Critical review of the Reading

Summary of the reading

Critical evaluation of the reading

It is tempting to assume that climate change equally influences the lives of women and men because the most visible effects occur on societal scales. Yet, because these effects are refracted through the economic, social, and political characteristics of the polity, the reality is that climate change poses a gender-specific set of risks that create disproportionate hardships for women. This paper eval- uates whether the unequal distribution of costs women bear as a result of climate change are reflected across broader macro-social institutions to the detriment of gender equality. While existing scholarship has evaluated climate change's effects on women, and on gendered dimensions of climate vulnerability and adapta- tion in specific locations, questions remain as to the extent that environmental processes associated with climate change affect gender equality and women's rights. This paper addresses this lacuna through an empirical investigation into the impact of cli- mate shocks and climatic natural disasters on women's social and economic rights across a sample of developing states from 1981 to 2010. Vulnerability provides a conceptual framework for evaluating the impact of climate change on gender equality. Vulnerability comprises exposure and sensitivity to environmental threats, and capacity to cope with environmental crises (IPCC, 2001). Impover- ished populations face higher levels of risk: they are more reliant on ecosystem services for livelihoods; more likely to live in envi- ronmentally exposed locations such as a flood plain or on a degraded hill slope; and possess fewer resources to adapt to chang- ing environmental conditions and to recover from disasters. How- ever, the poor are not a homogenous entity. Disproportionate household and familial burdens and a relative lack of control over productive assets can enhance female vulnerability beyond that of men (Goh, 2012). In many cases, discriminatory legal institutions and social customs exacerbate these vulnerabilities by heightening exposure and undermining coping capacity. The result is that women are more likely to be impoverished than men, less capable of adapting to present and future climate change impacts, and less likely to participate in and contribute knowledge to policy-making

Building on these insights, I argue that gender disparities in cli- mate change vulnerability not only reflect preexisting gender inequalities, they also reinforce them. Inequalities in the owner- ship and control of household assets and rising familial burdens due to male out-migration, declining food and water access, and increased disaster exposure can undermine women's ability to achieve economic independence, enhance human capital, and maintain health and wellbeing. Consequences for gender equality include reductions in intra-household bargaining power, as women become less capable of generating independent revenue. Outside the home, norms of gender discrimination and gender imbalances in socio-economic status should increase as women are less able to participate in formal labor markets, join civil soci- ety organizations, or collectively mobilize for political change. The outcome of these processes can reduce a society's level of gender equality by increasing constraints on the advancement of laws and norms that promote co-equal status. While we should not expect these findings to apply to all women in equal measure, those of lower socio-economic status and those who rely on agri- culture as a means of subsistence and production should be acutely vulnerable. Empirical analysis substantiates these arguments. I test the relationship between climate change and gender equality with panel data from 1981 to 2010 for all countries classified by the International Monetary Fund as ''Developing" or ''Emerging Market Economies" (IMF, 2017). The findings suggest that climate shocks and climatic disasters exert a broadly negative impact on gender equality in these countries, as deviations from long-term mean temperatures and increasing incidence of climatological and hydro-meteorological disasters are associated with declines in women's economic and social rights. These effects appear to be most salient in states that are relatively less-democratic, with greater dependence on agriculture, and lower levels of economic development. 2. Gendered vulnerability to climate change Scientists now agree with a high level of certainty that contem- porary changes to the Earth's climate are unparalleled in recorded human history (IPCC, 2014). Increases in average global tempera- tures are fueling environmental processes that decrease the pre- dictability of rainfall and moisture content of soils, elevate the intensity of environmental hazards, reduce biodiversity, and alter wildlife migration. Although some areas in far northern latitudes might experience benefits in the form of prolonged growing sea- sons, the macro-level impacts of climate changeincreasing sea- sonal variability, glacial melting, rising sea levels, and altered precipitation patternsare expected to increase proximate risks from storms, droughts, floods, landslides, fires, disease epidemics, and heat waves across much of the world. These effects will be especially pernicious in developing nations located in tropic and sub-tropic latitudes that rely on agriculture for subsistence and livelihoods, because these states are acutely vulnerable to climato- logical hazards that undermine agricultural productivity, and because they possess fewer resources to invest in adaptation. The poor, who tend to rely most heavily on ecosystem services, will be among the hardest hit, especially (but not exclusively) in rural areas where reduced water availability and agricultural output diminish livelihood options, undermine food security and subsis- tence capabilities, and necessitate outward labor migration. Although these processes affect both women and men, the nat- ure and degree of the impacts can vary accordingly. Prior scholar- ship in feminist political ecology and disaster studies offers insight into the causal drivers of gender disparities in climate change vul- nerability (Mies & Shiva, 1993; Mies, 1998; Rocheleau, Thomas- Slayter, & Wangari, 1996; Salleh, 1997; Terry, 2009). These works assess how social and economic structures that ascribe distinct roles to women in society also expose them to distinct constella- tions of environmental risk. Gender imbalances in the division of labor and asset ownership, and the persistence of discriminatory laws and social norms that restrict women's rights and opportuni- ties magnify the hardships women face in adapting to environmen- tal conditions that reduce livelihood opportunities and heighten resource scarcities. Gendered divisions of household labor common across the developing world disproportionately amplify women's vulnerabil- ity to climate change. While both sexes contribute to household preservation, men's responsibilities generally include cash- cropping or wage labor, while women's concern the management of resources necessary to ensure family nutrition and healthtend- ing subsistence crops and small livestock, collecting water, and gathering fuel wood (FAO, 2003). When a climate shock disrupts income flows or food cultivation, or necessitates changes to water supplies or the distribution of crops, women often face greater challenges with adaptation. Buechler (2009), for example, finds that reduced water supplies due to higher temperatures in Sonora, Mexico amplify women's household responsibilities, as their tradi- tional role as caregivers require them to remain stationary while men migrate to find employment. At the same time, decreasing employment opportunities in agricultural processing sectors diminish women's livelihoods, which undermine their economic self-sufficiency and reduce their ability to participate in customs such as gift giving that confer social status among female commu- nity members. In this case, the combination of rising household burdens, declining access to subsistence and financial resources, and limited opportunities to participate in the workforce and develop human and social capital serve to exacerbate gendered disparities in climate change vulnerability. Inequalities in the ownership and control of tangible assets such as land, housing, livestock, and agricultural inputs can magnify these effects. Naraya, Patel, Schafft, Rademacher, and Koch- Schulte (2000, 5) argue that: ''poor people rarely speak of income but focus instead on managing assetsphysical, human, social, and environmentalas a way to cope with their vulnerability, which in many areas takes on gendered dimensions." Assets, espe- cially land, are critical because ownership can provide physical protection, a way to mitigate and manage crises, and adapt liveli- hood strategies to changing environmental conditions (Deere and Doss, 2006). Across much of the developing world, land ownership is overwhelmingly male. In Africa, women are responsible for between 50 and 80% of agricultural production, but hold title to less than a 20% all agricultural land (FAO, 2016).1 This disparity can create hardships for women when climatic changes under- mine agricultural livelihoods, because ownership increases access to formal credit markets, which can enable individuals to cope with lost harvests, invest in new livelihood strategies, or purchase agricultural inputs that can reduce production volatility (Nuryartono, 2005). This ''gender-asset gap" can magnify female vulnerabilities in other ways as well (Deere & Doss, 2006; Deere & Leon, 2003). Dillon and Gill (2014) find that gender inequalities in access to agricultural technologies (irrigation equipment, motorized tillers, 1 There is considerable interstate variation in land ownership: in Mali, women hold only around 5% of agricultural land titles; in Zambia, the number is around 15%, and Malawi, 32% (FAO, 2016). Figures are similar for South and Southeast Asia, and Latin America, and much lower for the Middle East (FAO, 2016). However, even when women technically ''own" assets, husbands or extended family members can often mediate control See for example: Quisumbing, Roy, Njuki, Tanvin, and Waithanji et al. (2013). 290 J. Eastin/World Development 107 (2018) 289-305 etc.) can aggravate disparities in crop production during a climate shock. While men's control over these resources enables them to maintain output sufficient to almost offset the shock, women expe- rienced proportionally diminished harvests. The impacts of climate change on the gender-asset gap can be self-reinforcing when vulnerable populations are forced to sell their assets in response to declining livelihoods, and when degra- dation and sea-level rise decrease their use- and exchange- values. Although evidence for a rising gap between husbands and wives is limited (Quisumbing, Kumar, & Behrman, 2017) climate change is likely to have a significant effect on the gap among female-headed households and dual-parent or male-headed households (Flat, Muttarak, & Pelser, 2017). Female-headed households are more likely to be impoverished and chronically food insecure, more likely to have a higher number of non- working dependents, and more likely to experience greater limita- tions on mobility and earnings potential because women must per- form both domestic duties and act as primary breadwinners (Rosenhouse, 1989; Flat et al., 2017).2 A climatic shock that under- mines livelihoods should only amplify these burdens, leaving asset sales and savings expenditures as the most viable coping mechanisms.3 Finally, patriarchal social and legal institutions prevalent in many developing states exacerbate both gendered disparities in vulnerability as well as the relative difficulties women face in adapting to climate change. For example, traditional inheritance and marriage dissolution practices exacerbate the gender asset gap (Peterman, 2012). Greater inheritances for women are associ- ated with higher socio-economic status, indicating that this prob- lem is especially acute among the poor. Customs such as these can even conflict with formal laws: In Tajikistan, the law autho- rizes joint land ownership between husbands and wives, yet few women are registered landowners because many rural marriages are not documented, and therefore not covered under the law (Djusaeva, 2012). In India, men and women hold equal land own- ership rights, but women own less than 10% of all private landhold- ings. One reason for this disparity is that families view dowries paid to sons-in-law as a portion of daughters' inheritances even though the assets are transferred to the husband and his family (Scalise, 2009). The result is that in cases of disagreement, divorce, separation, or death, women stand to lose not only assets, but also future income sources and a hedge against disaster. The effects of discriminatory institutions on female vulnerabil- ity to climate change are also evident when assessing gender dis- parities in health and wellbeing. Because women's responsibilities in the household division of labor usually include gathering resources necessary for family subsistence, environmen- tal changes that decrease resource access and amplify domestic burdens can also undermine women's health, in some cases plac- ing them direct physical jeopardy. Reductions in the availability of water and fuel-wood increase the risk of sexual assault when women are forced to travel farther for collection (Brody, Demetriades, & Esplen, 2008). This scenario is especially prevalent in conflict zones, refugee camps, and other ungoverned or insecure spaces. Scholars have also found that drought-induced economic shocks and crop failures can increase the risk of disease when women are compelled to engage in transactional sex as a means of income generation, or when husbands expose wives to sexually transmitted infections upon returning from labor migrations (Burke, Gong, & Jones, 2015; See also: Swidler & Watkins, 2007; Dinkelman, Lam, & Leibbrandt, 2008; Robinson & Yeh, 2011). Natural disasters, which are expected to grow in magnitude as climate change progresses, pose similar challenges. Customs such as traditional dress codes and norms against teaching women to swim can heighten female death rates during floods (Alam & Collins, 2010), while women's responsibilities as familial caregivers can magnify the difficulties of self-rescue because of the need to attend to children and the elderly (Schwoebel & Menon, 2004). Sim- ilarly, discriminatory practices in a disaster's post-crisis phase can underminewomen'saccesstoreliefresourcesandincreaseinterper- sonalviolenceagainstthem(Neumayer & Plmper2007).Caseanal- yses provide supportive evidence. In Namibia, the tendency for wives to act submissively toward their husbands increases physical and psychological stress in times of food insecurity, because it is expected that women will attempt to exhaust all available means of resolving scarcities before speaking with their husbands (Angula, 2010). In Kenya, it is common for women to deliberately decrease food consumption in response to drought-induced food shortages, which increases health risks among women, children, and pregnant and lactating mothers (Serna, 2011). 3. Vulnerability and gender equality This paper argues that gendered disparities in climate change vulnerability not only reflect preexisting gender inequalities, they also reinforce and strengthen them. The result can perpetuate a ''vicious circle [whereby] the more women are affected negatively by climate change, the worse the inequalities get. And the worse the inequalities get, the worse the impact becomes" (Panitchpakdi, 2008, 107). Work in neoclassical and feminist eco- nomics on intra-household bargaining behavior lends insight into causal dynamics (Folbre, 1986, 1989). In household bargaining models, household decision-making outcomes are conceptualized as the product of preference contestation among household mem- bers. An individual's contribution to household income and their control over household assets determine the bargaining power they command (Rogers, 1990). For women, gains in bargaining power can promote empowerment by increasing their participa- tion in decisions over childbearing and care, allocation of house- hold resources, mobility, and occupational participation (Iversen & Rosenbluth, 2006; Ashraf, Karlan, & Yin, 2010; Aslam & Kingdon, 2012; Thomas, Contreras, & Frankenberg, 1997). In the context of climate change, gender disparities in vulnera- bility are likely to magnify inequalities in intra-household bargain- ing power, with implications for gender equality and women's rights at both the micro- and macro-levels. Rising familial burdens due to male out-migration and declining food and water security, coupled with a relative inability to employ productive assets to cope with climatic shocks, should decrease female income genera- tion capabilities, mobility, opportunities to build human capital, and access to formal credit markets. The result can reinforce gen- der disparities in the household division of labor and heighten female reliance on male income, which increases the opportunity costs of divorce and reduces women's independence. These effects can also increase the likelihood that gender discrimination and stereotypes will persist across generational divides. Because par- ents have an incentive to prepare children for what they expect their adult responsibilities to be, and because children learn via socialization, the greater the level of female subordination in the household, the more likely these norms will be passed along (Iversen & Rosenbluth, 2006; Eagly & Steffen, 1984). Environmental changes that undermine women's ability to gen- erate independent revenue streams can also affect gender equality 2 While the percentage of female-headed households varies significantly across the developing world, many of the least developed states possess the highest numbers, with states such as Swaziland, Zimbabwe, and Haiti approaching 50% of all households (World Bank WDI, 2016).3 Preexisting inequalities coupled with increased familial burdens can potentially undermine other possible coping mechanism such as credit market access or the opportunity to migrate to more prosperous areas. J. Eastin/World Development 107 (2018) 289-305 291 by reducing children's health and educational opportunities (Mishra & Sam, 2016; Menon, Van Der Meulen Rodgers, & Nguyen, 2014). Education is important because it enables girls to enhance their human capital and future earnings potential, delay marriage and childbirth, and encourage contraceptive use, and because female education can facilitate the development of social norms that discourage gender discrimination. The greater the share of women's contribution to household income, and the greater her control over household assets, the more likely girls will be to attend school. Qian (2008), for example, finds that increasing women's share of household revenue in China increases educa- tional participation for both male and female children. In contrast, when a man's share of household revenue increases relative to woman's, girls' educational participation declines and boys' remains static. Thus, while children of both sexes can benefit from increasing female empowerment, reductions in female income dis- proportionately hurt girls (See also: Foster, 1995). Additionally, when household livelihood and subsistence capabilities diminish overallwhen both male and female income declinesgirls also bear greater costs. In a study of the effects of climate variability on children's education in Uganda, Bjrkman-Nyqvist, (2013) finds that negative deviations in mean annual rainfall (droughts), have a significant adverse impact on primary school enrollment and test scores for older girls, but no impact on school attendance or perfor- mance for boys or younger girls. The findings suggest that during episodes of resource scarcity, 1) households prioritize resource access for male over female children, and 2) are more likely to use girls as supplementary domestic labor, which reduces their school attendance. Similar findings hold true for girls' health: Qian's (2008) study also finds that increasing a woman's share of household revenue improves daughters' survival rates. This out- come occurs because greater bargaining power enables women to demand a larger share of household resources for daughters, and because the potential that girls hold future value in the labor market can increase the opportunity costs of childhood gender dis- crimination in access to these resources. As this discussion indicates, the impact of increased vulnerabil- ity to climate change should not only affect gender relations at the household level, but should also have consequences for women's rights at the macro (societal) level. Factors that magnify domestic burdens and undermine women's status within the home should also reduce their ability to develop human capital, build robust social networks, and join civil society organizations. Consequences can include: increasing constraints on women's ability to achieve gainful employment, especially in non-sexually stratified occupa- tions; achieve authoritative positions in the workforce; and to col- lectively mobilize for public goods to support women's issues. The outcome can reduce public support for laws and norms that pro- mote coequal status, and increase impediments to the develop- ment of post-materialist values that discourage gender discrimination (Inglehart 1977, 1997). Taken together, these insights inform the following hypothesis: H1. Environmental processes associated with climate change should generate a negative effect on women's rights in vulnerable states. H1 enables an empirical assessment of the relationship between climate change and women's rights. It rests on the assumption that vulnerability and livelihood sensitivity to climatic stressors are important conditional determinants. Accordingly, structural conditions that bear on household and gendered vulner- ability disparities to climate change should influence the magni- tude of these impacts. Households in relatively poorer and more agriculturally dependent economies are likely to possess greater exposure and sensitivity to climatic processes and reduced capac- ity to cope with climatic disasters. Similarly, relatively less- democratic political institutions are likely to correlate with lower levels of responsiveness to the needs of a population, and lower governmental capabilities (or political will) to provide social ser- vices that can mitigate or alleviate livelihood shocks. While each of these characteristics should influence a population's absolute vulnerability to climate change, they might also affect disparities in vulnerability among women and men (Denton, 2002). Poorer economic prospects, greater reliance on agriculture, and less democratic governmental structures can also be associated with more substantial female household burdens, reduced opportuni- ties for female empowerment and human capital enhancement, greater prevalence of patriarchal norms that undermine women's status, and lower social support for women's issues that can decrease vulnerability (Goldin, 1995; Eastin & Prakash, 2013). The preceding logic informs hypotheses two (H2). H2. Climate change should have a greater impact on women's rights in less-developed states with greater dependence on agriculture as a means of production, and less-democratic political institutions, than thosethatarerelativelymoredeveloped,industrialized,anddemocratic. 4. Data and variables To test the impact of climate change on women's rights, I have compiled a cross-sectional, time-series dataset of developing states from 1981 to 2010. The unit of analysis is the country/year. The time period was selected because it corresponds to data availability of the key variables of interest. Countries were included in the sample if they fall into the ''Emerging Market" or ''Developing" economy categories of the International Monetary Fund's World Economic Outlook Database (IMF, 2017). These clas- sifications are based on countries' 1) per capita income levels, 2) their integration into world financial markets, and 3) their level of export diversification, a category which prevents countries with a high GDP per capita from classification as ''advanced" economies if they derive a significant share of their income from petroleum exports. In total, 151 countries are included in the IMF database, though the maximum number of countries included in any one estimation is 127 because of disparities in data availability among the variables included in the analysis. A full list of all countries included is the dataset is located in the Appendix. To the best of my knowledge, this analysis provides the most comprehensive (and perhaps the only) cross-national assessments of climate change on women's rights. 4.1. Dependent variables The dependent variables capture a society's adoption of women's rights and norms of co-equal status as an approximation of gender equality. Data is drawn from the Cingranelli-Richards Human Rights Dataset (CIRI), which has developed indices that evaluate these concepts across countries over time (Cingranelli, Richards, & Clay, 2014). Social Rights measures whether women in society hold rights to equal inheritance; to own and control property; to live where they like; to marry whom they choose and initiate a divorce; to obtain a passport and travel abroad; to confer citizenship to a spouse or child; to participate in social and community activities on an equal basis; to be educated; and to be free from the threat of genital mutilation and forced steriliza- tion. Economic Rights measures whether women enjoy rights to equal pay for equal work; to equality in hiring practices; to choose whether they work and their specific profession without male con- sent; to equal levels of job security; to non-discriminatory employ- ment practices; to enjoy a workplace free of harassment; and to work at night, in dangerous jobs, and in the military and police 292 J. Eastin/World Development 107 (2018) 289-305 forces. Both of these variables are indices that are ranked on an ordinal scale from zero to three. Higher numbers indicate greater equality. The rankings depend both on whether these rights exist in law (de jure), and the extent that the government actively enforces them (de facto). Further details on coding practices, as well as information on primary data sources can be found in Cin- granelli, Richards, and Clay (2013). The CIRI indices on women's social and economic rights are preferable to alternative measures of gender equality because they most closely approximate the outcomes discussed in the study, and because they are among the most comprehensive and robust measurements available. Other indices, such as the UNDP's Gender Development Index (GDI), Gender Equality Measure (GEM) and Gender Inequality Index (GII), have far lower data coverage, while, the GDI and GEM have been the subject to criticism for their unre- liability and lack of comparability across cases (Klasen, 2006). An alternative strategy might also have been to choose a variable that captures an isolated dimensions of gender equalitygender dispar- ities in infant mortality, for examplehowever measurements such as this are less comprehensive and much less precise for mea- suring women's rights across various legal and social institutions. Data for Social Rights is available from 1981 to 2005, while data for Economic Rights is available from 1981 to 2010. Although thirty years is a relatively short period, it is adequate to capture inter- temporal variation in CIRI scores sufficient to justify empirical analysis. For example, Guyana transitions between scores for Eco- nomic Rights fifteen times between 1981 and 2010; Sudan transi- tioned ten times, Turkey twelve times, Iraq six times, and Thailand ten times. 4.2. Key independent variables Key independent variables approximate environmental pro- cesses associated with climate change. The primary measurement gauges the occurrence of climatic shocks, which are expected to rise in frequency and magnitude as climate change progresses. Temperature measures standardized deviations in inter-annual temperature estimates from the panel mean for each country in the sample. The formula for calculating this measure is Xit Xi=ri, where Xi is the panel mean for country i, Xit is the temperature estimate for country i at time t, and ri is country i's standard deviation. This process results in a standardized temper- ature measure with a mean of approximately zero (1.49e-08), a range of 2.9 to 3.3, and a standard deviation of 0.98. I also include Precipitation, which measures standardized country-level rainfall deviations using the same formula. The Precipitation variable has a mean of approximately zero (7.34e-10), a range of 3.6 to 4, and a standard deviation of 0.98. Raw data for both these variables are drawn from Burke, Hsiang, and Miguel (2015). These authors derive annual country-level esti- mates from the ''Terrestrial Air Temperature and Precipitation: Monthly and Annual Time Series Dataset (v. 3.01), which provides global annual mean temperature and precipitation data at a 0.5 0.5 degree resolution (Matsuura & Willmott, 2012). Burke et al. aggregate these grid cell values to the country-year, and weight them by population density in the year 2000, using data from the Gridded Population of the World. Data for Temperature and Precip- itation, variables are available across the entire time period under study, from 1981 to 2010. While I include both Temperature and Precipitation measure- ments in the analysis to account for the possibility that one effect might mask the other, I anticipate Temperature to be the more robust measure, and positive temperature shocks to generate greater impact across cases. Rising temperatures and temperature variability are the most consistent environmental effect of climate change, and can generate an array of adverse effects for livelihoods and food security, particularly in developing nations in tropic and sub-tropic latitudes (IPCC, 2014). While rainfall shocks can also influence these processes, the directionality of rainfall's effects is not universal, which might affect the interpretation of their impact on the outcome (Eastin, 2016). In other words, in some locations such as the Sahara, Sahel, and some states in Sub-Saharan Africa, a positive rainfall shock can encourage agricultural growth, and therefore be a boon to food security and livelihoods. In contrast, too much rain in rice-producing states in Southeast Asia, especially at certain times of the year, can diminish production. Thus, the pre- dicted directionality of Rainfall is less consistent. Temperature and Precipitation provide a useful approximation of climate change because they are entirely exogenous to existing socio-economic and political processes that could influence gender equality independent of the climate. However, a challenge arises from attempts to measure these variables at the country/year level, because doing so carries the possibility of concealing intra-country and intra-year climatic variations. The problem is perhaps most acute with Precipitation, because of a greater potential for masking environmental hazards associated with rainfall variation over time. In other words, a year with significant floods and droughts could register as a ''normal" year if the overall rainfall totals hewed closer to long-term averages. To help offset this concern, I incorporate an additional measure of climate change: a count variable of the annual incidence of cli- matological and hydro-meteorological natural disasters occurring in a country. These include: droughts, floods, wildfires, storms, insect infestations, disease epidemics and heat waves. As climate change progresses, increasing prevalence of these disasters are anticipated to be a primary outcome (IPCC, 2014). Yet, while disas- ters provide an alternative measure of climate change and have been employed in prior studies (Neumayer & Plmper, 2007), they are arguably a more controversial choice because of the fact that impoverished populations, which are more likely to experience lower levels of gender equality, are also more vulnerable to disas- ter (Klasen, 2002). Nevertheless, when included alongside models with the climatic data, these variables can increase analytical robustness and inspire greater confidence in the findings. IdrawdataforthedisastervariablesfromtheWorldHealthOrga- nization Collaborating Center for the Epidemiology of Disasters (CRED). According to CRED, a disaster occurs when:''(1) Ten or more people reported killed. (2) One hundred or more people reported affected. (3) Declaration of a state of emergency. (4) Call for interna- tional assistance." Approximately half (54%) of the country/year observations had zero disaster incidents, 41% had between one andfive,3%hadbetweensixandten,and2%hadtenorgreaterThese dataareavailablefortheentiretimeperiodunderstudy,1981-2010. 4.2.1. Control variables I include a number of control variables in this analysis to account for alternative explanations that might affect the out- comes independent of the primary variables of interest. Variables were chosen based on both their theoretical relationship to the processes described in this study, and on findings from earlier research citing them as key influences on gender equality and women's rights. Data for all control variables are available for the entire time period under study, 1981-2010. Democracy. I control for democracy because it is often assumed that democratic regimes have greater respect for human rights, including women's rights, relative to authoritarian regimes (Inglehart, Norris, & Welzel, 2002). I draw data from the well- known Polity IV dataset, which ranks governments on a 21-point scale from 10, ''strongly autocratic" to 10, ''strongly democratic" (Marshall, Gurr, & Jaggers, 2016). J. Eastin/World Development 107 (2018) 289-305 293 Conflict. Gender equality suffers in conflict situations through direct victimization of women and through spousal and child loss. Prolonged periods of conflict can also limit a society's capacity to cultivate and develop social norms and invest in social programs that promote gender equality (Jansen, 2006). I control for a coun- try's involvement in international or domestic conflict by including a dichotomous variable coded one if a country was a participant in a conflict and zero if not. The data are from the Uppsala Conflict Data Program and International Peace Research Institute, which defines conflict as: ''A contested incompatibility that concerns gov- ernment and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths" (Gleditsch, Wallensteen, Eriksson, Sollenberg, & Strand, 2002). GDP/capita. I control for economic conditions that may influence women's rights. GDP/capita measures a country's GDP per capita, and is a conventional metric of economic development. A com- monly held assumption in the development literature is that more developed countries are also likely to have greater respect for women's rights. I also include GDP/capita2 to test for a curvilinear relationship between economic development and women's rights. Prior work has found the impact of development on women's rights to be contingent on the particular stage of development (Eastin & Prakash, 2013; Forsythe, Korzeniewicz, & Durrant, 2000; Boserup, 1970). Data for the GDP/capita variable are taken from the World Bank Databank. Foreign direct investment (FDI). Empirical work finds that foreign direct investment can be beneficial for a society's level of gender equality because investment flows can generate employment opportunities for women (Richards & Gelleny, 2007). However, some scholars contend that policies designed to increase FDI can also reduce state revenue, and therefore diminish state's capacity for social service provision (Hemmati, 2001; Rao & Kelleher, 2005). Because women are often the key beneficiaries of these ser- vices, FDI might undermine gender equality. Furthermore, eco- nomic integration can solidify gendered occupational segregation, which forces women into poorly paid jobs (UNDP, 1999). Thus, I include FDI, but am agnostic as to the directionality of its effect. Data for the FDI variable are taken from the World Bank Databank. Urbanization. Urbanization has been linked to increasing economic opportunities for women and greater independence relative to their rural counterparts (Tacoli & Satterthwaite, 2013). More numerous employment opportunities, lower fertility rates, and improved access to health care and social services are among the key causal drivers. However, research also suggests that urbaniza- tion can expose women to greater levels of violence and sexual harassment, and that many of the economic opportunities afforded women in urban settings reinforce occupational segregation and labor force stratification, which should diminish gender equality (Levy, 2013; Chant, 2013). In this analysis, I control for urbaniza- tion data that measure the percent of a country's population that resides in urban areas. Given the diverging expectations, I remain agnostic as to the potential directionality of effect. Data for the Urbanization variable are taken from the World Bank Databank. Agriculture. Finally, I control for a country's reliance on agriculture as a means of subsistence and production. It is possible that the greater the reliance on agriculture, the lower the opportunities for female formal labor force participation, which can undermine the potential for equality gains. Additionally, heavy reliance on agriculture might also be associated with the perpetuation of pre-existing gender stereotypes because of the pervasiveness of gender inequality in agricultural sectors. For these reasons, I expect a country's agricultural dependence to be negatively associated with social and economic rights for women. To ensure model robustness, I employ two measures of agricultural dependence in the models: 1) the percentage of a country's land area devoted to agriculture, and 2) the value added of agriculture to the economy as a percentage of its overall GDP. Data for the Agriculture variables are taken from the World Bank Databank. Table 1 displays the descriptive statistics for each of the vari- ables used in the analysis. 5. Models, analysis, and results To estimate the impact of climate change on women's rights, I employ a series of ordered-logistic regressions on a country-level panel data set covering the period 1981-2010. Ordered-logit is a preferred method for modeling independent variables that contain more than two categories of response and that are in sequential order (Agresti, 2010). All ordered-logit models were run with country-clustered robust standard errors to address possible auto- correlation at the unit level, and because these estimates tend to be more consistent across a broader range of possible forms of corre- lation than those derived from random effects models (Cameron & Trivedi, 2010). I include regional dummy variables in each model to account for the possibility that unobserved factors within a par- ticular geographic region affect the outcome independent of the primary variables of interest. To assess the robustness of these findings, I also estimate each model using ordered-probit regression and ordered-logistic regres- sion with random effects. Additionally, while there is no consistent, universally agreed upon method for applying fixed effects to an ordered-logistic model (Baetschmann, Staub, & Winkelmann, 2015), I experimented with ordinary least squares regressions that incorporate the fixed effects assumption. All of these techniques yield results in line with those presented here, which should increase confidence in the analysis. Furthermore, because the best predictor of a country's level of gender equality in year t is often that which is present in year t1, I also estimate models that include a lagged dependent variable as a predictor. The inclusion of this vari- able did not change the statistical significance of the variables of interest, though it did often result in higher AIC and BIC scores, indi- cating the variable's inclusion can be unnecessary. Finally, correla- tion statistics reported in the Appendix do not indicate that multi- collinearity is likely to be a significant danger in this analysis. A vari- able correlation matrix as well as results from these alternative esti- mation procedures can be found in the Appendix. Table 2 display the findings from ordered logistic regressions that estimate women's economic and social rights, respectively, across a sample of developing states from 1981 to 2010. The num- bers in these tables are odds ratios, which communicate the impact of a one-unit change in each independent variable on the odds of being in the highest category of the ordinal dependent variable, holding all other variables to their mean values. Ratios greater than one reflect a positive impact on the outcome, those lower than one reflect a negative impact, and a ratio of exactly one indicates no effect. The greater the distance from one, the greater the impact should be. The figures in the parentheses are z-scores. The chi- squared goodness of fit statistics indicate that all models presented in Table 2 are significantly distinct from the null; and likelihood ratio tests reveal that including the climate change variables improves model fit to a statistically significant degree. Variations in sample sizes in the models displayed these tables resultbothfromtemporaldisparitiesindatacoverageforthedepen- dent variables (1981-2005 for Social Rights, and 1981-2010 for Eco- nomic Rights), and spatial disparities in data coverage between the climate and disaster variables and between the two variables approximating agricultural dependence.

As these tables indicate, environmental processes associated with climate changetemperature shocks and climatological/hyd ro-meteorological disastersexert a significant and negative effect on women's social and economic rights, and are robust to a num- ber of alternative explanations. These findings provide support for H1, that climate change can reduce a society's level gender equality by undermining the prevalence of social and economic rights that promote coequal status. All of the odds ratios for Tem- perature and Disaster indicate the existence of a modest negative impact. The influence of temperature shocks appears to generate the strongest absolute effect on women's social rights among the primary models. In Model 1, the odds ratio for Temperature indi- cates that for each one-unit increase in standardized temperature means in a country, the odds of full respect for women's economic rights were only 0.854 times that which would be the case had the increase in temperature not occurred. In other words, temperature increases appear to be slowing progress towards the improvement of women's rights and gender equality. This relationship is robust across all model specifications at varying degrees. Destructive cli- matic or hydro-meteorological disasters exert a similar effect, although lower in magnitude, as the lowest odds ratio for Disaster across these models is 0.905. However, it should be noted that the Disaster variable is less robust to alternative estimation techniques than the Temperature variable, as revealed in Tables A8-A10 in the Appendix. Interestingly, Temperature appears to exert the strongest negative impact on women's rights for all variables other than the incidence of violent armed conflict. In contrast, Precipitation does not appear to have a substantive effect on the outcome, as the vari- able was statistically insignificant and ambiguous in its direction- ality of effect across all models. This finding might reflect the fact that the impact of precipitation on food security and livelihoods is less consistent across time and space than that of temperature or disaster, or that the country-year unit of analysis masks impor- tant intra-annual variations in precipitation patterns. To further probe the impact of climate change on gender equal- ity, Table 3 displays the marginal effects that variations in temper- ature and disaster incidence have on the predicted probability of a country achieving a score of 0, 1, 2, or 3 on the CIRI Economic and Table 1 Descriptive statistics. Variable Obs. Mean Median Std. Dev. Min Max Econ. Rights 3,662 1.15 1 0.60 0 3 Social Rights 2,877 0.99 1 0.67 0 3 Temperature 3,817 0.01 0 0.98 2.86 3.33 Precipitation 3,817 0.00 .041 0.98 3.63 4.10 Disaster 4,530 1.19 0 2.44 0 30 GDP/capita 4,034 2.88 1.197 4.76 0 54.48 Democracy 3,715 0.13 0 6.80 10 10 Ag. (% land) 4,270 39.51 39.27 21.73 0.45 86.01 Ag. (% GDP) 3,634 20.66 17.77 14.44 0.30 69.33 FDI 3,800 3.41 1.68 6.95 82.89 161.82 Urban pop. 4,521 44.96 43.36 21.16 4.50 98.26 Conflict 4,530 0.06 0 0.24 0 2 Table 2 Ordered logit regressions, economic rights. Economic Rights Social Rights 1 2 3 4 5 6 7 8 Temperature 0.854 0.855 0.861 0.881 (2.66)*** (2.56)** (2.78)*** (2.17)** Precipitation 0.997 0.989 0.939 0.955 (0.06) (0.23) (1.37) (0.92) Disaster 0.922 0.913 0.905 0.912 (2.45)** (3.1)*** (2.29)** (2.3)** GDP/cap. 1.12 1.136 1.113 1.127 1.041 1.163 1.041 1.107 (1.72)* (1.29) (1.71)* (1.27) (0.41) (0.9) (0.42) (0.61) GDP/cap.2 0.997 0.995 0.997 0.995 0.998 0.992 0.998 0.994 (1.74)* (1.43) (1.81)* (1.54) (0.7) (0.72) (0.79) (0.54) Democracy 1.034 1.032 1.032 1.028 1.039 1.036 1.038 1.038 (1.86)* (1.58) (1.75)* (1.45) (1.86)* (1.61) (1.84)* (1.7)* Ag. % land 0.993 0.994 0.994 0.995 (1.07) (0.82) (0.84) (0.68) Ag. % GDP 0.995 0.992 1.017 1.013 (0.4) (0.68) (1.21) (0.94) FDI 1.003 1.013 0.997 1.003 0.995 1.009 0.993 1.003 (0.29) (0.87) (0.32) (0.25) (0.57) (0.59) (0.86) (0.22) Urban pop. 0.999 0.996 1.001 0.998 1.001 1.005 1.003 1.005 (0.06) (0.33) (0.14) (0.23) (0.13) (0.34) (0.31) (0.37) Conflict 0.49 0.437 0.552 0.498 0.743 0.682 0.819 0.744 (3.25)*** (3.82)*** (2.69)*** (3.26)*** (1.41) (1.89)* (0.99) (1.53) Obs. 3032 2764 3091 2791 2404 2181 2450 2199 Countries 124 120 127 123 124 119 126 121 Prob. >v2 0 0 0 0 0 0 0 0 Ordered logit regressions with country-clustered robust standard errors. Displayed are proportional odds ratios with z-scores in parentheses. Ratios for regional binary variables are omitted for presentation. *p > .1, **p > .05, ***p > .01. J. Eastin/World Development 107 (2018) 289-305 295 Social Rights indices, with all other variables in the equations held to their central tendency. These predictions are derived from Mod- els 1, 3, 5, and 7 to maximize sample sizes. The predicted probabil- ities are calculated with the Temperature values set to two standard deviations below the mean (2SD), the mean Temperature value, and two standard deviations above the mean (+2SD); while the Disaster variable is set at zero disasters, or slightly less than one standard deviation below the mean value; three disasters, or approximately one standard deviation above the mean; and five disasters, or two standard deviations above the mean. One can draw at least two conclusions from Table 3: First, war- mer temperatures and greater incidence of disaster increase the probability that a country will achieve a low CIRI score (zero or one), and decrease the probability that a country will achieve a high CIRI score (two or three) in any given year. For example, for an ''average" country with a stable Temperature (value set to the mean), the probability of receiving a zero score on the women's Economic Rights index in any given year is 8%. An increase in tem- perature in a given year of two standard deviations above the mean increases this probability to 10.7%. In contrast, this same change decreases the probability of an average country achieving a score of two from 18.8% to 14.6%. Second, while changes in predicted probability for a given score at various values of Temperature or Disaster do not in isolation appear to be large, it is important to rec- ognize that these reveal the probability of such a change in any given year between 1981 and 2010 (or 2005, in the case of Social Rights); they do not reflect the cumulative effects of such changes over a number of years. Thus, an accumulation of years of increased temperatures and disaster incidence should exert a stronger impact on the outcome than these figures reflect. In order to assess H2, which concerns the conditionality of the findings, I split the sample set of developing countries according to their levels of economic development, agricultural dependence, and democracy. Specifically, I assess the impact of the climatic variables on samples of: 1) countries above and below the sample median for the percentage of agriculture in a country's GDP; 2) countries above and below the sample median for GDP/capita; and 3) democracies, which include those countries that rate above a five on the Polity IV scale; and non-democraciescountries that rate five and below.4 Additionally, in order to provide further context regarding the regional dimensions of the processes under study, I also estimate models on a 4) sample of countries located only the African conti- nent, and on a sample excluding African countries. There are a number of reasons to believe that countries located on the African continent worthy of consideration in isolation from the larger sam- ple set. Among them: a lack of stable, developed infrastructure; pervasiveness of violent armed conflict across the continent, including spillovers into non-conflict countries; a history of colo- nial exploitation and poor governance; the continent's location in tropic and sub-tropic latitudes, and, importantly, a heavy depen- dence on agriculture as a means of subsistence and economic pro- duction and a high degree of vulnerability to the environmental and economic effects of climate change (IPCC, 2014). Collectively, Africa also ranks last among all global regions on the Human Development Index, a comparative index that measures life expec- tancy, literacy, education, and living standards to determine a country's level of development (HDI 2017). Thus, according to the logic discussed previously, it is also possible that the impact of women's rights in Africa could be greater as well.5 For all conditionality tests, I estimate two ordinal logit models: a ''sparse" model containing only the climatic variables of interest and regional dummies, and a model containing the full suite of control variables (Salehyan & Hendrix, 2014). Both estimations yield consistent results on the key variables of interest. Figs. 1 and 2 report the odds ratios and 95% confidence intervals for the Temperature and Disaster variables drawn from sparse model esti- mations for each of the sub-samples. These figures allow assess- ment of the contingency of the relationship between climate change and women's rights. Results drawn from the full models are available in the Appendix. The results from regressions reported in Figs. 1 and 2 support H2, which suggests that climate change should exert the greatest impact on women's rights in relatively poorer, more agriculturally dependent states, with less-democratic political institutions. Com- parisons of the odds ratios for the paired sample subsets indicate that Temperature has a more substantive and statistically signifi- cant impact on Economic and Social Rights in each of these cate- gories than it does in countries with higher per capita GDPs, those that are less dependent on agriculture, or more democratic. The Disaster variables' odds ratios yield similar conclusions. The primary difference is that Disasters exert a more consistent effect on women's Social and Economic Rights across all eight of these sub-samples than Temperature, as least with respect to Economic Rights, though the impact of Temperature is slightly larger in the four sub-samples where the relationship is hypothesized to be the strongest

Countries located on the African continent also appear to be uniquely susceptible to the impacts of climatic variables on women's rights, as the odds ratios for these variables are negative and statistically significant in most models, and of greater impact than those in models run on the non-African sample set. These findings not only provide additional support for H2, but also sub- stantiate the theoretical logic supporting H1, that vulnerability and livelihood sensitivity to climate change, and in particular, gendered disparities in vulnerability, can manifest across macro- social institutions to the detriment of gender equality and women's rights. Results run on sub-samples of Sub-Saharan and Low-GDP African countries provide additional support for this argument. With respect to the control variables: First, the results reveal modest support for an inverted 'U'-shaped relationship between GDP/capita and Economic Rights in models 1 and 3, which suggests that the effect of development can be positive in very low stages of development, then turns negative as the level of development sur- passes a certain threshold. Scholars such as Forsythe et al. (2000) have suggested the inverted 'U' to be a product of exploitative effects of growth as women are forced into lower-paying discrim- .806 .863 .781 .854 .764 .805 .827 .805 1.04 .999 .992 .959 1.02 1.02 1.02 1.05 Temperature .6 .8 1 1.2 .6 .8 1 1.2 .6 .8 1 1.2 .6 .8 1 1.2 Agri. dep. GDP low Africa Nondemocracies Not agri. dep. GDP high Not Africa Democracies Fig. 1. Odds ratios and 95% confidence intervals for temperature shocks from sparse model ordered-logit estimations. Models run on sub-samples of countries above and below the sample median for agricultural dependence; above and below the sample median for GDP/capita; African and Non-African countries; and countries below and above a score of five on the Polity IV scale. Circles = Economic Rights; Squares = Social Rights. .885 .875 .895 .911 .862 .834 .907 .88 .916 .979 .927 .904 .936 .962 .913 .983 Disaster .7 .8 .9 1 1.1 .7 .8 .9 1 1.1 .7 .8 .9 1 1.1 .7 .8 .9 1 1.1 Agri. dep. GDP low Africa Nondemocracies Not agri. dep. GDP high Not Africa Democracies Fig. 2. Odds ratios and 95% confidence intervals for hydro-meteorological and climatic disasters from sparse model ordered-logit estimations. Models run on sub-samples of countries above and below the sample median for agricultural dependence; above and below the sample median for GDP/capita; African and Non-African countries; and countries below and above a score of five on the Polity IV scale. Circles = Economic Rights; Squares = Social Rights. J. Eastin/World Development 107 (2018) 289-305 297 inatory positions that reflect and reinforce labor market stratifica- tion. It is possible that as development progresses further, this rela- tionship might again turn positive, when competitive labor market forces begin to penalize unequal treatment, and social develop- ment ushers in a post-materialist values orientation that further discourages gender discrimination (Eastin & Prakash, 2013). How- ever, the inclusion of a cubic term in these models does not con- firm this effect. Second, these models find modest evidence of a positive relationship between democratic political institutions and women's Social and Economic Rights, which is also in line with expectations and prior research. Next, Agricultural Dependence and appears to exert relatively consistent negative effects on women's rights, but the variable fails to achieve statistical significance in any of the models. The same is true for FDI and Urbanization, though these variables are inconsistently signed. Finally, also in line with expectations, Conflict appears to impose the largest absolute effect on women's rights across models, however the variable is only sig- nificant for five of the eight models. 6. Conclusion This paper argues that income and asset inequality coupled with rising familial burdens due to male out-migration, declining subsistence resource access, and increasing vulnerability to natural disasters diminish women's ability to achieve economic indepen- dence, and enhance their human and social capital relative to men. The consequences include reduced bargaining power in the household, as women become less capable of generating indepen- dent revenue. Outside the home, discriminatory norms and unequal socio-economic status among men and women increase as women are less able to participate in formal labor markets, join civil society organizations, or collectively mobilize for political change. The overall impact magnifies constraints on the advance- ment of laws and norms that promote gender equality. In short, gendered disparities in climate change vulnerability not only reflect preexisting gender inequalities, they also reinforce and strengthen them. Empirical findings are consistent with this argu- ment, indicating that states that experience greater climatic tem- perature variability and increasing incidence of climatological and hydro-meteorological natural disasters are also more likely to experience lower levels of women's social and economic rights. An important implication of these findings might suggest that policies designed to facilitate climate change adaptation also include measures that address the gender inequalities in vulnera- bility, not only as a mechanism to promote public health, but gen- der equality and respect for women's rights as well. For example, a recent report issued through the Canadian International Develop- ment Agency (CIDA) outlines a number of recommendations that policymakers and non-governmental and inter-governmental organizations can take in facilitating gender-specific adaptation, including: conducting research that assesses the usefulness of adaptive technologies acceptable to both women and men; incor- porating gender analysis into projects designed to identify specific climate vulnerabilities; and targeting women's needs in vulnera- bility reduction initiatives alongside those of men. Across all inter- ventions, a vital means of reducing both female vulnerability would be to increasingly incorporate women into decision- making frameworks for climate adaptation and mitigation pro- grams. Such efforts could have the effect of not only reducing cli- mate vulnerability, but also creating another avenue for female empowerment in vulnerable contexts. A key issue for future research should be to evaluate in greater detail the causal mechanisms underpinning the results in this anal- ysis. This study offers a macro-story about the relationship between climate change and gender equality, and provides some of the first systematic cross-country evidence that environmental processes associated with climate change can have an adverse effect on women's social and economic rights. However, as men- tioned, a key challenge with a country/year unit of analysis is cap- turing precise estimates of causal processes. Future empirical work would do well to further probe the conditionality of these findings and address which of the hypothesized mechanisms are most sali- ent and under what circumstances, as they will likely vary across social, political, and economic contexts. While previous work has done an excellent job at evaluating particular mechanisms at the micro-level, there is a dearth of literature that addresses their com- parability across cases