The impacts of the microfinance multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda

Berendson, Ricardo Morel; Gassmann, Franziska; Martorano, Bruno; Tirivayi, Nyasha J.; Kamau, John Working Paper The impacts of the microfinance multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda UNU-MERIT Working Papers, No. 2024-014 Provided in Cooperation with: Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), United Nations University (UNU) Suggested Citation: Berendson, Ricardo Morel; Gassmann, Franziska; Martorano, Bruno; Tirivayi, Nyasha J.; Kamau, John (2024) : The impacts of the microfinance multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda, UNU-MERIT Working Papers, No. 2024-014, United Nations University (UNU), Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), Maastricht This Version is available at: https://hdl.handle.net/10419/326910 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/ #2024-014 The impacts of the Microfinance Multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda Ricardo Morel, Franziska Gassmann, Bruno Martorano, Nyasha Tirivayi and John Kamau Published 19 June 2024 Maastricht Economic and social Research institute on Innovation and Technology (UNU-MERIT) email: [email protected] | website: http://www.merit.unu.edu Boschstraat 24, 6211 AX Maastricht, The Netherlands Tel: (31) (43) 388 44 00 UNU-MERIT Working Papers ISSN 1871-9872 Maastricht Economic and social Research Institute on Innovation and Technology UNU-MERIT UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at UNU-MERIT to stimulate discussion on the issues raised. 1 The impacts of the Microfinance Multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda Ricardo Morel a, Franziska Gassmann b, Bruno Martorano b, Nyasha Tirivayi c, and John Kamau d June 2024 Abstract We study the impact of an innovative program that combines microfinance with farming extension services on food security outcomes in rural South-Western Uganda. For this purpose, we use experimental data and monthly panel data collected over two years to monitor seasonal changes. The results suggest that neither the combined approach of microfinance with farming extension services nor standalone microfinance demonstrated significant effectiveness in reducing food insecurity throughout seasons in the period of analysis. Households in both treatment groups experienced a reduction of dietary diversity mainly during land preparation approximately two years after the start of the interventions. Heterogeneous analysis revealed that households receiving MFM services and having better access to markets experienced occasional improvements in food security. Finally, households with higher food poverty levels in the MFM group experienced some improvements in food security, while those in the Microfinance group encountered sporadic negative outcomes in terms of dietary diversity. JEL codes: G21, I3, Q16, Q18 Keywords: dietary diversity, food security, microfinance, agriculture a Innovations for Poverty Action (IPA). Contact: [email protected] b UNU-MERIT – Maastricht University c UNICEF Innocenti-Global Office of Research and Foresight d Low-Income Financial Transformation (L-IFT) 2 1. Introduction Rural populations in Sub-Saharan Africa face significant challenges due to weather and economic shocks associated with seasonal agrarian cycles, leading to variations in income, food access, and consumption. Seasonal poverty, particularly during the lean season, adversely affects food security, dietary intake, and nutritional status (Hoddinott and Yohannes 2002; Sen and Drèze 1989; Janvry et al. 2016). Addressing these issues is a critical policy priority for national governments and the international community, reflected in initiatives like the Sustainable Development Goals and the 'zero hunger' agenda (FAO et al. 2020; FAO 1996; United Nations 2017). Achieving food security requires consistent access to sufficient, safe, and nutritious food that meets dietary requirements and preferences while considering the challenges of seasonality (FAO 2002). Using an experimental design, this study aims to measure the impact of an innovative approach that integrates Microfinance products with Agriculture, Poultry, and Livestock (APL) extension services on food security across farming seasons. This bundled approach, known as Microfinance Multiplied (MFM), posits that integrating programs addressing complementary outcomes can yield greater impacts than implementing them separately. MFM targets poor rural households that are grappling with seasonality constraints and can enhance household productivity through different mechanisms. Microfinance can improve seasonal income during the lean season, thereby increasing households’ liquidity for investments and (food) spending as well as savings capacity, which helps to smoothen consumption over these seasonal shocks (Khandker et al.2015; 2010). The APL program is directly linked to farming activities, and it offers agriculture, poultry, and livestock extension services, including training on modern farming methods in combination with distributing productive technologies, such as improved seeds or vaccines for chickens and livestock. Extension services can help boost the productivity and diversity of crops and animal products, and the surplus can be sold in the markets or saved for consumption in the lean season (Hawkes and Ruel 2006). While their mechanisms are different, both programs have the potential to complement each other to ensure food security. Cash liquidity can be used immediately to smooth consumption while learning modern farming practices and using improved productive inputs can increase productivity in the longer term. Additionally, credit can be injected into farming activities to boost productivity and commercialize products, which can lead to higher income, savings, and food to better cope with seasonal variations. Despite the potential complementarity of these programs in ensuring food security, there is limited evidence regarding their synergies in smoothing consumption and mitigating the negative impacts of seasonal shocks. To fill this evidence gap, this study employs a factorial randomised design that compares the impact of the bundled MFM approach and standalone Microfinance to a control group. The aim is to evaluate how these interventions protect households from seasonal variations in food security. The research assesses trends in food security and dietary diversity using a rich panel dataset collected monthly over 24 months in rural South-Western Uganda, predominantly comprising smallholder farmers engaged in rain-fed agriculture. 3 Results suggest that neither the bundled MFM approach nor standalone microfinance showed significant effectiveness in reducing food insecurity across seasons in the short term. However, there were noticeable negative impacts on dietary diversity during the harvesting and land preparation seasons at the end of the study period. The lack of significant effects on food insecurity outcomes and the negative effects on dietary diversity could be due to institutional and socioeconomic factors. The analysis shows that households that resided closer to markets and received MFM services experienced sporadic improvements in food security during harvesting, while dietary diversity remained unchanged. Households with higher food poverty levels in the MFM group experienced some improvements in terms of food security score in the first month and during land preparation phases in the second year, while those in the Microfinance group experienced a decrease in the dietary diversity score during one month of the second harvesting of year one. This study addresses a gap in the literature by comprehensively examining the impact of bundling microfinance and extension services on food security across agricultural seasons. The existing evidence on the effectiveness of microfinance and farming programs in addressing seasonal food insecurity and dietary diversity is inconclusive, and there is a lack of research on the combined effects of these interventions. This paper makes a valuable contribution to the study of seasonal food security by conducting an in-depth analysis based on a comprehensive panel survey covering four planting and harvesting seasons. Unlike many impact evaluations in the field of development economics that rely on limited data points – usually, two or occasionally three if there is a midline survey – this unique panel of monthly survey rounds over 24 months allows for a thorough exploration of the potential benefits of combining microfinance and technology adoption programs in reducing seasonal food insecurity. Testing the disaggregated impacts of a bundled approach is another contribution of this paper. The factorial randomised design allows for differentiating effects of microfinance services and an integrated multiplied approach that adds on agriculture and livestock extension services. The findings highlight the importance of enhancing program design to incorporate food security and nutrition components at its core, strengthening and adapting microfinance services to rural populations and seasonality, and considering local market access and dynamics. 2. Literature review Seasonality – understood as intra-annual, seasonal fluctuations in the agrarian cycle – is inherent to smallholder farming households and common in rural Sub-Saharan Africa. This dynamic means that income and consumption are prone to fluctuate depending on the season. In the growing season, when most of the food is produced, farm labour is in demand, food prices are cheaper, and, in general, more income is generated. This period is followed by the lean season when food stocks are depleted, labour is scarce, food prices are higher, and income tends to be lower. This seasonal variation, therefore, raises the opportunity cost of prices, labour, wages, and even migration patterns (Dercon and Krishnan 2007; Devereux, et al. 2011; Kaminski, et al. 2014; Stephens and Barrett 2011; Basu and Wong 2012; Bryan, et al. 2014). There is a large body of evidence that indicates that seasonal variation in household consumption is mainly due 4 to fluctuations in income, prices, labour, unreliable markets, and access to credit (Rosenzweig and Binswanger 1993; Chaudhuri and Paxson 2002; Paxson 1993). One of the most critical risks of this seasonal dynamic is that households that do not have the necessary coping mechanisms – for example, storing food, savings, assets, credit – will consume more calories and nutrients during the growing season but be unable to meet their an adequate dietary intake during the lean season, negatively affecting their food security and nutritional status, a key dimension of human wellbeing (Hoddinott and Yohannes 2002; Dercon and Krishnan 2007; Christian and Dillon 2016; Janvry et al. 2016). Achieving effective consumption smoothing is challenging for many poor rural households due to limited resources and external factors such as price fluctuations, labour demand, and health and climate shocks. As a result, rural households use different risk mitigation mechanisms to smoothen consumption and ensure food security, complementing proactive strategies (before seasonal shocks occur) with reactive responses (after seasonal shocks occur). For example, households can store food, use savings, accumulate assets, borrow money, or engage in off-farm labour (Morduch 1995; Fink, Jack, and Masiye 2020; Chambers, et al. 1981; Dercon and Krishnan 2007; Collins et al. 2009). However, degrees of success vary because of the uncertainty of several external factors, such as food price fluctuation, availability of off-farm work, and even the incidence of health and climate shocks, duration of the lean season, among others (Christian and Dillon 2016). The literature also emphasizes that the poorest households are more vulnerable to seasonal deprivation as they cannot afford most of these mechanisms, such as livestock and grain storage, significant savings, or access to flexible credit. As a result, their methods for smoothening consumption are inadequate and inefficient (Khandker, et al. 2015). Even if expected, seasonal variations present constant challenges to poor rural households as they have to constantly save and borrow money to have enough food to eat during the year (Collins et al. 2009). Certain interventions such as microfinance and farming extension services may have the potential to overcome these constraints and empower households to manage risks and shocks, enabling them to maintain food security and stabilize their consumption. The existent evidence related to these interventions is further discussed below. Microfinance. Access to credit provides liquidity and can be a tool for consumption smoothening during seasonal shocks (Collins et al. 2009; Morduch 1995; Zeller and Sharma 2000; Angelucci, et al. 2015). Microfinance can also help promote off-farm income-generating activities (Bandiera, et al. 2022; Banerjee, et al. 2015). These activities are less volatile during seasonal shocks and, in principle, provide households with a steadier income during the lean season, allowing more spending on food and helping to smooth consumption during critical moments (Pitt and Khandker 2002; Zaman 1999; 2004; Khandker 2012). However, the evidence base does not convincingly support this claim. Several experimental studies in different contexts show null effects on income, savings, or consumption (Meager 2019; Bandiera, et al. 2022; Banerjee, et al. 2015; Dahal and Fiala 2020). 1 Scholars also point out that standard microfinance 1 The analysis by Banerjee, Karlan, and Zinman covered Bosnia (Augsburg et al. 2015), Ethiopia (Tarozzi, Desai, and Johnson 2015), India (Banerjee, Karlan, and Zinman 2015), Mexico (Angelucci, Karlan, and Zinman 2015), Morocco (Crépon et al. 2015), and Mongolia (Attanasio et al. 2015) between 2003 and 2012. Meager’s research (Meager 2019) included the same studies and added an RCT by Karlan and Zinman (2011) in the Philippines. Dahal and Fiala (2020), finally, added a study by Fiala (2018) in Uganda. 5 schemes, requiring regular repayments regardless of the income seasonality, can be detrimental to borrowers, for example, by leading to over-indebtedness (Morduch 1999; Schicks 2014; Rahman 1999). Microfinance institutions have started to tailor their products targeted to farmers to the seasonal variations of the agrarian cycle and the evidence, in general, shows that flexible products have more potential than the standard schemes (Field et al. 2013). Evidence of the impact of flexible credit products in Bangladesh and Zambia finds improvements in food security (Khandker, Khalily, and Samad 2015; Fink, et al. 2020), especially among the poorest. Loans are used as a mechanism for consumption smoothing during the lean season and then repaid in small instalments over the year when income and food access increase again (Berg and Emran 2020). Farming extension services. Rural households in low-income settings face various challenges such as limited resources, high transaction costs, and market 'failures' like the lack of established credit and saving mechanisms. According to the agricultural household model theory, these households make decisions about their production and consumption that are interdependent, and non-separable from their resource allocation decisions (Janvry and Sadoulet 2006; Conceição et al. 2016; Jack 2013; Karlan et al. 2014; Morris et al. 2007; Muzari, et al. 2012; Udry 2010; Janvry et al. 2016). To address this challenge, farming extension services offer training on modern farming techniques and access to highly productive technologies, such as improved seeds or animal vaccination, to diversify cropping systems and boost the productivity of smallholder farmers, developing, at the same time, their capacity to respond to shocks and risks by smoothing consumption and diversifying diets potential to fight food insecurity (Hawkes and Ruel 2006; Barrett, Reardon, and Webb 2001; Loison 2015). However, the empirical evidence is inconclusive and there is still an important gap in the effects of seasonality. A meta-analysis on the association between the diversification of crop production and diet diversity shows that, while findings tend to be generally positive, few studies report consistent results across indicators (Sibhatu and Qaim 2018). One of the challenges highlighted is the high numbers of additional crops or animal species required for production by smallholder farmers in order to achieve the desired levels of dietary diversity (Sibhatu and Qaim 2018). A systematic review of the effectiveness of agricultural programs on child nutrition in developing countries shows overall positive impacts on productivity and food consumption, however, these effects did not lead to a substantive improvement in the diet of poor households (Masset et al. 2012). However, a recent evaluation in Uganda shows that crop diversification increased household diet diversity and consumption (Tesfaye and Tirivayi 2020). Additionally, there is still little causal evidence of the link between extension services and food security. Researchers found that the adoption of technologies that require low upfront investments achieved food security in Uganda (Pan, et al. 2018). In Tanzania, researchers found that farmer field schools encouraged farmers to adopt technologies that helped smooth production over seasons (i.e., produced food across all seasons) and reduced hunger (Larsen and Bie Lilleør 2014). However, in Cameroon, an evaluation found that extension services did not improve the household dietary diversity score (Ngomi et al. 2020). 6 3. Uganda and the BRAC program Description of the context. Uganda is particularly vulnerable to food security as agriculture is mainly rain-fed and thus, susceptible to seasonal and weather shocks (Tesfaye and Tirivayi 2020). Uganda ranks 95th out of 113 countries in the Global Food Security Index and, according to the Ugandan Bureau of Statistics, 37% of households are food-poor whereas rural households were as double as likely to be under this category (The Economist 2022). Agriculture accounts for the largest share of employment with nearly 40% of the working population, of which 43% reported that subsistence farming was their main source of income. However, the adoption of productive farming inputs and technologies remains low (Barua 2011; World Bank 2016; UBOS 2018). Access to credit was rapidly increasing at the time of the study, reaching approximately a quarter of households, but most of them in urban settings. The largest proportion of credit sources were still informal with exorbitant interest rates, particularly in rural areas (Chaia et al. 2009; FSD Uganda 2018; UBOS 2014; 2010; 2018). The location of the study is in South-Western Uganda, in the districts of Kabale and Rukungiri (Figure 1, Panel A). This setting has a bimodal seasonal calendar, with two rainy seasons (usually between March and June and mid-August to December) and dry seasons in between (WFP 2013). This seasonality means that there are also two main agricultural seasons every year; the first is typically between January to July, and the second is from August to December. Each season includes periods for land preparation, crop cultivation, weeding, and harvesting. The first agricultural season of the year tends to be longer and, therefore, the peak lean season usually takes place around the middle of this period, that is from April to May. Both Kabale and Rukungiri districts have agriculture as a main source of food and income, but they differ in the main crops they produce. In Kabale, the main crops are potatoes, sorghum, and beans, while in Rukungiri are bananas, beans, and cassava. Most households in both districts manage their food and cash needs during the planting and weeding period through a combination of strategies. They utilize their carry-over stocks from previous seasons, engage in livestock rearing, sell their labour either locally or by migrating to lowland areas, or seek support from relatives (Browne and Glaeser 2010). From 2011 to 2013, the region experienced generally positive patterns of agriculture and livestock production. During this time, bimodal areas in Uganda, including the South-West, overall experienced favourable rainfall, and above-average crop production, leading to replenished food stocks, stabilized prices, and no or minimal food insecurity. In 2014, the bimodal regions saw short dry spells and poor rainfall distribution across both rainy seasons, however, the overall harvests were close to average, and, in general, households were able to meet their food needs through market purchase and seasonal activities, such as crops sales, casual labour in harvest and post-handling activities, and petty trade. Throughout the period of this study, bimodal areas remained in Phase 1 of the Integrated Phase Classification (IPC), 2 indicating minimal acute food insecurity (FEWS NET 2023). 2 The IPC system was developed by a consortium of international organizations including FAO, WFP, UNICEF, and WHO, among others. 13 Figure 3. Household Dietary Diversity Score (HDDS) – OLS panel analysis by month Note: Labels next to the x-axis in each figure refer to the month number and the season. "Lean" refers to the lean season; "harv" stands for harvesting; "prep" refers to land preparation. Robustness check. Results on the main outcomes of interest are further explored using a difference-in-difference (DID) estimation. The DID model assesses effects of the treatment groups over time compared to the control group. The model compares the differences of month 1 with each of the subsequent months. In other words, the regression is repeated for the interaction on month 1 with each following month, starting from month 2 until month 24. This approach allows identify whether and how the treatments have an impact that varies across different time periods relative to the first month of the panel. The DID analysis yielded similar results to the OLS regression analysis in terms of the lack of significant impacts on the HFIAS score (Table A5). Regarding dietary diversity, the only impact observed in the DID analysis was a higher score during the first land preparation period, just before the lean season. This finding contrasts with the results of the OLS analysis, where negative significant effects were observed on adequate dietary diversity in the last three months of the panel, during the last land preparation season in the study period. While some differences were observed between the DID analysis and the main OLS analysis by panel month, the overall trends were not substantially different. These 14 findings provide additional robustness to the main results and support the conclusion that the interventions did not have significant impacts on food security and dietary diversity. 5.2 Intermediary Outcomes Exploration of the of impact on intermediary outcomes – namely borrowing, savings, and labour – do not show clear trends of significant differences between the treatment groups and the control (Table A6 to Table A10). The study findings show no significant changes in formal borrowing and savings over seasons, in line with the findings of multiple experimental studies in diverse contexts (Banerjee, et al. 2015; Meager 2019; Dahal and Fiala 2020; Bandiera, et al. 2022).The only exception are households in the MFM group who borrowed less, on average, than those in the control group during one month of the last harvest season (month 22), which coincides with the significant decrease in dietary diversity. As discussed in the interpretation of food security outcomes, structural and programmatic barriers, such as a standard repayment model that offered no flexibility to rural households may have reduced households' ability to engage (Morduch 1999; Schicks 2014; Rahman 1999). Additionally, there were no sustained changes in labour participation, but those in the microfinance group worked more on someone else’s land during the lean weeding season (month 13) and increased entrepreneurial engagement during the first land preparation season (months 5 and 6). 6. Heterogenous analysis This section focuses into heterogenous treatment effects by examining the influence of distance to the nearest market and wealth. By considering these factors, the study aims to uncover nuanced insights into the varying impacts of the interventions on food security and dietary diversity among different subgroups within the study population. 6.1. Distance to the nearest market Proximity to markets can shape opportunities and constraints for households as it provides opportunities to access productive inputs, financial products, commercial information as well as more and diverse foods. Households closer to markets may also be better positioned to access off-farm income sources, reducing agriculture dependence. Conversely, farther households face elevated transaction costs, diminishing market these opportunities (Suri 2011; Abay and Jensen 2020; Madzorera et al. 2021; Kuss, Gassmann, and Mugumya 2022). Therefore, analysing the impact of interventions such as microfinance and agriculture extension programs by distance to markets can provide insights into whether different households experience varying impacts, which is crucial for designing more targeted and effective development policies. The distance was calculated by asking respondents in the baseline survey for the time it took them to walk to the nearest trading centre or market. Using the median time at the community level, these were then divided into two groups based on their distance to the market, with the median distance serving as the cut-off point. We create a dummy variable with a value of one for households closer to markets (i.e. below the median reported distance) and interact this variable with the treatment variables to estimate the differential impact of the programs due to the distance to the nearest market. 15 Results show that households closer to markets experience a significant decrease in the household food insecurity score in the MFM group during the first harvesting season of the second year (month 15) (Figure 4, Table A11). By contrast, the analysis of the dietary diversity score does not show significant changes based on the distance to markets (Figure 5, Table A12). Hence, the observed improvements in food security outcomes among households closer to markets in the MFM group seem to be sporadic, only appearing in specific months but hardly creating a clear trend. Figure 4. Household Food Insecurity Access Scale (HFIAS) score by distance to market (Closer to the market) Note: Labels next to the x-axis in each figure refer to the month number and the season. "Lean" refers to the lean season; "harv" stands for harvesting; "prep" refers to land preparation. 16 Figure 5. Household Dietary Diversity Score (HDDS) by distance to market (Closer to the market) Note: Labels next to the x-axis in each figure refer to the month number and the season. "Lean" refers to the lean season; "harv" stands for harvesting; "prep" refers to land preparation. 6.2 Food poverty Food poverty is measured by a dummy variable based on the number of types of food consumed at baseline. Households who consume less than the average number of food types, indicating higher food poverty, get a value of one. We interact this variable with the treatment variables to estimate the differential impact of the programs due to their level of food poverty. The analysis reveals that food-poor households in the MFM group experience a significant decrease in the food insecurity score in the first month and during the land preparation phase in the second year (months 17 and 23), (Figure 6, Table A13). However, significant changes are not consistent over time and, therefore, levels of food consumption do not seem to have substantial and sustained impacts in food security outcomes. 17 Figure 6. Household Food Insecurity Access Scale (HFIAS) score by food consumption (Food poverty) Notes: Labels next to the x-axis in each figure refer to the month number and the season. "Lean" refers to the lean season; "harv" stands for harvesting; "prep" refers to land preparation. Results also indicate a decrease in the dietary diversity score among food-poor households in the Microfinance group during the second harvesting of year one (month 10) (Figure 7, Table A14). The observed decreases in dietary diversity among food-poor households are scattered and inconclusive. A possible interpretation of these changes, however, could be attributed to the fact that these households faced limitations in accessing a variety of foods due to their financial constraints. Despite the provision of microfinance, the available financial resources may still be insufficient to support diverse food purchases. Moreover, asset-poor households may have limited access to markets and may be constrained by higher food prices and limited availability of diverse food items. It is important to note that while microfinance interventions may improve access to financial resources, additional focused support activities on enhancing food access and nutrition education may be necessary to address the specific challenges faced by asset-poor households in improving their dietary diversity. 18 Figure 7. Household Dietary Diversity Score (HDDS) by food consumption (Food poverty) Note: Labels next to the x-axis in each figure refer to the month number and the season. "Lean" refers to the lean season; "harv" stands for harvesting; "prep" refers to land preparation. 7. Conclusions The study implements a randomised evaluation to examine the impacts of Microfinance and an integrated approach that combines Microfinance with Agriculture, Poultry, and Livestock extension programs on food security outcomes and dietary diversity over farming seasons. The results indicate that the interventions did not have significant effects on food security outcomes, as measured by the Household Food Insecurity Access Scale (HFIAS). Moreover, the analysis on dietary diversity indicates that households in both the MFM approach and the standalone Microfinance program experienced a significantly lower Households Dietary Diversity Score (HDDS) during the last harvesting season and subsequent land preparation months in the study period. Heterogeneous analysis employing distance to markets and levels of food poverty produced interesting results. The analysis reveals that households closer to markets in the MFM group experienced positive food security outcomes, particularly during the first harvesting season of the second year. Moreover, households with higher food poverty levels in the MFM group experienced some improvements in terms of 19 food security in the first month and during land preparation phases in the second year; however, households with higher food poverty in the Microfinance group experienced sporadic negative outcomes in terms of dietary diversity. The findings of this study provide valuable insights for policymakers and practitioners aiming to enhance the effectiveness of financial inclusion interventions in addressing food security and dietary diversity. One key takeaway is the importance to enhance the design of financial inclusion programs by building in components that explicitly target food security and nutrition outcomes. It is not enough to simply bundle it with another program – in this case, APL services – if it does not integrate mechanisms that directly enhance the intake of diverse and nutritious foods into the program design. In other words, the recommendation is to take a comprehensive stand by focusing not only on income generation and productivity but also on engraining other dimensions in the program design, such as food-related aspects that can more effectively contribute to improving food security and nutritional well-being among vulnerable populations. Additionally, it is crucial to tailor microfinance products and services to the specific needs of households, particularly for rural households with volatile income due to agricultural seasons (e.g., including flexible repayment terms) while taking into consideration the dynamics of local markets. More granular findings also suggest that proximity and access to markets can play a crucial role in mitigating food insecurity. Moreover, targeting poorer sections of the population is relevant to tackle food security-related outcomes. These recommendations can contribute to the improvement of financial inclusion interventions in addressing food security and dietary diversity. By enhancing program design, strengthening microfinance services, considering local market dynamics, effectively targeting less-wealthy populations, and prioritizing long-term measurement, policymakers and practitioners can refine interventions and promote sustainable improvements in food security and nutritional well-being among vulnerable populations. 20 References Abay, Kibrom A., and Nathaniel D. Jensen. 2020. “Access to Markets, Weather Risk, and Livestock Production Decisions: Evidence from Ethiopia.” Agricultural Economics 51 (4): 577–93. https://doi.org/10.1111/agec.12573. Angelucci, Manuela, Dean Karlan, and Jonathan Zinman. 2015. “Microcredit Impacts: Evidence from a Randomized Microcredit Program Placement Experiment by Compartamos Banco.” American Economic Journal: Applied Economics 7 (1): 151–82. https://doi.org/10.1257/app.20130537. Attanasio, Orazio, Britta Augsburg, Ralph De Haas, Emla Fitzsimons, and Heike Harmgart. 2015. “The Impacts of Microfinance: Evidence from Joint-Liability Lending in Mongolia.” American Economic Journal: Applied Economics 7 (1): 90–122. https://doi.org/10.1257/app.20130489. Augsburg, Britta, Ralph De Haas, Heike Harmgart, and Costas Meghir. 2015. “The Impacts of Microcredit: Evidence from Bosnia and Herzegovina.” American Economic Journal: Applied Economics 7 (1): 183–203. https://doi.org/10.1257/app.20130272. Ballard, Terri, Jennifer Coates, Anne Swindale, and Megan Deitchler. 2011. “Household Hunger Scale: Indicator Definition and Measurement Guide.” Bandiera, Oriana, Robin Burgess, Erika Deserranno, Ricardo Morel, Imran Rasul, Munshi Sulaiman, and Jack Thiemel. 2022. “Microfinance and Diversification.” Economica 1. https://doi.org/10.1111/ECCA.12424. Bandiera, Oriana, Robin Burgess, Erika Deserranno, Ricardo Morel, Munshi Sulaiman, and Imran Rasul. 2022. “Social Incentives, Delivery Agents and the Effectiveness of Development Interventions.” Journal of Political Economy Microeconomics 1 (1): 162–224. https://doi.org/10.1086/722898. Banerjee, Abhijit, Dean Karlan, and Jonathan Zinman. 2015. “Six Randomized Evaluations of Microcredit: Introduction and Further Steps.” American Economic Journal: Applied Economics 7 (1): 1–21. Barrett, C B, T Reardon, and P Webb. 2001. “Nonfarm Income Diversification and Household Livelihood Strategies in Rural Africa: Concepts, Dynamics, and Policy Implications.” Food Policy 26: 315–31. Barua, Proloy. 2011. “Assessment of the Short-Run Impact of BRAC’s Agriculture and Livestock Programme in Uganda.” BRAC Working Paper Series 1. Barua, Proloy, and Munshi Sulaiman. 2006. “Targeting the Poorest in Microfinance: Poverty Outreach of BDP Ultra Poor Programme.” CFPR/TUP Working Paper Series 13. 21 Basu, Karna, and Maisy Wong. 2012. “Evaluating Seasonal Food Security Programs in East Indonesia.” Munich Personal RePEc Archive Paper 51219. Beaman, Lori, Dean Karlan, Bram Thuysbaert, and Christopher Udry. 2014. “Self-Selection into Credit Markets: Evidence from Agriculture in Mali.” NBER Working Paper Series 20387: 1–37. https://doi.org/10.3386/w20387. Berg, Claudia, and M. Shahe Emran. 2020. “Microfinance and Vulnerability to Seasonal Famine in a Rural Economy: Evidence from Monga in Bangladesh.” Journal of Economic Analysis and Policy 20 (3). https://doi.org/10.1515/bejeap-2018-0359. Browne, Stephen, and Laura Glaeser. 2010. “Livelihood Mapping and Zoning Exercise: Uganda.” Famine Early Warning System Network. Bryan, Gharad, Shyamal Chowdhury, and Ahmed Mushfiq Mobarak. 2014. “Underinvestment in a Profitable Technology: The Case of Seasonal Migration in Bangladesh.” Econometrica 82 (5): 1671–1748. https://doi.org/10.3982/ECTA10489. Chaia, Alberto, Aparna Dalal, Tony Goland, Maria Jose Gonzalez, Jonathan Morduch, and Robert Schiff. 2009. “Half the World Is Unbanked.” Financial Access Initiative Framing Note October. Chambers, Robert, Richard Longhurst, and Arnold Pacey. 1981. Seasonal Dimensions to Rural Poverty. London: Frances Pinter Ltd. Chaudhuri, Subham, and Christina Paxson. 2002. “Smoothing Consumption under Income Seasonality :” Columbia Department of Economics Discussion Paper Series 0102 (54). Christian, Paul, and Brian Dillon. 2016. “Long Term Consequences of Consumption Seasonality.” African Development Bank Working Series 241. Coates, Jennifer, Anne Swindale, and Paula Bilinsky. 2007. “Household Food Insecurity Access Scale (HFIAS) for Measurement of Food Access: Indicator Guide.” Washington, D.C.: Food and Nutrition Technical Assistance Project. Collins, Daryl, Jonathan Morduch, Stuart Rutherford, and Orlanda Ruthven. 2009. Portfolios of the Poor: How the World’s Poor Live on $2 a Day. Princeton: Princeton University Press. Conceição, Pedro, Sebastian Levine, Michael Lipton, and Alex Warren-Rodríguez. 2016. “Toward a Food Secure Future: Ensuring Food Security for Sustainable Human Development in Sub-Saharan Africa.” Food Policy 60: 1– 9. https://doi.org/10.1016/j.foodpol.2016.02.003. 22 Coppel, Emily, and Munshi Sulaiman. 2015. “Key Findings from Randomized Evaluation of Microcredit and Agriculture Services in Uganda.” BRAC Research Review 2. Crépon, Bruno, Florencia Devoto, Esther Duflo, and William Parienté. 2015. “Estimating the Impact of Microcredit on Those Who Take It up: Evidence from a Randomized Experiment in Morocco.” American Economic Journal: Applied Economics 7 (1): 123–50. https://doi.org/10.1257/app.20130535. Dahal, Mahesh, and Nathan Fiala. 2020. “What Do We Know about the Impact of Microfinance? The Problems of Statistical Power and Precision.” World Development 128. https://doi.org/doi.org/10.1016/j.worlddev.2019.104773. Dercon, Stefan, and Pramila Krishnan. 2007. “Vulnerability, Seasonality and Poverty in Ethiopia.” The Journal of Development Studies 36 (6): 25–53. https://doi.org/10.1080/00220380008422653. Devereux, Stephen, Rachel Sabates-Wheeler, and Richard Longhurst. 2011. Seasonality, Rural Livelihoods, and Development. London: Routledge. Drescher, Larissa S, Silke Thiele, and Gert B M Mensink. 2007. “A New Index to Measure Healthy Food Diversity Better Reflects a Healthy Diet Than Traditional Measures.” The Journal of Nutrition 137: 647–51. FAO. 1996. “Rome Declaration on World Food Security.” World Food Summit. 1996. http://www.fao.org/3/W3613E/W3613E00.htm. ———. 2002. “Report of the World Food Summit: Five Years Later.” 2002. https://www.fao.org/3/Y4172M/y4172m01e.htm. FAO, IFAD, UNICEF, WFP, and WHO. 2020. The State of Food Security and Nutrition in the World 2020. Transforming Food Systems for Affordable Healthy Diets. Rome: FAO. https://doi.org/10.4060/ca9692en. FEWS NET. 2023. “Uganda - Food Security Outlook.” Famine Early Warning Systems Network. February 13, 2023. https://fews.net/east-africa/uganda. Fiala, Nathan. 2018. “Returns to Microcredit, Cash Grants and Training for Male and Female Microentrepreneurs in Uganda.” World Development 105: 189–200. https://doi.org/10.1016/j.worlddev.2017.12.027. Field, Erica, Rohini Pande, John Papp, and Natalia Rigol. 2013. “Does the Classic Microfinance Model Discourage Entrepreneurship Among the Poor? Experimental Evidence from India.” American Economic Review 103 (6): 2196– 2226. https://doi.org/10.1257/aer.103.6.2196. Fink, Günther, B Kelsey Jack, and Felix Masiye. 2020. “Seasonal Liquidity, Rural Labor Markets, and Agricultural Production †.” American Economic Review 110 (11): 3351–92. https://doi.org/10.1257/aer.20180607. 29 Table A2. Balance check at baseline (1) (2) (3) Variables All treatments MFM Microfinance Age of respondent 0.00182 -0.00109 0.00291* (0.00133) (0.00122) (0.00128) Marital status -0.0149 -0.00120 -0.0137 (0.0144) (0.0126) (0.0143) Primary education completed -0.0285 -0.000315 -0.0281 (0.0519) (0.0525) (0.0512) Borrowed from formal sources in the last year -0.00583 0.0211 -0.0270 (0.0372) (0.0403) (0.0359) Engaged in agriculture or livestock in the last year -0.0506 0.0118 -0.0624 (0.0740) (0.0780) (0.0811) Number of sleeping rooms in the house -0.0161 0.00786 -0.0240 (0.0194) (0.0189) (0.0205) Average number of meals per day 0.0220 0.0190 0.00299 (0.0288) (0.0281) (0.0310) Kept cash savings in the last year -0.0985 -0.0487 -0.0498 (0.0725) (0.0665) (0.0709) Worked in someone else’s farm (wages) 0.0765 0.0558 0.0207 (0.0799) (0.0799) (0.0868) Worked off-farm (self-employed) 0.101* -0.00536 0.107 (0.0506) (0.0468) (0.0578) Worked off-farm (wages) -0.0239 -0.0540 0.0301 (0.0890) (0.0755) (0.0924) Prevalence of household food insecurity access (based on HFIAS) -0.0835 0.0391 -0.123 (0.0680) (0.0592) (0.0681) Insufficient food access (based on HFIAS) 0.0146 0.0474 -0.0328 (0.0549) (0.0562) (0.0512) Inadequate dietary diversity (based on HDDS) -0.00149 0.0209 -0.0224 (0.0836) (0.0923) (0.0829) Constant 0.779** 0.237* 0.542** (0.152) (0.118) (0.155) Observations 792 792 792 R-squared 0.007 0.013 0.019 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 30 Table A3. HFIAS: Household Food Insecurity Scale Score (0-27) – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding 0.4766 [0.5730] 0.2364 [0.6072] 774 0.112 7.516 7.794 7.607 month 2 harvest 0.2834 [0.5583] 0.3688 [0.6000] 771 0.104 7.115 7.408 7.527 month 3 harvest -0.1426 [0.5338] 0.5384 [0.5800] 754 0.112 6.107 5.885 6.688 month 4 preparation -0.1140 [0.5890] 0.7379 [0.6152] 738 0.099 6.126 5.906 6.917 month 5 preparation -0.0023 [0.5875] 0.6106 [0.5982] 715 0.095 6.228 6.192 6.793 month 6 preparation 0.3827 [0.6012] 0.1288 [0.5759] 706 0.094 6.251 6.597 6.469 month 7 lean 0.4740 [0.5210] 0.4393 [0.5367] 712 0.079 6.093 6.529 6.645 month 8 lean 0.4484 [0.5396] 0.4685 [0.5555] 751 0.119 5.834 6.169 6.323 month 9 harvest 0.7185 [0.5438] 0.8833 [0.5515] 773 0.152 5.099 5.684 5.905 month 10 harvest 0.3333 [0.5493] 0.5894 [0.5890] 764 0.178 5.364 5.601 5.916 month 11 preparation 0.2936 [0.5388] 0.7895 [0.6009] 768 0.191 5.046 5.178 5.714 month 12 preparation 0.5538 [0.5083] 0.9846 [0.5545] 767 0.199 4.850 5.167 5.644 month 13 lean 0.0784 [0.5568] 0.1677 [0.5528] 627 0.147 5.797 5.797 6.057 month 14 harvest 0.2996 [0.5154] 0.5392 [0.5215] 772 0.113 5.267 5.526 5.763 month 15 harvest 0.0793 [0.4902] 0.3116 [0.4835] 778 0.103 4.830 4.926 5.162 month 16 preparation -0.0900 [0.5374] 0.2540 [0.5328] 778 0.112 5.003 4.912 5.234 month 17 preparation -0.1775 [0.5662] 0.0941 [0.5669] 768 0.146 5.245 5.072 5.387 month 18 preparation -0.5490 [0.5231] 0.1829 [0.5556] 776 0.117 5.475 4.898 5.579 month 19 lean 0.1713 [0.5495] 0.2697 [0.5319] 759 0.141 5.777 5.813 5.943 month 20 lean -0.0993 [0.5100] -0.2906 [0.5113] 782 0.105 5.372 5.327 5.076 month 21 harvest 0.4086 [0.5374] -0.3767 [0.5311] 780 0.126 5.303 5.728 4.917 month 22 harvest 0.2058 [0.5944] -0.5532 [0.5633] 780 0.097 5.339 5.492 4.649 month 23 preparation 0.5110 [0.5970] -0.4276 [0.5507] 772 0.139 5.589 5.988 5.023 month 24 preparation 0.5752 [0.6466] -0.2698 [0.5894] 687 0.174 5.353 5.511 4.767 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 31 Table A4. HDDS: Household Dietary Diversity Score (0-12) – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding -0.1916 [0.2358] -0.2434 [0.2231] 791 0.068 5.782 5.651 5.545 month 2 harvest 0.1072 [0.2096] 0.1914 [0.2126] 788 0.101 6.301 6.463 6.466 month 3 harvest -0.2044 [0.2681] -0.0251 [0.2436] 789 0.100 6.752 6.665 6.805 month 4 preparation -0.1692 [0.2716] -0.2545 [0.2287] 782 0.101 6.418 6.395 6.295 month 5 preparation -0.1961 [0.2874] 0.1164 [0.2397] 774 0.119 6.436 6.293 6.560 month 6 preparation -0.2019 [0.3254] 0.3741 [0.2675] 779 0.122 6.372 6.260 6.839 month 7 lean -0.1442 [0.2651] 0.0851 [0.2210] 765 0.085 6.242 6.207 6.392 month 8 lean -0.2558 [0.2532] -0.0327 [0.2253] 775 0.093 6.493 6.341 6.530 month 9 harvest -0.3817 [0.2622] 0.0925 [0.2203] 774 0.140 6.741 6.456 6.846 month 10 harvest -0.0976 [0.2574] -0.0155 [0.2184] 763 0.108 6.543 6.448 6.444 month 11 preparation -0.2208 [0.3274] -0.0434 [0.2473] 768 0.103 6.755 6.530 6.744 month 12 preparation -0.4317 [0.2866] 0.0656 [0.2259] 767 0.135 6.686 6.299 6.832 month 13 lean -0.4242 [0.3298] -0.0815 [0.2516] 627 0.160 6.356 5.934 6.295 month 14 harvest -0.2721 [0.3064] -0.0748 [0.2402] 772 0.156 6.719 6.455 6.662 month 15 harvest -0.6054 [0.3238] -0.3163 [0.2445] 778 0.134 7.018 6.527 6.812 month 16 preparation -0.5595 [0.3033] -0.3377 [0.2502] 778 0.126 7.028 6.458 6.734 month 17 preparation -0.3983 [0.2939] -0.3239 [0.2394] 768 0.128 6.918 6.618 6.693 month 18 preparation -0.3503 [0.3106] -0.3378 [0.2463] 776 0.171 6.975 6.675 6.675 month 19 lean -0.4961 [0.3176] -0.1758 [0.2419] 759 0.157 6.652 6.217 6.582 month 20 lean -0.4459 [0.3162] -0.2123 [0.2473] 782 0.125 6.872 6.512 6.747 month 21 harvest -0.6055 [0.3136] -0.1719 [0.2489] 780 0.161 6.952 6.461 6.890 month 22 harvest -0.6756* [0.3015] -0.0620 [0.2509] 780 0.169 7.014 6.417 7.010 month 23 preparation -0.6811* [0.3308] -0.4134 [0.2672] 772 0.157 7.354 6.797 7.066 month 24 preparation -0.8456* [0.3285] -0.4661 [0.2601] 687 0.177 7.378 6.689 7.030 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 32 Table A5. Food security indicators – DID analysis Group HFIAS Score HDD Score rvest MFM month_2 -0.2271 0.2691 [0.5508] [0.1944] Mfi month_2 0.1327 0.4071 [0.5499] [0.2129] MFM month_3 -0.6325 -0.0470 [0.5967] [0.2479] Mfi month_3 0.2808 0.2224 [0.6102] [0.2345] Preparation MFM month_4 -0.6349 -0.0098 [0.6307] [0.2719] Mfi month_4 0.5503 -0.0116 [0.6361] [0.2652] MFM month_5 -0.5301 -0.0378 [0.6151] [0.2874] Mfi month_5 0.4334 0.3680 [0.6026] [0.2494] MFM month_6 -0.1043 -0.0376 [0.6202] [0.3203] Mfi month_6 -0.0684 0.6323* [0.5758] [0.2849] Lean MFM month_7 -0.0806 0.0110 [0.6426] [0.3027] Mfi month_7 0.2017 0.3153 [0.6120] [0.2928] MFM month_8 -0.0784 -0.0797 [0.6820] [0.3033] Mfi month_8 0.2512 0.1960 [0.6497] [0.3017] Harvest MFM month_9 0.1552 -0.2069 [0.7126] [0.3214] Mfi month_9 0.6045 0.3230 [0.6502] [0.3046] MFM month_10 -0.2501 0.0647 [0.7272] [0.3021] Mfi month_10 0.3281 0.2133 [0.6697] [0.2765] Preparation MFM month_11 -0.2579 -0.0796 [0.6959] [0.3676] Mfi month_11 0.5669 0.1753 [0.6488] [0.3038] MFM month_12 0.0127 -0.2751 [0.6590] [0.3370] Mfi month_12 0.6862 0.2934 [0.6891] [0.2954] Lean MFM month_13 -0.4851 -0.2594 [0.7589] [0.3859] Mfi month_13 -0.0345 0.1476 [0.7901] [0.3200] Harvest MFM month_14 -0.2703 -0.1340 [0.7208] [0.3705] Mfi month_14 0.2675 0.1815 [0.7158] [0.3149] MFM month_15 -0.4836 -0.4534 [0.7136] [0.3785] Mfi month_15 0.0473 -0.0582 [0.7215] [0.3138] 33 Group HFIAS Score HDD Score Preparation MFM month_16 -0.6278 -0.3833 [0.7017] [0.3649] Mfi month_16 0.0106 -0.0745 [0.6987] [0.3329] MFM month_17 -0.7468 -0.2327 [0.7838] [0.3552] Mfi month_17 -0.1742 -0.0612 [0.7459] [0.3166] MFM month_18 -1.1088 -0.1904 [0.7352] [0.3625] Mfi month_18 -0.1166 -0.0773 [0.7471] [0.3099] Lean MFM month_19 -0.3748 -0.3383 [0.7336] [0.3886] Mfi month_19 -0.0028 0.0731 [0.7759] [0.3362] MFM month_20 -0.6302 -0.2975 [0.6906] [0.4099] Mfi month_20 -0.5354 0.0313 [0.7318] [0.3491] Harvest MFM month_21 -0.1329 -0.4444 [0.7291] [0.3853] Mfi month_21 -0.6518 0.0904 [0.7571] [0.3264] MFM month_22 -0.3371 -0.5195 [0.7561] [0.3763] Mfi month_22 -0.8044 0.1803 [0.7959] [0.3321] Preparation APL month_23 -0.0810 -0.5207 [0.7745] [0.3802] MFM month_23 -0.6836 -0.1512 [0.7843] [0.3178] Mfi month_24 0.0251 -0.6817 [0.7956] [0.3919] MFM month_24 -0.5673 -0.1969 [0.8388] [0.3243] * p<0.05, ** p<0.1 Robust standard errors in parentheses. 34 Table A6. Borrowing from formal sources – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding -0.0337 [0.0445] -0.0841 [0.0451] 792 0.071 0.364 0.337 0.292 month 2 harvest -0.0411 [0.0421] -0.0818 [0.0423] 792 0.097 0.309 0.267 0.224 month 3 harvest -0.0014 [0.0368] -0.0341 [0.0373] 791 0.093 0.213 0.191 0.175 month 4 preparation -0.0113 [0.0380] -0.0251 [0.0371] 792 0.075 0.220 0.194 0.185 month 5 preparation -0.0125 [0.0431] -0.0601 [0.0392] 790 0.078 0.275 0.242 0.205 month 6 preparation 0.0051 [0.0397] -0.0131 [0.0372] 791 0.089 0.224 0.218 0.198 month 7 lean -0.0073 [0.0476] -0.0455 [0.0396] 791 0.045 0.248 0.249 0.195 month 8 lean 0.0656 [0.0471] -0.0244 [0.0376] 791 0.087 0.245 0.315 0.218 month 9 harvest 0.0554 [0.0467] 0.0037 [0.0405] 790 0.123 0.263 0.315 0.260 month 10 harvest 0.0518 [0.0501] -0.0344 [0.0438] 791 0.095 0.307 0.354 0.256 month 11 preparation 0.0418 [0.0491] -0.0553 [0.0443] 790 0.089 0.297 0.340 0.237 month 12 preparation 0.0289 [0.0468] -0.0712 [0.0445] 791 0.083 0.293 0.319 0.224 month 13 lean -0.0390 [0.0447] -0.0690 [0.0443] 791 0.116 0.269 0.230 0.198 month 14 harvest -0.0774 [0.0438] -0.0398 [0.0484] 789 0.073 0.343 0.254 0.289 month 15 harvest 0.0080 [0.0429] 0.0465 [0.0444] 791 0.061 0.259 0.265 0.289 month 16 preparation -0.0019 [0.0414] 0.0253 [0.0434] 791 0.043 0.269 0.265 0.273 month 17 preparation -0.0072 [0.0443] 0.0188 [0.0438] 791 0.052 0.290 0.280 0.286 month 18 preparation -0.0301 [0.0452] 0.0165 [0.0467] 790 0.079 0.324 0.292 0.306 month 19 lean -0.0509 [0.0439] -0.0524 [0.0443] 791 0.086 0.309 0.272 0.257 month 20 lean -0.0590 [0.0436] -0.0304 [0.0476] 791 0.096 0.326 0.276 0.300 month 21 harvest -0.0450 [0.0417] -0.0359 [0.0447] 791 0.094 0.326 0.276 0.293 month 22 harvest -0.0808* [0.0404] -0.0287 [0.0451] 790 0.093 0.334 0.249 0.300 month 23 preparation -0.0443 [0.0379] 0.0028 [0.0453] 786 0.081 0.284 0.240 0.280 month 24 preparation -0.0690 [0.0409] 0.0024 [0.0456] 721 0.107 0.269 0.191 0.258 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 35 Table A7. Kept cash savings – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding -0.0325 [0.0591] -0.0494 [0.0606] 792 0.047 0.787 0.764 0.734 month 2 harvest 0.0794 [0.0552] 0.0362 [0.0549] 792 0.052 0.753 0.837 0.795 month 3 harvest 0.0290 [0.0470] 0.0007 [0.0480] 791 0.070 0.825 0.856 0.805 month 4 preparation 0.0025 [0.0524] -0.0111 [0.0506] 792 0.063 0.770 0.779 0.750 month 5 preparation -0.0627 [0.0542] -0.0189 [0.0554] 790 0.067 0.780 0.719 0.747 month 6 preparation -0.0129 [0.0548] -0.0196 [0.0549] 791 0.050 0.790 0.798 0.766 month 7 lean 0.0022 [0.0553] -0.0170 [0.0588] 791 0.051 0.766 0.770 0.734 month 8 lean 0.0275 [0.0458] 0.0066 [0.0442] 791 0.079 0.790 0.813 0.786 month 9 harvest -0.0221 [0.0526] -0.0245 [0.0469] 790 0.060 0.779 0.755 0.747 month 10 harvest 0.0110 [0.0497] 0.0526 [0.0425] 791 0.102 0.762 0.774 0.805 month 11 preparation -0.0439 [0.0498] 0.0452 [0.0425] 790 0.092 0.803 0.762 0.834 month 12 preparation -0.0515 [0.0514] 0.0216 [0.0460] 791 0.065 0.790 0.739 0.802 month 13 lean -0.0716 [0.0380] -0.0494 [0.0308] 791 0.533 0.707 0.611 0.640 month 14 harvest -0.0271 [0.0489] 0.0227 [0.0413] 789 0.099 0.782 0.766 0.818 month 15 harvest -0.0215 [0.0467] 0.0048 [0.0433] 791 0.090 0.797 0.786 0.805 month 16 preparation 0.0072 [0.0402] 0.0507 [0.0393] 791 0.092 0.814 0.813 0.851 month 17 preparation -0.0279 [0.0488] 0.0289 [0.0440] 791 0.064 0.793 0.770 0.825 month 18 preparation 0.0337 [0.0441] -0.0182 [0.0440] 790 0.123 0.797 0.856 0.805 month 19 lean -0.0678 [0.0566] -0.0230 [0.0548] 791 0.036 0.790 0.747 0.785 month 20 lean -0.0553 [0.0471] 0.0067 [0.0435] 791 0.117 0.804 0.770 0.824 month 21 harvest -0.0393 [0.0444] 0.0097 [0.0432] 791 0.183 0.777 0.759 0.798 month 22 harvest -0.0518 [0.0463] 0.0130 [0.0429] 790 0.152 0.779 0.751 0.798 month 23 preparation 0.0127 [0.0450] 0.0331 [0.0433] 786 0.157 0.765 0.768 0.788 month 24 preparation -0.0090 [0.0549] 0.0488 [0.0478] 721 0.097 0.762 0.753 0.801 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 36 Table A8. Worked on someone else's land (wages) – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding 0.0121 [0.0316] -0.0046 [0.0304] 792 0.058 0.0859 0.105 0.0779 month 2 harvest 0.0188 [0.0275] 0.0394 [0.0298] 792 0.054 0.0687 0.0891 0.104 month 3 harvest -0.0143 [0.0338] -0.0041 [0.0357] 791 0.042 0.117 0.0973 0.0974 month 4 preparation -0.0144 [0.0327] -0.0159 [0.0343] 792 0.053 0.124 0.112 0.0974 month 5 preparation -0.0072 [0.0323] -0.0016 [0.0333] 790 0.060 0.127 0.121 0.107 month 6 preparation -0.0028 [0.0339] -0.0194 [0.0333] 791 0.070 0.131 0.136 0.101 month 7 lean 0.0520 [0.0324] 0.0186 [0.0329] 791 0.058 0.124 0.171 0.130 month 8 lean -0.0304 [0.0359] 0.0277 [0.0441] 791 0.060 0.159 0.132 0.179 month 9 harvest 0.0050 [0.0341] 0.0345 [0.0397] 790 0.038 0.107 0.113 0.133 month 10 harvest -0.0196 [0.0356] 0.0657 [0.0436] 791 0.067 0.128 0.105 0.175 month 11 preparation 0.0238 [0.0286] 0.0450 [0.0338] 790 0.051 0.107 0.133 0.143 month 12 preparation -0.0073 [0.0351] 0.0449 [0.0396] 791 0.066 0.155 0.152 0.179 month 13 lean -0.0282 [0.0315] 0.0779* [0.0376] 791 0.107 0.145 0.117 0.201 month 14 harvest -0.0084 [0.0361] 0.0422 [0.0405] 789 0.057 0.166 0.156 0.182 month 15 harvest -0.0091 [0.0337] 0.0436 [0.0401] 791 0.045 0.152 0.136 0.172 month 16 preparation -0.0212 [0.0312] 0.0311 [0.0371] 791 0.060 0.172 0.144 0.172 month 17 preparation -0.0146 [0.0374] -0.0098 [0.0374] 791 0.066 0.186 0.167 0.156 month 18 preparation -0.0502 [0.0346] 0.0025 [0.0371] 790 0.075 0.210 0.160 0.189 month 19 lean -0.0316 [0.0380] -0.0292 [0.0372] 791 0.069 0.206 0.183 0.153 month 20 lean -0.0206 [0.0352] -0.0148 [0.0363] 791 0.071 0.168 0.148 0.134 month 21 harvest -0.0076 [0.0357] -0.0391 [0.0356] 791 0.046 0.172 0.167 0.111 month 22 harvest -0.0190 [0.0367] -0.0264 [0.0350] 790 0.061 0.159 0.148 0.114 month 23 preparation -0.0318 [0.0380] -0.0262 [0.0374] 786 0.077 0.176 0.146 0.134 month 24 preparation -0.0565 [0.0396] -0.0404 [0.0395] 721 0.078 0.185 0.119 0.122 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 37 Table A9. Worked off-farm (self-employed) – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding 0.0329 [0.0338] 0.0862* [0.0427] 792 0.172 0.206 0.256 0.315 month 2 harvest 0.0199 [0.0349] 0.0739 [0.0446] 792 0.172 0.223 0.252 0.289 month 3 harvest 0.0360 [0.0345] 0.0750 [0.0414] 791 0.165 0.210 0.261 0.299 month 4 preparation 0.0196 [0.0378] 0.0716 [0.0408] 792 0.145 0.227 0.256 0.308 month 5 preparation 0.0210 [0.0359] 0.0959* [0.0415] 790 0.161 0.206 0.230 0.315 month 6 preparation 0.0475 [0.0366] 0.1133** [0.0420] 791 0.132 0.190 0.233 0.308 month 7 lean -0.0130 [0.0351] 0.0076 [0.0381] 791 0.141 0.228 0.218 0.240 month 8 lean 0.0235 [0.0361] 0.0046 [0.0382] 791 0.167 0.224 0.249 0.237 month 9 harvest 0.0246 [0.0367] 0.0210 [0.0372] 790 0.142 0.215 0.245 0.240 month 10 harvest 0.0175 [0.0403] 0.0233 [0.0404] 791 0.148 0.224 0.257 0.256 month 11 preparation -0.0027 [0.0400] 0.0157 [0.0386] 790 0.155 0.238 0.238 0.263 month 12 preparation 0.0094 [0.0377] 0.0288 [0.0376] 791 0.166 0.221 0.226 0.260 month 13 lean -0.0163 [0.0324] 0.0154 [0.0344] 791 0.120 0.179 0.152 0.192 month 14 harvest 0.0039 [0.0393] 0.0404 [0.0404] 789 0.148 0.225 0.227 0.269 month 15 harvest 0.0056 [0.0405] 0.0342 [0.0423] 791 0.136 0.234 0.233 0.276 month 16 preparation -0.0089 [0.0418] 0.0428 [0.0417] 791 0.140 0.241 0.226 0.286 month 17 preparation -0.0125 [0.0384] 0.0604 [0.0410] 791 0.134 0.228 0.206 0.292 month 18 preparation -0.0071 [0.0397] 0.0477 [0.0419] 790 0.148 0.238 0.222 0.287 month 19 lean -0.0317 [0.0422] 0.0228 [0.0440] 791 0.140 0.247 0.210 0.267 month 20 lean -0.0274 [0.0406] 0.0596 [0.0433] 791 0.154 0.251 0.218 0.303 month 21 harvest -0.0207 [0.0407] 0.0433 [0.0435] 791 0.152 0.258 0.230 0.300 month 22 harvest -0.0363 [0.0407] 0.0335 [0.0456] 790 0.141 0.266 0.226 0.300 month 23 preparation -0.0237 [0.0407] 0.0367 [0.0442] 786 0.144 0.263 0.236 0.300 month 24 preparation -0.0563 [0.0423] 0.0355 [0.0447] 721 0.140 0.265 0.217 0.310 * p<0.05, ** p<0.1 Robust standard errors in parentheses. 38 Table A10. Worked off-farm (wages) – OLS panel analysis by month Month Season MFM [S.E.] Mfinance [S.E.] Obs R2 Control avg MFM avg Mfi avg month 1 weeding -0.0235 [0.0233] -0.0129 [0.0241] 792 0.060 0.0722 0.0543 0.0584 month 2 harvest 0.0019 [0.0245] -0.0003 [0.0240] 792 0.027 0.0584 0.0659 0.0617 month 3 harvest 0.0088 [0.0185] 0.0373 [0.0224] 791 0.055 0.0309 0.0428 0.0747 month 4 preparation -0.0170 [0.0173] 0.0027 [0.0224] 792 0.054 0.0447 0.0349 0.0584 month 5 preparation -0.0255 [0.0148] -0.0061 [0.0207] 790 0.071 0.0447 0.0195 0.0422 month 6 preparation -0.0312 [0.0174] 0.0060 [0.0230] 791 0.060 0.0552 0.0272 0.0714 month 7 lean -0.0342 [0.0190] -0.0117 [0.0218] 791 0.049 0.0586 0.0272 0.0487 month 8 lean -0.0197 [0.0149] 0.0049 [0.0218] 791 0.064 0.0414 0.0233 0.0487 month 9 harvest 0.0073 [0.0162] 0.0107 [0.0194] 790 0.064 0.0311 0.0350 0.0455 month 10 harvest -0.0179 [0.0164] -0.0181 [0.0173] 791 0.054 0.0448 0.0272 0.0325 month 11 preparation 0.0045 [0.0173] -0.0039 [0.0189] 790 0.067 0.0414 0.0391 0.0390 month 12 preparation -0.0150 [0.0208] -0.0182 [0.0211] 791 0.073 0.0552 0.0350 0.0390 month 13 lean -0.0146 [0.0155] -0.0206 [0.0158] 791 0.075 0.0379 0.0195 0.0227 month 14 harvest 0.0053 [0.0174] 0.0097 [0.0186] 789 0.085 0.0311 0.0391 0.0455 month 15 harvest -0.0167 [0.0175] -0.0010 [0.0206] 791 0.090 0.0414 0.0311 0.0487 month 16 preparation -0.0043 [0.0154] 0.0102 [0.0186] 791 0.090 0.0310 0.0311 0.0455 month 17 preparation 0.0016 [0.0173] 0.0026 [0.0174] 791 0.076 0.0345 0.0389 0.0487 month 18 preparation 0.0006 [0.0172] 0.0060 [0.0172] 790 0.086 0.0310 0.0389 0.0554 month 19 lean 0.0022 [0.0164] 0.0142 [0.0177] 791 0.075 0.0275 0.0350 0.0554 month 20 lean -0.0009 [0.0165] 0.0068 [0.0180] 791 0.075 0.0344 0.0350 0.0521 month 21 harvest -0.0200 [0.0173] -0.0006 [0.0187] 791 0.093 0.0447 0.0272 0.0489 month 22 harvest -0.0243 [0.0177] -0.0041 [0.0189] 790 0.091 0.0483 0.0272 0.0489 month 23 preparation -0.0135 [0.0176] 0.0033 [0.0188] 786 0.094 0.0415 0.0315 0.0489 month 24 preparation -0.0069 [0.0167] 0.0033 [0.0186] 721 0.081 0.0346 0.0340 0.0453 * p<0.05, ** p<0.1 Robust standard errors in parentheses.