. I don’t change my mind easily 11. I prefer to leave experimenting with new things to someone else 12. I postpone investments until they really need to be done 13. I am now more aware of risks in farming than I was a few years ago, and I am more concerned about risks than I was then. 14. Financially, I can afford to take a few risks and experiment with new ideas 15. Financially, I can’t afford to make a poor decision, even a small one PART B: SWEETPOTATO PRODUCTION b1. How often do you grow sweetpotato? 1=Every season; 2=Every other season (1 season every year); 3=Once in a while (say a whole year could pass) b2. In which seasons do you plant sweetpotato in a year? 1=April/May; 2=July/August; 3=Both b3. In which specific months do you normally plant sweetpotato? …… 1. Jan; 2. Feb; 3. Mar; 4. Apr; 5. May; 6. Jun; 7. Jul; 8. Aug; 9. Sep; 10. Oct; 11. Nov; 12. Dec b4. In which particular months did you plant sweetpotato this year 2022? 0. Never planted; 1. Jan; 2. Feb; 3. Mar; 4. Apr; 5. May; 6. Jun; 7. Jul; 8. Aug; 9. Sep; 10. Oct; 11. Nov; 12. Dec b5. How many varieties of sweetpotato have you planted on your farm in: a) April/May season 2022_____ b) July/September season 2022__________? ………. b6. Which variety(s) of sweetpotato have you planted in the last 2 years? 1. Acil Acil; 2. Adoch; 3. Alero; 4. Apaka paka; 5. Apul apul; 6. Araka Araka; 7. Ateseke; 8. Awie twon gweno; 9. Bulula; 10. Bunduguza; 11. Dimbuka; 12. Egang/Emalayan; 13. Ejumula; 14. Kakamega; 15. Kampala; 16. Kavunza; 17. Kiwoko; 18. Kyebandula; 19. La Gulu; 20. La Lira; 21. Milika; 22. Mukiga; 23. Muwulu Aduuduma; 24. Nabagereka; 25. Naspot 1 (Bwenje); 26. Naspot 10 (Kabode); 27. Naspot 11 (Tomulabula); 28. Naspot 12 (Gerald); 29. Naspot 13 (Jowelia); 30. Naspot 8; 31. Naspot 9 (Vita); 32. Nawamba; 33. New Dimbuka; 34. New Kawogo; 35. Nimilabana; 36. Ochol/ Ocuc; 37. Okonynedo; 38. Oluli; 39. Oyitodege; 40. Semanda; 41. Tanzania/Osukut; 42. Tera; 43. Wanyamu kwena; 88. Others (specify) 41 b7. Which one is your most preferred sweetpotato variety? 1. Acil Acil; 2. Adoch; 3. Alero; 4. Apaka paka; 5. Apul apul; 6. Araka Araka; 7. Ateseke; 8. Awie twon gweno; 9. Bulula; 10. Bunduguza; 11. Dimbuka; 12. Egang/Emalayan; 13. Ejumula; 14. Kakamega; 15. Kampala; 16. Kavunza; 17. Kiwoko; 18. Kyebandula; 19. La Gulu; 20. La Lira; 21. Milika; 22. Mukiga; 23. Muwulu Aduuduma; 24. Nabagereka; 25. Naspot 1 (Bwenje); 26. Naspot 10 (Kabode); 27. Naspot 11 (Tomulabula); 28. Naspot 12 (Gerald); 29. Naspot 13 (Jowelia); 30. Naspot 8; 31. Naspot 9 (Vita); 32. Nawamba; 33. New Dimbuka; 34. New Kawogo; 35. Nimilabana; 36. Ochol/ Ocuc; 37. Okonynedo; 38. Oluli; 39. Oyitodege; 40. Semanda; 41. Tanzania/Osukut; 42. Tera; 43. Wanyamu kwena; 88. Others (specify) b8. This most preferred variety of sweetpotato is 1. Local; 2. Improved; 99. Don’t know b9. What is the flesh color of this most preferred variety? 1. Orange-fleshed; 2. Purple-fleshed; 3. White-fleshed; 4. Yellow-fleshed; 5. Cream-fleshed b10. For how long (years) have you grown this most preferred variety?_______ b11. What is the most important reason for preferring this variety in Qb7? 1. High yield; 2. Root underground storage; 3. Disease resistance (SPVD); 4. Stress tolerance (drought, soil fertility); 5. Skin colour; 6. Sweetness; 7. Mealiness (softness/smoothness feel in the mouth); 8. Beta carotene content/ nutritional & health benefits (flesh color is orange); 9. Bulking ability; 10. Early maturity (days to maturity); 11. Smoothness of the roots; 12. Shape of the roots; 88. Other (specify) b12. What is the main purpose of growing sweetpotato? 1. Fresh roots for consumption only; 2. Fresh roots for sale only; 3. Fresh roots for consumption with surplus sales; 4. Fresh roots for processing; 5. Dried chips (Amukeke) for home consumption; 6. Dried roots/Amukeke for sale; 7. Animal feed; 88. Other (specify) b13. What is the total size of the land of your own you can cultivate? Quantity …………..Unit………….. 1. Acres; 2. Square Meters; 3. Square Yards; 4. Hectares b14. Do you rent land for farming? 1. Yes, 0. No Season 1 (March – May) 2022 Current season 1 (July – Sept) 2022 a) Quantity b)Unit a)Quantity b)Unit b15. What is the total size of the land under sweetpotato in the current season? b16. Size of your own land under sweetpotato in b17. If Yes to Qb14, Size of rented land under sweetpotato in: Code A: 1=Hectares; 2 = Acres(paces); 3 = Garden; 4 = Apuleju; 5 = Square Meters; 6 = Square Yards b21 Who in the household is involved in..... 1- man 2- woman 3 - both 4 - collective household 5- other (specify__) b21a. Management of sweetpotato plot from planting to harvesting b21b. Decision making on the sourcing of the sweetpotato planting materials 42 b21c. If b12 is 2, 3, 4, or 6, the sale of sweetpotato roots b21d. If b12 is 2, 3, 4, or 6, in charge of revenues gained from sweetpotato sales Man: means only the man in the household is involved. Woman: means only the woman in the household is involved. Both: means both the man and the woman in the household are involved. Collective household: if more household members than the man and woman are involved. Other: Any other person involved who is NOT a household member, family member or the landowner. PART C: ACCESS TO SWEETPOTATO VINES c1. Where do you NORMALLY obtain your sweetpotato vines/planting materials from? 1. Recycled vines/plants from previous crop; 2. Obtained vines/plants from neighbours/family; 3. From local markets and local stores; 4. From extension workers or projects/NGO; 5. Directly from National Sweetpotato Program (NSP); 6. From vine multipliers; 7. Schools; 8. Bought from market; 9. Farmer Field Schools; 10. Fellow Farmer; 11. Politicians; 12. Processors; 13. Religious (Church/Mosque); 88. Other (specify) c2a. Are you aware of a vine multiplier around your area? 1. Yes ,0. No c2b. If Yes Qc2a, what is the distance from your home to the nearest vine multiplier? …… walking minutes. c2c. If Yes Qc2a, How much (in UGX) do you spend to move from your home to the nearest vine multiplier using boda boda? c3a. Have you ever purchased sweetpotato vines to plant? 1. Yes ,0. No c3b. If Yes to Qc3a, how often have you purchased vines for your sweetpotato plots? 1. Every season; 2. Every other season; 3. Once in a while; 88. Other (specify) c3c. If Yes to Qc3a, on average, what quantity of vines do you normally purchase from this source? i) Quantity ……… ii) Unit of vine purchase 1=Bundles of 50 vines; 2=Bundles of 100 vines; Bundles of 200 vines; 4 = Half a bag (500 vines); 5 = Full bag (30kg – 1000 vines) c4. What is the price (UGX) of a 30-kg bag containing approx. 1000 vines…………………….. c5a. Are sweetpotato vines sold in your local markets? 1. Yes ,0. No c5b. If Yes Qc5a, how long will it take you to move from your home to this local market? ___ walking minutes. c5b_1. How much (in UGX) will you spend to move from your home to this local market (using a boda boda)? c5c. Have you ever wanted to grow a particular variety of sweetpotato and you failed the get the vines? (1=Yes; 0=No) c5d. Why couldn't you get the desired variety? 1=No money to buy; 2 = No vine multiplier; 3 = Vine multiplier didn't have it/failed to multiply it; 4 = Lost my own seed/planting material; 5= Don't know how to conserve vine; 88= Other (Specify) PART D: ACCESS TO AGRICULTURAL INFORMATION d1. Where do you normally get information on the quality and availability of seeds/planting materials of the crops that you grow? 1 = Community Leaders; 2 = Farmer groups/associations; 3 = Fellow farmers; 4 = Agricultural extension agent; 5 = Input dealers; 6 = Local Leaders; 7 = NAADS; 8 = NGO projects; 9 = Opinion leaders; 10 = Processors; 11 = 43 Religious Leaders; 12 = Researchers; 13 = Traders; 14 = Vine multipliers; 15 = Radio spots/adverts; 16 = Local agro-vet stores; 17 = Neighbours, family members or friends; 88 = Others (Specify); 999 = None d2a. Did you get information from any of these sources this year (2022 April to date)? 1=Yes, 0=No d2b. If Yes to Qd2a, from which one(s)? [use codes in Q d1]_____________________ (1 = Community Leaders; 2 = Farmer groups/associations; 3 = Fellow farmers; 4 = Agricultural extension agent; 5 = Input dealers; 6 = Local Leaders; 7 = NAADS; 8 = NGO projects; 9 = Opinion leaders; 10 = Processors; 11 = Religious Leaders; 12 = Researchers; 13 = Traders; 14 = Vine multipliers; 15 = Radio spots/adverts; 16 = Local agro-vet stores; 17 = Neighbours,family members or friends; 88 = Others (Specify); 999 = None) d3a. Do you own a functioning mobile phone? 1. Yes, 0. No d3b. If Yes to Qd3a, what type of mobile phone do you have? 0. Basic (call/voice only); 1. Smartphone (Voice & internet); 3. Landline d3c. If Yes to Qd3a, do you use your phone to do the following (select all that applies) 1 = Read text messages (SMS); 2 = Send text messages (SMS); 3 = Use WhatsApp messenger app; 4 = Use other social media platforms e =g = Facebook; 99 = Not Applicable) d3d. If Yes to Qd3a, have you ever used your phone to receive/access agricultural information? 1 = Yes ,0 = No, 99 = NA d4. If Yes Qd3d; If you have used your phone to get agricultural information, what type was the information? 1 = Type of variety; 2 = Source of planting material; 3 = Attributes of varieties; 4 = Where to sell output/roots; 5 = Pests and disease control/management; 6 = How to plant the seed/planting material (Agronomy); 7 = How to manage the variety in the field (management); 8 = Harvest related information; 9 = Post-harvest handling; 10 = Crop utilization; 11 = Market prices; 88 = Others (Specify) d5. If you have received agricultural information via your phone, how frequently do you do so? 1 = Daily; 2 = About twice a week; 3 = Once a week; 4 = Every two weeks; 5 = Once a month; 6 = Once every season(3months); 7 = Once a year; 8 = Once in lifetime; 99 = Not Applicable d6. What is the distance from your home to the nearest local agricultural inputs market in walking minutes? ……………………….. [If No to Qid6] PART E1: GENERAL NOLSTAGIA [Luganda: okuyaayaanira; Ateso: aiyitonor; Luo: para] I will now ask you some questions about how you feel about your past. We will be talking about your sentimental longing for the past, also called nostalgia. We will use a sliding scale from 1 to 7. [Enumerator: Explain how the sliding scale works using the demo card] e1. How valuable is nostalgia for you? 1 2 3 4 5 6 7 [Not at all<<<< >>>> very much] e2. How important is it for you to bring to mind nostalgic experiences? 1 2 3 4 5 6 7 [Not at all<<< >>> very much] 44 e3. How significant is it for you to feel nostalgic? 1 2 3 4 5 6 7 [Not at all<<<< >>>> very much] e4. How prone are you to feeling nostalgic? 1 2 3 4 5 6 7 [Not at all <<< >>>>very much] e5. How often do you experience nostalgia? 1 2 3 4 5 6 7 [Very rarely <<<< >>>Very frequently] e6. Generally speaking, how often do you bring to mind nostalgic experiences? 1 2 3 4 5 6 7 [Very rarely <<< >>>Very frequently] e7. Specifically, how often do you bring to mind nostalgic experiences? [Please check one only] 1. At least once a day; 2. Three to four times a week; 3. Approximately twice a week; 4. Approximately once a week; 5. Once or twice a month; 6. Once every couple of months; 7. Once or twice a year; 88=Others(specify) PART E2: NOSTALGIA & AGRICULTURE In reference to traditional agriculture and how our forefathers lived, please state the level to which you agree or disagree with the following statements [Use scale: 1-Strongly Disagree; 2-Disagree; 3. Neither Disagree nor Agree; 4-Agree; 5-Strongly Agree] e8. Farming helps me feel a connection to my parents/ancestors e9. I have a fondness for the agricultural production practices my parents/ancestors used e10. I have a fondness for the crops my parents/ancestors used to grow [If No to Qid6] PART F: LOSS AVERSION In this section, we will now present to you statements that could describe you or your nature. Please use the scale below to indicate the extent to which you agree or disagree with each statement, 1 = Strongly Disagree; 2 = Disagree; 3 = Neither Disagree nor Agree; 4 = Agree; 5 = Strongly Agree. [Enumerator: Please use the scale card] f1. I worry too much over something that really doesn’t matter. f2. I get in a state of tension or turmoil as I think over my recent concerns and interests. f3. I have disturbing thoughts. f4. I take disappointments so keenly that I can’t put them out of my mind. f5. It’s horrible if your boss thinks less of you than you’re really worth. f6. I get easily attached to material things (phone/radio/TV, my furniture, ..) f7. I would have problems with having to live in a smaller house. f8. I think eventually I could cope with losing the ability to walk. f9. I feel awful if someone talks bad about me behind my back. f10. I go crazy if I lose something, even when it’s not that important. f11. I think I could cope with losing all my belongings in a [house] fire. f12. Once I’ve acquired a position in the company [farmer/social group], I wouldn’t want to take a step back. f13. Losing your house to a fire is bad, but I would manage. f14. I would hate it if a colleague [fellow farmer] thought that I’m not as good in my job [farming] now as I was before. f15. I would feel very down if I got fired [in a farmer/social group], even if I know I will find a similar position in another. f16. I don’t care what people would think if I was suddenly unemployed [became poorer than I am]. 45 f17. I would feel very tense if the farmer/social group changed our way of operation. f18. I would have no problem accepting a [casual farm] job/employment / lejaleja that has less pay [wage] than my previous/current one. f19. I would be okay with trading my current bike/phone/radio/TV for a cheaper one. PART G1: VINE PURCHASE g1a. Were you aware that vines were being sold in your village? 1=Yes 0=No g1b. Were you aware of vine sales in another village? 1=Yes 0=No g1c. If Yes to g1b, Which village(s) is that? ___ g2. How did you get to know that there were vines being sold in your village? [Enumerator – DO NOT prompt] 1-I had visited the place/farmer for other reasons and noted that vines were being sold there 2-I was notified by co-villagers/friends/relatives that the vines were being sold at this place. 3-I received a text message that notified me that the vines were sold at this place. 4- I was passing by and got attracted by the sales stand structure that was at this persons place and later found out that they sold vines 5-I was attracted by the poster 6-I learned from agricultural officer 7-I learned from LC1/Parish chief or other local administrator 88-Others (Specify …) g3. Did you buy any of the improved vines that were being sold in your village in the last 2 seasons? 1=Yes 0=No g4. If Yes, when did you buy the vines? 1=April/May season; 2=July/September season; 3= Both seasons g5. If Yes to g1b. Did you buy the improved vines from a village outside your village? [Enumerator: not conditional on whether vines were bought from own village]1=Yes 0=No g6. How long (in walking minutes) does it take you to move from your home to the place where you bought the vines from?_________ g7. What is your relationship with the person that so