Women's Empowerment in Agriculture and Dietary Quality Across the Life Course: Evidence from Bangladesh

Using nationally representative survey data from rural Bangladesh, this paper examines the relationship between women’s empowerment in agriculture and indicators of individual dietary quality. Our findings suggest that women’s empowerment is associated with better dietary quality for individuals within the household, with varying effects across the life course. Women’s empowerment is associated with more diverse diets for children younger than five years, but empowerment measures are not consistently associated with increases in nutrient intake for this age group. Women’s empowerment is positively and significantly associated with adult men’s and women’s dietary diversity and nutrient intakes. Different empowerment domains may have different impacts on nutrition, but other characteristics, such as maternal schooling and household socioeconomic status, may play a more important role for younger children. The importance of maternal education in the dietary quality of young children, and the relatively greater importance of women’s empowerment for older children and adults, imply that policies designed to empower women and improve nutritional status should be informed by knowledge of which specific domains of women’s empowerment matter for particular nutritional outcomes at specific stages of the life course.

iv Tables   Table 3.1 Summary statistics for individual outcomes and characteristics 12 Table 3.2 The five domains of empowerment in the Women's Empowerment in Agriculture Index 16 Table 3.3 Summary statistics for household characteristics 19 Table 4.1 Women's empowerment and dietary quality of children younger than five: Selected coefficients from levels and fixed-effects estimates 22 Table 4.2 Women's empowerment and dietary quality of children ages 5-10 years: Selected coefficients from levels and fixed-effects estimates 27 Table 4.3 Women's empowerment and dietary quality of children ages 11-17 years: Selected coefficients from levels and fixed-effects estimates 31 Table 4.4 Women's empowerment and dietary quality of adults: Selected coefficients from levels and fixed-effects estimates 35 Table 4.5 Results summary for education, landownership, and dairy cow ownership and joint tests for coefficients from Model 1 (aggregate empowerment score) 39

INTRODUCTION
Recognition that sufficient dietary energy does not guarantee the consumption of enough protein and micronutrients to lead an active and healthy life has increased attention to improving dietary quality, particularly within low-income countries that are at risk of widespread food insecurity and micronutrient deficiencies. One such country is Bangladesh, where, despite increases in food production and reduction of child stunting in the past two decades, the rate of wasting (14 percent) remains alarmingly high (NIPORT 2015) and deficiencies in micronutrients such as iron, zinc, iodine, and vitamin A are prevalent, affecting young children and women of reproductive age, who are most vulnerable to the negative effects of micronutrient malnutrition. Anemia (in part due to iron deficiency) is estimated to affect 26 percent of nonpregnant, nonlactating women, whereas 42 percent suffer from iodine deficiency and 5.4 percent from vitamin A deficiency. In preschool-age children, the rates are 33 percent, 40 percent, and 20 percent for anemia, iodine deficiency, and vitamin A deficiency, respectively (ICDDR,B et al. 2013). In Bangladesh, such deficiencies reflect poor diets that are rice dominated and lacking in diversity ), together with high infection rates due to poor hygiene and sanitation (UNICEF and WHO 2015).
Micronutrient deficiencies are linked to a wide range of adverse health outcomes, such as growth restriction, impaired cognition, and increased morbidity and mortality (Black et al. 2008). Deficiency in vitamin A, a major factor in reducing excess mortality from infectious diseases in developing countries, is the most common cause of preventable childhood blindness. Zinc, together with other micronutrients, has a role in preventing both diarrhea and respiratory diseases (Darnton-Hill et al. 2005). The evidence that individual dietary diversity is strongly linked with nutritional outcomes in women and children (Rah et al. 2010; Ruel and Menon 2002;Arimond and Ruel 2004;Arimond et al. 2010) has led to efforts to increase dietary diversity through the design and implementation of nutrition-sensitive agricultural programs.
A key component of nutrition-sensitive agricultural programs is explicit recognition of women's role in their families' nutrition (Ruel, Alderman, and Maternal and Child Nutrition Study Group 2013).
Because women are primarily responsible for food preparation and distribution in the household in rural Electronic copy available at: https://ssrn.com/abstract=3132731 Bangladesh, policy interventions targeted toward improving women's status are often expected to contribute to the well-being of all members of the household, including women themselves. Previous studies examining women's role in agriculture in Bangladesh have found that women's education improves household protein intake and dietary diversity (Rashid, Smith, and Rahman 2011) and that women's empowerment in agriculture is associated with greater availability and diversity of food consumed at the household level ). However, the relationship between women's empowerment and the dietary quality of individuals has not yet been explored.
Using nationally representative data from the 2011-2012 Bangladesh Integrated Household Survey, or BIHS (Ahmed 2013), this paper examines the relationship between women's empowerment in agriculture and the dietary quality of individuals within the household. Although a number of studies have looked at determinants of dietary quality at the household level (Rashid, Smith, and Rahman 2011; and at under-five child and maternal dietary diversity in Bangladesh (Bhagowalia et al. 2012;Nguyen et al. 2013), this paper is a first attempt to link women's empowerment to the dietary quality of individual household members across the life course. Our indicators of dietary quality are the dietary diversity score (the number of food groups consumed) as well as individual intakes of macronutrients (calories and protein) and micronutrients (iron, zinc, and vitamin A). Because dietary requirements vary across the life cycle, we look at the dietary quality of four age groups: (1) children ages 6 months to 5 years, excluding breastfed children; (2) children between 5 and 10 years of age; (3) adolescents between 11 and 17 years of age; and (4) adults (ages 18 years and older). Given the welldocumented patterns of gender disparities in education, asset ownership, incomes, and nutritional status in South Asia (Quisumbing and Maluccio 2003;Ahmed and Maitra 2010;Miller 1981), we also examine whether women's empowerment has a differential impact on the dietary quality of individuals in these age groups according to the sex of the individual.
Our measure of women's empowerment in agriculture is a recently developed survey-based index, the Women's Empowerment in Agriculture Index (WEAI), which assesses women's empowerment in agriculture in five domains: decision making over production, ownership of and decisions over Electronic copy available at: https://ssrn.com/abstract=3132731 resources, control over use of income, leadership in the community, and time use (Alkire et al. 2013). The WEAI collects the same information for both the primary male and the primary female decision makers in the household, which can be used to compare the gender gap in empowerment within the household. This paper uses four measures of women's empowerment: the aggregate women's empowerment score, based on the five domains of empowerment in agriculture; two individual indicators in which women are most disempowered in rural Bangladesh-women's participation in groups and women's rights over assets (the extent to which a woman participates in decision making over the purchase, sale, and transfer of assets); and the gender gap in empowerment, that is, the difference between the empowerment scores of the primary male and female decision makers.
Overall, findings suggest that women's empowerment has different effects on dietary diversity and nutrient intakes across the life course. Women's empowerment appears to be weakly correlated with measures of dietary quality in the youngest children, ages 6-59 months, but strongly correlated with the dietary quality of older children, adolescents, and adults. Women's empowerment has differential effects on males and females within the household, with the number of groups to which the primary female belongs having a differential positive association with the dietary diversity of girls ages 5-10 and the nutrient intakes of adult women themselves. The findings indicate that the strength of the association between women's empowerment and the dietary quality of different individuals may vary across the life course, a fact that has implications for the design and targeting of interventions to improve the dietary quality of individuals, particularly women and children, in rural Bangladesh.
Electronic copy available at: https://ssrn.com/abstract=3132731 cash transfers (Yoong, Rabinovich, and Diepeveen 2012)-has found that these programs improve children's well-being, especially through investments in children's health and education, although none of the studies reviewed explored the counterfactual of targeting resources to men. However, whether programs that aim to empower women have a measurable impact on nutritional outcomes is not established. The findings of van den Bold, Quisumbing, and Gillespie's (2013) review of the evidence on three types of programs that aim to empower women-agricultural programs, microcredit programs, and cash transfer programs-did not overwhelmingly point toward positive or negative impacts on women's empowerment or nutrition. The authors identified significant gaps in the research, owing to evaluations that are not well designed and do not pay explicit attention to gender and empowerment, the concentration of studies in specific geographical regions, and the lack of explicit attention to the programs' "pathways of impact." More recent efforts to examine the relationship between women's empowerment in agriculture and nutritional outcomes draw from two strands of the literature. One strand is the emerging literature on agriculture-nutrition pathways (Kadiyala et al. 2014;Ruel, Alderman, and Maternal and Child Nutrition Study Group 2013), which highlights linkages between agriculture and nutrition as well as the important role of women in agriculture and nutrition, particularly through the effects of women's social status and empowerment on their access to and control over resources, and the impacts of women's participation in agriculture on their time allocation and their own health and nutritional status. The other strand of the literature draws on Kabeer's (1999) definition of empowerment as an expansion of people's ability to make strategic life choices, particularly in contexts in which this ability had been denied to them. The WEAI, inspired by Kabeer's definition, attempts to directly measure women's empowerment in agriculture, emphasizing agency and choice in both the productive and the reproductive spheres, not just control of resources or decision making over household expenditures (Alkire et al. 2013). Several studies have used the WEAI in the context of the agriculture-nutrition pathways to examine the association between women's empowerment in agriculture and household and individual welfare.  found that women in rural Ghana were most disempowered in terms of access to and Electronic copy available at: https://ssrn.com/abstract=3132731 decision making over credit, and that improvement in this domain was associated with greater dietary diversity for women and female children. In Nepal, overall empowerment of women, their membership in collective action groups, their greater control over income, and their hours worked were significantly associated with both maternal dietary diversity and body mass index. Sraboni and colleagues (2014) found that women's active membership in collective-action groups, their control over agricultural assets, and a narrowing gap in empowerment between men and women within households were positively associated with calorie availability and dietary diversity at the household level. The Ghana and Nepal studies, which analyzed both maternal and child outcomes, suggested that the domains of empowerment that are significant for women's and children's diet and nutrition outcomes may not always overlap, and pointed to the need to examine the impacts of women's empowerment on individuals across the life course.
This paper, by examining the association between various indicators of women's empowerment and the dietary quality of household members, attempts to fill this lacuna in knowledge about how women's agency and choice in the productive and reproductive spheres may influence the nutritional well-being of individuals in their households across the life course. Moreover, we seek to understand whether women's empowerment has a differential impact on the dietary quality of household members in various age groups, according to the sex of the individual, a contribution that is particularly important in the context of countries with well-documented gender disparities in resource allocation.

DATA, EMPIRICAL SPECIFICATION, AND VARIABLES Data
We use data from the first round of the BIHS, conducted by the International Food Policy Research Institute between December 2011 and March 2012 (Ahmed 2013), and nationally representative of rural Bangladesh. The survey includes a wide range of information on household demographics, education, employment, housing and assets, food and nonfood consumption and expenditures, agricultural production, and livestock holdings, well as separate WEAI modules for the self-identified primary male and female decision makers, interviewed separately and in private. The BIHS dataset consists of 1,608 nonfarm and 3,895 farm households; however, because our focus is on women's agency in the agricultural sector, we restrict our analysis to farm households (those that are involved in one or more agricultural activities-growing crops, or raising or engaging with livestock, poultry, or fisheries) to reduce the probability of individuals' being misclassified as disempowered when they do not participate in any agricultural activity. This procedure results in 1,608 observations' being dropped. Additionally, 325 observations are dropped because the primary female respondent was unavailable on the day of the interview, 227 because a female other than the primary female was interviewed, 250 more because of missing values for the empowerment measures, and an additional 197 because of data entry errors in the individual demographic and food consumption data. The final estimation sample consists of 2,896 households with 7,506 adults, 1,786 children ages 11-17 years, 2,015 children ages 5-10 years, and 1,024 children younger than 5 years. Analyses of the empowerment gap between men and women in the same households are conducted for a subsample of 2,851 dual-adult households.

Empirical Specification
To analyze the relationship between measures of individual dietary quality and women's empowerment, we estimate the following equation: where bi is the set of parameters to be estimated, and ν is an error term. For males, the relationship between women's empowerment and dietary quality is given by b2. For females, the relationship is given by (b2 + b3). If b3, the coefficient on the interaction term between empowerment and the female dummy, is significantly different from 0, this suggests that women's empowerment has differential effects on males and females. The reasoning is analogous for the coefficients on the other household characteristics, b5, and their interaction terms, b6.
One possible source of bias in this analysis is the endogeneity of the empowerment measures, which may be affected by the same factors that influence dietary quality. An instrumental variables technique would have been useful in correcting for endogeneity, but owing to a lack of suitable instruments, we estimate equation (2) using ordinary least squares (OLS) and interpret the results as associations rather than causal relationships. In addition, we estimate (2) with household fixed effects, 1 which control for household-level unobservables that may confound the impact of empowerment on dietary quality outcomes. Estimating (2) with fixed effects does not permit us to directly estimate the impact of women's empowerment on these outcomes, because empowerment is measured for a single woman in the household, the primary female, and does not vary within households. However, because our specification interacts empowerment with the sex of the individual, we are able to estimate the differential impact of empowerment on individual dietary quality insofar as its effects on individuals vary depending on their sex.

Outcome Variables
BIHS is the first nationally representative household survey in Bangladesh that collected data on intrahousehold food consumption. The analysis presented here is based on individual food intake data, collected in the dietary intake module of the household survey, using a combination of 24-hour food recall and food weighing methods to estimate consumption quantities of various food items by individual household members. This information was used to construct the dietary diversity score and individual intakes of the following nutrients: food energy in terms of calories, protein, vitamin A, iron, and zinc. 1 We choose households with at least 1 child of each sex so that the gender dummies are relevant in the household fixed-effects specification. This choice results in a drop in the sample size so that, for the fixed-effects models, there are 7,433 adults, 531 adolescents, 583 children ages 5-10, and 102 children younger than 5 years.
Appropriate conversion factors were used to calculate quantities of nutrients contained in foods eaten by individual household members.

Dietary Diversity for Four Age Groups
We use individual dietary diversity scores, which aim to reflect the nutrient adequacy of individuals, rather than household dietary diversity measures, which aim to reflect the economic ability of a household to access a variety of food groups (Hoddinott and Yohannes 2002). The individual dietary diversity scores have been validated for several age and sex groups as proxies for macro-and micronutrient adequacies (Hatloy, Torheim, and Oshaug 1998;Kennedy et al. 2007;Arimond et al. 2010), although not all of these scores have been validated for Bangladesh. There is currently no international consensus on which food groups to include in the scores at the individual level for different age or sex groups (Kennedy, Ballard, and Dop 2011). Individual age-specific dietary diversity scores are based on guidelines from the World Health Organization (WHO) and the Food and Agriculture Organization of the United Nations, with a 7food-group score for children ages 6-59 months and 5-10 years, and a 9-food-group score for children ages 11-17 years and adults age 18 and older. The two different sets of dietary diversity scores reflect the differences in micronutrient requirements across the life course (Ruel, Deitchler, and Arimond 2010;Arimond et al. 2010).

Seven food groups for children ages 5-10 years and 6-59 months.
Following WHO guidelines (WHO 2010), dietary diversity for these two age groups is measured as the number of food groups consumed during the last 24 hours out of 7 food groups: (1) cereals and tubers, (2) legumes and nuts, (3) dairy products, (4) flesh foods, (5) eggs, (6) vitamin A-rich fruits and vegetables, and (7) other fruits and vegetables.

Nutrient Intake
Individual nutrient intake measures include (1) food energy in kilocalories, (2) proteins in grams, (3) iron in milligrams, (4) zinc in milligrams, and (5) vitamin A in retinol activity equivalents. Summary statistics for the outcome variables, presented separately for males and females, are in Table 3.1. The low dietary diversity of individuals is noteworthy; for example, female children ages 5-10 years and those less than 5 years of age consume, on average, 3.87 and 3.57 groups, respectively, out of the recommended 7 groups; out of 9 food groups, male adults and all children ages 11-17 consume, on average, 4.18 and 4.23 groups, respectively. Gender differences in nutrient intake increase over the life course. Among children between 6 and 59 months old, there is no significant difference in dietary diversity scores nor in nutrient intakes, except for vitamin A intake, which is higher for boys (but this difference is only weakly significant). Boys and girls also do not differ significantly in terms of age, maternal age, or maternal schooling. Among children 5-10 years of age, despite an insignificant difference in the dietary diversity score, boys have significantly higher intakes of calories (p < 0.10), protein (p < 0.05), iron (p < 0.10), and zinc (p < 0.05).
Gender differences in allocations between boys and girls become more obvious in adolescence, with boys having significantly higher intakes of calories, protein, iron, and zinc. This may reflect both a combination of higher caloric needs of adolescent boys associated with greater involvement in agricultural labor and social factors discriminating against adolescent girls, who have higher biological needs for iron. Girls have higher vitamin A intakes, and this difference is statistically significant. This finding may be because vegetables, a source of vitamin A, are considered a lower-status food compared with protein and iron sources, such as fish, meat, and eggs, which tend to be preferentially allocated to 2 We test the robustness of the 9-food-group dietary diversity measure on a subsample of women of reproductive age (15-49 years old) by comparing OLS estimates for the 9-food-group score with those for the recently developed minimum dietary diversity score for women, or MDD-W (FAO and FHI 360 2016). Results using the MDD-W (available on request) are not substantively different from those using the dietary diversity score.
males (Gittelsohn and Vastine 2003;Sudo et al. 2004). Dietary diversity scores, however, are not significantly different between adolescent boys and girls. In contrast, adult men have a significantly more diverse diet than adult women, and men consume significantly more nutrients, across all types of macroand micronutrients, than women. Although men engaged in farming may have higher energy requirements, women have a greater biological need for micronutrients during their reproductive years.
Reflecting typical age differences at marriage, men in the sample are also significantly older than women and have completed more years of schooling.   indicators that make up the 5 domains of empowerment in agriculture. This measure is increasing in empowerment, ranges from 0 to 1, and reflects overall empowerment in agriculture.

Model 2. Number of groups in which the woman is an active member (leadership domain, group membership indicator):
the total number of groups in which the female respondent reports being an active member. Active membership in more groups implies a wider network of social ties and potentially greater empowerment. For example, Sanyal (2009) found that participation in microfinance groups often encouraged Indian women to participate more extensively in collective action such as mobilizing in response to domestic violence against women, annulling underage marriages, or acquiring public goods.

Model 3. Number of sole or joint decisions made by the woman concerning the purchase, sale, or transfer of assets (resources domain, rights over assets indicator)
: the total number of decisions made solely or jointly by the female respondent, summed over all asset types. For each asset type, the survey asks who can decide whether to sell, give away, mortgage/rent, or purchase the asset. As mentioned earlier, the existing literature has documented extensively that women who have greater control over household assets are more likely to direct resources toward household welfare.
Model 4. Gender parity gap: an indicator of women's relative empowerment within the household.
Considerable evidence exists in support of the need to pay attention to intrahousehold gender inequality for attaining development objectives Haddad, Hoddinott, and Alderman 1997).
Therefore, the fourth model employs the gender parity gap. According to Alkire and others (2013), a household enjoys parity if the woman's empowerment score is greater than or equal to that of the male in her household. Thus, the gender parity gap is 0 if the household enjoys gender parity. Otherwise, the gap equals the difference between the male and female aggregate empowerment scores. Higher numbers indicate a larger gap between male and female empowerment, favoring males. As mentioned above, Model 4 is estimated only for dual-adult households with valid responses from the primary male and female adults.

Other Independent Variables
Other independent variables include the household characteristics of household size and dependency ratio. The dependency ratio is defined as the ratio of the number of members between 1 and 10 years old plus those older than 64 to those between ages 10 and 64; a greater number of nonearning individuals in the household might be associated with constraints on food availability for each member. We also include the price of two staple food items, rice and pulses, and three protein sources, chicken, small fish, and large fish, as control variables, because food prices have previously been shown to exert a major influence on the consumption pattern of households and individuals (Rashid, Smith, and Rahman 2011;Villa, Barrett, and Just 2011). Ahmed and others (2013), for example, found that rice accounts for a fifth of all spending and 35 percent of the food expenditure of an average rural household in Bangladesh. The following variables are used as indicators of the socioeconomic status of the household: the amount, in decimals (100 decimals = 1 acre), of cultivable land owned by the household; 4 the number of dairy cows owned; a dummy for whether the household has access to electricity; and dummies indicating whether it owns a sanitary latrine and a hand-pumped tube well. Individual characteristics include age, age squared, a dummy variable for sex (= 1 if female), and own years of schooling for adults. The survey is unable to link adolescents and children ages 5-10 to their biological mothers, unless they are children of the household head. For these two age groups, we use the primary female respondent's age and education as proxies for their biological mother's characteristics. To control for any differences between children whose parents are more accurately measured (that is, the children of household heads) compared with other children, we use a dummy variable that equals 1 if the child is the offspring of the household head. For children younger than five, who can be matched to their biological mothers, we use their mother's age and years of schooling. We use years of schooling of different individuals in the age groupspecific regressions because different decision makers within the household decide on food and nutrition over the life course; this is typically the mother for the youngest children, but for adults, it would be the individual him-or herself. Division dummies for each of the seven administrative divisions in Bangladesh are included to control for location-specific effects. Summary statistics of the household characteristics, including the primary female's empowerment indicators, are presented in Table 3.3.

Women's Empowerment and Dietary Quality
Selected coefficients from levels and fixed-effects estimates for dietary quality outcomes of the four age groups are presented in Tables Table 4.1 examines the relationship between the empowerment indicators of the primary female (who is, in most cases, the mother of the specific child) and child dietary quality, controlling for child characteristics, the child's mother's characteristics, and a dummy variable indicating whether the child is the biological offspring of the primary female. 5 Recall that the effect of empowerment on boys is given by the coefficient on the empowerment indicator, whereas the effect on girls is the sum of the coefficient on the empowerment indicator and the interaction term between empowerment and the female dummy variable. The interaction term captures whether the empowerment indicator has a differential effect on females.    dietary diversity for girls. None of the coefficients using the indicator for women's group membership are significant, but the number of asset decisions is significantly associated with improvements in boys' dietary quality, but not that of girls. The coefficient on the gender parity gap is negative and significant for some outcomes, indicating that a reduction in the gap in empowerment between men and women is associated with improvement in the calorie, protein, and zinc intakes of boys. Reducing the gender parity gap is also associated with a significant increase in the calorie intake of girls (p < 0.05) and a weakly significant increase in girls' protein intakes. Intrahousehold differences in the effects of empowerment by sex are mostly insignificant, except for a negative impact of the number of asset decisions on girls in the levels estimates (which is not robust to the inclusion of fixed effects) and a negative association of overall empowerment on boys' vitamin A intake in the fixed-effects estimates.
These results suggest that, for the youngest age group, women's overall empowerment and women's rights over assets are positively and significantly associated with the dietary diversity of all children younger than five years. This finding is consistent with evidence that women's control over assets is important for child outcomes (Shroff et al. 2009). None of the other empowerment indicators have any significant impact on the dietary diversity of younger children, although reducing the gender parity gap improves intakes of calories for both boys and girls, and weakly improves protein and zinc intakes for boys. The literature on infant and young child nutrition may provide a possible explanation for why female empowerment, by itself, may not be sufficient to improve the dietary diversity of children in the younger age groups. Although several studies assert the importance of proper infant and young child feeding (IYCF) practices in improving the dietary diversity of young children (Saha et al. 2008;Zongrone, Winskell, and Menon 2012), the low level of IYCF knowledge and practices among mothers in Bangladesh could underlie the high levels of undernutrition among children (Hackett et al. 2012;Rasheed et al. 2011). IYCF knowledge is not necessarily gained with increased empowerment in agriculture; hence, empowerment may not, on its own, be sufficient to improve the dietary quality of the youngest members in the household, whose feeding patterns are markedly different from those of adults and older children. Table 4.2 presents summary results from levels and fixed-effects estimates for children ages 5-10 years. In contrast to the results for younger children, where women's empowerment had few associations with dietary quality outcomes, we see that, in the levels estimates, women's overall empowerment is associated with greater dietary diversity for both boys and girls, and improvements in a range of intake indicators (for calories, protein, iron, and zinc). These improvements are significant for boys, significant at conventional levels (p < 0.05) for iron intake for girls, and only weakly significant for calories, protein, and zinc for girls. The primary female's participation in groups is significantly correlated with a higher dietary diversity score for boys. The number of asset decisions made also has a significant positive correlation with both boys' and girls' dietary diversity, as well as girls' protein intake. Finally, a reduction in the gender parity gap improves a range of boys' dietary quality indicators (dietary diversity as well as intakes of calories, protein, iron, and zinc), but improves only the overall dietary diversity score for girls.
In terms of differential effects by sex, although levels estimates suggest that the number of groups in which the primary female participates has differential positive associations with girls' calorie and vitamin A intakes, these are not robust to the inclusion of fixed effects. The only interaction term that is significant in the fixed-effects estimates is that on the number of groups, which suggests that the primary female's belonging to more groups has a significant correlation with improved dietary diversity for girls. Electronic copy available at: https://ssrn.com/abstract=3132731 Relationships between women's empowerment indicators and the dietary quality of adolescents suggest the emergence of strong preferences for adolescent boys rather than girls (Table 4.3). The overall women's empowerment score is associated with higher calorie, protein, iron, and zinc intakes for boys, the number of groups is weakly associated with higher iron intake for boys, and a reduction in the gender parity gap is correlated with higher intakes of calories, iron, and zinc for boys. None of the empowerment indicators has a significant association with any measure of girls' dietary quality, as revealed by the Ftests on the sums of coefficients on the empowerment indicator and the interaction term. A closer look at the interaction terms with the sex of the individual in the levels estimates suggests that overall empowerment is weakly associated with lower iron intake for girls, whereas a reduction in the gender parity gap is associated with better protein, iron, and zinc intakes for boys. However, these significant interaction terms seem to be driven by household-level unobservables: the only fixed-effects estimate showing a significant differential association by sex suggests that a reduction in the gender parity gap differentially improves boys' vitamin A intakes. The findings suggesting that women's empowerment may improve adolescent boys'-but not girls'-dietary quality has worrisome implications. Given the prevailing marriage pattern in which girls leave the household to marry during adolescence, the apparent absence of a link between women's empowerment and investment in adolescent girls' nutritional status suggests that a son preference persists, with serious consequences for girls' own health and that of their offspring. Finally, Table 4.4 presents summary results from levels and fixed-effects estimates of the relationship between measures of women's empowerment and adult dietary quality. We find the strongest evidence of positive relationships between women's empowerment and dietary quality in this age group. 6

Table 4.3 Women's empowerment and dietary quality of children ages 11-17 years: Selected coefficients from levels and fixed-effects estimates
All indicators of women's empowerment (the overall empowerment score, the number of groups in which the primary female participates, the number of asset decisions in which she participates, and the gender parity gap) are associated with improvements in both men's and women's dietary quality across almost all dietary quality outcomes in the levels estimates. Examination of the interaction terms reveals that different indicators of empowerment have differential impacts by sex. In the levels estimates, the overall empowerment score is associated with a differential positive effect on men's zinc intakes, but this association is not robust to the inclusion of household fixed effects. In contrast, the fixed-effects estimates suggest a weak positive association between the aggregate women's empowerment score and women's vitamin A intakes. Interaction terms with the number of groups to which the primary female belongs are significant only in the fixed-effects estimates, suggesting that participation in groups has differential positive associations with women's nutrient intakes across macroand micronutrients. Except for a differential improvement in women's dietary diversity when the gender parity gap is lower, however, interaction terms with the gender parity gap suggest that men's nutrient intakes improve more than women's in households with a lower gender parity gap. Thus, although empowerment in terms of group membership appears to benefit women, it is possible that a reduction in the gender parity gap may involve some trade-offs or accommodations that favor men, for example in terms of their nutrient intakes. The positive association of women's empowerment with adult women's outcomes, but not adolescent girls', is a puzzle that deserves further investigation. 6 These results are similar to the results of studies using the WEAI in Nepal  and Ghana  that find women's aggregate empowerment score to be positively and significantly correlated with women's dietary diversity score. The Nepal study found significant associations with women's outcomes but weaker associations with child outcomes.  Electronic copy available at: https://ssrn.com/abstract=3132731   Electronic copy available at: https://ssrn.com/abstract=3132731

Other Correlates of Dietary Quality
In contrast to the insignificant associations of women's empowerment with most dietary quality outcomes among girls ages 6-59 months, maternal schooling has significant positive associations with the dietary diversity of both boys and girls, and with girls' intakes of protein, iron, and zinc. However, because the fixed-effects coefficients are not significant, we cannot reject the null hypothesis that the effects of maternal schooling on boys and girls are equal. The size of land owned has a weak positive correlation with the dietary diversity of both boys and girls, whereas the number of dairy cows owned has a weak negative association with the vitamin A intake of girls, but the latter is not robust to the inclusion of fixed effects. Joint tests of the coefficients on the gender interactions show that (1) intakes of vitamin A appear to be correlated with gender-differentiated impacts of household wealth and with prices, and (2) there may be variations in gender norms across divisions that affect intakes of calories, protein, zinc, and vitamin A for the youngest age group.
The schooling of the primary female improves some dietary quality indicators for boys ages 5-10 (dietary diversity, protein, and iron) as well as dietary diversity and iron for girls. None of the interaction terms with child sex are significant, indicating that the primary female's schooling affects boys and girls ages 5-10 similarly. Interestingly, the size of land owned is associated with more diverse diets and increased calorie, protein, and iron intakes for boys, but only greater dietary diversity for girls. There are no differential effects of landownership by sex of the individual. Household ownership of dairy cows improves dietary quality for boys and calorie intake for girls; however, these results are not robust to the inclusion of fixed effects. Fixed-effects estimates suggest compensatory effects: ownership of dairy cows differentially improves girls' dietary diversity score, but also worsens zinc intakes (albeit only at p < 0.10). There is some indication that prices have differential effects according to child sex.
Maternal schooling appears to significantly improve the dietary quality of adolescent boys and marginally improve the protein intake of adolescent girls. Maternal schooling appears to differentially favor boys, as shown by the sign of the interaction term, but this result is not robust to the inclusion of fixed effects. Very interestingly, the size of land owned is positively correlated with adolescent boys' dietary diversity and intake of nutrients, but it is significantly correlated with adolescent girls' vitamin A intakes. Although levels estimates suggest that boys in households with larger sizes of land owned have higher iron intakes, this effect disappears in the fixed-effects estimates. Dairy cow ownership is associated with higher calorie intake for adolescent boys and better dietary diversity and intakes of calories and protein for adolescent girls. None of the household characteristics, prices, or division dummies indicate that differential effects by child sex are jointly significant.
Finally, in the regressions for adults, one's own schooling is associated with better dietary quality outcomes (although not all coefficients are robust to the inclusion of household fixed effects). The levels estimates suggest that the size of land owned is positively correlated with men's dietary diversity and iron intake as well as with women's dietary diversity score. Fixed-effects estimates also suggest that larger land sizes have a differential positive impact on women's nutrient intakes, except for vitamin A. Dairy cow ownership is also associated with improved dietary quality for both men and women.

CONCLUSION
Previous studies on associations with women's empowerment in agriculture in rural Bangladesh have documented a positive association with household food security outcomes, such as per capita calorie intake and household dietary diversity (that is, the number of food groups consumed at the household level). Our findings suggest that women's empowerment in agriculture is similarly associated with better dietary quality for individuals within the household, but these effects vary across the life course. Although This study has several limitations. First, the analysis is based on 24-hour recall collected during one visit, which may not give as accurate a picture of dietary patterns as would multiple visits. Second, because we are unable to account for endogeneity of empowerment, owing to the lack of suitable instruments, our conclusions are indicative of associations, not causation. Third, even if household fixedeffects estimates control for household-level unobservables, the reduction in sample size owing to the need to restrict the estimation sample to families with at least one child of each sex could have reduced its statistical power, making our conclusions more conservative than they would be without the restriction.7 Fourth, we did not explore any other domains beyond those identified as areas where empowerment gaps were largest. 7 Re-estimation with household fixed-effects and without the restriction of having at least one child of each sex increased the sample size, with no substantive change in results. These results are available upon request.       Source: Estimated using data from the IFPRI Bangladesh Integrated Household Survey, 2011-2012 (Ahmed 2013). Note: Standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.