Women in Agriculture THE IMPACT OF MALE OUT-MIGRATION ON WOMEN’S AGENCY, HOUSEHOLD WELFARE, AND AGRICULTURAL PRODUCTIVITY Report No: AUS9147 May 2015 Women in Agriculture THE IMPACT OF MALE OUT-MIGRATION ON WOMEN’S AGENCY, HOUSEHOLD WELFARE, AND AGRICULTURAL PRODUCTIVITY Report No: AUS9147 May 2015 © 2016 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions.The findings, in- terpretations, and conclusions in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundar- ies, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concern- ing the legal status of any territory or the endorse- ment or acceptable of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax 202-522-2625; e-mail: pubrights@worldbank.org. A mother carries her child in a crop field in Chi- maltenango, Guatemala. Photo: Maria Fleischmann / World Bank Table of Contents Foreword v Acknowledgements vi Abbreviations and Acronyms vii 1. Study Overview 1 2. Introduction 2 Methodology 2 Background on Guatemala 4 Migration and Women’s Agency 5 3. Access to and Use of Endowments 5 Access to and Use of Endowments: Land 5 Access to and Use of Endowments: Labor 7 Access to and Use of Endowments: Knowledge 7 4. Impacts on Women’s Agency 8 Agricultural Agency Index 9 Non-Agricultural Agency 10 Soft Agency 11 5. Impacts on Household Welfare 12 Income: Amounts and Sources 13 Household Food Security and Diversity 14 6. Conclusions 15 References 16 Appendix A: Data Tables 18 Appendix B: Additional Agency Information and Results 20 Appendix C: Explanation of Variables 27 Appendix D: Regressions of Interest 30 Boxes Box 1: What is Agency? 10 Figures Figure 1:Year of Departure for Current Migration 4 Figure 2: Average Annual Agricultural Earnings per Hectare by Household Type (US$/ha) 6 Figure 3: Distribution of Agency Indexes by Household Type 9 Figure 4: Soft Agency Index by Household Type 12 Figure 5: Distribution of Autonomy Self-Rating (left panel) and Difference in Rating (right panel) by Household Type 12 Figure 6: Distribution of Annual Income Sources by Household Type (US$) 13 Figure 7: Distribution of Food Insecurity (left panel) and Food Diversity (right panel) by Household Type 14 Maps Map 1: Departments of Guatemala 3 Tables Table 1: Analytical Framework 3 Table 2: Agency Index Summary 8 Foreword This study explores the little studied phenomenon to feed its growing population. According to the of the impact of male out-migration on the wom- Food and Agriculture Organization (FAO), if ru- en and farms left behind, which is transforming ru- ral women in developing countries had the same ral economies, landscapes, and potentially, gender access to productive resources as men, they could relations. The objective of the study was to inves- increase yields on their farms by 20-30 percent.Yet, tigate the implications of male out-migration for very little data have been collected at the micro- women’s agency, household welfare, and agricul- economic level to analyze the impact on women tural productivity. Of fundamental interest is what left alone on farms after their partners’ migration decisions a woman is able to take, in agriculture, or the impact this has on agricultural yields, plant- the household, and the community in the absence ing decisions and other factors. of her male partner. The study focuses on Guatemala due to the relative The World Bank Group takes as its starting point importance of rural male out-migration there in that no country, community, or economy can recent decades and the significance of the Bank’s achieve its potential or meet the challenges of the current and future engagement in rural develop- 21st century without the full and equal participa- ment. A detailed survey questionnaire was devel- tion of women and men. Failure to fully unleash oped and piloted focused on migration, agricul- women’s productive potential meanwhile represents tural decision making and measures of agency. The a major missed opportunity with significant con- results yield important findings for policy makers, sequences for individuals, families, and economies. researchers, and others interested in the impact of male out-migration on the agriculture sector and Agriculture accounts for one-third of gross-do- on the women and families they leave behind. mestic product (GDP) globally and women’s par- ticipation is crucial if the world is going to be able Laurent Msellati Oscar Calvo-González Practice Manager Practice Manager Latin America and Caribbean Region Latin America and Caribbean Region Agriculture Global Practice Poverty and Equity Global Practice v Acknowledgements This study was financed by a grant from the Um- lo Chacon and Khanti Consultants, Abla Safir, brella Facility for Gender Equality. The study was Gero Carletto, Maria Beatriz Orlando, Sanna Liisa led by Victoria Stanley with a research team com- Taivalmaa, and Katherine M. Scott. Comments re- posed of Maira Emy Reimão, Barbara Coello, So- ceived during the session of the Land and Poverty phie Theis, and Marc Smitz. Guidance and review Conference 2015 and at the presentation organized were provided to the team by Holger Kray and by the Gender and Rural Development Themat- Martin Henry Lenihan. The team is also thankful ic Group are gratefully acknowledged. The ques- to Jason McMann and Mario Mendez for their tionnaire constructed for this study was built upon support with editing and finalizing the paper. The the Women’s Empowerment in Agriculture Index team also wishes to thank Elizaveta Perova, Pab- Questionnaire (WEIA) developed by IFPRI. vi Abbreviations and Acronyms FAO Food and Agriculture Organization (of the United Nations) GDP Gross domestic product NGO Nongovernmental organization WDR World Development Report WFP World Food Programme vii Women in Agriculture The Impact of Male Out-Migration on Women’s Agency, Household Welfare, and Agricultural Productivity1 1. Study Overview1 vast majority of households remain in agricul- ture even when the male head of household migrates. Migration is transforming rural economies, land- scapes, and, potentially, gender relations. Migration • The continuation of agriculture as a household is one of the drivers of the so-called “feminization livelihood strategy is characterized by the trans- of agriculture” in Latin America (Deere and León formation and expansion of the role of married de Leal 2001). This feminization has relevance for women in agricultural production. As men in everyone given agriculture’s role in regional food southeastern Guatemala now migrate for years security, national shared prosperity, and household at a time, their partners face greater responsibil- resilience to shocks. ities in agricultural production, both in decision making and in production itself. This little studied phenomenon is not yet well un- derstood and new evidence is needed. To fill this • These households, where the male partner has gap, financing was secured from the Umbrella Fa- migrated, are more likely than other types of cility for Gender Equality and household surveys households to employ non-household members were conducted in two departments in southeast- or paid workers for agricultural labor. This gap ern Guatemala. The objective of this study is persists even when controlling for the depen- to investigate the “feminization” of agricul- dency ratio2 and household size. ture as well as its implications for women’s • As agriculture is still seen as a traditionally male agency, household welfare, and agricultural endeavor, women reported having not only to productivity. Of fundamental interest is what take on farming, but also to learn how to farm forms of engagement a woman wants to take and is once their husbands migrated. But extension able to take in agriculture, the household, and the services and technical assistance generally fail to community in the absence of her male partner. In reach women in rural areas. particular, this analysis seeks: (i) to understand how male out-migration is influencing women’s agency • Households with a male partner who has mi- in agriculture; (ii) to understand if, when wom- grated have the highest levels of food security en are in control of farms, it changes the types of and food diversity relative to other groups. Giv- decisions they make and thus the results that they en the higher level of remittances received by obtain; and (iii) to get a better sense of how dif- these households and the fact that remittances ferences in agency (if any) lead to better or worse tend to go directly to women, this result is in outcomes for the farm household. line with literature showing that money con- trolled by women is allocated at greater rates Five key findings of interest from the study are as toward family nutrition than money controlled follows: by men (Thomas 1990). • Contrary to the popular belief held by local of- ficials, policy makers, and researchers alike, the 1 Unless otherwise noted, all tables and figures appearing 2 The dependency ratio refers to the ratio of the number of in this study derive from the original survey research de- household members under age 15 and over age 64 to the scribed herein. number of working-age household members. 1 2. Introduction Methodology This analysis seeks to investigate the impact This section provides an introduction to the role of male migration on agriculture and its im- of women in agriculture, lays out the study meth- plications for women’s agency and agricul- odology, and provides background information on tural productivity, as mediated by factors such migration, women, and agriculture in Guatemala. as land tenure and access to agricultural extension Women have a central role at the nexus of services. In particular, this analysis seeks: (i) to better rural development, food security, and agri- understand how male out-migration is influencing culture (FAO 2011b; World Bank 2011). Accord- women’s agency in agriculture; (ii) to understand ing to the Food and Agriculture Organization if, when women are in control of their farms, it (FAO), if rural women in developing countries had changes the types of decisions they make and thus the same access to productive resources as men, the results they obtain; and (iii) to get a better sense they could increase yields on their farms by 20-30 of how differences in agency (if any) lead to bet- percent. This could raise total agricultural output ter or worse livelihood outcomes for farm house- in developing countries by 2.5-4 percent, which holds. Table 1 outlines the framework used for the could in turn reduce the number of hungry people analysis, based on the analytical framework used for in the world by 12-17 percent (FAO 2011b; IFPRI the 2012 World Development Report (WDR) on 2003). According to a recent World Bank (2014) Gender and Development (World Bank 2011).The study, this access to inputs has to include access to WDR recognized the importance of access to and labor, technology, and knowledge and may need to use of endowments such as land, labor, and knowl- be tailored for women farmers. edge but also raised the profile of women’s agency as a key to economic development. Finally, data on Women’s role in agriculture is particularly household income and food security are analyzed crucial in Guatemala, which suffers from the to understand the impact of migration and wom- double burden of chronic malnutrition and en’s agency on household well-being. obesity. The country has a competitive agro-food sector, while at the same time the rate of chronic This study is based on a quantitative field malnutrition in its rural areas is one of the highest in survey conducted in August 2014, as well as the world. The agriculture sector represent 11 per- qualitative focus groups and interviews conducted cent of gross domestic product (GDP), with food in May 2014 to test the questionnaire. The study exports representing more than 44 percent of total was performed in two southeastern departments of exports (World Bank 2015). Despite this, Guatema- Guatemala, Jutiapa and Chiquimula (see Map 1), la’s chronic undernutrition rate is currently at 49.8 on a sample of 572 agricultural households.4 The percent among children under five (WFP 2015). sampling process ensured that the results presented here are representative of these two departments, Very little data have been collected at the mi- which are near the border with Honduras and El croeconomic level to analyze the impact on Salvador. women left alone on farms after their part- ners’ migration (FAO 2011b; World Bank 2012; The households interviewed are classified etc., among many others). Women seem to be large into three groups: but statistically invisible contributors to rural life • Type 1: Women whose male partners are cur- through paid and unpaid employment. According rently migrants. to FAOSTAT (2013), women in Guatemala repre- • Type 2: Women in households where both the sent almost 10 percent of the labor force in agricul- male and female heads are present (independent ture, while the International Labour Organization of possible migration history). reports that 12.6 percent of female employment in Guatemala is in the agriculture sector.3 • Type 3: Single female-headed households. 3 Data come from the World Bank’s (2015) World Develop- ment Indicators. 4 See Appendix A for additional details. 2 Women in Agriculture Table 1: Analytical Framework Access to/Use of Endowments Women’s Agency Livelihood Outcomes Land Soft Agency Food Security Labor Agricultural Income Knowledge Non-Agricultural Map 1: Departments of Guatemala 3 Figure 1: Year of Departure for Current Migration .08 Kernel Density Function .06 .04 .02 0 1990 1995 2000 2005 2010 2015 Year Note: Figure based on data from Type 1 households only. The use of three groups (one treatment and two they are relatively long. While only 16 percent control groups) allows the analysis to parse out the of partners in dual-headed households ever lived in effects of having a migrant partner versus being a the United States, those who did spent 50 months single head of household and gives an opportunity away on average.The sample confirms that out-mi- to see general social norms across households in a gration is largely a male phenomenon in rural given community. For simplicity, from this point southeastern Guatemala, as over the last 10 years forward, women/households are referred to by only 15 percent of women with migrant husbands their “type,” as classified above. have lived outside the locality in which they were interviewed, with most living elsewhere in the Background on Guatemala country rather than abroad. In Central America and particularly in Gua- The decision to migrate appears to fall most- temala, male out-migration is accelerating; ly within men’s domain. In speaking of their more than 70 percent of migrants are young partner’s most recent or current migration episode, males and almost 90 percent of these mi- 81 percent of women with migrant partners and 77 grants are in the United States. (Cohn, Gon- percent of women currently in dual-headed house- zalez-Barrera, and Cuddington 2013). This report holds said that the decision to migrate was made by draws on data collected in Chiquimula and Jutiapa, their partner alone. Only 15 percent and 18 per- as this region did not suffer as much displacement cent, respectively, said that their partner’s migration relative to other parts of the country due to the was a joint decision. civil war (1960-1996). Thus little evidence of mi- In southeastern Guatemala, agriculture is tra- gration is seen in previous generations here (only ditionally a male endeavor; although women six of the women interviewed said that either one participate in several areas of the production of their grandparents or their spouse’s grandparents process, men are the primary decision mak- had ever lived abroad). Nonetheless, as many as 10 ers. In 85 percent of Type 2 households in the sam- percent of women said that their father had lived ple, for instance, women do not participate in the in the United States, and those women are more decision of what to plant. Similarly, 88 percent do likely to be currently married to migrant husbands. not take part in deciding what inputs to use. Nev- These days, migration episodes tend to occur ertheless, about half of Type 2 women participate only a few times in a person’s life, though in some part of the agricultural production process, 4 Women in Agriculture with 27 percent purchasing inputs and 30 percent or access to the same economic support systems. taking part of the crop harvest. Further, qualitative Other researchers find that traditional gen- interviews revealed that women play a critical sup- der divisions of labor can be reinforced by porting role on a daily basis: as some of the land male migration. Given few labor opportunities used is hours away from the house by foot, men outside of agriculture, some studies show that it is may spend the day there, while women walk back rare for women to join the labor force in agricul- and forth to bring food and supplies as needed. ture or otherwise. Pessar (2005) notes that there are instances in which “women (commonly from more The vast majority of households remain economically secure households) are forbidden by in agriculture even when the male head of migrant husbands to work outside the home.” household migrates, contrary to the popular belief held by local officials, policy makers, But migration has the potential to change and researchers alike. Households’ persistence social norms within a community that pre- in agriculture has been defined by the transforma- scribe how women participate in agricul- tion and expansion of the role of married women ture, community groups, household decision in agricultural production. As men in southeastern making, and so on. The experience of heading Guatemala now migrate for years at a time, their the farm and household in the absence of her part- partners face greater responsibilities in agricultural ner may earn a woman more trust and authority production, both in decision making and in pro- from her partner, peers, and community – and pos- duction itself. In contrast to Type 2 households, sibly increase her own sense of self-efficacy, or the half of women in Type 1 households participate internal component of agency. Certainly, bargain- in the decision of what to plant and what inputs ing within the household is affected by structural to use (and the majority of these make the deci- conditions and institutions in which the household sion alone). Even more dramatically, 73 percent of is embedded (Agarwal 1997). Furthermore, women women in Type 1 households actually participate in communities with high levels of out-migration, in some part of agricultural production, 60 percent even if they themselves do not have a partner who purchase inputs, 50 percent harvest, 42 percent said has migrated, may experience changes in gender they participate in planting, and 44 percent partic- roles over time. ipate in cleaning the land. Migration and Women’s Agency 3. Access to and Use of The literature on the effects of male migra- tion on women’s agency and empowerment Endowments reveals a mixed picture (Menjivar and Agadja- This section looks at women’s access to and use nian 2007). First, agency is variable depending on of endowments such as land, labor, and knowledge the domain; undoubtedly, migration of a male part- and the differences between household types. The ner does not increase agency across all domains – study finds that women’s access to both labor nor does it decrease agency across the board. In and knowledge impacts their ability to farm some cases, women see an expansion of their tra- all of the land they own or have access to. In- ditional roles. Some studies find that out-migration terestingly, women’s access to land does not appear increases women’s participation in the labor force, to be the primary constraint and women are just as often even in traditionally masculine activities likely to have documents to their land as men. (Mummert 1988). The increase in responsibility is often not by Access to and Use of Endowments: choice but out of necessity, when remittances Land are insufficient or erratic (Pessar 2005). These new roles may represent an excessive time burden Land productivity is similar across all three with the loss of male labor, or represent obligations groups. Overall, Type 2 households’ annual farm- that are not always accompanied by social approval ing incomes (US$776) are almost twice as large as 5 Figure 2: Average Annual Agricultural Earnings per Hectare by Household Type (US$/ha) 1800 1600 1400 1200 1000 800 600 400 200 0 Type 1 Type 2 Type 3 Farming Farming + Animal Husbandry Note: Income reported here represents annual income. those of households in the other two types: US$415 others.Very few plots of land (less than 10 percent) and US$365 for Type 1 and Type 3 households, re- are rented for money or used for sharecropping. spectively. The former also use more land than the Though women participate in agricultural other two groups, so when considering agricultural production, their levels of land ownership income per hectare, most of this difference disappears. are relatively low and show evidence of male Type 2 households produce on average US$1,527 preference in inheritance. Over half of the land per year per hectare (including household con- owned by households in the sample was acquired sumption), but Type 1 households are not far be- through inheritance, and was much more likely to hind, at an annual rate of US$1,435 per hectare. come through male lineage than female lineage. Type 2 households are 16 percent more likely to sell Furthermore, in only 58 percent of cases in which at least some agricultural production. The lack of a plot was inherited through the woman’s side of economies of scale for Type 2 households may sig- the family (from her parents or relatives) was she nal labor and input constraints in rural Guatemala.5 listed as an owner of that plot.6 For Type 1 households, with one fewer male Other important differences in women’s member and a shift of the purpose of agri- ownership exist across groups, with more than culture towards subsistence or consumption 30 percent of women from Type 1 households smoothing during periods of lower remit- owning at least one plot (jointly or as sole owner), tances, agricultural production is lower than in contrast to 21 percent of women from Type 2 for Type 2 households. This is a function of land households, and the difference is statistically signif- use, not productivity. Type 2 households tend to use icant. Notably, 20 percent of women from Type 1 more agricultural land than the other two groups. households are the sole owner of at least one plot Nonetheless, there is no statistically significant dif- ference in the likelihood of owning agricultur- al land across groups (see Appendix A,Table A2). Two in five households in the sample own agricul- 6 In the survey, women were asked to list the owners of each tural land (three in four own some land, including plot of land used or owned by the household. As such, it their household plot), though in all three groups it favors “perceived” ownership over legal ownership. None- theless, the results indicate that in 42 percent of cases in is common to use both land owned and land from which the plot of land was inherited from the woman’s side of the family, the respondent did not consider the land 5 Note that land quality is not taken into account. to be hers. 6 Women in Agriculture of land, in contrast to 13 percent of women in The possibility of hiring outside workers was Type 2 households. consistently mentioned by women across all types during the qualitative interviews, and Importantly, 80 percent of agricultural land seems to be a very important constraint for owned by households in the sample has doc- women in Type 1 households. Women in Type umentation (51 percent have a deed; 32 percent 1 households have one less adult than do Type 2 are also registered). No differences were found households, but also a higher dependency ratio. in the likelihood of documentation between fe- The latter also explains why even though Type 2 male-owned and non-female-owned land, indicat- households tend to use more agricultural land than ing that in this context the documentation process the other two household groups, no statistically is no more inclusive of women than men. No dif- significant difference is found in the likelihood of ferences were found in the likelihood of documen- owning agricultural land across groups. tation across household groups. Access to and Use of Endowments: Access to and Use of Endowments: Labor Knowledge As agriculture is a traditionally male endeav- Type 2 households are less likely than the or, women reported having to not only take other two household types to employ on farming but also to learn how to farm once non-household members or paid workers for their husbands migrated. In focus groups, several agricultural labor. This gap persists even when women said they did not know how to farm when controlling for the dependency ratio and house- their husbands decided to migrate, learning just be- hold size, suggesting that in Type 2 households the fore they left or, once their partners left, from male male head of household may undertake a signifi- relatives or from their partners over the phone. cant portion of the agricultural tasks that cannot be easily done by women. Instead, Type 1 and Type Male relatives are an important source of 3 households rely on outside help (paid or information and advice on agriculture for unpaid) to replace this source of labor. women. The quantitative study found that the majority of women who know to farm first learned During the qualitative interviews, women from their fathers (70 percent). Partners are also a explained that one of the reasons they cannot principal teaching source, especially for women cultivate all their land is the lack of available with migrant partners: 24 percent of Type 1 wom- labor.Women also explained other difficulties with en first learned how to farm from their partners, in hiring laborers – weak negotiating power, inability contrast to 18 percent of Type 2 women. to monitor the quality of work, and women not being considered as a “real” farmer. Most women Extension services and technical assistance cope with these constraints by asking for help from generally fail to reach women in rural ar- another male in the household or community to eas. Only 13 women in the entire dataset said manage the hiring and supervision of workers. they received technical assistance in the last 12 months. Two-thirds of women noted that they do Households that can employ outside workers not currently learn about agriculture from anyone, have significantly higher agricultural income. including extension services or neighbors, parents, After the total amount of land is incorporated, this etc. About 25 percent of women learn from family is the second most important factor in the expla- members or neighbors. The scarcity of extension nation of agricultural income. It seems to reinforce services is corroborated by the fact that only six the idea that households in rural areas have a high- extension agents serve the entirety of the two de- er income – between US$160-200 higher – when partments in the study – with only three agents per they are able to hire an external worker to help department – two generally serving male groups them accomplish some of the agricultural tasks. and one serving women. The focus groups and consultations revealed that, with a few exceptions, 7 Table 2: Agency Index Summary Agency Index Component Variables Self-Determination/ Soft Autonomy Self-esteem/Aspiration Self-perception Decision in the Participation in the com- Non-Agricultural Access to financial services household munity Agriculture deci- Agricultural Agriculture actions Agriculture ownership sions the extension services offered to women focus on way, are women able to exercise their agency? nutrition and food preparation. These efforts stand Women living in the context of male out-migra- in stark contrast with women’s preferences and the tion represent an opportunity to empirically test role they play in agriculture, as 7 in 10 women in two important research questions that remain to the sample stated that they would like to receive be answered to better understand agency: (i) how extension services or training in agricultural pro- women’s self-evaluated agency and agency out- duction. The highest demand is for training on se- comes are associated; and (ii) how agency in differ- lecting seeds (42 percent of all women in the sam- ent domains potentially relate to one another. ple), animal immunizations (41 percent), and pest As seen above, male out-migration has trans- control (36 percent). formed agricultural roles, increasing women’s The lack of technical assistance for wom- participation in agricultural production. The en in agriculture is alarming, as households large influx of remittances also changes the distribu- in which women reported that they do not tion of money over which women have control, or learn about agriculture from anyone have are at least responsible for managing.Together, these lower agricultural and total incomes rela- shifts may influence women’s choices, self-percep- tive to other groups. Specifically, households tion, sense of empowerment, and ability to act. This in which women reported that they learned how section looks specifically at women’s agency in agri- to farm alone have agricultural incomes that are cultural decision making, non-agricultural decision US$371 lower than those of other households; after making, and ”soft agency” (see Box 1). controlling for other variables such as household What emerges is that women in Type 1 type and size, this difference decreases to US$240- households tend to have more agency – both 260, but is still significant. Similarly, total household agricultural and non-agricultural – than Type income is US$1,084-1,216 lower for those women 2 households, meaning that they are more involved who learned how to farm on their own, even when in decision making for both the farm and the controlling for covariates. This result highlights the household. Soft agency measures, however, re- high cost of the lack of extension services, borne veal that women in Type 1 households do not not only by women and their households but also necessarily see themselves as freer or more by the agriculture sector as a whole. autonomous than other household types. The distribution of agency measurements 4. Impacts on Women’s varies by household type and by the agen- cy measurement used. Nonetheless, in all four Agency agency indices, women in Type 3 households have a higher level of agency on average, as shown in The second realm of impact of male out-mi- Figure 3. Women in Type 1 households have a gration in Guatemala studied here is wom- higher agency level in agricultural and non-agri- en’s agency. When men leave their farms to mi- cultural dimensions of agency relative to those in grate internationally, to what extent, and in what Type 2 households, but the distribution of the soft 8 Women in Agriculture Figure 3: Distribution of Agency Indexes by Household Type Agricultural Agency Non-Agricultural Agency 8 4 Kernel Density Function Kernel Density Function 3 6 2 4 1 2 0 0 -2 -1 0 1 2 -2 -1 0 1 2 3 Type 1 Type 2 Type 3 Type 1 Type 2 Type 3 Soft Agency Overall Agency 5 6 Kernel Density Function 4 Kernel Density Function 4 3 2 2 1 0 0 -3 -2 -1 0 1 2 -2 -1 0 1 2 Type 1 Type 2 Type 3 Type 1 Type 2 Type 3 agency measurement is similar for all three groups. groups report being the sole decision maker. Even With respect to overall agency measurement, Type when several dimensions are combined, including 2 women have the lowest level of agency, followed how women participate in the decision on what to by Type 1 and then Type 3 women. plant, the decision on inputs, and more generally on agricultural production, the results show a sim- Agricultural Agency Index ilar trend of women in Type 2 households partici- pating less in these decisions. Women in Type 1 and Type 3 households have significantly more decision-making agency Among households with small animals, women in agriculture. Women in Type 1 households are tend to be responsible for them, though at a lower also more likely to participate in agricultural deci- rate among Type 3 households. The latter might be sion making, with 64 percent of women in Type 1 due to women’s responsibilities for everything else. versus 20 percent of women in Type 2 households The survey included questions about large animals, participating in the decision of what crops to plant. but very few households own them (16 percent), Only 2 percent of women in Type 2 households and when they do, women are usually not respon- report being the sole decision maker on that is- sible for them. Thus, the index does not cover this sue, while 50 percent of women in the other two dimension of animal ownership. 9 Box 1: What is Agency? This study draws upon the concept of agency to enrich the understanding of female empower- ment in agriculture. The questionnaire used in this study built upon the Women’s Empowerment in Agriculture Index (WEAI) (Alkire et al. 2013), and was designed specifically to focus on trends in the feminization of agriculture due to male out-migration. Agency, then, is a quality or capacity exercised when a person is able to capitalize on endowments and economic opportunities to lead to desired actions. Sen (1989) defines agency as an individual’s ability to act on behalf of what the individual values and has reason to value. Agency is not “global” but rather multidimensional in the sense that an individual can exercise dif- ferent levels of agency in pursuit of multiple aims (Alkire 2005). These aims can be very diverse and someone may have variation in her level of agency with respect to different aims (Alkire 2008). A woman may have significant say in decision making over what kinds of food to buy, but no control over the amount of income she is allocated by her husband out of their earnings. Consequently, the measure of agency used captures three different dimensions of rural women’s agen- cy related to agricultural and non-agricultural variables: (i) soft agency; (ii) non-agricultural agency; and (iii) agricultural agency. Each of these measurements is built using three components, as described in Table 2. A description of each variable is included in Appendix C, along with a brief explanation of the methodology used to construct the agency measurements (principal components score). This study also includes a broader agency index that combines all nine variables into a single index. One important part of the survey instrument was the deeper exploration of “soft agency.” Based on qualitative work, the survey questionnaire was developed with several modified psychosocial scales to measure women’s self-determination and self-esteem. This section included questions on women’s self-perception on specific qualities of agency in contrast to other women like her and her perception of cultural norms to contextualize what kinds of choice and behavior are perceived as possible in a community. This soft agency section serves to test some of the psychosocial scales in a new context amongst rural women and to test links between soft agency and other factors of agency, like decision making and access to endowments and economic opportunities. Non-Agricultural Agency women who stated they were not employed, for example, 45 percent of those in Type 1 households The non-agricultural agency index comprises and 35 percent of those in Type 2 households stat- three dimensions: the distribution of house- ed that their partners were the ones who decided hold decision making, participation in the that the women would not work outside the home. community, and access to financial services. In contrast, 61 percent of single/widowed women Household decision making comprises various who do not work made that decision themselves. realms, some of which may be traditionally within The latter group is also much more likely to de- women’s domains (e.g., food) and others that are not cide alone on any other activities they do outside (e.g., the household’s overall budget). The extent to the household: 89 percent of them decide in which which women participate in local groups, both as a activities to participate, in contrast to 47 percent member and in leadership positions, as well as their of women from Type 1 households and 34 percent access to banking are also considered. of women from Type 2 households. Notably, how- The roles played by individuals in decisions ever, over half of the women who said they played vary by household type, and married wom- no role in deciding on their activities outside the en are less likely to make decisions regard- house also stated that they did not wish they had ing their own time and employment. Among more decision-making power. 10 Women in Agriculture Women in Type 1 and Type 3 households have management and receipt of remittances. In fact, re- a greater say in the household budget than ceiving remittances increases the likelihood that a do women in Type 2 households. As many as woman has a bank account by 9 percentage points. 57 percent of women in Type 1 households and 77 Having a bank account is associated with percent of women in Type 3 households say they higher incomes. Households in which women decide and manage the household budget alone, have a bank account have earnings that are US$898 while only 13 percent of women in Type 2 house- more than households in which they do not; this holds do so. Another 36 percent of women in Type amount reaches US$1,023 for Type 1 households. 1 households share this responsibility, but 30 per- Given the overall low rates of credit use in this cent of women in Type 2 households have no say in context, only access to a bank account is used as a the household budget. This pattern also holds true proxy for access to financial services. for household food decisions: in 39 percent of Type 2 households the male partner decides alone how It should be noted that the level of agency mea- much to spend on food, while 75 percent and 83 sured for agricultural and non-agricultur- percent of women are the sole decision makers in al decisions is higher for women in Type 1 Type 1 and Type 3 households, respectively. households than those in Type 2 households. The participation of women in any type of productive group, other than church and Soft Agency sports activities, is very low, at around 22 The soft agency measurement designed percent. These results are somewhat surprising, through the survey comprises three variables. particularly considering the extremely low stated Specifically, it considers self-efficacy (sense of free- participation in productive groups (less than 10 dom and choice), aspirations (abilities and goals), and percent). It is possible, however, that the question autonomy. Figure 4 shows average scores in each of on “belonging to a group” may not have been well the three dimensions by household type. More de- understood or interpreted by the respondent as tail on the questions used to elicit these psychosocial envisioned in the survey design, as the qualitative measurements is included in Appendix C. work in several communities showed higher levels A greater share of Type 3 women relative to of women’s participation in groups organized by women in other groups perceive themselves local NGOs. as very autonomous. In the “autonomy” ques- Very few women have a leadership position tion, women were asked to position themselves on in their community. As expected,Type 3 women a ladder with 10 rungs, with the first rung repre- are slightly more likely to be leaders (23 percent) senting someone without any freedom and the top on average than Type 1 and Type 2 women, at 18 rung (i.e., the tenth) representing someone who percent and 17 percent, respectively. is completely free. Figure 5 shows that women in Type 3 households tend to perceive themselves as The use of credit and insurance is low in this more free than others. region of Guatemala. Less than 10 percent of households in the sample have any credit and fewer It is interesting to note that women in Type 1 than 7 percent have formal credit (with a bank or households are more likely to give themselves NGO). Around 7 percent of households also carry the same rating of freedom as they assign to some form of life insurance. the rest of the women in their community. A follow-up to the autonomy question asked women However, 33 percent of women in Type 1 to state the rung on which they thought most of households have an independent bank ac- the women in their community would be. Figure count. This is significantly more than women in 6 shows the distribution of the difference between Type 2 (11 percent) and Type 3 (14 percent) house- the woman’s own rung and the rung she assigned holds. Women in Type 1 households might enjoy to women in her community, so that zero indicates a secondary effect due to their higher familiarity she placed both of them on the same rung; a pos- with financial institutions provided by the necessary itive number indicates that the woman thinks she 11 Figure 4: Soft Agency Index by Household Type 10 8 6 4 2 0 Type 1 Type 2 Type 3 Self-Efficacy Aspirations Autonomy Figure 5: Distribution of Autonomy Self-Rating (left panel) and Difference in Rating (right panel) by Household Type Distribution of Self-Rating Distribution of Difference in Rating .25 Kernel Density Function .3 .2 .2 .15 .1 .1 .05 0 0 0 2 4 6 8 10 -10 -5 0 5 10 Type 1 Type 2 Type 3 Type 1 Type 2 Type 3 Note: The difference reported in the right panel represents respondents’ self-rating minus their rating of other women. has more freedom than the rest of the women in her community, while a negative number indicates 5. Impacts on Household less freedom. Welfare The survey finds that women in Type 3 house- holds are more likely not only to place them- In the context of the high levels of malnutri- selves high, but also to consider themselves tion found in Guatemala, two principal mea- to be freer relative to the rest of women in the surements of family welfare are household community. The high concentration of Type 1 food security and food diversity. This section women at zero is inconsistent with higher levels of explores the differences in income sources across agency in agricultural and non-agricultural mea- the three groups, and the differences in food securi- sures, and raises the possibility that their responses ty and food diversity between them. Type 1 house- to the autonomy question were biased in an at- holds have higher levels of food security and food tempt to “fit in.” diversity compared to the other household types. 12 Women in Agriculture Figure 6: Distribution of Annual Income Sources by Household Type (US$) 3,000 33 68 223 79 2,000 404 1659 1645 1559 1,000 506 873 488 449 0 Type 1 Type 2 Type 3 Agriculture Labor Remitances Other Income: Amounts and Sources Type 1 households use remittances to make up for losses in agricultural and wage income. Contradicting common belief in Guatemala, migrant households are not richer than the Notably, among households in the latter two rest. Some of the new social programs being de- groups that do receive remittances, the trans- signed at the time of fieldwork excluded migrant fers are also fairly large: on average, Type 2 re- households, assuming that they were always better mittance-recipient households receive US$1,023 off than other types of households, given that they per year; Type 3 households receive an average of had a supplementary income source in the form US$1,158. Nonetheless, these are around half the of remittances. Instead, this study finds that house- amount received by Type 1 households that receive holds in the three groups have, on average, the same remittances (79 percent), for which the average an- amount of total income (Figure 6).The average an- nual amount is US$2,192. Interestingly, no differ- nual income for Type 1 households is US$2,715; ence exists across households in women’s participa- for Type 2 households, US$2,769; and for Type 3 tion rate in deciding what to do with remittances. households, US$2,437.7 Type 2 households that receive remittances Not surprisingly, women from Type 1 house- are 21.6 percentage points less likely than holds have a higher share of income from Type 1 households to use remittances for remittances. On average, Type 1 households re- food. Type 3 households are 9 percentage points ceive US$1,659 in remittances per year, in contrast less likely to do so. No difference exists in the like- to US$223 for Type 2 households and US$404 for lihood of spending remittances on education, even Type 3 households. While total income across the when accounting for the number of children. Type three groups varies little, the composition differs, as 1 and Type 2 households are just as likely to use re- mittances for agriculture (13-15 percent), but Type 3 households are less likely to do so. 7 Migrants earn more income than what they remit, and the total amount was not taken into account in computing Type 2 households are more likely to be en- “household income,” including only the amount received gaged in wage/salaried work. About two in in remittances. (It was not possible to collect data on mi- three (67 percent) Type 2 households have income grants’ total earnings, as interviews were carried out with from non-agricultural work, along with 55 percent their spouses, many of whom may not know or want to report their partner’s earnings abroad.) In this sense, for of Type 3 households but only 26 percent of Type economic purposes a “household” is considered as the 1 households. family members and other individuals living in the same house and sharing meals, with remittances an additional Government transfers represent a very small source of income. amount of total income (i.e., other income). 13 Figure 7: Distribution of Food Insecurity (left panel) and Food Diversity (right panel) by Household Type Household Food Insecurity Household Food Diversity .06 .2 Kernel Density Function Kernel Density Function .04 .15 .02 .1 0 .05 0 9 18 27 0 2 4 6 Type 1 Type 2 Type 3 Type 1 Type 2 Type 3 Note: In the left panel, higher values on the x-axis indicate more food insecurity. In the right panel, higher values on the x-axis indicate more food diversity. Three in ten (30 percent) households receive gov- as 1990). A surprising and perhaps alarming result, ernment cash transfers, but, as corroborated by the however, is that Type 3 households (female-headed qualitative interviews conducted and the explicit households) have the most precarious nutritional exclusion of households with migrants from so- status, particularly with respect to their levels of cial programs, Type 1 households have much lower food insecurity. rates of transfers (16 percent) compared to Type 2 Households with a higher share of agricul- (32 percent) and Type 3 (42 percent) households. tural income to total income are slightly less The amount of the transfer is quite small, howev- likely to be food insecure but also less likely er, so Type 1 households receive on average US$13 to have food diversity. That is, while agricultural per year, compared to US$30 among Type 2 house- production stabilizes access to food so that house- holds, and US$49 among Type 3 households. The holds are less likely to go days without eating or most common type of in-kind transfer in rural with little food, for instance, they are also less likely Guatemala is fertilizer: 45 percent of Type 1 house- to experience diversity in their food, as they rely holds on average receive fertilizer, compared to 65 on their own production for food and that produc- percent of Type 2 households. tion is limited in diversity. Households that rely on remittances or other sources of income may buy a Household Food Security and wider range of foods. As expected, higher income Diversity is correlated with lower food insecurity and higher Type 1 households have the highest levels of food diversity. food security and food diversity compared Households that receive remittances have to the other groups, as indicated in Figure 7. higher levels of food diversity, though not Given the higher level of remittances received necessarily food security. For households that by these households and the fact that remittances receive remittances, the amount of remittances has tend to go directly to women, this result is in line a small but significantly positive effect on food with literature showing that money controlled by security and diversity. Interestingly, and perhaps women is allocated at greater rates towards family contrary to the literature on women’s allocation nutrition than money controlled by men (Thom- of resources, the study does not find evidence that 14 Women in Agriculture women’s participation in the decision of how re- en may not see themselves as freer or may feel bur- mittances are allocated affects food security or di- dened by the need to make more decisions alone. versity. This may be due to sample size limitations, While land productivity is similar across all three or may be attributed to the fact that the major- groups of households, farming income varies across ity of households (79 percent) allocate some of households, with households in which a male head their remittances towards food anyway, regardless is present reporting the highest farm income. But of whether the woman participates in the deci- when considering agricultural income per sion-making process. hectare, most of this productivity difference disappears. 6. Conclusions The lack of economies of scale for migrant households may signal labor, input, and The research yields important findings for policy knowledge constraints in rural Guatemala. makers, researchers, and others interested in the im- The lower farm income reported by these agricul- pact of male out-migration on the agriculture sector tural households appears to have less to do with de- and on the women and families they leave behind. cision making and more to do with the high infor- Contrary to popular belief, the vast majori- mational and labor barriers faced by women.While ty of households remain in agriculture after women may wish to stay in agriculture, their lack the migration of the male head of household. of knowledge and access to labor and other inputs However, they tend to shift the purpose of agricul- hampers them from becoming more productive. ture towards subsistence and consumption smooth- Diversifying risk in the household by diversifying ing during periods of lower remittances. agricultural production is an important factor of When men out-migrate, women report hav- higher agricultural income. Remittances should ing more agricultural agency and become not impact households’ access to social more involved in agricultural and household transfers, as remittances do not contribute decision making. However, improved household to higher overall family income. Food securi- welfare reported among migrant households aris- ty and food diversity could be achieved at a faster es primarily due to remittance flows and decisions pace if women had not only more economic em- about income allocation, rather than to improve- powerment but also more “soft agency.” ments in productivity. At the same time, these wom- 15 References Agarwal, B. 1997. “Bargaining and Gender Rela- ———. 2013. FAOSTAT Database. Accessed at tions within and beyond the Household.” http://faostat.fao.org/. Feminist Economics 3(1). International Food Policy Research Institute (IF- Alkire, S. 2008. Concepts and Measures of Agen- PRI). 2003. Household Decisions, Gender, cy, OPHI Working Papers ophiwp009, and Development.Washington, DC.: IFPRI. Queen Elizabeth House, University of Menjivar, C, and V. Agadjanian. 2007. “Men’s Mi- Oxford. gration and Women’s Lives: Views from Alkire, S. 2005. Why the Capability Approach? Rural Armenia and Guatemala.” Social Journal of Human Development, Volume Science Quarterly 88(5). 6, Issue 1, pages 115-135 Mummert, G. 1988. “Mujeres de migrantes y mu- Alkire, S., Dick R. Meinzen, A. Peterman, A. Qui- jeres migrantes de Michoacán: Nuevos sumbing, G. Seymour, and A. Vaz. 2013. papeles para las se quedan y para las que The Women’s Empowerment in Agriculture se van.” In Movimientos de población en el Index. World Development. occidente de México. Eds. T. Calvo and G. López. El Colegio de Michoacán/CEM- Cohn, D., A. Gonzalez-Barrera, and D. Cudding- CA. ton. 2013. Remittances to Latin America Re- cover—But Not to Mexico. Hispanic Trends Pessar, P. R. 2005. “Women, Gender, and Interna- Project, Pew Research. tional Migration Across and Beyond the Americas: Inequalities and Limited Em- De Schutter, O. 2012. Women’s Rights and the Right powerment.” Expert Group Meeting on to Food. United Nations General Assem- International Migration and Develop- bly A/HRC/22/5. ment in Latin America and the Carib- Deere, Carmen Diana, and Magdalena León de bean. Mexico City, Department of Eco- Leal. 2001. Empowering Women: Land and nomic and Social Affairs, United Nations Property Rights in Latin America. University Secretariat. of Pittsburgh Press. Pittsburgh, PA. Amartya Sen. 1989. “Development as Capability Food and Agriculture Organization of the United Expansion,”  Journal of Development Plan- Nations (FAO). 2011a. The Role of Women ning 19: 41–58. in Agriculture. Agricultural Development Thomas, D. 1990. “Intra-household allocation: An Economics Division. ESA Working Paper Inferential Approach.” The Journal of Hu- 11-02. Rome. man Resources 24(4): 635-664. ———. 2011b. The State of Food and Agriculture: World Bank. 2011. World Development Report 2012: Women in Agriculture: Closing the Gender Gender Equality and Development. Wash- Gap for Development. Rome: FAO. ington, DC: World Bank. 16 Women in Agriculture ———. 2012. Women’s Economic Empowerment in Latin America and the Caribbean. Washing- ton, DC: World Bank. ———. 2014. Migration & Remittances Data. Recent Trends and Outlook: 2013-2016. Washing- ton, DC: World Bank. ———. 2015. World Development Indicators Data- base. “Key Indicators of the Labour Mar- ket Database.” Washington, DC: World Bank. World Food Programme (WFP). 2015. “Guatema- la: Country Page (Overview).” Accessed at https://www.wfp.org/countries/guatemala/ overview. 17 Appendix A: Data Tables Table A1: Descriptive Statistics by Household Type Household Type Migrant Single- Dual-Headed Husband Female-Headed Marital Status of Woman (interviewee)8 Single (%) 0 0 23.40 Married (%) 64.85 75.56 0 Common-law married (%) 35.15 24.44 0 Divorced (%) 0 0 1.42 Separated (%) 0 0 21.99 Widowed (%) 0 0 53.19 Woman’s Age (mean) 35.4 40.4 43.8 Partner’s Age (mean) 38.6 44.8 - Woman’s Literacy Can read and write (%) 68.48 60.15 39.01 Can read or write with difficulty (%) 15.76 13.16 17.73 Not literate 15.76 26.69 43.26 Partner’s Literacy Can read and write (%) 86.06 69.17 - Can read or write with difficulty (%) 3.03 8.27 - Not literate 10.91 22.56 - Woman’s Schooling 9 None/Less than primary (%) 11.52 23.31 36.88 Some primary (%) 51.52 43.98 43.97 Completed primary (%) 28.48 24.44 13.48 Secondary or more (%) 7.87 6.77 5.67 Partner’s Schooling None/Less than primary (%) 10.30 24.44 - Some primary (%) 36.36 39.47 - Completed primary (%) 36.97 26.32 - Secondary or more (%) 8.48 9.03 - 8 9 8 Note that the survey did not make a distinction between de jure and de facto marital status, and instead simply asked women to select an option as they felt fit. 9 May not add up to 100 percent because of “do not know” answers. The same applies to “Partner’s Schooling” below. 18 Women in Agriculture Table A1: Descriptive Statistics by Household Type Household Type Migrant Single- Dual-Headed Husband Female-Headed Household Size 4.5 5.6 5.0 Number of Children (age<=12) in Household 1.6 1.8 1.5 Number of Woman’s Children 10 2.9 4.1 3.7 Dependency Ratio 11 1.18 0.75 0.86 10 11 Table A2: Descriptive Statistics on Agricultural Land Household Type Migrant Single- Dual-Headed Husband Female-Headed Number of plots used, managed, or rented out by the 2.29 2.42 ** 2.04 *** household Average plot size (per plot), m2 4,716 7,060 *** 4,702 Total land (across all plots), m 2 9,403 15,316 *** 8,603 Share of plots used that are owned by household members 57.14% 46.10% *** 50.00% * Households that own at least one plot 76.36% 74.06% 76.60% Total land owned by household (across all plots), m 2 6,507 7,087 4,664 Share of plots owned by woman 19.15% 11.27% *** 32.17% *** Women that own at least one plot 30.30% 21.80% ** 49.65% *** Total land owned by woman (across all plots), m 2 4,334 5,451 4,939 10 Includes all living children of all ages, whether living in the household or elsewhere. 11 The dependency ratio refers to the ratio of the number of household members under age 15 and over age 64 to the number of working-age household members. 19 Appendix B: Additional Agency Information and Results Measuring Agency household in which the woman participates or is the sole decision maker. It is worthwhile to measure agency since it Proxies in this approach include whether the is an intrinsically valuable expression of free- woman participates in/is the sole decision mak- dom and choice, and a pathway to gender er in the household’s expenditure decisions, equality. Because of the complex nature of agen- schooling decisions, etc. or, more comprehen- cy, approaches for measuring it are varied. In the sively, is based on her share of participation in literature, three general proxies exist for measuring various decisions. Nonetheless, sole decision agency, though each has shortcomings: making alone is not a perfect measurement of agency, as female heads of households and 1. Endowments. The amount or share of goods other women may actually prefer to share deci- owned by the woman. sion-making duties with another person. This traditional method considers proxies such 4. Elicited psychosocial measurements. This as the amount of land owned by the woman or method uses questions to elicit women’s per- the land received by the couple as a dowry. But ceptions of their own level of agency or a sim- agency is not only a matter of a woman’s en- ilar notion. dowments and economic opportunities, despite the influence they have on her capacity to exer- A survey question asking a woman how she cise agency. Two individuals with the same en- rates her level of freedom relative to other peo- dowments and economic opportunities do not ple in her household or village, for instance, can necessarily have the same goals or equal ability be a proxy for a woman’s degree of agency in to advance the goals they value and have reason this context. Similarly, women may be asked to value (Alkire 2008). whether they think their opinions are heard or whether they feel capable to do what they set 2. Actions. The woman’s behavior, with assump- out to do. These newer measurements rely on tions about what one’s behavior might be if free the assumption that complex questions are ad- to choose. equately understood. However, individuals may Here, the proxies used include participation in have incorrect perceptions, such that a woman the labor force or a having a lower number of who says she has more freedom relative to an- children. The difficulty with this measurement other may not in reality; her low level of agency is that agency is not equivalent to action and may have led her to expect and accept a lower should not be measured by a list of actions that a level of freedom. third party deems as expressing agency. A wom- In this report, all four approaches are taken into an who does not participate in agricultural la- consideration, building indices that combine more bor, for example, may be exercising her agency than one of them. This offers more holistic mea- in the decision not to work and divide her labor surements of agency and mitigates the shortcom- strategically with her spouse. ings of each measurement by supplementing each 3. Decision-making responsibilities. The with others. Reassuringly, however, positive cor- share or number of decisions pertaining to the relations are found between all of the measure- ments used. 20 Women in Agriculture Further Agency Results ing to increase other sources of income, rather than just agricultural income.This is corroborated by the fact that among Type 3 women, greater levels of Using the calculated agency indices, regres- autonomy are associated with more time spent on sions were estimated to understand the rela- productive activities other than agriculture. tionship between agency and other outcomes of interest. Some of these results are included in A surprising result is the negative correlation Appendix D, but this subsection highlights some between the agricultural agency index and interesting and/or policy-relevant findings. total agricultural income. This might be a sign that women experience significantly less access to Perhaps not surprisingly – but reassuring inputs, or that higher agricultural incomes are asso- for the validity of the measurements – high- ciated with larger production, with more labor and er levels in the soft agency index and in the a lower participation level from women in both de- non-agricultural agency index are associated cisions and actions. with higher income. For Type 1 women, higher levels of agency and particularly aspiration lev- Having a high level of soft agency is correlat- els are associated with higher incomes. ed with higher levels of food security and food diversity in the household. Higher auton- When including each measurement of agency omy seems to explain higher levels of food security, separately, the soft agency index is positively whereas higher self-determination is associated with correlated with better agricultural income. higher levels of food diversity. This is again consis- For Type 1 women, a higher autonomy rating is tent with the literature on the greater tendency of very strongly associated with higher agricultural women to allocate money towards food; women earnings. In contrast, for Type 3 women, a higher with higher levels of agency, who thus feel more ca- autonomy rating may be associated with lower ag- pable to take control of and allocate resources, may ricultural earnings. This might be explained by the be more successful at channeling income towards fact that when these women are more autonomous food expenditures for their household. than the average, they may invest more effort in try- Figure B1: Soft Agency Index by Household Type 10 8 6 4 2 0 Type 1 Type 2 Type 3 Self-Efficacy Aspirations Autonomy 21 Figure B2: Decide on Activities Outside the House 100% 80% 60% 40% 20% 0% Type 1 Type 2** Type 3*** Someone else Together Alone Figure B3: Decide on Household Budget 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3** Someone else Together Alone Figure B4: Decide on Food 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Someone else Together Alone 22 Women in Agriculture Figure B5: Decide on Own Medical Attention 60% 40% 20% 0% Type 1 Type 2*** Type 3* Someone else Together Alone Figure B6: Belong to a Social Group (Excluding Church and Sports) 100% 80% 60% 40% 20% 0% Type 1 Type 2 Type 3 Yes No Figure B7: Woman has a Leadership Role in a Social Group 100% 80% 60% 40% 20% 0% Type 1 Type 2 Type 3 Yes No 23 Figure B8: Woman has a Bank Account Alone 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3*** Yes No Agricultural Decision Making Figure B9: Woman Participates in Decisions on Plantation (Sembrar) 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No Figure B10: Woman Participates in Decisions on Inputs (Insumos) 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No 24 Women in Agriculture Figure B11: Woman Participates in Decisions Regarding Agricultural Production 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No Figure B12: Woman is Responsible for Small Animals 100% 80% 60% 40% 20% 0% Type 1 Type 2 Type 3*** Yes No HH does not have small animals Agricultural Actions Figure B13: Woman Participates in Cultivation 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No 25 Figure B14: Woman Participates in Agricultural Production 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No Figure B15: Woman Was the Respondent for the Agriculture Module 100% 80% 60% 40% 20% 0% Type 1 Type 2*** Type 3 Yes No Figure B16: Woman Owns Agricultural Land 100% 80% 60% 40% 20% 0% Type 1 Type 2 Type 3* Yes No 26 Women in Agriculture Appendix C: Explanation of Variables Agency Variables Non-Agricultural Agency Variables Participation in budget: Measure of woman’s participation in two facets of the household bud- Soft Agency Variables get: deciding on the overall budget and managing Self-efficacy: A composite of answers to two the budget. It is coded as done alone (2), together questions regarding self-efficacy. The first asks the with someone else (1), or by someone else/no par- respondent to choose from four sentences the one ticipation (2). that most describes her situation (e.g., On one ex- Participation in food expenditures: Measure treme, “I always feel free to do whatever I decide of woman’s participation in three facets of food to do,” and on the other, “Almost always what I expenditures: deciding on the overall amount allo- do is not what I would have chosen to do”). The cated towards food, deciding on what food to buy, second question offers an alternative set of four and making the purchase. It is coded as done alone sentences (e.g., “I always choose the way in which (2), together with someone else (1), or by someone I do things,” and at the other end of the spectrum, else/no participation (2). “I never choose for myself the way in which I do things”). The answer to each of the two questions Participation in decisions on own activities is given a score from 0 to 3 that are added, for a outside the household: Measure of woman’s self-efficacy score ranging from 0 (least self-effica- participation in deciding the activities she carries cy) to 6 (most). out outside the household. It is coded as done alone (2), together with someone else (1), or by Aspirations/Self-esteem: A composite of an- someone else/no participation (2). swers to four questions regarding aspirations, all of which ask whether the respondent completely Participation in decisions regarding own disagrees, disagrees, agrees, or completely agrees health care: Measure of woman’s participation in with the description of herself. The statements are: two facets of her own health care: when feeling “Sometimes I think I am not good at anything,” ill, whether to get care and where to get care. It is “I am capable of doing things just as well as most coded as done alone (2), together with someone people,” “I generally do not dare share my ideas,” else (1), or by someone else/no participation (2). and “I think I am capable of fulfilling some of my Participation in non-agricultural decisions: dreams.”The answer to each of the questions is giv- The sum of the previous four variables, ranging en a score from 0 to 3 that are added together, for from 0 to 8. an aspirations score ranging from 0 (fewest aspira- tions) to 12 (most). Social participation: A composite of two indi- cators: whether a woman participates in a group Autonomy: This variable comes directly from in her community (excluding church and sports a question asking respondents to imagine a lad- groups, due to high participation in the latter), der with 10 rungs, “where people with the least and whether she holds a leadership position in any amount of freedom are at the bottom rung and group (including church and sports groups). This people with the most freedom are at the top rung,” variable ranges from 0 (no participation or leader- and to state which rung they believe they are on. ship) to 2 (participation and leadership in at least This question was aided by a visual representation one group). of a ladder. Answers range from 0 (lowest rung) to 10 (highest). Woman has a bank account alone: Single vari- able based on answers to whether anyone in the 27 household has a bank account, and who owns that module, they were first asked if they believed they account. Coded as “no” (0) or “yes” (1). could answer a module on the household’s land use and agricultural production. If not, they could in- Agricultural Agency Variables dicate a different respondent for that module. This variable records whether the woman was the re- Participation in decisions on what to plant: spondent for the agriculture module (1) or not (0). Single variable based on the listing of household members who participate in the decision of what Participation in agricultural actions: The sum to plant for agricultural production. Coded as “no” of the previous three variables, ranging from 0 to 3. (0) if the woman is not listed among the partici- Woman owns agricultural land: Single variable pants or “yes” (1) otherwise. based on the listing of each plot of land used or Participation in the decisions on inputs: Single owned by the household, and the listing of indi- variable based on the listing of household members viduals who own each plot. Coded as “no” (0) if who participate in the decision of which inputs to the woman is not listed as an owner for any of the use in agricultural production. Coded as “no” (0) if plots listed or “yes” (1) if she owns at least one of the woman is not listed among the participants or the plots listed. “yes” (1) otherwise. Participation in decisions on agricultural production: Women were asked whether they Agency Indices participate in the household’s agricultural produc- Soft agency: Principal-component factor using tion decisions or not. Coded as “no” (0) or “yes” the three variables listed in the soft agency variables (1). category above, each rescaled to range from 0 to 1. Responsible for small animals: Single variable Factor analysis considers the correlation between based on the listing of household members who the included variables and creates a composite of are responsible for the small animals owned by them, giving weights according to the correlation the household, by type of animal (rabbits, chicken, matrix. It has a mean of 0 and standard deviation roosters, turkeys, and ducks). Coded as “no” (0) if of 1. the woman is not responsible for any of the small Non-agricultural agency: Principal-component animals owned by the household or “yes” (1) if she factor using the last three variables listed in the is responsible for at least one type of small animal. non-agricultural agency variables category above, Participation in agricultural decisions: The each rescaled to range from 0 to 1. Factor analy- sum of the previous four variables, ranging from sis considers the correlation between the included 0 to 4. variables and creates a composite of them, giving Participation in cultivation: Single variable weights according to the correlation matrix. It has based on the listing of household members who a mean of 0 and standard deviation of 1. participate in crop cultivation. Coded as “no” (0) if Agricultural agency: Principal-component fac- the woman is not listed among the participants or tor using the variables “participation in agricul- “yes” (1) otherwise. tural decisions,” “participation in agricultural ac- Participation in agricultural production: tions,” and “woman owns agricultural land” listed Women were asked whether they participate in the in the agricultural agency variables category above, household’s agricultural production or not. Coded each rescaled to range from 0 to 1. Factor analy- as “no” (0) or “yes” (1). sis considers the correlation between the included variables and creates a composite of them, giving Woman answered agriculture module: In the weights according to the correlation matrix. It has survey implementation, the default protocol was for a mean of 0 and standard deviation of 1. selected interviewees to answer all of the modules in the questionnaire. However, for the agriculture Agency: Principal-component factor using all nine variables included in the three indices above, 28 Women in Agriculture each rescaled to range from 0 to 1. Factor analy- sis considers the correlation between the included variables and creates a composite of them, giving weights according to the correlation matrix. It has a mean of 0 and standard deviation of 1. Other Variables of Interest Dependency ratio: The dependency ratio refers to the ratio of the number of household members under age 15 and over age 64 to the number of working-age household members. Food insecurity: Variable based on a standard set of nine questions measuring household food inse- curity (for an example of the questions, see: http://www.unscn.org/layout/modules/resourc- es/files/Household_food_insecurity_Sp.pdf, p. 6). The scores range from 0 (least food insecurity) to 27 (most food insecurity). Food diversity: Households were asked whether, over the previous 24 hours, anyone in the house- hold consumed vegetables, fruits, meat (beef, chick- en, or pork), fish/seafood, eggs, and milk or milk products. One point was given for each “yes” and zero for “no,” such that this composite score ranges from 0 to 6. Time spent on agriculture: Number of hours the respondent spends working in agriculture “on an average working day.” Time spent on other income-generating ac- tivities: Number of hours the respondent spends “on an average working day” on income-generat- ing activities other than agriculture. 29 Appendix D: Regressions of Interest Table D1: Determinants of Annual Household Agricultural Income from Farming and Animal Husbandry (US$) (1) (2) (3) (4) (5) Type 2 Household 125.5* 109.9 70.17 46.38 97.84 (62.70) (65.43) (63.73) (67.39) (63.50) Type 3 Household -48.95 -21.90 -28.14 -64.43 -16.50 (60.40) (59.85) (58.94) (60.24) (59.67) Soft Agency 56.66* 67.00* (28.34) (29.52) Non-Agricultural Agency -19.09 -16.18 (35.13) (36.14) Agricultural Agency -70.67* -75.55* (30.91) (31.26) Agency Index -33.61 (33.02) Land size (1000m ) 2 52.52*** 53.28*** 51.59*** 51.00*** 52.80*** (9.443) (9.400) (9.330) (9.509) (9.295) Land size (1000m2) squared -0.282** -0.288** -0.275** -0.272** -0.284** (0.104) (0.102) (0.0993) (0.101) (0.101) Woman learned alone how to farm -232.6* -252.7** -233.4* -200.9* -250.6* (96.67) (96.72) (99.32) (99.07) (97.85) Woman does not know how to farm 4.765 -13.62 -65.45 -57.22 -30.22 (115.7) (115.7) (119.3) (118.7) (119.1) Number of crops cultivated 92.88 87.51 101.2* 105.5* 91.31 (48.03) (47.68) (47.46) (49.48) (46.64) Farm labor includes workers 186.7** 188.2** 196.6** 198.9** 189.9** from outside the household (70.16) (70.75) (70.32) (70.36) (70.92) Household owns a plot of land 58.02 53.94 57.22 57.15 53.70 (58.61) (59.06) (59.61) (58.39) (59.32) Household size -25.57* -27.36* -29.40* -27.79* -28.15* (11.80) (12.17) (12.36) (12.08) (12.31) N 572 572 572 572 572 Note: Standard errors in parentheses; * p<0.05, ** p<0.01, *** p<0.001. These regressions also include control variables for distance to market, the dependency ratio, respondent age, and respondent literacy. Coefficients have been omitted here to conserve space. Full results are available upon request. 30 Women in Agriculture Table D2: Determinants of Total Annual Household Income (US$) (1) (2) (3) (4) (5) Type 2 Household -470.4 -163.2 -247.9 -113.6 -156.9 (249.4) (281.4) (277.7) (294.8) (284.2) Type 3 Household -563.3 -464.0 -455.8 -515.7 -511.9 (301.4) (298.0) (299.8) (297.2) (302.4) Soft Agency 201.6* 140.6 (102.4) (102.4) Non-Agricultural Agency 295.9 214.7 (150.9) (159.4) Agricultural Agency 229.7 143.7 (132.8) (136.1) Agency Index 303.7* (143.3) Household owns a plot of land 46.49 93.20 53.99 77.36 72.42 (239.3) (232.6) (239.1) (233.8) (237.3) Time spent in other 330.0*** 319.1*** 349.9*** 321.1*** 335.6*** productive activities (75.34) (73.80) (74.63) (74.27) (74.94) Household size 210.2*** 205.9*** 211.9*** 212.7*** 211.9*** (45.73) (45.28) (45.62) (45.25) (45.47) Dependency ratio -449.4*** -451.2*** -462.2*** -449.8*** -456.2*** (125.3) (122.7) (125.6) (123.9) (125.1) N 503 503 503 503 503 Note: * p<0.05, ** p<0.01, *** p<0.001. These regressions also include control variables for distance to market, number of crops, size of agricultural land, size of agricultural land squared, respondent age, and respondent literacy. Coefficients have been omitted here to conserve space. Full results are available upon request. 31 Table D3: Determinants of Food Insecurity (1) (2) (3) (4) (5) Type 2 Household 1.228 1.278 2.762** 2.839** 1.981* (0.875) (0.947) (0.950) (0.962) (0.958) Type 3 Household 2.465** 2.020* 2.256** 2.805** 1.957* (0.898) (0.874) (0.848) (0.869) (0.865) Soft Agency -1.016** -1.190*** (0.335) (0.326) Non-Agricultural Agency 0.0759 -0.177 (0.365) (0.372) Agricultural Agency 1.587*** 1.808*** (0.351) (0.368) Agency Index 0.773* (0.368) Number of crops cultivated 0.358 0.436 0.167 0.0329 0.380 (0.486) (0.488) (0.488) (0.490) (0.488) Time spent in other -0.462* -0.539** -0.488** -0.383* -0.556** productive activities (0.189) (0.191) (0.184) (0.189) (0.185) Household receives remittances -0.933 -0.927 -0.792 -0.766 -0.917 (0.696) (0.705) (0.677) (0.679) (0.692) Household owns a plot of land -0.336 -0.331 -0.364 -0.383 -0.292 (0.734) (0.747) (0.722) (0.707) (0.741) Household size 0.119 0.146 0.201 0.175 0.169 (0.139) (0.137) (0.135) (0.138) (0.136) N 555 555 555 555 555 Note: * p<0.05, ** p<0.01, *** p<0.001. These regressions also include control variables for size of agricultur- al land, size of agricultural land squared, dependency ratio, respondent, and respondent literacy. Coefficients have been omitted here to conserve space. Full results are available upon request. Note that a higher dependent variable indicates higher food insecurity. 32 Women in Agriculture Table D4: Determinants of Food Diversity (1) (2) (3) (4) (5) Type 2 Household -0.167 -0.130 -0.183 -0.243 -0.0560 (0.202) (0.216) (0.225) (0.226) (0.225) Type 3 Household -0.365 -0.230 -0.236 -0.380 -0.240 (0.210) (0.210) (0.211) (0.211) (0.210) Soft Agency 0.298*** 0.308*** (0.0706) (0.0711) Non-Agricultural Agency 0.0343 -0.0156 (0.0778) (0.0768) Agricultural Agency -0.0225 -0.0623 (0.0797) (0.0803) Agency Index 0.105 (0.0855) Number of crops cultivated 0.0968 0.0785 0.0796 0.107 0.0691 (0.105) (0.108) (0.108) (0.106) (0.107) Time spent in other 0.0307 0.0482 0.0506 0.0296 0.0481 productive activities (0.0430) (0.0421) (0.0419) (0.0432) (0.0420) Household receives remittances 0.601*** 0.588*** 0.593*** 0.600*** 0.594*** (0.157) (0.161) (0.160) (0.159) (0.159) Household owns a plot of land 0.0244 0.0287 0.0255 0.0238 0.0315 (0.162) (0.166) (0.166) (0.162) (0.166) Household size -0.0935** -0.101*** -0.102*** -0.0956*** -0.0983*** (0.0285) (0.0290) (0.0290) (0.0285) (0.0289) N 555 555 555 555 555 Note: * p<0.05, ** p<0.01, *** p<0.001. These regressions also include control variables for size of agricultural land, size of agricultural land squared, dependency ratio, respondent age, and respondent literacy. Coefficients have been omitted here to conserve space. Full results are available upon request. 33