Impact Evaluation Associated with the Bangladesh AVC Project Midline Report May 2018 Alan de Brauw, Berber Kramer, Hazel Malapit, Elena Martinez, and Mike Murphy 1 Impact Evaluation Associated with the Bangladesh AVC Project: Midline Report Evaluation team: Alan de Brauw, Berber Kramer, Hazel Malapit, Elena Martinez, and Mike Murphy Submitted by the International Food Policy Research Institute (IFPRI) to the U.S. Agency for International Development’s Bureau for Food Security Agreement No. AID-BFS-IO-14-00002 Washington, DC May 31, 2018 2 CONTENTS Executive Summary 1 1 Background 9 2 Interventions 10 2.1 NGO Trainings 10 2.2 NAAFCO Promotions 11 2.3 Research Questions 12 2.3.1 Understanding Adoption 12 Key Question: Can targeting financial and informational constraints improve take-up of improved inputs and practices and reduce knowledge gaps? 12 2.3.2 Trust in the Value Chain 12 Key Question: How does trust between smallholders and input traders affect transactional behavior within the market system, and how does trust influence smallholder productivity? 12 2.3.3 Labor Scarcity and Women’s Roles in Agriculture 13 Key Question: As mobility is increasing and household labor availability is declining, is the role that women are playing in jute production changing? 13 2.3.4 Further Research Questions 13 3 Data Collection 13 3.1 Sample Selection 13 3.2 Survey Instruments 15 3.2.1 Household Listing 15 3.2.2 Output Purchasers Listing 15 3.2.3 Market Surveys 15 3.2.4 Household Surveys 16 3.2.5 Women’s Empowerment in Agriculture Index (WEAI) 17 3.2.6 Measuring Expectations 18 3.2.7 Assessing Input Knowledge 18 3.2.8 Trust Experiment 19 3.2.9 Photobooks 20 3.3 Enumeration Team and Training 20 3.4 Fieldwork 21 3.4.1 Household Interviews 21 3.4.2 Trust Game Payouts and GPS Verification Form 21 3.5 Data capture and Monitoring 21 4 Sample Characteristics 22 4.1 Survey Sample 22 4.2 Attrition 22 4.3 Respondent Characteristics 23 3 4.4 Treatment Assignment 24 4.5 Treatment Compliance 25 4.6 Treatment Balance 27 5 Analysis 30 5.1 Understanding adoption 30 5.1.1 Farmer Yield Expectations 31 5.1.2 Knowledge about Inputs 33 5.1.3 Improved Practices 35 5.1.4 Cultivation Decisions 37 5.1.5 Production outcomes 38 5.2 Understanding Trust in the Value Chain 42 5.2.1 Relationships between farmers and input sellers 42 5.2.2 Trust Experiment 43 5.2.3 Heterogeneity by Trust 45 5.3 Migration, Labor Scarcity and Women’s Role in Agriculture 47 5.3.1 Women’s Empowerment in Agriculture Index (WEAI) 47 5.3.2 Women’s Role in Jute Production 49 5.3.3 Migration and Labor Demand 51 5.3.4 Labor Scarcity and Female Labor Usage 52 6 Conclusion and Next steps 54 6.1 Conclusions 54 6.2 Next Steps 55 7 References 57 8 Appendices 58 8.1 Estimation Strategy (Understanding Adoption) 58 8.2 Summary Statistics 59 8.3 Comparison of WEAI & Pro-WEAI Indicators 60 8.4 Design Protocol 61 8.5 Midline Questionnaire 89 4 Executive Summary Project Background The USAID-funded Bangladesh Agricultural Value Chains (AVC) project implemented by Development Alternatives International (DAI) is focused on improving food security and nutrition by strengthening agricultural value chains in the Feed the Future Zone of Influence in Bangladesh. AVC follows a market systems approach, working with private sector firms to co-develop commercial strategies to more effectively engage and incentivize value chain actors, including smallholder farmers, input suppliers, output buyers, and service providers. The impact evaluation associated with the AVC project neither attempts to measure impacts of all the market systems interventions being conducted, nor all the value chains with which AVC works. Market systems interventions are necessarily too complex and adaptive to attempt to measure impacts of the interventions as a whole; such a measurement would at best provide an average treatment effect on all the adaptive interventions at once. However, particularly with rain-fed agriculture, some interventions must take place at a specific point in time each year, directly targeting large numbers of farmers in a uniform way, and farmers would theoretically adapt to these interventions in the following agricultural season, or would have to wait a year to do so. In this context, in this report the impact evaluation focuses on interventions that took place in the jute value chain in early 2016.1 This report explores research questions in three primary areas: ● Understanding adoption Can targeting financial and informational constraints improve take-up of improved inputs and practices and reduce knowledge gaps? How does take-up affect outcomes at the farm-level? ● The Role of Trust in the Value Chain USAID has identified trust as a key influencer of interactions between market actors. How does trust between smallholders and input traders affect transactional behavior within the market system, and how does trust influence smallholder productivity? ● Empowerment, Labor Availability & Women’s Role in Agriculture As mobility is increasing and household labor availability is declining, is the role that women are playing in jute production changing? The analysis takes a gender sensitive approach through the collection of both gender disaggregated data and household level outcomes. Data Sources The report relies on two main data sources. First, it uses a portion of the baseline survey conducted for the impact evaluation with a representative sample of 20 jute farming households per village in 50 villages from Faridpur, Jhenaidah, Madaripur and Narail districts. The baseline survey covered multiple topics, including modules making up the Women’s Empowerment in Agriculture Index (WEAI). The WEAI modules are completed independently and privately by the main male and female respondents in the household, allowing the team to complement aggregate data on household agricultural outcomes with specific data on topics such as male and female roles in household decision-making and self-perception. The midline survey was collected in the same households, and added 500 jute farming households (10 in each village) representative of jute farming households with more land. The latter survey includes innovations on standard data collection, including a trust game derived from the behavioral economics literature, to improve measures of trust from the baseline. 1 The endline report on the impact evaluation will compare impacts of interventions in the jute value chain with impacts from interventions in the mung bean value chain. However, activities in mung bean producing areas were not at a scale to measure impacts at the time of midline data collection. 5 Understanding Adoption Interventions At midline, the impact evaluation measured the effects of two interventions carried out in the study area: NGO Trainings and NAAFCO Promotions. The NGO trainings were a remnant from the previous focus of the AVC, and followed a standard extension model, with local experts providing training to farmers on best practices for improving productivity and output quality. The trainings aimed at generally improving knowledge about the jute production process, including seeding techniques, fertilizing, and dealing with weeds and pests. The promotions were organized with NAAFCO, a private sector input provider. Farmers in this treatment group were exposed to fairs promoting NAAFCO products, describing why their products are useful relative to other jute inputs. The fairs also included a series of raffles through which local smallholders could win discount coupons to purchase NPKS, a fertilizer product designed to be optimal for high-quality jute production. The two interventions were cross-randomized at the village level, so groups intended to be exposed to the NGO trainings, the NAAFCO promotions, and both can be examined against a control group with no concern of confounding factors affecting the analysis. Outcomes Both interventions were associated with changes in farmers’ subjective expectations around improved inputs, and positive gains in scores on a series of knowledge questions. The NGO Trainings enhanced knowledge around inputs and practices not directly related to NPKS, while NAAFCO promotions were associated with gains in knowledge around practices and outcomes associated with improved fertilizer use. However, the evaluation did not find increases in measures of average productivity or gross margins per hectare. Measured gains in productivity were small and as agricultural yields are highly variable, they were not statistically significant. The NAAFCO promotions treatment resulted in an increase in take-up on NPKS fertilizer among treatment households, suggesting that private sector models for forms of input or extension service provision could be considered within the context of a market systems approach to foster adoption of new agricultural technologies or practices (Table 1). Table 1- Effect of Interventions on Key Outcomes Intervention Any NPKS Knowledge? Fertilizer Knowledge Score Other Input Knowledge Score Used NPKS Used IPM Jute yield (log) Trainings Only 0.032 (0.048) 0.096 (0.105) 0.208* (0.117) -0.029 (0.043) 0.133*** (0.040) 0.03 (0.04) Promotions Only 0.081* (0.047) 0.225* (0.115) 0.012 (0.095) 0.096** (0.040) 0.209*** (0.040) 0.05 (0.04) Training & Promotions 0.217*** (0.043) 0.271*** (0.080) 0.074 (0.092) 0.144*** (0.029) 0.092** (0.036) 0.01 (0.03) Mean, Control Group 0.132 -0.137 -0.037 0.016 0.311 7.58 Note: Table presents regression coefficients (estimated magnitudes of effects), with standard errors (precision of estimates) in parentheses. Stars indicate statistical significance: *10% level; **5% level; ***1% level (for example ** indicates a less than 5% probability that the effect would be observed through random chance). In sum, the NAAFCO promotions increase narrow knowledge of NPKS as an input, and drive the adoption of NPKS, which was quite low among the control group at baseline (less than 2 percent). The evaluation finds that knowledge gains among farmers in villages exposed to NGO trainings are broader than among the control villages. Their knowledge about input use beyond fertilizer increases, reflecting the breadth of training topics covered. Among the promotions group, there are positive changes in fertilizer knowledge relative to the control group, which are not reflected among the NGO trainings group. Therefore, the findings suggest that each type of training was at least somewhat effective at what it was meant to do; the NGO trainings increased knowledge broadly about the jute production process, while the promotions group increased knowledge specifically about fertilizer. As such gains 6 Figure 1- Labor Allocation by Gender and Stage of Jute Crop Production would be the goal of a private sector firm operating promotions in this context, the promotions appear to have been successful in doing what they set out to do. That said, there is no evidence of statistically significant differences in productivity attributable to either treatment. However, the positive coefficients suggest that there could be a potential effect that cannot be identified due to a lack of statistical power. The Role of Trust in the Value Chain To study trust in the value chain, the survey instrument incorporated subjective questions on trust between farmers and input retailers, and included an incentivized trust game conducted with farmers and retailers. This game follows a standard experimental design, with some procedural modifications to facilitate implementation in the context of a large-scale household survey (Kramer et al., 2017). In the game, farmers were shown pictures of six local retailers, and were asked for each retailer how much of 150 Taka to keep for himself, and how much to send to the retailer; with the understanding that the research team would triple the amount sent to the retailer, who would in turn decide on an amount to send back to the farmer. For each farmer and for each retailer, the research team selected one game for which the result was paid out, creating real stakes in each game. In terms of the directly reported survey measure (reported on a 1-10 scale), farmers generally rate input sellers highly (the average score is 8.9 in the midline), and in the incentivized experimental trust game, on average farmers sent retailers more than half of their initial endowment of 150 Taka. Using this experimental measure also provided an interesting finding relating to productivity: farmers who sent more than the median amount among the NAAFCO promotions group produce 15 percent more jute than the control group, while the same result is not present among lower trust farmers. The implication is that interventions that can build trust in their products via trusted input retailers, potentially through improving the market systems, can have productivity effects. Empowerment, Labor Availability & Women’s Role in Agriculture Analysis of the WEAI data show that empowerment is relatively low in the sample; according to the data, only 38 percent of women achieve empowerment in agriculture. The two primary drivers of disempowerment in the sample were autonomy in production and workload. The next part of the report, focusing on gender differences in labor allocation in jute production, identified during which phases of the production cycle women are involved. It finds that women—whether in the household or hired—tend to participate in post-harvesting tasks but not in production or marketing (Figure 1). As migration is increasing, the next step was to examine how tasks are changing in households facing more labor scarcity. Although women could, in theory, take on more tasks in jute production to cope with labor shortages, the report finds that in households reporting that labor became less available between baseline and midline, households became more likely to hire female labor, rather than using more female household labor in the production process. Among women within households reporting more labor scarcity, the only change in labor use is that they spend more time within tasks they were already doing. Further, while the daily wage increased on average for male laborers, there is no evidence that wages are increasing similarly among women. Because of the strict gender differentiation of tasks, it may be that opportunities for women remain limited within jute production. Though increased labor scarcity does appear to be changing the composition of labor in jute production, it is changing towards the use of more hired labor rather than changing gender roles within production. 7 Next Steps The impact evaluation team is completing a qualitative report, which will focus specifically on questions related to women’s role in jute production, as well as an endline report based on data collected in April and May, 2018. The data collection will contain a detailed migration module to better explore its potential role in changing women’s roles in agriculture, and another trust game that will test whether associating specific brands (NAAFCO and Partex) with input retailers selling those brands improves trust. The endline survey will be conducted in both the jute and mung bean growing areas. A final report covering the entire impact evaluation will incorporate data from the baseline, the midline, the qualitative phase, and the endline survey. 8 1 BACKGROUND Funded under the Feed the Future (FTF) Initiative, the Bangladesh Agricultural Value Chains (AVC) project being implemented by Development Alternatives International (DAI) is working to improve food security and nutrition through strengthened agricultural value chains. Agricultural value chains in Bangladesh are typically fragmented, and lack investment, inclusion of vulnerable groups, and critical linkages. AVC is focusing on a portfolio of food and non-food crops to facilitate growth in the agricultural sector. AVC is working on improving value chains in six classes of food crops (pulses, tomato, mango, ground nuts, potatoes, and a summer vegetable basket) and two classes of non-food crops (natural fibers and floriculture). The geographic focus for the AVC project is 20 southern districts in Barisal, Dhaka, and Khulna Divisions. Some of the specific value chain interventions are more focused, since not all the focus crops are grown in all project districts. The AVC project has four main components, which are called intermediate results en route to improved food security through stronger agricultural value chains. These intermediate results are: ● Intermediate Result 1: Sustainable, diversified agricultural productivity increased ● Intermediate Result 2: Agricultural market systems strengthened ● Intermediate Result 3: In7novation and value chain upgrading increased ● Intermediate Result 4: Local capacities and systems strengthened In addition, the project has several elements that are cross-cutting in each of the main components. These elements include: ● Nutritional practices improved; ● Effective gender integration and youth participation enhanced; and ● Environmental sustainability and resilience to climate change strengthened. In mid-2015, AVC began to change its focus from primarily contracting smallholder farmer trainings to a market systems approach, in which they work with selected private sector firms to co-develop commercial strategies that more effectively engage and incentivize value chain actors, including smallholder farmers, input suppliers, output buyers, and service providers. The goal of these strategies is to increase transactions, build trust, build industry networks, strengthen market systems, and enable increased investment, competition, and positive development outcomes for communities in FTF target areas. As such, the impact evaluation aims to generate useful insights on constraints to market development in selected value chains with generalizability to other value chains in low-income countries and Bangladesh specifically. The evaluation will go beyond just trying to understand the impacts of some of the specific, selected AVC interventions on improving agricultural productivity. First, the analysis will evaluate whether and why the “mass market” approach facilitated by AVC, with input suppliers moving from a wholesale to a retail approach for distributing their inputs, can improve outcomes for producers, and enhance mutual trust among both input suppliers and smallholder farmers; to that end, the evaluation team are designing experiments related to repeated interactions, quality signals, and trust to play with sampled market participants. To understand the main market constraints to the use of high quality inputs in production, the team will examine whether specific interventions at the farmer level help improve the productivity and quality of specific products and assess what value chain actors are willing to pay for high-quality products in order to learn more about where the market structure fails to incentivize quality. In studying all these aspects of the AVC project, the analysis will take a gender sensitive approach by collecting gender disaggregated data as well as household level outcomes. In addition to incorporating individual level questions, the main survey instruments also incorporate modules from the Women’s Empowerment in Agriculture Index (WEAI) instrument. These sections are completed independently and privately by the main male and female respondents in the household, allowing the team to complement aggregate 9 data on household agricultural outcomes with specific data on topics such as male and female roles in household decision-making and self-perception. The analysis will also explore patterns of participation in the markets studied by gender, and disaggregate the results of the associated experiments by gender. The cross-cutting component of the AVC project on gender will be an important outcome for the impact evaluation. The impact evaluation will not assess the impacts of all the market system interventions being conducted by AVC. Rather, it will focus on interventions pertaining to one food crop and one non-food crop. In the selection of the food crop, the team only considered crops that are inherently nutritious. Mung beans have nutritional value in terms of iron content. Iron deficiency anemia remains a major health problem in Bangladesh, whereas micronutrients covered by other crops that are potentially being studied by AVC (e.g. orange sweet potato and mango, which are rich in vitamin A) are not as large of a deficiency. Moreover, mung beans are grown by a large number of farmers within a specific part of the FTF Zone of Influence (ZOI), which will enable the impact evaluation to sample a large and relatively representative group of farmers. Finally, mung beans have a reasonably simple value chain (that can therefore be traced more easily), as mung beans are roasted and sold as an individual product; therefore, mung beans are the most appropriate choice for a value chain product for exploration in this impact evaluation. For the non-food crop, jute was selected because the crop is also grown by a significant proportion of farmers in the FTF ZOI and appears to have similar traits to the mung bean value chain. As trust issues between input sellers, farmers and processors appear to exist in both the mung bean and jute value chains, and farmers do not tend to use certified seeds and other improved inputs in both value chains, there are similar issues that exist in both value chains, allowing for an interesting comparison. That said, jute has a long history as a particularly politicized crop in Bangladesh (e.g. Ali, 2012), and as such the possibilities for expansion of the crop may be limited relative to those for mung beans. Other non-food crops were also considered but the team concluded that these are not viable for an impact evaluation; in the case of cut flowers, the AVC was quite advanced at the time the impact evaluation began, but the number of farmers growing cut flowers in Bangladesh overall is small, so it would be difficult to identify a potential comparison group growing or potentially growing cut flowers for market. Even if a comparison group were identified, it would be difficult to identify a large enough treatment group to make meaningful statistical comparisons. In the case of coir, the value chain is quite underdeveloped, and as such any interventions are quite prospective. Hence, jute is the remaining possibility. This report focuses on the jute value chain rather than both value chains. The interventions in the mung bean value chain which began under the market systems approach during the calendar year 2016 were too small in scope to include in an impact evaluation, but by late 2017 they will cover a large enough number of farmers to attempt to detect impacts following on our research questions. The following sections focus specifically on analyzing data collected around the interventions which took place in the jute growing areas in 2016. 2 INTERVENTIONS 2.1 NGO TRAININGS The NGO trainings were a series of informational training sessions organized in February 2016 and targeted training 4,000 farmers per district in four districts (Faridpur, Jhenaidah, Madripur, and Narail). The trainings were undertaken by four AVC subcontractors, non-governmental organizations (NGOs), who were competitively selected for their track records of successful training in agricultural topics.2 Trainings provided information on a range of best practices relating to jute including improved and certified seed varieties, optimal fertilizer types and application practices, pest identification and management, modern retting techniques for fiber separation, and grading of jute fiber quality. The 2 The four subcontractors were Gono Unnayan Prochesta (GUP); Society Development Committee (SDC); Sheba Manab Kallyan Kendra (SMKK); and Prove Society. 10 goal of the trainings was to provide farmers with information enabling them to produce higher quality jute, and command higher prices. The extension trainings followed a standard outreach model, with AVC contracting outside organizations to provide training to farmer groups. Representatives travelled to each village in advance of each session to inform farmers of the coming training and register those interested in participating. Participants were required to own at least a small area of cultivated land and have experience growing jute. The goal of these registration criteria was to ensure trainings were focused on farmers most likely to benefit from the information provided, and to introduce participants to peer farmers in their area prior to beginning the trainings. Following registration, training groups were formed in each area, each consisting of approximately 30 farmers living close to one another. Each group met with a representative of an AVC subcontractor for two full days. Participants were not compensated for attending, but were provided with lunch, tea, snacks and transportation to and from the training. The training leaders were agriculture experts such as former government agriculture officers or freelance consultants, hired by the subcontractor via a competitive interview process. In addition to the informational component, AVC or its subcontractors also arranged linkage meetings as part of, or shortly after the trainings. The goal of these sessions was to reduce informational gaps within the market system for farmers at various points of the value chain. For these sessions, a public or private sector representative would be invited, such as a government official or output purchaser. Public sector representatives would often discuss legislation related to the crop in question, and explain what sorts of institutions the Bangladesh government already has in place that farmers could take advantage of. Private sector representatives typically explained what goods and/or services their organization provides that could be of interest to the farmers. For example, they would explain the quality and grading of the types of jute they require, and make suggestions as to how farmers can improve their output quality to meet these standards. 2.2 NAAFCO PROMOTIONS In addition to the informational intervention provided by the group trainings, a second treatment intervention was carried out by USAID|AVC’s private sector partner NAAFCO to organize a series of input fairs in which suppliers’ representatives would organize crop clinics to provide farmers with additional extension advice, give information on NAAFCO products, including NPKS fertilizer, and provide discounts on NPKS fertilizer. NPKS improves on existing fertilizers because it offers a composition of Nitrogen (N), Phosphorus (P), Potassium (K) and Sulfur (S) that is optimal for jute production (Khanom, Hossain, & Hossain, 2012). The latter intervention follows a market systems approach, targeting input sellers as well as farmers, and adjusting the interventions as seen fit to improve the efficiency of the entire market for jute fertilizer. For the farmer, the aim was to provide both tailored input information (via the crop clinics) and access to promotional discounts, reducing the entry cost to trying NPKS. The overall objective is to attempt to enhance demand later, but also to help farmers understand what makes a fertilizer worth purchasing, including assurances it is not counterfeit. For the input seller, the fairs were designed to not only foster interactions with local smallholder farmers but to provide NAAFCO with insights into the business case for targeting farmers with promotional marketing. From the research perspective, it was important to isolate a component of the market systems intervention that would or could only take place once a year. Consequently, the promotional component was designed by AVC and IFPRI in collaboration with NAAFCO. During the fairs, a raffle was implemented in which farmers would be eligible to win discount coupons for a 25kg pack of NPKS fertilizer for jute cultivation: first prize winners would receive an 80% discount; second prize winners a 50% discount; and third prize winners a 20% discount. The raffle was designed such 11 that approximately 50% of participants would win a discount coupon.3 Discount coupons could be used at local input sellers that were already part of NAAFCO’s distribution network. This design enables the impact evaluation team to evaluate both the effect of lowering the costs associated with taking up NPKS fertilizer, as well as to assess the level of take-up associated with each of the respective discounts. In addition to enabling the comparison of adoption rates between those who did and did not receive a discount, by incorporating variation in the levels of discount the intervention could potentially provide important information to traders on potential marketing strategies for improved products. If, for example, a low discount level induced relatively large take-up, this could indicate high potential demand for improved products. Conversely, if a high discount rate was required to induce take-up, it could be the case that promotional marketing is not a cost-effective strategy for traders to adopt. This design also can help to understand if the one-time discounts led to higher use of NPKS more generally over a longer period, which will be explored in subsequent reports. The objective of the intervention was to deliver extension advice in a manner which facilitated interactions between farmers and input vendors, and in the process build trust between sellers and clients. In addition to assessing the effect of the respective treatments (information and promotions), the interventions were intended to foster interactions between different actors within the market system. As part of the informational intervention, the farmer registration process was intended to bring together a group of experienced jute farmers, helping to encourage knowledge sharing among peers related to best practices. Similarly, the linkage meetings were intended to encourage information sharing between different value chain actors who might not typically communicate directly. The input promotions fairs followed a similar approach. Advice was provided both on specific products, and in response to farmers’ specific queries through the crop clinics. 2.3 RESEARCH QUESTIONS 2.3.1 Understanding Adoption Key Question: Can targeting financial and informational constraints improve take-up of improved inputs and practices and reduce knowledge gaps? A primary focus of the evaluation is to assess the extent to which the interventions were successful in promoting input adoption among smallholder farmers. The goal is to understand whether constraints related to price, addressed by the discounts through the fairs, or by knowledge through trainings, affect farmer adoption of improved inputs. By testing the NGO training model alongside and in addition to fairs, the impacts both on input use and then agricultural production can be measured. In line with the research recommendations set out by USAID’s Leveraging Economic Opportunities (LEO) project, this approach can evaluate the extent to which a market systems facilitation approach leads to increased take-up of technologies, in turn resulting in improved outcomes across the value chain. In the case of the promotions treatment, the analysis will assess the efficacy of a private sector driven input delivery model and provide lessons learned for future program development. 2.3.2 Trust in the Value Chain Key Question: How does trust between smallholders and input traders affect transactional behavior within the market system, and how does trust influence smallholder productivity? Lack of trust between smallholders and input suppliers has been identified as a key influencer of market actor behavior, leading to sub-optimal outcomes (USAID, 2015). The main research question here is whether trust issues 3 Some complaints were received by AVC, and relayed to the impact evaluation team, that some participant farmers were unhappy about the discounts because they were inadequate to cover the participation cost. However, note that no farmers were coerced to participate in the fairs, and this report demonstrates some average benefits to farmers in villages that were exposed to the fairs. 12 either within smallholder farmers or between smallholders and traders affect the transactional behavior of farmers, and if so whether market systems programming can be designed to overcome those trust issues. An innovative feature of the data collected as part of the evaluation is that it contains measures of trust between buyers and input sellers obtained from both parties. Farmers and input sellers both participated in an incentivized experiment designed to elicit levels of trust between both groups. This analysis, along with additional lab-in-the-field experiments conducted between farmers and input sellers, provides unique insights into patterns of trust and reciprocity between these groups of market actors, and how these patterns affect farmers’ transactional behavior. 2.3.3 Labor Scarcity and Women’s Roles in Agriculture Key Question: As mobility is increasing and household labor availability is declining, is the role that women are playing in jute production changing? A key concern for evaluating market systems interventions is to understand how they affect and interact with existing customs and norms, and influence outcomes for sub-populations within the sample. In the context of Bangladesh where urbanization is proceeding rapidly (Ellis & Roberts, 2016), the population remaining in rural areas is changing, and the consequences of urbanization for rural labor market outcomes, by gender, have not been widely studied. In this report, the analysis explores the role that women are playing the jute value chain as labor scarcity increases through migration. 2.3.4 Further Research Questions In this report, the focus will be on progressing towards answers related to these first three research questions proposed. The fourth question, which deals with trying to disentangle impacts of different training modalities, will be studied in the endline report, to be completed at the end of FY2018. The fifth question, related to consumer acceptance of processed foods, was conditional on AVC making enough progress on processed mung beans, which did not occur. 3 DATA COLLECTION 3.1 SAMPLE SELECTION For the jute intervention, an original sampling frame was developed by IFPRI in collaboration with DAI in January 2016. The sample frame identified 59 potential jute producing villages for potential expansion of ongoing interventions, of which 50 were selected for study inclusion. The 50 villages were selected to exclude sub-districts in which only one or two villages were included, and to prevent the selection of villages very close to one another that could cause spillover effects, affecting the ability to draw inferences about intervention effects. The jute villages selected for the study are in four districts: Faridpur, Jhenaidah, Madripur, and Narail. In line with the study design, treatment status was assigned by village. Given that there were two treatments, each village could therefore fall into one of four categories: receiving both training visits and direct input marketing; receiving training visits only; receiving direct input marketing only; or the control group, who received neither treatment (Figure 3.1). 13 Figure 3.1- Treatment Design For all villages, a household listing was carried out in early 2016 to develop a sample frame for each village; it is described in more detail in the next section. Using this data, the team applied two eligibility criteria for potential sample inclusion. First, households had to have either grown jute in the previous season, or intend to plant jute in the coming season. This criterion excluded non-agricultural households and ensured inclusion of households growing jute for market. Second, households owning more than 500 decimals (5 acres) of land were excluded, to ensure the sample was drawn from smallholders rather than larger commercial farmers. From the remaining households, the sample first selected all households with a female household head, defined as households in which the primary jute farmer was female, to maximize statistical power for gender related questions. Remaining households were selected at random from among those eligible to create a list of twenty households per village, with a median area of 66 decimals for jute cultivation. For each village, replacements were drawn at random from the remaining unselected households in case the chosen household was not available or refused to participate in the study, though in practice these were seldom used. Enumerators from DATA used this sample frame to conduct a baseline survey in 1,000 households between February and April 2016. The data collection team also conducted a market survey in each village, as well as a buyer pre￾screening to identify local value chain actors who purchased jute output from the farm households. Details on each of these survey instruments are presented in the next section. Following baseline data collection, AVC updated their selection criteria for program participants to focus on more experienced farmers with slightly larger holdings. To account for these farmers, the impact evaluation team developed an additional sampling framework to capture more of these participants. Within each sampled village, the impact evaluation team sampled an additional 10 households. Additional households were required to have at least ten years of jute farming experience, and have cultivated at least 66 decimals in the previous season. While these households were already eligible to receive the NGO trainings in villages with such trainings, logistical constraints prevented their inclusion in the initial raffle distribution in villages with promotions, and therefore a second round of raffles were organized for these farmers. As a result, additional households received comparable exposure to NGO trainings to the original sample. While baseline survey data is not available for these households, the nature of the analysis design still allows for statistical inference to be drawn between the respective treatment and control groups in each, since their ex ante characteristics are expected to be similar. In collaboration with IFPRI, the DATA field team conducted a second round of household surveys in February-March 2017. The survey included both those in the original sample who were interviewed at baseline (20 x 50 = 1,000 households) and the additional households from the second sample, not interviewed at baseline (10 x 50 = 500 households). The team also conducted interviews with local input sellers in each village (256 households) to better 14 understand how local input market systems function, and again conducted a market survey on local food prices. For ease of understanding, the remainder of the report will refer to each set of households in terms of these groups, where: - Group 1 households are farm households from the original sample, interviewed in both the 2016 and 2017 survey rounds; - Group 2 households are farm households from the additional sample, interviewed only in the 2017 survey round; and - Group 3 households are input seller households interviewed in the 2017 survey round. Unless otherwise stated, the analysis will consider the effects of the interventions on a combined sample of Group 1 and Group 2 households. As input sellers often provide services in multiple villages (belonging to different treatment categories), Group 3 households are not considered in the analysis in terms of treatment effects for the households themselves, but will be primarily used to explore the nature of linkages and interactions between farm households and other value chain actors to better understand the process of adoption by farmers. 3.2 SURVEY INSTRUMENTS All survey data collection was carried out by DATA staff using SurveyCTO, a computer-assisted personal interview (CAPI) program using Android tablet computers. All survey instruments were designed by IFPRI, with DATA providing translation to Bengali and advice on ensuring questions and response options were appropriate to the local context. The AVC monitoring and evaluation team was provided with draft survey forms early in the process for both the baseline and midline, and provided input on the baseline questionnaire. Survey programming was done by IFPRI. IFPRI carried out internal testing prior to trainings, and conducted additional testing in collaboration with DATA during trainings to check for errors and make adjustments to ensure translations were accurate and properly adapted. At baseline, additional support for survey programming was initially provided by ikapadata, a market research company based in Cape Town, South Africa. Both survey rounds included standard data validation features, automatic skip logic and preloading of household-specific information (such as names of household members, which were obtained during the listing exercise). 3.2.1 Household Listing The household listing was a very short survey which gathered basic information on households in each village. It collected demographic information on the head of each household (age, gender and identifying information), identified their spouse (if married or cohabiting), and whether the household had grown jute in the preceding agricultural season, or planned to grow it in the coming season. The listing was conducted on the first day of survey activities at baseline in each village, and was designed to allow enumerators to reach nearly all households in each village. Information from the listing was then preloaded into the baseline household survey, to allow enumerators to rapidly verify they were at the correct household when commencing each interview. 3.2.2 Output Purchasers Listing For each village, an output purchaser survey was conducted along with the baseline survey to identify the principal jute buyers in the area. A supervisor from the DATA team met with local officials, village leaders, or other individuals knowledgeable about jute sales in each village. After identifying at least one knowledgeable informant, the supervisor collected basic identifying data for each buyer. Names of local buyers were then included as a preloaded option list in the main baseline household survey, and used as choice options for relevant questions on jute sales. 3.2.3 Market Surveys Along with the baseline and midline household surveys, for each village a DATA supervisor carried out a food market survey at the principal market used by village residents. They gathered information on the availability and price of food items listed in the consumption module of the household survey. Price data from the market survey was then used to estimate household level consumption expenditures. 15 3.2.4 Household Surveys The baseline and midline household surveys were conducted in person by DATA enumerators and consisted of a common format of three forms: a main form focused on agricultural production, and a male and a female form to capture aspects of empowerment at the household level, consisting of modules developed for the Women’s Empowerment in Agriculture Index (WEAI). The main respondent was the primary jute farmer in the household, who also answered the individual form for their gender, while their spouse would answer the other individual form. 4 For a household in which the main jute farmer was male, that farmer would answer the main form and male form, while his spouse would answer the female form. The male and female form interviews were conducted by different interviewers, with no other household members present where possible. For input seller (Group 3) households at midline, a subset of the main questionnaire was asked, including the following modules: household identification; household roster; housing; household expenditures; beliefs; and consumption. Input sellers were also asked subsets of the modules on agricultural production and value chain interactions, with questions adapted to be relevant to the respondent. As with the other sample groups, modules were targeted by gender, though the main respondent (the input seller) was exclusively male. Table 3.1 presents a summary of modules included, by group. 5 4 As identified in the household listing exercise, though for both surveys the respondent could be updated if the member from the listing exercise was no longer the primary jute farmer in the household. 5 For example, rather than ask if the respondent purchased NPKS fertilizer in the previous season, they were asked whether they had sold it in the previous year. 16 Table 3.1- Household Survey Modules by Respondent Group, Survey Round Group 1 Group 2 Group 3 Module Description 2016 2017 2017 2017 A Household Identification X X X X B Roster X X X X C Agricultural Production X X X Jute Production X X X Production of Other Crops X X X Inputs and Labor Allocation X X X X Extension X X X X Technology Use X X X X Input Knowledge X X X D Value Chain Interactions X X X X Output purchasers X Input Sellers / Purchasers X X X E Housing X X X F Access to Capital* X X Assets X X X Credit X X X G Household Expenditures X X X X H WEAI X X X A-WEAI X Pro-WEAI X I Beliefs X X X X Expectations X X X X Risk Preferences X Trust X X X J Consumption X X X X Expenditures X X X X Food Security** X K Health X X X * Data on access to capital was collected at midline as part of the Pro-WEAI modules ** The food security module was removed for midline after <1% of households reported hunger at baseline For each section, enumerators aimed to have the most informed person in the household act as the respondent, hence the main form included questions on agricultural production and sales; input use; labor allocation and adoption of new technologies; and interactions with other value chain actors, while the female form included questions on household composition; expenditures on consumption items and health. Both forms also included several additional modules, which are described in the following subsections. 3.2.5 Women’s Empowerment in Agriculture Index (WEAI) At baseline, the survey included modules from IFPRI’s A-WEAI version of the index. 6 Modules were asked of both male and female respondents. The A-WEAI includes questions on control, decision-making, mobility, and time use, as well as questions on asset ownership and access to credit. At midline, Group 2 households completed modules from the updated Pro-WEAI version of the index.7 Thus, for Group 1 and Group 2 the A-WEAI scores can be generated, 6 The A-WEAI was co-developed by IFPRI, USAID, and the Oxford Poverty and Human Development Initiative. 7 The A-WEAI encompasses five domains of empowerment: Input in Production Decisions; Control over Resources; Control over Use of Income; Group Membership; and Workload. The Pro-WEAI expands the range of indicators used to construct the first three domains (adding questions on: access to information; autonomy in production; decision-making over land; access to a financial account; and autonomy in income). In addition, the Pro-WEAI incorporates two new domains: Mobility (ability to visit 17 while additionally a Pro-WEAI score can be generated for Group 2. Given the index is focused on agricultural producers, the WEAI modules were not administered to input seller households. 3.2.6 Measuring Expectations A key outcome for measuring the effectiveness of the informational component of the NPKS intervention was to understand whether the farmer perceived NPKS as likely to increase their own production. If a farmer believes that a new product or practice is not likely to benefit their crop, they are unlikely to use it on their entire crop, if they use it at all. However, asking farmers directly whether they believe an intervention will be good for their jute production is likely to bring about biased responses, since they may fear that negative answers could affect the likelihood of their receiving future interventions, or they may be more likely to try to answer what they believe the interviewer wants to hear. Furthermore, it is important to consider how risky the farmer perceives the intervention to be, since risk averse farmers may avoid adopting technologies they expect to provide higher average but more variable production than their current methods. To provide a measure of the effectiveness of the informational component, a method of eliciting the likelihood of different outcomes developed by Luseno et al. (2003) was adapted to the context and included in the survey form. This method has the dual advantage of being “low tech,” making it easy to implement in the field, as well as intuitive, allowing the team to measure expectations without having to attempt to explain complex concepts of probability to respondents with limited formal education. The method proceeds as follows. The interviewer presents the respondent with 5 bowls, each corresponding to a different range of potential yields.8 The respondent is given 10 beans, and asked to allocate them each to a range of yields, according to how likely the respondent thought they were to generate that range of production if applying a specified package of inputs to their jute field that season. After the interviewer modeled different scenarios with the respondent and asked comprehension questions to ensure that they understood the exercise, it was carried out twice. First, the farmer was asked to allocate the beans as if they were using a standard package of fertilizers. Second, they were asked for their expectations if they replaced the standard package with recommended applications of NPKS fertilizer. Using this information, the team could construct measures capturing both the perceived average effect of NPKS use, relative to conventional fertilizer, and how much uncertainty the farmer associated with the outcome. To construct these outcome variables, summary statistics were calculated across ten observations for each farmer (corresponding to the choice associated with each of the ten beans) and the midpoint of each category. For example, a farmer placing 1 bean in the 0-600 bin, 3 beans in the 601-900 bin and 6 beans in the 901-1050 bin would have an average expected yield of 1/10*300 + 3/10*750.5 + 6/10*975.5 = 840.45 kg/acre. A similar procedure was used to calculate the standard deviation, which was used in calculating the coefficient of variation (standard deviation/mean), providing a measure of dispersion that could be compared across the two scenarios. 3.2.7 Assessing Input Knowledge In addition to capturing subjective perceptions of the effects of NPKS, the midline survey sought to directly measure how much participants learned from the trainings. To do so, the midline survey included a module in the section on agricultural production with specific knowledge questions regarding input use and other best practices for jute cultivation. The questions used were developed by IFPRI based on an English translation of the NGO trainings manual, and encompassed questions on: seed sowing and appropriate varieties; fertilizer application, timing, and important locations) and Intrahousehold Relationships (which includes questions on respect among household members and attitudes to domestic violence). A tabular comparison of the two indicators is presented in Appendix 8.3. 8 The ranges (in kg/acre) were: 0-600;601-900;901-1050;1051-1350;1350+. To construct the mean and coefficient of variation, an upper bound of 1800kg/acre (corresponding to the 99th percentile of the yield distribution at baseline) was used. The method, based on a review of studies measuring expectations and perceptions in developing countries (Attanasio, 2009), was pilot-tested extensively for measuring agricultural yield perceptions in the context of Nigeria as part of the Financial and Health Diaries project (Janssens, Kramer, Van der List, & Pap, 2013). It was adjusted by the impact evaluation team for the context of the present study, and further piloted and refined prior to—and during—the baseline survey training. 18 types; weeding and pest management; harvest practices; appropriate retting techniques; and how to identify high quality jute fiber. IFPRI defined the correct response(s) to a given question based on the training manual. The number of correct questions in each category was then totaled, and standardized (since the number of questions varied by topic) to calculate an overall knowledge score. Since the promotions treatment focused specifically on NPKS fertilizer, the overall score was disaggregated into a fertilizer component, and a component on all topics other than fertilizer. 3.2.8 Trust Experiment To better understand how trust may affect market interactions between farmers and input sellers, the IFPRI team added an incentivized trust experiment as the final section of the main midline form, in which each farmer was randomly paired with six input sellers from their local area. This trust experiment followed a standard design, first developed by Berg et al. (1995) involving two players. The first player receives a small amount of money from the experimenter, and has the option to keep all or some portion of it, sending the remaining amount to the second player. The experimenter then triples the amount sent, and gives it to the second player, who then decides how much to keep for themselves, and how much to return to the first player. Under classical economic assumptions with self-interested players, the first player would be expected to send nothing, because the second player would have no reason to return any money. Yet, in practice, Berg et al. (1995) find large majorities of participants do engage in trusting and reciprocal behavior, with first players sending money and second players returning money, even though the games are played anonymously. They argue the amount sent by the first player is a measure of the extent to which the first player trusts the second player to reciprocate, providing a measure of the degree of trustworthiness in the subject pool. Subsequent studies using the same experimental set-up to explore trustworthiness in different social and cultural contexts have found very similar results.9 In this implementation, at the end of each survey each farmer was invited to participate in the experiment. They received an envelope containing play money representing 150 Taka (approx. $1.80) associated with a given input seller. They could then decide to either keep 0, 50, 100 or all 150 Taka, and place the remainder in the envelope to send to the input seller. Farmers were told that DATA would triple any money in the envelope, meaning that the input seller would receive 0, 150, 300 or 450 Taka depending on whether the farmer sent 0, 50, 100 or 150 Taka, respectively. This process was repeated for each of the six input sellers. Following the interview with the input seller, input sellers received four envelopes containing each potential amount sent and had to decide how much to return in each scenario. Both farmers and input sellers additionally received a flat fee of 150 Taka to participate; input sellers had the option to include some or all of the fee in the amount sent to farmers. For each of the six input sellers that a farmer was matched with, the value to be returned was calculated, using the input seller’s decision. One of the six was then chosen at random to determine an actual payout for the farmer. To ensure the experiment was implemented consistently in each interview, the IFPRI team developed a standard script using the tablet software which the DATA enumerator read to the farmer or input seller in the field.10 Similar to the expectations module, the script included examples and comprehension questions to ensure the respondent understood how to play the game. The randomized matching of farmers to input sellers was carried out by IFPRI and preloaded into the tablets used for the interviews. Similarly, the decision amounts were submitted to the server as part of data collection, and used by the IFPRI team to generate anonymous payment lists, to ensure that neither the DATA interviewer nor the interviewee would be able to identify the individual they were paired with whose choice was selected for payment. The method borrowed heavily on the standard trust game, but was adjusted for implementation in a large-scale household survey based on scripts and procedures developed a similar laboratory field experiment in Nigeria (Barr, 9 For a detailed discussion of the trust game, including a summary of contexts in which it has been applied, see Brulhart and Usunier (2008). 10 The English language version of this script is included as part of the Midline Questionnaire in Appendix 8.4, under the heading ”Trust Game”. 19 Dekker, Janssens, Kebede, & Kramer, 2017). The Nigeria study found that giving participants real money and asking them how much of that to send to another player by placing the money in an envelope made the task more tangible, easier to understand and more anonymous (hence less vulnerable to social desirability bias) compared with how the trust game is typically implemented in household surveys. The research team therefore followed a similar procedure in Bangladesh, but because each farmer played with six different sellers, and only one of these games was randomly selected for payment, participants were not given real money in the envelopes; instead, the experiment used plastic game money, and told participants that the game money for the selected game would be replaced by real money. This procedure worked well in the context of the lab-in-the-field experiments conducted as part of the impact evaluation in late 2016. The procedures were piloted by field coordinators and supervisors prior to the enumerator training, and adjustments were made to the script after the pilot test. For example, the trust game was reduced to include only four choice options for the farmer (sending 0, 50, 100 or 150 Taka instead of sending any amount between 0 and 150 Taka), to simplify the game and make it easier for enumerators to explain. After a field test of the simplified trust game occurred during training, the script was further refined and finalized for implementation to ensure that it was appropriate to the community context and easy for farmers to understand. 3.2.9 Photobooks In addition to these specialized modules, at the end of each farmer household interview, the main respondent was asked for their permission to have their picture taken by the enumerator. For each Group 3 interview, a photobook was generated containing pictures of each respondent organized by village. The photobook was used for input sellers to identify farmers with whom they had previously transacted by sight, in addition to having their name provided. Similarly, photobooks of input sellers were generated prior to beginning fieldwork of each of the input sellers to be interviewed (using previously collected pictures), so that farmers could likewise visually identify individuals with whom they recalled having previously transacted. 3.3 ENUMERATION TEAM AND TRAINING Enumerators were organized into 10 teams of six, each led by a supervisor, who in turn reported to two senior DATA staff members overseeing the fieldwork. All team supervisors had significant previous experience implementing surveys in rural areas and underwent additional internal training in managed logistics, reporting procedures and troubleshooting. All supervisors and enumerators underwent an extensive two-week survey training, developed by DATA in collaboration with IFPRI. The training was delivered by senior staff members at DATA with an IFPRI representative present to provide information on survey instrument design and to answer questions. The goal of the training was both to ensure all enumerators had a strong understanding of the questionnaire structure, proper procedures for each module, and knowledge of how to use the tablets for electronic data collection. Throughout the training, the data collection program was displayed on a projector so that each enumerator could verify they were collecting the same information as the trainer. Sections requiring specific procedures, such as the module on expectations and the trust experiment, were first modelled by DATA and IFPRI staff members to the group, then role-played by different groups of enumerators, with time devoted after each role play for enumerators to ask questions and provide peer feedback. The training was also used to extensively test the electronic questionnaire, which was updated daily prior to each session to remove errors, improve translations, and add additional notes or instructions to enumerators based on issues which arose during the training. 20 3.4 FIELDWORK 3.4.1 Household Interviews Household data collection for the midline survey began on February 10, 2017 and concluded on March 9. Each team carried out interviews in 5 villages, beginning with Group 1 and Group 2 households. For Group 1 households, the team was instructed to find the household interviewed at baseline. In cases where the household had moved, or its composition had changed, the team was to identify the main respondent from the baseline survey if that respondent was still present within the union and interview their household. If the main respondent had left the union or died, the team were advised to locate the remaining household members and use the individual who was now the principal jute farmer as the main form respondent. Overall, the DATA team was highly successful in locating households from the baseline, successfully re-interviewing 987 of the 1000 baseline households. For Group 2 households, the team followed the sampling procedure described above, attempting to locate listed households, then proceeding through the list of assigned replacements. After completing the assigned quota of Group 1 and Group 2 households within each village in a union (section of a union), the team then carried out Group 3 interviews, while market surveys were carried out by supervisors alongside household data collection. Figure 3.2- Bangladesh AVC Midline Evaluation Districts 3.4.2 Trust Game Payouts and GPS Verification Form Following completion of the household survey by all farmer and input seller households within a union, DATA supervisors notified IFPRI. The impact evaluation team would then download submitted survey data from the server and run an automated check to ensure that the trust experiment questions had been completed for each participating household, and that no data was missing. IFPRI then combined this data with the randomly pre￾assigned data matching farmers to a given input seller. The combination was used to calculate the amount to be paid out to the farmer. This payment data was entered into a spreadsheet and sent to DATA to enable them to manage the logistics of distributing payments. The data was also uploaded into a custom SurveyCTO form which was used by DATA supervisors in each area to confirm the payment amount to each household, and disburse the appropriate amount. The CTO form enabled DATA supervisors to independently check the amount, without having to contact another team member, to ensure that the randomization procedure was carried out automatically and correctly. The form was also used to collect GPS points for households where a point could not be successfully captured during the initial interview, typically due to weather conditions. 3.5 DATA CAPTURE AND MONITORING Each team of enumerators was provided with a 3G mobile hotspot, which was used nightly by supervisors to upload the day’s interviews to the CTO server. Submitted surveys were then batch downloaded by IFPRI and processed 21 using a series of automated data quality checks, and feedback was provided to the field team on response rates and interview duration. Each team supervisor also updated an Excel spreadsheet daily, tracking completed interviews and adding any comments or specific feedback. The spreadsheet was processed daily by IFPRI to ensure server submissions matched interviews recorded as completed, and that the status of every household in the sample was recorded accurately. 4 SAMPLE CHARACTERISTICS 4.1 SURVEY SAMPLE The sample design called for the field team to locate the 1,000 Group 1 households and 500 Group 2 households, along with the 259 previously identified input seller households. The field team was able to successfully locate 987 Group 1 households (Table 4.1), representing an attrition rate of 1.3%. Attrition was similarly low among Group 3 households, was the field team interviewed 256 of the 259 input seller households. For the Group 2 households, 501 interviews were carried out.11 Table 4.1- Households Interviewed at Midline, by Sample Group Sample Group Group 1 Panel farmer households Group 2 New farmer households Group 3 Input seller households Number of households 987 501 256 To properly assess the impact of the interventions, the analysis will focus primarily on the sample of farmer households (Group 1 and Group 2). To focus on the jute value chain, the sample is further restricted to include only households which report planting jute in each season that they were interviewed. From the 987 Group 1 households that were successfully re-interviewed at midline, an additional 27 households are excluded for reporting not planting jute either or both survey periods (Table 4.2). Of the Group 2 households interviewed, all report growing jute in the most recent season. As a result, the main sample used in the analysis will be 960 + 501 = 1461 households. Table 4.2- Number of Group 1 (Panel) Households Baseline sample 1000 of which, located at midline 987 of which, planted at baseline 972 of which, planted at midline 960 4.2 ATTRITION While the overall attrition rate is quite low, it could still pose a problem for inference if households and individuals who were not re-interviewed have different unobservable characteristics from those who were. If household attrition is non-random, it could bias impact estimates. Therefore, Table 4.3 compares baseline characteristics among Group 1 households who were successfully re-interviewed with households that left the sample. First, it is important to study differences in treatment assignment, a prerequisite for potentially finding differences. There is no difference in treatment assignment for the NGO training intervention (row 1). However, there is a statistically difference for the NAAFCO promotions treatment; a slightly higher proportion of those leaving the sample are from 11 One additional interview was carried out with a replacement household after a technical glitch prevented the data from a listed interview from being exported during the field work. The IFPRI team were subsequently able to successfully extract this data, and decided to retain both interviews in the dataset. 22 control villages. However, the magnitude of this difference is small (equivalent to eight fewer control households than in expectation) and unlikely to meaningfully influence results. Table 4.3- Comparison of Returning to Non-Returning Baseline Households Returning Households Non-Returning Households Difference Mean SD Obs. Mean SD Obs. Mean p-value Sig. Treatment Status NGO Trainings 0.50 0.50 960 0.50 0.51 40 0.00 0.998 NAAFCO Promotions 0.51 0.50 960 0.32 0.47 40 0.18 0.037 ** Household Characteristics Is female 0.03 0.18 960 0.08 0.27 40 -0.04 0.332 Age in years 47.3 12.7 960 50.1 12.5 40 -2.8 0.111 Is Muslim 0.69 0.46 960 0.68 0.47 40 0.01 0.898 Can read & write 0.50 0.50 960 0.52 0.51 40 -0.03 0.774 Completed primary education 0.40 0.49 960 0.45 0.50 40 -0.05 0.552 Completed secondary education 0.09 0.29 960 0.15 0.36 40 -0.06 0.370 Has secondary income source 0.73 0.45 960 0.70 0.46 40 0.03 0.715 Asset Index 0.00 2.02 960 -0.36 2.05 39 0.36 0.304 Planting Decisions Jute area planted (Decimals) 87.3 66.1 960 74.6 67.0 24 12.7 0.399 Inputs Expenditure (Taka) 4849 4827 960 4094 4850 24 755 0.479 Expenditure on inputs excl. fertilizer (Taka) 2284 2916 960 1989 2766 24 295 0.623 Days of household labor prior to harvest 14.6 13.4 960 9.4 9.9 24 5.2 0.926 Days of hired labor prior to harvest 18.0 22.5 960 28.5 66.3 24 -10.5 0.220 Expenditure on hired labor prior to harvest (Taka) 4683 6227 960 8053 19532 24 -3370 0.393 Production Outcomes Jute harvested (Kg) 755 609 960 580 383 24 175 0.069 * Jute yield (Kg/decimal) 887 308 960 893 252 24 -6 0.923 Jute price (Taka/Kg) 45.2 7.5 960 44.8 5.1 24 0.5 0.651 Total revenue (Taka) 34758 44862 960 27635 20170 24 7123 0.180 Gross margin (USD/Hectare) 505 2977 960 282 2007 24 223 0.490 Notes: Summary statistics from baseline (Group 1) households, split by those included/excluded in midline analysis due to attrition. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. In terms of household characteristics, attrited households appear quite similar to retained households in terms of household characteristics. Among those who planted jute at baseline, there are no statistically significant differences in terms of area cultivated, expenditures on inputs or use of labor. In terms of production, there is a marginally statistically significant difference in terms of total quantity harvested, which is slightly lower in the attrited group than returning households (580 kg on average compared to 755 kg for returning households). However, both groups have similar measures of productivity, price received for their jute, and total revenue from jute sales. Furthermore, given the number of statistical tests in Table 4.3, one would expect at least one test to be significant at the 10 percent level or better. 4.3 RESPONDENT CHARACTERISTICS Within the farm households in the sample, there are slightly more female than male members (52% female). However, men are far more likely to be identified as the primary farmer: in 95% of households, the main survey respondent was male (Table 4.4). Within households, adult men are much more likely to report agricultural labor on the family farm as their primary source of income (79%) than adult women (22%), though this proportion is unsurprisingly far higher in households in which the main respondent is female. Men in the sample report slightly higher rates of primary and secondary education than women overall, with female household heads being substantially less likely to have completed secondary education (4%) than male household heads (9%) or adult 23 women in general (12%). Women in the sample are also slightly younger than men on average, though the difference appears to be smaller than in other parts of Bangladesh, where male out-migration rates generally exceed female rates. That said, there is evidence that the type of agricultural activity undertaken varies by gender. Women are much less likely than men to report working on the household’s jute crop as their primary activity, but more likely to report taking care of livestock, providing suggestive evidence that some agricultural activity is quite gender specific. Table 4.4- Form Respondent Characteristics HH Members Main Respondent WEAI Respondent All Male Female All Male Female Male Female Demographic Characteristics Is female 51.6% - - 4.7% - - - - Reports farm work 49.4% 79.1% 21.5% 96.7% 97.9% 72.1% 99.4% 30.9% Is married 78.2% 76.5% 79.8% 94.9% 96.3% 66.2% 97.2% 98.8% Completed primary education 49.5% 51.8% 47.4% 39.7% 39.8% 36.8% 40.1% 40.3% Completed secondary education 14.1% 16.0% 12.3% 9.1% 9.3% 4.4% 10.0% 5.8% Age 40.6 42.0 39.4 47.2 47.4 42.5 47.7 39.0 16-25 24.1% 22.1% 26.0% 2.5% 2.6% 1.5% 1.8% 10.8% 26-40 29.9% 27.2% 32.4% 30.1% 29.2% 48.5% 27.5% 44.7% 41-60 33.9% 36.7% 31.3% 54.6% 55.1% 42.6% 59.3% 43.9% 60+ 12.1% 14.1% 10.3% 12.8% 13.1% 7.4% 11.4% 0.6% Primary Occupation Crop Farming 37.2% 64.5% 11.7% 85.7% 86.9% 61.8% 94.2% 15.8% Day Labor 1.2% 1.6% 0.8% 1.6% 1.5% 2.9% 0.0% 0.8% Livestock 18.5% 0.9% 35.0% 1.3% 0.2% 23.5% 0.2% 62.5% Helping on Farm 3.6% 2.8% 4.4% 0.1% 0.1% 0.0% 0.0% 5.8% Observations 4,583 2,220 2,363 1,461 1,393 68 499 501 Notes: Demographic and occupational categories of individuals from midline survey households. “HH Members” and “Main Respondent” columns include Group 1 and Group 2 households, while “WEAI respondent” only includes Group 2 households. 4.4 TREATMENT ASSIGNMENT Recall that due to the randomization, a given farm household, within a village, could belong to one of four assignment categories: control households, who received neither trainings nor access to the NAAFCO promotions; training only households, who could attend trainings but did not have access to the NAAFCO promotions; promotions only households who did not have access to trainings, but did have access to the NAAFCO promotions; and households which had both access to the NGO trainings and the NAAFCO promotions. Consequently, there are approximately a balanced number of households in treatment and control villages (Table 4.5). In the case of the discount treatment, due to logistical difficulties, NAAFCO were not able to complete raffles in all the assigned villages. As a result, the number of households who did not have access to the raffle is somewhat greater than the number that did: 680 households could participate in the raffles, compared to 781 who did not have access.12 For Group 1 households, raffles were carried out in 23 of the 25 assigned villages, while for Group 2 households the rate was lower with raffles being carried out in 17 of 25 villages. 12 The difference in the number of households that were exposed versus not exposed to the raffles creates a very small loss in statistical power relative to the case in which the sample was split evenly between treatment and control. Considering a variable distributed with a standard normal distribution, this sample can identify an effect size of 0.288 standard deviations, whereas a balanced sample would be able to identify an effect size of 0.287 standard deviations. 24 Table 4.5- Number of Households, by Treatment Sample Group: Group 1- Panel farmer households Group 2- Additional farmer households NGO Trainings In village with no NGO training 480 251 In village with NGO training 480 250 NAAFCO Marketing In village with no NAAFCO promotions 511 330 In village with NAAFCO promotions 449 171 Combined Treatments In control village 267 170 In village with NGO trainings 244 160 In village with NAAFCO promotions 213 81 In village with trainings & promotions 236 90 4.5 TREATMENT COMPLIANCE The analysis will largely be completed using intent-to-treat (ITT) effects, meaning that it will not only focus on individuals or households who took up the interventions, but rather the emphasis is on understanding what the offer of trainings or raffles and fairs did to change outcomes. The main benefit of using an ITT approach is that it represents the average effect of the intervention as a whole, taking into account that some people are potentially not interested in participation. Non-participation among the treatment group is not random, so removing them from the treatment group would mean that the remaining members are no longer directly comparable to control members, since it is not known which control members would not choose to participate. For example, consider the possibility that treatment group members who do not participate have lower levels of intrinsic motivation than treatment group members who do chose to participate. If only participants were considered as the treatment group, the decision to participate would introduce bias, as the analysis cannot isolate those who would have participated within the control group as the proper comparison. A drawback of using ITT to measure treatment effects is that the primary interest may be in the average treatment effect on the treated (ATT). An estimate of the ATT is simply the ITT divided by the participation rate. However, it is not possible to do statistical inference with that estimate of the ATT. Consequently, one has to use an alternative statistical technique, called instrumental variables estimation, to estimate standard errors on the ATT. However, it should be noted that instrumental variables estimators are quite variable relative to standard estimation techniques (like ordinary least squares estimation, which can be used in estimating the ITT). So with relatively low compliance rates, it may not be possible to estimate an ATT that is statistically significant. Since it is not possible for the treatments to affect outcomes without participation, it is first important to understand whether or not households participated in trainings or the promotions. The survey instrument asked respondents to report whether they had participated in the interventions. However, self-reports can be affected by recall problems (not remembering participation) or attribution (misremembering who provided the intervention). As a result, self￾reports may provide very noisy estimates, and other data sources are typically preferred to measure participation. In the case of the NGO trainings intervention, administrative data is available, allowing the impact evaluation team to match individuals in the survey data to contemporaneous participation records. Although matching survey data with participation data might sound simple, it was necessary to match both villages and individuals by name and/or phone number. The team used a fuzzy matching algorithm to match names of individuals with the administrative data. As a result, while the administrative data is expected to be more accurate than self-reports, it may still somewhat underestimate training attendance overall. Though the algorithm was designed to be relatively conservative, there is still likely to be some incorrect attribution of attendance to control group members, particularly as they may have forgotten in which year they may have attended a training. 25 In the case of the NAAFCO promotions intervention, unfortunately administrative data on fair attendance is not available, as the fairs were typically held in open spaces as promotional events, making it impossible to track who attended. The study sample formed only a small proportion of fair participants, given that anyone from the surrounding villages was invited to attend. Hence, even if attendance had been taken it would have been measured with substantial error. As a result, self-reported data from the midline survey must be used to estimate attendance. Note, however, that while farmers needed to attend the fairs to benefit from the advice offered, the raffle component was not conditioned on attendance. Farmers from treatment villages where raffles were carried out were entered into the raffle, whether or not they actually attended the fair. Conversely, even if farmers from control villages attended the fairs, they were not eligible for the raffle. While both groups could benefit from the advice by attending the fair, only treated households could receive a discount from the raffle. Attendance may therefore be a less important concern in evaluating the NAAFCP promotions intervention than in the case of the NGO trainings (Table 4.6). Table 4.6- Attendance by Treatment Assignment Attendance: Treatment Villages Attendance: Control Villages Villages Households Villages Households Rate Villages Households Rate NGO Trainings (Admin data) All 25 730 24 222 30.4% 6 37 5.4% Group 1 25 480 23 145 30.2% 6 19 4.0% Group 2 25 250 23 77 30.8% 5 18 7.2% NAAFCO Events (Self-report) All 25 738 25 350 47.4% 18 58 8.0% Group 1 25 487 25 255 52.4% 18 54 11.4% Group 2 25 251 23 95 37.8% 2 4 1.6% Notes: Assignment status from IFPRI sample design, NGO training attendance based on matching survey respondents to administrative data provided by DAI. NAAFCO event participation based on midline household survey self-report. Group 1 & Group 2 households. For the NGO trainings, in eligible villages the overall attendance rate was 30%, with little difference in attendance levels between Group 1 and Group 2 households. As noted above, this estimate may somewhat undercount training attendance, as not all variations in names may have been captured in the matching procedure. Additionally, there are a few cases in which members of control villages attended trainings, particularly among Group 2 households, but the overall rate is relatively low. Because the difference in training participation, it may be challenging to ascertain impacts on outcomes from the trainings. Moreover, the participation rates could be taken to suggest that AVC’s reluctance to continue to rely on NGO trainings to deliver interventions seems justified, as does the agreement from USAID to shift to a market systems approach. We return to this justification in the results section. In terms of the promotions treatment, attendance at the fairs is somewhat higher, at 47.4% of respondents. Attendance was somewhat higher among Group 1 households than Group 2 households- 52.4% versus 37.8%- reflecting the fact that the second round of events covered fewer villages. The rate of control group households attending the fair is also somewhat higher than for the control group household attending the trainings at 8.0%, and particularly so among Group 1 households, where 16.7% of households report attending the fairs. In part this difference may reflect the typically noisy nature of self-reported data. As described above however, eligibility for the 26 raffle was independent of whether the individual attended the fair, and was strictly conditioned on village-level treatment status. As a result, the high rate of contamination (particularly in the Group 1 sample) does not present a problem in measuring impacts of the promotions. While contamination is not a concern in evaluating the effects of the discounts, the overall rate of households receiving discounts was lower than in the original design. As described above, this situation results from the fact that raffles were not carried out in some treatment villages. As a result, the overall rate of households receiving discounts was lower than originally designed (Table 4.7). Among households in treatment villages (ie. who were intended to be treated), 84% were in villages in which a raffle took place. As the raffles were carried out separately for the two sample groups, the treatment rate was somewhat higher for Group 1 households (92%) than Group 2 households (68%). As a result, the NAAFCO promotions treatment is less intense than intended in the original design. Whereas half of households were supposed to receive discount coupons in the original design, as implemented 39% of households assigned to treatment villages received discount coupons, within which 46% of Group 1 households and 25% of Group 2 households received discounts. In terms of the level of the discount received, of the 285 winners in both groups: 57 (20% of winners) received a discount of 80% of the retail price; 88 (31%) received a 50% discount, and 140 (59%) received a 20% discount. Table 4.7- Proportion of Treated Households Receiving Discount Assignment Actual Raffle Winners Villages Households Villages Households % Households % NAAFCO Promotions All 25 738 23 620 84.0% 285 38.6% Group 1 25 487 23 449 92.2% 222 45.6% Group 2 25 251 17 171 68.1% 63 25.1% Notes: Assignment status from IFPRI sample design, self-reported attendance taken from midline survey report of main respondent. Group 1 & Group 2 households. In the context of the analysis, the fact that the proportion of households receiving the treatments is lower than the proportion assigned each treatment does not prevent unbiased estimation of a treatment effect. Since the analysis of treatment effects follows an ITT approach, it considers any household in a treatment village as treated, independent of actual participation. As discussed above, the strength of this approach is that it prevents unobservable characteristics correlated with compliance from driving the results. However, lower rates of take-up of treatment do pose a concern in terms of statistical power. While the coefficients obtained from ITT estimation will be unbiased, as the rate of treatment compliance decreases, it becomes more difficult to demonstrate the change is statistically different from no change. Intuitively, the weaker the signal the harder it is to distinguish from statistical noise. As a result, the analysis cannot exclude the possibility that there may be effects associated with each of the respective treatments which are not identified, but could have been if the treatment compliance rate had been higher. That said, the low participation rates in the NGO trainings may also speak to the efficacy of that intervention model. 4.6 TREATMENT BALANCE In any evaluation following a randomized control trial (RCT) design, it is important to ensure that the assignment of the treatment is indeed random. The reason for this is to ensure that the group receiving the treatment is comparable to the group which did not in terms of observed characteristics. Even though assignment to the treatment is done randomly, it could nevertheless be the case that the chance selection procedure could generate treatment and control groups which differ in important respects which would have to be accounted for in subsequent analysis. 27 To check that the two sets of samples are balanced, Tables 4.8 and 4.9 present summary statistics on gendered household type, primary farmer characteristics, previous jute production, and previous input use, for both treatment and control groups for each randomized intervention. For the statistics relating to jute production and input use, which are likely to change over time, the sample is restricted to Group 1 households, since Group 2 households were not interviewed at baseline. The gendered household type variable and main farmer characteristics were predetermined prior to the intervention or are unlikely to have been affected by the intervention and hence Groups 1 and 2 using midline survey data are both included in these variables.13 13 For example, it is extremely unlikely that the intervention would induce a farmer to complete secondary education, especially after just one year of intervention. However, it is worthwhile to compare the education levels of both treatment assignment groups to ensure that one does not contain a disproportionate number of well-educated farmers. 28 Table 4.8- Treatment Balance: Households in Control Villages vs. NGO Training Villages No Intervention Village NGO Training Village Mean SD Obs. Mean SD Obs. p-value Gendered Household Type Adult male only household 0.00 0.05 379 0.00 0.05 730 0.975 Adult female only household 0.06 0.23 379 0.01 0.12 730 0.096 * Adult male & female household 0.94 0.24 379 0.98 0.13 730 0.092 * Main Respondent Characteristics Is female 0.08 0.27 379 0.03 0.16 730 0.088 * Age in years 46.0 11.9 379 47.2 12.0 730 0.213 Is Muslim 0.74 0.44 379 0.67 0.47 730 0.619 Can read & write 0.48 0.50 379 0.50 0.50 730 0.691 Completed primary education 0.38 0.49 379 0.39 0.49 730 0.760 Completed secondary education 0.07 0.26 379 0.10 0.29 730 0.289 Primary income: Farming 0.88 0.33 379 0.88 0.32 730 0.954 Primary income: Agricultural labor 0.02 0.15 379 0.02 0.12 730 0.447 Primary income: Trader 0.02 0.15 379 0.03 0.18 730 0.560 Has secondary income source 0.72 0.45 379 0.75 0.43 730 0.530 Asset Index -0.02 1.37 379 -0.03 1.78 730 0.952 Planting Decisions (Group 1 Only) Planted Area 86.8 60.4 249 85.3 63.6 480 0.860 Inputs Expenditure (Taka) 4917 4828 249 4773 4606 480 0.863 Expenditure on inputs excl. fertilizer (Taka) 2233 2610 249 2199 2624 480 0.943 Days of household labor prior to harvest 15.5 14.3 249 13.8 12.8 480 0.446 Days of hired labor prior to harvest 17.6 22.6 249 19.0 23.0 480 0.656 Expenditure on hired labor prior to harvest (Taka) 4647 6409 249 4919 6451 480 0.783 Jute Production (Group 1 Only) Jute harvested (Kg) 717.8 532.3 249 758.8 635.4 480 0.589 Jute yield (Kg/decimal) 861.0 341.7 249 884.8 299.9 480 0.589 Jute price (Taka/Kg) 45.5 5.3 249 44.6 5.3 480 0.337 Total revenue (Taka) 30890 23324 249 33156 29313 480 0.506 Gross margin (USD/Hectare) 434.0 2219.0 249 688.1 2514.7 480 0.183 Notes: Summary statistics from baseline households, split by village assignment to receive NGO trainings. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Group 1 & Group 2 households. Table 4.8 shows that the households in the villages where NGO trainings were not offered are comparable on most observed dimensions to households in villages which were eligible to receive trainings. The primary farmers from these households in both groups are on average of similar age, religious affiliation, education level, and occupation. There is a marginally statistically significant difference in the proportion of female headed households between the two samples, with a slightly higher proportion present in the control than in the treatment group. This difference is likely due to the clustered sample and the low proportion of female headed households, which can cause a statistical difference if female headed households are clustered in a few villages. Consequently, an indicator variable for female household heads is included in the set of controls in subsequent regression analysis to ensure that this difference does not bias results estimated for treatment outcomes. In terms of production outcomes, the households in both groups on average report similar levels of both area cultivated and quantity of jute harvested, and hence similar yields at baseline. Similarly, both report selling comparable amounts of jute and similar values of both the amount they report selling and the total quantity harvested. Both groups report similar levels of input utilization and expenditures. Table 4.9 presents the same comparison for the marketing promotions treatment. As for the NGO trainings, the primary farmers in the household are on average very similar in terms of demographic characteristics and education levels, though there is a marginally statistically significant difference in terms of ages, with farmers in the treatment 29 villages reporting being slightly older than those in control villages. While this difference is likely to be a product of random chance, as before the age of the farmer is included as an explanatory variable in subsequent regressions where controls are included. Both groups have balanced characteristics in terms of input usage, expenditures on inputs, and jute production. Table 4.9- Treatment Balance: Households in Control Villages vs. NAAFCO Promotions Villages No Intervention Village NAAFCO Promotions Village Mean SD Obs. Mean SD Obs. p-value Gendered Household Type Adult male only household 0.00 0.05 379 0.00 0.05 738 0.982 Adult female only household 0.06 0.23 379 0.02 0.15 738 0.205 Adult male & female household 0.94 0.24 379 0.97 0.16 738 0.202 Main Respondent Characteristics Is female 0.08 0.27 379 0.05 0.21 738 0.289 Age in years 46.0 11.9 379 48.0 11.9 738 0.037 ** Is Muslim 0.74 0.44 379 0.68 0.47 738 0.676 Can read & write 0.48 0.50 379 0.48 0.50 738 0.903 Completed primary education 0.38 0.49 379 0.40 0.49 738 0.629 Completed secondary education 0.07 0.26 379 0.09 0.29 738 0.294 Primary income: Farming 0.88 0.33 379 0.87 0.33 738 0.885 Primary income: Agricultural labor 0.02 0.15 379 0.02 0.15 738 0.962 Primary income: Trader 0.02 0.15 379 0.04 0.20 738 0.371 Has secondary income source 0.72 0.45 379 0.75 0.43 738 0.420 Asset Index -0.02 1.37 379 -0.01 1.90 738 0.937 Planting Decisions (Group 1 Only) Planted Area 86.8 60.4 249 89.7 65.6 487 0.747 Inputs Expenditure (Taka) 4917 4828 249 5000 5233 487 0.923 Expenditure on inputs excl. fertilizer (Taka) 2233 2610 249 2415 3350 487 0.727 Days of household labor prior to harvest 15.5 14.3 249 15.1 14.1 487 0.848 Days of hired labor prior to harvest 17.6 22.6 249 18.4 23.6 487 0.810 Expenditure on hired labor prior to harvest (Taka) 4647 6409 249 4683 6191 487 0.970 Jute Production (Group 1 Only) Jute harvested (Kg) 717.8 532.3 249 804.7 634.6 487 0.254 Jute yield (Kg/decimal) 861.0 341.7 249 903.8 299.8 487 0.343 Jute price (Taka/Kg) 45.5 5.3 249 45.2 5.6 487 0.767 Total revenue (Taka) 30890 23324 249 35265 30108 487 0.197 Gross margin (USD/Hectare) 434.0 2219.0 249 736.2 3030.6 487 0.181 Notes: Summary statistics from baseline households, split by village assignment to receive NGO trainings. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Group 1 and Group 2 households are both included unless otherwise specified. 5 ANALYSIS 5.1 UNDERSTANDING ADOPTION To understand how the interventions influenced farmer behavior, the survey instrument was designed to capture a range of outcomes beyond just take-up of improved inputs. The analysis of the effects of the interventions is therefore structured to follow the potential impact pathways of the two treatments. Specifically, it will explore how the treatments affected: farmers’ subjective expectations for the coming agricultural season; their knowledge of 30 how to use inputs and apply improved practices; their take-up of NPKS fertilizer and overall input expenditures; their allocation of household and hired labor; and their overall agricultural productivity and revenue from jute sales. 5.1.1 Farmer Yield Expectations An important concern in understanding adoption by farmers is to understand how they perceive a given technology. If farmers do not believe that a given technology provides them with consistent benefits, they are unlikely to continue to use it in future. As described in Section 3.2.6, the primary survey form contained a module designed to elicit farmers’ expectations for the coming year under two different potential input scenarios: applying a standard fertilizer mix, and applying NPKS. For each potential scenario, two variables were constructed from the approximate distribution created by the farmer’s allocation: the mean and the coefficient of variation of the expected yields. The first measures the average expected outcome, while the second measures the dispersion of expected outcomes. Together, these two measures capture the farmer’s expectations over yields under standard fertilizers versus NPKS. Table 5.1- Summary Statistics for Yield Expectations Due to Fertilizer Use Conventional Fertilizer NPKS Mean SD Obs. Mean SD Obs. Mean (kg/acre) 1117 264 1,461 1242 248 1,461 Coefficient of Variation 0.17 0.12 1,461 0.15 0.12 1,461 Implied probability (%): < 600 kg /acre 5.63 14.47 1,461 2.57 8.39 1,461 601-900 kg /acre 13.50 19.40 1,461 8.01 14.42 1,461 901-1050 kg/acre 24.53 24.06 1,461 18.95 21.75 1,461 1051-1350 kg /acre 33.94 29.43 1,461 31.91 28.03 1,461 > 1350 kg/acre 22.40 32.82 1,461 38.56 38.63 1,461 Notes: Constructed from midline reports of yield expectations module; refer to Section 3.2.6 for details on variable construction. Group 1 and Group 2 households are both included. Under the conventional input scenario, the mean average expectation was 1,117 kg/acre, which is reasonably close to the realized yield of 1,027 kg/acre from self-reports of actual production at midline (Table 5.1). For NPKS the figure was somewhat higher, with an average reported expected yield of 1242 kg/acre. Responses for NPKS were also slightly less dispersed than for the conventional fertilizer example, as illustrated through the lower coefficient of variation. Respondents appear to perceive less downside risk with NPKS compared to conventional fertilizers: their individual allocations imply a probability of almost 20 percent of realizing a yield of less than 900kg/acre using conventional fertilizers, relative to only a 11 percent risk using NPKS. While Table 5.1 provides suggestive evidence that the interventions affected respondents’ subjective perceptions of the effectiveness of these inputs, to test this hypothesis statistically requires a regression framework. The hypothesis is tested in the following way: Panel A shows estimates of the effect of being in an NGO training village relative to villages receiving no interventions; Panel B shows estimates of the effect of being in a NAAFCO promotions village relative to villages receiving no interventions; and Panel C estimates the effects of being in a training-only village, a promotions-only village, and a training and promotions village respectively, relative to villages which were not eligible for trainings or promotions (the pure control group). Estimates are shown both with and without controls for a set of household characteristics. Full details of the estimation strategy employed are provided in Appendix 8.1. 31 Table 5.2- Effect of Interventions on Subjective Outcomes Any NPKS Knowledge (Self-report) Trust Game: Amount Sent Mean expected yield (Typical inputs) Mean expected yield (NPKS) Difference in mean expected yield Difference in CV expected yield (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Panel A NGO Trainings 0.161*** 0.151*** -3.6 -2.6 46.9** 41.8* 61.0** 51.0** 14.1 9.2 -0.021** -0.022** (0.045) (0.045) (2.9) (3.1) (21.3) (22.2) (23.2) (24.2) (22.6) (22.7) (0.008) (0.008) Observations 1081 1081 1109 1109 1109 1109 1109 1109 1109 1109 1109 1109 Panel B NAAFCO Promotions 0.201*** 0.197*** -4.3* -4.0 40.7** 40.2** 38.7** 33.5 -2.0 -6.7 -0.014 -0.015 (0.051) (0.050) (2.6) (2.8) (17.6) (18.3) (18.7) (20.6) (25.3) (26.1) (0.010) (0.010) Observations 1059 1059 1117 1117 1117 1117 1117 1117 1117 1117 1117 1117 Panel C NGO Training Only 0.051 0.032 -5.9 -5.7 53.2* 50.5* 87.8*** 82.7*** 34.6 32.2 -0.023** - 0.026*** (0.052) (0.048) (3.7) (3.9) (27.1) (28.4) (26.9) (28.1) (23.0) (22.8) (0.009) (0.010) NAAFCO Promotions Only 0.101** 0.081* -6.2* -5.8 44.0** 38.8** 23.4 14.9 -20.6 -23.9 -0.003 -0.005 (0.050) (0.047) (3.3) (3.6) (18.7) (18.1) (20.7) (20.6) (23.9) (24.5) (0.012) (0.011) Training & Promotions 0.219*** 0.217*** -3.1 -2.4 48.1** 45.1** 47.8** 44.0* -0.3 -1.0 -0.016* -0.018** (0.045) (0.043) (2.7) (2.8) (19.6) (19.6) (21.4) (22.4) (24.1) (24.6) (0.009) (0.009) p-value (Train Only = Promo Only) 0.248 0.223 0.935 0.991 0.691 0.621 0.005 0.003 0.000 0.000 0.030 0.021 p-value (Train Only = Train + Promo) 0.000 0.000 0.328 0.258 0.830 0.816 0.112 0.127 0.045 0.058 0.349 0.260 p-value (Promo Only = Train + Promo) 0.006 0.001 0.274 0.261 0.820 0.718 0.243 0.152 0.242 0.182 0.183 0.191 Observations 1433 1433 1461 1461 1461 1461 1461 1461 1461 1461 1461 1461 Mean (No interventions) 0.132 0.132 78.1 78.1 1097 1097 1205 1205 108 108 -0.020 -0.020 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes No Yes Notes: Panel A regresses the outcome variable on an indicator variable which takes the value 1 if the household was eligible to receive trainings and 0 if the household was not eligible to receive either intervention; Panel B regresses the outcome on an indicator which takes the value 1 if the household was eligible to receive NAAFCO promotions and 0 if the household was not eligible to receive either intervention; Panel C regresses the outcome on three indicators, for training eligible households only; promotions eligible households only and households eligible for both trainings and promotions, with households eligible for neither treatment as the excluded category. Numbers in parentheses identify the set of regression specifications (A-C) associated with each dependent variable. Even numbered specifications include a vector of controls for household characteristics: these include demographic and educational characteristics from midline, and agricultural production variables from baseline, for Group 2 households the latter variables are assigned the median value of their village at baseline as well as an indicator variable for Group 2 status. Odd numbered specifications do not include these controls. For binary outcome variables, logistic regressions are estimated, all other outcomes are estimate using OLS regression. Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. 32 First, respondents’ self-assessed level of NPKS knowledge is used as a dependent variable (Table 5.2, columns 1-2). The coefficients on each of the treatment outcomes are positive and statistically different from zero at the one percent level, suggesting that individuals who were in treatment villages perceive themselves as more knowledgeable about NPKS than individuals not eligible for either treatment. The next outcome considered is the average amount that the farmer sent in the trust game (described in Section 3.2.8). The goal of measuring this outcome was to determine whether either of the treatments affected the level of trust which farmers were willing to place in input sellers, which could have important implications for the promotion of new technologies or practices. There are no statistically significant differences in trust levels for either intervention controlling for baseline characteristics. Since trust is only supposed to be built by market systems interventions over longer periods of time, it is not surprising that the interventions did not affect trust between farmers and input dealers. Turning to the yield expectations (columns 5-8), households in NGO training villages report significantly higher average expected yields when using NPKS, relative to households who were not eligible to receive trainings. While training households do appear to have somewhat more positive expectations from inputs in general, when comparing conventional inputs to NPKS, they perceive a greater benefit associated with NPKS take-up than non￾training households (columns 7-8). These households also perceive less variation in outcomes when using NPKS as compared to a conventional fertilizer mix, suggesting that NPKS is not perceived as more risky than standard input allocations. In terms of the NAAFCO promotions, in comparing treatment to non-treatment villages, eligibility for the discount treatment does not appear to affect perceptions, though there is marginal evidence to suggest that households report a smaller gap in expectations between conventional fertilizer and NPKS. Exploring this idea further with Panel C, farmers in villages were exposed to the NAAFCO promotions were less optimistic about gains that NPKS can bring them compared to farmers from villages with NGO trainings; farmers from NAAFCO villages without NGO trainings do not differ significantly from farmers in the control group. 5.1.2 Knowledge about Inputs While expectations are clearly important to understand farmers’ perceptions of improved inputs, in evaluating the effects of the interventions it is also important to understand whether the informational components of the interventions actually increased farmers’ knowledge about inputs and improved practices. In the case of the NGO trainings, improving farmers’ knowledge of best practices for jute production was one of the primary goals of the intervention, while the NAAFCO promotions included direct information on different types of inputs and how to apply them. To provide some context to the information sources on inputs available to farmers, the midline survey included questions on interactions outside of the trainings with input providers (Table 5.3). Farmers were asked whether they received information on inputs in general from their input seller (other than fertilizers) and, if so, how long the discussion lasted. Farmers were then asked the same questions specifically referring to fertilizer, as well as more detailed questions on the topics discussed. 33 Table 5.3- Summary Statistics on Input Information Provided by Sellers Mean SD Min. Max Obs. Information provision Received information on other inputs 0.40 0.49 0 1 1,460 Minutes spent on other inputs 17.99 20.46 1 240 581 Received information on fertilizer 0.41 0.49 0 1 1,460 Minutes spent on fertilizer 16.18 12.76 1 120 595 Fertilizer topics covered Type of fertilizer to use 0.97 0.18 0 1 595 Quantity to apply 0.94 0.23 0 1 595 When to apply 0.83 0.38 0 1 595 Best for certain weather 0.19 0.39 0 1 595 Best to improve yields 0.44 0.50 0 1 595 Best for fiber quality 0.43 0.50 0 1 595 Best for disease resistance 0.35 0.48 0 1 595 Notes: Self-report of access to input information of main survey respondent at midline. Group 1 and Group 2 households are both included. For both fertilizer and other inputs, approximately 60 percent of farmers do not report receiving information from the vendor at the time of purchase. Among those that do, discussions are typically short: the median discussion time for fertilizer was 10 minutes, while the median time for other inputs was 15 minutes. In terms of the topics discussed for fertilizers, the focus is primarily on the type of product, and how much and when to apply it, rather than which fertilizer to use for different conditions or for different outcomes. For most farmers, the role of the input vendor as a source of information is clearly limited at present. There is therefore scope for interventions to increase both the amount and type of information regarding inputs available to farmers. The survey questionnaire included a series of questions on topics relating to inputs covered during the input trainings to measure any such increases, as well as specific questions on NPKS and other fertilizers. As distinct from the expectations module, which sought to measure subjective beliefs, the knowledge questions require farmers to demonstrate objective knowledge (for example, the correct variety of jute seed to use in highland areas, or how many days after planting a farmer should commence weeding). These questions allow the impact evaluation team to assess whether respondents in treatment groups know more about input use than those who did not receive the interventions. Estimated impacts of the NGO trainings and NAAFCO promotions suggest some objective knowledge gains (Table 5.4). In the case of the overall knowledge score, the coefficient on the indicator variable for access to the NGO trainings is positive and statistically significant in both regressions in Panel A (with and without the inclusion of control variables). While the coefficient on the promotions treatment is positive however, it appears to only change knowledge for fertilizer (Panel B), as the first two coefficients are significant, but not the coefficient on the knowledge score with fertilizer excluded. Panel C helps clarify these effects, by separating outcomes into all three potential treatment statuses (relative to the pure control group). The gains to overall knowledge are being driven by participation in the NGO trainings. Disaggregating the overall knowledge score into one score for the fertilizer questions and one for all the other questions, the NAAFCO promotions treatment is associated with a statistically significant increase in scores for fertilizer knowledge, while there is no significant gain for fertilizer knowledge for those who only were assigned to NGO trainings. Conversely, when the fertilizer questions are excluded from the knowledge score, there is a positive and statistically significant effect for the NGO trainings only group, but no significant effect associated with the NAAFCO promotions treatment. 34 Table 5.4- Effect of Interventions on Knowledge of Inputs & Practices Knowledge Score (Standardized) Knowledge Score for Fertilizer (Standardized) Knowledge Score excl. Fertilizer (Standardized) (1) (2) (3) (4) (5) (6) Panel A NGO Trainings 0.213** 0.161* 0.215** 0.188** 0.153 0.106 (0.099) (0.089) (0.087) (0.088) (0.097) (0.087) Observations 1109 1109 1109 1109 1109 1109 Panel B NAAFCO Promotions 0.264*** 0.220*** 0.322*** 0.292*** 0.168 0.131 (0.091) (0.076) (0.084) (0.081) (0.102) (0.089) Observations 1117 1117 1117 1117 1117 1117 Panel C NGO Training Only 0.259** 0.220* 0.137 0.096 0.235** 0.208* (0.126) (0.128) (0.105) (0.105) (0.114) (0.117) NAAFCO Promotions Only 0.144 0.090 0.257** 0.225* 0.059 0.012 (0.091) (0.084) (0.121) (0.115) (0.099) (0.095) Training & Promotions 0.199** 0.162* 0.306*** 0.271*** 0.102 0.074 (0.095) (0.083) (0.080) (0.080) (0.102) (0.092) p-value (Train Only = Promo Only) 0.232 0.199 0.284 0.251 0.058 0.047 p-value (Train Only = Train + Promo) 0.565 0.595 0.061 0.070 0.175 0.188 p-value (Promo Only = Train + Promo) 0.492 0.350 0.663 0.679 0.644 0.489 Observations 1461 1461 1461 1461 1461 1461 Mean (No interventions) -0.082 -0.082 -0.137 -0.137 -0.037 -0.037 Union Fixed Effects Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. For full description of regression specifications, refer to Table 5.2. These results suggest that NGO led trainings and a market systems approach will impart different knowledge upon farmers. NGO led trainings, even with relatively low attendance, cause broadly improved knowledge among farmers about farming jute, whereas the farmers “trained” by the market systems approach learn more about fertilizer, as designed. The effects on knowledge therefore appear to be in line with the objectives of the respective treatments. Farmers with access to input trainings demonstrate higher levels of knowledge of inputs that those without, while farmers receiving access to the promotional sessions demonstrate higher levels of fertilizer knowledge than those who did not, but similar levels of knowledge about other inputs. In contrast with the idea that a market systems approach might be clearly dominant over an NGO subcontracting approach, they each appear to have different knowledge benefits. 5.1.3 Improved Practices Given differences in how the interventions influenced both farmers’ subjective perceptions and their knowledge of NPKS and improved practices, the analysis next explores how the treatments influenced respondents’ take-up of both NPKS and improved practices (Table 5.5). First, whether or not interventions affected adoption of JRO-524, the jute variety promoted by the NGO trainings, is explored (columns 1-2). There is no significant association between the NGO trainings and adoption of JRO-524. However, general improved seed use in the sample is high (95% of control households report using improved seeds), even if a small proportion of households are using JRO-524. 35 Households may not differentiate strongly between different seed types. For NPKS use, there is a large, statistically significant positive effect of the intervention in both Panels A and B. As demonstrated in Panel C, the estimated effect is zero when considering households who only received NGO trainings, but there is a strong complementarity in villages receiving both trainings and promotions where the effect is large and statistically different from the effect for villages in which only NAAFCO Promotions were offered. These findings are reflected in the variation in take-up rates between training and non-training villages. In raffle villages in which trainings were offered, 25 percent of those who won a discount report using NPKS, while in raffle villages where trainings were not available, only 12 percent of winners report take-up. Therefore, strong evidence exists that the NAAFCO Promotions induced take-up of NPKS, and that this effect was strongest in villages in which NGO trainings were also offered. 36 Table 5.5- Effect of Interventions on Improved Practices Used JRO-524 seed Used NPKS Used Improved Pest Management Used Composting Used Sorting & Grading (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Panel A NGO Trainings 0.008 0.004 0.125*** 0.128*** 0.115*** 0.119*** 0.024 0.016 0.015 -0.006 (0.030) (0.029) (0.036) (0.035) (0.030) (0.031) (0.021) (0.020) (0.058) (0.058) Observations 1030 1030 840 840 1081 1081 736 736 980 931 Panel B NAAFCO Promotions -0.027 -0.031 0.168*** 0.169*** 0.138*** 0.137*** 0.006 0.008 0.077 0.074 (0.036) (0.035) (0.039) (0.034) (0.031) (0.030) (0.027) (0.028) (0.072) (0.076) Observations 975 975 826 826 1089 1089 767 767 997 934 Panel C NGO Training Only -0.014 -0.020 -0.029 -0.029 0.132*** 0.133*** 0.046* 0.043* 0.034 0.018 (0.039) (0.040) (0.045) (0.043) (0.038) (0.040) (0.024) (0.022) (0.059) (0.061) NAAFCO Promotions Only -0.028 -0.036 0.099** 0.096** 0.220*** 0.209*** 0.031 0.027 0.045 0.020 (0.046) (0.045) (0.043) (0.040) (0.040) (0.040) (0.033) (0.033) (0.057) (0.062) Training & Promotions 0.003 0.002 0.141*** 0.144*** 0.096*** 0.092** -0.001 -0.002 0.019 0.002 (0.031) (0.031) (0.032) (0.029) (0.034) (0.036) (0.026) (0.026) (0.053) (0.057) p-value (Train Only = Promo Only) 0.663 0.611 0.004 0.003 0.086 0.144 0.540 0.516 0.780 0.973 p-value (Train Only = Train + Promo) 0.462 0.338 0.000 0.000 0.378 0.311 0.073 0.074 0.658 0.647 p-value (Promo Only = Train + Promo) 0.340 0.204 0.200 0.138 0.014 0.023 0.315 0.369 0.466 0.666 Observations 1296 1296 1105 1105 1433 1433 1000 1000 1332 1266 Mean (No interventions) 0.087 0.087 0.016 0.016 0.311 0.311 0.034 0.034 0.190 0.190 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. For full description of regression specifications, refer to Table 5.2. There is further evidence that both interventions increased the proportion of farmers employing improved pest management techniques, with a larger impact in the promotions only group. For use of composting and sorting and grading, there is some weak evidence to suggest a positive effect for the trainings and promotions treatments respectively, but in general though positive the coefficients are small and statistically indistinguishable from zero. In particular for the promotions treatment, such impacts would not have been expected. 5.1.4 Cultivation Decisions The results thus far have demonstrated that the interventions affected farmers’ perceptions, increased their knowledge and increased the rate at which they adopted NPKS. The analysis will next proceed to explore whether changes in knowledge and perceptions led to changes in the decisions households made about allocating other inputs and labor for jute production (Table 5.6). Neither the NGO training nor the NAAFCO promotions treatment were associated with a statistically significant change in the amount of land which farmers used for jute cultivation. This finding is not surprising, but it will be useful in thinking about production effects, since if the intervention had caused farmers to expand into less productive land (or conversely to cultivate a smaller, more productive area more intensively) it could influence the interpretation of results. 37 Table 5.6- Effect of Interventions on Planting Decisions Planted Area Expenditure on inputs (Taka) Expenditure on inputs excl. fertilizer (Taka) Days of household labor prior to harvest Days of hired labor prior to harvest Expenditure on hired labor prior to harvest (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Panel A NGO Trainings -2.83 -3.73 35 -42 -84 -111 0.25 0.06 4.69** 4.64** 1267** 1236*** (7.00) (3.53) (454) (443) (172) (151) (1.42) (0.98) (2.13) (1.99) (562) (444) Observations 1109 1109 1109 1109 1109 1109 1109 1109 1109 1109 1109 1109 Panel B NAAFCO Promotions 2.63 -2.58 609 389 264 159 1.52 0.86 3.34 2.14 1030 714 (8.73) (4.40) (607) (556) (317) (268) (1.44) (1.02) (2.35) (1.57) (676) (437) Observations 1117 1117 1117 1117 1117 1117 1117 1117 1117 1117 1117 1117 Panel C NGO Training Only -5.64 -2.11 -16 120 5 73 -0.03 0.22 6.74** 7.30*** 1842** 2001*** (8.20) (3.90) (507) (451) (200) (177) (1.58) (1.21) (2.72) (2.39) (698) (530) NAAFCO Promotions Only 16.74 5.67 1476** 1031* 803* 620 3.00 2.00 5.31 2.35 1479 706 (11.55) (6.63) (712) (604) (436) (393) (1.86) (1.47) (3.44) (2.55) (950) (668) Training & Promotions 0.05 -3.44 295 124 46 -18 0.49 -0.01 4.29* 3.72** 1156* 997** (7.92) (4.10) (500) (476) (202) (169) (1.37) (0.92) (2.19) (1.67) (607) (415) p-value (Train Only = Promo Only) 0.027 0.198 0.011 0.021 0.023 0.062 0.055 0.169 0.688 0.107 0.672 0.067 p-value (Train Only = Train + Promo) 0.225 0.686 0.388 0.989 0.724 0.406 0.643 0.805 0.256 0.070 0.141 0.018 p-value (Promo Only = Train + Promo) 0.097 0.168 0.051 0.069 0.053 0.066 0.090 0.098 0.750 0.598 0.682 0.648 Observations 1461 1461 1461 1461 1461 1461 1461 1461 1461 1461 1461 1461 Mean (No interventions) 96.6 96.6 4069 4069 1518 1518 12.4 12.4 20.3 20.3 5394 5394 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. For full description of regression specifications, refer to Table 5.2. When considering expenditures, there is not strong evidence to suggest that farmers varied their expenditure on inputs as a result of the treatment. While there is a weakly significant result for overall input expenditures in Panel C associated with the promotions treatment, accounting for differences in baseline covariates, there are no differences in overall expenditure excluding fertilizer in treatment villages relative to control villages. Thus, this result appears to be driven by increased fertilizer use, indicating that there is not evidence to suggest that the promotions treatment caused farmers to substitute NPKS for other types of inputs, such as herbicides or insecticides. Equally, it does not suggest that the treatments caused farmers to increase their expenditures on inputs other than fertilizer. In terms of allocation of labor prior to harvest, neither intervention was associated with a change in the amount of household labor used by the household. Households in NGO training villages however, did increase their usage of hired labor- on average using an additional 5 days of hired labor in the previous season, resulting in their spending an additional 1240 Taka on average for hired labor relative to non-NGO training villages. This result is not driven by differences in the availability of labor locally (based on respondent’s reports). 5.1.5 Production outcomes As there were positive impacts on NPKS use and some inputs, the analysis next turns to jute production outcomes. The following outcome variables are measured: the total quantity of harvested; jute yields; the price received for 38 jute sales (an indicator of output quality); and total sales revenue (Table 5.7). These outcomes are reported using the natural log of the outcome variable, allowing coefficients to be interpreted in terms of percentages.14 In addition to these variables, an additional outcome is included for the gross margin of production per hectare. This measure incorporates both the total value of production (including output used as payments rather than sold directly) and the costs associated with producing that output for a given hectare of land. As such, it allows the overall benefit of the intervention to be assessed. Following USAID’s Feed the Future Indicator Handbook (2016), the gross margin is calculated using the following formula: M = ((TP * VS / VQ) – IC) / UP Where TP is the total production in metric tons: VS/VQ is the ratio of the value of sales to the quantity of sales, or the effective unit price; IC is the total cost of inputs; and UP is the area of cultivated land used. Since the coefficient for this outcome has an intuitive interpretation (the additional dollars of benefit per hectare) this value is not reported in logs. Table 5.7- Effect of Interventions on Production Outcomes Jute harvested (log) Jute yield (log) Jute price (log) Sales revenue (log) Gross margin (USD/Hectare) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Panel A NGO Trainings -0.01 -0.03 0.02 0.01 -0.01 -0.00 -0.07 -0.10 18.1 9.1 (0.10) (0.05) (0.04) (0.03) (0.01) (0.01) (0.13) (0.06) (42.1) (43.3) Observations 1109 1109 1109 1109 1089 1089 1109 1109 1109 1109 Panel B NAAFCO Promotions 0.06 0.00 0.03 0.03 -0.01 -0.01 0.04 -0.04 35.3 32.9 (0.11) (0.05) (0.04) (0.04) (0.01) (0.01) (0.14) (0.07) (47.0) (46.4) Observations 1117 1117 1117 1117 1095 1095 1117 1117 1117 1117 Panel C NGO Training Only -0.02 0.02 0.04 0.03 -0.00 0.00 -0.09 -0.06 20.8 21.2 (0.11) (0.05) (0.05) (0.04) (0.01) (0.01) (0.14) (0.07) (44.0) (45.4) NAAFCO Promotions Only 0.24* 0.10* 0.07 0.05 0.02 0.02 0.22 0.07 28.4 20.2 (0.13) (0.06) (0.05) (0.04) (0.01) (0.01) (0.16) (0.08) (67.8) (63.2) Training & Promotions 0.02 -0.02 0.02 0.01 -0.01 -0.01 -0.02 -0.08 35.4 27.0 (0.10) (0.05) (0.04) (0.03) (0.01) (0.01) (0.13) (0.06) (44.4) (44.3) p-value (Train Only = Promo Only) 0.018 0.136 0.472 0.586 0.209 0.283 0.011 0.048 0.909 0.986 p-value (Train Only = Train + Promo) 0.580 0.224 0.702 0.532 0.347 0.251 0.423 0.645 0.721 0.879 p-value (Promo Only = Train + Promo) 0.043 0.023 0.312 0.269 0.053 0.051 0.041 0.021 0.920 0.912 Observations 1461 1461 1461 1461 1432 1432 1461 1461 1461 1461 Mean (No training, no promotions) 7.34 7.34 7.58 7.58 4.42 4.42 11.0 11.0 654 654 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. For full description of regression specifications, refer to Table 5.2. 14 For example, if a statistically significant coefficient of β=0.05 were obtained for a given treatment, that would imply that the treatment was associated with an approximately 0.05 x 100% = 5% increase in that outcome. 39 For households in the NGO training villages, there is no evidence of an effect of the intervention on household jute production, with no statistically significant differences in outcomes in either Panel A or Panel C. For the promotions intervention, the sign of the coefficient on the overall quantity of jute harvested, is positive and weakly significant, though this effect is only observed in Panel C when comparing the promotions-only villages to control villages for harvested jute. This effect is not robust to the inclusion of additional controls for use of NPKS or for input expenditures, suggesting that input use is driving the outcome in Panel C. This result for the promotions treatment is driven by the lower portion of the distribution while the upper half converges with the other treatment indicators (Figure 5.1). Figure 5.1- Cumulative Distribution of Log Harvested Quantity, by Treatment Status For the other production outcomes considered, none of the treatment variables are statistically significant. Given that the promotions treatment did foster adoption of NPKS, it is important to explore why this did not lead to gains in production outcomes. One potential concern, considering the low levels of take-up discussed in Section 4.5, is that there was a positive impact of the treatment for those who complied (i.e. attended the trainings or received a discount) but that there is insufficient statistical power to identify the effect using the ITT approach. As an alternative approach, it is also possible to try to estimate a local average treatment effect (LATE) using an instrumental variable approach to measure the effects of compliance with the treatment.15 To do so, it is necessary to identify a valid instrumental variable which is both sufficiently strongly correlated with complying with the treatment, and uncorrelated with other determinants of the outcome variable. For the NAAFCO promotions treatment, raffle winner status is a good candidate to use, since it is both correlated with increased NPKS use and, due to the random assignment, is unlikely to be correlated with unobservable characteristics. Unfortunately for the NGO trainings, there does not appear to be a valid instrument available to use, since compliance (i.e. attending the trainings) is very likely correlated with unobservable outcomes (such as an individual’s latent motivation) which are also likely to be correlated with the outcomes under consideration. Therefore, in estimating the IV regressions, only the effects of the NAAFCO promotions treatment are considered (Table 5.8). By focusing on raffle winners, the analysis can exclude the hypothesis that the lack of results is simply 15 For a detailed description of LATE estimation, see Imbens and Angrist (1994). 40 being driven by low take-up of the discount. In line with the previous result for treatment status, winning the discount is significantly and positively associated with adoption of NPKS. When focusing on these winners however, there is no observable effect on any production outcomes; the coefficients on the treatment indicator in each case are not statistically different from zero. Though all the estimated coefficients are positive, as expected the standard errors are much larger than in the ITT model. As a result, though the coefficients are all positive they are not statistically different from zero. As a result, it is not possible to conclude the promotions treatment had a positive effect on average production. Table 5.8- IV Regression, Effect of Treatment Compliance on Production Jute harvested (log) Jute yield (log) Jute price (log) Sales revenue (log) Gross margin (USD/Hectare) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Used NPKS = Won Discount 0.76 0.17 0.20 0.07 0.07 0.05 0.90 0.30 319 155 (0.85) (0.37) (0.32) (0.27) (0.14) (0.11) (0.96) (0.43) (486) (425) Mean (Non-winners) 7.32 7.32 7.58 7.58 4.42 4.42 11.01 11.01 664 664 F-statistic (First stage) 16.20 16.80 16.20 16.80 16.58 17.18 16.20 16.80 16.20 16.80 p-value (First stage) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Baseline controls No Yes No Yes No Yes No Yes No Yes N 1461 1461 1461 1461 1432 1432 1461 1461 1461 1461 Notes: Even numbered specifications include a vector of baseline controls for household characteristics: these include demographic and educational characteristics from midline, and agricultural production variables from baseline, for Group 2 households the latter variables are assigned the median value of their village at baseline as well as an indicator variable for Group 2 status. Two-stage least squares IV regression of listed outcome variable on NPKS use, instrumented using an indicator variable which takes the value 1 if the respondent was a NAAFCO raffle winner and 0 otherwise. Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Over the sample of Group 1 households excluding adopters of NPKS, there is a fairly large increase in production between baseline and midline (Figure 5.2). Average production increased by approximately 200kg per household, consistent with the increase found among the control group. It seems that overall conditions were highly favorable for jute production in 2016, so the scope for NPKS adoption to further improve on these gains may have been limited or difficult to measure, as they would have been subtle than in an average or bad year. Figure 5.2- Changes in Aggregated Jute Production from Baseline to Midline In sum, the NGO trainings and NAAFCO promotions both appear to have helped improve knowledge, albeit in different ways. Trainings were effective at enhancing knowledge around inputs and practices that were not directly 41 related to NPKS, whereas the NAAFCO promotions were effective at improving knowledge and adoption of NPKS fertilizer. Although the promotions improved NPKS fertilizer usage, and were associated with some increases in harvested quantity, no strong effects were found on other outcomes. This finding can partially be explained through relatively low attendance in trainings, and partially because the difference in yields between using urea only versus NPKS in potentially quite subtle and difficult to find statistically. Though the analysis does not find strong gains in productivity, it does demonstrate that private sector models can work to induce adoption of improved practices. This is an important consideration in thinking about market systems interventions, since potentially private sector activities for other input types or services could be promoted in a similar manner. For such interventions to succeed and be sustained, a necessary condition is for there to be trust on the part of farmers that the products being offered are genuine and will benefit them. To explore this issue, the next section of the analysis uses information from the household surveys to provide some context on the role of trust in the context of the jute value chain. 5.2 UNDERSTANDING TRUST IN THE VALUE CHAIN Trust is an important concern in market interactions between smallholder farmers and input vendors. Unlike in other types of markets in which a buyer may make frequent, repeated purchases from a given seller, smallholders typically purchase inputs once a year, in advance of the planting season. Such “one-off” transactions provide weaker incentives to input sellers to supply high quality inputs, since there is less likelihood that providing poor quality products will harm profits in future. Moreover, since the effects of inputs take some time to be realized, it may be more difficult for farmers to recognize low quality inputs relative to other types of goods. As a result, farmers and input vendors may exist in a low trust equilibrium. The inclusion of input vendors in the survey sample enabled the impact evaluation team to gather detailed information on input sales from both farmers and sellers, and to match these samples to one another. The use of photobooks enabled respondents to clearly identify buyers/sellers whom they may not otherwise have been able to recognize based on name alone. In addition to gathering retrospective information on transactions and relationships between buyers and sellers, the survey also ended with the respondent participating in the incentivized trust game (Section 3.2.8), in which the amount of money that a farmer sent to a given input seller was an indication of the farmer’s trust in that input seller to reciprocate and return money. Unlike survey measurements, this incentivized game allowed the team to observe farmer trust and seller reciprocity under conditions in which a modest, but non￾trivial, amount of money was at stake. While the treatment interventions were not associated with a statistically significant difference in how farmers played the game, the measure nonetheless provides an opportunity to explore descriptively the results of the trust game and potential insights it may yield for understanding market interactions between farmers and input sellers. 5.2.1 Relationships between farmers and input sellers As described in Section 3, as part of the main survey, Group 1 and 2 respondents were presented by the enumerator with a photobook of different local input sellers, and asked in turn whether they knew a given input seller, and if they had ever transacted with that seller. If they had, they were asked when they had first transacted with that seller, and if they had done so in the previous season. If they had purchased inputs in the previous season, they were asked a series of questions about the transaction(s). They were then asked to rate how trustworthy and how knowledgeable about inputs they considered the input seller to be, on a scale of 1 to 10. For the input sellers in Group 3, a similar procedure was carried out, with the input seller first being presented with a photobook of farmers in the area to identify, then answering a series of questions about farmers they had transacted with (Table 5.9). 42 Table 5.9- Summary Statistics: Farmers’ Interactions with Input Vendors Mean SD Min. Max. Obs. Number of sellers known 4.3 3.1 0 41 1461 Proportion of sellers known 0.20 0.21 0 1 1461 Number ever bought from 2.6 1.8 0 13 1461 Proportion of sellers ever bought from 0.11 0.11 0 1 1461 Number bought from last year 1.5 1.0 0 9 1461 Proportion of sellers bought from last year 0.06 0.06 0 0.5 1461 Proportion purchasing from one seller 0.53 0.50 0 1 1461 Years known 8.5 4.6 1 27 1429 Minutes to reach (All) 15.6 11.0 0 120 1367 Minutes to reach (Preferred) 15.4 11.8 0 120 1285 Satisfaction rating (All) 8.7 1.5 1 10 1367 Satisfaction rating (Preferred) 9.1 1.3 1 10 1285 Knowledge rating (All) 7.1 2.0 1 10 1455 Knowledge rating (Preferred) 8.8 1.6 1 10 1422 Trust rating (All) 7.2 1.8 1 10 1455 Trust rating (Preferred) 8.9 1.4 1 10 1438 Notes: Self-reported data from midline survey, using input seller photobooks. Group 1 and Group 2 households are both included. The number of input sellers that farmers could identify suggests that competition among input providers is generally limited in the sample area. Farmers could recognize just four input sellers from their local area on average￾equivalent to 20% of the surveyed input sellers in their local area. Market concentration is an important concern: one third of farmers were able to recognize fewer than three input sellers, with 13% able to identify only one, suggesting that for many the market can be characterized as mono- or duopolistic. There does appear to be some degree of spatial heterogeneity: the median respondent in Mulia district was able to identify only two input sellers, compared to five in Faridpur district. The low level of competition is reflected in purchasing behavior: more than half of the farmers surveyed reported buying from one sole input provider in the previous year, with less than 5% of those surveyed reporting making purchases from more than two different vendors. In terms of realized purchases, more than half of the farmers surveyed reported buying from one sole input provider in the previous year, with less than 5% of those surveyed reporting making purchases from more than two different vendors. Farmers reported having known input sellers for a relatively long time, 8.5 years on average, and having known their preferred seller for slightly longer: an average of 9.2 years. In general, farmers do not report long travel times to reach input sellers. On average sellers are just over 15 minutes away, and the maximum travel time reported was two hours. There is little variation in average travel times between the farmer’s preferred input seller, and other input sellers, suggesting that distance is unlikely to play a considerable role in determining whether a seller is a farmer’s preferred choice to make purchases from. In general, farmers report being very satisfied with their input seller- on a scale from one to ten, the average rating was 8.7 for sellers with whom they had transacted that year, and 9.1 for their preferred seller. These averages are not too different from the baseline, and are not suggestive of large trust problems. However, perceptions do appear to matter in choosing an input seller, as farmers typically view their preferred input seller as both more knowledgeable and slightly more trustworthy than local sellers in general. 5.2.2 Trust Experiment To better understand how trust affects interactions between farmers and input sellers, the midline survey also included an incentivized trust game (described in detail in Section 3). For this game, each farmer was randomly paired with six input sellers from their locality. For each pairing, the farmer received 150 Taka and could opt to send 43 0, 50, 100 or 150 Taka to each seller. The seller then received triple that amount, and could decide how much to return to the farmer (Table 5.10). Table 5.10- Trust Decisions & Farmer Outcomes Farmer Sent Percentage of transactions Average Amount Returned Percentage of Total Received Returned Average Farmer Payout Average Seller Payout Share to Farmer 0 12.9% 0 - 150 0 100.0% 50 33.2% 105 70.2% 205 45 82.1% 100 36.5% 187 62.5% 237 113 67.9% 150 17.5% 265 59.0% 265 185 59.0% Notes: Summary of farmers’ decisions in first stage of incentivized trust game (six decisions per farmer). For details of the trust experiment, refer to Section 3.2.8. Group 1 and Group 2 households are both included. As can be seen, a large majority of farmer choices indicated that they were willing to trust the input seller at least somewhat, as they chose to send at least some money 87% of the time. Similarly, sellers typically chose to reciprocate. Average payouts to both farmers and sellers are increasing with the amount the farmer chose to send, with farmers who elected to send the full allocation receiving a 77% higher payout on average relative to those who did not send anything, though the share of the total payout going to the seller is increasing with the amount sent. Interestingly there also appeared to be some evidence of altruistic behavior on the part of sellers, with some sellers electing to send some or all of their participation fee to farmers (these additional payments are not included in calculating the averages above). The farmers knew which six input sellers they were paired with, allowing the potential relationship between the incentivized trust measure and the characteristics of a given input seller to be analyzed (Table 5.11).16 Table 5.11- Farmer Trust Decision, by Seller Characteristics Mean Amount Sent % Sending Nothing % Sending Maximum Obs. Mean Amount Sent % Sending Nothing % Sending Maximum Obs. p-value Farmer does/has… Yes No …recognize seller 82 12.0% 19.6% 1687 79 13.1% 17.0% 7079 0.028 …ever purchased from seller 84 11.2% 20.9% 915 79 13.0% 18.1% 772 0.038 …purchased from this year 86 9.1% 21.9% 506 80 13.2% 18.6% 1181 0.008 …prefer seller 89 8.0% 24.2% 376 79 13.1% 18.3% 1311 0.000 Seller does… Yes No …sell at local market 87 7.6% 25.2% 119 86 9.6% 20.9% 387 0.715 …provide input information 86 11.3% 21.6% 222 86 7.4% 22.2% 284 0.833 …offer discounts 81 11.3% 16.8% 936 79 13.1% 17.6% 7830 0.393 …offer credit 81 11.4% 18.4% 2454 79 13.5% 17.2% 6312 0.011 Notes: Summary of farmers’ decisions in first stage of incentivized trust game (six decisions per farmer). Sellers characteristics are assigned based on farmer responses to input seller module in midline survey. Group 1 and Group 2 households are both included. 16 To prevent potential negative outcomes for participants, the design was such that decisions would remain anonymous; no input seller knew which farmer they had been paired with, and no farmer knew which input seller’s choice had been selected for payment. 44 Table 5.11 presents the average amount farmers sent to input sellers, based on the characteristics of that seller and the p-value obtained by performing a two-sided t-test of the two means. This allows salient characteristics of input sellers to be identified. The trust measure appears to be increasing with the degree of familiarity between the farmer and input seller.17 The average amount sent by farmers to sellers they recognize is larger than the average sent to those that they do not. The gap grows when comparing sellers from whom the farmer has ever purchased to those with whom they have never transacted, and is larger again when comparing input sellers whom they purchased from in the previous season to those whom they did not purchase from. The gap is largest when comparing the amount farmers sent to their preferred input supplier to other sellers, with farmers trusting approximately 10 Taka more on average, and six percentage points more likely to send the maximum allocation. Looking at other seller characteristics, among sellers with whom farmers transacted in the previous year, farmers display similar levels of trust toward local sellers relative to other sellers, and to sellers who provided information on inputs than to those who did not, though in both cases the sample sizes are small. Trust does not appear to vary based upon whether input sellers offer discounts to some customers, however, farmers do display somewhat higher levels of trust toward input sellers who offer credit for input purchases. 5.2.3 Heterogeneity by Trust In the context of the interventions, the results from the incentivized trust game present an opportunity to explore whether intervention effects vary based upon the degree of trust which farmers demonstrate towards input sellers. To do so, the sample is split between “low trust” individuals, who are defined as farmers who send less than the median amount to input sellers, and “high trust” individuals (those who send the median amount or higher). Results are first presented among high trust farmers (Table 5.12). 17 Note that the test here tests for a statistically significant correlation, rather than identifying a causal relationship: farmers may place less trust in sellers with whom they have not transacted in the past, equally they may be less likely to transact with sellers whom they trust less. 45 Table 5.12- Effect of Interventions on Production (High Trust Individuals) Jute harvested (log) Jute yield (log) Jute price (log) Sales revenue (log) Gross margin (USD/Hectare) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Panel A NGO Trainings -0.05 -0.04 0.05 0.04 -0.02 -0.02 -0.1 -0.1 38 41 (0.12) (0.09) (0.06) (0.06) (0.01) (0.01) (0.2) (0.1) (71) (75) Observations 551 551 551 551 538 538 551 551 551 551 Panel B NAAFCO Promotions 0.06 0.03 0.09 0.09 -0.02 -0.02 0.0 0.0 69 77 (0.14) (0.10) (0.07) (0.07) (0.02) (0.02) (0.2) (0.1) (79) (82) Observations 559 559 559 559 546 546 559 559 559 559 Panel C NGO Training Only -0.06 -0.02 0.04 0.04 -0.01 0.00 -0.1 -0.1 42 58 (0.13) (0.09) (0.07) (0.07) (0.02) (0.02) (0.2) (0.1) (66) (71) NAAFCO Promotions Only 0.24 0.14 0.15** 0.15** 0.01 0.01 0.2 0.1 109 119 (0.16) (0.10) (0.07) (0.07) (0.02) (0.02) (0.2) (0.1) (89) (84) Training & Promotions 0.00 -0.01 0.07 0.06 -0.03* -0.02 -0.1 -0.1 52 57 (0.12) (0.09) (0.06) (0.06) (0.01) (0.01) (0.2) (0.1) (73) (78) p-value (Train Only = Promo Only) 0.006 0.005 0.022 0.024 0.486 0.497 0.012 0.010 0.353 0.297 p-value (Train Only = Train + Promo) 0.326 0.732 0.500 0.555 0.062 0.035 0.406 0.519 0.797 0.996 p-value (Promo Only = Train + Promo) 0.023 0.002 0.051 0.023 0.035 0.014 0.040 0.013 0.446 0.318 Observations 725 725 725 725 708 708 725 725 725 725 Mean (No interventions) 7.32 7.32 7.57 7.57 4.43 4.43 11.0 11.0 653 653 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Regression specifications, including outcome variables, are the same as those in Table 5.7, with the sample restricted to farmers who sent the median amount or higher on average in the incentivized trust experiment. Effects of the promotions treatment appear to be stronger than in the full sample, the point estimate for these treatments is uniformly larger, and is statistically significant for the promotions only group in Panel C at the 5% level. Overall the take-up rate of NPKS was slightly higher among high trust individuals in the promotions group (12% to 10%) which may account for a portion of this gain, however it alone is insufficient to account for the full difference. For the other production measures, there remains no evidence for statistically significant effects, though the point estimates for effects of the interventions on the gross margin of production appear to be uniformly higher than in the full sample. Conversely, for the low trust sample there appear to be no statistically significant differences between control and treatment groups across all specifications (Table 5.13). 46 Table 5.13- Effect of Interventions on Production (Low Trust Individuals) Jute harvested (log) Jute yield (log) Jute price (log) Sales revenue (log) Gross margin (USD/Hectare) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Panel A NGO Trainings 0.02 -0.02 -0.02 -0.02 0.01 0.01 -0.0 -0.1 -10 -26 (0.13) (0.06) (0.04) (0.03) (0.01) (0.01) (0.1) (0.1) (52) (44) Observations 558 558 558 558 551 551 558 558 558 558 Panel B NAAFCO Promotions 0.05 -0.03 -0.04 -0.04 0.01 0.00 0.0 -0.1 -10 -20 (0.14) (0.07) (0.04) (0.03) (0.02) (0.01) (0.1) (0.1) (56) (48) Observations 558 558 558 558 549 549 558 558 558 558 Panel C NGO Training Only 0.00 0.05 0.02 0.02 0.01 0.01 -0.1 -0.0 -7 -13 (0.14) (0.06) (0.04) (0.04) (0.02) (0.02) (0.1) (0.1) (65) (57) NAAFCO Promotions Only 0.22 0.05 -0.03 -0.06 0.03 0.02 0.2 0.0 -59 -92 (0.15) (0.08) (0.05) (0.04) (0.02) (0.02) (0.1) (0.1) (81) (76) Training & Promotions 0.04 -0.03 -0.02 -0.04 0.01 0.01 0.0 -0.1 20 -4 (0.13) (0.07) (0.04) (0.03) (0.02) (0.01) (0.1) (0.1) (54) (45) p-value (Train Only = Promo Only) 0.091 0.941 0.258 0.055 0.185 0.412 0.038 0.670 0.496 0.304 p-value (Train Only = Train + Promo) 0.695 0.110 0.273 0.088 0.833 0.882 0.346 0.561 0.631 0.845 p-value (Promo Only = Train + Promo) 0.174 0.336 0.876 0.677 0.245 0.453 0.173 0.434 0.327 0.253 Observations 736 736 736 736 724 724 736 736 736 736 Mean (No training, no promotions) 7.36 7.36 7.58 7.58 4.42 4.42 11.1 11.1 655 655 Union Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes No Yes No Yes Notes: Standard errors are clustered at the village level. Asterisks *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Regression specifications, including outcome variables, are the same as those in Table 5.7, with the sample restricted to farmers who sent less than the median amount on average in the incentivized trust experiment. 5.3 MIGRATION, LABOR SCARCITY AND WOMEN’S ROLE IN AGRICULTURE As described in Section 2, rural-urban migration is increasing in Bangladesh, resulting in a decline in the availability or agricultural laborers in rural areas. In addition to affecting market systems as a whole, changes in local labor markets may provide additional opportunities or changes in circumstances for women in rural areas. To explore this issue, this section will first provide an overview of the results from the Women’s Empowerment in Agriculture Index (WEAI), and focus on specific domains relating to the role that women play in decisions around jute production. It will then provide some context on migration from surveyed households, and the demand for agricultural labor within the sample. Finally, the analysis will proceed to explore the relation between the two, by exploring how changes in labor demand may affect the roles played by women in jute production. 5.3.1 Women’s Empowerment in Agriculture Index (WEAI) As described in Section 3.2.5, as part of the midline survey Group 2 households completed a series of modules based on IFPRI’s updated pro-WEAI design.18 The WEAI score is constructed using two sub-indices.19 The first sub-index, the five domains of empowerment (5DE), is a measure of empowerment in production, resources, income, 18 Group 1 households completed a smaller set of modules from the original A-WEAI design of the index, results from which are included in the baseline report. 19 For a full description of the design of the WEAI and construction of sub-indices, see Alkire et al. (2012). 47 leadership, and time. A respondent is considered empowered if they achieve adequacy in 80 percent or more of the weighted indicators that make up the 5DE.20 The second sub-index, the Gender Parity Index (GPI), measures women’s relative empowerment compared to the primary male respondent in the household (Table 5.14). Table 5.14- WEAI Score & Sub-Index Scores (Group 2) Indicator Women Men Number of observations 500 498 5DE score 0.74 0.83 Disempowerment score (1 – 5DE) 0.26 0.17 % achieving empowerment 38% 53% % not achieving empowerment 62% 47% Mean 5DE score for not yet empowered 0.58 0.64 Mean disempowerment score (1 – 5DE) 0.42 0.36 Gender Parity Index (GPI) 0.88 Number of dual-adult households 498 % achieving gender parity 53% % not achieving gender parity 47% Average empowerment gap 0.25 WEAI score 0.75 Notes: Aggregate WEAI score and sub-domains for male and female respondents to midline survey. Includes only Group 2 households. The WEAI score for Group 2 households in the AVC midline sample was 0.75. Thirty-eight percent of women and 53 percent of men achieved empowerment. Of women who were not yet empowered, the mean 5DE score was 0.58, meaning that these women achieved adequacy in an average of 58 percent of the domains. Of men who were not yet empowered, the mean 5DE score was 0.64, meaning that these men achieved adequacy in an average of 64 percent of the domains. The Gender Parity Index (GPI) was 0.88, and 53 percent of households achieved gender parity. The average empowerment gap between women who did not achieve gender parity and adult males in their household was 25 percent. Excluding group membership (which is low for both men and women, suggesting there are few civil society groups in the area), the primary drivers of disempowerment in the sample were autonomy in production, and workload (Figure 5.3). The next section of the analysis will therefore focus on the role women play in jute production, before proceeding to look at the role migration plays in influencing the demand for female labor. 20 As the aggregate measure for the pro-WEAI is still under development, and to ensure comparability with the baseline measures, the original WEAI was calculated. Three of the original ten indicators are no longer included in the pro-WEAI (rights over assets, public speaking, and leisure) and so were not used in this calculation. The remaining indicators were re-weighted accordingly. 48 Figure 5.3- Contribution of WEAI Indicators to Total Disempowerment 5.3.2 Women’s Role in Jute Production As part of the pro-WEAI modules, both male and female respondents from Group 2 households were asked which members of the household were involved in decision-making around jute production and other farming activities. These questions were asked independently to male and female respondents, to provide insight both into participation rates, and how perceptions of decision-making may differ between male and female respondents (Table 5.15). Table 5.15- Involvement in Decision-Making, by Activity Jute Production Other Crop Production Poultry / Small Livestock Male Report Female Report Male Report Female Report Male Report Female Report Male Respondent 99.5% 98.1% 100.0% 99.3% 56.6% 23.8% Female Respondent 54.9% 61.3% 51.0% 57.2% 94.9% 97.0% Observations 421 421 416 416 332 332 Notes: Summary of responses to pro-WEAI module questions on agricultural decision-making. Includes only Group 2 households. There are two important trends in the data. First, the male respondent is typically seen as a primary decision-maker regarding jute production by both genders, while female respondents are only involved in decisions about jute in approximately half of sample households. Second, women play more of a role in decision-making in taking care of small livestock such as poultry where there appears to be a clear gender norm. It is also notable that men are somewhat less likely to report that the female respondent is involved in decision-making over these activities (55% of male respondents report that the female respondent participates in jute production, while 51% report that the female respondent participates in other crop production) and much more likely to report themselves as involved in decision-making over livestock, than the female respondent reports.21 However, even considering differences in reporting, there appears to be clear gender differences in the extent of men and women’s roles in decision-making over agriculture with men much more involved in decisions over the household’s primary cash crop. In addition to the detailed questions on decision-making collected for Group 2 respondents using the pro-WEAI modules, the main household survey included a detailed labor allocation module for the full sample. This module 21 Disagreements between spousal reports are not uncommon in household surveys. See Ambler et al. (2017) for a detailed treatment in the context of Bangladesh. 0 0.05 0.1 0.15 0.2 0.25 0.3 Women Men Disempowerment score (1-5DE) Workload Group membership Control over use of income Access to and decisions on credit Asset ownership Autonmy in production Input in productive decisions 49 collected disaggregated statistics on how households allocated male and female labor. including labor by both household members and hired laborers, at different stages of jute production. Figure 5.4 Gendered Labor Allocation, by Production Stage (Groups 1 & 2) In line with the results on decision-making, jute production is primarily carried out by male household members, with limited use of hired labor at key time-sensitive points: weeding and input application; harvesting; bundling and retting; and stripping and washing. Household female labor is rarely used prior to harvest, but becomes much more prevalent in post-harvest activities to prepare and process the jute crop. Hired labor is predominantly male, though female hired labor is used by a large proportion of households for the stripping and washing of jute (and to a lesser extent for weeding and input application). These summary statistics, consistent with the more aggregated data on labor allocation collected at baseline, suggest there are gendered norms around stages of the production cycle, with more manually demanding tasks such as land preparation being viewed as “male” tasks, while less physically intensive tasks such as stripping jute fibers being considered more appropriate for women. Differences in gendered time allocation are similarly apparent when comparing the proportion of the total days of labor spent working on jute spent on each activity for each type of laborer. Given that workload accounts for a large share of the drivers of disempowerment among women (even though the survey was conducted at the end of the jute season), increasing uptake of labor-saving technologies could have important benefits in terms of women’s empowerment. Similarly, interventions focused on jute processing could particularly benefit women, however a more detailed analysis of local labor markets would be needed to maximize benefits. Labor by male household members is allocated across a broad range of tasks, with a large proportion of time taken up by input application and weeding, and harvesting. Meanwhile, female household members spend more than 70% of their total time allocation on two post-harvest activities: drying jute, and stripping and washing fibers. When hired male labor is used, similarly to male household labor, it is primarily used for assistance with weeding and input application, or with the harvest. Female hired labor however is overwhelmingly used for assistance with stripping and washing of jute fiber. While some women are hired to assist with weeding and input application, the use of female hired labor is extremely rare for all other tasks. Given these differences, the analysis will next proceed to 50 explore the changing dynamics of the demand for, and supply of, agricultural laborers and how this may affect women’s role in the jute value chain. 5.3.3 Migration and Labor Demand While the survey instrument was primarily designed to provide data on outcomes relating to the intervention, it included questions which tracked household members in Group 1 households between rounds (Table 5.16). Specifically, if a member had left the household between baseline and midline the respondent was prompted to provide the reason why they had left. While this data cannot provide a comprehensive analysis of internal and external migration in the survey areas, it can provide some insight into how the pool of household labor available to households within the sample changed between rounds, and how this change may have affected labor outcomes for the household. Table 5.16- Household Migration (Group 1 Households) Percentag e Observation s Individuals (Any Migration) 6.0% 4440 Individuals (Temporary Migration) 2.9% 4440 Individuals (Permanent Migration) 3.1% 4440 Of migrants, female 45.5% 268 Households with 1+ Migrant 18.3% 960 Households with 1+ Temporary Migrant 9.1% 960 Households with 1+ Permanent Migrant 9.7% 960 Notes: Reported incidence of migration as primary reason for individuals leaving household, reported at midline in reference to baseline household roster. Group 1 households only (4440 individuals, comprising 960 households). The overall rate of migration within the survey sample is high, reflecting the high levels of migration within Bangladesh as a whole. Over 18 percent of Group 1 households report having one or more person leave due to migration in the year between the 2016 and 2017 surveys22, with approximately 6 percent of all members listed at baseline reported as no longer household members due to migration, or 268 individuals out of 4,440 total household members. Migrants are somewhat more likely to be male (54 percent) than female (46 percent), while slight more than half are reported as having permanently left the household. Table 5.17- Migration by gender, age group Age Migrated Permanently Migrated Temporarily Male Female Male Female Under 18 34.0% 42.0% 30.0% 35.0% 18-25 27.0% 39.0% 37.0% 38.0% 26-35 23.0% 8.0% 27.0% 15.0% 36-60 11.0% 3.0% 4.0% 8.0% Over 60 5.0% 8.0% 2.0% 4.0% Notes: Reported age of individual at baseline, listed as having migrated at baseline. Group 1 households only (4440 individuals, comprising 960 households). Migration is being heavily driven by younger household members (Table 5.17). Among permanent migrants, 61 percent of males and 81 percent of females were twenty-five years old or younger. A similar pattern is found among temporary migrants. Given that the data shows a decline in the average household size between baseline and 22 In comparison, of the households tracked in IFPRI’s Bangladesh Integrated Household Survey dataset, 26% of households report that one or more person out-migrated in the three-year period between the 2012 and 2015 survey rounds (Authors’ own calculation, IFPRI Bangladesh Integrated Household Survey 2015). 51 midline among the panel sample, the availability of household members who can work as productive laborers is also likely to have declined. As a result, households may have to adjust their production strategies. To do so, households may respond in one of three different ways. First, households may increase household labor use by other members, either by intensifying the work done by some household members already working on jute, or by having household members who had not worked on jute in the past to start. Second, they could hire additional labor (or exchange with other households); however, it is well known that external labor is not a perfect substitute for family labor, due to monitoring difficulties. Third, households could have simply reduced total labor inputs, which would either imply increasing other inputs or reduced production. If farmers choose either of the first two options, there could be important consequences for women’s labor, since either female household members could spend more time working on jute production, or there could be increased options to work as hired laborers on other farms. Additionally, recall that jute production in the sample increased from baseline to midline, with Group 1 households increasing their output by approximately 20% at midline, compared to what they reported in the 2015 season (Figure 5.2). Though somewhat larger, this increase is in line with the national trend for Bangladesh which saw an overall increase of 9.1% in jute production from 2015-2016 to 2016-2017 (Bangladesh Bureau of Statistics, 2017). Since production increased, it would seem sensible that demand for labor during harvest and post-harvest processing should have increased, while the data suggest that the overall pool of available household labor is declining. 5.3.4 Labor Scarcity and Female Labor Usage In line with these trends, households in the sample report increasing difficulty in finding adequate labor for jute production. In both survey rounds, as part of the module on labor allocation, respondents were asked whether they had had trouble finding labor for a particular stage of the jute production process (Table 5.18). Between baseline and midline, there is a large increase in Group 1 households reporting labor scarcity, increasing from 27 to 42 percent of the sample reporting difficulties. The increase in scarcity is somewhat smaller in Faridpur, but is large in the other three districts. The scale of this change provides an opportunity to analyze how households altered labor allocation for jute production. The remainder of this section will use this change to explore changes in labor outcomes by gender. Table 5.18- Households Reporting Labor Scarcity (Group 1) Baseline Midline N % N % Group 1 255 26.6% 405 42.2% Faridpur 110 47.2% 132 56.7% Jhenaidah 93 30.4% 149 48.7% Madaripur 28 1.3% 61 28.8% Narail 24 11.5% 63 30.1% Notes: Percentage of households reporting difficulty finding labor in baseline and midline survey. Includes only Group 1 households N=960. However, the difficulty a household experiences in finding labor is not independent of other household characteristics or local trends. In the preceding analysis of intervention effects, treatments were randomly assigned, which led to causal interpretation to the results. Since the variation observed in labor scarcity is not exogenously determined, the following analysis is necessarily descriptive in nature. Nevertheless, the observational analysis may demonstrate potentially interesting correlations between labor scarcity and utilization of female labor, which could inform the design of future programming and research around these issues. To explore this, households are categorized into three types, based on their responses to the question of whether they had encountered difficulties hiring labor in the previous season, for each survey round. The availability of labor for households which did not report difficulty hiring labor at baseline, but did report difficulty hiring labor at midline is categorized as “Became scarce”. Conversely, for households reporting difficulties hiring labor at baseline but not at 52 midline, it is categorized as “No longer scarce”. Lastly, if a household reports either difficulty in both periods, or no difficulty in both periods, the categorization is “No change”. Using these categories allows a comparison to be made by households experiencing different trends in the availability of labor (Table 5.19). Table 5.19- Household Labor Use, by Scarcity Trend (Group 1) Labor Availability Male Household Labor Male Hired Labor Use Days Use Days Daily Wage (Taka) Baseline Midline Baseline Midline Baseline Midline Baseline Midline Baseline Midline All Households 97.1% 96.7% 40.2 35.9 95.3% 94.8% 41.7 46.2 278 313 Became scarce 96.6% 97.0% 42.3 39.7 97.8% 98.5% 45.2 52.3 281 319 No change 96.9% 96.2% 40.0 34.5 93.8% 93.4% 40.8 44.1 273 311 No longer scarce 99.1% 98.3% 36.6 34.1 97.4% 93.2% 38.0 42.1 294 314 Female Household Labor Female Hired Labor Use Days Use Days Daily Wage (Taka) Baseline Midline Baseline Midline Baseline Midline Baseline Midline Baseline Midline All Households 80.9% 87.2% 11.6 12.1 50.8% 52.8% 8.3 13.3 173 152 Became scarce 78.3% 88.0% 10.9 14.4 44.2% 55.1% 7.1 15.1 178 158 No change 81.8% 87.3% 12.2 11.4 54.0% 52.8% 9.0 12.7 170 151 No longer scarce 82.9% 84.6% 10.6 9.6 50.4% 47.9% 7.9 12.1 177 141 Notes: Summary statistics for labor outcomes at baseline and midline, disaggregated by changes in whether household reported difficulty finding labor at baseline and endline. Group 1 households only. Overall, the use of male labor whether from household members or hired workers is uniformly high with little variation across survey rounds, though farmers use fewer days of male household labor at midline than in the previous year, and more days of male hired labor. Nominal wages offered by households to hired male laborers increase at midline, though at a lower rate in households which shift from reporting difficulty hiring workers to not finding workers scarce (‘no longer scarce’). There is a concurrent increase in the use of female household labor in the sample, driven mainly by households who report difficulty in finding labor at midline but not at baseline (”became scarce”), while other households (with “no change” or ”no longer scarce”) report an overall decline in the total number of days that female household members worked in jute production. Similarly, there is an overall slight increase in the use of hired female labor in the sample, because of a large increase in the use of female hired laborers in households newly reporting scarcity (”became scarce”). All groups of households report hiring female laborers for longer periods, but again this increase is largest for households newly reporting scarcity (”became scarce”). They more than double the number of days for which they hire female workers. However, the fact that an overall decrease in the daily wage earned by hired females is observed is concerning. So households appear to hire female laborers for longer periods, but pay them less per day worked. This disparity does not appear to be driven simply by the types of work that males and females do, for the activity with the highest rate of use of hired female labor (stripping and washing jute fibers) the average daily wage for males at midline was 400 Taka, compared to only 150 Taka for female laborers. This comparison can be explored more formally using a regression framework. Table 5.20 presents the regression of the outcome variables for labor on two indicator variables, for households who change their reported scarcity status from baseline to midline, with those reporting no change comprising the excluded category. 23 The specification includes a control for the baseline level of the outcome of interest (with an additional vector of household level controls included in the even numbered regressions). This allows for a more precise exploration of the correlations between changes in labor scarcity between survey rounds and female labor utilization since it allows the analysis to control for village and household level characteristics. 23 Households that report no difficulty finding labor at baseline but do so at midline are treated as reporting labor “became scarce”, while those reporting difficulty at baseline but not at midline are treated as reporting labor as “no longer scarce”. 53 Table 5.20- Changes in Labor Scarcity & Use of Female Labor (Group 1) Rate Used Days Worked Wages Household Labor Hired Labor Household Labor Hired Labor Hired Labor (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Became scarce <0.001 <0.001 0.070** 0.06** 3.71*** 3.45** 5.58** 5.11** -0.61 -0.32 (0.02) (0.02) (0.03) (0.029) (1.38) (1.36) (2.18) (2.23) (1.79) (1.82) No longer scarce -0.035 -0.037 -0.02 -0.019 -1.58 -1.55 -0.22 -0.23 -1.54 -1.88 (0.03) (0.03) (0.04) (0.04) (1.32) (1.28) (2.16) (2.06) (1.56) (1.83) Baseline Controls N Y N Y N Y N Y N Y Mean (No change) 0.873 0.873 0.528 0.528 11.45 11.45 12.67 12.67 151.31 151.31 Observations 960 960 960 960 960 960 960 960 348 348 Notes: Ordinary least squares regression of labor outcomes on changes in reported difficulty in finding labor from baseline to midline. Excluded category is households reporting no change. Standard errors are clustered at the village level. Asterisks *,** and *** denote significance at the 10%, 5% and 1% levels respectively. Using this framework, there does not appear to be evidence that when labor became scarce, households became more likely to use female labor. This may be because among the reference group (no change in scarcity), 87 percent of households already use female household labor. However, there is evidence that such households become more likely to hire labor; with controls, they are six percentage points more likely to hire female labor if they report that female labor became scarce. Further, there are changes along the intensive margin; households reporting an increase in labor scarcity also increase the days of female household labor and hired labor, by about 3.5 and 5.1 days, respectively. There are no clear changes in wages conditional on hiring labor. As a result, there are potentially subtle patterns of changes occurring in jute production for women. There are no changes in the types of tasks women are doing as labor becomes scarce, but there are some changes in the number of days worked by women on those tasks. Increased labor scarcity does appear to provide opportunities for women to enter the labor market. However, it is unclear whether simple entry is beneficial to women’s empowerment, as wages are not concurrently increasing. Future interventions should also consider the potential benefits to women’s empowerment of labor saving technologies for post-harvest activities, and attempt to address gender wage gaps more generally. 6 CONCLUSION AND NEXT STEPS 6.1 CONCLUSIONS This report has discussed results from the midline survey for the impact evaluation of the AVC project. This particular study provides mid-term evidence on the effects of interventions in the jute value chain, which included NGO trainings and fairs run by NAAFCO, a private sector input provider, along with a raffle for a discount on NAAFCO’s NPKS fertilizer, which can improve yields compared to the more commonly used urea applications by providing a more balanced nutritional composition. In this section, the findings are summarized and next steps for the evaluation are discussed. A first goal of the mid-term evaluation is to understand whether constraints related to price or knowledge affect farmer adoption of improved inputs. The NGO trainings and NAAFCO promotions both helped improve knowledge, albeit in different ways. Trainings were effective at enhancing knowledge around inputs and practices not directly related to NPKS, whereas the NAAFCO promotions were mainly effective at improving knowledge and adoption of NPKS fertilizer. Although the promotions improved NPKS fertilizer usage, and were associated with small increases in harvested quantity, there are no strong effects on outcomes related to productivity. There are two plausible explanations for 54 why stronger impacts were not observed. First, the intervention asked farmers to make a subtle shift between types of fertilizer, rather than going from no fertilizer to some fertilizer, and the yield increases from this subtle shift will have been modest. Second, participation rates in the NGO trainings and raffles, and adoption rates of fertilizer amongst those winning a discount, were somewhat lower than expected. As a result, the treatment effect estimated here is the average of the effect on farmers who were actually exposed to the intervention, and the effect for those who were not actually exposed to an intervention, which may have limited the treatment effect for the average respondent. A second evaluation goal is to better understand trust between farmers and input sellers. The market systems approach to changing value chains recognizes that trust issues exist between smallholders and value chain participants. The study seeks to understand whether trust issues either within smallholder farmers or between smallholders and traders affect the transactional behavior of farmers, and if so whether market systems programming can be designed to overcome those trust issues. Farmers appear to trust input dealers that they know through transactions more than dealers that they have never used in the past. Further, having participated in the raffles appears to improve jute yields more among farmers who strongly trust their input sellers than among others, who place less trust in their input sellers. This result suggests that future programming design should consider the role of trust not simply as a component of take-up but an important factor which may influence the degree of compliance with an intervention. Third, a key concern for evaluating market systems interventions is to understand how they affect both existing customs and norms while other changes are also occurring. The research question explores the role that women play in the jute value chain as labor scarcity increases through migration. There are no changes in the types of tasks women are doing as labor becomes scarce, but there are some changes in the intensity with which women do those tasks. There are apparently more opportunities for women within labor markets as well, as hired labor appears to increase a bit more than household labor intensity, though not enough to exert upward pressure on wages as well. Meanwhile, average wage rates are falling for women while they increase for men. As a result, an observation is that the gender wage gap—within jute production—is actually rising even while women are working more on the intensive margin. This finding may be detrimental to women’s welfare as their time allocation is already filled with activities such as child rearing, cooking, and otherwise managing households. Moreover, a potential solution to help empower women would be to develop labor-saving technologies during the jute processing stage and reduce their workload. 6.2 NEXT STEPS There are five primary next steps for both further analysis and the impact evaluation. First, the IFPRI team will complete analysis of the lab-in-the-field work that was conducted simulating interventions in a market game played by input dealers and farmers in many of the sample villages. The analysis of that data was started, and will continue again after this report has been submitted. These experiments go in more depth on how trust and reciprocity between farmers and input sellers develops over time, at least within the experiment, and to what extent interventions in market systems could help improve the quality of inputs sold in the market, and outcomes for farmers versus input sellers. Second, in 2018 an endline survey is planned in both jute and mung bean areas. The mung bean areas took longer to materialize, but interventions have taken place in a large enough geographic area to conduct research there as well, which will be conducted using quasi-experimental analysis. There are potentially quite interesting differences between the mung bean and jute value chains, as issues are likely to be slightly different, despite the evaluation team’s efforts to choose relatively similar value chains. In jute, the trust in relationships is likely to be important on both sides of the value chain, between farmers and input dealers, and also between farmers and buyers. In the mung bean chain, it appears to be equally important to encourage farmers to consider the profitability of purchased seed, since most farmers use saved seed in growing mung beans. 55 Third, more analysis will be conducted through the endline survey on women’s empowerment and the interaction between AVC interventions and women’s empowerment. Some of this analysis will be conducted using data collected through ongoing qualitative fieldwork, which will be completed in the second quarter of FY 2018. From a program design perspective, there are interesting questions related to women’s empowerment that have arisen; for example, the data demonstrate that group membership is relatively infrequent; it could be that by forming women’s self-help groups, women have a stronger voice in making agricultural decisions. The endline will also include more detailed questions on labor market participation, including migration, to firm up some of the analysis of changes in labor markets that are ongoing. Fourth, in the endline, analysis of the fourth research question will be conducted. Specifically, by comparing the NGO trainings that were conducted with jute with more diffuse training methods through AVC sub￾contracts with businesses; though direct comparisons will be difficult, some analysis towards answering this question can be conducted. Fifth, the IFPRI impact evaluation team have been discussing a potential final lab-in-the-field experiment with AVC, potentially focused on trust in the output market rather than the input market, in one of the two value chains. These discussions should take place in January and any such idea would be proposed to USAID before moving forward. 56 7 REFERENCES Alkire, S., Meinzen-Dick, R., Peterman, A., A.R., Q., Seymour, G., & Vaz, A. (2012). The Women's Empowerment in Agriculture Index. International Food Policy Research Institute Discussion Paper 01240. Ambler, K., Doss, C., Kieran, C., & Passarelli, S. (2017). He Says, She Says: Exploring Patterns of Spousal Agreement in Bangladesh. IFPRI Discussion Paper Series, 37. Bangladesh Bureau of Statistics. (2017). Estimates of Jute 2016-2017. Dhaka: Bangladesh Bureau of Statistics. Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, Reciprocity, and Social History. Games and Economic Behavior, 122- 142. Bergquist, L. F. (2016). Pass-through, Competition and Entry in Agricultural Markets: Evidence from Kenya. Job Market Paper. Brulhart, M., & Usunier, J.-C. (2008). Verified Trust: Reciprocity, Altruism, and Randomness in Trust Games. Institute of Research in Management Working Paper #0809. Ellis, P., & Roberts, M. (2016). Leveraging Urbanization in South Asia: Managing Spatial Transformation for Prosperity and Livability. Washington DC: World Bank Group. Imbens, G., & Angrist, J. (1994). Identification and Estimation of Local Average Treatment Effects. Econometrica, 467- 475. Khanom, S., Hossain, S., & Hossain, S. (2012). Effects of N,P,K and S Application on Yield and Quality of White Jute (Chorchorus Capsularis L.) Var. BJC-2197. Dhaka University Journal of Biological Science, 109-116. Luseno, W. K., McPeak, J. G., Barrett, C. B., & Little, P. D. (2003). Assessing the Value of Climate Forecast Information for Pastoralists: Evidence from Southern Ethiopia and Northern Kenya. World Development, 1477-1494. Sparkman, T., Field, M., & Derks, E. (2016). Practical Tools for Measuring System Health. USAID. USAID. (2015). Scaling Impact: Extending Input Delivery to Smallholder Farmers at Scale (LEO Report 5). USAID. USAID. (2016). Feed the Future Indicator Handbook Definition Sheets. Washington, DC: USAID. USAID. (2016). Guidelines for Monitoring, Evaluation and Learning in Market Systems Development (LEO Report 51). USAID. 57 8 APPENDICES 8.1 ESTIMATION STRATEGY (UNDERSTANDING ADOPTION) To explore the impact of the different treatment assignments, the analysis employs a standard randomized control trial approach. The goal of the analysis is to estimate the effect of receiving one (or both) of the treatments on a given household, relative to what the outcome would have been had they not received the treatment. The effect of the intervention for that household would then be simply: 𝛿𝛿 = 𝑌𝑌𝑇𝑇 − 𝑌𝑌𝑈𝑈, where 𝑌𝑌𝑇𝑇 is the outcome having received the treatment, 𝑌𝑌𝑈𝑈, is the outcome without the treatment, and δ is the effect for that household. However, since both 𝑌𝑌𝑇𝑇 and 𝑌𝑌𝑈𝑈 cannot be observed for the same household (i.e. a household cannot both receive and not receive a given program), for any given household the effect of the treatment cannot be observed directly. To overcome this problem, the estimation strategy relies on the random assignment of treatment status to calculate an average treatment effect for the sample. Randomization allows for the construction of a comparison group which is similar to the treatment group in terms of its observable characteristics, such that the expected value of a given outcome is the same for the comparison group as it would be for the treatment group had it not received the treatment. As a result, the average treatment effect can be calculated as the difference in outcomes between the two groups: 𝛿𝛿̂ = 𝐸𝐸(𝑌𝑌𝑡𝑡|𝑡𝑡 ∈ 𝑇𝑇) − 𝐸𝐸(𝑌𝑌𝑐𝑐|𝑐𝑐 ∈ 𝐶𝐶)where t represents a member of the treatment group, and c represents a member of the comparison, or control group. Less formally, since both groups are observed to share similar characteristics before the intervention, and because receiving the intervention is independent of observable characteristics, the difference in average outcomes between groups may be interpreted as being caused by the intervention. To do so, the following model is estimated: 𝑦𝑦𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜕𝜕𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜇𝜇𝑈𝑈 + 𝜀𝜀𝑖𝑖𝑖𝑖 where 𝑦𝑦𝑖𝑖𝑖𝑖 is the outcome of interest for a household i, in a given union, j; 𝑇𝑇𝑖𝑖𝑖𝑖 is an indicator variable which takes the value 1 if the household received the treatment; 𝑋𝑋𝑖𝑖𝑖𝑖 is a vector of household covariates; 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 is a union-level fixed effect; and 𝜀𝜀𝑖𝑖𝑖𝑖 is a mean-zero error term. The key parameter of interest is therefore 𝛽𝛽 since this represents the amount of average change in the outcome which can be attributed to the treatment group, i.e. if the outcome under consideration were jute sales in USD and 𝛽𝛽 = 72.43 and statistically significant, this would be interpreted as meaning that being in the treatment group was associated with an average increase in jute sales of $72.43 relative to the control group during the period of the evaluation. Note that the vector of covariates 𝑋𝑋𝑖𝑖𝑖𝑖 is independent of treatment status and not required to estimate an unbiased estimate of 𝛽𝛽. However, since its inclusion may improve the precision of the estimate, and to ensure that estimates are robust to any minor differences in baseline controls, results are provided both with and without baseline covariates included. For the two treatment interventions under consideration the following models are estimated: (A) 𝑦𝑦𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽𝑇𝑇𝑇𝑇𝑇𝑇 𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜕𝜕𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜇𝜇𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 + 𝜀𝜀𝑖𝑖𝑖𝑖 (B) 𝑦𝑦𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽𝑃𝑃 𝑃𝑃 𝑃𝑃𝑖𝑖𝑖𝑖 + 𝜕𝜕𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜇𝜇𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 + 𝜀𝜀𝑖𝑖𝑖𝑖 In addition, to explore the effect of receiving both interventions, an additional model is estimated: (C) 𝑦𝑦𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽1𝑇𝑇𝑇𝑇𝑇𝑇 𝑇𝑇 𝑂𝑂 𝑂𝑂 𝑖𝑖𝑖𝑖 + 𝛽𝛽2𝑃𝑃 𝑃𝑃 𝑂𝑂 𝑂𝑂 𝑖𝑖𝑖𝑖 + 𝛽𝛽3𝐵𝐵 ℎ𝑖𝑖𝑖𝑖 + 𝜕𝜕𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜇𝜇𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 + 𝜀𝜀𝑖𝑖𝑖𝑖 This enables comparisons to be drawn between the effect of each possible treatment allocation (only having access to NGO trainings; only having access to NAAFCO promotions; having access to both NGO trainings and NAAFCO promotions) relative to the control group. In addition, Wald tests are performed to test the equality of each of the treatment coefficients to determine if there are statistically significant differences between the effects for each of the three potential treatment assignments. In line with the assignment of treatment status, standard errors are clustered by village. 58 8.2 SUMMARY STATISTICS Full Sample Mean SD Min Max Obs. Respondent Characteristics Is female 0.05 0.21 0 1 1461 Age in years 47.2 11.9 18 98 1461 Is Muslim 0.69 0.46 0 1 1461 Can read & write 0.50 0.50 0 1 1461 Completed primary education 0.40 0.49 0 1 1461 Completed secondary education 0.09 0.29 0 1 1461 Subjective Expectations Any NPKS Knowledge (Self-report) 0.22 0.41 0 1 1461 Trust Game: Amount Sent 79.3 35.5 0 150 1461 Mean expected yield (typical inputs) 1118 264 300 1576 1461 Mean expected yield (NPKS) 1243 248 300 1576 1461 Difference in mean expected yield 125 249 -1276 1276 1461 Difference in CV expected yield -0.03 0.11 -1 1 1461 Knowledge of Inputs & Practices Knowledge Score (Standardized) 0.01 1.04 -5 3 1461 Knowledge Score for Fertilizer (Standardized) 0.01 0.94 -4 3 1461 Knowledge Score excl. Fertilizer (Standardized) 0.01 1.06 -4 3 1461 Input Use & Improved Practices Used JRO-524 seed 0.08 0.28 0 1 1461 Used NPKS 0.06 0.24 0 1 1461 Used Improved Pest Management 0.33 0.47 0 1 1461 Used Composting 0.04 0.20 0 1 1461 Used Sorting & Grading 0.15 0.36 0 1 1461 Production Decisions Planted Area 97.4 63.7 14 336 1461 Expenditure on inputs (Taka) 4132 4600 226 104335 1461 Expenditure on inputs excl. fertilizer (Taka) 1578 2326 0 66675 1461 Days of household labor prior to harvest 12.6 13.1 0 140 1461 Days of hired labor prior to harvest 22.9 31.3 0 500 1461 Expenditure on hired labor prior to harvest 6013 8210 0 114000 1461 Production Outcomes Jute harvested (kg) 990 718 20 5200 1461 Jute yield (Kg/decimal) 1027 319 22 4348 1461 Jute price (Taka/Kg) 42.0 4.5 17 77 1432 Total revenue (Taka) 41508 34102 0 431429 1461 Gross margin (USD/Hectare) 663.0 640.5 -10723 3302 1461 Notes: Summary statistics of respondent characteristics and outcome variables from midline household survey. Group 1 & Group 2 households. 59 8.3 COMPARISON OF WEAI & PRO-WEAI INDICATORS Domains and indicators of empowerment in the WEAI and pro-WEAI Domain WEAI indicator Pro-WEAI indicator Production Input in productive decisions Input in productive decisions Access to information Autonomy in production Autonomy in production Resources Ownership of assets Ownership of assets Rights over assets^ Decision-making over land Access to and decisions on credit Access to and decisions on credit Access to a financial account Income Control over use of income Control over agricultural income Control over non-agricultural income Control over household purchases Autonomy in income Leadership Group membership Group membership Speaking in public^ Time Workload Workload Leisure^ Mobility* Ability to visit important locations Intrahousehold relations* Intrahousehold respect Attitudes about domestic violence *Domains that are not part of the original WEAI ^Not included in the WEAI calculation for this project 60 8.4 DESIGN PROTOCOL Impact Evaluation Associated with the Bangladesh AVC Project Evaluation team: Alan de Brauw, Berber Kramer, Hazel Malapit, and Eric Yen 1 BACKGROUND Funded under the Feed the Future (FTF) Initiative, the Bangladesh Agricultural Value Chains (AVC) project is working to improve food (and nutrition) security through strengthened agricultural value chains. Agricultural value chains in Bangladesh are typically fragmented, and lack investment and inclusion of vulnerable groups and critical linkages. AVC is focusing on a portfolio of food and non-food crops to facilitate growth in the agricultural sector. AVC is working on improving value chains in six classes of food crops (pulses, tomato, mango, ground nuts, potatoes, and a summer vegetable basket) and two classes of non-food crops (natural fibers and floriculture). The geographic focus for the AVC project is 20 southern districts in Barisal, Dhaka, and Khulna Divisions. Some of the specific value chain interventions are more focused since specific crops are not grown in all districts. The AVC project has four main components, which are called intermediate results en route to improved food security through stronger agricultural value chains. These intermediate results are: • Intermediate Result 1: Sustainable, diversified agricultural productivity increased • Intermediate Result 2: Agricultural market systems strengthened • Intermediate Result 3: Innovation and value chain upgrading increased • Intermediate Result 4: Local capacities and systems strengthened Finally, the project has a number of cross-cutting elements. These elements include: • Nutritional practices improved; • Effective gender integration and youth participation enhanced; and • Environmental sustainability and resilience to climate change strengthened. The impact evaluation will not assess the impacts of all of the interventions being conducted by AVC. Rather, it will focus on one food and one non-food crop, ideally with both having comparable attributes and growing patterns. The evaluation will go beyond just trying to understand the impacts of some of the specific, selected AVC interventions on improving agricultural productivity. First, we want to understand whether and why the “mass market” approach facilitated by AVC, with input suppliers moving from a wholesale to a retail approach for distributing their inputs, can enhance mutual trust among both input suppliers and 61 | Page smallholder farmers; to that end, we are designing experiments related to repeated interactions, quality signals and trust to play with sampled market participants. Second, we want to understand the main market constraints to production of high-quality products. We will examine whether specific interventions at the farmer level help improve the productivity and quality of specific products, and we will assess what value chain actors are willing to pay for high-quality products in order to learn more about where the market structure fails to incentivize quality. Third, we plan to study all of these points in a gender disaggregated manner, both by using a project level version of the Women’s Empowerment in Agriculture Index, and by ensuring that we analyze patterns of participation in the markets we are studying by gender, as well as examining results from experiments by gender. Therefore, the cross-cutting components of the AVC project will be key to the impact evaluation and its results. Value Chains IFPRI is focusing on two value chains in the impact evaluation, value chains for mung beans and for jute. As such the impact evaluation does not constitute an impact evaluation of the AVC in its entirety. We considered other crops, but only those that are inherently nutritious from a food perspective. Mung beans have nutritional value in terms of iron content. Iron deficiency anemia remains a major health problem in Bangladesh, whereas micronutrients covered by other crops that are potentially being studied by AVC (e.g. orange sweet potato and mango, which are rich in vitamin A) are not as large of a deficiency. Moreover, mung beans are grown by a large number of farmers and have a reasonably simple value chain (that can therefore be traced), as mung beans are roasted and sold as an individual product, sometimes with some flavoring, offering an alternative to less healthy snacks. Mung beans are therefore the most appropriate choice. Jute makes an interesting comparison value chain because it is also grown by a significant proportion of farmers in the FTF intervention zone and appears to have similar traits and constraints as the mung bean value chain. We considered other non-food crops but concluded that these are not viable for an impact evaluation; in the case of cut flowers, the AVC is quite advanced, but the number of farmers growing cut flowers in Bangladesh overall is small, and AVC had almost reached the target number of beneficiary farmers during the inception phase of the impact evaluation. In the case of coir, the value chain is in its infancy and as such any interventions are quite prospective. Hence, jute is the remaining possibility. As trust issues between input sellers, farmers and processors appear to exist in both the mung bean and jute value chains, and farmers do not tend to use improved seeds and other inputs in both value chains, there are similar issues that exist in both value chains, allowing for an interesting comparison. That said, jute has a long history as a particularly politicized crop in Bangladesh (e.g. Ali, 2012), and as such the possibilities for expansion of the crop may be limited relative to those for mung beans. Here, it is worth noting that AVC’s focus is changing from primarily contracting trainings to trying to better understand what is not working well for smallholders in their interactions with input suppliers, aggregators, and traders, and building or improving those relationships. In the first two project years, the main approach was to create producer groups at the village level, with 62 | Page the goals of delivering training on best practices and technologies; creating aggregated demand for inputs and larger aggregated outputs; and creating more, lower-risk opportunities for access to finance. Participating farmers were selected via open meetings with AVC subcontractors, as described in Annex 1. Groups of 30 participating farmers were established based on geographic proximity. These groups then participated in project activities, including training and linkage meetings. AVC has shifted to a market systems approach, and works with selected private sector firms to co-develop commercial strategies that more effectively engage and incentivize value chain actors, including smallholder farmers, input suppliers, output buyers, and service providers. The ultimate goal of these strategies are to increase transactions, build trust, build industry networks, strengthen market systems, and enable increased investment, competition, and positive development outcomes for communities in FTF target areas. As such, the impact evaluation aims to generate useful insights on constraints to market development in the mung bean and jute value chains that will ideally generalize to other food and non-food value chains in low-income countries and Bangladesh specifically. 2 THEORY OF CHANGE 2.1 ASSESSMENT AVC has recently shifted from a more traditional value chains approach to a market systems approach, though this shift is currently in process. The concept of a market systems approach is to understand how market interactions take place or do not take place all along the value chain, to identify bottlenecks or constraints, and then to design interventions that attempt to change attitudes of actors toward engaging in transactions or in making market transactions less risky and costly for all involved parties. As such, according to the market systems approach, it is important to understand the way the value chain is or is not working for specific crops before attempting to nudge actors to change behaviors to make the market system work better for all involved. Therefore some of the work in the impact evaluation is trying to assess interventions that might overcome such constraints so that actors within the system start interacting with one another more efficiently. To develop an impact evaluation around market systems, there are a few particular challenges. The organizing principle behind the market systems approach is that if a bottleneck to a well￾functioning value chain can be identified and removed, then actors within the value chain will identify the new profitable opportunities and take advantage of them throughout the market system. If so, it would be very difficult to attribute such changes to the AVC versus a general trend in the market system that would also have occurred in the absence of AVC activities. As such, we do not attempt to conduct an impact evaluation of the AVC as a whole, or even the specific value chains being studied. 63 | Page A second challenge is that there might be small changes that can be made within a market system that are both preferred by purchasers or a good or service and by the providers. If such mutually profitable opportunities exist and become possible, one can envision that change spreading rapidly through a market system. For example, one can imagine traders learning about an arbitrage opportunity between villages, and prices equilibrating once traders learn to take advantage of that opportunity (similar to Jensen, 2007). Similarly, if someone comes up with a mutually beneficial insurance contract, one can imagine actors quickly coming in and mimicking the contract. Any opportunity that can be taken advantage of quickly also is a challenge for evaluation, since the control group would almost certainly be contaminated from the point-of￾view of research.24 The goal of the impact evaluation is to be able to attribute change to specific interventions. Thus, we had to overcome the challenges above by finding interventions or components of interventions that can be implemented for (groups of) randomly selected farmers and for which benefits would accrue more slowly, allowing us to attribute changes to the interventions. Specifically, we looked for interventions in the jute and mung bean market systems that were 1) already being planned by AVC to be fielded by partners within the specific chains; 2) where randomization would not adversely affect the pace of market system change, if occurring, because otherwise, actors within the value chain would take advantage of profitable opportunities throughout the market system, which could include the control group; and 3) that could answer interesting questions to AVC, to USAID, and to the research team. Lab-in-the-field experiments take place in a more controlled and observable environment, thereby allowing us to investigate the impacts of market systems interventions for which attribution is difficult outside the lab. To design an evaluation around AVC activities, we honed in on three major bottlenecks in market systems for jute and mung bean production. A first constraint in these market systems is a lack of farmers’ knowledge regarding proper cultivation and post-harvesting practices. Although knowledge constraints can be overcome by an informational intervention or training, and could potentially significantly increase productivity as well as demand for inputs and technologies, we assume that private willingness to pay for training alone is limited. As a consequence, farmers may not be trained in the use of specific types of inputs, which reduces demand for those inputs. Consequently, one might ask why the private sector, for instance input sellers who could promote their products through training, do not either outright provide such trainings as a service to increase their customer base, or bundle training with input sales. There are several possible reasons. First, they may not be perceived a respected and effective voice, in particular when farmers are worried that private input sellers bias their trainings towards the inputs they provide themselves, in other words, farmers may not trust private input sellers to provide accurate information. Second, knowledge is a public good. If one private input seller trains farmers, it is not guaranteed that this will increase farmers’ loyalty to that particular private input seller; input sellers may distrust farmers and worry that they will purchase their inputs from 24 From the market systems point of view, quickly reducing arbitrage opportunities or developing new insurance contracts that quickly spread is advantageous; the challenge remains attribution, as is described in the previous paragraph. 64 | Page other sellers after training. Third, farmers in Bangladesh are mainly smallholder farmers, which means that costs of training an additional farmer are high relative to the potential surplus that this transaction can generate, that is, input sellers will face high transaction costs. This rationale highlights the second constraint in the jute and mung bean market systems analyzed in the impact evaluation, namely, a lack of trust between farmers and input sellers. The market has converged on an equilibrium where farmers do not trust inputs to be of high quality, and indeed, the quality of locally provided inputs is low, with market prices being unable to signal quality. A lack of trust is common in contexts such as Bangladesh, where contracts are often not enforceable. In such contexts, as long as contracts remain unenforceable, relational contracts will play an important role in improving the quality of and demand for inputs. Relational contracts are informal agreements sustained by the value of future relationships. In a relational contract, a farmer and input seller may have the agreement that the input seller provides high quality along with additional services such as credit or agricultural extension, as long as the farmer buys from the input seller. As long as the value that both parties can obtain from future trade in this relationship is sufficiently high, the informal agreement will sustain. The question, addressed by the impact evaluation, is to what extent the AVC interventions give rise to an emergence of such relational contracts. A third constraint, also already highlighted above, is high transaction costs associated with working with smallholder farmers, because production per farmer is low. In a context with perfect enforceability of contracts between farmers, it is possible to generate economies of scale by working with farmer groups to provide inputs, trainings or buy their output. However, such horizontal coordination fails when farmers face private incentives to purchase cheaper inputs outside the group, or cannot commit themselves to sell their outputs through the group because spot markets may offer better prices. Moreover, without strong trust between group members some may try to provide lower quality output to group sales, reducing returns for all group members. A lack of trust between farmers that they will all commit themselves to the group agreement will jeopardize groups’ ability to generate economies of scale or to meet quality standards for either export markets or local processing markets. Unreliability of farmer groups will also reduce trust in farmers among other actors in the value chain. Thus, a third question addressed by the impact evaluation is whether the AVC interventions give rise to horizontal coordination, for instance collective marketing, collective purchasing of inputs, and other services provided through farmer groups, and whether input sellers’ trust in farmers depends on whether they are organized as a group. 2.2 INTERVENTIONS In describing interventions implemented by the AVC, we distinguish between two types of interventions: those being implemented by NGOs and those implemented by the private sector. For the jute value chain, AVC is working primarily through NGOs or subcontractors to improve access to inputs, production, and post-harvest management. In Year 3, AVC subcontractors are training 4,000 farmers each for a total of 16,000 producers across 4 districts with improved capacity and speed the transition to use of imported, certified FO-524 jute seeds (rather than 65 | Page using non-certified seeds or seeds retained from the previous year). Farmers are being trained on how improved seeds create higher yields; how to grade jute, and the value of different grades. Both improved and traditional seeds will exhibit improved yields in degraded soils if fertilizer is added to the soil at specific times during the growing process, but the yield gains will depend on the fertilizer quality. Because quality of fertilizer and other inputs provided in local markets is low, for instance due to counterfeiting or repackaging, and because certification is a way to signal high quality, an additional important issue discussed in the trainings is whether or not fertilizer is certified. Trainings are being conducted by four subcontractors (Gono Unnayan Prochesta (GUP), Society Development Committee (SDC), Sheba Manab Kallyan Kendra (SMKK), and Prova Society), who were competitively selected for their track records of successful training in agricultural topics. Trainings and messaging being conveyed are also related to gender mainstreaming, access to finance, and fundamental business strategy. In addition, through linkage meetings, AVC is engaging high-value input suppliers (Konika Seed Company, NAAFCO Group, and Padma Seed Company) and aggregators and introducing them to farmer groups formed at the village level as a new mass market for sales and supply. However, these input suppliers do not have an active role in terms of training on modern cultivation and post-harvesting practices. To promote market systems development, AVC is also supporting the input suppliers in creative promotional campaigns to reach this market. In the jute value chain, AVC is facilitating a series of informative events, including one fair per district and additional small group sessions, organized and co-funded by NAAFCO, in which representatives of this input supplier will provide information on the products that NAAFCO is selling. The fair will also include a crop clinic in which farmers can seek additional extension advice. In addition, selected farmers will participate in a raffle through which they will receive discounts on a 25 kg pack of jute fertilizer. First-prize winners will receive an 80% discount, second-prize winners a 50% discount and third-prize winners a 20% discount. For the mung bean value chain, AVC is investing in strengthening the ability of producers and processors to respond to market demand. AVC will facilitate promotional campaigns and other interventions that expand on market linkages between producers and high-quality input and seed suppliers (ACI, Laldeer), and leverage value chain actors’ market incentives to expand production. AVC will focus on increasing access to high quality inputs, particularly BARI-6 and BARI-7 (in lieu of current seeds), and modern cultivation practices. The project will work with seed companies and processors to establish preferred supplier networks with producer groups, strengthening market relationships, and support small/medium enterprises (SMEs) and entrepreneurs who are interested in investing in upgrading sorting and processing equipment. Currently, the plan for training prior to the 2016-2017 growing season is that the training will cover the use of improved seed varieties and pulse cultivation practices. The trainings will encourage producers and aggregators to continue investing in upgrades that will drive expanded premium market access and processing. AVC is linking the training program to ongoing promotional campaigns further connecting farmers to seed and input sellers, to create a 66 | Page competitive market for inputs that links growing farmer demand for improved inputs with quality suppliers to expand access for smallholders.25 A description of the project locations is provided in Table 1. The districts/unions in which the project is active were selected because of their high concentrations of jute and mung bean farmers, respectively. Specific villages were identified by AVC in collaboration with their subcontractors as having a high concentration of relevant farmers, having acceptable proximity to subcontractors and key private sector actors, and being as far removed from other projects as possible to minimize contamination of the sample for this study. Table 1. Overview of AVC Project Locations for Jute and Mung Bean Value Chains. Jute Value Chain Locations Mung Bean Value Chain Locations No. of Registered Farmers: 16,000 11,000 Districts: • Faridpur • Gopalgonj • Jessore • Jhinaidah • Barguna • Puthuakhali Local Implementing Partners: • Gono Unnayan Prochesta (GUP) • Society Development Committee (SDC) • Sheba Manab Kallyan Kendra (SMKK) • Prova Society • To be determined as the interventions occur at the end of calendar year 2016 Promoted Technology and Management Practices: • Crop genetic (FO-524 seeds, mulching) • Certified inputs such as fertilizer, weedicide, herbicides • Post-harvest management (ribbon retting, improved retting, grading and sorting, handling, storing) • Pest management (ICM/IMP physical methods) • Marketing and distribution (improved input purchase, market information) • NAAFCO: Promotional fairs and raffles of promotional discounts • Crop genetic (improved and certified BARI-6 and BARI-7 seeds) • Cultivation practices • Pest management (ICM/IPM methods) • Post-harvest management (decay and insect control, temperature and humidity control, improved quality control technology, sorting and grading, value added processing) • Handling (improved packing house technology and practices, improved transportation) 25 Other cross-cutting interventions will also take pace in these value chains; we make the assumption that when randomized, the treatment and control groups will have equal exposure to these additional cross-cutting interventions. Surveys taking place in the middle and end of the project can help test that assumption. 67 | Page Jute Value Chain Locations Mung Bean Value Chain Locations • Partex 2.3 OUTCOMES AND INDICATORS Table 2 provides an overview of the constraints, which randomized interventions address these constraints, outcomes these interventions are expected to have, and what indicators we will use to measure these outcomes. We discuss the indicators in more detail in Section 5, along with indicators of impacts that may be brought about by changes in the relevant outcomes and indicators listed here. Note that we do not list outcomes or indicators that follow from the changes we expected to see in household level outcomes by alleviating these constraints. Table 2. Constraints, Interventions, Outcomes, and Indicators Constraints Interventions addressing the constraint Relevant outcomes Relevant indicators Lack of knowledge regarding improved inputs Training provided by NGOs Training provided by input sellers Discounts / free trial packs Improved knowledge of the benefits of improved inputs Knowledge of improved inputs; Adoption of improved inputs; Yield of targeted crops (jute and mung bean). Lack of trust between farmers and input sellers Linkage meetings Wholesale sellers of improved inputs moving to retail system. Increased prevalence of relational contracts Inputs purchased on credit; Repeat purchases; Extension/training from private sector; Adoption of improved inputs. Yield of targeted crops (jute and mung bean). High transaction costs of working with smallholders Group formation done by NGOs Improved coordination between farmers Quantity of inputs purchased and output sold via group. To highlight how these indicators can help us understand the causal chain, consider the example of how trainings can affect the market system (Figure 1) through group formation; trainings that teach the proper use of inputs and any potential improvements in growing techniques; and 68 | Page linkages to input sellers and crop buyers. Assuming that the information given in trainings is relevant and that farmers attend trainings, they will become more aware of improved techniques and inputs. Assuming that trainings are convincing enough to try these new techniques or inputs, the next step is that farmers will consider making changes to practices or input use. Assuming again that they have access to these new inputs or techniques, that they can finance these changes, that they are willing to take risks, and that they perceive that investments or changes in investments are profitable, they will adopt those new techniques and will obtain higher levels of output and crop sales if the investments are indeed profitable, if markets exist or can handle additional output, and if shocks do not negatively affect production too much. Figure 1. Theory of Change, Training Intervention A potential intervention that could improve farmers’ trust in private sellers of quality inputs is a promotional campaign, including for instance discounts or free trial packets of inputs. From an economic theory standpoint, a promotional campaign provides information to consumers that changes their perception of the product and hence the demand; further, if promotional/initial discounts or trial packs are used in combination with the information channel above, the price of improved inputs is low enough to fit within a consumer’s demand set, encouraging the farmer to “try” those inputs. Either this trial or information provided through promotional campaigns or Impact: Increased output and sales of target crops; improved household outcomes Intervention • Farmer group formation • Trainings that teach use of new growing techniques and modern inputs (seeds/fertilizer/other chemicals) • Linkage meetings with input sellers and buyers Reach/reaction: Farmers become aware of improved farming practices/uses of inputs Capacity changes: Farmers consider making changes to practices or input use Practice changes: Farmers adopt new techniques and inputs Assumptions • Value Chain actor/NGO is respected and effective voice • Information is relevant to farming practices in area • Farmers attend trainings • For female farmers, trainings are gender iti Assumptions • Trainings are convincing enough to make farmers consider making changes to practices Assumptions • Farmers can finance any changes to practices • Some farmers willing to take risks of trying new practices (others can follow by watching initial farmers) • Farmers perceive investments or changes will be profitable Assumptions • Adoption leads to higher output and farmers can find markets for output • Negative shocks do not adversely affect output 69 | Page agricultural training will lead the consumer to demand more in the long run at the normal, non￾discounted, price. The key insight is that the consumer lacks information about or trust in the product (or product category) that the company providing the promotions is selling, and with information, trust and knowledge of potential benefits is enhanced, which results in higher demand for improved inputs. There is a slight difference between information disseminated via promotional campaigns versus promotional discounts and/or trial packs, in that if a farmer receives a discount, then the production of a specific output could be profitable with the discount but not without; if so, one would assume the farmer would not continue to use the input. In other words, the promotional discounts will only lead to increased demand for normal-priced inputs in the long run if the inputs increase yields sufficiently to offset the normal cost of that input. 3 RESEARCH QUESTIONS The study has several research questions. These largely revolving around issues of understanding adoption in the value chain; understanding trust in the value chain; women’s empowerment; and training modalities. We are also including a fifth set of questions here related to consumer acceptance of processed foods; we will include that component if it is possible and aligns with AVC’s future workplans. Understanding Adoption A basic question about the growth of agriculture in developing countries centers around enhancing adoption of new technologies and practices. In both jute and mung bean crops, farmer returns could increase by using better planting material (e.g. improved and varieties of certified seeds that are better for processing) and by using (higher quality) inputs such as certified fertilizer, weedicide and herbicide. AVC is planning to use several types of promotional tools in attempting to upgrade input supply networks. These promotional tools include mass media campaigns (which can include local fairs) to encourage smallholders to use improved seeds (such as BARI 6 and BARI 7) and certified fertilizer as opposed to lower-yielding fertilizers from local markets; comparative demonstration plots; the development of seed seller networks; and promotional discounts or starter packets of high quality seeds and inputs. The promotional discounts are a good candidate for randomization. AVC, through partnering input suppliers, will randomize whether farmers receive these discounts through public raffles in promotional events. These discounts can induce initial adoption of an input or new variety. The question of whether initial promotional discounts can be effective in helping induce technology uptake is a research question of primary interest, understanding that such adoption takes place in an environment in which a positive adoption message is being broadcast from several directions. We therefore randomize the size of the discount to be able to test the research question: • Can promotional discounts be used to catalyze input demand and increase production of food and non-food crops, relative to just using mass marketing techniques? What are the effects of increasing that production on household income, individual diets, gendered time use, and other outcomes? 70 | Page Note that with a short mid-term survey we can ask whether households were exposed to the mass media messages, and that can help us understand the relative efficacy of mass marketing, mass marketing plus promotions, and no marketing. For randomization to identify the impacts of promotional discounts, one would only require that the proportion of households exposed to mass media messages be statistically equivalent across the treatment and control groups. This research question relates to the first key question in the Feed the Future Learning Agenda under “agricultural productivity”: “What are characteristics of effective, efficient, and sustainable vehicles for promoting adoption of innovation (technology, practices, behaviors) and diffusion of products and new technologies among the poor, women and socially marginalized? What are the most binding constraints in promoting technology adoption and the most effective interventions for dealing with these constraints?” Understanding Trust in the Value Chain A recurring theme that emerged from our discussion with AVC staff is the lack of trust between farmers and traders, which may be a key constraint to the development of many value chains. We propose to measure trust in vertical relationships (e.g., farmer-trader) using trust games and analyze what type of interventions – including stylized elements of the retailer model that AVC is aiming to facilitate – help improve trust in these relationships. We will design these trust games to also involve horizontal coordination, i.e. the need for market their produce or purchase inputs collectively, and we will analyze whether trust in vertical relationships affects coordination in horizontal relationships and vice versa. We will develop these experiments to take place after the baseline survey, since they either involve problems with the coordination of sales between farmers or trust issues related to farmers. The research question we plan with this lab-in-the-field experiment is: • Do trust issues within smallholder farmers or between smallholders and traders affect the transactional behavior of farmers, and how can programming be designed or modified to overcome such trust issues? Answering this question in a lab framework may help AVC design better interventions to improve the market system, and speaks to the FTF Learning Agenda question under “Expanded markets, value chains and increased investment”: “Have interventions in agricultural value chain development led to development of local institutions and systemic behavior change? What are effective pathways for generating that change?” Women’s Empowerment The migration trends in Bangladesh present an opportunity for learning about women’s empowerment. Labor costs and wages are increasing since there is depopulation in rural areas, with more land being available for farming. Land leasing is hence increasing, with cash transfers from land tenants to the landholders in the city; sharecropping is reducing. We thus propose to stratify by areas of migration, which affects the extent of labor shortages and creates pressure for 71 | Page women to step into men’s roles in agriculture. The stratification will allow us to ask a secondary research question: • To what extent are women taking on more of men’s roles in agriculture? If we observe labor shortages in specific tasks that women manage, are there specific technologies or practices that can be marketed to women to increase productivity and alleviate their time burden? Answering this question may be helpful to AVC in identifying opportunities for targeting agricultural productivity interventions to women. Training An initial aspect of the AVC interventions that we can potentially evaluate is the effectiveness of the training by modality. AVC is interested in understanding whether, for example, a trader would be more effective in conducting farmer trainings because they have an interest in farmer’s adoption of improved seeds, as compared to an NGO, who do not have future interactions with the farmers. The change in training modalities will take place between FY2016 and FY2017, which does not give us a chance to do temporal tests, but it will potentially give us a chance to measure impacts of both modalities individually, even if in a non-randomized setting. Whereas we were able to randomize training by NGOs for jute farming (versus no training) in FY2016, it is not yet clear how training will occur in FY2017 for mung beans, and whether training in FY2017 for jute will be conducted by the private sector. It is likely that any jute training will take place through the private sector in FY2017, which would allow for at least a comparison between learning from NGO and private sources; nonetheless, we caution that it would be difficult to control the flow of information through private sources in the sense of attempting to randomize, so we will likely have to use non-random methods (e.g. propensity scores) to examine this hypothesis. Assuming that the change in modalities takes place as planned, we can test the research question: • Which training modalities are most effective in increasing farmer adoption of improved technologies and practices? Consumer Acceptance of Processed Foods We propose to measure the revealed preference for a processed product (e.g, roasted mung beans) that has nutrition labeling relative to an equivalent product with no labels through a lab￾in-the-field experiment conducted as part of our fieldwork. Enumerators can carry packets of the processed mung beans during fieldwork and conduct the experiment as part of the mid-line survey. The price could be varied to learn about the price premium. AVC realizes that increased income may lead to an unhealthy diversification of diets (“If you can fry it, they will buy it.”), so this experiment aims to measure whether nutrition labeling can help consumer make healthier food purchases. If AVC is able to reach this point in their workplans, we will attempt to answer the research question: 72 | Page • Under what conditions can nutrition labeling affect consumers’ food purchases in rural areas? This experiment would help us understand the consumer approach to products from value chains, and to understand whether labelling has an impact on consumer decisions in rural areas. Additionally, we can conduct such an experiment quite easily with both men and women to learn about gender differentials in the willingness to pay. It is important to remember that all of the selected interventions for evaluation take place within the context of other on-going interventions. IFPRI’s impact estimates will represent average treatment effects on the treated of the specific interventions, averaging over whether or not other households are exposed to other interventions in the value chain. The impact evaluation will not attempt to measure the overall impacts of the AVC. The research team will need to understand either from project data or household level data how other households are exposed, both within the treatment and control groups. The team will design data collection instruments that will cover other interventions within specific value chains to ensure this essential heterogeneity is captured. 4 METHODOLOGY 4.1 CLUSTER RANDOMIZED TRIALS Much of the impact evaluation builds on a cluster randomized trial (CRT). CRTs are a type of randomized controlled trial (RCT) in which groups of subjects, rather than individual subjects, are randomized in order to evaluate the impact of a particular intervention, policy or project. Relative to individual level RCTs, CRTs reduce the risk of biases in the evaluation due to spillover and crossover effects between individuals, and are a natural fit when the intervention is applied at the group level. The latter is definitely the case with this project, since both the design and implementation plan rely heavily on actions taken at the farmer group level such as information sharing, community mobilization, trainings and extension, and access to inputs and agents. In this context a CRT is the most appropriate method for the impact evaluation. CRTs also reduce any probability of spillovers or envy between groups. We use villages as the unit of randomization as training groups were not well formed at the time the baseline survey needed to occur. However, the desirable properties of CRTs come at a statistical cost. Since randomization in the CRT will occur at the village level, the study requires a large number of villages to attain sufficient statistical power to demonstrate impacts. Typically this will be more challenging, administratively and financially, than conducting a study randomizing at the individual level. More importantly, since the unit of analysis of the study is the individual (household) while the unit of randomization is the village, the dependence between individuals of the same village reduces the effective sample size (each additional individual in the same cluster adds little additional information, since outcomes are correlated among individuals within the same village), which must be accounted for when calculating the study sample size through power 73 | Page calculations. Therefore, only 20 randomly selected farmers from each village will be invited to participate in the study. To begin the evaluation, we will sample 50 jute growing villages and 50 mung bean growing villages within the Feed the Future intervention area in southern Bangladesh. The idea is to cluster randomize the availability of promotions, to observe whether these improve adoption or productivity at the household level. Ideally, we will be able to cross-randomize access to training; cross-randomization did occur for jute, but is still unclear whether randomization will be possible for the mung bean areas. Assuming that it is possible, the sample for each component will be broken up as follows: Training? Yes No Promotions? Yes T1 T2 No T3 C (Control group) This randomization was possible for the subsample of jute farmers; it will definitely also be possible to randomize promotions among mung bean farmers and at worst we will attempt to examine training using quasi-experimental methods. In each village, we will include 20 farmers at least initially who are supposed to be targeted for the interventions; e.g. farmers that fit any criteria given to us by AVC. If some of the households in the baseline survey retrospectively do not “look” appropriate for the intervention, based on either adjusted criteria or alternative criteria, we will plan to supplement both the treatment and control samples and stop following inappropriate households.26 With this framework, we can compare treatments with the control and potentially with other treatments, to learn about which is most effective; comparisons of training will likely take place across crops. We base power calculations in section 6 on this type of a framework. To actually design the sample, AVC shared a list of sites where their subcontractors potentially planned to provide training for both jute and mung beans, but where they had not yet begun to register farmers. Subcontractors listed approximately twice as many villages as they needed to train. Based on this input, we did an initial sampling of 50 villages (25 treatment to receive training, 25 control to receive no training) for each of jute and mung bean, and shared the list with AVC (for mung bean, villages have not yet been assigned to treatment groups). We collaborated with AVC on the final selection of sites, then – in the case of jute – AVC shared their training schedule so that we could administer our baseline survey within one month before training was initiated. In the case of jute, IFPRI then sampled for promotions. We sampled half of each of the training treatment and control villages for promotions, to identify villages that would receive: training and promotions, training and no promotions, no training and promotions, and no training and no 26 By criteria, for example, consider land holdings, age of the farmer, and number of years they have been growing the crop of interest. 74 | Page promotions. A similar approach is envisioned for mung bean. The final list of sampled villages for jute is provided in Appendix Table 1, and the list of mung bean villages is in Appendix Table 2. 4.2 LAB IN THE FIELD EXPERIMENTS In the first year of the impact evaluation, we will use lab-in-the-field experiments to analyze trust and trustworthiness among different actors in the market system, and test interventions to potentially improve signals of quality and lead to repeated interactions in the value chains. We are concentrating modeling on input markets. The basic set-up of the experiment is as follows. Farmers and local input sellers (some of whom may have been identified as retailers by the large input providers that collaborate with AVC) will be invited to come to a session in their village or union. A small group of farmers will be matched to a group of two input sellers. Input sellers will decide on a price at which to offer inputs, and will decide quality but will not reveal it to the farmer. Then farmers can choose which seller from which to buy inputs, or can refrain from purchasing from either of them. The joint payoff to either high or low quality inputs is the same; farmers gain more from high quality inputs, whereas input dealers gain more from either setting high prices or from volume sales. Farmers earn least when they buy and the input seller provides low quality, whereas the input seller earns most when providing low quality, although they can potentially make more by selling high quality to a larger number of farmers. The experiment is set up so that the farmer may not trust the input seller to provide high quality, and refrain from buying inputs, leading to inefficient outcomes. The goal of lab-in-the-field experiments is to test whether the interventions have a positive effect on trust in these retailers relative to other local input suppliers. The advantage of measuring trust in this way, as opposed to measuring trust in surveys, is that the choices made in the experiments are incentivized. That is, farmers and input sellers face real payoffs in the experiments; a farmer who buys from an untrustworthy input seller will for instance earn less from the experiment than a farmer who does not buy from an unworthy input seller. By making the payoffs real, participants will act more closely to how they would act in ‘real life’, which is a major advantage over hypothetical survey questions. After setting up the basic experiment, we will experiment with interventions to attempt to improve trust in the marketplace. First, we will allow input dealers to signal if they plan to sell the high quality input to farmers. The signal is not binding, but gives input sellers a method of telling farmers they plan to provide high quality inputs. In a second experiment, we will test allowing different types of promotional treatments, simulating to the type of promotions that AVC is conducting. The main advantage of doing so via an experimental approach is that the laboratory provides a test bed in which we can construct a market with the desired specifications and then sharply manipulate the factors we wish to study, to see how they interact and ultimately affect trade. In this way, we gain insight into what makes for a better trading system. As such, the experiments will help identify points in the value chain where, and conditions under which, actors lack trust in each other—whether that be between input sellers and farmers, or between farmers themselves (hindering group formation or effectiveness). Identifying such points will 75 | Page allow making recommendations on where and how to build trust along the beginning of the value chain. Of course, if the input market improves either through the interventions taking place, it does not necessarily mean that farmer welfare will improve, since the output markets could still work against farmer interests. We plan to work with AVC and USAID to consider whether it makes sense to add lab-in-the-field work for output markets in 2017, which would need to be designed. 5 OUTCOMES OF INTEREST The research questions posed in Section 3 will be answered using the following outcome variables: Input Purchases: We will measure input purchases through follow-up surveys. We will measure indicators for the purchase of any improved input (seed, fertilizer, precision services), as the total value and amount of specific inputs purchased, which will lead to the total value and amount of inputs purchased. We expect measures from the baseline and follow-up surveys to be better than administrative records. To be able to measure group level outcomes, we will include in surveys whether or not purchases were made through groups. Input Purchases on Credit: We will learn about how people purchased inputs, and whether or not they used credit. If farmers purchase on credit, they must trust that the inputs are genuine and the creditor must trust that the farmer will pay them back, so this measure will serve as a measure of trust. Repeat Purchasers, Same Seller: We want to also understand whether farmers purchase from the same seller year-to-year, since this behavior would also suggest a level of trust in the system. This measure also serves as an indirect measure of trust. Input Use: That said, it is not necessarily clear that once farmers have purchased inputs, that they will also use them. We will ask about input use on baseline and follow-up surveys, which will allow us to measure indicators for the use of any improved input (seed, fertilizer), the total value and amounts of specific inputs used as well as the total value and amount of overall inputs used. Agricultural Productivity: We will measure productivity primarily as the productivity for the specific crop of interest; e.g. either jute or mung beans. As a secondary measure, we will use the gross value of agricultural output, using procedures for generating prices for goods that are not sold well established in the development economics literature, which will reflect overall gains to the household rather than just specific gains within one crop. Output Sales: For the primary cash crops (jute and mung beans), we want to understand whether crop sales increase both for the specific crops for which AVC is developing groups, and to understand whether those sales were conducted through groups to improve prices received for the crops. 76 | Page Expenditures: We will measure expenditures through a simplified expenditure module in the household survey which will give us an approximate measure of changes in levels and composition of household expenditures. We will look at expenditures in order to assess whether the agricultural input use and the potential increase in productivity are reflected in increased consumption, and to determine whether access to savings through mobile money affects the composition of expenditures in the household, particularly moving away from investments in durable assets. Food Security: We will use measures of food security that are relatively standardized and were included in the survey. Food Frequency: One of the final outcome goals of the activity is to understand whether nutritional outcomes improve as a result of interventions. We expect that any nutritional outcomes among jute farmers can only occur through the “income” channel (e.g. through improvements in income), whereas for mung bean farmers they would occur through either the production channel (which makes mung beans cheaper for the household), or through the income channel. Pro-WEAI:We will measure men’s and women’s empowerment in agriculture using the project WEAI (pro-WEAI) indicators, which is being developed under the second phase of the Gender, Agriculture and Assets Project (GAAP2). As part of the GAAP2 portfolio, the impact evaluation associated with the AVC will collect information on different domains of empowerment in agriculture, including but not limited to, production decisions, access to productive capital, control over income, group membership, and time allocation. The WEAI methodology will allow us to identify the key domains where men and women are disempowered, as well as measure empowerment gaps between men and women within households. 6 STATISTICAL POWER The CRT sample includes 50 villages for the jute value chain and 50 villages for the mung bean value chain. Each village represents a single cluster and has a cluster size of 20 households. Jute samples villages are located in Faridpur, Jhenaidah, Madaripur, and Narail districts; mung bean sample villages are located in Barguna and Putuakhali districts. For jute, villages were first randomized into two groups: 25 villages that will receive training and 25 that will not. Villages were then additionally randomized into promotion and non-promotion groups, yielding the following four treatment groups: 1. Training + Promotion (13 villages) 2. Training + No Promotion (12 villages) 3. No Training + Promotion (13 villages) 4. No Training + No Promotion (12 villages) 77 | Page For mung beans, we will randomize in a similar way once interventions are chosen; at minimum there will be an intervention that covers promotional activities for mung beans, and a training type intervention may take place as well. This scenario is covered by the training only power calculations. In order to identify farmers to be surveyed within a village, survey teams conducted a census (or a household listing exercise) in each village of all available jute/mung bean farmers. We then used the census-generated roster to randomly select 20 households per village for our sample. All female-headed households were automatically included in our sample to ensure sufficient statistical power when answering our gender-related questions. Farmers with more than 500 decimals of land are not considered smallholder farmers and were hence excluded from the sample. We next conduct ex post power calculations to determine the minimum detectable treatment effect (MDE) that can be statistically detected given our predetermined sample size.27 Here we discuss MDEs in adoption of new technologies and practices, agricultural productivity, and household consumption, and across our treatment groups for the following comparisons: 1. Training versus non-training (pooling promotion and non-promotion groups) 2. Promotion versus non-promotion (pooling training and no training groups) 3. Training + promotion versus training-only or promotion-only 4. Winning a prize in the promotional raffle (per village, 10 out of 20 farmers receive a prize) 5. Winning 1st prize in the raffle vs not winning (2 treatment, 10 control farmers per village) 6. Winning 2nd prize in the raffle vs not winning (3 treatment, 10 control farmers per village) 7. Winning 3rd prize in the raffle vs not winning (5 treatment, 10 control farmers per village) The latter comparisons are valid for the jute growing groups; they may differ for mung bean groups depending upon the type of promotional activity that occurs, as raffles may not be chosen by PARTEX. Our discussion of statistical power begins with discrete variables (adoption of technologies and practices) before continuing with continuous variables (household consumption, agricultural productivity, and generic variables with a standard-normal distribution). Calculations in the latter were based on data from the 2011-2012 Bangladesh Integrated Household Survey (BIHS), a nationally representative survey of over 6,500 households that includes coverage of the Feed the Future zone of influence. At any given sample size, village-level randomization will generally exhibit lower statistical power when compared to individual-level randomization due to the existence of intra-cluster correlation (ICC). ICC is essentially a measure of the homogeneity of units (households) within the same cluster (village), and is inversely related to statistical power. Given our fixed sample size, this impact is transformed into larger MDEs. We assume an ICC of 27 The sample size was determined ahead of time considering the number of villages that were being included in activity expansion for each value chain; it was not possible to collect a much larger sample as a result. 78 | Page 0.1 and corresponding design effect of 2.9.28 The MDEs between comparison groups also depends on initial adoption rates of improved agricultural technologies and practices. We do not have preliminary data here and estimate the MDEs based on a range of possible initial adoption rates. The following table presents the estimated MDEs in each of our specified comparison groups:29 Table 3. Power Calculations, Discrete Variables, AVC Jute and Mung Bean Interventions Control adoption level 25 clusters/arm, 20 HH/cluster, ICC = 0.1 T: 10 ; C: 10 T: 2 ; C: 10 T: 3 ; C: 10 T: 5 ; C: 10 Training vs Non￾Training Promotion vs Non￾Promotion Training & Promotion vs Training/Promotion Only Win Prize 1st Prize 2nd Prize 3rd Prize ( - ) ( + ) ( - ) ( + ) ( - ) ( + ) ( + ) ( + ) ( + ) ( + ) 0.10 0.07 0.11 0.07 0.11 0.10 0.17 0.11 0.19 0.16 0.13 0.20 0.11 0.13 0.11 0.13 0.15 0.20 0.13 0.23 0.20 0.16 0.30 0.13 0.15 0.13 0.15 0.18 0.22 0.15 0.25 0.21 0.18 0.40 0.14 0.15 0.14 0.15 0.20 0.22 0.15 0.26 0.22 0.18 0.50 0.15 0.15 0.15 0.15 0.22 0.22 0.15 0.25 0.22 0.18 0.60 0.15 0.14 0.15 0.14 0.22 0.20 0.14 0.24 0.21 0.17 0.70 0.15 0.13 0.15 0.13 0.22 0.18 0.13 0.21 0.18 0.16 0.80 0.13 0.11 0.13 0.11 0.20 0.15 0.11 0.17 0.15 0.13 For example, if farmers who do not receive training have a 10 percent initial adoption rate of improved fertilizer, the impact evaluation is able to detect with 80 percent power a difference of at least 11 percentage points for training compared to non-training villages. If the non-training group’s initial adoption rate is 30 percent, we will be able to detect a difference of at least 15 percentage points with sufficient statistical power. Note that these results may be generalized to any binary outcome that we expect to impact with our intervention. Next, consider household consumption and agricultural productivity, both continuous variables. Household consumption is measured by monthly expenditure per capita in taka. Agricultural productivity is measured by yield in kilograms of output per cultivated acre of the target crop. We estimate the MDEs in the tables below, again assuming an ICC of 0.1 and taking standard deviations from BIHS data to be representative of our study population. Changes in crop yield are presented separately for jute and mung bean due to differences in productivity variance. The impact evaluation is able to detect an effect of training plus promotion, compared to only training, or only promotion, with 80 percent power if the bundled intervention increases yield by at least 1.76 kg/decimal compared to trainings or promotions 28 Source: 2011-2012 Bangladesh Integrated Household Survey (BIHS). An assumption of an ICC of 0.1 is quite standard in the literature on cluster corrected power calculations. 29 MDEs at the 90% control level are omitted from the table because adoption rates are generally lower and because the MDEs are often outside the feasible range. 79 | Page alone. Shifting our focus to mung bean, if farmers who receive both training and promotions have a yield that is at least 0.79 kg/decimal higher compared to training-only farmers or promotions-only farmers, we will be able to detect this change with 80 percent power. Table 4. Power Calculations, Jute and Mung Bean Yields Yield: Jute (kg/decimal) 25 clusters/arm, 20 HH/cluster, rho = 0.1 T: 10 ; C: 10 T: 2 ; C: 10 T: 3 ; C: 10 T: 5 ; C: 10 Mean, SD: (8.8, 3.87) Trainin g vs Non￾Trainin g Promotio n vs Non￾Promotio n Training & Promotion vs Training/Prom otion Only Win Prize 1st Prize 2nd Prize 3rd Prize units 1.19 1.19 1.76 1.17 2.04 1.74 1.44 % 13.54 13.54 20.00 13.32 23.23 19.73 16.33 Yield: Mung bean (kg/decimal) 25 clusters/arm, 20 HH/cluster, rho = 0.1 T: 10 ; C: 10 T: 2 ; C: 10 T: 3 ; C: 10 T: 5 ; C: 10 Mean, SD: (3.3, 1.73) Trainin g vs Non￾Trainin g Promotio n vs Non￾Promotio n Training & Promotion vs Training/Prom otion Only Win Prize 1st Prize 2nd Prize 3rd Prize units 0.53 0.53 0.79 0.52 0.91 0.78 0.64 % 16.14 16.14 23.85 15.88 27.70 23.52 19.47 The table for household consumption is interpreted in the same manner. The impact evaluation will be able to detect a change in monthly per capita household consumption of at least 662.43 Taka (representing about 27 percent of income) for farmers who win a prize in the raffle compared to farmers who enter the raffle but win nothing. This change is fairly substantial, and reflects the challenge of actually finding impacts on variables with larger variances, like consumption aggregates. Table 5. Power Calculations, Consumption Yields Household consumption (taka/person/month) 25 clusters/arm, 20 HH/cluster, rho = 0.1 T: 10 ; C: 10 T: 2 ; C: 10 T: 3 ; C: 10 T: 5 ; C: 10 Mean, SD: (2457.7, 2186.57) Training vs Non￾Training Promotion vs Non￾Promotion Training & Promotion vs Training/Prom otion Only Win Prize 1st Prize 2nd Prize 3rd Prize units 673.38 673.38 994.65 662.43 1155.18 980.99 812.08 % 27.40 27.40 40.47 26.95 47.00 39.91 33.04 Finally, we provide MDEs for a generic standard normal continuous variable. Table values represent the required difference in standard deviations between each respective comparison group. We can detect differences of at least 0.31 standard deviations when comparing training 80 | Page versus non-training and promotion versus non-promotion villages; differences of at least 0.45 standard deviations when comparing villages with promotions plus training versus villages with only trainings, or only promotions; and differences of at least 0.30 standard deviations when comparing prize winners with non-prize winners in the promotion villages. These are considered medium effect sizes; we believe they are achievable for outcomes for which the ratio of the standard deviation to the mean is not very large. Table 6. Power Calculations, General Continuous Variable Standard normal, continuous (# of SD) 25 clusters/arm, 20 HH/cluster, rho = 0.1 T: 10 ; C: 10 T: 2 ; C: 10 T: 3 ; C: 10 T: 5 ; C: 10 Mean, SD: (0, 1) Training vs Non￾Training Promotion vs Non￾Promotion Training & Promotion vs Training/Promotion Only Win Prize 1st Prize 2nd Prize 3rd Prize units 0.31 0.31 0.45 0.30 0.53 0.45 0.37 7 QUALITATIVE ANALYSIS Qualitative work will be completed after the first harvest post randomization and will be designed to focus on two objectives – (1) understanding the factors that affect the implementation and take-up of the intervention, and (2) understanding the different constraints to men’s and women’s empowerment in agriculture in this setting, which could also help identify entry points for engaging women in AVC interventions. Insights from qualitative work can potentially feed directly into programming if the results are suggestive of changes that would enhance impacts, as well as help elucidate whether completing additional modules in the endline survey would be worthwhile. Factors that can inform the implementation and design of the intervention Key informant interviews will be conducted with farmers, input suppliers, trainers and implementers to explore the following topics: • Farmer feedback on trainings, including perceptions on the effectiveness of trainings and challenges in participating • Farmer feedback on the gender and nutrition aspect of the trainings • Challenges faced by farmers who used or attempted to use improved seeds or cultivation practices • Challenges faced by subcontractors/trainers in conducting trainings • Challenges faced by input suppliers, aggregators, and implementers in conducting promotional campaigns • Reasons for non-adoption of improved seeds and/or cultivation practices • Factors that affect the diffusion of technology to other households in the same communities 81 | Page • Unintended effects of the interventions Understanding men’s and women’s empowerment in agriculture As a participant in GAAP2, the impact evaluation associated with the AVC is committed to carrying out qualitative research to complement and interpret the data collected through the pro￾WEAI module of the impact evaluation surveys. The qualitative work on understanding women’s empowerment will be conducted with the support of the GAAP2 qualitative team and will follow their qualitative research strategy, currently under development, which has two purposes – to validate the pro-WEAI and to develop evidence-based strategies for empowering women through agricultural programs. We will conduct key informant interviews, as well as two male and two female focus group discussions composed of individuals belonging to jute- and mung bean-growing households (4 FGDs in total). These interviews and discussions will explore the following themes: • Men’s/women’s perceptions on the most important dimensions of empowerment • The types of activities and decisions men/women participate in in the jute and mung bean value chains • How do men/women make decisions about what foods to consume • Changes in men’s/women’s time allocation over different seasons, whether these patterns have shifted in the last (x) years, and if so, why • Acceptability for women/men to take on new roles and tasks within the household and within agriculture, especially in households that experience labor shortages • Challenges faced by women in earning and controlling their own income • Reasons for women’s physical mobility restrictions, whether these have changed in the last (x) years and why 8 MAJOR TASKS, DELIVERABLES/MILESTONES, AND ESTIMATED TIMELINE (TENTATIVE) The key sowing seasons in southern Bangladesh are February and March for jute, and December and January for mung beans. The key project activities began with jute in February and March this year (2016), and large-scale promotional activities were timed to occur as farmers started sowing. The timing of specific and important project activities are described in Table 7. Table 7. Important Project Activities for Impact Evaluation Activity Timing Registration of jute farmers January 2016 Jute sowing season in Khulna/Jessore districts February – March 2016 Raffle tickets distributed March 2016 82 | Page Activity Timing NAAFCO-sponsored jute promotional events held March – April 2016 Registration of mung bean farmers TBD, 2016 Mung bean season in Barisal October 2016 – March 2017 Mung bean promotional events October- November, 2016 We structure survey and research activities around these seasons/activities, as well as around important dates in Bangladesh (e.g. Ramadan). IFPRI conducted a baseline survey in all villages in February-April 2016; we will thereafter break up surveys to mirror important seasons, to best measure agricultural outputs. Lab-in-the field experiments will take place during July and August of 2016. We will plan a jute midline in September-October 2016 (at which time we will field the pro-WEAI); we will plan a mung bean midline after mung bean sales in 2017. Qualitative work is currently scheduled for October 2016, and an endline will be conducted in September-October 2017 (as presently scheduled). The timing of specific research activities is described in Table 8. Table 8. IFPRI Research Activities Activity Timing Baseline Survey February-April 2016 Visit, pre-test and set up Lab-in-the-Field work May 2016 Lab experiments July – August 2016 Jute midline survey (includes pro-WEAI) August-September 2016 Qualitative Fieldwork Late September (after Eid)-October 2016 Final Impact Evaluation Survey August-September 2017 Finally, we present a task and deliverable schedule in Table 9. Table 9. Task and Deliverable Schedule 83 | Page Task Deliverable or Milestone Date of Deliverable Submission Complete Concept Note Final Concept Note 1 January 2016 Baseline survey instrument Baseline survey instrument 31 October 2015 Develop IE protocol First Draft IE Protocol 30 April 2016 Complete IE protocol Final IE protocol plus register project on AEA registry 31 May 2016 Develop Lab-in-the-Field Experiments Protocols for L-in-F Experiments 30 June 2016 Develop baseline report First draft baseline report 30 June 2016 Draft midline survey form Midline survey form 15 July 2016 Develop Qualitative Question Guide Qualitative Guide 31 August 2016 Complete baseline report Final baseline report 15 September 2016 Develop Qualitative Report Report on Qualitative Work+Lab-in-the-Field work 31 December 2016 Baseline data anonymized and ready for public release Data Release Plan 31 January 2017 Draft Midline Report (Jute) Draft midline report 28 February 2017 Finalize Midline Report (Jute) Final midline report 31 May 2017 Draft Endline Survey form Endline survey form 15 July 2017 Midline data anonymized and ready for public release Data Release Plan 31 January 2018 Draft Endline Report Draft Endline Report 31 March 2018 Final Endline Report Final Endline Report 30 June 2018 Final Data set anonymized and ready for public release Released data 30 September 2018 9 REFERENCES Jensen, Robert (2007). “The Digital Provide: Information (Technology), Market Performance and Welfare in the South Indian Fisheries Sector,” Quarterly Journal of Economics, 122(3), p. 879 − 924. Ali, Tariq Omar. 2012. The Envelope of Global Trade: The Political Economy and Intellectual History of Jute in the Bengal Delta, 1850s to 1950s. Doctoral Dissertation, Harvard University. ANNEX A: RANDOMIZED ASSOCIATIONS 84 | Page 85 | Page Appendix Table 1. Sampled Villages for Jute Value Chain by Training and Promotions District Upazila Union Village Trainings Promotions Faridpur Boalmari Satiore Koyra Yes Yes Faridpur Faridpur Sadar Koijuri Akunvatpara Yes Yes Faridpur Faridpur Sadar Koijuri Ghoradah Yes Yes Faridpur Boalmari Satiore Patitapara Yes No Faridpur Faridpur Sadar Koijuri Kasnail Yes No Faridpur Nagarkanda Talma Kathia Yes No Faridpur Faridpur Sadar Koijuri Muralidaho No Yes Faridpur Faridpur Sadar Koijuri Sreedhorpur No Yes Faridpur Nagarkanda Ramnagar Kunjanagar No Yes Faridpur Boalmari Goshpur Ralandia No No Faridpur Faridpur Sadar Majchar Bakchar No No Faridpur Nagarkanda Talma Rosulpur No No Jhenaidah Horinakundu Capashatia Valki Yes Yes Jhenaidah Jhenaidah Sadar Horisankorpur Chadrajani Yes Yes Jhenaidah Shailakupa Digngar Achintapur Yes Yes Jhenaidah Shailakupa Umedpur Bistupur Yes Yes Jhenaidah Horinakundu Chadpur Hamirhati Yes No Jhenaidah Jhenaidah Sadar Horisankorpur Norohoridra Yes No Jhenaidah Jhenaidah Sadar Horisankorpur Poylanpur Yes No Jhenaidah Shailakupa Digngar Horora Yes No Jhenaidah Horinakundu Chadpur Zadobpur No Yes Jhenaidah Jhenaidah Sadar Horisankorpur Sitarampur No Yes Jhenaidah Shailakupa Monohorpur Mohishadanga No Yes Jhenaidah Shailakupa Monohorpur Sondah No Yes Jhenaidah Horinakundu Capashatia Kesmot No No Jhenaidah Jhenaidah Sadar Dogachi Goalpara Putia No No Jhenaidah Jhenaidah Sadar Dogachi Mdhunathpur No No Jhenaidah Shailakupa Umedpur Umedpur No No Madaripur Rajoir Pouroshava Vanna Bari Yes Yes Madaripur Shivchar Sheroil Kulatola Yes Yes Madaripur Shivchar Sheroil Palpara Yes Yes Madaripur Rajoir Pouroshava Chouari Bari Yes No Madaripur Shivchar Sheroil Solapur Yes No Madaripur Rajoir Khalia Macha Rong No Yes Madaripur Rajoir Khalia Soro Mongol Pashim Par No Yes Madaripur Shivchar Sheroil Sadeka Bad No Yes Madaripur Rajoir Badar Pasa Umar Khali No No Madaripur Shivchar Sheroil Cawar Hat No No Madaripur Shivchar Sheroil Char Kakoir No No Narail Narail Sadar Mulia Bongram Yes Yes Narail Narail Sadar Mulia Durbajuri Yes Yes Narail Narail Sadar Mulia Korgram Yes Yes Narail Narail Sadar Mulia Baliadanga Yes No Narail Narail Sadar Mulia Goalbari Yes No Narail Narail Sadar Mulia Shaliarvita Yes No Narail Narail Sadar Mulia Echorbaha No Yes Narail Narail Sadar Mulia Hijoldanga No Yes Narail Narail Sadar Mulia Basvita No No Narail Narail Sadar Mulia Borendar No No Narail Narail Sadar Mulia Goaldanga No No 86 | Page Appendix Table 2. Sampled Villages for Mung Bean Value Chain by Training and Promotions District Upazila Union Village Barguna Amtali Amtali Dakshin and Poshchim Amtoli Barguna Amtali Amtali Mohisdanga Barguna Amtali Amtali Nilgonj Barguna Amtali Amtali Uttor & Dakshin Nachnapara Barguna Amtali Gulishakhali Angulkata Barguna Amtali Gulishakhali Baingunia Barguna Amtali Gulishakhali Bazarkhali Barguna Amtali Gulishakhali Dalachara Barguna Amtali Gulishakhali Fokirkhali Barguna Amtali Gulishakhali Goskhali Barguna Amtali Gulishakhali Gulishakhali Barguna Amtali Gulishakhali Horidrabaria Barguna Amtali Gulishakhali Kalibari Barguna Amtali Gulishakhali Khekuani Barguna Amtali Gulishakhali Kolagachia Barguna Amtali Kukua Amragachia Barguna Amtali Kukua Horimirtunjoy Barguna Amtali Kukua Kalipura Barguna Amtali Kukua Keorabunia Barguna Amtali Kukua Khakdan Barguna Amtali Kukua Krishnanagar Barguna Amtali Kukua Kukua Barguna Amtali Kukua Poshchim & Purbo Chunakhali Barguna Amtali Kukua Purbo kukua Barguna Amtali Kukua Raybala Putuakhali Dosmina Bashbaria Bashbaria Putuakhali Dosmina Bashbaria Char Hosnabad Putuakhali Dosmina Bashbaria Dakkin Daspara Putuakhali Dosmina Bashbaria Dhandania Putuakhali Dosmina Bashbaria Gasani Putuakhali Dosmina Betagi Sankipur Betagi Putuakhali Dosmina Betagi Sankipur Betagi Sankipur Putuakhali Dosmina Betagi Sankipur Borogopaldi Putuakhali Dosmina Betagi Sankipur Chinguria Putuakhali Dosmina Betagi Sankipur Dabari Betagi Putuakhali Dosmina Betagi Sankipur Jafarabad Betagi Putuakhali Dosmina Betagi Sankipur Kharija Betagi (Kamrabaz) Putuakhali Dosmina Betagi Sankipur Rambollov Putuakhali Dosmina Betagi Sankipur Shamermardana Putuakhali Dosmina Betagi Sankipur Sohertaluk Putuakhali Dosmina Dosmina Arojbegi Uttar and Dakshin 87 | Page Putuakhali Dosmina Dosmina Charhadi Uttar and Dakshin Putuakhali Dosmina Dosmina Dosmina Putuakhali Dosmina Dosmina Hajirhut Putuakhali Dosmina Dosmina Katakhali Putuakhali Dosmina Dosmina Kaunia Putuakhali Dosmina Dosmina Nijabad Putuakhali Dosmina Dosmina Poshchim & Purbo Laxmipur Putuakhali Dosmina Dosmina Sayeed Jafor Putuakhali Dosmina Dosmina Uattar Laxmipur 88 | Page 8.5 MIDLINE QUESTIONNAIRE Bangladesh Agricultural Value Chain (AVC) Impact Evaluation Midline Survey with Jute Farmers and Input Sellers (January – February 2017) Survey designed and supervised by: International Food Policy Research Institute (IFPRI) Survey administered by: Data Analysis and Technical Assistance Limited (DATA) HOUSEHOLD QUESTIONNAIRE 1. Table of Contents FEMALE FORM MODULE B: HOUSEHOLD ROSTER (ALL GROUPS) CONSENT OF RESPONDENT MODULE J: CONSUMPTION (ALL GROUPS) MODULE K: HOUSEHOLD ILLNESSES (GROUP 2 ONLY) MODULE G2: ROLE IN HOUSEHOLD DECISION-MAKING AROUND PRODUCTION AND INCOME (GROUP 2 ONLY) MODULE G3(A): ACCESS TO PRODUCTIVE CAPITAL (GROUP 2 ONLY) MODULE G3(B): ACCESS TO FINANCIAL SERVICES (GROUP 2 ONLY) MODULE G4: TIME ALLOCATION (GROUP 2 ONLY) MODULE G5: GROUP MEMBERSHIP (GROUP 2 ONLY) MODULE G6. PHYSICAL MOBILITY (GROUP 2 ONLY) MODULE G7: INTRAHOUSEHOLD RELATIONSHIPS (GROUP 2 ONLY) MODULE G8(A): AUTONOMY IN DECISION-MAKING (GROUP 2 ONLY) MODULE G8(B): NEW GENERAL SELF-EFFICACY SCALE (GROUP 2 ONLY) MODULE G8(C): LIFE SATISFACTION (GROUP 2 ONLY) MODULE G9. ATTITUDES ABOUT DOMESTIC VIOLENCE (GROUP 2 ONLY) MODULE X: CLOSING MODULE C: AGRICULTURE (ALL GROUPS) MODULE C1: PRODUCTION OF CROPS OTHER THAN JUTE (GROUPS 1 & 2 ONLY) MODULE C2: JUTE PRODUCTION IN KHARIF-1 (AUS) 2016 (GROUP 1 & GROUP 2 ONLY) MODULE C3: JUTE PRODUCTION COSTS: INPUTS (ALL GROUPS) MODULE C4 KNOWLEDGE AND KNOWLEDGE SHARING OF FERTILIZERS (ALL GROUPS) MODULE C5: EXTENSION SERVICES FOR JUTE PRODUCTION (ALL GROUPS) MODULE C6: USE OF IMPROVED TECHNOLOGIES AND PRACTICES IN JUTE PRODUCTION (ALL GROUPS) MODULE D: MARKETING 89 | Page MODULE D2: RELATIONSHIPS WITH LOCAL JUTE INPUT SELLERS (GROUP 1 & GROUP 2) / BUYERS (GROUP 3) MODULE E: HOUSING (ALL GROUPS) MODULE G: HOUSEHOLD EXPENDITURES ON GOODS OTHER THAN FOOD (ALL GROUPS) MODULE G1: MONTHLY RECALL MODULE G2: ANNUAL RECALL (ALL GROUPS) MODULE I: BELIEFS (ALL GROUPS) MODULE I1A: EXPECTATIONS FOR JUTE PRODUCTION AND SALES (ALL GROUPS) TRUST GAME (ALL GROUPS) END OF MAIN FORM MALE FORM MODULE G2: ROLE IN HOUSEHOLD DECISION-MAKING AROUND PRODUCTION AND INCOME (GROUP 2 ONLY) MODULE G3(A): ACCESS TO PRODUCTIVE CAPITAL (GROUP 2 ONLY) MODULE G3(B): ACCESS TO FINANCIAL SERVICES (GROUP 2 ONLY) MODULE G4: TIME ALLOCATION (GROUP 2 ONLY) MODULE G5: GROUP MEMBERSHIP (GROUP 2 ONLY) MODULE G6. PHYSICAL MOBILITY (GROUP 2 ONLY) MODULE G7: INTRAHOUSEHOLD RELATIONSHIPS (GROUP 2 ONLY) MODULE G8(A): AUTONOMY IN DECISION-MAKING (GROUP 2 ONLY) MODULE G8(B): NEW GENERAL SELF-EFFICACY SCALE (GROUP 2 ONLY) MODULE G8(C): LIFE SATISFACTION (GROUP 2 ONLY) MODULE G9. ATTITUDES ABOUT DOMESTIC VIOLENCE (GROUP 2 ONLY) 90 | Page Respondents The midline questionnaire will be administered to three distinct groups of households: a) Jute farmer households interviewed at baseline (Group 1) b) Jute farmer households not interviewed at baseline (Group 2) c) Agricultural input seller households (Group 3) For Group 1 households: There are 20 households per village with whom we did the baseline survey. 1. Jute farmer still cultivating jute and still in the same village: Interview the jute farmer and his household, even if he split from the baseline household and now lives in another household. 2. Jute farmer still in the same village but no longer cultivating jute, or moved to another village: Interview the member of the baseline household who is now jute farming (if nobody is jute farming then interview the primary decision-maker when it comes to agriculture). 3. Only if the entire household moved away from the village, then we can mark that household as ‘cannot be interviewed’ because of the reason ‘moved away’ For Group 2 households: In Group 2, we should consider two groups of villages: Raffle villages and non-raffle villages. In raffle villages, farmers entered a raffle in which NAAFCO gave out discounts on fertilizer (20%, 50% and 80%). In all raffle villages, raffle tickets were distributed among all 20 households from the baseline (Group 1) as well as 10 additional households who did not participate in the baseline, but who satisfied the following criteria: 1. At least 66 decimals of land at the time of the census 2. Agriculture is the main source of income 3. The farmer is between 18 and 65 years old These additional farmers are “Group 2” in the raffle villages and we will interview them in the current survey. For that, follow the household that received the raffle ticket, not the individual, then identify the jute farmer. If no longer in the same village, then take replacement households. Also take replacement households when there are less than 10 households that you can interview, so that we have on average 10 households per village. For that, we will provide a list of replacement farmers within the village and a list of replacements from other villages for cases in which you deplete the own-village replacement list. We also have non-raffle villages. In these villages, we will conduct interviews with 10 households who were not in the baseline. These households should satisfy the same criteria as the farmers in the raffle villages (at least 66 decimals of land at the time of the census, agriculture being the main source of income, and the farmer being between 18 and 65 years old). However, we do not know which households satisfy the last two criteria. Hence, before interviewing Group 2 households in non-raffle villages, the enumerators will always have to ask: (1) is agriculture the main source of income, and (2) is the farmer between 18 and 65 years old. For Group 3 households: Interview the person who participated in the lab in the field experiment. If the owner of the enterprise sent a representative to the field experiment, that individual should be the person interviewed. You will be prompted to specify whether the respondent is the owner of the business or their representative. If you cannot find or interview the person who participated in the lab-in-the-field experiments, then interview the most knowledgeable available person, and indicate why you are interviewing someone else. Structure The survey is divided into three forms: (1) a main form which should be asked to the primary jute farmer (Group 1 & Group 2 households) or to the input seller (Group 3 households); (2) a male form which should be asked to the respondent of the main form (if male) or another male household member (if respondent main form is female); and (3) a female form, which should be asked to the spouse of the main form respondent (if male) or to the main form respondent (if female). If there is no female in the household, then interview the person responsible for cooking. Always start with the female form. Not all parts of the survey will be asked to all respondents. All households should complete the main form and female form. Group 1 households do not complete some sections of the main form or female form, and do not complete the male form. Group 2 households complete all sections of all three forms. Group 3 households do not complete some sections of the main form or female form and do complete the male form. Under normal circumstances, the interviewer should first complete the female form with the appropriate 91 | Page respondent, then complete the main form with the jute farmer or input seller, then complete the male form with the appropriate male respondent (Group 2 & Group 3 households only). Form Module Title Group 1 Baseline Jute Farmers Female Form (Respondent: Respondent main form if female; if not, interview spouse of main form respondent) B- Household roster x J- Consumption x K- Household Illnesses WEAI- Empowerment & Decision-making Main Form (Respondent: Jute farmer / Input seller) A- Coversheet & Respondent Consent x C1 - Agriculture x C2 – Jute Production x C3 – Jute Inputs x C4 – Knowledge and Knowledge Sharing x C5 – Extension Services for Jute Production x C6 – Use of Improved Technologies and Practices in Jute Production x D2- Relationships with Input Sellers/Farmers x E- Housing x G- Non-food Expenditures x I1- Expectations for Jute Yield x Trust Game x Male Form (Respondent: Respondent main form if male; if not, interview other male household member) WEAI- Empowerment & Decision-making A number of modules require farmers (input sellers) to indicate for different input sellers (farmers) whether they have ever done business with that person. For that purpose, we will prepare photo albums at the union/district level (TBD in consultation with the lab-in-the-field team). For Group 2, we will not yet have pictures; their pictures are only needed for the input sellers, so we will start in a union with interviewing Groups 1 & 2, then print the photo album of farmers, and then interview the input sellers. ◻ FEMALE FORM MODULE A: HOUSEHOLD IDENTIFICATION (ALL GROUPS) Variable IDENTIFICATION Response date DATE interviewer TEAM [SUPERVISOR NAME/CODE] & INTERVIEWER [NAME/CODE] hhid HOUSEHOLD IDENTIFICATION NUMBER [NUMERICAL] Household verification (1): HOUSEHOLD INFORMATION FOR ID ${hhid} A_02 DISTRICT / UPAZILA [CODED] This information wi Interviewers should they are interviewing If the preload informa respondent, or if ther person to be interview with the interview an A_03 UNION [CODED] A_04 VILLAGE [CODED] HOUSEHOLD HEAD: A_08 NAME A_12 GENDER HOUSEHOLD HEAD’S FATHER A_10 NAME A_05 HOUSEHOLD LOCATION/LANDMARK [STRING] MAIN RESPONDENT IN PREVIOUS ROUND [PRELOADED N 92 | Page CHECK THE ABOVE INFORMATION CAREFULLY AND CHOOSE "YES" TO PROCEED IF YOU ARE SURE YOU ARE AT THE RIGHT HOUSEHOLD. ◻ 1 – Yes A_06 CAN YOU START THE INTERVIEW? ◻ 1 – Yes -> A_09 * YOU CANNOT PROCEED UNLESS YOU HAVE FOUND THE CORRECT HOUSEHOLD. EITHER SPEAK TO THE RESP CANCEL THE INTERVIEW AND CONTINUE TO LOOK FOR THE RIGHT HOUSEHOLD. A_07 WHY CAN YOU NOT START THE INTERVIEW? AFTER SELECTING AN ANSWER -> END ◻ 1 – Moved ◻ 2 – Not found ◻ 3 – Refused ◻ 4 – Unavailable ◻ 5 – Other (speci RESPONDENT CHECK resp_check Is the main respondent (jute farmer / input seller) still [PRELOADED NAME]? If the family stopped producing jute / selling, but the main respondent is still present in the household, select “Yes”. Only select “No” if the person is no longer doing that activity and another household member took over Yes / No resp_why [IF resp_check == NO] Why is [PRELOADED NAME] no longer the main respondent? 1. Deceased 2. Health/illness/injury 3. Moved out of this vil resp_alt [IF resp_check == NO] Who is now the jute farmer/input seller in this household? relation_alt [IF resp_check == NO] How is ${resp_alt} related to [PRELOADED NAME]? 1. Spouse/Partner 2. Son/daughter 3. Parent 4. Grandchild 5. Grandparent 6. Employer 7. Employee 8. Other relation, spe 9. Other, specify MAIN FORM RESPONDENT resp_ml If resp_check = Yes, respondent is the listed individual If resp_check = No, respondent is the person recorded under ${farmer_alt} Household verification (2): PLEASE CONFIRM WHETHER THE FOLLOWING IS STILL CORRECT. YOU CAN COLLECT CO INCORRECT. 93 | Page A_21 Name of the household head Yes / No A_09 Head’s National ID #: … Yes / No phone Phone number: …. Yes / No A_22 Name of the household head’s father Yes / No A_23 Household location / landmark Yes / No RECORD CORRECTIONS FOR ITEMS SELECTED “NO”. phone2 Is there another phone number we can use to contact you? ENTER 9999 IF NO 94 | Page ◻ Module B: Household Roster (All groups) Instructions for Group 1 Households The survey program will load a list of members who were listed in the household at baseline. For each member, the respondent will be asked to confirm if they are still a member of the household. The respondent should provide the number of additional members (for example: newborns) who were not listed in the baseline survey. The surveyor should then complete one row for each member. After completing the list of names, you will be prompted to select the male and female form respondents from the list of household members. Instructions for Group 2 & Group 3 Households Group 2 & Group 3 households will not have any roster information loaded in the program. One row of questions B1_01-B1_24 should be completed for each of the members listed under B1_00_1. Question B1_00_2 is not asked to Group 2 or Group 3 households. After completing the list of names, you will be prompted to select the male and female form respondents from the list of household members. Household definition A household is a group of individuals who eat from the same pot, and sleep under the same roof. Someone is a member of the household if he or she was in the household at least 3 out of the last 6 months, and at least 4 days per week. Exceptions: if someone just moved out with the intention of not moving back soon, then this is not a household member. If someone recently moved in or was born, and the intention for the person is to stay in the household, then this is a household member. hhid Please enter the household ID A_13 Religion of the household head ◻ 1 – Muslim ◻ 2 – Hindu ◻ 3 – Christian ◻ 4 – Other A_14 Ethnic group of the household head ◻ 1 – Bengali ◻ 2 – Bihari ◻ 3 – Tribal (specify) ◻ 4 – Other (specify) B1_00_1 Groups 2 & 3 only How many people are currently members of this household? B1_00_2 Group 1 Households only Please select anyone who is NOT currently a member of the household ◻ Member 1 ◻ … B1_00_3 Group 1 Households only How many household members were NOT included in this list? (include any adults or children who are members of the household but were not listed) roster_list Please list the names of all the household members For Group 1 Households, the names of returning members will be displayed automatically (who are not excluded from the previous list of baseline members) Member 1 Member 2 … main_resp From the list of members, please select the respondent for the MAIN FORM Group 1 & Group 2 Households- Select the jute farmer. Group 3 Households- Select the input seller ◻ Member 1 ◻ Member 2 ◻ Member 3 ◻ Member 4 … female_resp From the list of members, please select the respondent for the FEMALE FORM ◻ Member 1 95 | Page Group 1 & Group 2 Households- If the jute farmer is male, select their spouse or a female adult relative. If the jute farmer is female, select the jute farmer. Group 3 Households- If the input seller is male, select their spouse or a female adult relative. If the input seller is female, select the input seller. ◻ Member 2 ◻ Member 3 ◻ Member 4 … male_resp From the list of members, please select the respondent for the MALE FORM Group 1 & Group 2 Households- If the jute farmer is male, select the jute farmer. If the jute farmer is female, select her male spouse or adult male relative. Group 3 Households- If the input seller is male, select the input seller. If the input seller is female, select her male spouse or adult male relative ◻ Member 1 ◻ Member 2 ◻ Member 3 ◻ Member 4 ◻ … THE FEMALE RESPONDENT IDENTIFIED (${female_resp}) SHOULD BE THE PERSON WHO ANSWERS ALL SUBSEQUENT QUESTIONS IN THIS FORM. THE JUTE FARMER / INPUT SELLER SHOULD BE THE PERSON WHO ANSWERS THE MAIN FORM 96 | Page Consent of Respondent Good morning/afternoon. I am ________ from the Data Analysis and Technical Assistance Limited (DATA), a Bangladeshi research organization based in Dhaka. Together with the International Food Policy Research Institute (IFPRI), we are conducting a survey that will provide IFPRI with necessary information to carry out research that is designed to help promote the welfare of Bangladeshis; particularly, to improve food consumption and nutrition of the people and women’s status, and to enhance agricultural development and income generation. Your household has been chosen by a random selection process. We are inviting you to be a participant in this study. We value your opinion and there are no wrong answers to the questions we will be asking in the interview. We will use approximately 3-4 hours of your time to collect all the information. If you prefer, we can do the interview in two visits. There will be no cost to you other than your time. There will be no risk as a result of your participating in the study. Your participation in this research is completely voluntary. You are free to withdraw your consent and discontinue participation in this study at any time. This study is conducted anonymously. You will only be identified through code numbers. Your identity will not be stored with other information we collect about you. Your responses will be assigned a code number, and the list connecting your name with this number will be kept in a locked room and will be destroyed once all the data has been collected and analyzed. Any information we obtain from you during the research will be kept strictly confidential. Your participation will be highly appreciated. The answers you give will help provide better information to policy-makers, practitioners and program managers so that they can plan for better services that will respond to your needs. The researcher read to me orally the consent form and explained to me its meaning. I agree to take part in this research. I understand that I am free to discontinue participation at any time if I so choose, and that the investigator will gladly answer any question that arise during the course of the research. Contact Person: Name of the Principal Investigator (PI): _____________ Address: Tel: __________; E-mail of PI: _________ Signature of the Enumerator: _____________ Date: /_____/_____/_____/ consent_f Do you agree to be interviewed for the purposes of this study? fem_alone Ability to be interviewed 1. Alone 2. With adult female members present 3. With adult male members present 4. With adults of both sexes present 5. With children present 6. With adults of both sexes and children present 97 | Page Membe r ID Relation to MALE form responden t Relation to FEMALE form responden t Relation to MAIN form respoden t Preloaded Gender Preloade d Age Preloaded Marital status If B1_05a != 1 Who chose your current spouse for you? Preloaded Age at marriage Preloaded Level of literacy Currently attending school or college? Hours spent in school (last 7 days) 1. Male 2. Femal e (WHOLE YEARS) 1. Unmarried (never married) 2. Married 3. Widow/widowe r 4. Divorced Separated 1. Chose each other 2. You chose them and they agreed 3. Family arrange d and you agreed 4. Family arrange d and you did not agree 5. They chose and you agreed 6. They chose and you did not agree (WHOLE YEARS) 1. Can read and write 2. Can read only 3. Can sign only 4. Cannot read or write 1. Yes 2. No NO B_11 (HOURS) B1_02_1 B1_02_2 B1_02_3 B1_03 B1_04 B1_05a B1_05b B1_06 B1_07 B1_08 B1_09 1 2 … 98 | Page Member ID Hours spent doing homework (last 7 days) Level of education Did he/she work for income, profit, or family gain in the last 12 months? Main source of income or profit in the last 12 months IF DID NOT EARN INCOME, SELECT “NON￾EARNING OCCUPATION” Describe main source of income or profit in the last 12 months Secondary source of income or profit in the last 12 months Describe secondary source of income or profit in the last 12 months (HOURS) ${CODE2} 1. Yes 2. No NO  NEXT MEMBER 1. Farming 2. Wage laborer 3. Salaried worker 4. Self-employment 5. Trade 6. Production 7. Livestock/poultry 8. Non-earning occupation 9. Other (specify) 10. Don’t know ${CODE3} 1. Farming 2. Wage laborer 3. Salaried worker 4. Self￾employment 5. Trade 6. Production 7. Livestock/poultry 8. Non-earning occupation 9. Other (specify) 10. Don’t know ${CODE3} B1_10 B1_11 B1_12 B1_13 B1_14 B1_16 B1_17 1 2 … Member ID Number of months spent doing main income activity (last 12 months) Hours spent doing main income activity (last 7 days) Number of months spent doing secondary income activity (last 12 months) Hours spent doing secondary income activity (last 7 days) (MONTHS) (HOURS) (MONTHS) (HOURS) B1_18 B1_20 B1_22 B1_24 1 2 … 99 | Page CODE1: Relationship CODE2: Education CODE3: Occupation Relationship with primary respondent 11. Husband/wife 12. Son/daughter 13. Grandson/granddaughter 14. Father/mother 15. Brother/sister 16. Niece/Nephew 17. Cousin 18. Father-in-law/mother-in￾law 19. Daughter-in-law/son-in-law 20. Brother-in-law/sister-in-law 21. Wife’s niece/nephew 22. Wife’s cousin 23. Other relative 24. Other non-relative 25. Permanent servant 1. Never attended school 2. Preschool (before class I) 3. Reads in class I 4. Completed class I 5. Completed class II 6. Completed class III 7. Completed class IV 8. Completed class V 9. Completed class VI 10. Completed class VII 11. Completed class VIII 12. Completed class IX 13. Completed Secondary School/Dakhil 14. Completed Higher Secondary/Alim 15. BA/BSC pass/Fazil Wage Laborer 1. Agricultural day labor 2. Earth work (govt. program) 3. Sweeper 4. Scavenger 5. Tea garden worker 6. Construction laborer 7. Factory worker 8. Transport worker 9. Apprentice 97. Other (specify) Salaried worker 10. Government/ parastatal 11. Service (private sector) 12. NGO worker 13. House maid 14. Teacher (GoB Primary school) 15. Teacher (Non-GoB Primary school) 16. Teacher (GoB High school) 17. Teacher (Non-GoB High school) 18. Teacher (college, university) 97. Other (specify) Self-employment 19. Rickshaw/van pulling 20. Driver of motor vehicle 21. Tailor/seamstress 22. Blacksmith 23. Potter 24. Cobbler 25. Hair cutter 26. Clothes washer 27. Porter 28. Goldsmith/silversmith 29. Repairman (appliances) 30. Mechanic (vehicles) 31. Plumber 32. Electrician 33. Carpenter 34. Mason 35. Doctor 36. Rural physician 37. Midwife 38. Herbal doctor/Kabiraj 39. Engineer 40. Lawyer/deed writer/Moktar 41. Religious leader 42. Lodging master 43. Private tutor/house tutor 44. Beggar 97. Other (specify) Trader 45. Small INPUT trader (roadside stand or stall) 46. Small NON-INPUT trader (roadside stand or stall) 47. Medium INPUT trader (shop or small store) 48. Medium NON-INPUT trader (shop or small store) 49. Large INPUT trader (large shop or wholesale) 50. Large NON-INPUT trader (large shop or wholesale) 51. Fish Trader 52. Contractor 97. Other (specify) Production 53. Food Processing 54. Small industry 55. Handicrafts 97. Other (specify) Livestock/poultry 56. Milk collector 57. Livestock Vet medicine seller 58. Livestock Feed supplier 59. Commercially feed producer 60. Animal Breeder 61. Veterinary/paravet doctor 97. Other (specify) Farming 62. Working own farm (crop) 63. Share cropper/tenant 64. Homestead farming 65. Fisherman (using non owned/not leased water body) 66. Raising fish / fish pond 67. Raising poultry 68. Raising livestock 69. Dairy production/ dairy farming 97. Other (specify) Non-earning occupation 70. Caring for children 71. Caring for sick/disabled family member(s) 72. Caring for elderly family member(s) 73. Housework 74. Helping on family farm 75. Helping with family INPUT business 76. Helping with family NON-INPUT business 97. Other (specify) Don’t know 100 | Page No. Enumerator Instructions Male form roster Group 2 Households ENUMERATOR: You have now completed the household roster for the female form. This information will later be used in completing the male form for this household. From this point onward, you should not make changes to the household roster. If you do need to make a change, you must re-complete the household roster section, copy down the updated information, then re-complete modules G2 onward. Do not edit the roster section then skip forward- doing so may cause serious data discrepancies and you may be required to redo the interview. The following information is a summary of the name, age, sex & PID of each member listed in the roster. Using the sheet provided make a careful copy of this information. This information should then be entered in to the first section of the Male Form to identify the household members. Please take care to ensure all data is copied and re-entered correctly. PID NAME AGE SEX Is male respondent? Is female respondent? 1 Member 1 32 M Yes No 2 Member 2 28 F No Yes 3 Member 3 61 F No No 4 Member 4 3 M No No … Please make a careful note of this information, ensuring that each field is copied correctly. Then use this information to complete the coversheet for relevant Male/Female forms 101 | Page Module J: Consumption (All Groups) MODULE J SHOULD BE ASKED TO THE MEMBER OF THE HOUSEHOLD WHO IS MOST INVOLVED IN COOKING FOOD FOR THE FAMILY. FOR HOUSEHOLDS WITH ADULT MALE AND ADULT FEMALE MEMBERS, THIS WILL TYPICALLY BE THE ADULT FEMALE SPOUSE OF THE MAIN RESPONDENT. IF THIS INDIVIDUAL IS UNAVAIBLE, THE SECTION MAY BE COMPLETED BY THE MAIN RESPONDENT. In the last 7 days, did your household consume any of the following types of food? 102 | Page Cereals, pulses, and edible oils Item name Item no. In the last 7 days, what types of cereals and pulses did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Course rice (parboiled) 1 Coarse rice (non-parboiled) 2 Fine rice 3 Rice flour 4 Atta 5 Semai/noodles 6 Chira (flattened rice) 7 Muri/Khoi (puffed rice) 8 Other 9 Lentil 10 Anchor daal 11 Mung bean 12 Soybean 13 Groundnuts 14 Mustard 15 Soybean oil 16 Mustard oil 17 103 | Page Non-leafy vegetables Item name Item no. In the last 7 days, what types of non-leafy vegetables did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Bitter gourd 18 Eggplant 19 Tomato 20 Sweet gourd 21 Water gourd 22 Bottle gourd 23 Pumpkin 24 Cucumber 25 Sheem 26 Radish 27 Cauliflower 28 Green banana 29 Green chili 31 Kachu (arum) 32 Potato 33 Onion 34 Garlic 35 Cabbage 36 Leafy vegetables Item name Item no. In the last 7 days, what types of leafy vegetables did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Shalgom 37 Pui (Indian spinach) 38 Lal Shak (red amaranth) 39 Bathua 40 Kachu Shak 41 Lau Shak 42 Dhania Shak 46 104 | Page Palang Shak (spinach) 47 Onion/garlic stalk 48 Radish leaves 49 Mixed leafy vegetables 50 Meat, fish, eggs, and dairy Item name Item no. In the last 7 days, what types of meat, fish, eggs, and dairy did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Beef/buffalo 51 Chicken 52 Fish (large) 53 Fish (small) 54 Egg 55 Milk 56 Powdered Milk 57 Fruit Item name Item no. In the last 7 days, what types of fruit did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Mango 58 Banana 59 Papaya 60 Orange 61 Apple 62 Coconut 63 Jack Fruit 64 Litchis 65 Black berry 66 Bel 67 Pomelo 68 Grapes 69 105 | Page Amra 70 Karambola 71 Guava 72 Jujube/dried jujube 73 Olive 74 Tamarind 75 Dalim 76 Lemon 77 Dates 78 Sugarcane 79 Green Coconut 80 Ata (bullock’s heart) 81 Chalta 82 Tarmuj (Watermelon) 83 Bangi (Musk melon) 84 Pineapple 85 Sobeda 86 Jaamrul 87 Myrobalan/ Indian Gooseberry 88 Water Caltrop 89 Spices and condiments Item name Item no. In the last 7 days, what types of spices and condiments did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Dried chili 90 Turmeric (dried) 91 Jira 92 Salt 93 Panchforan 94 Coriander 95 Ginger 96 Garam Masala 97 106 | Page Tejpata 98 Sugar 99 Gur 100 Tea leaves 101 Packaged food and drink Item name Item no. In the last 7 days, what types of packaged food and drink did your household consume? How many days (out of the last 7 days)? Quantity per day (Average) Unit J1_01 J1_02 J1_03 J1_04 J1_05 Tea- prepared 102 Coke/Pepsi, etc. 103 Packaged Juice 104 Biscuit 105 Cake 106 Chips 107 Chocolate 108 Chewing gum 109 Sweets 110 Chanachur 111 Tobacco 112 Betel Leaf 113 Supari 114 Dishes prepared outside the home Item name Item no. In the last 7 days, what types of dishes prepared outside the home did your household consume? How many days (out of the last 7 days)? Number of dishes/pieces per day Price paid per dish/piece J1_01 J1_02 J1_03 J1_04 J1_06 Bonroti/paoroti 115 Bhaji 116 Jhol curry 117 Singara 118 Puri 119 Piaju 120 107 | Page Any fried food 121 Any boiled food 122 No. Question Answer J1_07 Were most of the food items purchased, self-produced or received as a gift? SELF-PRODUCED OR RECEIVED AS A GIFT  END MODULE 1. Purchased 2. Self-produced 3. Received as a gift J1_08 Where did you acquire most of the food items from? 1. Farm/home 2. Village market (within own village) 3. Village market (outside own village) 4. City market 5. Other (specify) J1_09 Who paid for most of the food items? 1. Self 2. Spouse 3. Other household member 4. Other non-household member J1_10 How were most of the food items paid for? 1. Cash 2. Credit 3. Advance 4. Other (specify) ◻ 108 | Page Module K: Household Illnesses (Group 2 ONLY) MODULE K SHOULD BE ASKED TO THE MEMBER OF THE HOUSEHOLD WHO TYPICALLY TAKES CARE OF MEMBERS WHEN THEY ARE SICK. FOR HOUSEHOLDS WITH ADULT MALE AND ADULT FEMALE MEMBERS, THIS WILL TYPICALLY BE THE ADULT FEMALE SPOUSE OF THE MAIN RESPONDENT. IF THIS INDIVIDUAL IS UNAVAIBLE, THE SECTION MAY BE COMPLETED BY THE MAIN RESPONDENT. Membe r ID Membe r Name In the last 4 weeks, has this household member suffered from any of the following? READ ALL OPTIONS ALOUD Significan t weight loss If yes, for how many days? (last 4 weeks ) Prolonge d fever If yes, for how many days? (last 4 weeks ) Diarrhe a If yes, for how many days? (last 4 weeks ) Persisten t cough If yes, for how many days? (last 4 weeks ) Generalize d skin rash If yes, for how many days? (last 4 weeks ) Mouth or throat infectio n If yes, for how many days? (last 4 weeks ) Any other illness or injury? (specify ) If yes, for how many days? (last 4 weeks ) K_02a K_02 b K_03a K_03 b K_04a K_04 b K_05a K_05 b K_06a K_06 b K_07a K_07 b K_08a K_08 b 1 2 … K_09 In the last 4 weeks, how many days has ${NAME} been unable to perform his/her main activity due to *any* illness or injury? _________ days K_10 In the last 4 weeks, has ${NAME} contacted or visited a health care facility to seek treatment or advice K_11 In the last 4 weeks, how many times has ${NAME} contacted or visited a health care facility to seek treatment or advice? ________ times K_12 In the last 4 weeks, what type of health care facility did ${NAME} contact or visit? 109 | Page MODULE G2: ROLE IN HOUSEHOLD DECISION-MAKING AROUND PRODUCTION AND INCOME (Group 2 only) Now I’d like to ask you some questions about your participation in certain types of work activities and on making decisions on various aspects of household life. Did you [NAME] participate in [ACTIVITY] in the past 12 months (that is, during the last [one/two] cropping seasons), from [PRESENT MONTH] last year to [PRESENT MONTH] this year? When decisions are made regarding [ACTIVITY], who is it that normally takes the decision? ENTER UP TO THREE (3) MEMBER IDs IF RESPONSE IS MEMBER ID (SELF) ONLY  G2.05 OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98  NEXT ACTIVITY How much input did you have in making decisions about [ACTIVITY]? USE CODE G2↓ To what extent do you feel you can participate in decisions regarding [ACTIVITY] if you want(ed) to? CIRCLE ONE To what extent are you able to access information that you feel is important for making informed decisions regarding [ACTIVITY]? CIRCLE ONE How much input did you have in decisions about how much of the outputs of [ACTIVITY] to keep for consumption at home rather than selling? USE CODE G2↓ How much input did you have in decisions about how to use income generated from [ACTIVITY]? USE CODE G2↓ ACTIVITY G2.01 G2.02 G2.03 G2.04 G2.05 G2.06 G2.07 ID #1 ID #2 ID #3 A Staple grain farming and processing of the harvest: grains that are grown primarily for food consumption (rice, maize, wheat, millet) YES…...1 NO…….2  ACTIVITY B NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT B Horticultural (gardens) or high value crop farming and processing of the harvest YES…...1 NO…….2  ACTIVITY C NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT C Large livestock raising (cattle, buffaloes) and processing of milk and/or meat YES…...1 NO…….2  ACTIVITY D NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT D Small livestock raising (sheep, goats, pigs) and processing of milk and/or meat YES…...1 NO…….2  ACTIVITY E NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT E Poultry and other small animals raising (chickens, ducks, turkeys) and processing of eggs and/or meat YES…...1 NO…….2  ACTIVITY F NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT 110 | Page CODE G2 LITTLE TO NO INPUT IN DECISIONS 1 INPUT INTO SOME DECISIONS 2 INPUT INTO MOST OR ALL DECISIONS 3 NOT APPICABLE / NO DECISION MADE 98 Did you [NAME] participate in [ACTIVITY] in the past 12 months (that is, during the last [one/two] cropping seasons), from [PRESENT MONTH] last year to [PRESENT MONTH] this year? When decisions are made regarding [ACTIVITY], who is it that normally takes the decision? ENTER UP TO THREE (3) MEMBER IDs IF RESPONSE IS MEMBER ID (SELF) ONLY  G2.05 OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98  NEXT ACTIVITY How much input did you have in making decisions about [ACTIVITY]? USE CODE G2↓ To what extent do you feel you can participate in decisions regarding [ACTIVITY] if you want(ed) to? CIRCLE ONE To what extent are you able to access information that you feel is important for making informed decisions regarding [ACTIVITY]? CIRCLE ONE How much input did you have in decisions about how much of the outputs of [ACTIVITY] to keep for consumption at home rather than selling? USE CODE G2↓ How much input did you have in decisions about how to use income generated from [ACTIVITY]? USE CODE G2↓ ACTIVITY G2.01 G2.02 G2.03 G2.04 G2.05 G2.06 G2.07 ID #1 ID #2 ID #3 F Fishpond culture YES…...1 NO…….2  ACTIVITY G NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT G Non-farm economic activities (running a small business, self￾employment, buy-and-sell) YES…...1 NO…….2  ACTIVITY H NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT H Wage and salary employment (work that is paid for in cash or in￾kind, including both agriculture and other wage work) YES…...1 NO…….2  ACTIVITY I NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT 111 | Page I Large, occasional household purchases (bicycles, land, transport vehicles) NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT J Routine household purchases (food for daily consumption or other household needs) NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT CODE G2 LITTLE TO NO INPUT IN DECISIONS 1 INPUT INTO SOME DECISIONS 2 INPUT INTO MOST OR ALL DECISIONS 3 NOT APPICABLE / NO DECISION MADE 98 112 | Page MODULE G3(A): ACCESS TO PRODUCTIVE CAPITAL (Group 2 ONLY) G3.01-G3.05 (GROUP 2 ONLY) Now I’d like to ask you specifically about your household’s land. QUESTION RESPONSE G3.01. Does anyone in your household currently own or cultivate land? YES……..1 NO………2  G3.06, ITEMA G3.02. Who generally makes decisions about what to plant on this land? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER 94 NOT APPLICABLE 98 ID #1 ID #2 ID #3 G3.03. Do you [NAME] solely or jointly cultivate any land? CIRCLE ONE YES, SOLELY YES, JOINTLY YES, SOLELY AND JOINTLY NO G3.04. Who generally makes decisions about what to plant on the land that you yourself cultivate? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER 94 NOT APPLICABLE 98 ID #1 ID #2 ID #3 G3.05. Do you own any of the land owned or cultivated by your household? CIRCLE ONE YES, SOLELY YES, JOINTLY YES, SOLELY AND JOINTLY NO 113 | Page Now I’d like to ask you about a number of items that could be used to generate income. Does anyone in your household currently have any [ITEM]? Do you [NAME] own the [ITEM] in your household? CIRCLE ONE ITEM G3.06 G3.07 A Large livestock (cattle, buffaloes) YES……..1 NO………2  ITEM B YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 B Small livestock (sheep, goats, pigs) YES……..1 NO………2  ITEM C YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 C Poultry and other small animals (chickens, ducks, turkeys) YES……..1 NO………2  ITEM D YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 D Fish pond or fishing equipment YES……..1 NO………2  ITEM E YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 E Non-mechanized farm equipment (hand tools, animal-drawn plough) YES……..1 NO………2  ITEM F YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 F Mechanized farm equipment (tractor-plough, power tiller, treadle pump) YES……..1 NO………2  ITEM G YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 G Non-farm business equipment (solar panels used for recharging, sewing machine, brewing equipment, fryers) YES……..1 NO………2  ITEM H YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 H House or building YES……..1 NO………2  ITEM I YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 I Large consumer durables (refrigerator, TV, sofa) YES……..1 NO………2  ITEM J YES, SOLELY 1 114 | Page YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 Does anyone in your household currently own any [ITEM]? Do you [NAME] own any [ITEM]? CIRCLE ONE ITEM G3.06 G3.07 J Small consumer durables (radio, cookware) YES……..1 NO………2  ITEM K YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 K Cell phone YES……..1 NO………2  ITEM L YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 L Other land not used for agricultural purposes (pieces/plots, residential or commercial land) YES……..1 NO………2  ITEM M YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 M Means of transportation (bicycle, motorcycle, car) YES……..1 NO………2  MODULE G3(B) YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 115 | Page MODULE G3(B): ACCESS TO FINANCIAL SERVICES (GROUP 2 ONLY) Next I’d like to ask about your household’s experience with borrowing money or other items (in-kind) in the past 12 months. Would you or anyone in your household be able to take a loan or borrow cash/in-kind from [SOURCE] if you wanted to? Has anyone in your household taken any loans or borrowed cash/in-kind from [SOURCE] in the past 12 months? CIRCLE ONE Who made the decision to borrow from [SOURCE] most of the time? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 Who makes the decision about what to do with the money or item borrowed from [SOURCE] most of the time? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 Who is responsible for repaying the money or item borrowed from [SOURCE]? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 LENDING SOURCES G3.08 G3.09 G3.10 G3.11 G3.12 ID #1 ID #2 ID #3 ID #1 ID #2 ID #3 ID #1 ID #2 ID #3 A Non-governmental organization (NGO) YES...…….1 NO………..2  SOURCE B MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE B DON’T KNOW 97 B Formal lender (bank/financial institution) YES...…….1 NO………..2  SOURCE C MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE C DON’T KNOW 97 C Informal lender YES...…….1 NO………..2  SOURCE D MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE D DON’T KNOW 97 D Friends or relatives YES...…….1 NO………..2  SOURCE E MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE E DON’T KNOW 97 E Group based micro-finance or lending including VSLAs / SACCOs YES...…….1 NO………..2  SOURCE F MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE F DON’T KNOW 97 F Informal credit / savings groups (.e.g., merry-go￾rounds, tontines, funeral societies, etc.) YES...…….1 NO………..2  G3.13 MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 G3.13 DON’T KNOW 97 116 | Page G3.13 An account can be used to save money, to make or receive payments, or to receive wages or financial help. Do you, either by yourself or together with someone else, currently have an account at any of the following places: a bank or other formal institution (e.g., post office)? YES NO DON’T KNOW 117 | Page MODULE G4: TIME ALLOCATION (GROUP 2 ONLY) G4.01: PLEASE RECORD A LOG OF THE ACTIVITIES FOR THE INDIVIDUAL IN THE LAST COMPLETE 24 HOURS (STARTING YESTERDAY MORNING AT 4 AM, FINISHING 3:59 AM OF THE CURRENT DAY). THE TIME INTERVALS ARE MARKED IN 15 MIN INTERVALS. MARK ONE ACTIVITY FOR EACH TIME PERIOD BY ENTERING THE CORRESPONDING ACTIVITY CODE IN THE BOX. G4.02: CHECK THE BOX BELOW IF THE RESPONDENT WAS CARING FOR CHILDREN WHILE PERFORMING EACH ACTIVITY. Now I’d like to ask you about how you spent your time during the past 24 hours. We’ll begin from yesterday morning, and continue through to this morning. This will be a detailed accounting. I’m interested in everything you did (i.e. resting, eating, personal care, work inside and outside the home, caring for children, cooking, shopping, socializing, etc.), even if it didn’t take you much time. I’m particularly interested in agricultural activities such as farming, gardening, and livestock raising whether in the field or on the homestead. I’m also interested in how much time you spent caring for children, especially if it happened while you did some other activity (e.g., collecting water while carrying a child or cooking while watching after a sleeping child). Night Morning Day 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 G4.01 Activity (WRITE ACTIVITY CODE) G4.02 Did you also care for children? YES..…CHECK BOX NO…LEAVE BLANK YES CHECK BOX NO LEAVE BLANK □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Day Evening Night 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 24:00 1:00 2:00 3:00 G4.01 Activity (WRITE ACTIVITY CODE) G4.02 Did you also care for children? YES CHECK BOX NO LEAVE BLANK □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ ACTIVITY CODES FOR G4.01 1. Sleeping and resting 2. Eating and drinking 3. Personal care 4. School (including homework) 5. Work as employed 6. Own business work 8. Farming (other than jute) 9. Large livestock raising (cattle, buffaloes) 10. Small livestock raising (sheep, goats, pigs) 11. Poultry and raising other small animals (chickens, ducks turkeys) 12. Fishpond culture 13. Shopping/getting service (including health 15. Weaving, sewing, textile care 16. Cooking 17. Domestic work (including fetching wood and water) 18. Post-processing for jute 19. Post-processing for crops other than jute 20. Caring for children 23. Exercising 24. Social activities and hobbies 25. Religious activities 97. Other (specify) 118 | Page 7. Farming (jute) service) 14. Commuting (to/from work or school) 21. Caring for adults/elderly 22. Travelling (Not for work or school) G4.03.In the last 24 hours did you work (at home or outside of the home including chores or other domestic activities) less than usual, about the same as usual, or more than usual? FOR FEMALES ONLY: DOES RESPONDENT HAVE A CHILD UNDER 5 YEARS OLD? YES...…….1  G4.04 NO………..2  MODULE G5 G4.04. If you wanted to do something (livelihood-related, training-related, self￾care) and could not take your child with you, is there someone who could care for your child in your absence? YES...…….1  G4.05 NO………..2  MODULE G5 G4.05. Who? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 ID #1 ID #2 ID #3 LESS THAN USUAL ABOUT THE SAME AS USUAL MORE THAN USUAL IF RESPONDENT IS MALE  MODULE G5 119 | Page MODULE G5: GROUP MEMBERSHIP (GROUP 2 ONLY) Now I’m going to ask you about groups in the community. These can be either formal or informal and customary groups. Is there a [GROUP] in your community? Is this group composed of all male or female or mixed-sex members? Are you an active member of this [GROUP]? To what extent do you feel like you can influence decisions in this [GROUP]? To what extent does this [GROUP] influence life in the community beyond the group activities? GROUP CATEGORIES G5.01 G5.02 G5.03 G5.04 G5.05 A Agricultural / livestock / fisheries producer’s group (including marketing groups) YES 1 NO 2 DON’T KNOW 97 GROUP B ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP B NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT B Water users’ group YES 1 NO 2 DON’T KNOW 97 GROUP C ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP C NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT C Forest users’ group YES 1 NO 2 DON’T KNOW 97 GROUP D ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP D NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT D Credit or microfinance group (including Grameen, SACCOs / merry-go-rounds / VSLAs) YES 1 NO 2 DON’T KNOW 97 GROUP E ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP E NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT E Mutual help or insurance group (including burial societies) YES 1 NO 2 DON’T KNOW 97 GROUP F ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP F NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT F Trade and business association group YES 1 NO 2 DON’T KNOW 97 GROUP G ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP G NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT G Civic group (improving community) or charitable group (helping others) YES 1 NO 2 DON’T KNOW 97 GROUP H ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP H NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT H Religious group YES 1 NO 2 DON’T KNOW 97 GROUP I ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP I NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT I Other (specify): _______________________ YES 1 NO 2 DON’T KNOW 97 MODULE G6 ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  MODULE G6 NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT 120 | Page If G5.01A = YES, COMPLETE THE QUESTIONS ON PRODUCER GROUPS BELOW (Group 2 only) No. Question Answer G5_06 Is this producer group related to jute? 1. Yes 2. No NO  END MODULE G5_07 How many other active members does this producer group have? (QUANTITY) G5_08 What year did you join this producer group? (YEAR) G5_09 Did you ever engage in any of the following activities for jute through this producer group? IF AT MOST ONE ACTIVITY SELECTED  G5_10 1. Yes 2. No G5_10 Which of these activities is your main reason for actively participating in this producers group? ☐${code2} G5_11 In the last ${period} season, did you ${CODE2} through this producer group? ONLY SELECT FROM ACTIVITIES SELECTED IN D2_07 IF AT LEAST ONE ACTIVITY SELECTED  D2_11 1. Yes 2. No SELL OUTPUT (PROCESSED OR UNPROCESSED) D2_11 G5_12 What is the main reason why farmers do not always sell or bulk their output via this producer group?" 1. Does not bulk group members' output 2. Does not provide a good price for sold output 3. May not be able to sell the output 4. Corruption / money from sales disappears 5. Does not pay out timely for output sold 6. Does not reward quality / combines output with different quality levels 7. Other (specify) G5_13 In the last ${period} season, what type of buyer did this producer group mainly sell to? 1. Village collector 2. Wholesaler 3. Cold storage owner 4. Wholesaler to cold storage 5. Collection center of company 6. Processing farm 7. Cooperative society 8. Farmer society 9. Retailer 10. Consumer 11. Hotel/restaurant 12. Other (specify) 121 | Page 122 | Page No. Question Answer G5_14 In the next 12 months, are you planning to engage in any of the following activities through your producer group? ◻ Buy inputs for jute ◻ Sell jute (before post-harvest processing) ◻ Process jute after harvest ◻ Sell processed jute (after post-harvest processing) ◻ Participate in agricultural training for jute ◻ Any other service for jute (specify) 1. Yes 2. No Now I’d like you to think about how much jute you and other farmers would sell through your producer group, depending on the price the group offers. In the next ${period} season, of all the jute you are planning to sell, what percentage are you planning to sell though your producer group if the price the group offers is… G5_15 …higher than the price that other buyers offer for jute? (PERCENTAGE) G5_16 …the same as the price that other buyers offer for jute? (PERCENTAGE) G5_17 …lower than the price that other buyers offer for jute? (PERCENTAGE) In the next ${period} season, on average, what percentage of their jute do you think other members will sell through your producer group if the price the group offers is… G5_18 …higher than the price that other buyers offer for jute? (PERCENTAGE) G5_19 …the same as the price that other buyers offer for jute? (PERCENTAGE) G5_20 …lower than the price that other buyers offer for jute? (PERCENTAGE) G5_21 Do you expect this producer group will offer a higher, lower, or the same price for jute as other buyers you can sell to? 1. Higher price 2. Lower price 3. Same price G5_22 Does this producer group have any rules or by-laws on how much of your jute members should sell through the group? 1. Yes 2. No NO  END MODULE 123 | Page G5_23 DESCRIBE RULES/BY-LAWS. PROBE FOR MINIMUM QUANTITY AND WHAT HAPPENS IF QUANTITY IS NOT DELIVERED (E.G. FINES). 124 | Page MODULE G6. PHYSICAL MOBILITY (Group 2 only) QUESTION RESPONSE FOR G6.01 - G6.06: USE CODE G6↓ G6.01 How often do you visit an urban center? G6.02 How often do you go to the market / haat / bazaar? G6.03 How often do you go to visit family or relatives? G6.04 How often do you go to visit a friend / neighbor’s house? G6.05 How often do you go to the hospital / clinic / doctor (seek health service)? G6.06 How often do you go to a public village gathering / community meeting / training for NGO or programs? G6.07. In the last 12 months, how many times have you been away from home for one or more nights (in other words, sleeping somewhere else for the night)? G6.08. In the last 12 months, have you been away from home for more than one month at a time? YES NO IF RESPONDENT IS MALE MODULE G7 CODE G6 EVERYDAY EVERY WEEK AT LEAST ONCE EVERY 2 WEEKS AT LEAST ONCE EVERY MONTH AT LEAST ONCE LESS THAN ONCE A MONTH NEVER 125 | Page REMAINDER OF MODULE (G6.09-G6.12) SHOULD ONLY BE ASKED IF RESPONDENT IS FEMALE Now I’d like to ask you some questions about different places you might visit. Who usually decides whether you can go to [PLACE]? ENTER UP TO THREE (3) MEMBER IDs IF RESPONSE IS MEMBER ID (SELF) ONLY  NEXT PLACE OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 Does your husband/partner or other household member object to you going alone to [PLACE]? Under what circumstances would this person NOT object to your going to [PLACE] alone? CIRCLE ALL APPLICABLE Do these objections prevent you from going alone to [PLACE]? PLACE G6.09 G6.10 G6.11 G6.12 ID #1 ID #2 ID #3 A Urban center YES……1 NO..……2  PLACE B IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE B YES……1 NO..……2 B Market / haat / bazaar YES……1 NO..……2  PLACE C IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 YES……1 NO..……2 126 | Page UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE C C Visit family or relatives YES……1 NO..……2  PLACE D IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE D YES……1 NO..……2 D Visit a friend / neighbor’s house YES……1 NO..……2  PLACE E IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE E YES……1 NO..……2 E Hospital / clinic / doctor (seek health service) YES……1 NO..……2  PLACE F IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE F YES……1 NO..……2 Who usually decides whether you can go to [PLACE]? ENTER UP TO THREE (3) MEMBER IDs Does your husband/partner or other household member object to you going alone to [PLACE]? Under what circumstances would this person NOT object to your going to [PLACE] alone? CIRCLE ALL APPLICABLE Do these objections prevent you from going alone to [PLACE]? 127 | Page IF RESPONSE IS MEMBER ID (SELF) ONLY  NEXT PLACE OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 PLACE G6.09 G6.10 G6.11 G6.12 ID #1 ID #2 ID #3 F Temple / church / mosque YES……1 NO..……2  PLACE G IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE G YES……1 NO..……2 G Public village gathering or community meeting YES……1 NO..……2  PLACE H IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE H YES……1 NO..……2 H Training for NGO / programs YES……1 NO..……2  PLACE I IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  PLACE I YES……1 NO..……2 128 | Page I Outside your community or village YES……1 NO..……2  MODULE G7 IF I HAVE COMPANY (RELATIVES, CHILDREN)………………………..….1 IF I CAN ARRANGE MY OWN EXPENSES (FOR TRANSPORT)………....2 IF I FOLLOW PURDAH / DRESS ACCEPTABLY…………………………...3 OTHER (SPECIFY)………………………………………………………………4 UNDER NO CIRCUMSTANCES WOULD I BE ALLOWED TO GO………..5  MODULE G7 YES……1 NO..……2 129 | Page MODULE G7: INTRAHOUSEHOLD RELATIONSHIPS (Group 2 only) Now I’d like to ask you some questions about how you feel about some of other people in your household or family group and how you think they feel about you. ENTER MEMBER ID FOR EACH RELATION OTHER CODES: NON-HH MEMBER...….94 Do you [NAME] respect your [RELATION]? Does your [RELATION] respect you? Do you trust your [RELATION] to do things that are in your best interest? When you disagree with your [RELATION], do you feel comfortable telling him/her that you disagree? IS [RELATION] THE OTHER RESPONDENT WITHIN THIS HOUSEHOLD? Is there a co-wife within your household? RELATION G7.02 G7.03 G7.04 G7.05 G7.06 G7.07 A Husband / wife ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER YES……1  RELATION C NO..……2 B Other adult male or female member of the household ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER C IF RESPONDENT IS MALE: Father (or adapt this category to capture other important relationship) IF RESPONDENT IS FEMALE: Mother-in-law ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER IF RESPONDENT IS MALE  MODULE G8(A) YES……1 NO..……2  MODULE G8(A) 130 | Page D Most senior co-wife (the person who was in the household just before you, or, if you are the senior wife, the one who married into the household after you) ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER 131 | Page MODULE G8(A): AUTONOMY IN DECISION-MAKING (Group 2 only) Now I am going to read you some stories about different farmers and their situations regarding different agricultural activities. This question format is different from the rest so take your time in answering. For each I will then ask you how much you are like or not like each of these people. We would like to know if you are completely different from them, similar to them, or somewhere in between. There are no right or wrong answers to these questions. READ ALOUD EACH STORY, SUBSEQUENT QUESTIONs, AND RESPONSE CODES. NAMES SHOULD BE ADOPTED TO LOCAL CONTEXT AND BE MALE/FEMALE DEPENDING ON THE SEX OF THE RESPONDENT. THE ORDER OF TOPICS A-D SHOULD BE RANDOMIZED, AND WITHIN EACH TOPIC, THE ORDER OF STORIES 1-4 SHOULD BE RANDOMIZED. Are you like this person? CIRCLE ONE Are you completely the same or somewhat the same? CIRCLE ONE Are you completely different or somewhat different? CIRCLE ONE STORY G8.01 G8.02 G8.03 The types of crops to grow or raise for consumption and sale in market A1 “[PERSON’S NAME] cannot grow other types of crops here for consumption and sale in market. Beans, sweet potato and maize are the only crops that grow here.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A2 SOMEWHAT THE SAME…....2  A2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A2 “[PERSON’S NAME] is a farmer and grows beans, sweet potato, and maize because her spouse, or another person or group in her community tells her she must grow these crops. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A3 SOMEWHAT THE SAME…....2  A3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A3 “[PERSON’S NAME] grows the crops for agricultural production that her family or community expect. She wants them to approve of her as a good farmer.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A4 SOMEWHAT THE SAME…....2  A4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A4 “[PERSON’S NAME] chooses the crops that she personally wants to grow for consumption and sale in market and thinks are best for herself and her family. She values growing these crops. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B1 SOMEWHAT THE SAME…....2  B1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 Livestock raising B1 “[PERSON’S NAME] cannot raise any livestock other than what she has. These are all that do well here.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B2 SOMEWHAT THE SAME…....2  B2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B2 “[PERSON’S NAME] raises the types of livestock she does because her spouse, or another person or group in her community tell her she must use these breeds. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B3 SOMEWHAT THE SAME…....2  B3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B3 “[PERSON’S NAME] raises the kinds of livestock that her family or community expect. She wants them to approve of her as a good livestock raiser.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B4 SOMEWHAT THE SAME…....2  B4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B4 “[PERSON’S NAME] chooses the types of livestock that she personally wants to raise and thinks are good for herself and her family. She values raising these types. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C1 SOMEWHAT THE SAME…....2  C1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 132 | Page READ ALOUD EACH STORY, SUBSEQUENT QUESTIONs, AND RESPONSE CODES. NAMES SHOULD BE ADOPTED TO LOCAL CONTEXT AND BE MALE/FEMALE DEPENDING ON THE SEX OF THE RESPONDENT. Are you like this person? CIRCLE ONE Are you completely the same or somewhat the same? CIRCLE ONE Are you completely different or somewhat different? CIRCLE ONE STORY G8.01 G8.02 G8.03 Taking crops or livestock (incl. eggs or milk) to the market (or not) C1 “There is no alternative to how much or how little of her crops or livestock [PERSON’S NAME] can take to the market. She is taking the only possible amount.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C2 SOMEWHAT THE SAME…....2  C2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C2 “[PERSON’S NAME] takes crops and livestock to the market because her spouse, or another person or group in her community tell her she must sell them there. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C3 SOMEWHAT THE SAME…....2  C3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C3 “[PERSON’S NAME] takes the crops and livestock to the market that her family or community expect. She wants them to approve of her.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C4 SOMEWHAT THE SAME…....2  C4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C4 “[PERSON’S NAME] chooses to take the crops and livestock to market that she personally wants to sell there, and thinks is best for herself and her family. She values this approach to sales. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D1 SOMEWHAT THE SAME…....2  D1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 How to use income generated from agricultural and non￾agricultural activities D1 “There is no alternative to how [PERSON’S NAME] uses her income. How she uses her income is determined by necessity.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D2 SOMEWHAT THE SAME…....2  D2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D2 “[PERSON’S NAME] uses her income how her spouse, or another person or group in her community tell her she must use it there. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D3 SOMEWHAT THE SAME…....2  D3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D3 “[PERSON’S NAME] uses her income in the way that her family or community expect. She wants them to approve of her.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D4 SOMEWHAT THE SAME…....2  D4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D4 “[PERSON’S NAME] chooses to use her income how she personally wants to, and thinks is best for herself and her family. She values using her income in this way. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME...1G8.04 SOMEWHAT THE SAME….2 G8.04 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 133 | Page MODULE G8(B): NEW GENERAL SELF-EFFICACY SCALE (Group 2 only) Now I’m going to ask you some questions about different feelings you might have. Please listen to each of the following statements. Think about how each statement relates to your life, and then tell me how much you agree or disagree with the statement on a scale of 1 to 5, where 1 means you “strongly disagree” and 5 means you “strongly agree.” (Note: Randomize order of statements) STATEMENTS G8.04 A I will be able to achieve most of the goals that I have set for myself. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE B When facing difficult tasks, I am certain that I will accomplish them. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE C In general, I think that I can obtain outcomes that are important to me. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE D I believe I can succeed at most any endeavor to which I set my mind STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE E I will be able to successfully overcome many challenges. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE F I am confident that I can perform effectively on many different tasks. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE G Compared to other people, I can do most tasks very well. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE H Even when things are tough, I can perform quite well. STRONGLY DISAGREE DISAGREE 134 | Page NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE MODULE G8(C): LIFE SATISFACTION (Group 2 only) The following questions ask how satisfied you feel with your life as a whole, on a scale from 1 to 5, where 1 means you feel “very dissatisfied” and 5 means you feel “very satisfied.” STATEMENTS G8.05 A Overall, how satisfied are you with life as a whole these days? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED B Overall, how satisfied with your life were you 5 years ago? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED C As your best guess, overall how satisfied with your life do you expect to feel 5 years from today? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED 135 | Page MODULE G9. Attitudes about Domestic Violence (Group 2 only) Now I would like to ask about your opinion on the following issues. Please keep in mind that I am not asking about your personal experience or whether the following scenarios have happened to you. I would only like to know whether you think the following issues are acceptable. In your opinion, is a husband justified in hitting or beating his wife in the following situations? SITUATION G9.01 A If she goes out without telling him? YES NO DON’T KNOW B If she neglects the children? YES NO DON’T KNOW C If she argues with him? YES NO DON’T KNOW D If she refuses to have sex with him? YES NO DON’T KNOW E If she burns the food? YES NO DON’T KNOW 136 | Page Module X: Closing No. Question Field X1_01 May we take a photograph of you? The photo will only be used for our research. 1. Yes 2. No NO >> gps Photo PLEASE TAKE A PHOTOGRAPH OF THE RESPONDENT. (PHOTO) Gps PLEASE RECORD THE CURRENT GPS COORDINATES. END OF FEMALE FORM 137 | Page MAIN FORM MODULE A: HOUSEHOLD IDENTIFICATION (GROUP 1, 2 & 3) Variable IDENTIFICATION Response date DATE interviewer TEAM [SUPERVISOR NAME/CODE] & INTERVIEWER [NAME/CODE] hhid HOUSEHOLD IDENTIFICATION NUMBER [NUMERICAL] Household verification (1): HOUSEHOLD INFORMATION FOR ID ${hhid} A_02 DISTRICT / UPAZILA [CODED] This information will be pre-filled by the survey program. Interviewers should review the information and ensure that they are interviewing the correct respondent. If the preload information does not conform to the correct respondent, or if there is any doubt over the status of the person to be interviewed, the interviewer should not proceed with the interview and immediately contact their supervisor. A_03 UNION [CODED] A_04 VILLAGE [CODED] HOUSEHOLD HEAD: A_08 NAME A_12 GENDER HOUSEHOLD HEAD’S FATHER A_10 NAME A_05 HOUSEHOLD LOCATION/LANDMARK [STRING] MAIN RESPONDENT IN PREVIOUS ROUND [PRELOADED NAME] CHECK THE ABOVE INFORMATION CAREFULLY AND CHOOSE "YES" TO PROCEED IF YOU ARE SURE YOU ARE AT THE RIGHT HOUSEHOLD. ◻ 1 – Yes ◻ 2 – No * A_06 CAN YOU START THE INTERVIEW WITH THE MAIN RESPONDENT IDENTIFIED IN FEMALE FORM? ◻ 1 – Yes -> A_09 ◻ 2 – No -> A_07 138 | Page * YOU CANNOT PROCEED UNLESS YOU HAVE FOUND THE CORRECT HOUSEHOLD. EITHER SPEAK TO THE RESPONDENT AGAIN TO CHECK THE DETAILS OR CANCEL THE INTERVIEW AND CONTINUE TO LOOK FOR THE RIGHT HOUSEHOLD. A_07 WHY CAN YOU NOT START THE INTERVIEW? AFTER SELECTING AN ANSWER -> END ◻ 1 – Moved ◻ 2 – Not found ◻ 3 – Refused ◻ 4 – Unavailable ◻ 5 – Other (specify) RESPONDENT INFORMATION main_resp Please enter the NAME of the respondent for this form, identified in the female form. The respondent should be the primary jute farmer/input seller in the household. resp_age Please provide the age of the respondent for this form resp_sex Please provide the sex of the respondent for this form 139 | Page CONSENT OF RESPONDENT Has consent been obtained in the female form? 1.Yes -> NEXT MODULE 2. No -> Obtain informed consent Good morning/afternoon. I am ________ from the Data Analysis and Technical Assistance Limited (DATA), a Bangladeshi research organization based in Dhaka. Together with the International Food Policy Research Institute (IFPRI), we are conducting a survey that will provide IFPRI with necessary information to carry out research that is designed to help promote the welfare of Bangladeshis; particularly, to improve food consumption and nutrition of the people and women’s status, and to enhance agricultural development and income generation. Your household has been chosen by a random selection process. We are inviting you to be a participant in this study. We value your opinion and there are no wrong answers to the questions we will be asking in the interview. We will use approximately 3-4 hours of your time to collect all the information. If you prefer, we can do the interview in two visits. There will be no cost to you other than your time. There will be no risk as a result of your participating in the study. Your participation in this research is completely voluntary. You are free to withdraw your consent and discontinue participation in this study at any time. This study is conducted anonymously. You will only be identified through code numbers. Your identity will not be stored with other information we collect about you. Your responses will be assigned a code number, and the list connecting your name with this number will be kept in a locked room and will be destroyed once all the data has been collected and analyzed. Any information we obtain from you during the research will be kept strictly confidential. Your participation will be highly appreciated. The answers you give will help provide better information to policy-makers, practitioners and program managers so that they can plan for better services that will respond to your needs. The researcher read to me orally the consent form and explained to me its meaning. I agree to take part in this research. I understand that I am free to discontinue participation at any time if I so choose, and that the investigator will gladly answer any question that arise during the course of the research. Contact Person: Name of the Principal Investigator (PI): _____________ Address: Tel: __________; E-mail of PI: _________ Signature of the Enumerator: _____________ Date: /_____/_____/_____/ consent_m “Do you agree to be interviewed for the purposes of this study?” male_alone Ability to be interviewed 1. Alone 2. With adult female members present 3. With adult male members present 4. With adults of both sexes present 5. With children present 6. With adults of both sexes and children present no_consent WHY DOES THE RESPONDENT NOT GIVE CONSENT?  END INTERVIEW 140 | Page Module C: Agriculture (All Groups) Module C1: Production of Crops Other Than Jute (Groups 1 & 2 Only) C1_01 Did you harvest any crops besides jute in the last 3 seasons (Boro 2015/2016, Aus 2016, and Aman 2016? What other crops did you harvest? In which season of 2016 did you harvest ${other}? How much land did you use to grow the …that you harvested in 2016? How much … did you harvest in 2016 (total production)? Of the … that you harvested in 2016, how much did you sell? Sale price per KG 2. Rabi (Boro: December 2015 – April 2016) 3. Kharif-I (Aus: May – July 2016) 4. Kharif-II (Aman: August – November 2016) (DECIMALS) (KG) (KG) (TAKA) C1_02 C1_03 C1_04 C1_05 C1_06 C_08 Rice (aman) Rice (aus) Rice (boro) Wheat Potato Tomato Pulses Groundnuts Mango Coriander Chili peppers Garlic Onion Mustard Sesame Other (specify) Total sales value instead of price per kg 141 | Page Module C2: Jute Production in Kharif-1 (Aus) 2016 (Group 1 & Group 2 ONLY) C2_01. Did you grow jute in Kharif-I (Aus: May-July 2016)? 1. Yes 2. No >> MODULE C4 To grow jute, in what month did you... How much land did you use to grow jute May –July ‘16)? Fiber quantity harvested (total production) Fiber quantity received from leased out plot Fiber quantity sold Who did you sell the fiber to? start land preparations? … finish harvesting? If >2 buyers To which two buyers did you sell most of the jute? How many kg did you sell in total to [BUYER 1]? If >1 buyers And to [BUYER 2]? What price did [BUYER 1] pay you per kg of jute? If >1 buyers And [BUYER 2]? May be up to 3 months before start of Kharif-I (Aus: May – July 2016) season. May be up to 2 months after end of Kharif-I (Aus: May – July 2016) season. (DECIMALS) (KG) (KG) (KG) - List of buyers - Other (specify) List of buyers Kg Taka/Kg C2_02 C2_03 C2_04 C2_05 C2_06 C2_08 C2_09 C2_10 C2_11 C2_12 142 | Page Where did you sell your jute to [BUYER 1]? If >1 buyers And to [BUYER 2]? 1. Farm gate (home) 4. Town market (own union) 2. Village market (own village) 5. Town market (other union) 3. Village market (other village) 6. Other (specify) C2_13 != 1 (Farm gate) How many minutes did you travel on average per transaction with [BUYER 1]? If >1 buyers And to [BUYER 2]? Main Travel Mode How much were the total transportation costs to sell to [BUYER 1]? If >1 buyers And to [BUYER 2]? WRITE 0 IF NO COSTS Code Minutes 1. Walking 2. Rickshaw 3. Van 4. Push Van 5. Tractor 6. Truck 7. Motorcar 8. Bicycle 9. Motor bicycle 10. Horse cart 11. Bullock cart 12. Other (specify) Taka C2_13 C2_14 C2_15 C2_16 Did you sell the jute sticks? C2_17= 1 What percentage of your jute sticks did you sell? How much did you receive in total for the sale of your sticks? 1. Yes 2. No % (Taka) C2_17 C2_18 C2_19 143 | Page Module C3: Jute Production Costs: Inputs (All Groups) Please answer the following questions about your use (sales) of inputs for jute production in Kharif-I (Aus: May – July 2016). Input Did you use (sell) …for jute production in Kharif-I (Aus: May – July 2016)? READ OUT ALL OPTIONS How many kg/liters of … did you use (sell) in total? Unit From what type of seller did you mainly purchase the…? What price did you pay per kg / liter to purchase …? Were you offered any … for free, at a subsidy, or discount compared to the market value? PROBE FOR PROMOTIONS FROM TRAINING PROGRAMS, DEALERS & INPUT SELLERS INCLUDE PROMOTIONS EVEN IF NOT USED What type of organization provided the discount? Name of organization What did you mainly do with the subsidy/discount? What was the value of the gift, subsidy or discount per kg/liter? If AVC, specify % discount received Did you receive a loan or other form of credit to purchase …? 1. Yes 2. No NO  C3_04 1. Kilos 2. Litres 1. Commercial agricultural company 2. Local distributor 3. Agricultural input store 4. Cooperative/farmer association 5. Government 6. Intermediaries 7. NGOs 8. Trader 9. Own production -> Next input 10. Other (specify) AMOUNT SPENT, EXCL. SUBSIDIES / DISCOUNTS BUT INCL. CREDIT 1. No  C3_06 2. Yes, for free 3. Yes, at a subsidy/discount 1. Input seller (C3_03) 2. AVC/NAAFCO 3. Other local commercial agricultural company 4. Other local distributor 5. Other agricultural input store 6. Other coop./farmer organization 7. Other government institution 8. Other intermediares 9. Other NGOs 10. Other trader 97. Other, specify 1. Used it for own jute cultivation 2. Used it for other crop 3. Sold it to someone else 4. Gave it to someone else 5. Saved it for future use 6. Discarded it / did not use it 7. Other, specify 1. Yes 2. No  C3_10 C3_01 C3_02 C3_02u C3_03 C3_04 C3_05 C3_06 C3_07 C3_08 C3_09 C3_10 Seed Only ask in Group 3 Urea 144 | Page TSP DAP NPKS MOP/MP Gypsum Magnesium sulph. Zinc sulph. Manure Compost Other, spec. Insecticide Herbicide Fungicide Irrigation Total cost Input C3_10=Yes Source(s) C3_10=Yes How much credit did you receive or how much did you borrow in total to purchase …? C3_10=Yes Could you have obtained more credit to purchase …? C3_10=No Could you have obtained a loan or credit to purchase …? C3_13 or C3_14=Yes From what source(s)? ALLOW FOR MULTIPLE RESPONSE C3_01==Yes & Group==3 At what price did you sell … on average? C3_01==Yes &Group==3 Did you provide any … for free, at a subsidy, or discounted price compared to the market value? C3_17==Yes &Group==3 What was the value of the gift, subsidy or discount per kg/liter? C3_01==Yes Group==3 In the past 12 months, to what percentage of buyers did you sell … on credit? 145 | Page 1.Purchase on credit - ${C3_04} 2.Bank/Formal lender 3.Informal lender 4.Friend/relative 5.Group-based MFI/VSL/Coop. 6. Informal savings group (Taka) 1. Yes  C3_11 2. No  C3_12 1. Yes 2. No  C3_12 1.Purchase on credit - ${C3_04} 2.Bank/Formal lender 3.Informal lender 4.Friend/relative 5.Group-based MFI/VSL/Coop. 6. Informal savings group PRICE PER KG/LITER, NET OF PROMOTIONS, SUBSIDIES OR DISCOUNTS INCL. CREDIT 1. No  C3_16 2. Yes, for free 3. Yes, at a subsidy / discount (Taka) WRITE ZERO IF NEVER SOLD … ON CREDIT C3_11 C3_12 C3_13 C3_14 C3_15 C3_16 C3_17 C3_18 C3_19 Seed Urea TSP DAP NPKS MOP Gypsum Magnesium sulph. Zinc sulph. Manure Compost Other, spec. Insecticide Herbicide Fungicide Irrigation IF GROUP 3 AND NPKS FERTILIZER SELECTED IN C3_01 C3_20 How many 25kg bags of NAAFCO NPKS fertilizer did you sell with a coupon for a …. a) 20% discount? b) 50% discount? c) 80% discount? C3_21 So in total, in Aus 2016, X farmers purchased NPKS from you with a coupon. How many of them: a) Purchased more than one 25kg bag of NPKS from you? b) Purchased any other fertilizers from you? 146 | Page c) Purchased any other inputs, other than fertilizer, from you? C3_22 If these X farmers weren’t given this coupon, how many of them do you think … a) Would have purchased NPKS from you? b) Would have purchased another fertilizer from you? c) Would have purchased inputs other than fertilizer from you? THE REMAINDER ONLY FOR GROUPS 1 & 2 EXCEPT FOR THE INPUT BUSINESS LINE 147 | Page Activity ONLY ASK LABOR QUANTITIES (DAYS, HOURS, WAGES) FOR TYPES OF LABOR USED IN EACH TASK Household Hired Difficulties finding labor? Labor days of 8 hours Actual days Hours per day Wage per day Adult male Adult female Child Male Female Male Female Male Female Yes/No C3_31 C3_32 C3_33 C3_34 C3_35 C3_36 C3_37 C3_38 C3_39 C3_40 C3_27_1 For the ${c2_04} decimals used to cultivate jute in the last Kharif-I season, what type of labor did you use for cultivation? E.g. ploughing, land leveling, seeding, weeding, applying fertilizers. READ ALL CHOICES ALOUD. (GROUPS 1 & 2 ONLY) ◻ 1 – Male adult household members (incl. farmer himself) ◻ 2 – Female adult household members ◻ 3 – Child household members ◻ 4 – Hired male labor ◻ 5 – Hired female labor ◻ 6 – Hired machinery ◻ 7 – Did not use any labor Ploughing, harrowing, and land leveling Seeding Weeding, applying fertilizer and other chemicals C3_27_2 For the ${c2_04} decimals used to cultivate jute in the last Kharif-I season, what type of labor did you use for harvesting? E.g. harvesting, drying, curing, threshing. READ ALL CHOICES ALOUD. (GROUPS 1 & 2 ONLY) SEE ABOVE CATEGORIES Harvesting C3_27_3 For the ${c2_04} decimals used to cultivate jute in the last Kharif-I season, what type of labor did you use for drying, curing and threshing? READ ALL CHOICES ALOUD. (GROUPS 1 & 2 ONLY) Drying Curing Threshing C3_27_4 For the ${c2_04} decimals used to cultivate jute in the last Kharif-I season, what type of labor did you use for fiber separation? E.g. bundling stalks, retting, stripping, washing, bailing. THIS QUESTION AND ACTIVITIES FOR JUTE ONLY. READ ALL CHOICES ALOUD. (GROUPS 1 & 2 ONLY) ◻ 1 – Male adult household members (incl. farmer himself) ◻ 2 – Female adult household members ◻ 3 – Child household members ◻ 4 – Hired male labor ◻ 5 – Hired female labor ◻ 6 – Hired machinery ◻ 7 – Did not use any labor Bundling stalks and retting 148 | Page Stripping and washing Bailing C3_27_5 For the ${c2_04} decimals used to cultivate jute in the last Kharif-I season, what type of labor did you use for post-harvest processing? E.g. sorting, grading packing, transporting. READ ALL CHOICES ALOUD. READ ALL CHOICES ALOUD. (GROUPS 1 & 2 ONLY) ◻ 1 – Male adult household members (incl. farmer himself) ◻ 2 – Female adult household members ◻ 3 – Child household members ◻ 4 – Hired male labor ◻ 5 – Hired female labor ◻ 6 – Hired machinery ◻ 7 – Did not use any labor Sorting, grading, and packing Transporting to market C3_27_6 Last Kharif-I season, what type of labor did you use in your business for selling jute inputs? READ ALL CHOICES ALOUD (GROUP 3 ONLY) ◻ 1 – Male adult household members (incl. farmer himself) ◻ 2 – Female adult household members ◻ 3 – Child household members ◻ 4 – Hired male labor ◻ 5 – Hired female labor ◻ 6 – Hired machinery ◻ 7 – Did not use any labor Labor used for selling jute inputs (excl. labor used for selling other inputs) 149 | Page Module C4 Knowledge and Knowledge Sharing of Fertilizers (All Groups) C4_01 Last jute season, when you sold (purchased) your jute inputs other than fertilizer, did you provide (receive) information on how the inputs should be used? 1. Yes 2. No C4_02 [IF C4_01=1] How many minutes did you (the input seller) spend explaining how the input should be used (per transaction)? (Minutes per transaction) C4_03 Now we have a few questions about fertilizers. Last jute season, when you sold (purchased) your jute inputs, did you provide (receive) information on what type of fertilizers to use and/or how to apply fertilizers? 1. Yes 2. No C4_04 [IF C4_03=1] How many minutes did you (the input seller) spend explaining what type of fertilizers to use and/or how to apply fertilizers (per transaction)? (Minutes per transaction) C4_05 [IF C4_03=1] What type of information did you provide (receive) on fertilizer? Did you provide (receive) information about … 1. What type of fertilizer to use? Yes / No 2. How much fertilizer to be applied? Yes / No 3. When to apply fertilizer? Yes / No 4. What fertilizers are best given expected weather conditions? Yes / No 5. What fertilizers improve your yields the most? Yes / No 6. What fertilizers improve the fiber quality of your jute the most? Yes / No 7. What fertilizers improve disease resistance the most? Yes / No C4_06 How much do you know about the effects of NPKS? 1. Nothing 2. Some but not much 3. Quite a lot 4. Everything C4_07 [IF C3_01 == 1 for NPKS] Last jute season, when you sold (purchased) your jute inputs, did you provide (receive) any information about NPKS fertilizers? 1. Yes 2. No C4_08 [IF C4_07 == 1] Last jute season, what type of information did you (the input seller) provide on NPKS? Did you talk about … Basal Top 1. Whether to use NPKS? Yes / No Yes / No 2. How much NPKS to apply? Yes / No Yes / No 3. When to apply NPKS? Yes / No Yes / No 4. What effects NPKS has on your yields? Yes / No Yes / No 5. What effects NPKS has on the fiber quality of your jute? Yes / No Yes / No 6. What effects NPKS has on disease resistance? Yes / No Yes / No C4_09 Last jute season, how much time did you (the input seller) spend giving this information on NPKS? (Minutes per transaction) C4_10 a) Basal dressing b) Top dressing 1) [IF C4_08.2=1] How much NPKS did you (the input seller) say should be applied? Kg/acre Kg/acre 2) [IF C4_08.3=1] When did you (the input seller) say NPKS should be applied? (# days before planting) (# days after planting) C4_11 Do you think the farmer did (Did you) follow this advice? 1. Yes 2. No 150 | Page C4_12 [C4_08.4==1] What yield did you (the input seller) tell the farmer (you) to expect with NPKS? Kg/acre C4_13 [C4_08.4==1] What yield did you (the input seller) tell the farmer (you) to expect with a standard fertilizer application of TSP, MP and Manure? Kg/acre The following questions are about your thoughts on how to best use fertilizers in jute production. So this is not about what you do, or what you talked about with farmers (input sellers), but what you think would give you the highest profit. C4_14 For the following fertilizers, how much should one apply? WRITE 0 IF NOTHING, 99 IF DOES NOT KNOW Basal (kg/acre) Top dressing (kg/acre) Urea TSP DAP NPKS MOP Gypsum Magnesium sulphate Zinc sulphate Other (specify) C4_15 What should you do with your jute seeds before sowing? DO NOT READ OUT. TICK ALL THAT APPLY. 1. Nothing 2. Treat with Provex 3. Treat with crushed garlic 4. Don’t know C4_16a When sowing desh jute, how many kg of seed should you apply per acre... … when using line sowing? (Don’t know = 9999) C4_16b … when using broadcast sowing? (Don’t know = 9999) C4_17a Now we would like to know when it is better to grow tossa jute, or deshi jute. What is better when you have high lands, where rain water does not stand? 1. Tossa jute 2. Deshi jute 3. Does not matter 4. Don’t know C4_17b What is better when you have low lands, where rain water does stand? 1. Tossa jute 2. Deshi jute 3. Does not matter 4. Don’t know C4_18a Can you apply urea when there is heavy sun? 1. Yes 2. No 3. Don’t know C4_18b Can you apply urea when there is heavy rain? 1. Yes 2. No 151 | Page C4_19 When you use organic manure, should you apply more chemical fertilizer, less chemical fertilizer, or the same amount? 1. More chemical fertilizer 2. Less chemical fertilizer 3. The same amount of chemical fertilizer 4. It doesn’t matter 5. Don’t know C4_20 How many times should you do weeding on your jute plots? C4_21 Which of the following insect attacks can you control by using Diazinon? 1. Bichha insect 2. Letha insect 3. Ghora insect 4. Chelle insect 5. Makor insect 6. Diazinon does not help control any of these insect attacks 7. Don’t know C4_22a Now please tell me when it is best to cut your jute plants. After how many days of sowing… … should you cut your jute plants? (days) C4_22b … should you cut the jute plants to achieve the highest yield? (days) C4_22c … should you cut the jute plants to achieve the highest fiber quality? (days) C4_23a Now I have some questions about the difference between traditional retting and ribbon retting. Which is better for the environment? 1. Traditional retting 2. Ribbon retting 3. Does not matter 4. Don’t know C4_23b Which is better for your health? 1. Traditional retting 2. Ribbon retting 3. Does not matter 4. Don’t know C4_23c Which is better in cases where water is very scarce? 1. Traditional retting 2. Ribbon retting 3. Does not matter 4. Don’t know C4_24 Which is the best to use for retting: banana plant, soil clod, or sack cloth? 1. Banana plant 2. Soil clod 3. Sack cloth 4. Don’t know C4_25 Should you be retting thick and thin plants together or separately? 1. Together 2. Separately 3. Does not matter 4. Don’t know C4_26 What should your jute look like in order to get paid a higher price? 1. Length must be 8-10 feet 152 | Page DO NOT READ OUT, PROBE IF THERE IS ANYTHING ELSE. TICK ALL THAT APPLY 2. Fiber color must be bright 3. Fiber must be spotless and stick less 4. Fiber must be fine and strong 5. Don’t know 153 | Page Module C5: Extension Services for Jute Production (All groups) THIS MODULE HAS TWO OBJECTIVES: MEASURE THE EXPOSURE TO AVC ACTIVITIES, AND MEASURE EXPOSURE TO OTHER NGO ACTIVITIES THAT MAY HAVE IMPACTED FARMERS, SO THAT WE CAN CONTROL FOR THAT WHEN ESTIMATING HOW EXPOSURE TO AVC ACTIVITIES AFFECTS OUR OUTCOME VARIABLES. No. Question Field C5_01 In 2016, did you or anyone from your household participate in the following activities related to [production of jute]/[selling inputs for jute production]? 1. AVC-sponsored jute training (provided by Gono Unnayan Prochesta (GUP), Society Development Committee (SDC), Sheba Manab Kallyan Kendra (SMKK), or Prova Society). 2. AVC/NAAFCO raffle in the village 3. AVC/NAAFCO promotional fair 4. Any other AVC/NAAFCO activities 5. Information about jute production from the private sector, e.g. input sellers, dealers, wholesalers, traders or seed companies (but not AVC/NAAFCO) [Include input seller trainings from seed company/dealers in this category] 6. Other extension services for jute production not provided by AVC (including any agricultural advice you may have received in person or over the phone) 1. Yes 2. No C5_02 How many hours did you spend in total in this …? RECORD MINUTES AS FRACTIONS OF AN HOUR, ie. 15 minutes = 0.25 (Hours) C5_03 Which topics were covered as part of this …? CATEGORIES FROM MODULE C6 C5_04 Who provided this extension service for jute production in the Kharif-I (Aus: May – July 2016) season? [ONLY ASK FOR C5_01 ITEM (5)/(6)] 1. Department of Agricultural Extension 2. NGOs (specify) 3. Private Business (specify) 4. Other (specify) C5_05 Extension service administration level [ONLY ASK FOR C5_01 ITEM (4)/(5)] 1. Upazila 2. Union 3. Village 4. Farmer association group 5. Group of households 6. Individual household 7. Other (specify) C5_06 How much did you have to pay the organizers to participate in this … (Taka) C5_07 How much cash did you receive in compensation for participating in this … INCL. TRANSPORT ALLOWANCE (Taka) C5_08 How many hours did you spend with input suppliers as part of this …? INCL. LINKAGE MEETINGS (Hours) C5_09 How many hours did you spend with jute buyers as part of this …? INCL. LINKAGE (Hours) 154 | Page MEETINGS C5_10 How useful was this activity for you as a farmer (input seller)? 1 – Activity was useless 2 – Activity was not useless, but also not very useful 3 – Activity was a little bit useful 4 – Activity was very much useful C5_11 What topic was the most useful for you in this extension service? SHOW CODES SELECTED IN C5_03 C5_12 What topic was the second-most useful for you in this extension service? IDEM 155 | Page Module C6: Use of Improved Technologies and Practices in Jute Production (All Groups) ONLY IF C2_02 == 1 [PLANTED JUTE IN KHARIF-I (AUS: MAY – JULY 2016).] OR IF GROUP 3 Interventions Did you use (sell) … in the Kharif-I (Aus: May – July 2016) season for jute production? C6_01==1 What variety did you use (sell) for jute production? Groups 2/3 Did you ever use (sell) …for jute production? Groups 2/3 What year did you start using (selling) …for jute production? Groups 2/3 What year did you stop using (selling) … for jute production? C6_01==2 Why are you not using (selling) … for jute production? Groups 2/3 What year was … first available in your village? 1. Yes 2. No -> C6_03 List of NPKS brands, and seeds from seller census MULTIPLE CHOICE -> C6_04 1. Yes 2. No -> C6_06 (YEAR) IF C6_01 = 1 -> C6_07 (YEAR) 1. Don’t know about it 2. Too expensive 3. No cash/cannot get credit at time of purchase 4. Uncertain of benefits 5. Too risky 6. Other (specify) (YEAR) C6_01 C6_02 C6_03 C6_04 C6_05 C6_06 C6_07 NKPS fertilizers Also Group 1 Also Group 1 Also Group 1 Also Group 1 Also Group 1 Improved and/or certified seed (Products related to) Pest management (percing, IPM/ICM, physical method) (Products related to) Disease management (fungicide, bactericide, nematicide, virucide, IPM/ ICM, water control management) (Products related to) Composting Irrigation (Products related to) Climate change mitigation or adaptation Improved commodity sales technologies and practices (GROUPS 1 & 2 ONLY) Improved market information systems technologies and practices (GROUPS 1 & 2 ONLY) 156 | Page Improved packing house technologies and practices (GROUPS 1 & 2 ONLY) Improved transportation (GROUPS 1 & 2 ONLY) Temperature and humidity control (GROUPS 1 & 2 ONLY) Improved quality control technologies and practices (GROUPS 1 & 2 ONLY) Sorting and grading (GROUPS 1 & 2 ONLY) Value added processing (GROUPS 1 & 2 ONLY) 157 | Page Module D: Marketing Module D2: Relationships with Local Jute Input SELLERS (GROUP 1 & GROUP 2) / BUYERS (GROUP 3) ${seller } Have you ever purchased/sol d jute seeds, fertilizer or other inputs to cultivate jute from ${seller}/to ${buyer}? First year you purchase d from ${seller}/ sold to ${buyer} Did you purchase from/sell to them seller in the last Kharif-I season? Which inputs did you purchase from/ sell to them in the last Kharif-I season? Did you purchase/sell any inputs at a discount or for free from/to them? Did you receive/give credit from them to purchase inputs? Group 1 / Group 2 Only Where do you usually buy from ${seller}? Group 1 / Group 2 Only How many minutes does it take to travel to ${location} from your home? Group 1 / Group 2 Only How much did it cost to travel there (per trip)? Group 1 / Group 2 Only Travel Mode 1. Yes 2. No >> D2_10 3. Don’t know him >> Next seller (YEAR) 1. Ye s 2. No 1. Seed 2. NPKS fertilizer 3. Other fertilizer 4. Insecticid e 5. Fungicide 6. Herbicide 1. Yes 2. No 1. Yes 2. No 1. Farm gate (home) 2. Village market (within own village) 3. Village market (outside own village) 4. Town market (within own union) 5. Town market (outside own union) 6. Other (specify) (MINUTES) SKIP IF SOLD AT “Farm gate (home)” (Taka) 1. Walking 2. Bicycle 3. Ricksha w 4. CNG 5. Van 6. Push Van 7. Tractor 8. Truck D2_01 D2_02 D2_03 D2_04a D2_04b D2_04c D2_05 D2_06a D2_06b D2_07 Seller 1 Seller 2 158 | Page … The remaining questions in D2 should only be asked to Group 1 & Group 2 households ${seller } Reasons for buying from ${input seller} READ OUT ALL OPTIONS AND SELECT ALL THAT APPLY IF AT MOST ONE ELEMENT SELECTED  D2_12 Most important reason for buying from ${input seller}  D2_12 Reasons for not buying from ${input seller} Most important reason for not buying from ${input seller} Did ${input seller} provide any information or service related to jute production? NO>>  D2_14 How many minutes did ${input seller} spend providing information related to jute production (per transaction)? 1. Lives nearby 2. Only available seller 3. Buys jute fiber 4. Provides good quality 5. Sells on credit 6. Gives good information 7. Gave a discount 8. Other (specify) OPTIONS BASED ON REASONS SELECTED IN D2_08 1. Too far away 2. Prices too high 3. Provides low quality 4. Does not sell on credit 5. Does not give good information 6. Did not give discount/promotion 7. Other (specify) OPTIONS BASED ON REASONS SELECTED IN D2_10  D2_15 Yes/No Minutes per transaction D2_08 D2_09 D2_10 D2_11 D2_12 D2_13 Seller 1 Seller 2 … ${seller } On a scale of 1 to 10, where 1 is very unsatisfied and 10 is very satisfied, how would you rate your transaction with ${seller}? Consider a 10-step ladder, where step 1 is the least trustworthy and step 10 is the most trustworthy. On which step does ${seller} stand? Overall, on a scale of 1 to 10 how would you rate ${seller}’s knowledge of inputs 159 | Page (1 – 10) (1 – 10) (1 – 10) D2_14 D2_15 D2_16 Seller 1 Seller 2 … D2_18 Who is your preferred input seller? POPULATE CHOICES FROM LIST ABOVE D2_19 Who is your preferred input seller if ${D2_18} is not available? POPULATE CHOICES FROM LIST ABOVE. DO NOT INCLUDE ${D2_18} 160 | Page Module E: Housing (All Groups) No. Question Answer No. Question Answer E_01 Do you own this house, rent this house, or use it for free? 1. Own 2. Rent 3. Use for free 2  E_03 3  E_04 E_0 8 What is your main source of cooking fuel? 1. Electricity 2. Supply gas 3. LPG 4. Kerosene 5. Firewood 6. Dried cow dung 7. Coal 8. Rice bran/sawdust 9. Dried leaves 10. Other (specify) E_0 2 If you chose to rent this house to another person, how much would you be able to receive in rent per month? (TAKA)  E_04 E_0 9 What is your main source of lighting fuel? 1. Electricity (solar panel) 2. Electricity (power grid) 3. Supply gas 4. LPG 5. Kerosene 6. Firewood 7. Dried cow dung 8. Coal 9. Rice bran/sawdust 10. Dried leaves 11. Other (specify) E_0 3 How much rent do you pay per month for this dwelling? (TAKA) E_10 What type of latrine do you use? 1. None (open field) 2. Kutcha (fixed place) 3. Pucca (unsealed) 4. Sanitary without flush (water sealed) 5. Sanitary with flush (water sealed) 6. Community latrine 7. Other (specify) Do not read this question aloud. Please observe the dwelling and enter your response based on your observation. E_11 What is your main source of drinking water? 1. Supply water (piped) inside house 2. Supply water (piped) outside house 3. Own tube well 4. Other’s tube well 5. Community tube well 6. Rainwater 7. Ring well/indara 8. Pond/river/canal 9. Bottled water 10. Shallow tube well for irrigation 11. Deep tube well for irrigation 12. Other (specify) E_0 4 WHAT IS THE CONDITION OF THE DWELLING? 1. No sign of damage 2. Slightly damaged 3. Somewhat damaged 4. Very damaged 13. E_0 5 WHAT IS THE PRIMARY CONSTRUCTION MATERIAL OF THE MAIN DWELLING’S OUTER WALLS? 1. Concrete/brick 2. Tin/CI sheet 3. Wood 4. Mud 5. Bamboo 6. Jute straw 7. Plastic 8. Cardboard/pape r 9. Golpaata/palm leaf 10. Grass/straw 5. Other (specify) 14. 161 | Page E_0 6 WHAT IS THE PRIMARY CONSTRUCTION MATERIAL OF THE MAIN DWELLING’S ROOF? SEE E_05 E_12 What is your main source of water not used for drinking? SEE E_11 E_07 WHAT IS THE PRIMARY CONSTRUCTION MATERIAL OF THE MAIN DWELLING’S FLOOR? SEE E_05 E_13 How do you dispose of garbage? 1. Local authority collects 2. Private firm collects 3. Public garbage pit/hole 4. Own garbage pit/hole 5. Burned/buried 6. Own garbage heap (not pit) 7. Gather in an open place 8. Throw in pond/khaal/beel 9. Other (specify) 162 | Page Module G: Household Expenditures on Goods Other Than Food (All Groups) Module G1: Monthly Recall "I will now ask you about any goods that you or other household members may have purchased or received for free. Please exclude any expenses or investments to run your business or farm. We are interested in items that you obtained for yourself or the household." * In the last 30 days, * did your household obtain any of the following items? Item name Item No. Amount spent on ${item} (in the past 30 days) Value of ${item} received for free (in the past 30 days) G1_01 G1_02_a G1_02_b Fuel, etc. Firewood, cow dung, pit coal, gas, agri by-fuel, etc. used for fire 1 Kerosene, candles, matches, etc. 2 Electricity (MONTHLY BILL. IF SOLAR, ENTER ESTIMATED VALUE) 3 Beauty and hygiene Women's cosmetics and accessories 4 Hair cutting, shaving, razor blades, shaving cream, parlour, salon, etc. 5 Bath soap, shampoo, toothpaste, etc. 6 Washing soap, powder for cloths 7 Vim, dish cleaning supplies, finis, phenyl, etc. 8 Washing / laundry expenses, bleaching powder, soda, etc. 9 Toilet papers 10 Mosquito coil or spray 11 Transportation Long-distance transportation (bus fares, CNG taxis, train, etc.) 12 Short-distance transportation (rickshaw, van, etc.) 13 Bicycle maintenance, tires, tubes repairs etc. 14 Maintenance, repairs etc. for other vehicles (motor-cycle, car, boat, etc.) 15 Petrol, diesel, motor oil, CNG, etc. 16 Other Airtime for mobile phone 17 Other telecom (telephone charges, telegram, postal, courier, etc.) 18 Salaries and wages of guards, gardeners, housekeepers, etc. 19 163 | Page Other contingencies expenses 20 164 | Page Module G2: Annual Recall (All Groups) In the last 12 months did your household obtain any of the following items? Item name Item no. Amount spent on ${item} (in the past 12 months) Value of ${item} received for free (in the past 12 months) G2_01 G2_02a G2_02b Clothing and footwear Ready-made clothing for adults 1 Ready-made clothing for children 2 Napkins, diapers, disposable napkins 3 Clothing material, cloth, fabric, silc, and tailoring expenses 4 Shoes or sandals 5 Household operations Household textiles (bedsheets, quilt, blanket, Katha, Toshok, mosquito netting) 6 Household appliances (refrigerator, vacuum cleaner, stove, fan, A/C, etc.) 7 Household cleaning equipment (brooms, dusters, mops, buckets, etc.) 8 Furniture (bed, chowki, table, chair, sofa, almirah, etc.) 9 Trunks, suitcases 10 Household utensils (glass, china, cutlery, glassware, plates, pots, pans, etc.) 11 Housing Water/sewerage charges 12 Home additions, improvements or painting 13 Maintenance and repair (disaster-related or routine) 14 165 | Page Item name Item no. Amount spent on ${item} (in the past 12 months) Value of ${item} received for free (in the past 12 months) G2_01 G2_02a G2_02b Health Maternity care (midwives, maternity homes, etc.) 15 Medicines, contraceptives, condoms 16 Medical tests (X-ray, blood, urine, etc.) 17 Fees for hospitals, clinics or practitioners (doctors, nurses, homeopath, etc.) 18 Traditional medical services (traditional care, native doctors, etc.) 19 Health-related travel (incl. ambulance) 20 Education School fees 21 Private tutoring 22 Other fees (registration, examination, etc.) 23 Textbooks, note books, stationary 24 Gifts and ceremonies Remittances to others living separately 25 Zakat 26 Fitra 27 Donation / Sadqa 28 Qurbani 29 Expenditure on ceremonies (weddings, births, deaths/funerals, milad, etc.) 30 Photography 31 Parties and presents for a marriage day or birthday 32 Durables Jewellery, clocks, watches 33 Purses, money bags, vanity bags and other personal use items 34 Electronics (radio, TV, camera, phone, etc.) 35 Other Bank / interest charges 36 Other regular fees (cable fees, legal expenses for practitioner fees, etc.) 37 166 | Page Insurance (health, life, general, pension, etc.) 38 167 | Page Module I: Beliefs (All groups) Module I1a: Expectations for Jute Production and Sales (All Groups) Now we would like to play a game in which we ask you some questions about your jute production in the next season. We would like to know two things: how much jute you expect to harvest in the next season and how much money you expect to sell your jute harvest for. We are now going to start with an example. Suppose I planted jute using normal fertilizer (for example: 36kg of urea (N) and 90kg of manure per acre) The minimum amount of jute I could harvest this season is 0 kilograms per acre, in case something very unexpected happens and I cannot produce anything. What would be the maximum amount of jute you could harvest this season if conditions were perfect and nothing bad happened? [X] So your yield would be between 0 and X kg/acre, ok? Here, I have split the potential yields into five different boxes. The first box means that I yielded less than 600 kg of jute per acre. [POINT AT BOX] The second box means that I yielded 601-900 kg of jute per acre. [POINT AT BOX] The third box means that I yielded 901-1050 kg of jute per acre. [POINT AT BOX] The fourth box means that I yielded 1051-1350 kg of jute per acre. [POINT AT BOX] The fifth box means that I yielded more than 1350 kg of jute per acre. [POINT AT BOX] MAKE SURE THAT TO THE ENUMERATOR’S RIGHT, YOU ALWAYS SHOW THE LOWEST AMOUNT, AND TO THE LEFT, ALWAYS THE HIGHEST AMOUNT. Can you tell me how much jute I yield if I choose this box? [POINT AT BOX 5] [CONFIRM THAT THE RESPONDENT ANSWERED "more than 1350 kg/acre". IF HE DOES NOT, REPEAT INSTRUCTIONS TO HELP HIM UNDERSTAND.] Can you tell me how much jute I yield if I choose this box? [POINT AT BOX 3] [CONFIRM THAT THE RESPONDENT ANSWERED "900-1050 kg/acre". IF HE DOES NOT, REPEAT INSTRUCTIONS TO HELP HIM UNDERSTAND.] I am going to use these marbles to represent the likelihood that each of the yields per acre will actually occur. [SHOW MARBLES] The more marbles in each box, the more likely it is that I will harvest that amount of jute. For example, if I put all ten marbles in the first box (less than 600 kg per acre), [PLACE THE MARBLES IN THE BOX] it means that I believe I will harvest less than 600kg per acre of jute with absolute certainty. [CLEAR MARBLES FROM BOX] If I put 5 marbles in the first box (less than 600 kg per acre ) [PLACE THE MARBLES] and 5 marbles in the second box (601-900 kg per acre), [PLACE THE MARBLES] it means that I believe it is equally likely that I will harvest less than 600 kg of jute per acre or 601-900 kilograms per acre of jute, and that I do not expect to harvest more than 900 kg per acre at all. [CLEAR MARBLES FROM BOXES] 168 | Page Now let’s try another example. I will put 1 marble in the first box (less than 600 kg/acre ) [PLACE THE MARBLES], Kilograms per acre Distributio n Marbles Less than 600 kg/acre 1/10 • 600-900 kg/acre 2/10 • • 900-1050 kg/acre 4/10 • • • • 1050-1350 kg/acre 3/10 • • • More than 1350 kg/acre 0/10 2 marbles in the second box (600-900 kg/acre) [PLACE THE MARBLES], 4 marbles in the third box (900-1050 kg/acre) [PLACE THE MARBLES], 3 marbles in the fourth box (1050-1350 kg/acre ) [PLACE THE MARBLES], and no marbles in the fifth box (more than 1350 kg/acre ). This means I believe there is a 1 in 10 chance that I will harvest less than 600 kg per acre [POINT AT THE BOX], and 2 in 10 chance that I will harvest 600-900 kg per acre [POINT AT THE BOX], There is also a 4 in 10 chance that I will harvest 900-1050 kg per acre [POINT AT THE BOX], a 3 in 10 chance that I will harvest 1050-1350 kg per acre [POINT AT THE BOX], a no chance that I will harvest more than 1350 kg per acre. [POINT AT THE BOX] Can you tell me how likely it is that I yield 600-900 kg per acre? [POINT AT BOX 2] [CONFIRM RESPONDENT ANSWERED “2 in 10.” IF HE DOES NOT, REPEAT INSTRUCTIONS TO HELP HIM UNDERSTAND] There is also a 4 in 10 chance that I will harvest 900-1050 kg per acre [POINT AT THE BOX], a 3 in 10 chance that I will harvest 1050-1350 kg per acre [POINT AT THE BOX], a no chance that I will harvest more than 1350 kg per acre. [POINT AT THE BOX] Can you tell me how likely it is that I yield 1050-1350 kg per acre? [POINT AT BOX 4] [CONFIRM RESPONDENT ANSWERED “3 in 10.” IF HE DOES NOT, REPEAT INSTRUCTIONS TO HELP HIM UNDERSTAND] Most likely, I will harvest between 900-1050 kg/acre, which is why I put 4 out of 10 marbles in this box. [POINT AT THE BOX] The chance that I harvest between 1050-1350 kg/acre is a little bit less, so I am putting 1 marble less here. [POINT AT THE BOX] The chance of harvesting between 600-900 kg/acre is less, so I am putting 2 marbles here. [POINT AT THE BOX] And it is very unlikely that I harvest less than 600 kg/acre, but it may happen, so I am putting only 1 marble here. [POINT AT THE BOX] I don’t think there is any chance that I will harvest between more than 1350 kg/acre, so I did not put any marbles here. [POINT AT THE BOX] Is this clear? [CONFIRM RESPONDENT UNDERSTANDS] Now I will ask you questions regarding your expectations for next season’s jute harvest. 169 | Page Yield with urea fertilizer Yield with NPKS fertilizer No. Script Yield per Acre Distrib ution No. Script Yield per Acre Distri butio n I1_01a Now I would like to ask you how much jute you would expect to yield if you carried out your normal activities and applied 36kg of urea fertilizer and 90kg of manure per acre I have five boxes and ten marbles here. The values shown in each box represent different yields per acre of jute with 36 kg of urea fertilizer and 90kg of manure. Every marble represents a possibility that the amounts shown in each block will happen in real life. Your task is to divide the marbles over the five blocks. The amount on the left represents the worst case scenario with the lowest yields per acre. The amount on the right represents the best case scenario with the highest yields per acre. The three amounts in the middle represent yields somewhere in-between the worst and the best case scenarios. Notice that as we move from left to right, we move from the lowest to the highest yields. Remember, these are yields that you would expect with urea fertilizer and manure. Please divide the marbles over the five blocks now. RECORD THE DISTRIBUTION OF THE MARBLES 0- 600 kg I1_02a Now I would like to ask you how much jute you would expect to yield if you carried out your normal activities and applied 75 kg of NPKS fertilizer per acre and 39 kg of urea I have five boxes and ten marbles here. The values shown in each box represent different yields per acre of jute with application of 75 kg NPKS fertilizer and 39 kg of urea Every marble represents a possibility that the amounts shown in each block will happen in real life. Your task is to divide the marbles over the five blocks. The amount on the left represents the worst case scenario with the lowest yields per acre. The amount on the right represents the best case scenario with the highest yields per acre. The three amounts in the middle represent yields somewhere in-between the worst and the best case scenarios. Notice that as we move from left to right, we move from the lowest to the highest yields. Remember, these are yields that you can expect with application of NPKS fertilizer and urea. Please divide the marbles over the five blocks now. RECORD THE DISTRIBUTION OF THE MARBLES 0- 600kg I1_01 b 600– 900 kg I1_02 b 600– 900 kg I1_01c 900 – 1050 kg I1_02c 900 – 1050 kg I1_01 d 1050 – 1450 kg I1_02 d 1050 – 1450 kg I1_01e 1450kg + I1_02e 1450kg+ 170 | Page Trust Game (All Groups) GENERAL INSTRUCTIONS NOW WE WILL BE PLAYING A GAME FOR REAL MONEY. YOU SHOULD UNDERSTAND THAT THIS IS NOT DATA’S OWN MONEY. IT IS MONEY GIVEN TO DATA BY RESEARCHERS TO USE FOR THIS STUDY. YOU SHOULD ALSO UNDERSTAND THAT THIS IS A NEW GAME THAT NONE OF YOU HAVE PLAYED BEFORE. SO IF YOU HAVE HEARD ABOUT A GAME LIKE THIS BEFORE YOU SHOULD TRY TO FORGET EVERYTHING THAT YOU HAVE BEEN TOLD. THIS IS A COMPLETELY DIFFERENT GAME. I WILL EXPLAIN THE GAME TO YOU. IT IS IMPORTANT THAT YOU LISTEN AS CAREFULLY AS POSSIBLE, BECAUSE YOU WILL BE ABLE TO MAKE THE DECISIONS THAT ARE RIGHT FOR YOU ONLY WHEN YOU UNDERSTAND THE GAME. YOU CAN ASK ME QUESTIONS TO BE SURE THAT YOU UNDERSTAND HOW TO PLAY. [FLIPCHART 1] This game is played by pairs. Each pair is made up of a farmer [POINT] and an input seller [POINT]. You will play this game with an input seller (a farmer) from your same union. However, you will not know exactly with whom you are playing, and the input seller (farmer) will not know exactly with whom he is playing either. Only the researchers and my supervisor know who is to play with whom and they will never tell anyone else. [FLIPCHART 2] At the start of the game, you both get some money. Right now, we will use play money, but in a few days, we will replace whatever you earnt with actual money. Ok? It works as follows. We will give both of you 150 Taka [POINT]. Then, you (the farmer) can send some or all of the 150 Taka to the input seller (you). This is done by putting the portion to be sent in an envelope for the seller [POINT]; either all 150 Taka [POINT], or 100 Taka [POINT], or 50 Taka [POINT], or nothing [POINT]. The other money has to be put in another envelope, which is for the farmer. Ok? Both the seller envelope and the farmer envelope are going to my supervisor. My supervisor will take the money from the seller envelope, triple it, and put the tripled amount back in the seller envelope; so then the seller envelope will contain whatever you (the farmer) put in there, times three. Then, the seller envelope goes to the input seller (you). Ok? My supervisor will keep the farmer envelope safe. The money in the farmer envelope will stay the same. So it will not be tripled. Ok? [FLIPCHART 3, START FROM BOTTOM] Now let’s explain what we mean by tripling the money in the seller envelope. - If the farmer puts 150 Taka in the seller envelope, how much does the input seller receive? [450] - If the farmer puts 100 Taka in the seller envelope, how much does the input seller receive? [300] - If the farmer puts 50 Taka in the seller envelope, how much does the input seller receive? [150] - If the farmer puts nothing in the seller envelope, how much does the input seller receive? [0] [FLIPCHART 4] We then ask the input seller (you) whether to return any of this money to you (the farmer). We will give the input seller 150 Taka, then one envelope with the amount from you (the farmer) and another envelope for the input seller (you). We will count the money from you (the farmer), then the input seller (you) will decide how 171 | Page much money to put back in the farmer envelope to send back to you (the farmer) and how much to put in the seller envelope for the input seller (you) to keep: the input seller (you) can choose to return any portion to you (the farmer). [FLIPCHART 5] Now we have the final step: my supervisor will return to you (the farmer) the amount that the input seller (you) decided to send back to you (the farmer). So you (the farmer) will receive the amount that they decided to keep plus the amount that the input seller (you) decides to send back. You (the farmer) will receive all of this money back from my supervisor. [FLIPCHART 6] So the input seller (you) will receive the total amount the input seller (you) placed in the seller envelope (of the 150 taka we gave the input seller (you) plus whatever you (the farmer) sent) minus the amount that the input seller (you) decided to send back to you (the farmer). Is that clear? Here are some examples [FLIP CHART 7]: 1. Imagine that the farmer gives 150 to the input seller. How much does the farmer keep? [NOTHING]. And how much does the farmer put in the seller envelope? [150 TAKA]. The researchers triple this amount, so the input seller gets how much? [150 X 3 = 450, plus the 150 Taka received at the start of the game = 600 Taka in total]. Then the input seller decides how much to send back to the farmer. - Suppose he decides to return 450 Taka. How much will the farmer then receive? [450 Taka] And the input seller? [600-450 = 150 Taka] - Suppose he decides to return 150 Taka. How much will the farmer then receive? [150 Taka] And the input seller? [600-150=450 Taka] - Suppose he decides to return 0 Taka. How much will the farmer then receive? [0 Taka] And the input seller? [600-0=600 Taka] Is this clear? Now let’s try another example. [FLIP CHART 8]: 2. Imagine that the farmer gives 100 to the input seller. How much does the farmer keep? [50 TAKA] And how much does the farmer put in the envelope? [100 TAKA] The researchers triple this amount, so the input seller gets how much? [100 X 3 = 300, plus the 150 Taka from the start of the game, so 450 Taka in total]. Then the input seller decides how much to send back to the farmer. - Suppose he decides to return 400 Taka. How much will the farmer then receive? [400 + 50 = 450 Taka] And the input seller? [450 – 400 = 50 Taka] - Suppose he decides to return 250 Taka. How much will the farmer then receive? [250 + 50 = 300 Taka] And the input seller? [450 – 250 = 200 Taka] - Suppose he decides to return 50 Taka. How much will the farmer then receive? [50 + 50 = 100 Taka] And the input seller? [450 – 50 = 400 Taka] Is this clear? Now let’s try another example. 172 | Page [FLIP CHART 9]: 3. Imagine that the farmer gives 50 to the input seller. How much does the farmer keep? [100 TAKA] And how much does the farmer put in the envelope? [50 TAKA] The researchers triple this amount, so the input seller gets how much? [50 X 3 = 150, plus the 150 Taka from the start of the game, so 300 Taka in total]. Then the input seller decides how much to send back to the farmer. - Suppose he decides to return 100 Taka. How much will the farmer then receive? [100 + 100 = 200 Taka] And the input seller? [300 – 100 = 200 Taka] - Suppose he decides to return 50 Taka. How much will the farmer then receive? [50 + 100 = 150 Taka] And the input seller? [300 – 50 = 250 Taka] - Suppose he decides to return 0 Taka. How much will the farmer then receive? [0 + 100 = 100 Taka] And the input seller? [300 – 0 = 300 Taka] Is this clear? Now let’s try a final example. [FLIP CHART 10]: 4. Imagine that the farmer gives 0 to the input seller. How much does the farmer keep? [150 TAKA] And how much does the farmer put in the envelope? [0 TAKA] So the input seller gets how much? [0, plus the 150 Taka from the start of the game, so 150 Taka in total]. Then the input seller decides how much to send back to the farmer. - Suppose he decides to return 100 Taka. How much will the farmer then receive? [150 + 100 = 250 Taka] And the input seller? [150 – 100 = 50 Taka] - Suppose he decides to return 50 Taka. How much will the farmer then receive? [150 + 50 = 200 Taka] And the input seller? [150 – 50 = 100 Taka] - Suppose he decides to return 0 Taka. How much will the farmer then receive? [150 + 0 = 150 Taka] And the input seller? [150 – 0 = 150 Taka] Is this clear? Note that if the farmer puts more in the seller envelope, the two of you will earn more together. The more the farmer puts in the seller envelope, the more money there is to be tripled by the researchers. So combined, the two of you earn more together. However, it is not necessarily the case that the farmer earns more when putting more money in the seller envelope; it is entirely up to the input seller to decide what to return. So the farmer could end up with more than 150 Taka or less than 150 Taka, depending on how much the input seller returns. Is that clear? 173 | Page Group 1 / Group 2 Households Only You will play this game with 3 different input sellers, so there is a set of seller and farmer envelopes for 3 different input sellers. But the researchers will randomly pick only one of the three sets of envelopes, then keep the farmer envelope from that set, and triple the amount in the seller envelope from that set, and send only that seller envelope to the input seller. So the other 2 input sellers will not get an envelope from you; they will get envelopes from a different farmer, also randomly picked. Ok? Each of your sets of buyer and farmer envelopes has an equal chance of being picked, and it will not depend on how much money you put in the envelope. I will ask you to put money in the envelope for each input seller separately. We will do this as follows. I give you the first set of seller and farmer envelopes and tell you to which input seller the seller envelope will go if randomly picked. Then you decide how much money to put in the seller envelope and how much in the farmer envelope while I turn away – so you can make your choice in private. Then, once you are ready, I will give you the second set of seller and farmer envelopes and tell you to which input seller that seller envelope will go if randomly picked. Then again you decide how much money to put in that second seller envelope while I turn away. Finally, once you are ready with the second set of envelopes, I will tell you to which input seller the third set of envelopes will go if randomly picked. Then again you decide how much money to put in the third seller envelope, again while I am turned away. And then we are done. The researchers will then randomly pick one of these three sets of farmer and seller envelopes. My supervisor will then triple the money in the seller envelope. The seller envelope is then sent to the input seller. We will not tell him that the money is coming from you. So he will never find out that you were playing with him. Ok? DO THIS FOR SELLER 1, SELLER 2 AND SELLER 3 (READ OUT THEIR NAMES ONLY JUST BEFORE GIVING THEM THE ENVELOPE SETS, SEQUENTIALLY; WHEN DOING ENVELOPE 1, THEY SHOULD NOT YET KNOW WHO IS INPUT SELLER 2 OR 3). You will now play this game with [SELLER X]. Here is your 150 Taka. [At this point 150 dummy Takas/play money is placed in front of the farmer.] While I am turned away, you must put the amount of money you want to keep in this farmer envelope, and you must put the money you want to be tripled in this seller envelope. You can give [SELLER X] nothing, 50 Taka, 100 Taka, or 150 Taka. If we select this set of envelopes, the researchers will triple the amount in the seller envelope and give it to [SELLER X]. He will then choose how much to return to you. We give you the amount that he returns, plus the amount that is in your farmer envelope. Ok? [NOW TURN AWAY] [CONFIRM THAT THE RESPONDENT HAS MADE A DECISION, THEN TURN BACK AROUND- CHECK TO MAKE SURE THAT THE ENVELOPE MATCHES THE SELLER NAME] Ok we are done now. The researchers will randomly pick one set of farmer and seller envelopes and send the seller envelope from that set to an input seller. The input seller will then decide how much to return to you. In a few days from now, we will let you know how much you earnt and give you the money. Do you have any questions for me now? 174 | Page Group 3 Households Only You will play this game with 4 different farmers, so there is a seller envelope for 4 different farmers, but the researchers will randomly pick only one of the four sets of envelopes. The money that you put in the farmer envelope for the randomly selected farmer will go to that farmer; and the money you decided to keep from that farmer will go to you. This is the money that you put in the selected seller envelope. So only your choice for that set will be paid by my supervisor. Ok? Each of your sets of buyer and farmer envelopes has an equal chance of being picked, and it will not depend on how much money you put in the envelope. I will ask you to put money in the envelope for each farmer separately. We will do this as follows. I give you the first envelope and tell you how much is in it. Then you decide how much money to put back in the envelope while I turn away – so you can make your choice in private. Then, once you are ready, I will give you the second envelope and tell you again how much is in it. Then again you decide how much money to put back in that second envelope while I turn away. We follow this procedure for all envelopes. And then we are done. Ok? The researchers will then give the farmer the money in the envelope, plus whatever the farmer decided to put in the farmer envelope. We will not tell him that the money is coming from you. So he will never find out that you were playing with him. Ok? DO THIS FOR ALL FOUR ENVELOPES. (FOR EACH ENVELOPE SAY THE AMOUNT IN THE ENVELOPE ONLY JUST BEFORE GIVING THEM THE ENVELOPE IN SEQUENCE; WHEN DOING ENVELOPE 1, THEY SHOULD NOT YET KNOW HOW MUCH IS IN ENVELOPE 2 ETC.). Let’s do the game for the first / second / third / etc. envelope. Here is your 150 Taka to start with. And here is the envelope with from farmer 1/2/3/4. There is [… insert amount …] Taka in the envelope. So the farmer sent you [… insert amount/3 …] Taka out of his initial 150 Taka. Here is an envelope for you to decide what to send back to the farmer. While I am turned away, please put the amount you want to keep back into the first envelope and the amount you want to send to the farmer into the second envelope. You can give the farmer as much as you like. [NOW TURN AWAY] [CONFIRM THAT THE RESPONDENT HAS MADE A DECISION, THEN TURN BACK AROUND- CHECK TO MAKE SURE THAT THE ENVELOPE MATCHES THE FARMER NUMBER] Ok we are done now. One of the farmers will be randomly selected and you will receive real money in the amount that you decided to keep when playing with that farmer. My supervisor will also give the farmers real money for the amount they earnt; so that is the amount that you gave back in the envelope for the farmer, plus the amount the farmer decided to keep. Do you have any questions for me now? All Households ENUMERATOR: Were you able to complete the trust game successfully for all rounds? ENUMERATOR: Describe any problems you had in completing the trust game. END OF MAIN FORM 175 | Page MALE FORM MODULE A: HOUSEHOLD IDENTIFICATION (GROUP 1, 2 & 3) Variable IDENTIFICATION Response date DATE interviewer TEAM [SUPERVISOR NAME/CODE] & INTERVIEWER [NAME/CODE] hhid HOUSEHOLD IDENTIFICATION NUMBER [NUMERICAL] Household verification (1): HOUSEHOLD INFORMATION FOR ID ${hhid} A_02 DISTRICT / UPAZILA [CODED] This information will be pre-filled by the survey program. Interviewers should review the information and ensure that they are interviewing the correct respondent. If the preload information does not conform to the correct respondent, or if there is any doubt over the status of the person to be interviewed, the interviewer should not proceed with the interview and immediately contact their supervisor. A_03 UNION [CODED] A_04 VILLAGE [CODED] HOUSEHOLD HEAD: A_08 NAME A_12 GENDER HOUSEHOLD HEAD’S FATHER A_10 NAME A_05 HOUSEHOLD LOCATION/LANDMARK [STRING] MAIN RESPONDENT IN PREVIOUS ROUND [PRELOADED NAME] CHECK THE ABOVE INFORMATION CAREFULLY AND CHOOSE "YES" TO PROCEED IF YOU ARE SURE YOU ◻ 1 – Yes ◻ 2 – No * 176 | Page ARE AT THE RIGHT HOUSEHOLD. A_06 CAN YOU START THE INTERVIEW WITH THE MALE RESPONDENT IDENTIFIED IN FEMALE FORM? ◻ 1 – Yes -> A_09 ◻ 2 – No -> A_07 * YOU CANNOT PROCEED UNLESS YOU HAVE FOUND THE CORRECT HOUSEHOLD. EITHER SPEAK TO THE RESPONDENT AGAIN TO CHECK THE DETAILS OR CANCEL THE INTERVIEW AND CONTINUE TO LOOK FOR THE RIGHT HOUSEHOLD. A_07 WHY CAN YOU NOT START THE INTERVIEW? AFTER SELECTING AN ANSWER -> END ◻ 1 – Moved ◻ 2 – Not found ◻ 3 – Refused ◻ 4 – Unavailable ◻ 5 – Other (specify) RESPONDENT INFORMATION male_resp Please enter the NAME of the respondent for this form, identified in the female form. The respondent should be the primary jute farmer/input seller in the household. resp_age Please provide the age of the respondent for this form resp_sex Please provide the sex of the respondent for this form roster_list Group 2 Households only Please enter the name, age, sex & PID of each member of the household Member 1 Member 2 … 177 | Page male_resp Group 2 Households only From the list of members, please select the ${male_resp} 178 | Page CONSENT OF RESPONDENT Has consent been obtained in the female form or main form? 1.Yes -> NEXT MODULE 2. No -> Obtain informed consent Good morning/afternoon. I am ________ from the Data Analysis and Technical Assistance Limited (DATA), a Bangladeshi research organization based in Dhaka. Together with the International Food Policy Research Institute (IFPRI), we are conducting a survey that will provide IFPRI with necessary information to carry out research that is designed to help promote the welfare of Bangladeshis; particularly, to improve food consumption and nutrition of the people and women’s status, and to enhance agricultural development and income generation. Your household has been chosen by a random selection process. We are inviting you to be a participant in this study. We value your opinion and there are no wrong answers to the questions we will be asking in the interview. We will use approximately 3-4 hours of your time to collect all the information. If you prefer, we can do the interview in two visits. There will be no cost to you other than your time. There will be no risk as a result of your participating in the study. Your participation in this research is completely voluntary. You are free to withdraw your consent and discontinue participation in this study at any time. This study is conducted anonymously. You will only be identified through code numbers. Your identity will not be stored with other information we collect about you. Your responses will be assigned a code number, and the list connecting your name with this number will be kept in a locked room and will be destroyed once all the data has been collected and analyzed. Any information we obtain from you during the research will be kept strictly confidential. Your participation will be highly appreciated. The answers you give will help provide better information to policy-makers, practitioners and program managers so that they can plan for better services that will respond to your needs. The researcher read to me orally the consent form and explained to me its meaning. I agree to take part in this research. I understand that I am free to discontinue participation at any time if I so choose, and that the investigator will gladly answer any question that arise during the course of the research. . Contact Person: Name of the Principal Investigator (PI): _____________ Address: Tel: __________; E-mail of PI: _________ Signature of the Enumerator: _____________ Date: /_____/_____/_____/ consent_m “Do you agree to be interviewed for the purposes of this study?” no_consent WHY DOES THE RESPONDENT NOT GIVE CONSENT?  END INTERVIEW 179 | Page MODULE G2: ROLE IN HOUSEHOLD DECISION-MAKING AROUND PRODUCTION AND INCOME (Group 2 only) Now I’d like to ask you some questions about your participation in certain types of work activities and on making decisions on various aspects of household life. Did you [NAME] participate in [ACTIVITY] in the past 12 months (that is, during the last [one/two] cropping seasons), from [PRESENT MONTH] last year to [PRESENT MONTH] this year? When decisions are made regarding [ACTIVITY], who is it that normally takes the decision? ENTER UP TO THREE (3) MEMBER IDs IF RESPONSE IS MEMBER ID (SELF) ONLY  G2.05 OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98  NEXT ACTIVITY How much input did you have in making decisions about [ACTIVITY]? USE CODE G2↓ To what extent do you feel you can participate in decisions regarding [ACTIVITY] if you want(ed) to? CIRCLE ONE To what extent are you able to access information that you feel is important for making informed decisions regarding [ACTIVITY]? CIRCLE ONE How much input did you have in decisions about how much of the outputs of [ACTIVITY] to keep for consumption at home rather than selling? USE CODE G2↓ How much input did you have in decisions about how to use income generated from [ACTIVITY]? USE CODE G2↓ ACTIVITY G2.01 G2.02 G2.03 G2.04 G2.05 G2.06 G2.07 ID #1 ID #2 ID #3 A Staple grain farming and processing of the harvest: grains that are grown primarily for food consumption (rice, maize, wheat, millet) YES…...1 NO…….2  ACTIVITY B NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT B Horticultural (gardens) or high value crop farming and processing of the harvest YES…...1 NO…….2  ACTIVITY C NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT C Large livestock raising (cattle, buffaloes) and processing of milk and/or meat YES…...1 NO…….2  ACTIVITY D NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT D Small livestock raising (sheep, goats, pigs) and processing of milk and/or meat YES…...1 NO…….2  ACTIVITY E NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT E Poultry and other small animals raising (chickens, ducks, turkeys) and processing of eggs and/or YES…...1 NO…….2  ACTIVITY F NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT 180 | Page meat ◻ CODE G2 LITTLE TO NO INPUT IN DECISIONS 1 INPUT INTO SOME DECISIONS 2 INPUT INTO MOST OR ALL DECISIONS 3 NOT APPICABLE / NO DECISION MADE 98 Did you [NAME] participate in [ACTIVITY] in the past 12 months (that is, during the last [one/two] cropping seasons), from [PRESENT MONTH] last year to [PRESENT MONTH] this year? When decisions are made regarding [ACTIVITY], who is it that normally takes the decision? ENTER UP TO THREE (3) MEMBER IDs IF RESPONSE IS MEMBER ID (SELF) ONLY  G2.05 OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98  NEXT ACTIVITY How much input did you have in making decisions about [ACTIVITY]? USE CODE G2↓ To what extent do you feel you can participate in decisions regarding [ACTIVITY] if you want(ed) to? CIRCLE ONE To what extent are you able to access information that you feel is important for making informed decisions regarding [ACTIVITY]? CIRCLE ONE How much input did you have in decisions about how much of the outputs of [ACTIVITY] to keep for consumption at home rather than selling? USE CODE G2↓ How much input did you have in decisions about how to use income generated from [ACTIVITY]? USE CODE G2↓ ACTIVITY G2.01 G2.02 G2.03 G2.04 G2.05 G2.06 G2.07 ID #1 ID #2 ID #3 F Fishpond culture YES…...1 NO…….2  ACTIVITY G NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT G Non-farm economic activities (running a small business, self￾employment, buy-and-sell) YES…...1 NO…….2  ACTIVITY H NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT H Wage and salary employment (work that is paid for in cash or in￾kind, including both agriculture and other wage work) YES…...1 NO…….2  ACTIVITY I NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT 181 | Page I Large, occasional household purchases (bicycles, land, transport vehicles) NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT J Routine household purchases (food for daily consumption or other household needs) NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUMEXTENT TO A HIGH EXTENT CODE G2 LITTLE TO NO INPUT IN DECISIONS 1 INPUT INTO SOME DECISIONS 2 INPUT INTO MOST OR ALL DECISIONS 3 NOT APPICABLE / NO DECISION MADE 98 182 | Page MODULE G3(A): ACCESS TO PRODUCTIVE CAPITAL (GROUP 2 ONLY) G3.01-G3.05 (GROUP 2 ONLY) Now I’d like to ask you specifically about your household’s land. QUESTION RESPONSE G3.01. Does anyone in your household currently own or cultivate land? YES……..1 NO………2  G3.06, ITEMA G3.02. Who generally makes decisions about what to plant on this land? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER 94 NOT APPLICABLE 98 ID #1 ID #2 ID #3 G3.03. Do you [NAME] solely or jointly cultivate any land? CIRCLE ONE YES, SOLELY YES, JOINTLY YES, SOLELY AND JOINTLY NO G3.04. Who generally makes decisions about what to plant on the land that you yourself cultivate? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER 94 NOT APPLICABLE 98 ID #1 ID #2 ID #3 G3.05. Do you own any of the land owned or cultivated by your household? CIRCLE ONE YES, SOLELY YES, JOINTLY YES, SOLELY AND JOINTLY NO 183 | Page Now I’d like to ask you about a number of items that could be used to generate income. Does anyone in your household currently have any [ITEM]? Do you [NAME] own any [ITEM]? CIRCLE ONE ITEM G3.06 G3.07 A Large livestock (cattle, buffaloes) YES……..1 NO………2  ITEM B YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 B Small livestock (sheep, goats, pigs) YES……..1 NO………2  ITEM C YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 C Poultry and other small animals (chickens, ducks, turkeys) YES……..1 NO………2  ITEM D YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 D Fish pond or fishing equipment YES……..1 NO………2  ITEM E YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 E Non-mechanized farm equipment (hand tools, animal-drawn plough) YES……..1 NO………2  ITEM F YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 F Mechanized farm equipment (tractor-plough, power tiller, treadle pump) YES……..1 NO………2  ITEM G YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 G Non-farm business equipment (solar panels used for recharging, sewing machine, brewing equipment, fryers) YES……..1 NO………2  ITEM H YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 H House or building YES……..1 NO………2  ITEM I YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 I Large consumer durables (refrigerator, TV, sofa) YES……..1 NO………2  ITEM J YES, SOLELY 1 YES, JOINTLY 2 184 | Page YES, SOLELY AND JOINTLY 3 NO 4 Does anyone in your household currently own any [ITEM]? Do you [NAME] own any [ITEM]? CIRCLE ONE ITEM G3.06 G3.07 J Small consumer durables (radio, cookware) YES……..1 NO………2  ITEM K YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 K Cell phone YES……..1 NO………2  ITEM L YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 L Other land not used for agricultural purposes (pieces/plots, residential or commercial land) YES……..1 NO………2  ITEM M YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 M Means of transportation (bicycle, motorcycle, car) YES……..1 NO………2  MODULE G3(B) YES, SOLELY 1 YES, JOINTLY 2 YES, SOLELY AND JOINTLY 3 NO 4 185 | Page MODULE G3(B): ACCESS TO FINANCIAL SERVICES (GROUP 2 ONLY) Next I’d like to ask about your household’s experience with borrowing money or other items (in-kind) in the past 12 months. Would you or anyone in your household be able to take a loan or borrow cash/in-kind from [SOURCE] if you wanted to? Has anyone in your household taken any loans or borrowed cash/in-kind from [SOURCE] in the past 12 months? CIRCLE ONE Who made the decision to borrow from [SOURCE] most of the time? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 Who makes the decision about what to do with the money or item borrowed from [SOURCE] most of the time? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 Who is responsible for repaying the money or item borrowed from [SOURCE]? ENTER UP TO THREE (3) MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 LENDING SOURCES G3.08 G3.09 G3.10 G3.11 G3.12 ID #1 ID #2 ID #3 ID #1 ID #2 ID #3 ID #1 ID #2 ID #3 A Non-governmental organization (NGO) YES...…….1 NO………..2  SOURCE B MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE B DON’T KNOW 97 B Formal lender (bank/financial institution) YES...…….1 NO………..2  SOURCE C MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE C DON’T KNOW 97 C Informal lender YES...…….1 NO………..2  SOURCE D MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE D DON’T KNOW 97 D Friends or relatives YES...…….1 NO………..2  SOURCE E MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE E DON’T KNOW 97 E Group based micro-finance or lending including VSLAs / SACCOs YES...…….1 NO………..2  SOURCE F MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 SOURCE F DON’T KNOW 97 F Informal credit / savings groups (.e.g., merry-go￾rounds, tontines, funeral societies, etc.) YES...…….1 NO………..2  G3.13 MAYBE.….3 YES, CASH 1 YES, IN-KIND 2 YES, CASH AND IN-KIND 3 NO 4 G3.13 DON’T KNOW 97 186 | Page ◻ G3.13 An account can be used to save money, to make or receive payments, or to receive wages or financial help. Do you, either by yourself or together with someone else, currently have an account at any of the following places: a bank or other formal institution (e.g., post office)? YES NO DON’T KNOW 187 | Page MODULE G4: TIME ALLOCATION (GROUP 2 ONLY) G4.01: PLEASE RECORD A LOG OF THE ACTIVITIES FOR THE INDIVIDUAL IN THE LAST COMPLETE 24 HOURS (STARTING YESTERDAY MORNING AT 4 AM, FINISHING 3:59 AM OF THE CURRENT DAY). THE TIME INTERVALS ARE MARKED IN 15 MIN INTERVALS. MARK ONE ACTIVITY FOR EACH TIME PERIOD BY ENTERING THE CORRESPONDING ACTIVITY CODE IN THE BOX. G4.02: CHECK THE BOX BELOW IF THE RESPONDENT WAS CARING FOR CHILDREN WHILE PERFORMING EACH ACTIVITY. Now I’d like to ask you about how you spent your time during the past 24 hours. We’ll begin from yesterday morning, and continue through to this morning. This will be a detailed accounting. I’m interested in everything you did (i.e. resting, eating, personal care, work inside and outside the home, caring for children, cooking, shopping, socializing, etc.), even if it didn’t take you much time. I’m particularly interested in agricultural activities such as farming, gardening, and livestock raising whether in the field or on the homestead. I’m also interested in how much time you spent caring for children, especially if it happened while you did some other activity (e.g., collecting water while carrying a child or cooking while watching after a sleeping child). Night Morning Day 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 G4.01 Activity (WRITE ACTIVITY CODE) G4.02 Did you also care for children? YES..…CHECK BOX NO…LEAVE BLANK YES CHECK BOX NO LEAVE BLANK □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ Day Evening Night 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 24:00 1:00 2:00 3:00 G4.01 Activity (WRITE ACTIVITY CODE) G4.02 Did you also care for children? YES CHECK BOX NO LEAVE BLANK □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ □ ACTIVITY CODES FOR G4.01 A B C D E F G H Horticultural (gardens) or I Large livestock r J Small livestock rais K Poultry and oth (chickens, ducks, turkeys) L M Commuting N Shopping / getting service (incl. he O Weaving / sewing P Q Domestic work (incl. fetching water and c R Carin S Caring for adults T Traveling (not for w U V W X G4.03.In the last 24 hours did you work (at home or outside of the home including chores or other domestic activities) less than usual, about the same as usual, or more than usual? FOR FEMALES ONLY: DOES RESPONDENT HAVE A CHILD G4.04. If you wanted to do something (livelihood-related, training-related, self￾care) and could not take your child with you, is there someone who could care G4.05. Who? ENTER UP TO THREE (3) ID #1 ID #2 ID #3 188 | Page UNDER 5 YEARS OLD? YES...…….1  G4.04 NO………..2  MODULE G5 for your child in your absence? YES...…….1  G4.05 NO………..2  MODULE G5 MEMBER IDs OTHER CODES: NON-HH MEMBER...….94 NOT APPLICABLE….…98 LESS THAN USUAL ABOUT THE SAME AS USUAL MORE THAN USUAL IF RESPONDENT IS MALE  MODULE G5 189 | Page MODULE G5: GROUP MEMBERSHIP (GROUP 2 ONLY) Now I’m going to ask you about groups in the community. These can be either formal or informal and customary groups. Is there a [GROUP] in your community? Is this group composed of all male or female or mixed-sex members? Are you an active member of this [GROUP]? To what extent do you feel like you can influence decisions in this [GROUP]? To what extent does this [GROUP] influence life in the community beyond the group activities? GROUP CATEGORIES G5.01 G5.02 G5.03 G5.04 G5.05 A Agricultural / livestock / fisheries producer’s group (including marketing groups) YES 1 NO 2 DON’T KNOW 97 GROUP B ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP B NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT B Water users’ group YES 1 NO 2 DON’T KNOW 97 GROUP C ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP C NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT C Forest users’ group YES 1 NO 2 DON’T KNOW 97 GROUP D ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP D NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT D Credit or microfinance group (including Grameen, SACCOs / merry-go-rounds / VSLAs) YES 1 NO 2 DON’T KNOW 97 GROUP E ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP E NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT E Mutual help or insurance group (including burial societies) YES 1 NO 2 DON’T KNOW 97 GROUP F ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP F NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT F Trade and business association group YES 1 NO 2 DON’T KNOW 97 GROUP G ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP G NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT G Civic group (improving community) or charitable group (helping others) YES 1 NO 2 DON’T KNOW 97 GROUP H ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP H NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT H Religious group YES 1 NO 2 DON’T KNOW 97 GROUP I ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  GROUP I NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT I Other (specify): _______________________ YES 1 NO 2 DON’T KNOW 97 MODULE G6 ALL MALE ALL FEMALE 2 MIXED SEX DON’T KNOW YES……1 NO..……2  MODULE G6 NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT NOT AT ALL SMALL EXTENT MEDIUM EXTENT HIGH EXTENT 190 | Page If G5.01A = YES, COMPLETE THE QUESTIONS ON PRODUCER GROUPS BELOW (Group 2 only) No. Question Answer G5_06 Is this producer group related to jute? 3. Yes 4. No NO  END MODULE G5_07 How many other active members does this producer group have? (QUANTITY) G5_08 What year did you join this producer group? (YEAR) G5_09 Did you ever engage in any of the following activities for jute through this producer group? IF AT MOST ONE ACTIVITY SELECTED  D2_10 3. Yes 4. No G5_10 Which of these activities is your main reason for actively participating in this producers group? ☐${code2} G5_11 In the last ${period} season, did you ${CODE2} through this producer group? ONLY SELECT FROM ACTIVITIES SELECTED IN D2_07 IF AT LEAST ONE ACTIVITY SELECTED  D2_11 3. Yes 4. No SELL OUTPUT (PROCESSED OR UNPROCESSED) D2_11 G5_12 What is the main reason why farmers do not always sell or bulk their output via this producer group?" 8. Does not bulk group members' output 9. Does not provide a good price for sold output 10. May not be able to sell the output 11. Corruption / money from sales disappears 12. Does not pay out timely for output sold 13. Does not reward quality / combines output with different quality levels 14. Other (specify) G5_13 In the last ${period} season, what type of buyer did this producer group mainly sell to? 13. Village collector 14. Wholesaler 15. Cold storage owner 16. Wholesaler to cold storage 17. Collection center of company 18. Processing farm 19. Cooperative society 20. Farmer society 21. Retailer 22. Consumer 23. Hotel/restaurant 24. Other (specify) 191 | Page 192 | Page No. Question Answer G5_14 In the next 12 months, are you planning to engage in any of the following activities through your producer group? ◻ Buy inputs for jute ◻ Sell jute (before post-harvest processing) ◻ Process jute after harvest ◻ Sell processed jute (after post-harvest processing) ◻ Participate in agricultural training for jute ◻ Any other service for jute (specify) 3. Yes 4. No Now I’d like you to think about how much jute you and other farmers would sell through your producer group, depending on the price the group offers. In the next ${period} season, of all the jute you are planning to sell, what percentage are you planning to sell though your producer group if the price the group offers is… G5_15 …higher than the price that other buyers offer for jute? (PERCENTAGE) G5_16 …the same as the price that other buyers offer for jute? (PERCENTAGE) G5_17 …lower than the price that other buyers offer for jute? (PERCENTAGE) In the next ${period} season, on average, what percentage of their jute do you think other members will sell through your producer group if the price the group offers is… G5_18 …higher than the price that other buyers offer for jute? (PERCENTAGE) G5_19 …the same as the price that other buyers offer for jute? (PERCENTAGE) G5_20 …lower than the price that other buyers offer for jute? (PERCENTAGE) G5_21 Do you expect this producer group will offer a higher, lower, or the same price for jute as other buyers you can sell to? 4. Higher price 5. Lower price 6. Same price G5_22 Does this producer group have any rules or by-laws on how much of your jute members should sell through the group? 3. Yes 4. No NO  END MODULE 193 | Page G5_23 DESCRIBE RULES/BY-LAWS. PROBE FOR MINIMUM QUANTITY AND WHAT HAPPENS IF QUANTITY IS NOT DELIVERED (E.G. FINES). 194 | Page MODULE G6. PHYSICAL MOBILITY (Group 2 only) QUESTION RESPONSE FOR G6.01 - G6.06: USE CODE G6↓ G6.01 How often do you visit an urban center? G6.02 How often do you go to the market / haat / bazaar? G6.03 How often do you go to visit family or relatives? G6.04 How often do you go to visit a friend / neighbor’s house? G6.05 How often do you go to the hospital / clinic / doctor (seek health service)? G6.06 How often do you go to a public village gathering / community meeting / training for NGO or programs? G6.07. In the last 12 months, how many times have you been away from home for one or more nights (in other words, sleeping somewhere else for the night)? G6.08. In the last 12 months, have you been away from home for more than one month at a time? YES NO CODE G6 EVERYDAY EVERY WEEK AT LEAST ONCE EVERY 2 WEEKS AT LEAST ONCE EVERY MONTH AT LEAST ONCE LESS THAN ONCE A MONTH NEVER 195 | Page MODULE G7: INTRAHOUSEHOLD RELATIONSHIPS (Group 2 only) Now I’d like to ask you some questions about how you feel about some of other people in your household or family group and how you think they feel about you. ENTER MEMBER ID FOR EACH RELATION OTHER CODES: NON-HH MEMBER...….94 Do you [NAME] respect your [RELATION]? Does your [RELATION] respect you? Do you trust your [RELATION] to do things that are in your best interest? When you disagree with your [RELATION], do you feel comfortable telling him/her that you disagree? IS [RELATION] THE OTHER RESPONDENT WITHIN THIS HOUSEHOLD? Is there a co￾wife within your household? RELATION G7.02 G7.03 G7.04 G7.05 G7.06 G7.07 A Husband / wife ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER YES……1  RELATION C NO..……2 B Other adult male or female member of the household ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER 196 | Page C IF RESPONDENT IS MALE: Father (or adapt this category to capture other important relationship) IF RESPONDENT IS FEMALE: Mother-in-law ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER IF RESPONDENT IS MALE  MODULE G8(A) YES……1 NO..……2  MODULE G8(A) D Most senior co-wife (the person who was in the household just before you, or, if you are the senior wife, the one who married into the household after you) ID # MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MOST OF THE TIME SOMETIMES RARELY NEVER MODULE G8(A): AUTONOMY IN DECISION-MAKING (Group 2 only) Now I am going to read you some stories about different farmers and their situations regarding different agricultural activities. This question format is different from the rest so take your time in answering. For each I will then ask you how much you are like or not like each of these people. We would like to know if you are completely different from them, similar to them, or somewhere in between. There are no right or wrong answers to these questions. READ ALOUD EACH STORY, SUBSEQUENT QUESTIONs, AND RESPONSE CODES. NAMES SHOULD BE ADOPTED TO LOCAL CONTEXT AND BE MALE/FEMALE DEPENDING ON THE SEX OF THE RESPONDENT. THE ORDER OF TOPICS A-D SHOULD BE RANDOMIZED, AND WITHIN EACH TOPIC, THE ORDER OF STORIES 1-4 SHOULD BE RANDOMIZED. Are you like this person? CIRCLE ONE Are you completely the same or somewhat the same? CIRCLE ONE Are you completely different or somewhat different? CIRCLE ONE STORY G8.01 G8.02 G8.03 197 | Page The types of crops to grow or raise for consumption and sale in market A1 “[PERSON’S NAME] cannot grow other types of crops here for consumption and sale in market. Beans, sweet potato and maize are the only crops that grow here.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A2 SOMEWHAT THE SAME…....2  A2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A2 “[PERSON’S NAME] is a farmer and grows beans, sweet potato, and maize because her spouse, or another person or group in her community tells her she must grow these crops. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A3 SOMEWHAT THE SAME…....2  A3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A3 “[PERSON’S NAME] grows the crops for agricultural production that her family or community expect. She wants them to approve of her as a good farmer.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  A4 SOMEWHAT THE SAME…....2  A4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 A4 “[PERSON’S NAME] chooses the crops that she personally wants to grow for consumption and sale in market and thinks are best for herself and her family. She values growing these crops. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B1 SOMEWHAT THE SAME…....2  B1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 Livestock raising B1 “[PERSON’S NAME] cannot raise any livestock other than what she has. These are all that do well here.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B2 SOMEWHAT THE SAME…....2  B2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B2 “[PERSON’S NAME] raises the types of livestock she does because her spouse, or another person or group in her community tell her she must use these breeds. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B3 SOMEWHAT THE SAME…....2  B3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B3 “[PERSON’S NAME] raises the kinds of livestock that her family or community expect. She wants them to approve of her as a good livestock raiser.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  B4 SOMEWHAT THE SAME…....2  B4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 B4 “[PERSON’S NAME] chooses the types of livestock that she personally wants to raise and thinks are good for herself and her family. She values raising these types. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C1 SOMEWHAT THE SAME…....2  C1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 READ ALOUD EACH STORY, SUBSEQUENT QUESTIONs, AND RESPONSE CODES. NAMES SHOULD BE ADOPTED TO LOCAL CONTEXT AND BE MALE/FEMALE DEPENDING ON THE SEX OF THE RESPONDENT. Are you like this person? CIRCLE ONE Are you completely the same or somewhat the same? CIRCLE ONE Are you completely different or somewhat different? CIRCLE ONE STORY G8.01 G8.02 G8.03 198 | Page Taking crops or livestock (incl. eggs or milk) to the market (or not) C1 “There is no alternative to how much or how little of her crops or livestock [PERSON’S NAME] can take to the market. She is taking the only possible amount.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C2 SOMEWHAT THE SAME…....2  C2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C2 “[PERSON’S NAME] takes crops and livestock to the market because her spouse, or another person or group in her community tell her she must sell them there. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C3 SOMEWHAT THE SAME…....2  C3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C3 “[PERSON’S NAME] takes the crops and livestock to the market that her family or community expect. She wants them to approve of her.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  C4 SOMEWHAT THE SAME…....2  C4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 C4 “[PERSON’S NAME] chooses to take the crops and livestock to market that she personally wants to sell there, and thinks is best for herself and her family. She values this approach to sales. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D1 SOMEWHAT THE SAME…....2  D1 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 How to use income generated from agricultural and non-agricultural activities D1 “There is no alternative to how [PERSON’S NAME] uses her income. How she uses her income is determined by necessity.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D2 SOMEWHAT THE SAME…....2  D2 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D2 “[PERSON’S NAME] uses her income how her spouse, or another person or group in her community tell her she must use it there. She does what they tell her to do.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D3 SOMEWHAT THE SAME…....2  D3 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D3 “[PERSON’S NAME] uses her income in the way that her family or community expect. She wants them to approve of her.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME….1  D4 SOMEWHAT THE SAME…....2  D4 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 D4 “[PERSON’S NAME] chooses to use her income how she personally wants to, and thinks is best for herself and her family. She values using her income in this way. If she changed her mind, she could act differently.” YES...1 NO.....2  G8.03 COMPLETELY THE SAME...1G8.04 SOMEWHAT THE SAME….2 G8.04 COMPLETELY DIFFERENT....1 SOMEWHAT DIFFERENT.......2 199 | Page MODULE G8(B): NEW GENERAL SELF-EFFICACY SCALE (Group 2 only) Now I’m going to ask you some questions about different feelings you might have. Please listen to each of the following statements. Think about how each statement relates to your life, and then tell me how much you agree or disagree with the statement on a scale of 1 to 5, where 1 means you “strongly disagree” and 5 means you “strongly agree.” (Note: Randomize order of statements) STATEMENTS G8.04 A I will be able to achieve most of the goals that I have set for myself. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE B When facing difficult tasks, I am certain that I will accomplish them. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE C In general, I think that I can obtain outcomes that are important to me. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE D I believe I can succeed at most any endeavor to which I set my mind STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE E I will be able to successfully overcome many challenges. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE F I am confident that I can perform effectively on many different tasks. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE G Compared to other people, I can do most tasks very well. STRONGLY DISAGREE DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE H Even when things are tough, I can perform quite well. STRONGLY DISAGREE 200 | Page DISAGREE NEITHER AGREE NOR DISAGREE AGREE STRONGLY AGREE MODULE G8(C): LIFE SATISFACTION (Group 2 only) The following questions ask how satisfied you feel with your life as a whole, on a scale from 1 to 5, where 1 means you feel “very dissatisfied” and 5 means you feel “very satisfied.” STATEMENTS G8.05 A Overall, how satisfied are you with life as a whole these days? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED B Overall, how satisfied with your life were you 5 years ago? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED C As your best guess, overall how satisfied with your life do you expect to feel 5 years from today? VERY DISSATISFIED DISSATISFIED NEITHER SATISFIED NOR DISSATISFIED SATISFIED VERY SATISFIED 201 | Page MODULE G9. Attitudes about Domestic Violence (Group 2 only) Now I would like to ask about your opinion on the following issues. Please keep in mind that I am not asking about your personal experience or whether the following scenarios have happened to you. I would only like to know whether you think the following issues are acceptable. In your opinion, is a husband justified in hitting or beating his wife in the following situations? SITUATION G9.01 A If she goes out without telling him? YES NO DON’T KNOW B If she neglects the children? YES NO DON’T KNOW C If she argues with him? YES NO DON’T KNOW D If she refuses to have sex with him? YES NO DON’T KNOW E If she burns the food? YES NO DON’T KNOW END OF MALE FORM 202 | Page