Context and measurement: An analysis of the relationship between intrahousehold decision making and autonomy

World Development 111 (2018) 97–112

Contents lists available at ScienceDirect

World Development journal homepage: www.elsevier.com/locate/worlddev

Context and measurement: An analysis of the relationship between intrahousehold decision making and autonomy Greg Seymour a,⇑, Amber Peterman b a b

International Food Policy Research Institute, Environment and Production Technology Division, 1201 I Street NW, Washington, DC 20005, United States UNICEF Office of Research—Innocenti, Piazza della Santissima Annunziata 12, 50121 Florence, Italy

a r t i c l e

i n f o

Article history: Accepted 28 June 2018

JEL classification: D10 J16 Z13 Keywords: Autonomy Decision making Women’s empowerment Measurement Bangladesh Ghana

a b s t r a c t Using data from two culturally distinct locales, Bangladesh and Ghana, we investigate whether men and women who report sole decision making in a particular domain experience stronger (or weaker) feelings of autonomous motivation—measured using the Relative Autonomy Index (RAI)—compared to those who report joint decision making. Used primarily in psychology, the RAI measures the extent to which an individual’s actions are intrinsically or extrinsically motivated, where higher scores indicate greater autonomy. On aggregate, we find differences between men and women, and across countries, in the significance of association between the individual’s level of participation in decision-making and autonomy. In addition, we find heterogeneity in the strength of this association, depending on the domain (e.g., productive versus personal decisions) and whether partners agree on who normally makes decisions. These findings imply that details related to context and measurement matter for understanding individual decision-making power. We argue that all research using information on decision-making should include a careful analysis of men’s and women’s perceptions of decision making within the household, which may be useful for calibrating indicators to suit specific contexts. Ó 2018 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction The past decade has seen increased attention to measuring women’s empowerment and autonomy, motivated largely by the goal of identifying promising programs and policies for reducing gender inequalities. For the first time, the empowerment of women and girls is included in the Sustainable Development Goals as a stand-alone target. Yet, a lack of high-quality sex-disaggregated data—as well as ambiguity about how best to define and measure empowerment—makes it difficult to confidently measure gender inequalities and to assess the impact of development interventions on girls and women in many settings (Gammage, Kabeer, & Rodgers, 2016; Hanmer & Klugman, 2016; Klein, 2016; Peterman, Schwab, Roy, Hidrobo, & Gilligan, 2015; Richardson, 2017). In the social sciences, most approaches to defining and measuring empowerment are based on the concept of agency, defined by Sen as the ‘‘ability to use those capabilities and opportunities to expand the choices they have and to control their own destiny” (1999, 10), and focus on women’s ability to participate in decision ⇑ Corresponding author. E-mail addresses: [email protected] (G. Seymour), [email protected] (A. Peterman).

making over certain important matters (e.g., major household purchases, personal healthcare, or visits with friends and relatives). Questions about decision making are routinely collected in several large-scale surveys and contribute to a large body of evidence on how socioeconomic, health, and demographic outcomes are linked with women’s empowerment and agency.1 However, despite their widespread use, uncertainty persists about how to construct indicators of women’s empowerment based on these questions (Agarwal, 1997; Basu & Koolwal, 2005; Peterman et al., 2015). In particular, it is unclear to what extent sole and join decision making, respectively, should be considered different expressions of individual decision-making power and to what extent joint decision making reflects a consistent understanding of decision-making power within households. This paper takes a first step toward bridging these gaps by interrogating several of the most common critiques of household decision-making indicators using comparative information on women’s autonomy. Following psychologists working on a theory of motivation known as Self-Determination Theory (SDT), we 1 A set of decision-making questions has been included in Demographic and Health Surveys since the late 1990s, with the most recent round covering more than 40 developing countries globally.

https://doi.org/10.1016/j.worlddev.2018.06.027 0305-750X/Ó 2018 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

98

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

depart from the standard approach of treating autonomy as interchangeable with empowerment and instead conceptualize autonomy in terms of the motivations behind a person’s actions (Deci & Ryan, 2012; Ryan & Deci, 2000).2 In the parlance of SDT, ‘‘motivational” autonomy is defined as behavior that is experienced as willingly enacted and fully endorsed by a person. Thus, just as with Sen’s notion of agency, this definition emphasizes a person’s ability to act on behalf of his or her own personal values. Given this similarity, greater understanding of the relationship between motivational autonomy and decision making may provide insights into the robustness and validity of utilizing decision-making data to measure women’s empowerment. To this end, using data from two culturally distinct locales, Bangladesh and Ghana, we investigate whether respondents who report sole decision making in a particular domain tend to experience stronger (or weaker) feelings of autonomous motivation than those who report joint decision making. Specifically, we use multivariate regression models to estimate the association between a quantitative measure of motivational autonomy—the Relative Autonomy Index (RAI) proposed by Ryan and Deci (2000)—and, respectively, sole and joint decision-making outcomes. The RAI assigns a score to each decision domain based on survey questions that measure the extent to which an individual’s actions within the domain are intrinsically or extrinsically motivated, where higher scores indicate greater autonomy. On aggregate, we find that the significance of association between feelings of autonomous motivation and sole and joint decision-making, respectively, differs between men and women. Furthermore, we find heterogeneity in the strength of this association, depending on the domain and whether or not partners provide the same answers to questions about who normally makes decisions within the domain. Hence, the main lesson from our study is that the relationship between autonomy and sole or joint decision is heterogenous, depending largely on cultural context and the domain of decision making. To the extent that we believe autonomy is correlated with empowerment—as suggested by the literature, but not empirically confirmed in our analysis—our findings contribute to the discourse on measuring women’s empowerment and have implications for the broader use of decision-making indicators in development research. In particular, our analysis provides evidence of significant gender- and domain-specific variation in the association between autonomy and sole and joint decision making, respectively. This suggests, on one hand, that the common practice of treating sole and joint decision making as equivalent indicators of individual decision-making power may be inappropriate in some contexts. On the other hand, our results caution that, as a field, we are still far from understanding how generalized measures of autonomy and decision making relate to each other and to broader development objectives, such as empowerment and agency. Note that although we frame the policy and programming relevance of our findings in terms of women’s outcomes, our analysis utilizes data from both men and women. By doing so, we are able to add insight as to whether men’s and women’s reports agree on decision-making dynamics, as well as on how taking this heterogeneity into account affects our conclusions. Last, we expand our analysis beyond decision-making domains typically attributed to women and consider as well traditionally male-dominated productive and economic domains. The paper proceeds as follows. Section 2 discusses measurement issues in intrahousehold decision making and further develops the concept of motivational autonomy. Section 3 describes the data and offers context for our analysis. Section 4 reviews the

methodology used in the analysis. Section 5 presents our results. Section 6 concludes with a discussion of policy and research implications.

2 Anderson and Eswaran (2009), Cheong et al. (2017), Eswaran and Malhotra (2011); Jejeebhoy and Sathar (2001) all measured autonomy, at least partly, in terms of women’s ability to make decisions within their households.

3 Although not all of these studies focus on women’s ‘‘empowerment” per se, the underlying concepts under scrutiny are similar (agency, autonomy, bargaining power, status, etc.).

2. Review of decision making and autonomy in development research 2.1. Measuring intrahousehold decision making Women’s participation in intrahousehold decision making is frequently used as a metric of empowerment. The most common approach to operationalizing decision making in this manner involves condensing sole and joint decision making into a single (binary) indicator: having a ‘‘say” in a particular decision (Alkire et al., 2013; Allendorf, 2007; Anderson & Eswaran, 2009).3 Others have separated these dynamics, or analyzed only whether decisions are made solely or jointly with spouses or other household members (Bonilla et al., 2017; de Brauw, Gilligan, Hoddinott, & Roy, 2014; Handa, Peterman, Davis, & Stampini, 2009; Kishor & Subaiya, 2008). A potential problem with both approaches is that they implicitly assume that either sole and joint decision making are equally empowering for women, or alternatively that sole decisions are more empowering as compared to joint decisions. A lack of empirical evidence as to the conditions under which these assumptions hold or does not hold has led to ambiguity about how empowering having a say in a decision actually is for women (Deere & Twyman, 2012; Heckert & Fabic, 2013; Peterman et al., 2015). This uncertainty stems from several limitations, which have heretofore received insufficient attention in the literature. First and foremost, it is unclear to what extent being a joint participant in a decision reflects having a meaningful voice in the decision-making process, and how indicators might be constructed to capture any subtle differences. This concern stems, in part, from a lack of contextual details about the decision-making process itself—knowing who made a decision does not reveal everything about the mechanics of how a decision was made. For instance, joint decision making when all participants agree may reflect a different dynamic than joint decision making when there is conflict. In such cases, knowing what tends to happen when participants in a decision disagree with one another can provide valuable insight into the extent to which joint decision making reflects compromise among participants or capitulation by some participants to the wishes of another (dominant) participant. Although compromise may reflect empowerment, capitulation may or may not. These questions are further complicated by the fact that decisions are not discrete, and are often made iteratively. For example, spouses may make a joint decision in which one party undertakes ‘tacit agreement’ while not completely accepting the agreement and planning to open and contest the decision at a later date (Agarwal, 1997). Another factor that complicates the interpretation of joint decision making concerns household composition. In households with several adult members, decisions are more likely to be made jointly due to sharing of resources and responsibilities among household members. In such households, it may be especially important to consider with whom joint decisions are made, because the implications for empowerment may be very different if a woman makes a decision jointly with her spouse or with her father, mother-in-law, or son (Doss, 2013; Heckert & Fabic, 2013; Peterman et al., 2015). Similarly, the interpretation of sole decision making can also vary depending on the extent to which women are

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

constrained by the preferences of other household members or prevailing social norms (Doss, 2013). Relatedly, it is possible for individuals to leverage support or influence on decisions from others outside the households (e.g. referred to as ‘‘interest coalitions”), further complicating the interpretation of reported sole or joint decisions (Agarwal, 1997). Second, if the goal is to assess women’s empowerment, then we should also consider women’s preferences about decision making and the extent to which their preferences align with how they actually make decisions. This concern stems from the notion that the expression of agency requires that a person’s actions reflect the pursuit of goals that she personally values. Unfortunately, determining how involved women wish to be in decision making on a particular matter, and relatedly, which decision-making outcomes they would consider ideal, is rarely a straightforward matter. For a particular decision, several outcomes are possible, any one of which a woman may prefer for different reasons. For example, she may prefer to (1) solely make decisions if she places high value on the freedom to make decisions without consultation, (2) jointly make decisions if she derives utility from cohesion within the household (Doss & Meinzen-Dick, 2015), or (3) not be involved at all in decision making. Moreover, her preferences are likely to vary across domains depending on her available time or the level of responsibility she wishes to have (Basu & Koolwal, 2005). For example, decisions about children’s education may be the responsibility of one spouse, decisions about children’s health may be the purview of the other spouse, and decisions about when and whom children should marry may be made jointly by both spouses. Empirical research on decision making provides evidence of significant variation across domains (de Brauw et al., 2014; Doss, Kim, Njuki, Hillenbrand, & Miruka, 2014; Kishor & Subaiya, 2008; Mabsout & van Staveren, 2010; Oduro, Boakye-Yiadem, & BaahBoateng, 2012; Twyman, Useche, & Deere, 2015). Most notably, Kishor and Subaiya (2008), using Demographic and Health Surveys data for married women ages 15–49 in 23 developing countries,

99

found that the correlates of both sole and joint decision making varied by country and by domain in decisions over personal healthcare, large household purchases, household purchases for daily needs, and visits to family or friends. Similarly, how women experience empowerment can vary greatly across domains (Alkire, 2007; Ibrahim & Alkire, 2007). A decision-making arrangement that is empowering for women in one domain may not necessarily be as empowering if experienced in another domain. For instance, a woman might be empowered inside her household (e.g., household decision making or asset ownership) without experiencing improvements outside her household (e.g., mobility or the right to vote). Yet even within a given domain, the same set of behaviors and attributes that signify empowerment in one context may mean something entirely different in another context (Malhotra, Schuler, & Boender, 2002). For example, an increase in a woman’s ability to visit her parents without her husband’s permission may reflect empowerment in rural Bangladesh but not in urban Ghana. Moreover, personal interpretations of what it means to be empowered can vary greatly from one woman to another even among those living in relative close proximity to each other, as Klein (2014, 2016) found to be the case among women from different neighborhoods in Bamako, Mali. Accurately measuring empowerment requires information about the relative value a person places on participating in decision making within a domain. This paper suggests an approach to analyzing sole and joint decision making that leverages information about the motivations behind a person’s actions in a way that helps us to assess how each type of decision making relates to empowerment. Details of this approach are discussed in the following section. 2.2. Measuring autonomy According to SDT, human behavior is driven by intrinsic and extrinsic motivations (Fig. 1) (Deci & Ryan, 2012; Ryan & Connell,

Fig. 1. Self-determination continuum. Source: Adapted from Vaz et al. (2016) and Ryan and Deci (2000). Notes: Respondents were asked to rate (on a Likert scale ranging from 1, ‘‘never true,” to 4, ‘‘always true”) how true it would be to say that their actions within a specified domain were motivated by: (1) a desire to avoid punishment or gain reward (external), (2) a desire to avoid blame or so that other people speak well of you (introjected), and (3) your own values and/or interests (identified, integrated, and intrinsic). For each domain, a person’s RAI score was calculated by summing the answers to the three questions using the following weighting scheme:
2 for external motivation (question 1),
1 for introjected motivation (question 2), and +3 for autonomous motivation (question 3), where the weights correspond to the relative positions of each motivation on the self-determination continuum (Fig. 1). Thus, the RAI score is domain-specific and ranges from
9 to +9. Positive scores are interpreted as evidence of autonomous motivation, and negative scores as an indication of controlled behavior.

100

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

1989; Ryan & Deci, 2000). Intrinsic motivation refers to engaging in an activity for the satisfaction and fulfillment of the activity itself. It reflects the purest expression of autonomy. Extrinsic motivation refers to doing an activity to achieve some instrumental reward, separate from the experience of the activity itself, and can also reflect autonomy. SDT further categorizes extrinsic motivation into four subtypes: external regulation, introjected regulation, identified regulation, and integrated regulation. External regulation encompasses behaviors that are controlled or coerced by some external force, such as actions taken to achieve a tangible reward or to avoid a threatened punishment. Introjected regulation entails the internalization of external regulations and manifests as behaviors that are influenced by the beliefs and expectations of others, such as acting to avoid feelings of shame or anxiety or to attain ego enhancements, such as pride. Identified regulation involves conscious recognition and acceptance of the underlying value of a goal or activity as personally important, as in the decision to exercise regularly due to positive health effects. Integrated regulation occurs as identified regulations are integrated into a person’s identity or self-image. External regulation and introjected regulation are the least autonomous motivations, and identified regulation and integrated regulation the most autonomous. SDT recognizes that a person’s actions are the product of a continuum of several different motivations (Fig. 1). For instance, a woman in rural Ghana cares for her children every day because she (1) enjoys the experience (intrinsic), (2) views herself as caretaker and identifies as a ‘‘good mother”, (3) appreciates the impact her care has on her children’s lives (identified), (4) does not want to disappoint family members who expect her to do it (introjected), and (5) fears reprisal from others in her community if she did not do it (external). Similar examples could be given for decision domains around productive activities or around decisions regarding personal domains. The RAI explicitly takes these nuances into account by measuring the extent to which a person experiences each type of motivation with respect to her actions within a given domain. Specifically, we asked respondents to rate (on a Likert scale ranging from 1, ‘‘never true,” to 4, ‘‘always true”) how true it would be to say that their actions within a specified domain fit each of the following descriptions: 1. Motivated by a desire to avoid punishment or gain reward (external). 2. Motivated by a desire to avoid blame or so that other people speak well of you (introjected). 3. Motivated by and reflective of your own values and/or interests (identified, integrated, and intrinsic). Note that the last question captures all three forms of autonomous motivation (identified, integrated, and intrinsic). For each domain, an RAI score is calculated by summing the answers to the three questions using the following weighting scheme:
2 for external motivation (question 1),
1 for introjected motivation (question 2), and +3 for autonomous motivation (question 3), where the weights correspond to the relative positions of each motivation on the self-determination continuum (Fig. 1). Thus, the RAI score is domain-specific and ranges from
9 to +9. Positive scores are interpreted as evidence of autonomous motivation, and negative scores as an indication of controlled behavior. 2.3. Relating decision making to autonomy The comparison of a person’s RAI score to his or her level of participation in decision making within a given domain can be exploited to provide information about the decision-making process that avoids many of the limitations associated with measuring decision making and enhances our ability to make judgments

about how empowering sole and joint decision making are to the person(s) involved. Because many of the limitations around decision-making indicators have to do with the motivations behind a person’s decisions, they can be addressed by using the RAI score to tease out the level of involvement in decision making that the person might prefer, while also accounting for the impact of gender norms on the decision-making process. The intuition behind this argument is as follows. If the actual decision-making outcome that occurs within a given domain does not align with a person’s preferred outcome—say, if a woman makes decisions over her children’s education alone, but ideally would prefer input from her spouse on the matter—then we expect this discrepancy to be reflected in a lower RAI score because her actions would not, in this case, reflect her own values and interests (i.e., would not reflect autonomous motivation). Conversely, if the actual decision-making outcome that occurs within a given domain is a close match with a person’s preferences, then we expect the RAI score to be higher because the person’s actions would, in fact, reflect his or her personal values and interests. Thus, we argue that the correlation between the RAI score and decisionmaking outcomes provides a de facto ranking of men’s and women’s decision-making preferences and insight into how empowering a person perceives each outcome to be. Additionally, our approach is capable of handling cases in which men’s and women’s roles in decision making over certain matters within the household are mostly determined by gender norms. This ability stems from the fact that the internalization of gender norms reflects a form of introjected regulation. If decisionmaking responsibilities within a household are largely distributed among members according to gender norms (as opposed to more autonomous forms of motivation), then we expect household members’ RAI scores to be low. Thus, the judgments about decision making and empowerment that emerge from our analysis should be robust to the effects of gender norms on decision-making processes.

3. Data and context 3.1. Data The data used in our analysis come from the 2011–2012 Bangladesh Integrated Household Survey (BIHS) and the 2012 Feed the Future Ghana Population Baseline Survey (GPBS).4 The BIHS is statistically representative of rural Bangladesh (as a whole); the GPBS is statistically representative of the primarily rural Feed the Future zone of influence in northern Ghana (the Upper West, Upper East, and Northern regions, and areas in Brong Ahafo Region above the eighth parallel). The total sample sizes for the BIHS and GPBS are 5500 and 4340 households, respectively.5 Sampling for both surveys captured single-adult (adult[s] of only one sex) and dual-adult (adults of both sexes) household living arrangements. To avoid biases due to the effects of household composition on decisionmaking outcomes, we restrict our analysis to dual-adult households only. The resulting sample sizes for our analysis are 4750 households in Bangladesh and 3616 households in Ghana. Both surveys included the RAI questions (discussed above) and intrahousehold decision-making questions with respect to several different domains. The decision-making questions take the following format, without reference to a specific time frame: ‘‘When deci4 Both datasets are publicly available online: the BIHS from the IFPRI Dataverse, http://hdl.handle.net/1902.1/21266, and the GPBS from USAID’s Development Data Library, https://www.usaid.gov/data/dataset/f722cc77-3e00-4a78-921190044db5740a. 5 See Malapit and Quisumbing (2015) and Sraboni et al. (2014) for details on the sampling design in the GPBS and BIHS, respectively.

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

sions are made regarding [DOMAIN], who is it that normally takes the decision?” Response options to the decision-making questions included: (1) main male or husband, (2) main female or wife, (3) husband and wife jointly, (4) someone else in the household, (5) jointly with someone else in the household, (6) jointly with someone else outside the household, (7) someone outside the household, and (8) decisions not made. The decision-making questions directly preceded the RAI questions in the survey interview, and thus, if respondents indicated that decisions in a particular domain were not made in their household, then the RAI questions with respect to that domain were skipped. Note that although the two surveys administered the RAI questions using slightly different wording, the questions were similarly structured and captured the same forms of motivation. The core set of domains included in both surveys captures the essential elements of life within an agricultural household (agricultural production, purchase of inputs, crop choice, taking crops to market, livestock raising, nonfarm business activity, own employment, and minor household expenditures). In addition, the BIHS includes several nonagricultural domains (own health problems, protection from violence, expression of religion, daily tasks, and family planning), and the GBPS includes major household expenditures. 3.2. Country context Despite recent improvements in women’s access to education, the labor force, and health services, considerable progress remains to be made toward achieving gender equality in both Bangladesh and Ghana (World Bank, 2013, 2016a). Both countries ranked relatively poorly on the United Nations Development Programme’s most recent gender inequality index: Bangladesh was 111th and Ghana was 127th out of 188 countries in total (UNDP, 2015). Women in both countries have constrained autonomy stemming from numerous institutional and sociocultural factors, in particular norms around the intrahousehold division of labor. Traditionally, the roles of household head and primary income provider are assumed to be men’s responsibilities, and domestic matters, such as preparing family meals and caring for children, are considered women’s domain (Clark, 1994; White, 1992). In effect, such arrangements limit women’s ability to engage in work outside the home by restricting both their mobility and their available time. This is specifically the case in Bangladesh where the social norm of female seclusion (‘‘purdah”) is still practiced, whereby women’s mobility is restricted and women are likely to be accompanied by men, and/or covered when working outside the home or in public spheres (Heath, 2014; Uddin, Hossin, & Pulok, 2017). Yet norms around the intrahousehold division of labor within each country are not immutable and often vary over space and time, as well as by religion, class, education, and age. Agriculture is central to men’s and women’s livelihoods in both countries, but the opportunities afforded to women in agriculture in each country are very different. In Ghana, as in other parts of West Africa, men and women within the same household tend to cultivate separate plots (Doss, 2002). Yet a large gender gap exists in agricultural landownership: 64 percent of plots are owned solely by men, 29 percent by women, and 3% jointly by men and women (Doss et al., 2011). In Bangladesh, the vast majority of agricultural plots are cultivated solely by men, with an even larger gender gap in ownership: 86 percent of plots are owned solely by men, 12 percent by women, and 2 percent jointly by men and women (Kieran, Sproule, Doss, Quisumbing, & Kim, 2015; Seymour, 2017). The selection of Bangladesh and Ghana for this study was largely a practical choice, motivated by the fact that the BIHS and GPBS were among the first publicly available, large-scale Women’s Empowerment in Agriculture Index (WEAI) datasets with both the RAI questions and decision-making questions across numerous

101

agricultural decision domains. Yet, differences in women’s role in agriculture in each country—which might lead to corresponding differences in women’s preferences about agricultural decision making—provide a compelling context for studying the relationship between autonomy and decision making. Women living in Bangladesh, faced with few opportunities to exert authority over agricultural decision making, may be less inclined to differentiate between sole and joint decision making in agriculture. For these women, having any say in productive decision making may be viewed as an improvement in their bargaining position within the household, particularly given the social norm of purdah common in rural areas. In contrast, in Ghana, where it is relatively more common for women to own or operate land on their own, women’s view of joint decision making may not be as favorable—some might even view joint decision making as a reduction in their bargaining position. 4. Empirical methodology 4.1. Regression modeling To determine whether respondents who report sole decision making experience stronger (or weaker) feelings of autonomous motivation compared with those who report joint decision making, we investigate how autonomy correlates with participation in decision making. We begin by calculating each respondent’s RAI score and indicators of sole and joint decision making with respect to each domain. We then convert all three of these variables into indexes using the first factor from a factor analysis aggregating responses across two subsets of domains, broadly corresponding to agricultural decisions and ‘‘personal” (or noneconomic) decisions:6  Agricultural decisions (Bangladesh and Ghana) include agricultural production, purchase of agricultural inputs, types of crops to grow, who takes crops to market and when, and livestock raising.  Personal decisions (Bangladesh only) include: own health problems, expression of religion, daily tasks, and family planning. We choose an aggregate approach to minimize the risk of false positives or negatives due to multiple hypothesis testing. Nonetheless, we recognize that aggregating responses in this way potentially masks interesting and important heterogeneity across domains. Hence, as a robustness check, to understand underlying sources of variation, we also conduct a domain-specific analysis of autonomy and decision making. We estimate the following equation for each subset of domains via ordinary least squares (OLS) regression:

RAIi ¼ a þ solei b þ jointi c þ X0i d þ e;

ð1Þ

where RAIi, solei, and jointi are aggregate indexes of autonomy, sole decision making, and joint decision making, respectively, for person i, and X is a vector of control variables capturing relevant individual and household characteristics that may influence decision making (Table A.1 in the appendix provides descriptives statistics). Given that we are primarily interested in whether sole and joint decision making relate to autonomy similarly or differently within each domain, our focus is on how b and c compare with each other. To 6 Relative to simply summing the RAI scores and counting the incidence of sole and joint decision-making across domains, this approach provides a more nuanced measure of the individual indicators, by accounting for the joint distribution of responses across domains (Kline, 1994). See Table A.2 in the appendix for an alternate set of regression results, in which the RAI scores, sole decision making, and joint decision making are aggregated following this simpler approach. With the exception of a single hypothesis test, our findings are robust to this alternative approach.

102

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

this end, we conduct a hypothesis test comparing the two coefficient estimates (H0: b = c). A statistically significant difference between b and c indicates that, on average, sole and joint decision making do not relate to autonomy in the same way. Because this relationship may vary according to gender and cultural context, we estimate Eq. (1) separately for men and for women within each country. One critique of decision-making questions is that individuals within the same household may not necessarily perceive the decision-making process in the same way. Existing research suggests that couples disagree frequently when asked about decision making (Ambler, Doss, Kieran, & Passarelli, 2016; Donald et al., 2017; Twyman et al., 2015). For instance, using nationally representative data from Ecuador, Twyman et al. (2015), found that men tended to report their wives as participating less in several agricultural decisions than indicated by the wives themselves. Although the decision-making questions in the BIHS and GPBS are phrased to capture how decisions about certain aspects of household life are ‘‘typically” made within the household, without reference to a particular time frame, it is possible that when formulating their answers respondents may focus on different specific instances of decision making from the past. For example, conflicting reports of decision making on crop choice from a husband and wife might arise naturally if each considers a different plot—perhaps one that he or she cultivates individually—or a different point in time as the reference point for his or her answers. Spousal conflict may also reveal something about the underlying power dynamics within the household. For example, Ambler et al. (2016), using the same data from Bangladesh used in this paper, found that agreement between couples—particularly the husband’s acknowledgment of his wife’s involvement in decision making— was correlated with better development outcomes for women.7 From an empirical perspective, it is unclear how to deal with this sort of inconsistency. Are these differences a result of measurement error or social desirability bias, and thus should partners who give conflicting accounts be dropped from the sample? Should one partner’s response be used as a proxy for the other’s (as is typically done when questions are asked only of one individual within the household)? To investigate this issue—and more specifically, whether a person’s perception of autonomy depends on a consistent understanding of decision-making power within the household—we create an indicator of within-household agreement on decision making, whereby couples are judged as agreeing on decision making for a particular domain if (1) both report joint decision making or (2) one reports sole decision making and the other reports not participating in the decision. We aggregate responses across the same subsets of domains as before. However, to permit a more straightforward interpretation, we eschew a factor analytic approach in favor of simply defining a couple as being in agreement if they agree on the process used for the majority of decisions made within each subset of domains. Similar to before, we estimate the following equation for each subset of domains:

RAIij ¼ a þ soleij bDISAGREE þ jointij cDISAGREE þ agreementj hþ

soleij  agreementj bAGREE þ jointij  agreementj cAGREE þ X0i d þ e: ð2Þ where we add agreementj, denoting couples’ agreement on decision making in household j, and its interaction with the sole and joint 7

The outcomes considered in Ambler et al. (2016) include including (1) whether the wife worked more than 10.5 h per day, (2) the wife’s body mass index, (3) whether the wife has ever used birth control, (4) the number of groups in which the wife is an active participant, and (5) whether the wife currently has a loan.

decision-making indexes for person i. For ease of interpretation, we treat the RAI score as continuous and estimate Eq. (2) using OLS regression.8 Our primary focus is on whether sole and joint decision making relate to autonomy similarly or differently when respondents agree on how decisions are made versus when they disagree. Hence, we conduct hypothesis tests comparing the coefficient estimates associated with sole and joint decision making for couples who tend to agree (H0: bAGREE = cAGREE) and those who tend to disagree (H0: bDISAGREE = cDISAGREE). 4.2. Validity of the Relative Autonomy Index A final concern is the validity of the RAI itself. Although the RAI has been extensively tested across several different contexts (Alkire & Chirkov, 2007; Alkire, 2005; Chirkov, Ryan, Kim, & Kaplan, 2003), the majority of these exercises have been carried out among well-educated populations in more developed countries. Two exceptions—Vaz, Alkire, Quisumbing, and Sraboni (2013) and Vaz, Pratley, and Alkire (2016)—found support for the overall reliability and validity of the RAI (albeit with mixed results) using data from Bangladesh and Chad, respectively.9 Bearing in mind concerns about the performance of the RAI among less educated populations, we conduct an analysis of the validity of the RAI following a similar approach as Vaz et al. (2016). The results and further discussion of the validation exercise are presented in the appendix (Tables B.1–B.7). First, we conduct an exploratory factor analysis (EFA) to test whether the dimensional structure of the data reflects the latent characteristics that the questions aim to measure, i.e., external, introjected, and autonomous motivations. We generally find support for this hypothesis. The EFA results indicate that given reasonably sized samples—obtained by restricting our analysis to the two subsets of domains described above—three dimensions emerge from the data, each corresponding to one of the motivation subscales. The lone exception occurs among men in the Bangladesh sample, for whom only two factors emerge when we consider the subset of domains pertaining to personal decisions, which suggests that for men in Bangladesh the distinction between these two types of motivation, while relevant when making agricultural decisions, may not be as relevant in a non-agricultural context. Next, we calculate Spearman correlation matrices to test whether there is ordered correlation among the motivation subscales that conforms to the self-determination continuum—i.e., whether adjacent subscales, such as external and introjected motivation, are more correlated than nonadjacent subscales, such as external and autonomous motivation. We find mixed results for this hypothesis. Although the patterns of correlation observed in the data for men and women in Ghana and men in Bangladesh are generally a good fit for the self-determination continuum, we find evidence of positive correlation between nonadjacent subscales (external and autonomous motivation) among Bangladeshi women in the sample, particularly in domains involving income generation of some sort. One potential explanation for this pattern, offered by Vaz et al. (2013), is that it may reflect an internalization of social norms in Bangladesh around women’s participation in economic activities that makes it hard for women to discern between doing something because it aligns with their own values and interests and doing something to avoid punishment or gain reward. Another possible explanation is measurement error, perhaps stemming from respondents’ confusion over the differences 8 Treating RAI scores as ordinal and estimating Eq. (2) using ordered probit regression does not significantly change our results (Tables A.10–A.13 in the appendix). 9 Vaz et al. (2013) analyzed the same data for Bangladesh that we use in this paper, though with slight differences in specification.

103

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

among the three motivations asked about in the questions. Indeed, evidence from two studies assessing the cognitive validity of the WEAI, which includes the RAI, in Bangladesh, Uganda, and Haiti suggests that the RAI questions can be particularly difficult for some respondents to comprehend (Johnson & Rosell, 2015; Malapit, Sproule, & Kovarik, 2016). Concerns about its validity notwithstanding, we believe that analysis of the RAI can offer valuable insights into men’s and women’s behaviors. Vaz et al. (2013) reached a similar conclusion, contending that the RAI provides new information about individual behavior that is not captured by other indicators. Nevertheless, we exercise caution in the interpretation of our results and revisit in later discussion many of the concerns raised in this section.

5. Results 5.1. Decision-making outcomes Tables 1 and 2 provide a breakdown of men’s and women’s decision making in Bangladesh and Ghana, respectively. Several trends are evident. Overall, men in both countries are more likely than women to report sole decision making and less likely to

report having no input into decision making. Nevertheless, there are a few domains in which the gender gap in sole decision making is comparatively narrower, such as decisions about the expression of religion or daily tasks (in Bangladesh) and nonfarm business activity (in Ghana). Although this narrowing may reflect actual outcomes, it is also possible that questions about decision making in these domains were interpreted as applying specifically to the respondent’s own activities, as opposed to household activities in general. A different pattern is apparent in the rates of joint decision making reported by men and women in Bangladesh. Women reported joint decision making at more than twice the frequency of men for several domains (agricultural production, purchase of agricultural inputs, crop choice, taking crops to market, nonfarm business activity, own employment, and expression of religion). In comparison, men and women in Ghana reported joint decision making at roughly the same frequency for every domain. Tables 1 and 2 also show the mean RAI scores of men and women in Bangladesh and Ghana, respectively. RAI scores are positive—signifying that, on average, respondents’ behaviors are primarily motivated by personal values and interests rather than controlled by external factors—and range from 3.4 to 5.2, with slightly lower mean values for Bangladesh than for Ghana. In Ghana, women reported significantly lower levels of autonomy

Table 1 Decision-making outcomes and Relative Autonomy Index scores by gender, Bangladesh. Percentage of respondents reporting . . .

Mean RAI score (range
9 to +9)

Number of respondents

Sole decision making

Joint decision making

No input in decision making

Domain

Men

Women

Men

Women

Men

Women

Men

Women

Men

Women

Agricultural production Purchase of inputs Crop choice Taking crops to market Livestock raising Nonfarm business activity Own employment Minor household expenditures Own health problems Protection from violence Expression of religion Daily tasks Family planning

77.70 77.53 77.34 77.27 44.83 76.26 87.17 55.74 44.13 58.75 89.74 84.16 15.46

2.97 2.81 2.75 3.20 21.97 6.94 20.61 13.80 5.23 12.51 69.31 74.61 23.96

19.91 19.95 20.18 20.23 44.27 17.81 10.66 41.19 52.38 31.37 9.61 14.13 78.36

47.41 48.54 50.81 50.67 65.91 49.05 43.29 61.20 70.09 49.22 22.93 22.86 74.23

2.39 2.52 2.49 2.50 10.90 5.92 2.17 3.08 3.49 9.88 0.65 1.71 6.18

49.63 48.65 46.45 46.13 12.12 44.01 36.10 25.00 24.68 38.27 7.76 2.53 1.81

4.34 4.35 4.42 4.42 4.30 4.33 4.35 4.44 4.03 3.41 3.70 4.46 3.84

4.15 4.08 4.06 4.04 4.18 3.43 3.86 4.32 4.16 4.22 4.39 4.65 4.29

2848 2817 2815 2644 2817 2195 2626 4516 4038 1629 3820 4269 3447

2934 2919 2913 2753 2963 1629 1892 4652 4416 1351 3376 4419 3640

Source: Authors’ calculations based on 2011–2012 BIHS data. Note: Total number of households in sample: 4735. The percentages shown in columns 2–7 and totals shown in columns 10 and 11 exclude respondents who reported that decisions about a given domain were not made in their households. Gender differences are statistically significant at a 90 percent or greater confidence level unless indicated in bold. RAI = Relative Autonomy Index.

Table 2 Decision-making outcomes and Relative Autonomy Index scores by gender, Ghana. Percentage of respondents reporting . . . Sole decision making

Joint decision making

No input in decision making

Mean RAI score (range
9 to +9)

Number of respondents

Domain

Men

Women

Men

Women

Men

Women

Men

Women

Men

Women

Agricultural production Purchase of inputs Crop choice Taking crops to market Livestock raising Nonfarm business activity Own employment Major household expenditures Minor household expenditures

70.44 70.36 67.74 58.07 71.61 44.31 62.11 50.00 32.55

5.89 5.78 6.01 13.80 5.16 44.32 19.08 6.30 25.27

26.22 26.30 28.70 30.48 23.73 27.21 27.54 43.71 45.26

27.87 27.47 30.07 30.58 25.06 29.91 30.48 42.06 46.98

3.34 3.33 3.56 11.45 4.66 28.48 10.35 6.29 22.19

66.24 66.74 63.93 55.62 69.79 25.76 50.44 51.64 27.75

4.78 4.95 5.05 4.88 5.06 4.96 4.97 5.19 4.92

3.96 4.14 4.27 4.29 4.27 4.29 4.36 4.54 4.31

2967 2939 2951 2638 2339 1415 541 1256 3186

2734 2697 2714 2384 2095 1471 456 1158 2944

Source: Authors’ calculations based on 2012 GPBS data. Note: Total number of households in sample: 3616. The percentages shown in columns 2–7 and totals shown in columns 10 and 11 exclude respondents who reported that decisions about a given domain were not made in their households. Gender differences are statistically significant at a 90 percent or greater confidence level unless indicated in bold. RAI = Relative Autonomy Index.

104

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

Table 3 OLS regression results predicting Relative Autonomy Index scores by gender and country. Bangladesh

Ghana

Agricultural decisions

Personal decisions

Variable

Women

Men

Women

Men

Women

Men

Sole decision making (b)

0.048 (0.031) 0.155*** (0.041)
0.024 (0.015) 0.339y (0.177) 0.018 (0.069)
0.011 (0.171) 0.110 (0.155) 0.001 (0.002)
0.034 (0.050) —

0.152* (0.076) 0.136 (0.084) 0.003 (0.012) 0.039 (0.125) 0.055 (0.056)
0.122 (0.102)
0.030 (0.134)
0.002 (0.002)
0.020 (0.042) —


0.054 (0.066)
0.181** (0.060) 0.004 (0.012)
0.020 (0.149) 0.018 (0.056) 0.144 (0.144)
0.276 (0.211)
0.001 (0.002)
0.016 (0.041) —

0.349*** (0.089) 0.113 (0.083)
0.012 (0.010) 0.162 (0.105) 0.002 (0.048)
0.196* (0.095)
0.321 (0.235)
0.003y (0.002) 0.014 (0.038) —

Ethnicity dummies

0.080 (0.189) 0.036 (0.130) 0.067 (0.066) No


0.224y (0.117) 0.205* (0.096) 0.219*** (0.058) No


0.200 (0.134)
0.004 (0.111) 0.085 (0.065) No


0.218* (0.108) 0.125 (0.097) 0.210** (0.066) No

0.031 (0.026) 0.015 (0.030)
0.015 (0.010) 0.177 (0.107) 0.277* (0.136)
0.032 (0.196)
0.239 (0.149) 0.003y (0.002)
0.000 (0.058) 0.139* (0.064)
0.077 (0.077) 0.057 (0.050)
0.135*** (0.038) Yes

0.055 (0.066)
0.005 (0.065) 0.006 (0.008)
0.034 (0.075)
0.001 (0.103) 0.188y (0.111)
0.188y (0.114) 0.000 (0.002)
0.059 (0.073) 0.128y (0.068) 0.008 (0.076) 0.103* (0.048)
0.077* (0.036) Yes

Adjusted R2 H0: b = c (F statistic) Observations

0.020 6.147* 1558

0.031 0.201 1950

0.023 14.267*** 2325

0.071 59.086*** 2453

0.029 0.242 1785

0.025 4.351* 1992

Joint decision making (c) Age Age-squared/1000 Completed primary education Higher than primary education Married Age gap: Husband – wife Husband more educated than wife Christian Muslim (log) Area of cultivated land (log) Per capita consumption

Agricultural decisions

Source: Author’s calculations based on 2011–2012 BIHS and 2012 GPBS data. Note: Standard errors clustered at the primary sampling unit level in parentheses. yp < 0.10 *p < 0.05, **p < 0.01, and

than men across all domains. In Bangladesh, a more complicated pattern is apparent. Although women reported significantly less autonomy than men for 6 out of 13 domains (agricultural production, purchase of inputs, crop choice, taking crops to market, nonfarm business activity, and own employment), they report similar or higher levels of autonomy for the remaining 7 domains (livestock raising, minor household expenditures, own health problems, protection from violence, expression of religion, daily tasks, and family planning).

5.2. Regression analysis Table 3 presents the regression results from our aggregate analysis. We estimate the model with and without control variables (see Table A.3 in the appendix for the latter results). Although in the discussion below we emphasize the conditional results (i.e., with controls), our findings are consistent across both specifications. This implies that the correlations observed between autonomy and decision making cannot be explained by the underlying socio-economic factors included in the model.10 When comparing coefficient estimates, it is advisable to focus on the statistical significance of the difference between estimates, rather than the statistical significance of the estimates themselves (Gelman & Stern, 2006). Hence, our discussion emphasizes the results of the hypothesis test of whether the coefficients associated with sole and joint decision making are equal, rather than the size and significance of the coefficient estimates themselves. 10 This can also be seen in the fact that the correlates of the RAI are distinct from the correlates of sole and joint decision making (see Tables A.4 and A.5 in the appendix).

***

p < 0.001. — data not available.

For agricultural decisions, we see a clear distinction in the way in which women in each country experience sole and joint decision making. Notably, the difference in the coefficients associated with sole and joint decision making is statistically significant in Bangladesh, but not in Ghana. More specifically, women in Bangladesh associate joint decision making on agricultural matters more strongly with autonomy than sole decision making. Hence, at least in terms of agricultural decisions, joint decision making may be a stronger predictor of autonomy than sole decision making for women in Bangladesh, but not for women in Ghana, where neither mode of decision making seems to predict autonomy. Interestingly, we see something different for women in Bangladesh for personal decisions. Although the difference between the coefficients associated with sole and joint decision making is still statistically significant, both are negatively correlated with autonomy (though only statistically significantly for joint decision making). Thus, in contrast to agricultural decisions—where joint decision making is associated with greater autonomy—joint decision making on personal matters is associated with a reduction in women’s autonomy. Among men, yet another pattern emerges. In Bangladesh, the difference in the coefficients associated with sole and joint decision making is statistically insignificant for agricultural decisions, but statistically significant for personal decisions. In Ghana, the difference in the coefficients is statistically significant, yet just as with women in Ghana, neither is significantly associated with autonomy. Hence, neither sole nor joint decision making on agricultural matters seems to be a strong predictor of men’s autonomy in either country. For men’s personal decisions in Bangladesh, however, sole decision making appears to a much stronger predictor of men’s autonomy than joint decision making.

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

5.3. Sensitivity analysis to domain-specific heterogeneity Figs. 2 and 3 show the coefficient estimates associated with sole and joint decision making obtained when we repeat the above analysis separately for each domain (significance at or above p < 0.10 denoted by a black point estimate marker, with the significance level reported alongside the domain labels). Using the Benjamini-Hochberg method, we adjust the p values used in the hypothesis tests to control for multiple comparisons (Benjamini & Hochberg, 1995). Full regression results are presented in Tables A.6–A.9 in the appendix. Domain-specific analysis allows us to perceive several nuances that are masked in the aggregate analysis. Most notably, the stark contrast in the relationship of sole and joint decision making to women’s autonomy in Bangladesh, apparent when we consider agricultural decisions in aggregate, is less evident when we consider each decision separately (Fig. 2, Panel A). Although joint decision making still appears nominally more correlated with autonomy than sole decision making in all agricultural decisions except taking crops to market, differences between the coefficients are not statistically significant. For personal decisions, the domainspecific results reveal that the negative correlation initially observed between joint decision making and autonomy is largely driven by statistically significant differences in just two domains:

105

expression of religion and daily tasks. These are also domains where women report overall higher autonomy as compared to men, which may reflect gendered social norms which ascribe these domains to women or women’s own perception that they should have full control over these domains of personal nature. Among women in Ghana, however, the domain-specific results largely echo the aggregate results (Fig. 3, Panel A). Differences between sole and joint decision making are statistically significant in only one agricultural domain: taking crops to market. We do, however, observe significant differences in two out of four nonagricultural domains: nonfarm business activity and minor household expenditures. This indicate there are several productive domains in which women relate taking sole decisions with higher autonomous feelings. For men in Bangladesh, the aggregate and domain-specific results are, again, quite similar to the aggregate results (Fig. 2, Panel B). Differences in the coefficients associated with sole and joint decision making are generally small and statistically insignificant for agricultural decisions (taking crops to market is the only exception) but larger and statistically significant for personal decisions (nonfarm business activity is the only exception). The domains where we observe the largest differences—all in favor of sole decision making—are expression of religion, daily tasks, and family planning. Interestingly, these differences are also observed

Fig. 2. Decision-making coefficient estimates predicting Relative Autonomy Index scores with 95 percent confidence intervals by gender, Bangladesh. Source: Author’s calculations based on 2011–2012 BIHS data. Note: Coefficients from OLS models with standard errors clustered at the primary sampling unit level. 95 percent confidence intervals indicated by error bars. P values adjusted to control for multiple comparisons using Benjamini-Hochberg method (Benjamini & Hochberg, 1995). Statistically significant differences between the coefficients associated with sole and joint decision making indicated by black markers. yp < 0.10, *p < 0.05, **p < 0.01, and ***p < 0.001. Full results provided in Tables A.6 and A.7 in the appendix.

106

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

Fig. 3. Decision-making coefficient estimates predicting Relative Autonomy Index scores with 95 percent confidence intervals by gender, Ghana. Source: Author’s calculations based on 2012 GPBS data. Note: Coefficients from OLS models with standard errors clustered at the primary sampling unit level. 95 percent confidence intervals indicated by error bars. P values adjusted to control for multiple comparisons using Benjamini-Hochberg method (Benjamini & Hochberg, 1995). Statistically significant differences between the coefficients associated with sole and joint decision making indicated by black markers. yp < 0.10, *p < 0.05, **p < 0.01, and ***p < 0.001. Full results are provided in Tables A.8 and A.9 in the appendix.

for women, in favor of sole decision making—in two of these domains: expression of religion and daily tasks—which suggests these domains may be seen as personal in nature and those over which individuals should have full control. For men in Ghana, the domain-specific analysis reveals small but statistically significant differences in the coefficients associated with sole and joint decision making in every agricultural domain except one (agricultural production), which in all cases favor sole decision making (Fig. 3, Panel B). Coefficient differences in nonagricultural decisions are statistically significant in two of four domains (own employment and minor household expenditures). Similar to women in Ghana, this indicate there are men relate taking sole decisions with higher autonomous feelings across a number of different domains across productive activities.

5.4. Sensitivity analysis to couples’ agreement on decision making Figs. 4 and 5 provide breakdowns of decision-making outcomes within each domain according to whether men and women in the same household agree or disagree on who normally takes decisions. Overall, agreement between couples is much more common in Ghana (ranging from 67 to 82 percent of couples) compared to Bangladesh (ranging from 6 to 64 percent of couples). The predominant form of disagreement is similar across both countries. When couples disagree, it tends to be the case that the woman reports decisions as being made jointly, with the man reporting the same decisions as being made solely by him. Particularly striking are two exceptions to this pattern (expression of religion and daily tasks in Bangladesh) in which more than 90 percent of couples give conflicting reports of decision making, mostly owing to contradictory reports of sole decision making. This is likely less indicative of an actual conflict in couples’ understanding of decision-making within these domains, than it is a reflection that decisions within

these domains tend to be made solely by the individual, without consultation between partners. Given the high frequency of conflicting reports, the lack of significant differences observed between the coefficients associated with sole and joint decision making could be due to converging ambiguities in how decisions are made vis-à-vis the decisionmaking preferences of each individual. We therefore attempt to rule out the possibility that including conflicting responses in our sample biases the results, by re-estimating our model to include interaction effects between sole or joint decision making and agreement between partners (Eq. (2)). Table 4 shows the results. We find that agreement between partners—or the sharing of a common understanding of decision-making power—does have a bearing on how each associates sole and joint decision making with autonomy. Indeed, if couples agree on decision making, the difference in the coefficients associated with sole and joint decision making tends not to be statistically significant. This finding is true for nearly all of our subsamples, regardless of gender, country, or domain. The only exception is personal decisions among men in Bangladesh, where the difference in the coefficients is statistically significant regardless of whether couples agree or disagree. Disagreement between partners also appears to affect the association between sole or joint decision making and autonomy, but in the opposite way as agreement and only under certain circumstances. Among couples in Bangladesh who disagree on decision making, the difference in the coefficients associated with sole and joint decision making is statistically significant for women, regardless of domain; for men, the difference is statistically significant for personal decisions but not for agricultural decisions. In Ghana, differences are not statistically significant among people who disagree with their partner, regardless of gender. In light of these results, our earlier findings may, in fact, be explained by relative differences in the frequency of agreement observed in each country, with higher rates of agreement among

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

107

Fig. 4. Breakdown of decision-making outcomes according to whether men and women in the same household agree or disagree on who normally takes decisions by domain, Bangladesh. Source: Author’s calculations based on 2011–2012 BIHS data. Note: Percentages relative to the total number of households for which we have responses from both partners, and thus, sum to 100 if you include instances of agreement and disagreement between partners. Full results are provided in Table A.14 in the appendix.

couples in Ghana translating to fewer significant differences between the coefficients associated with sole and joint decision making, and vice versa in Bangladesh. This conclusion is somewhat intuitive if we expect that couples who agree on the way decisions are made also spend more time discussing each decision, understanding each other’s motivation for preferences and reconciling differences.

6. Discussion and conclusion Despite widespread agreement on the importance of women’s empowerment, ambiguity exists about how best to define and measure it, particularly when based on intrahousehold decision making (Peterman et al., 2015). While some aggregate indexes have been developed and validated to capture the multidimensional nature of empowerment across different contexts, such as the WEAI (Alkire et al., 2013) and the Sexual Relationship Power Scale (McMahon, Volpe, Klostermann, Trabold, & Xue, 2015; Pulerwitz, Gortmaker, & DeJong, 2000), there is increasing need

for adaptable indicators that can be applied to project-specific domains of interest. Using data on men and women from Bangladesh and Ghana, we analyze whether respondents who report sole decision making in a particular domain tend to experience stronger (or weaker) feelings of autonomous motivation than those who report joint decision making. On aggregate, we find differences between men and women in the significance of association between feelings of autonomous motivation and each of the two decision-making outcomes analyzed. Whereas men and women in Bangladesh tend to clearly distinguish between sole and joint decision making, this is less the case for those in Ghana. In particular, we find that joint decision making on agricultural matters is a stronger predictor of autonomy than sole decision making for women in Bangladesh, whereas in Ghana, neither mode of decision making seems to predict autonomy. This pattern may be attributable to several factors, such as differences in social norms governing the organization of agricultural production, social interactions, and family responsibilities. For example, given the male-dominated farming system and restrictions on women’s mobility in Bangladesh, women may

108

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

Fig. 5. Breakdown of decision-making outcomes according to whether men and women in the same household agree or disagree on who normally takes decisions by domain, Ghana. Source: Author’s calculations based on 2012 GPBS data. Note: Percentages relative to the total number of households for which we have responses from both partners, and thus, sum to 100 if you include instances of agreement and disagreement between partners. Full results are provided in Table A.15 in the appendix.

believe they will achieve better productive outcomes if their partners have some input and decision-making authority over productive domains and thus associate autonomy more strongly with joint decision making. A similar conclusion was reached in a study examining decisions over contraceptive use in Bangladesh, whereby women who were able to engage in ‘‘egalitarian” decision making (defined as involving discussion and agreement with their partners) were more likely to have positive contraceptive use outcomes as compared to women who made decisions independently (Uddin et al., 2017). Our analysis also reveals that the relationship between sole or joint decision making and autonomy varies according to the specific decision domain under scrutiny. Most notably, we find that sole and joint decision making predict autonomy differently depending on whether the decision is agricultural or personal in nature. In particular, results from Bangladesh suggest that as decisions become more personal in nature autonomy begins to be more strongly associated with making decisions by oneself rather than with someone else. We also find that agreement on decision making among couples—evidence of a consistent understanding of decision-making power within the household—does, in certain instances, affect the way in which individuals associate sole or joint decision making with autonomy. How decisions are made—whether solely or jointly—appears to matter more for the prediction of autonomy when couples disagree on decision making than when they agree.

Thus, the level of agreement on decision making may reveal something important about the underlying power dynamics within the household, and as such, is a subject worthy of future attention among researchers trying to measure individual-level empowerment or agency. Grouping together similar domains in our aggregate analysis provided insights into the differences between agricultural and personal decisions that may have been obscured had we followed only a domain-specific approach. Thus, when looking at decisions across multiple domains it may be advantageous to complement domain-specific analysis with analysis based on different groupings of domains to highlight intergroup heterogeneity. It is also important to consider bias due to multiple hypothesis testing, which may result in skewed or spurious results. Considered together, our results show that the degree to which men and women associate different decision-making outcomes with autonomous behavior is idiosyncratic and likely to vary from one context to another, depending on the particular decision being made, as well as sociocultural norms and other local features. These results are corroborated by studies in Ghana and Bangladesh using the WEAI—which is composed, in part, of decision-making and RAI indicators similar to the ones used in our study—that find the degree to which different domains of empowerment matter for favorable nutrition outcomes varies between contexts (Malapit & Quisumbing, 2015; Sraboni, Malapit, Quisumbing, & Ahmed, 2014).

109

G. Seymour, A. Peterman / World Development 111 (2018) 97–112 Table 4 OLS regression results predicting Relative Autonomy Index scores with agreement interaction terms by gender and country. Bangladesh

Ghana

Agricultural decisions

Personal decisions

Agricultural decisions

Variable

Women

Men

Women

Men

Women

Men

Sole decision making (bDISAGREE)

0.034 (0.043) 0.181** (0.060)
0.040 (0.072) 0.029 (0.046)
0.076 (0.077)
0.025y (0.015) 0.358* (0.178) 0.015 (0.069)
0.004 (0.170) 0.085 (0.158) 0.001 (0.002)
0.033 (0.050) —

0.175* (0.079) 0.205* (0.088) 0.002 (0.052)
0.214 (0.226)
0.293 (0.224) 0.004 (0.012) 0.022 (0.124) 0.052 (0.056)
0.111 (0.102)
0.023 (0.135)
0.002 (0.002)
0.021 (0.042) —

0.014 (0.093)
0.125 (0.086) 0.298** (0.097)
0.049 (0.110)
0.035 (0.109) 0.002 (0.012) 0.007 (0.153) 0.013 (0.055) 0.117 (0.146)
0.183 (0.214)
0.000 (0.002)
0.018 (0.041) —

0.341*** (0.089) 0.117 (0.083) 0.533*** (0.119) 0.027 (0.953)
0.221 (0.924)
0.012 (0.011) 0.159 (0.110)
0.005 (0.048)
0.190y (0.096)
0.329 (0.236)
0.002 (0.002) 0.015 (0.038) —

Ethnicity dummies

0.075 (0.189) 0.037 (0.130) 0.067 (0.066) No


0.222y (0.116) 0.197* (0.097) 0.219*** (0.058) No


0.206 (0.134) 0.014 (0.112) 0.085 (0.065) No


0.213* (0.107) 0.133 (0.097) 0.209** (0.065) No

0.008 (0.044)
0.062 (0.054)
0.002 (0.084) 0.031 (0.052) 0.098 (0.061)
0.015 (0.010) 0.174 (0.106) 0.275* (0.138)
0.038 (0.196)
0.232 (0.152) 0.003y (0.002) 0.003 (0.058) 0.134* (0.064)
0.074 (0.077) 0.058 (0.050)
0.132*** (0.038) Yes

0.082 (0.125) 0.126 (0.126) 0.256*** (0.076)
0.065 (0.144)
0.176 (0.147) 0.006 (0.008)
0.039 (0.075)
0.010 (0.102) 0.179 (0.112)
0.179 (0.113) 0.000 (0.002)
0.045 (0.073) 0.133y (0.069) 0.002 (0.077) 0.100* (0.047)
0.076* (0.037) Yes

Adjusted R2 H0: bAGREE = cAGREE (F statistic) H0: bDISAGREE = cDISAGREE (F statistic) Observations

0.021 1.569 5.387* 1558

0.032 2.434 0.472 1950

0.027 0.031 16.568*** 2325

0.078 11.184** 49.519*** 2453

0.028 1.098 1.495 1785

0.031 2.120 0.473 1992

Joint decision making (cDISAGREE) Agreement on decision making Sole decision making  agreement on decision making (bAGREE) Joint decision making  agreement on decision making (cAGREE) Age Age-squared/1000 Completed primary education Higher than primary education Married Age gap: Husband – wife Husband more educated than wife Christian Muslim (log) Area of cultivated land (log) Per capita consumption

Source: Author’s calculations based on 2011–2012 BIHS and 2012 GPBS data. Note: Standard errors clustered at the primary sampling unit level in parentheses. yp < 0.10, *p < 0.05, **p < 0.01, and

Given the strong theoretical ties between the concepts of autonomy and empowerment (Kabeer, 1999; Sen, 1999), our study contributes to the discourse on measuring empowerment—the context in which decision-making indicators are most often applied—and has implications for the broader use of decision-making indicators in development research. We do not, however, empirically test the hypothesis that autonomy and empowerment are correlated, and thus, any linkages we discuss between our study and the measurement of empowerment should be viewed as suggestive, rather than empirically-driven. One of the main implications of our study is to reject the common practice of treating sole and joint decision making as equivalent indicators of individual decision-making power. Instead, measurement of individual decision-making power—and to some extent, women’s empowerment—may be improved by constructing indicators that capitalize on differences in men’s and women’s perceptions of sole and joint decision making. A similar case, based on our results, can be made for constructing indicators of couples’ agreement on decision making, which may reveal something important about the underlying power dynamics within households. We advise all researchers interested in using decision making indicators undertake an analysis of how men and women perceive sole and joint decision making. Our study provides one example of a survey-based approach. Alternative methods, such as qualitative fieldwork, are also possible and may even yield

***

p < 0.001. — data not available.

clearer insights given the nuances involved in individual perceptions of decision making. Results of these analyses may be invaluable as means of calibrating indicators of empowerment to suit specific contexts. Our findings also underscore the importance of considering an expanded set of decision-making domains, which crosscut personal and economic decisions and fall into both traditionally male and female domains within any given cultural setting, to more holistically capture empowerment. Further, although we do not fully interrogate the types of women and men who tend to agree or disagree on decisions, nor the sources of disagreement, a clear implication is the importance of considering who in the household is being asked questions and how the choice of respondent might change the outcomes or conclusions of a given research investigation. Our recommendations are largely in line with the findings of a recent review of current practices for measuring women’s empowerment which recommends that researchers use (1) theory to guide selection of domains and indicators, (2) analytic models which minimize subjective or implicit judgements of what constitutes empowerment, and (3) multiple methods to triangulate findings (Richardson, 2017). Our study reveals how far development research, as a field, still has to go in understanding how generalized measures of autonomy and decision making relate not only to each other but also to broader development objectives, such as empowerment and agency.

110

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

Several methodological limitations, including those of commonly used indicators such as decision making, extensively discussed in earlier sections, should be kept in mind when considering our results. One is the interpretation and validation of the RAI in development settings—which has been questioned in the literature and also shows some weaknesses according to our own tests of validity (see Appendix). In addition, the RAI deeply relies on each person’s self-perception of autonomy. Thus, a person may feel him- or herself to be autonomous—and thus register a high RAI score—yet outside observers may have reason to question the accuracy of the self-assessment. A person’s values may be shaped so deeply by the circumstances of his or her life that he or she cannot conceive of what it would mean to live a truly autonomous life (see discussion of adaptive preferences in Nussbaum (2000) and Sen (1985, 1990)). Indeed, Alkire and Chirkov (2007) found that among women living in Kerala, India, adaptive preferences often arose with respect to their household responsibilities, which many felt to be an integral part of their identity as good and dutiful wives and mothers. Another limitation stems from the inherent difficulty individuals may have in disentangling the diverse motivations behind their actions, inasmuch as the RAI scores we observe may overestimate men’s and women’s actual levels of autonomy. For example, using qualitative methods, Lentz (2018) shows that women in Bangladesh exhibit ‘‘burdened agency,” whereby they trade off sub-optimal decisions in one domain (e.g., food security and nutrition) in an effort to negotiate better outcomes in another (e.g., intimate partner violence). Unpacking the mechanisms and meaning attached to decision making to determine the extent to which this may be true in our data would require similar qualitative work. In addition, given the response structure used in our decision-making questions, we are not able to fully analyze differences between joint decisions made between partners and those that may or may not include both partners as well as someone else—for example, decisions made with a brother or spouse may be very different than decisions made with someone outside the household, like an employer or village leader. However, given that very few respondents report that decisions are made jointly with someone else inside or outside the household, we do not believe the inability to fully differentiate joint decisions is likely to strongly influence our results. Finally, the GPBS does not include domains of personal decisions and, further, the sample is made up of dual-adult households selected specifically for the purpose of Feed the Future monitoring and evaluation. Thus, our ability to determine whether the pattern of differences found in Bangladesh between personal and agricultural decisions extends to other settings or to make population-level generalizations is limited. Measuring gender relations is complex, and encompasses not only the division of decision making and resources, but also encompasses ideological concepts of ‘‘abilities, attitudes, desires, personality traits, behavior patterns” among others (Agarwal, 1997, p. 1). Despite the limitations of current constructions and analysis of decision-making indicators, we document large gender disparities in average reporting on decision making and autonomy—whereby men consistently report higher sole decision making and autonomy than women. These disparities underscore the need for continued investigation into both programmatic and policy actions to reduce gender gaps, accompanied by rigorous analysis and innovation in measurement. Particularly promising are economic-based programs which have the potential to increase women’s control of resources, labor force participation and financial standing, with implications for intrahousehold bargaining (Buvinic´ & Furst-Nichols, 2014; Duflo, 2003; Fafchamps & Quisumbing, 2005). Based on the still unresolved sources of measurement and other biases in traditional decision-making indicators, it is unlikely that such measures alone will be able to

accurately capture culturally specific gender barriers in the way necessary to advance the field. For example, utilizing reports of decision making from a single household member—as is standard practice in many surveys—may mask large variations in autonomy and, perhaps, in the underlying concept of empowerment. Several lines of research are already underway investigating alternative measures of autonomy and empowerment, often utilizing vignettes and randomly controlled trials (e.g., Horne, Dodoo, and Dodoo (2013); Malapit, Sproule, and Kovarik (2017)). We see vignettes as a particularly promising approach that may provide greater accuracy than traditional survey questions by eliciting more truthful responses in the face of respondents’ potential reluctance to discuss such a sensitive topic as intrahousehold power. The next step is to test variations in the choice of indicator or method to understand if shifts in decision making and autonomy correlate or explain favorable changes in outcomes. Although there have been some promising examples of research exploring variation in indicator choice or cooperation within the household (e.g., Basu and Koolwal (2005) in India; Peterman et al. (2015) in Ecuador, Uganda and Yemen; McCarthy and Kilic (2017) in Malawi; Ambler et al. (2016) in Bangladesh), greater investments by researchers—unpacking, interrogating, and innovating around measurement in different contexts—are still needed to understand how measurement matters for making gendered-programmatic and policy recommendations, and to better contribute to reducing gender inequalities and enhancing the empowerment and agency of all individuals. Ideally, this work should be carried out in the context of longitudinal data or experimental (or quasiexperimental) designs, which allow for more robust conclusions drawn from causal relationships. Acknowledgements Funding support for this study was provided by the CGIAR Research Program on Policies, Institutions, and Markets (PIM). We are grateful for helpful discussions with Agnes Quisumbing, Cheryl Doss, Caitlin Kieran, and participants at the 2016 Midwest International Economic Development Conference. We would also like to thank three anonymous reviewers for their valuable comments and suggestions on the earlier versions. Both authors participated equally in the research and article preparation and have approved the final article. Conflicts of interest None. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.worlddev.2018.06. 027. References Agarwal, B. (1997). ‘‘Bargaining” and gender relations: Within and beyond the household. Feminist Economics, 3(1), 1–51. Alkire, S. (2005). Subjective quantitative studies of human agency. Social Indicators Research, 74(1), 217–260. Alkire, S. (2007). Measuring agency: Issues and possibilities. Indian Journal of Human Development, 1(1), 169–178. Alkire, S., & Chirkov, V. (2007). Measuring agency: Testing a new indicator in Kerala. In V. Pillai & S. Alkire (Eds.), Measuring individual agency or empowerment: A study in Kerala. Kerala, India: Centre for Development Studies Thiruvananthapuram. Alkire, S., Meinzen-Dick, R., Peterman, A., Quisumbing, A. R., Seymour, G., & Vaz, A. (2013). The women’s empowerment in agriculture index. World Development, 52, 71–91.

G. Seymour, A. Peterman / World Development 111 (2018) 97–112 Allendorf, K. (2007). Do Women’s land rights promote empowerment and child health in Nepal? World Development, 35(11), 1975–1988. Ambler, K., Doss, C. R., Kieran, C., & Passarelli, S. (2016). He says, she says: exploring patterns of spousal agreement in Bangladesh (IFPRI Discussion Paper 1616). Washington, DC: International Food Policy Research Institute. Anderson, S., & Eswaran, M. (2009). What determines female autonomy? Evidence from Bangladesh. Journal of Development Economics, 90, 179–191. Basu, A. M., & Koolwal, G. B. (2005). Two concepts of female empowerment: Some leads from DHS data on women’s status and reproductive health. In S. Kishor (Ed.), a focus on gender: Collected papers on gender using DHS data (pp. 15–53). Calverton, MD: OCR Macro. Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B (Methodological), 57(1), 289–300. Bonilla, J., Zarzur, R. C., Handa, S., Nowlin, C., Peterman, A., Ring, H., & Seidenfeld, D. (2017). Cash for women’s empowerment? A mixed-methods evaluation of the government of Zambia’s Child Grant Program. World Development, 95, 55–72. Buvinic´, M., & Furst-Nichols, R. (2014). promoting women’s economic empowerment: What works? The World Bank Research Observer, 31(1), 59–101. https://doi.org/10.1093/wbro/lku013. Cheong, Y. F., Yount, K. M., & Crandall, A. (2017). Longitudinal measurement invariance of the women’ s agency scale. Bulletin de Methodologie Sociologique, 134(1), 24–36. Chirkov, V., Ryan, R. M., Kim, Y., & Kaplan, U. (2003). Differentiating autonomy from individualism and independence: A self-determination theory perspective on internalization of cultural orientations and well-being. Journal of Personality and Social Psychology, 84(1), 97–110. Clark, G. (1994). Onions are my husband: Survival and accumulation by West African Market Women. Chicago, IL: University of Chicago Press. de Brauw, A., Gilligan, D. O., Hoddinott, J., & Roy, S. (2014). The impact of Bolsa Família on Women’s Decision-Making Power. World Development, 59, 487–504. Deci, E. L., & Ryan, R. M. (2012). Motivation, personality, and development within embedded social contexts: An overview of self-determination theory. In R. M. Ryan (Ed.), Oxford handbook of human motivation (pp. 85–107). Oxford, UK: Oxford University Press. Deere, C. D., & Twyman, J. (2012). Asset ownership and egalitarian decision making in dual-headed households in Ecuador. Review of Radical Political Economics, 44 (3), 313–320. Donald, A., Koolwal, G., Annan, J., Falb, K., & Goldstein, M. (2017). Measuring Women’s Agency. World Bank Policy Research Working Paper 8148. Doss, C. R. (2002). Men’s crops? Women’s crops? The gender patterns of cropping in Ghana. World Development, 30(11), 1987–2000. Doss, C. R. (2013). Intrahousehold bargaining and resource allocation in developing countries. World Bank Research Observer, 28(1), 52–78. Doss, C. R., Deere, C. D., Oduro, A. D., Swaminathan, H., Suchitra, J. Y., Lahoti, R., ... Hillesland, M. (2011). The gender asset and wealth gaps: Evidence from Ecuador, Ghana, and Karnataka, India. Bangalore: Indian Institute of Management Bangalore. Doss, C. R., Kim, S. M., Njuki, J., Hillenbrand, E., & Miruka, M. (2014). Women’s individual and joint property ownership: Effects on household decisionmaking (IFPRI Discussion Paper 1347). Washington, DC: International Food Policy Research Institute. Doss, C. R., & Meinzen-Dick, R. (2015). Collective action within the household: Insights from natural resource management. World Development, 74, 171–183. Duflo, E. (2003). Grandmothers and granddaughters: Old-age pensions and intrahousehold allocation in South Africa. The World Bank Economic Review, 17 (1), 1–25. https://doi.org/10.1093/wber/lhg013. Eswaran, M., & Malhotra, N. (2011). Domestic violence and women’s autonomy in developing countries: Theory and evidence. Canadian Journal of Economics, 44 (4), 1222–1263. Fafchamps, M., & Quisumbing, A. (2005). Assets at marriage in rural Ethiopia. Journal of Development Economics, 77(1), 1–25. https://doi.org/10.1016/J. JDEVECO.2004.02.003. Gammage, S., Kabeer, N., & Rodgers, Y. (2016). Voice and agency: Where are we now? Feminist Economics, 22(1), 1–29. Gelman, A., & Stern, H. (2006). The difference between ‘‘significant” and ‘‘not significant” is not itself statistically significant. The American Statistician, 60(4), 328–331. Handa, S., Peterman, A., Davis, B., & Stampini, M. (2009). Opening up Pandora’s Box: The effect of gender targeting and conditionality on household spending behavior in Mexico’s Progresa Program. World Development, 37(6), 1129–1142. https://doi.org/10.1016/J.WORLDDEV.2008.10.005. Hanmer, L., & Klugman, J. (2016). Exploring women’s agency and empowerment in developing countries: Where do we stand? Feminist Economics, 22(1), 237–263. Heath, R. (2014). Women’s access to labor market opportunities, control of household resources, and domestic violence: Evidence from Bangladesh. World Development, 57, 32–46. Heckert, J., & Fabic, M. S. (2013). Improving data concerning women’s empowerment in Sub-Saharan Africa. Studies in Family Planning, 44(3), 319–344. Horne, C., Dodoo, F. N., & Dodoo, N. D. (2013). The shadow of indebtedness: Bridewealth and norms constraining female reproductive autonomy. American Sociological Review, 78(3), 503–520. Ibrahim, S., & Alkire, S. (2007). Agency and empowerment: a proposal for internationally comparable indicators. Oxford Development Studies, 35(4), 379–403.

111

Jejeebhoy, S. J., & Sathar, Z. A. (2001). Women’s autonomy in India and Pakistan: The influence of religion and region. Population and Development Review, 27(4), 687–712. Johnson, K. B., & Rosell, P. D. (2015). Assessing the cognitive validity of the women’s empowerment in agriculture index instrument in the haiti multi – sectoral baseline survey. Social Practice, 8(3), 1–6. Kabeer, N. (1999). Resources, agency, achievements: Reflections on the measurement of women’s empowerment. Development and Change, 30(3), 435–464. https://doi.org/10.1111/1467-7660.00125. Kieran, C., Sproule, K., Doss, C. R., Quisumbing, A. R., & Kim, S. M. (2015). Examining gender inequalities in land rights indicators in Asia. Agricultural Economics, 46, 1–20. Kishor, S., & Subaiya, L. (2008). Understanding women’s empowerment: a comparative analysis of Demographic and Health Surveys (DHS) data (DHS Comparative Reports 20). Calverton, MD: Macro International. Klein, E. (2014). Psychological agency: Evidence from the Urban Fringe of Bamako. World Development, 64, 642–653. Klein, E. (2016). Women’s agency and the psychological domain: Evidence from the Urban Fringe of Bamako, Mali. Feminist Economics, 22(1), 106–129. Kline, P. (1994). An easy guide to factor analysis. New York, NY: Routledge – Taylor & Francis Group. Lentz, E. C. (2018). Complicating narratives of women’s food and nutrition insecurity: Domestic violence in rural Bangladesh. World Development, 104, 271–280. https://doi.org/10.1016/J.WORLDDEV.2017.11.019. Mabsout, R., & van Staveren, I. (2010). Disentangling bargaining power from individual and household level to institutions: Evidence on women’s position in Ethiopia. World Development, 38(5), 783–796. Malapit, H. J. L., & Quisumbing, A. R. (2015). What dimensions of women’s empowerment in agriculture matter for nutrition in Ghana? Food Policy, 52, 54–63. Malapit, H. J. L., Sproule, K., & Kovarik, C. (2016). Using cognitive interviewing to improve the women’s empowerment in agriculture index survey instruments (IFPRI Discussion Paper 1564). Washington, DC: International Food Policy Research Institute. Malapit, H. J. L., Sproule, K., & Kovarik, C. (2017). Using cognitive interviewing to improve the Women’s Empowerment in Agriculture Index survey instruments: Evidence from Bangladesh and Uganda. Journal of Gender, Agriculture and Food Security, 2(2), 1–22 https://doi.org/10.19268/JGAFS.222017.1. Malhotra, A., Schuler, S. R., & Boender, C. (2002). Measuring women’s empowerment as a variable in international development (Background Paper Prepared for the World Bank Workshop on Poverty and Gender: New Perspectives). Washington, DC: World Bank. McCarthy, N., & Kilic, T. (2017). Stronger together: Intra-household cooperation and household welfare in Malawi (Policy Research Working Paper 8043). Washington, DC: World Bank. McMahon, J. M., Volpe, E. M., Klostermann, K., Trabold, N., & Xue, Y. (2015). A Systematic review of the psychometric properties of the sexual relationship power scale in HIV/AIDS research. Archives of Sexual Behavior, 44(2), 267–294. Nussbaum, M. C. (2000). Women and human development: The capabilities approach. Cambridge, UK: Cambridge University Press. Oduro, A. D., Boakye-Yiadem, L., & Baah-Boateng, W. (2012). Asset ownership and egalitarian decision-making among couples: Some evidence from Ghana (The Gender Asset Gap Project Working Paper 14). Bangalore, India: Indian Institute of Management Bangalore. Peterman, A., Schwab, B., Roy, S., Hidrobo, M., & Gilligan, D. O. (2015). Measuring women’s decisionmaking: Indicator choice and survey design experiments from cash and food transfer evaluations in Ecuador, Uganda, and Yemen (IFPRI Discussion Paper 1453). Washington, DC: International Food Policy Research Institute. Pulerwitz, J., Gortmaker, S. L., & DeJong, W. (2000). Measuring sexual relationship power in HIV/STD research. Sex Roles, 42(7), 637–660. Richardson, R. A. (2017). Measuring women’s empowerment: A critical review of current practices and recommendations for researchers. Social Indicators Research. in press. Ryan, R. M., & Connell, J. P. (1989). Perceived locus of causality and internalization: Examining reasons for acting in two domains. Journal of Personality and Social Psychology, 57(5), 749–761. Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78. Sen, A. (1985). Well-being, agency and freedom: The Dewey Lectures 1984. The Journal of Philosophy, 82(4), 169–221. Sen, A. (1990). Gender and cooperative conflicts. In I. Tinker (Ed.), Persistent inequalities: women and world development (pp. 123–149). New York, NY: Oxford University Press. Sen, A. (1999). Development as freedom. New York: Alfred A. Knopf. Seymour, G. (2017). Women’s empowerment in agriculture: Implications for technical efficiency in Rural Bangladesh. Agricultural Economics, 48(4), 513–522. Sraboni, E., Malapit, H. J. L., Quisumbing, A. R., & Ahmed, A. U. (2014). Women’s empowerment in agriculture: What role for food security in Bangladesh? World Development, 61, 11–52. Twyman, J., Useche, P., & Deere, C. D. (2015). Gendered perceptions of land ownership and agricultural decision-making in Ecuador: Who are the farm managers? Land Economics, 91(3), 479–500. Uddin, J., Hossin, M. Z., & Pulok, M. H. (2017). Couple’s concordance and discordance in household decision-making and married women’s use of modern

112

G. Seymour, A. Peterman / World Development 111 (2018) 97–112

contraceptives in Bangladesh. BMC Women’s Health, 17, 107–117. https://doi. org/10.1186/s12905-017-0462-3. UNDP (2015). Human development report 2015: Work for human development. New York: United Nations Development Programme. Vaz, A., Alkire, S., Quisumbing, A. R., & Sraboni, E. (2013). Measuring autonomy: Evidence from Bangladesh (OPHI Research in Progress 38a). Oxford, UK: Oxford Poverty & Human Development Initiative. Vaz, A., Pratley, P., & Alkire, S. (2016). Measuring women’s autonomy in chad using the Relative Autonomy Index. Feminist Economics, 22(1), 264–294.

White, S. C. (1992). Arguing with the Crocodile: Gender and class in Bangladesh. London, UK: Zed Books Ltd. World Bank (2013). Country partnership strategy for The Republic of Ghana for the period FY13–FY16. Washington, DC: World Bank. World Bank (2016a). Country partnership framework for Bangladesh for the period FY16–FY20. Washington, DC: World Bank. World Bank (2016b). World Development Indicators Retrieved from http://data.worldbank.org/data-catalog/world-development-indicators.