Policy Research Working Paper 10851 When Does Decision-Making Reflect Agency? Evidence from the Rural Philippines Aries Arugay Aletheia Donald Forest Jarvis Hillary C. Johnson Aletheia Valenciano East Asia and Pacific Region & A verified reproducibility package for this paper is Africa Region available at http://reproducibility.worldbank.org, July 2024 click here for direct access. Policy Research Working Paper 10851 Abstract Decision-making is often used as a proxy for agency—the make personal decisions if desired are strongly associated ability to set goals and act on them—although there are with the RAI for both genders. The quantitative and qual- several theoretical critiques of this approach. Using unique itative data indicate that these concepts better capture the data from the rural Philippines, this paper empirically tests ability to make choices in line with one’s personal goals, the extent to which different aspects of decision-making are while being a decision-maker instead reflects being respon- correlated with the Relative Autonomy Index, a measure sible for the outcome or managing the execution of a task, of agency that has been validated for use in lower-income often in the face of limited options. The findings caution countries. Being a decision-maker (as asked in common against focusing on being a decision-maker as a sole indi- survey questions) is only weakly related to the Relative cator of agency and have practical implications for both Autonomy Index for women, and not at all for men. Having conceptualizing and measuring agency. input into decisions and, to a greater extent, the ability to This paper is a product of the East Asia and Pacific Region Gender Innovation Lab and the Africa Region Gender Innovation Lab. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at fjarvis1@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank.org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team When Does Decision-Making Reflect Agency? Evidence from the Rural Philippines Aries Arugay, Aletheia Donald, Forest Jarvis, Hillary C. Johnson and Aletheia Valenciano* Keywords: measurement, intra-household, decision-making, gender, Philippines JEL codes: C8, D13, J16, I32, O12 ∗ Arugay: aaarugay@up.edu.ph; Donald: adonald@worldbank.org; Jarvis: fjarvis1@worldbank.org; Johnson: hjohnson1@worldbank.org; Valenciano: avalenciano@up.edu.ph. The authors would like to thank the study participants for sharing their time and information with us, Innovations for Poverty Action for excellent support with data collection, and Connie Bayudan-Dacuycuy, Sundas Liaqat, Elizaveta Perova, Greg Seymour, Peter Srouji, and the participants of the World Bank East Asia and Pacific Chief Economist Office Seminar for helpful comments. This paper is a product of the World Bank’s East Asia and Pacific Gender Innovation Lab, the Africa Gender Innovation Lab and of the Measures for Advancing Gender Equality (MAGNET) initiative. We gratefully acknowledge financial support from the World Bank's Umbrella Facility for Gender Equality, in partnership with the Bill and Melinda Gates Foundation (INV-005620). The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 1. Introduction The act of being involved in making decisions has often been used as a proxy for individual agency, that is—one’s ability to set goals and act on them. This is true for the measurement of agency in both children and adults across fields and is especially prominent within the women’s empowerment literature (Alkire et al. 2013; Cavazzoni et al. 2022). Women’s self-reported status as decision-maker in a variety of activities is assumed to mean that they have more agency in that particular domain. 1 However, there is little empirical evidence on whether this measure is indeed reflective of agency or which aspects of decision-making may be more closely related to agency. Despite the relatively sparse empirical research up to this point, accurately capturing individual agency is essential to measure progress toward development goals, as enabling individuals to exercise their reasoned agency is both the principal means and the primary ends of development (Sen 1999). While numerous studies have linked decision-making modes—such as women’s involvement in decision-making—to positive outcomes within the household, there are multiple theoretical and methodological critiques of using being a decision-maker as a proxy for agency. Individuals may exercise their agency by having input into decisions without necessarily having to be the sole or final decision-maker. Focusing on the decision-maker may be too restrictive, and not capture the discussions and negotiations that occur before a final decision is made (Donald et al. 2020). In addition, being a decision-maker in any given realm may be due to socially prescribed roles rather than reflective of greater bargaining power or agency (Agarwal 1994; Eissler et al. 2021; Heckert and Fabic 2013). Moreover, day-to-day decision-making may be more about “task management” (including being held responsible for the outcome) and minor choices rather than consequential decisions, indicating a high degree of cognitive labor but not necessarily agency (Daminger 2019). This is particularly the case in the face of scarcity, when decisions are made over restricted choice sets. This paper empirically tests the extent to which different aspects of decision-making are reflective of agency among a sample of 570 farming households in the southern Philippines. We capture agency through the Relative Autonomy Index (RAI), a psychometric scale proposed by 1 See, for example, De Brauw et al. 2014, Ghose et al. 2017, Grillo 2018, Karimli et al. 2021, and Majlesi 2016. 2 Ryan and Deci (2000) that has been validated for use in lower-income countries. To capture three theoretical aspects of decision-making, we use three different decision-making measures to construct our outcome variables: (i) a binary variable of whether or not an individual reports being a decision-maker, (ii) whether or not the respondent has input in the decision-making process, and (iii) whether or not the respondent could make their own personal decisions if they wanted to. 2 Because decision-making can be subject to differing interpretations of what it means to make a decision (Liaqat et al. 2021), we also test a binary variable indicating that spouses agree the respondent is a decision-maker. Though our analysis is focused on agricultural decision-making and agency, most rural households in lower-income countries rely on agriculture as their primary source of income, and agricultural decision-making patterns in our households match those of overall decision-making patterns. Finally, we supplement our quantitative analysis with the results of qualitative interviews with farmers similar to our quantitative sample, through which couples provided more in-depth information on the decision- making process within their households, as well as their preferences regarding intra-household bargaining and their conceptualizations of what it means to be a decision-maker. We find that making decisions is not strongly correlated with agency; however, having input into decisions and the ability to make personal decisions if one wanted to show a strong association with agency for both men and women. A one standard deviation increase in the latter variable is associated with an increase of 0.19 standard deviations in RAI for women and 0.13 standard deviations for men. Self-reported status as a decision-maker is only weakly associated with agency for women, and not at all for men. When the three indicators are compared against each other in a multivariate regression, we find that only the ability to make personal decisions if desired maintains a strong relationship with agency. Our results are not due to decision-making being too restrictive a measure (only accounting for the main person shaping the choice) or a particularly noisy measure (i.e., prone to measurement error due to diverging interpretation across respondents). Indeed, we find that having input over decisions is less common than being a decision-maker. Our results also show that making decisions alone is not more strongly associated with agency than making decisions jointly, and that both spouses agreeing that an individual is a decision-maker does not meaningfully strengthen the relationship to agency. 2 These are drawn from the Women’s Empowerment in Agriculture Index (Alkire et al. 2013). 3 Our quantitative findings suggest that being a decision-maker does not always come with greater agency or even a high degree of input. This ‘decision-making without meaningful input’ may occur because of external constraints, such as limited choice in the respondent’s environment due to poverty, remoteness or other structural barriers. However, it can also manifest because of constraints internal to the household, with women in particular reporting being decision-makers even in cases where they are not able to sway their husband’s preference. Our qualitative work identifies three ways in which respondents may consider themselves as decision-makers while having a low degree of agency. First, respondents reported that the choice sets available to them were often constrained by economic conditions or other limitations. Even if respondents were able to make decisions, they were not necessarily able to make choices that were in line with their values or aspirations. Second, we find a strong normative preference within couples in our sample for consultative decision-making, even when spouses are only perfunctorily consulted and have no true input in the decision. This joint decision-making was seen as a way of sharing responsibility between spouses in order to avoid backlash for the consequences of decisions made without consultation. Third, we find that respondents reported themselves as decision-makers when engaging in minor, day-to-day task management rather than major, consequential decisions. In this last conceptualization, decision-making indicates that individuals are responsible for making sure tasks are completed—and the cognitive burden that accompanies this responsibility—even if these tasks do not align with their preferences and goals. Taken together, our quantitative and qualitative findings point to the fact that decision- making is conceptually and empirically a weak proxy for agency. In our context, it captures responsibility for the outcome of a decision—including managing a task—without meaningful input, often to conform to social norms or to make the only possible decision given limited external options. Other measures that are equally easy to ask respondents, but are less commonly included in questionnaires, perform better as proxies for men’s and women’s agency. Capturing the amount of input respondents may have in the decision-making process and the extent to which they can make their own personal decisions narrows the scope of decision-making toward cases where respondents can change the outcome of a decision to match their own goals and values. Our results also highlight the importance of understanding the local context, including 4 social norms and preferences regarding decision-making and intra-household bargaining, when analyzing quantitative data meant to capture social phenomena. Our paper adds to the increasing body of literature on decision-making and agency. We provide a quantitative and qualitative analysis of the measurement of decision-making in the Philippines context, including the first detailed examination of the validity of various decision- making measures in the country. We contribute to the broader literature on agency by providing a unique comparison of a series of decision-making measures, supplemented by in-depth qualitative work explaining why some aspects of decision-making are more predictive of agency than others. Thus, while we echo the findings of other researchers that caution is necessary when interpreting data on decision-making, we also offer a way forward: the addition of one or two alternative variables may add a great deal of predictive value for researchers wishing to measure decision-making in a way that is reflective of agency. The remainder of this paper proceeds as follows. Section 2 provides conceptual grounding for the concepts discussed in this paper, particularly empowerment, agency, and autonomy. Section 3 provides additional information about the local context and describes our data and empirical strategy. Section 4 provides our results, and Section 5 discusses the implications of our findings. 2. Conceptual Framework 2.1 Agency and decision-making Agency is the “ability to define one’s goals and act on them” (Kabeer 1999). It is a core component of the process of empowerment 3 and is crucial for allowing individuals to benefit from development. Definitionally, agency requires the ability to define goals in line with one’s own values and the ability to make strategic life choices—i.e., the ability to act on goals. The ability to make strategic life choices is most commonly captured through an individual’s involvement in decision-making within their households and/or communities. For example, the Demographic and Health Survey question on “who usually makes major decisions over large 3 Defined as “the process by which those who have been denied the ability to make strategic life choices acquire such an ability” (Kabeer 1999, p. 437). See Appendix Figure A1 for a graphical depiction of the different components of empowerment. 5 household purchases?” is taken as a measure of ability to make strategic choices in the household. However, capturing agency through the act of making decisions has multiple drawbacks (Gammage, Kabeer, and Rodgers 2013). First, research has shown that making decisions is a fuzzy concept, and it can be interpreted by respondents in different ways. This can introduce measurement error that limits the extent to which being a decision-maker is a reliable proxy for agency. Indeed, spousal surveys almost universally find high levels of discrepancy between spouses’ reports of decision-making (e.g., Ambler et al. 2020; Annan et al. 2019; Ghuman, Lee, and Smith 2006; Liaqat et al. 2021; Twyman, Useche, and Deere 2015), suggesting that these issues are widespread regardless of region or context. Second, decision-making may be too restrictive of a measure if it omits ways in which respondents are able to exercise their agency and influence the outcome of decisions even if they do not consider themselves final decision-makers. This view conceptualizes decision-making as a complex, iterative process between individuals with different access to social and economic power (Agarwal 1994), and highlights that even if women are not able to make the final decision in a certain domain, they may exercise their agency in more subtle ways during the decision- making process to change the outcome, or publicly acquiesce to men’s decisions but privately make their own choices (Kabeer 1999). Moreover, if responsibility over a certain domain is normatively expected to fall on women, a woman’s choice not to be the primary decision-maker may in fact be indicative of higher agency (Agarwal 1997). Capturing only the identity of the decision-maker may thus lead to agency being underestimated, particularly if respondents do not consider mere involvement in the decision-making process as sufficient to make one a decision- maker. Alternatively, rather than being unnecessarily restrictive, decision-making as a concept may be a poor proxy for agency if it includes cases where respondents consider themselves decision-makers but cannot make decisions in line with their goals and values. Decision-making would then be too broad of a concept. Indeed, in Uganda, Acosta et al. (2020) find that there is a great deal of heterogeneity between couples in what is considered a “joint” decision, but that these are virtually always dominated by men, limiting the empowerment potential of joint decision-making. Individuals may also be involved in the decision-making process, or choose to avoid making sole decisions, to avoid creating discord or incurring blame (Hindin and Adair 6 2002, Ebrahim and Atteraya 2019). Participation in decision-making may thus be a form of risk reduction (Donald et al. 2023), even if the ability to influence the decision is limited. Decision-making may also be an imprecise measure of agency if the decisions being made are misaligned with individuals’ goals and preferences, because of economic constraints or gender norms (Heckert and Fabic 2013). If decision-making is constrained by poverty, norms, or other structural barriers, then its empowerment potential may be limited (Gammage, Kabeer and Rodgers 2016). Moreover, individuals may be decision-makers in a given domain due to prescribed gender roles, even if they would prefer not to be involved. For instance, men’s ability to make decisions in a male-coded domain such as agriculture may not necessarily be indicative of higher agency if they specialize in that domain due to socialized gender roles and are otherwise constrained in their choices. In cases where the available choices are limited, involvement in decision-making may indicate “task management” (including responsibility for the outcome) rather than the ability to make consequential decisions that are in line with one’s goals and values. Decision-making may thus come with a high cognitive load (Adamkovič and Martončik 2017), particularly since trade- offs between different choices can be emotionally difficult (Battman, Luce, and Payne 1998; Luce 1998). More broadly, beyond involvement in decisions, the relationship between decision- making and agency depends on the types of decisions being made. If this is the case, then measurement of agency may benefit from alternative variables that narrow the scope of decision- making to reflect agency rather than simply task management. 2.2 Set Up Using the Relative Autonomy Index (RAI) as a benchmark for agency, 4 we first compare its relationship to being a decision-maker. We then explore the extent to which alternative measures of decision-making (agreement on decision-maker status within the couple, having input into decisions and being able to make personal decisions if desired) that address the conceptual concerns articulated in Section 2.1 correlate with agency as measured by the RAI. 4 Within psychology and philosophy, autonomy relates to being a causal agent over one’s life. Specifically, a person is autonomous when her behavior is experienced as willingly enacted and when she fully endorses the actions in which she is engaged and/or the values expressed by them. Along these lines, an autonomous individual is “able to act on one’s values and goals”—which perfectly overlaps with agency, although individuals’ ability to initiate transformative changes in their environment is usually not included under autonomy as it sometimes is under agency. 7 The RAI captures the extent to which choices are made from a sense of agency. Agency is conceptualized as a continuum, ranging from the most agentic form—intrinsic motivation, where behaviors are motivated by their sense of fulfillment—to the least agentic form—external motivation, where behavior is motivated by an external authority, fear of punishment, or rule compliance. Introjected reasons fall between these two extremes and consist of esteem-based pressures to act such as avoidance of guilt and shame or concerns about approval (Ryan and Connell 1989). 5 The scale has been validated in multiple contexts in lower-income countries, including in Bangladesh (Seymour and Peterman 2018; Vaz et al. 2013), Ghana (Seymour and Peterman 2018) and Chad (Vaz, Pratley, and Alkire 2016). Some versions administer the scale using anchoring vignettes, which have been validated in ten countries in Sub-Saharan Africa and South Asia (Horne, Dodoo, and Dodoo 2013; Malapit et al. 2019; Quisumbing et al. 2022; Sproule and Kovarik 2014) and have been used in the Philippines (Malapit et al. 2020). Results from Uganda suggest that vignettes may aid in understanding the meanings of different RAI subscales (Donald et al. 2017; Sproule and Kovarik 2014). To the best of our knowledge, this approach has previously only been used in one study. Seymour and Peterman (2018) use the RAI as a benchmark for evaluating respondents’ preferences for sole versus joint decision-making as expressions of agency. They argue that the relative strength of association between the RAI and different modes of decision-making is indicative of preferences—in other words, if a woman prefers to make sole decisions over her children’s education but instead makes them jointly with her spouse, her RAI score would be lower as it does not reflect her own goals and values (Seymour and Peterman 2018). Similarly, if a given decision-making variable is capturing autonomous decision-making, we expect that it will be associated with the RAI. Furthermore, given that decision-making and motivational autonomy are integral and highly related elements of the broader concept of agency (Donald et al. 2020; see Figure A1), we argue that decision-making variables should be considered as indicators of agency only if they show a significant association with RAI. To better understand respondents’ definitions, norms and preferences surrounding these different aspects of decision-making, we conduct qualitative work. This approach is similar to Acosta et al. (2020), who combine quantitative data with qualitative interviews and participant 5 While further divisions are possible, the above three categories of motivations are most used in surveys in the developing world, including the WEAI. 8 observation to shed light on the ways in which joint decision-making is understood by their target population. 3. Context and Empirical Strategy 3.1 Country context The Philippines ranks 101st out of 170 countries in the Gender Inequality Index. 6 Pre- colonial Philippine society was considered highly gender-equal, with women given equal social and legal power to men, but patriarchal norms became more engrained in the Spanish era (Medina 2001). In the modern day, women have higher overall educational attainment than men, with parents preferentially favoring daughters over sons with regards to education (Okabe 2016) while sons are preferred to inherit land (Estudillo, Quisumbing, and Otsuka 2001). However, women have a relatively low labor force participation rate despite high levels of education, in part due to gendered expectations of women’s responsibilities in the household, which are not captured in traditional measures of economic activity (Abrigo and Francisco-Abrigo 2019). A review of the social science literature on decision-making in the Philippines reveals complex dynamics of gender norms and relations. While the man is stated to be the head of the household, in practice many households may be dual-headed, as women can be active participants in decision-making and household management and have the power to revoke their husbands’ commitments (Illo 1989)— though gender roles are complex and there is a great deal of variation between families (Medina 2001). Joint, consultative decision-making is the most common manner of decision-making in Filipino families, with sole decision-making generally reserved for smaller decisions or areas where one spouse is perceived to have decision-making priority (David 1994). However, ultimate decision-making power usually skews more towards one spouse. Bautista (1977) finds that, while women control money, men still have the final say in how money is spent. Contado (1981) finds that joint decision-making is the most prevalent form of decision-making, but that farm decisions tilt toward the husband while household decisions tilt towards the wife. She finds a strong norm against conflict, with husbands more likely to concede to their wives in order to end a conflict, even as they are still considered the head of the household. 6 http://data.un.org/DocumentData.aspx?q=Gender+inequality+indexandid=471 9 In agricultural households, decision-making power tends to skew heavily towards men, with agriculture remaining a largely male-dominated sector (Tapia et al. 2019). Land is considered “conjugal property”, with both spouses having equal legal rights, although the Family Code of the Philippines states that in the case of disagreement over conjugal property that the husband’s decision will be final (World Bank 2012). Women commonly take an active role in agriculture, albeit with different responsibilities than men (Akter et al. 2017; Malapit et al. 2020). While women often possess a great deal of agricultural knowledge, this may be discounted by their husbands and themselves (Christie, Parks, and Mulvaney 2016). In a study of agricultural households in rural Mindanao Parks, Christie, and Baganes (2015) find that women’s role in agriculture is frequently constrained by patriarchal gender norms and social roles, with women expected to be generalists, managing both household and agricultural affairs, while men specialize in agriculture. Despite these specificities of the Filipino context, overall decision- making patterns are in line with those found in other lower-income countries and regions: approximately 34 percent of men and 41 percent of women report joint decision-making in any given domain in our sample, rates similar to those found in other contexts. 7 3.2 Quantitative Data and Analysis Our quantitative data were collected through a spousal survey carried out between February and May 2018 8 with 993 respondents, including 423 matched monogamous couples. Approximately 85 percent of the sample was in the island of Mindanao, while a small minority of households were in the Bicol Region in southeastern Luzon. 9 All households in our sample had received parcels of agricultural land through the Comprehensive Agrarian Reform Program (CARP) at least ten years prior to the baseline survey. As such, couples in our sample are almost universally engaged in agriculture, and skew older and poorer than the Philippine population as a whole. Descriptive statistics of our sample can be found in Appendix Table A1. 7 In an aggregated sample of 20 Sub-Saharan African Countries, Donald et al. (2017) find that 42 percent of men and 44 percent of women report joint decision-making over household purchases. 8 This spousal survey was collected during the baseline of an impact evaluation of a land tenure intervention carried out with the Philippines Department of Agrarian Reform. It was a follow-on to a longer baseline survey that had been carried out earlier with only the household member listed as an agrarian reform beneficiary. 9 Respondents in Mindanao were a mix of Cebuano, Ilonggo, Ilocano, Mandaya, and Manobo ethnicities, while respondents in Luzon were Bicolano. Data on specific ethnicities were not collected during the survey, and the dataset lacks the statistical power to compare ethnicities, due to the sample size and diversity of respondents. 10 The spousal survey was designed specifically to capture spousal participation in household decision-making, particularly related to agriculture. The questions asked to respondents are shown in Table 1 (respondents were only asked about activities in which they reported at least one person in the household participated) and are taken from the Women’s Empowerment in Agriculture Index (Alkire et al. 2013). We construct the following set of decision-making indicators as our explanatory variables: 1. An indicator variable taking the value of 1 if the respondent considers themselves a decision-maker, either solely or jointly (DMi in the equations below). 10 2. An indicator variable taking the value of 1 if both the respondent and their spouse agree that the respondent is a decision-maker. (AgreeDMi ) 3. An indicator variable taking the value of 1 if the respondent reports having input on most or all decisions (Inputi). 4. An indicator variable taking the value of 1 if the respondent reports the ability to make their own personal decisions if desired to a high extent (Abilityi). Table 1: Decision-making variables analyzed Question: Answer options: When decisions are made regarding Self; Spouse; Self and spouse; Other [activity], who is generally responsible for these household member; Other non-household decisions? member How much input would you say you have No input or input in a few decisions; in decisions regarding [activity]? Input on some decisions; Input on most or all decisions; No decision made To what extent could you make your own Not at all; To a small extent; To a large personal decisions about [activity] if you wanted extent; Don’t know/not sure to? Agricultural activities covered: Growing crops for consumption or selling on the market; Choosing seeds for food and cash crops; Buying agricultural inputs such as fertilizers; Hiring or paying laborers; Buying or renting farm equipment such as carabaos or hand tractors; Rearing livestock 11 As robustness checks, we verify whether our results are sensitive to indicator construction (defining decision-making as sole in 1 and 2, considering input on some decisions in 3, and the ability to make one’s own decisions to a medium extent in 4). As decision-making 10 These indicators are averaged across domains of agricultural decision-making. For instance, if a respondent reports being a decision-maker in 3 out of 6 domains, DMi would appear as 0.5 (50 percent). 11 The survey included a variety of non-agricultural activities, including non-farm self-employment, wage work, small household expenditures, and major and minor household expenditures. However, given that the RAI computed in the survey applies only to agricultural decisions, these variables were omitted from our analysis. 11 questions were asked about several agricultural domains, we use the percentage of domains in which the respondent reports decision-making power to only run one regression per equation, thereby limiting issues with multiple hypothesis testing. However, we also show disaggregated results by decision using p-value corrections in Appendix C. To facilitate interpretation, our final indicators are the z-scores of these percentage variables. Our main outcome variable is the RAI for three main types of agricultural decisions: 1) choosing seeds or crops; 2) choosing where to sell crops; and 3) decisions to sell or lease agricultural parcels. The RAI module was administered in vignette format, where respondents were presented vignettes of fictional people with motivations aligning to autonomous motivations, introjected motivations, and external motivations. 12 Answer options ranged from 1, “completely different” to 4, “completely the same”. 13 An example set of vignettes used in the RAI module can be found in Table 2, while the full set can be found in Appendix Table A2. Validation of the RAI in our sample finds that responses best align with a 2-factor scale, with introjected and autonomous motivations clustering together (see Appendix B for a detailed discussion). This suggests that for respondents in our sample, the difference between making decisions based on one’s own goals and motivations and based on internalized social norms may not be a relevant one. As such, we code the RAI variable in two subscales: a “Coerced” subscale, with a weight of -3, and an “Autonomous/Introjected” subscale, with a weight of 3, resulting in a range of -9 to 9 as with the original RAI. This approach is analogous to that adopted in Vaz, Pratley, and Alkire (2016), which adopts a 2-factor RAI scale that best represents their data. 14 12 In accordance with the 2016 WEAI questionnaire, a fourth category, where decisions were made because there was no other option, was included. However, following other analyses of the RAI scale, only the three categories above are included when computing the scale. 13 Respondents were first asked whether they were the same or different, then asked if they were “mostly” or “completely” the same or different in a follow-up question. 14 Using data from Chad, Vaz, Pratley, and Alkire (2016) do not find a meaningful distinction between external and introjected motivations and compute an alternative RAI based on two broad types of motivation, controlled and autonomous. 12 Table 2: Example vignettes from Relative Autonomy Index (RAI) module Vignette: Answer options: Fe (Juan) plants the crops on her parcel that she does because she 1. Completely thinks those are the best crops to be planting, and she thinks that's what's best different for her parcel. If she changed her mind on what to plant, she would plant 2. Somewhat other crops. Are you like this person? (Autonomous motivation) different Maria (Pedro) plants the crops on her parcel that she does because 3. Somewhat most other people in her community plant them and she wants to be seen as the same making good decisions with her parcel. Are you like this person? (Introjected 4. Completely motivation) the same Analyn (Oking) plants the crops on her parcel that she does because her spouse or someone else tells her that she must plant those. She does what they tell her to do. Are you like this person? (Coerced motivation) Note: Names in vignettes were changed depending on the gender of the respondent. We test which decision-making variables are predictors of agency using the following regressions: = + 1 + ′ + + (1) = + 2 + ′ + + (2) = + 3 + ′ + + (3) = + 4 + ′ + + (4) where , , , and are aggregate z-score indices of our decision-making variables (status as decision-makers, agreement on status as decision-maker, input into decisions, and the ability to make decisions if desired, respectively). represents respondent i’s Relative Autonomy Index, X is a vector of LASSO-selected baseline covariates (Belloni, Chernozhukov and Hansen 2014), and represents province (p) fixed effects. Results are reported with robust standard errors, both with controls and province fixed effects (adjusted) and without controls (unadjusted). Although we aim to evaluate how each indicator individually relates to agency, we also have an interest in comparing the four types of indicators against each other to see which has the strongest relationship. As such, we run two additional OLS regression taking the form of: 13 = + 1 + + + ′ + + (5) = + 2 + + + ′ + + (6) where the indicators are as specified for Equations (1) – (4). A significant coefficient on a given measure of decision-making is interpreted to mean that that indicator is significantly associated with agency when controlling for the other two measures. As relationships between decision-making and agency may differ by gender, we run regressions separately for men and for women. 3.3 Qualitative Data and Analysis Following the initial analysis of the quantitative data, we carried out a round of qualitative interviews focused on intra-household decision-making. Notably, the interviews explored respondents’ preferences about decision-making, perceived advantages and disadvantages of different types of intra-household decision-making, and how different ways of making decisions makes them feel. 15 Qualitative interviews were conducted with a separate but similar sample to the quantitative survey. 16 A total of 40 individual interviews and 20 joint spousal interviews were conducted in 40 households across three provinces from July-August 2019. The individual interviews allowed the respondents to explain the reasons behind their answers without having to consider how their spouse might react to their answers. The joint spousal interviews, on the other hand, shed light on the dynamics between husband and wife and their perceptions of what constitutes a joint decision and the processes that lead to this kind of decision-making. The questionnaire was composed of 10 thematic sections for the individual interviews, and 8 thematic sections for the joint spousal interviews. The transcripts of the qualitative interviews were coded based on a coding scheme developed from the research questions and main hypotheses of the study with the help of Atlas.ti, a qualitative text analysis tool. 15 The interviews also explored how men and women think about decision-making and what constitutes making a decision and the role that land titles play in decision-making on agricultural land. The full interview guide is available in Arugay and Valenciano (2022). 16 Quantitative data was drawn from the baseline survey of an impact evaluation, which meant that several rounds of data were already conducted or planned with the original quantitative sample. Using the original sample raised concerns of interview fatigue. 14 4. Results 4.1 Agency and decision-making Descriptive statistics suggest that our decision-making variables capture different phenomena (Table 3). As is common in the rural Philippines and as reviewed in Section 3.1, men and women in our sample tend to exercise decision-making power in separate spheres. Men are more likely to be decision-makers in all agriculture-related areas, while women are more likely to be decision-makers regarding small expenditures and non-farm employment. We also see that not all who report that they are decision-makers say that they have input into decisions or are able to make their own personal decisions, suggesting the indicators are capturing different aspects of decision-making. Fewer men and women say they have input on most or all decisions than consider themselves a decision-maker (pooling sole and joint decision- making). More women report the ability to make personal decisions than being sole decision- makers, whereas fewer men report the ability to make personal decisions than being sole decision-makers. 17 This may suggest that at least some women in our sample could make their own personal decisions but choose not to decide on their own, and at least some men in our sample may be constrained in the choices they can make even if they dominate a certain domain within their households. 17 The population averages are shown in Table 3. When looking at an individual level, we also find similar patterns; for example, women reporting they are decision makers without having input and men reporting they are sole decision makers who cannot make personal decisions. For example, only 56 percent of women who describe themselves as decision-makers over growing crops for sale and consumption state that they have input over “most or all decisions”, while 43 percent state they could make their own personal decisions to a high degree if they wanted to. Full results are available upon request. 15 Table 3: Descriptive statistics of decision-making by gender Spouses agree respondent Reports ability to make Considers self decision- Considers self sole is sole or joint decision- Reports input on most or all own decisions to a high maker (sole or joint) decision-maker maker decisions extent Domain Women Men Women Men Women Men Women Men Women Men Growing crops 56.8% 93.7% 21.1% 65.6% 21.9% 76.7% 46.1% 69.8% 34.4% 51.3% Buying/selling farm equipment 31.9% 94.3% 6.6% 65.4% 16.7% 84.4% 31.9% 71.8% 20.9% 51.9% Choosing where to sell crops 58.7% 89.1% 17.4% 53.2% 31.4% 78.7% 43.7% 65.9% 35.6% 51.9% Rearing livestock 65.4% 89.8% 18.1% 47.3% 41.1% 75.1% 51.6% 69.8% 36.7% 55.5% Choosing seeds 41.5% 91.3% 11.7% 68.2% 20.1% 80.6% 36.6% 74.7% 30.3% 65.1% Buying agricultural inputs 41.5% 92.0% 8.0% 66.7% 22.2% 84.0% 35.8% 77.3% 28.1% 62.1% Hiring and paying laborers 49.4% 93.8% 11.4% 63.9% 21.8% 85.4% 43.1% 76.0% 34.9% 61.4% Paid employment 55.9% 80.3% 30.6% 65.0% 16.9% 55.9% 51.5% 71.4% 41.0% 51.7% Non-farm self- employment 84.1% 51.9% 57.9% 20.3% 80.2% 26.0% 71.3% 42.9% 57.9% 41.4% Major expenditures 61.5% 82.4% 10.3% 32.0% 47.3% 72.7% 55.1% 71.2% 41.7% 54.2% Small expenditures 86.5% 71.3% 46.9% 30.6% 65.0% 40.3% 71.7% 53.4% 57.8% 40.1% Note: N = 993 respondents and 423 matched couples. Not all respondents participated in all activities, and thus sample size varies for each individual domain. 16 4.2 Regression analysis Our regression results, following equations (1) to (4), are shown in Table 4. Panel A presents results for women, while Panel B presents results for men. We find that the most commonly-used indicator of decision-making—being a decision-maker (solely or jointly)—is only weakly associated with motivational autonomy (Columns 1 and 2). The relationship is only statistically significant for women, and only at the 10 percent level when introducing controls. The relationship between being a decision-maker and agency is weak regardless of whether the variable is coded to include both sole and joint decisions or sole decisions only, and the latter coding is not significant for either gender (see Appendix C2). Table 4a: Relationship between decision-making measures and RAI (women) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Decision-maker 0.100 0.079 (DMi) (0.044)** (0.045)* Agree woman is 0.124 0.082 decision-maker (0.049)** (0.048)* (AgreeDMi) High decision 0.161 0.143 input (Inputi) (0.045)*** (0.045)*** High decision 0.223 0.185 autonomy (0.045)*** (0.046)*** (Abilityi) N 496 496 412 412 496 496 496 496 R2 0.010 0.078 0.017 0.099 0.026 0.091 0.050 0.103 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 The decision-maker index represents the percentage of agricultural domains where the respondent reports being a decision-maker (either solely or jointly). High input represents the percentage of agricultural domains where the respondent reports having input on most or all decisions. High decision ability represents the percentage of agricultural domains where the respondent reports being able to make their own decisions to a high extent. All indices are transformed into z-scores. N corresponds to the total number of observations for which the outcome is not missing. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 17 Table 4b: Relationship between decision-making measures and RAI (men) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Decision- 0.015 0.012 maker (DMi) (0.044) (0.044) Agree man is 0.114** 0.089 decision- (0.052) (0.055) maker (AgreeDMi) High decision 0.113 0.070 input (Inputi) (0.043)*** (0.042)* High decision 0.184 0.127 autonomy (0.046)*** (0.047)*** (Abilityi) N 462 462 410 410 462 462 462 462 R2 0.000 0.064 0.010 0.063 0.013 0.069 0.050 0.103 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 The decision-maker index represents the percentage of agricultural domains where the respondent reports being a decision-maker (either solely or jointly). High input represents the percentage of agricultural domains where the respondent reports having input on most or all decisions. High decision ability represents the percentage of agricultural domains where the respondent reports being able to make their own decisions to a high extent. All indices are transformed into z-scores. N corresponds to the total number of observations for which the outcome is not missing. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. For women, we find a slightly stronger relationship between RAI and decision-making when both spouses agree that the respondent is a decision-maker (Table 4, Columns 3 and 4). For men, the association between being a decision-maker and agency grows in size and significance; however, it disappears again once introducing controls. This indicates that measurement error or noise in our decision-making variable is not what is driving our observed pattern. In contrast, we see a clear relationship between agency and having input in decisions (Table 4, Columns 5 and 6). The relationship is significant for both men and women and robust to the inclusion of controls. It is also the case for both possible constructions of the “input” variable, although strongest when restricted to input in “most or all decisions” (see Appendix C2). This finding is aligned with the descriptive evidence presented in Section 4.1. A woman may not frequently make substantive inputs into the decision-making process even when she considers herself a decision-maker, while a man who considers himself a decision-maker in agriculture may not necessarily have input in all choices, particularly if his options are practically constrained. 18 We find an even stronger relationship between the ability to make one’s own personal decisions if desired and the RAI (Table 4, Columns 7 and 8). The coefficients are significant at the 1 percent level for both men and women, regardless of controls. Additionally, the magnitude of the coefficients is more than 29% higher for women and more than 81% higher for men than coefficients for input into decisions. A one standard deviation increase in the percentage of domains wherein a respondent can make their own personal decisions is associated with a 0.13 to 0.22 standard deviation increase in RAI. As with input into decisions, descriptive data suggests that this variable captures a different phenomenon than simple decision-making, with substantially fewer respondents reporting being able to make their own personal decisions in a given domain than report being decision-makers (Table 3). Comparing all four aspects of decision-making, the ability to make one’s own personal decisions if desired has the strongest relationship with agency. We confirm this by showing the results of equation (5) in Table 5, and the results of equation (6) in Table 6. In both multivariate regressions including three dependent variables, only the ability to make one’s own personal decisions maintains statistical significance when controlling for the other decision-making variables. While having a high degree of input into decisions shows a strong relationship with RAI when it is considered on its own (Table 4), this relationship is completely crowded out once the ability to make one’s own decision, the best-performing measure, is included. Overall, our findings suggest that asking respondents to what extent they can make their own, personal decisions in a certain domain if they want to is the best way to capture agency through decision- making questions. 19 Table 5: Relative association of decision-making measures with RAI Women Men Decision-maker 0.003 -0.020 (0.048) (0.048) High decision input 0.045 -0.006 (0.060) (0.055) High decision ability 0.154 0.136 (0.059)*** (0.57)** N 496 462 R2 0.104 0.082 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 The decision-maker index represents the percentage of agricultural domains where the respondent reports being a decision-maker (solely or jointly). High input represents the percentage of agricultural domains where the respondent reports having input on most or all decisions. High decision ability represents the percentage of agricultural domains where the respondent reports being able to make their own decisions to a high extent. All indices are transformed into z-scores. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. Table 6: Relative association of decision-making measures with RAI Women Men Agree on decision-maker 0.036 0.060 (0.050) (0.057) High decision input 0.054 -0.033 (0.065) (0.056) High decision ability 0.148 0.170 (0.062)** (0.061)*** N 412 410 R2 0.127 0.082 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Agreement on decision-maker represents the percentage of domains where both spouses consider the respondent a decision-maker (sole or joint). High input represents the percentage of agricultural domains where the respondent reports having input on most or all decisions. High decision ability represents the percentage of agricultural domains where the respondent reports being able to make their own decisions to a high extent. All indices are transformed into z-scores. LASSO-selected controls are used in this model and include ARB status, age difference between spouses, and province fixed effects. 4.3 Qualitative interview results Our qualitative work reveals three potential mechanisms for why self-reported status as a decision-maker may not capture agency, while having input and (especially) the ability to make personal decisions if desired can. First, decisions for respondents in our sample were often constrained by outside factors that made it difficult for households to align their choices with their preferences. In one case, a 20 couple described the decision over whether or not to lease their land; their choice was either to allow another farmer to till it and risk losing access to their own land, or keeping it in their possession even if they lacked inputs to till it. Several other households cited the decisions of whether to plant additional crops or sell their possessions due to the loss of their main crops from a typhoon, or how to get loans to pay for agricultural inputs or daily living. Other respondents struggled to think of major decisions as they saw their households as having few options. A female respondent stated that their agricultural decisions were “to plant whatever we have, as long as we have something to plant… There’s nothing else because that’s all we can do here.” Another male respondent cited the household’s major agricultural decisions as being “about the meager food, our difficulty in farming… Where to find resources because upland farming is once a year. We do corn farming but enough for consumption, not enough for selling.” Such decisions were described as difficult and stressful due to desperate circumstances. When describing the decision to take out a loan, a male respondent stated “Even if we have already agreed on what to do we will be worried. We must think of ways to pay back the debt, especially if we get sick.” A female respondent stated “Our problems are financial. I don’t think we will be able to eat. That is my life now. I cannot dream of anything anymore.” Being a decision-maker matters little for agency if respondents’ agency is constrained by external factors such as extreme poverty or natural disasters. Second, beyond external constraints, constraints internal to the household also mattered. Respondents stated that virtually all major or consequential decisions involved some level of consultation between spouses, even “sole” decisions. Some respondents stated that the purpose was to obtain the input of the other spouse and improve the outcome of the decision. Others reported that making decisions without consultation was undesirable because it could lead to misunderstandings or discord within the household and would result in the decision-making spouse taking sole blame for any negative consequences. As one male interviewee stated, “The ideal decision-making is done by both of us. This is to avoid trouble. When the decision is final no one is to be blamed. But when the decision is made alone, everyone will be pointing fingers when things get bumpy.” The need to involve spouses to avoid blame was frequently mentioned by both men and women when interviewed, and similar patterns emerged when respondents were interviewed individually and jointly with their spouses. 21 Although most couples preferred consultative decision-making, the amount of consultation that occurred during the decision-making process varied a great deal between couples. In many households, the process was dominated by the husband, with the wife only perfunctorily consulted or giving “rubber stamp” approval to her husband’s decisions. One female respondent, though insisting that she and her husband made all decisions together, stated that “he is the pillar of the household, if you are the wife it is enough that you concur.” In households where there were more lengthy conversations, women most often deferred to their husbands in matters of farming, either because they saw him as the head of the household or because they viewed him as having greater knowledge of agriculture. Even if conversations were brief or perfunctory, the wife’s approval was often sought so that both spouses would share responsibility for the decision. Lastly, when asked about decision-making, respondents often gave examples of relatively inconsequential day-to-day decisions. Examples included paying utility bills, raising livestock, daily budgeting, or planting vegetables on the farm for household consumption. This was particularly the case when respondents were asked to name decisions they made without their spouse’s involvement, as it was insisted that all major decisions had to be made with both husband and wife present. Together with respondents’ reports above that agricultural decisions were “to plant whatever we have…because that’s all we can do here”, as well as a core feature of decision-making being taking responsibility for the outcome of the decision, a picture emerges of decision-making as more akin to task management for many households, not as an expression of bargaining power or the ability to set goals and act on them. 5. Conclusion and Discussion Decision-making is often used as a proxy for agency, in particular to understand women’s empowerment. Nevertheless, the literature highlights several theoretical and methodological reasons why decision-making may not accurately capture agency. This paper provides some of the first empirical evidence testing these concerns. Using a spousal survey of farming households in the rural Philippines, we examine the extent to which measures that capture different aspects of decision-making correlate with agency for both men and women, as measured by the Relative 22 Autonomy Index. To better unpack our quantitative findings, we conduct qualitative work with a similar sample. We find that self-reported status as decision-maker has no significant association with agency for men and only a weak one for women in our sample, regardless of whether joint or sole decision-making is considered or the domain of agricultural decision-making. Our qualitative work suggests several reasons for this. Respondents regard themselves as a decision- maker due to a social norm of consultative decision-making, even when the level of consultation is perfunctory or they cannot influence the outcome of the decision. Moreover, respondents report that they have limited input in their decisions due to poverty or scarcity in their environment, and many of the decisions they report being involved in are often relatively inconsequential, such as paying workers or tending to livestock. Put together, these results suggest that “decision-making” is an overly broad category that includes cases where respondents are not able to meaningfully influence the decision-making process or make the decisions they might desire to. This echoes findings from other studies that find that restricting decision-making to “sole” or “joint” indicators masks a great deal of variability, and that joint decision-making is not necessarily egalitarian (Acosta et al. 2020). It also echoes theoretical work suggesting that decision-making is not necessarily empowering when it is constrained by gender norms or economic circumstances (Heckert and Fabic 2013). We test these qualitative patterns in our quantitative data. We find that having a high amount of input in decision-making is significantly associated with higher agency for both men and women. Measuring the amount of input respondents have in decisions may narrow the broad concept of decision-making to those cases where respondents are able to substantively influence the decision-making process and make their voice heard, making it a more effective proxy for agency. However, the strongest predictor of agency in our sample is the extent to which respondents can make their own personal decisions if they want to. This variable is very strongly associated with the RAI for both men and women, and for every domain of agricultural decision- making. Indeed, when all three variables tested are compared in a single regression, only this variable retains a significant relationship with the RAI. We provide empirical support for the conceptual criticisms of using decision-making as a direct proxy for agency in intra-household bargaining. In our sample, it would not be accurate to conclude that respondents who report being decision-makers in agricultural activities have 23 more agency over agricultural decision-making than those that do not. Rather, it is critical to consider the extent to which individuals can make decisions in line with their own preferences. Given the costs of decision-making, such as high cognitive load or the fear of blame, promoting women’s decision-making may not bolster their agency. Moreover, while work carried out in Bangladesh and Ghana (Seymour and Peterman 2018) suggests that the relationship between decision-maker status and autonomy is highly context-specific, our quantitative and qualitative results map closely to theoretical understandings of agency across disciplines. While recognizing that decision-making is associated with important development outcomes, including reproductive health (Bankole and Singh 1999; Ghuman, Lee and Smith 2006), nutrition (Amugsi et al. 2015; Bussolo et al. 2021; Malapit and Quisumbing 2015) and other women’s empowerment outcomes such as earnings and land ownership (Peterman et al. 2015; Annan et al. 2019; Donald et al. 2020), our findings suggest that relatively low-cost additions to household surveys can lead to large improvements in the measurement of agency. Asking respondents about the amount of input they have in decisions and the extent to which they can make their own personal decisions if desired in addition to questions about who makes decisions can allow for better measurement of agency. These two questions would have minimal costs to interviewers or respondents and are already used in surveys such as the Women’s Empowerment in Agriculture Index (WEAI), but are less commonly adopted compared to questions on “who makes decisions”. Our findings point to several areas for future research. First, replicating these findings in other contexts is needed to assess the extent to which they can be generalized. Second, further work to refine the RAI or develop alternative psychometric tools for measuring agency directly could improve the measurement of this important concept. Lastly, future research should also try to better understand why decision-making is associated with positive welfare outcomes in some contexts, and whether these alternative measures provide additional precision. 24 6. References Abrigo, Michael R.M. and Kris Francisco-Abrigo. 2019. “Counting women’s work in the Philippines.” PIDS Discussion Paper Series No. 2019-02. Acosta, Mariola, Margit van Wessel, Severine Van Bommel, Edidah L. Ampaire, Jennifer Twyman, Laurence Jassogne, and Peter H. Feindt. 2019. “What does it Mean to Make a 'Joint' Decision? Unpacking Intra-household Decision Making in Agriculture: Implications for Policy and Practice.” The Journal of Development Studies 56(0): 1210-1229. Adamkovič, Matúš and Marcel Martončik. 2017. “A Review of Consequences of Poverty on Economic Decision-Making: A Hypothesized Model of a Cognitive Mechanism.” Frontiers in Psychology 8: 1-13. Agarwal, Bina. 1994. “Gender and command over property: A critical gap in economic analysis and policy in South Asia.” World Development 22(10): 1455-1478. Agarwal, Bina. 1997. “Bargaining and gender relations: Within and beyond the household.” Feminist Economics, 3(1): 1-51. Akter, Sonia, Pieter Rutsaert, Joyce Luis, Nyo Me Htwe, Su Su San, Budi Raharjo, and Arlyna Pustika. 2017. "Women's empowerment and gender equity in agriculture: A different perspective from Southeast Asia." Food Policy 69: 270-279. Alkire, Sabrina, Ruth Meinzen-Dick, Amber Peterman, Agnes Quisumbing, Greg Seymour, and Ana Vaz. 2013. “The women’s empowerment in agriculture index.” World Development, 52: 71-91. Ambler, Kate, Cheryl Doss, Caitlin Kieran, and Simone Passarelli. 2021. “He Says, She Says: Spousal Disagreement in Survey Measures of Bargaining Power.” Economic Development and Cultural Change 69(2): 1-24. Amugsi, Dickson A., Anna Lartey, Elizabeth Kimani-Murage, and Blessing U. Mberu. 2016. “Women's participation in household decision-making and higher dietary diversity: Findings from nationally representative data from Ghana.” Journal of Health, Population and Nutrition, 35(1): 1-18. Annan, Jeannie, Aletheia Donald, Markus Goldstein, Paula Gonzalez Martinez, and Gayatri Koolwal. 2019. "Taking Power: Women's Empowerment and Household Well-being in Sub- Saharan Africa." Policy Research Working Paper, World Bank. Arugay, Aries and Aletheia K. Valenciano. 2022. “Understanding the Gender Dynamics of Decision-making in Agrarian Households of the Rural Philippines.” Washington, D.C.: World Bank 25 Group. http://documents.worldbank.org/curated/en/099524411212211006/IDU082a5feb505d17 04e7909d8a0bc5fdfb6accc. Bankole, Akinrinola, and Susheela Singh. 1999. "Couples' Fertility and Contraceptive Decision-Making in Developing Countries: Hearing the Man's Voice." International Perspectives on Sexual and Reproductive Health 24(1): 15-24. Battman, J. R., Mary F. Luce, and J. W. Payne. 1998. “Constructive consumer choice process.” Journal of Consumer Research 25(4): 187-217. Bautista, Germelino. 1977. “Socio-economic conditions of the landless rice workers in the Philippines: The landless of Barrio Sta. Lucia as a case in point.” Hired Labor in Rural Asia. Tokyo, Institute of Developing Economies: 106-126. Belloni, Alexandre, Victor Chernozhukov, and Christian Hansen. 2014. "Inference on treatment effects after selection among high-dimensional controls." Review of Economic Studies 81(2): 608-650. Bussolo, Maurizio, Nayantara Sarma, and Anaise Williams. 2021. “It Takes Two (To Make Things Right): Women’s Empowerment and Couple Concordance in South Asia.” Policy Research Working Paper No. 9545. World Bank, Washington D.C. Cavazzoni, Federica, Alec Fiorini and Guido Veronese. 2022. “How Do We Assess How Agentic We Are? A Literature Review of Existing Instruments to Measure Individuals’ Agency.” Social Indicators Research: An International and Interdisciplinary Journal for Quality-of-life Measurement. 159(3): 1125-1153. Christie, Maria Elisa, Mary Parks, and Michael Mulvaney. 2016. “Gender and local soil knowledge: Linking farmers’ perceptions with soil fertility in two villages in the Philippines.” Singapore Journal of Tropical Geography 37(1): 6-24. Contado, Mina. 1981. “Power Dynamics of Rural Families: The Case of a Samar Barrio.” Philippine Sociological Review 29(1): 73-85. Daminger, Allison. 2019. “The cognitive dimension of household labor.” American Sociological Review, 84(4), 609-633. David, Fely P. 1994. “The Roles of Husbands and Wives in Household Decision-Making.” Philippine Sociological Review 42(1): 78-93. Donald, Aletheia, Gayatri Koolwal, Jeannie Annan, Kathryn Falb, and Markus Goldstein. 2020. “Measuring women’s agency.” Feminist Economics, 26(3): 200-226. Donald, Aletheia, Cheryl Doss, Markus Goldstein, and Sakshi Gupta. 2023. “Sharing responsibility through joint decision-making and implications for intimate-partner violence: 26 evidence from 12 Sub-Saharan African Countries.” Review of Economics of the Household, 22(1): 1-32. Ebrahim, Nasser B. and Madhu S. Atteraya. 2019. “Women’s Household Decision-Making and Intimate Partner Violence in Ethiopia.” Academic Journal of Interdisciplinary Studies 8(2): 285-292. Eissler, Sarah, Jessica Heckert, Emily Myers, Gregory Seymour, Sheela Sinharoy, and Kathryn M. Yount. 2021. Exploring gendered experiences of time-use agency in Benin, Malawi, and Nigeria as a new concept to measure women’s empowerment. Vol. 2003. Washington, D.C.: International Food Policy Resource Institute. Estudillo, Joanna P., Agnes Quismbing, and Keijiro Otsuka. 2001. “Gender differences in land inheritance and schooling investment in the rural Philippines.” Land Economics 77 (1): 130- 143. Gammage, Sarah, Naila Kabeer, and Yana van der Meulen Rodgers. 2016. "Voice and Agency: Where Are We Now?" Feminist Economics 22(1): 1-29. Ghuman, Sharon J., Helen J. Lee, and Herbert L. Smith. 2006. “Measurement of women’s autonomy according to women and their husbands: Results from five Asian countries.” Social Science Research 35(1): 1-28. Hindin, Michelle and Linda S. Adair. 2002. “Who’s at risk? Factors associated with intimate partner violence in the Philippines.” Social Science and Medicine, 55(8): 1385-1399. Heckert, Jessica and Madeleine Short Fabic. 2013. "Improving Data Concerning Women's Empowerment in Sub-Saharan Africa". Studies in Family Planning 44(3): 319-344. Horne, Christine, F. Nii-Amoo Dodoo, and Naa Dodua Dodoo. 2013. “The shadow of indebtedness: Bridewealth and norms constraining female reproductive autonomy.” American Sociological Review 78(3): 503-520. Illo, Jeanne Frances. 1989. “Who heads the household? Women in households in the Philippines.” The Filipino woman in focus: a book of readings. Bangkok, UNESCO: 244-266. Kabeer, Naila. 1999. "Resources, agency, achievements: Reflections on the measurement of women's empowerment." Development and Change 30: 435-464. Liaqat, Sundas, Aletheia Donald, Forest Jarvis, Elizaveta Perova, and Hillary C. Johnson. 2021. “Lost in Interpretation: Why Spouses Disagree on Who Makes Decisions.” World Bank Policy Research Working Paper WPS9883. Luce, Mary Frances. 1998. “Choosing to Avoid: Coping with Negatively Emotion-Laden Consumer Decisions.” Journal of Consumer Research 24(4): 409-433. 27 Malapit, Hazel Jean L., Agnes R. Quisumbing. 2015. "What dimensions of women's empowerment in agriculture matter for nutrition in Ghana?" Food Policy 52(4): 54-63. Malapit, Hazel, Catherine Ragasa, Elena M. Martinez, Deborah Rubin, Greg Seymour, and Agnes Quisumbing. 2020. "Empowerment in agricultural value chains: Mixed methods evidence from the Philippines." Journal of Rural Studies 76: 240-253 Malapit, Hazel, Agnes Quisumbing, Ruth Meinzen-Dick, Greg Seymour, Elena M. Martinez, Jessica Heckert, and Deborah Rubin. 2019. "Development of the project-level Women’s Empowerment in Agriculture Index (pro-WEAI)." World Development 122: 675-692. Medina, Belen T.G. 2001. The Filipino Family. Diliman, Quezon City: University of the Philippines Press. Okabe, Masayoshi. 2016. "Gender-Preferential Intergenerational Patterns in Primary Educational Attainment: An Econometric Approach to a Case in Rural Mindanao, the Philippines." International Journal of Educational Development 46: 125-142. Parks, Mary Harman, Maria Elisa Christie, and Isidra Bagares. 2015. "Gender and conservation agriculture: constraints and opportunities in the Philippines." GeoJournal 80: 61- 77. Quisumbing, Agnes R., Ruth Suseela Meinzen-Dick, Hazel J. Malapit, Greg Seymour, Jessica Heckert, Cheryl Doss, and Nancy Johnson. 2022. “Can Agricultural Development Projects Empower Women? A Synthesis of Mixed Methods Evaluations Using pro-WEAI in the Gender, Agriculture, and Assets Project (Phase 2) Portfolio.” IFPRI Discussion Paper Series 2137. International Food Policy Research Institute (IFPRI). Richards, Esther, Saslly Theobald, Asha George, Julia C. Kim, Christiane Rudert, Kate Jehan, and Rachel Tolhurst. 2012. "Going beyond the surface: Gendered intra-household bargaining as a social determinant of child health and nutrition in low and middle income countries." Social Science and Medicine 95: 24-33. Ryan, Richard M. and James P. Connell. 1989. “Perceived Locus of Causality and Internalization: Examining Reasons for Acting in Two Domains.” Journal of Personality and Social Psychology 57: 749–61. Ryan, Richard M., and Edward L. Deci. 2000. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist 55(1): 68- 78. Sen, Amartya. 1999. Development as Freedom. New Yor: Alfred Knopf Seymour, Greg and Amber Peterman. 2018. "Context and measurement: An analysis of the relationship between intrahousehold decision making and autonomy". World Development 111: 97-112. 28 Sproule, Katie and Kovarik, Chiara. 2014. “Cognitive Testing and vignettes.” Gender Methods Seminar, IFPRI. December 12, 2014. Tapia, Maricel, Juan Pulhin, Gloria Luz Nelson, Canesio Predo, Aileen Peria, Rose Jane Peras, Rodel Lasco, Florencia Pulhin, and Regine Joy Evangelista. 2018. “Involvement of Women in Farm Decision-making and Adaptive Capacity to Extreme Events of Farming Households in Ligao City, Albay, Philippines.” Journal of Environmental Science and Management 21(2): 70-81. Twyman, Jennifer, Pilar Useche, and Carmen Diana Deere. 2015. "Gendered Perceptions of Land Ownership and Agricultural Decision-making in Ecuador: Who Are the Farm Managers?" Land Economics 91(3): 479-500. Vaz, Ana, Sabina Alkire, Agnes Quisumbing, and Esha Sraboni. 2013. “Measuring Autonomy: Evidence from Bangladesh.” Research in Progress Report 38a, Oxford Poverty and Human Development Initiative (OPHI). Vaz, Ana, Pierre Pratley, and Sabina Alkire. 2016. “Measuring Women’s Autonomy in Chad Using the Women’s Empowerment in Agriculture Index.” Feminist Economics 22(1): 264-294. World Bank. 2012. Republic of the Philippines: Gender and Development Mainstreaming. Country Gender Assessment. 29 7. Appendix Appendix A: Tables and Figures Appendix Figure A1: Situating decision-making and autonomy within agency and empowerment (Donald et al. 2020) EMPOWERMENT AGENCY RESOURCES OUTCOMES Ability to define Sense of control goals in line with Ability to make over one’s life and one’s values strategic life choices life choices (autonomy) Measured with… Measured with… Measured with… Relative Autonomy Decision-making Locus of Control (LOC) Index (RAI) scale, Self-efficacy Scales… 30 Appendix Table A1: Basic Demographic Characteristics of Sample Variable Obs % Mean Std. Dev. Women 514 51.8% Men 479 48.2% Households with matched monogamous couples 423 73.7% Number of households by province 18: Davao Oriental 119 28.1% Davao del Sur 92 21.5% Sarangani 26 6.2% North Cotabato 15 3.6% Sultan Kudarat 44 10.4% Misamis Oriental 41 9.7% Bukidnon 28 6.6% Surigao del Sur 2 0.5% Albay 19 28 6.6% Camarines Sur19 28 6.6% Husband age 20 54.14 12.05 Wife age 49.86 12.15 Age difference (husband’s age – wife’s age) 4.15 6.01 Couples in formal marriages 91.0% Couples in common-law marriages 8.8% Agrarian reform beneficiary’s 21 education level (years) 6.98 3.33 Per capita monthly household income (PHP) 5706.61 14272.73 Household below national food poverty income 194 47.8% threshold Women with non-farm income 223 52.7% Men with non-farm income 244 57.7% Female agrarian reform beneficiary 107 25.30% Parcels tilled by household 1.37 0.81 Parcels formally owned by household 22 1.16 0.44 Household size 4.84 2.15 18 All statistics are limited to matched couples in our data set. 19 Province is in the island of Luzon. 20 Data on age and income were gathered in the initial baseline survey carried out in 2016 and 2017. 21 Agrarian reform beneficiaries are the spouse who was the primary recipient of the agricultural parcel awarded through CARP. 22 We consider parcels “formally owned” when the household possesses an official title document for the land. This includes all parcels received through CARP. 31 Appendix Table A2: Full RAI module of spousal survey Now I'm going to read you some descriptions of various people. Can you tell me if you're like this 0. Completely person or not like this person? the same 1. Somewhat the same 2. Completely different 3. Somewhat different 1 Pedro (Maria) plants the crops on his parcel that he does because most other people in his community plant them and he wants to be seen as making good decisions with his parcel. Are you like this person? └─┴─┘ 2 Marcelino (Niña) decides whether or not to lease out the land that he owns because that is what he thinks is the best decision for his family and for their money. If he decided to change his mind about what to do with his parcel, he would. Are you like this person? └─┴─┘ 3 Pedro (Carmen) decided whether or not to sell or lease his parcel by looking at what other people in the community were doing with their own parcels. He doesn't want them to see him as being irresponsible with his land. Are you like this person? └─┴─┘ 4 Juan (Fe) plants the crops on his parcel that he does because he thinks those are the best crops to be planting, and he thinks that's what's best for his parcel. If he changed his mind on what to plant, he would plant other crops. Are you like this person? └─┴─┘ 5 Boy (Marivic) takes the crops to market because his spouse or someone else in the community tells him that he should. He does what they tell him to do. Are you like this person? └─┴─┘ 6 Martin (Fely) sells his crops at the market because that's what's expected in his community. He wants people to see his as a good businessman. Are you like this person? └─┴─┘ 7 Juan (Judith) chooses what produce and how much to take to the market because he thinks that's what's best for his family and his business. He could change how much he brings if he wanted to. Are you like this person? └─┴─┘ 8 If Ding (Jenilyn) decides to sell or lease his parcel, it will be because his spouse or someone in the community told him that she should. He does what they tell him to do. Are you like this person? └─┴─┘ 9 Oking (Analyn) plants the crops on his parcel that he does because his spouse or someone else tells him that he must plant those. He does what they tell him to do. Are you like this person? └─┴─┘ 32 Appendix B: Validating the RAI in our survey The RAI has been administered in a variety of contexts, including in the developing world (e.g., Vaz et al 2013; Vaz, Pratley and Alkire 2016; Malapit, Sproule, and Kovarik 2016; Seyour and Peterman 2018), as well as in the Philippines (Malapit et al 2020). Since we use the scale as a benchmark for validating other variables in our sample, here we conduct a psychometric analysis of the validity of the RAI in our survey. We first test the scale using factor analysis to examine whether the expected three factors of autonomous, introjected, and coerced motivations appear. An exploratory factor analysis (EFA) of our data finds that responses do, in fact, cluster into three distinct factors (Table B1). However, the factors do not all match the constructs that our RAI scale aims to capture: while coerced motivations cluster neatly into a single factor (Factor 3 for women and Factor 1 for men), autonomous and introjected motivations cluster together into the same factor (Factor 2). This suggests that for respondents in our sample, the difference between making decisions based on one’s own goals and motivations and based on internalized social norms may not be a relevant one. Moreover, a third, extraneous factor appears in our exploratory factor analysis that correlates with all questions related to selling or leasing parcels (Factor 1 for women and Factor 3 for men). For our respondents, decisions to sell or lease their parcels are more associated with each other than to any of the categories of motivation. The context of the study suggests a potential explanation: all households in our sample were the owners of at least one collectively- owned parcel distributed under the Comprehensive Agrarian Reform Program. Agrarian Reform Beneficiaries are not legally able to sell or lease their parcels, although this occurs very frequently on an informal level. 23 Thus, the sensitivities associated with this topic may have introduced a separate factor. 23 Of the agrarian reform beneficiaries originally contacted as potentially qualified for the impact evaluation, over 200 were disqualified for having informally sold their parcels, while approximately 20 percent of our sample leased out their parcels to other tillers. 33 Table B1: Results of an Exploratory Factor Analysis of the RAI Women Men Domain Subscale Factor 1 Factor 2 Factor 3 Factor 1 Factor 2 Factor 3 Choosing crops External 0.5525 0.6176 Choosing where to sell crops External 0.3978 0.3966 0.5086 Choosing to sell or lease parcel External 0.3842 0.4631 0.5802 0.3068 Choosing crops Introjected 0.3639 0.3489 Choosing where to sell crops Introjected 0.5198 0.5078 Choosing to sell or lease parcel Introjected 0.6049 0.4918 Choosing crops Autonomous 0.4260 0.3436 Choosing where to sell crops Autonomous 0.4688 0.5487 Choosing to sell or lease parcel Autonomous 0.5814 0.4976 Notes: Sample includes 510 women and 472 men. Table displays principle components results of an exploratory factor analysis of the RAI scale. Only factor loadings greater than 0.3 are shown. In line with our EFA finding that the expected three factors of the RAI do not appear for our population, a confirmatory factor analysis (CFA) of the data finds that a three-factor model (consisting of coerced, introjected, and autonomous factors) does not fit well for our data. However, a model with two motivational factors (an autonomous-introjected factor and a coerced factor), displays a satisfactory fit when taking into account the correlated measurement errors of the parcel-selling category, with an RMSEA of 0.056 (0.045, 0.068) and a Tucker- Lewis Index of 0.920. This model holds within our sample population for both men and women. Thus, similar to Vaz, Pratley and Alkire (2016), who similarly find two dimensions of motivations within their sample in Chad rather than three, we use an alternative Relative Autonomy Index in our analysis consisting of two subscales. This alternative RAI consists of a “Coerced” subscale, with a weight of -3, and an “Autonomous/Introjected” subscale, with a weight of 3, resulting in a range of -9 to 9 as with the original RAI. Several possible explanations present themselves for why a two-dimension RAI displays a better fit for our population. The first is that social norms have been internalized enough by our population that the distinction between introjected and autonomous motivations is not particularly meaningful. Put more concretely, choosing to plant a certain crop out of a desire to 34 be seen as a good farmer and doing so because one thinks it is the best crop to plant may be essentially the same concept, if one assumes that community consensus is more or less accurate. Indeed, in our qualitative work, respondents often struggled to distinguish their own ideal from a community ideal, and personal preference almost always aligned with their interpretation of the community’s norms. In a relatively collectivist society where defying community expectations isn’t seen as normatively superior to following them, introjected and autonomous motivations may be harder to distinguish and may cluster together. Alternatively, this finding may be related to aspects of the study and the questionnaire: the relatively small sample and small number of domains included in the RAI (one of which, selling or leasing parcels, was more sensitive than expected). While this questionnaire was extensively piloted in the field and revised based on field experiences, further cognitive testing of the RAI scale, including the use of vignettes, to better capture introjected motivation in the Philippine context is desirable. 35 Appendix C: Robustness Checks C1. Relationship between autonomy and measures of decision-making by domain Table C1a: Decision-maker status by domain (Women only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.197 0.129 0.165 0.161 0.074 0.292 0.116 (0.094)** (0.146) (0.100)* (0.096)* (0.104) (0.109)*** (0.105) RAI (Adjusted) 0.145 0.069 0.096 0.138 0.080 0.295 0.034 (0.094) (0.148) (0.100) (0.095) (0.103) (0.110)*** (0.106) N 453 180 404 440 399 298 348 R2 0.084 0.111 0.071 0.073 0.079 0.121 0.089 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent considers themselves a decision-maker in the domain, either solely or jointly. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 0.412, 0.747, 0.516, 0.412, 0.538, 0.059, 0.750. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 36 Table C1b: Decision-maker status by domain (Men only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.008 0.033 0.050 -0.127 0.016 0.068 -0.037 (0.169) (0.320) (0.146) (0.164) (0.163) (0.219) (0.256) RAI (Adjusted) -0.021 0.206 0.038 -0.155 0.065 0.147 -0.153 (0.151) (0.276) (0.148) (0.165) (0.161) (0.228) (0.279) N1 423 156 384 426 383 264 320 R2 0.066 0.160 0.053 0.063 0.067 0.064 0.073 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent considers themselves a decision-maker in the domain, either solely or jointly. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 1.000, 1.000, 1.000, 1.000, 1.000, 1.000, 1.000. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. Table C1c: Decision input by domain (Women only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.238 0.256 0.263 0.244 0.281 0.263 0.309 (0.094)** (0.149)* (0.098)*** (0.092)*** (0.104)*** (0.115)** (0.104)*** RAI (Adjusted) 0.197 0.221 0.225 0.224 0.265 0.301 0.244 (0.095)** (0.153) (0.099)** (0.092)** (0.103)** (0.116)** (0.106)** N2 453 180 404 440 399 298 348 R2 0.088 0.122 0.082 0.081 0.094 0.120 0.103 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent reports having some or all decisions in the domain. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 0.034 , 0.046, 0.034, 0.034, 0.034, 0.034, 0.034. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 37 Table C1d: Decision input by domain (Men only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.225 0.436 0.055 0.060 0.349 0.046 0.037 (0.102)** (0.160)*** (0.105) (0.105) (0.114)*** (0.148) (0.132) RAI (Adjusted) 0.168 0.266 -0.021 -0.004 0.298 -0.035 -0.060 (0.102)* (0.159)* (0.106) (0.104) (0.112)*** (0.145) (0.137) N 423 156 384 426 383 264 320 R2 0.071 0.170 0.053 0.061 0.082 0.063 0.073 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent reports having some or all decisions in the domain. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 0.248 , 0.248, 1.000, 1.000, 0.062, 1.000, 1.000. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. Table C1e: Ability to make decisions by domain (Women only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.376 0.467 0.389 0.325 0.348 0.349 0.327 (0.100)*** (0.153)*** (0.102)*** (0.095)*** (0.108)*** (0.117)*** (0.108)*** RAI (Adjusted) 0.272 0.350 0.306 0.266 0.308 0.423 0.257 (0.102)*** (0.160)** (0.103)*** (0.096)*** (0.109)*** (0.120)*** (0.109)** N 453 180 404 440 399 298 348 R2 0.095 0.131 0.090 0.084 0.096 0.135 0.103 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent reports being able to make their own personal decisions in the domain if they wanted to. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 0.010 , 0.012, 0.010, 0.010, 0.010, 0.004, 0.011. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 38 Table C1f: Ability to make decisions by domain (Men only) Growing Buying and Choosing Rearing Choosing Buying Hiring and crops for sale selling farm where and livestock seeds for agricultural paying and household equipment how to sell food and inputs such laborers consumption crops cash crops as fertilizer RAI (Unadjusted) 0.385 0.260 0.228 0.298 0.339 0.235 0.321 (0.097)*** (0.165) (0.103)** (0.096)*** (0.106)*** (0.128)* (0.117)*** RAI (Adjusted) 0.300 0.162 0.119 0.223 0.225 0.160 0.247 (0.099)*** (0.163) (0.105) (0.095)** (0.109)** (0.134) (0.123)** N 423 156 384 426 383 264 320 R2 0.085 0.163 0.056 0.073 0.077 0.068 0.085 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes are indicator variables for whether the respondent reports being able to make their own personal decisions in the domain if they wanted to. RAI has been converted into standard deviations. R2 of the adjusted model is reported. BKY (2006) Sharpened Q-Values of Impact (Adjusted Model): 0.020 , 0.160, 0.147, 0.062, 0.072, 0.147, 0.072. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 39 C2. Alternative codings of decision-making variables C2a: Decision-maker Status (Women only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Decision-maker 0.100 0.079 index (z-score) (0.044)** (0.045)* Decision-maker 0.274 0.217 index (0.121)** (0.122)* (percentage) Sole decision- 0.055 0.040 maker (z-score) (0.048) (0.050) Sole decision- 0.191 0.139 maker (0.167) (0.173) (percentage) N 496 496 496 496 496 496 496 496 R2 0.010 0.078 0.010 0.078 0.003 0.073 0.003 0. 073 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 The decision-maker index represents the percentage of agricultural domains where the respondent reports being a decision-maker (either solely or jointly). RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. C2b: Decision-maker Status (Men only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Decision-maker 0.015 0.012 index (z-score) (0.044) (0.044) Decision-maker 0.070 0.055 index (0.204) (0.203) (percentage) Sole decision- 0.030 0.038 maker (z-score) (0.045) (0.046) Sole decision- 0.079 0.099 maker (0.120) (0.121) (percentage) N 462 462 462 462 462 462 462 462 R2 0.000 0.064 0.000 0.064 0.001 0.065 0.001 0.065 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 The decision-maker index represents the percentage of agricultural domains where the respondent reports being a decision-maker (either solely or jointly). RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 40 C2c: Input into Decisions (Women only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Input on all 0.161 0.143 decisions (z-score) (0.045)*** (0.045)*** Input on all 0.436 0.386 decisions (0.123)*** (0.123)*** (percentage) Input on most or 0.120 0.116 all decisions (z- (0.045)*** (0.046)** score) Input on most or 0.371 0.358 all decisions (0.138)*** (0.141)** N 496 496 496 496 496 496 496 496 R2 0.026 0.091 0.026 0.091 0.015 0.085 0.015 0.085 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes represent the percentage of agricultural domains in which the respondent reports having input on decisions. RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. C2d: Decision-maker Status (Men only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) Input on all 0.113 0.070 decisions (z-score) (0.043)*** (0.042)* Input on all 0.340 0.210 decisions (0.129)*** (0.127)* (percentage) Input on most or 0.121 0.096 all decisions (z- (0.041)*** (0.041)** score) Input on most or 0.450 0.355 all decisions (0.152)*** (0.152)** N 462 462 462 462 462 462 462 462 R2 0.013 0.069 0.013 0.069 0.015 0.073 0.015 0.073 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes represent the percentage of agricultural domains in which the respondent reports having input on decisions. RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 41 C2e: Ability to make a decision (Women only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) High decision 0.223 0.185 autonomy (z-score) (0.045)*** (0.046)*** High decision 0.637 0.528 autonomy (0.129)*** (0.130)*** (percentage) High or medium 0.193 0.170 decision autonomy (0.045)*** (0.045)*** (z-score) High or medium 0.516 0.457 decision autonomy (0.120)*** (0.120)*** N 496 496 496 496 496 496 496 496 R2 0.050 0.103 0.050 0.103 0.037 0.099 0.037 0.099 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes represent the percentage of agricultural domains in which the respondent reports being able to make their own personal decisions if they wanted. RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. C2f: Ability to make a decision (Men only) RAI (no RAI RAI (no RAI RAI (no RAI RAI (no RAI controls (controls) controls) (controls) controls) (controls) controls) (controls) (1) (2) (3) (4) (5) (6) (7) (8) High decision 0.184 0.127 autonomy (z-score) (0.046)*** (0.047)*** High decision 0.477 0.328 autonomy (0.119)*** (0.121)*** (percentage) High or medium (0.043)*** (0.042)** decision autonomy 0.411 0.268 (z-score) High or medium 0.411 0.268 decision autonomy (0.133)*** (0.131)** N 462 462 462 462 462 462 462 462 R2 0.034 0.078 0.034 0.078 0.017 0.071 0.017 0.071 OLS models with marginal effects. Robust standard errors in parentheses. * p<0.1 ** p<0.05; *** p<0.01 Outcomes represent the percentage of agricultural domains in which the respondent reports being able to make their own personal decisions if they wanted. RAI has been converted into standard deviations. LASSO-selected controls for the adjusted model include ARB status, education, age difference between spouses, and province fixed effects. 42