Do women state legislators and administrators make any difference in preventing natural disaster mortality? Evidence from India

Abstract

We investigate the effect of women's political representation in the state legislative assembly and public administration on natural disaster mortality in 20 Indian states from 1981 to 2019. The paper combines two critical dimensions: political and administrative representation of women and disaster risk reduction. Results suggest that women's political representation reduces total disaster mortality after controlling socioeconomic and political covariates; however, the effects are statistically insignificant for the current and lag periods. We find that a one standard deviation increase in women's representation in public administration lowers total disaster mortality by 20.6 percentage points, which is 9.8 per cent of the sample mean. We observe the impacts of female administrative representation on gender-specific human development outcomes through reduced male and female disaster mortality, and we explain some mechanisms. Thus, women's political and administrative representation is crucial for addressing disaster mortality as it has major public health consequences.

1. Introduction

The adverse effects of climate change are felt through the increasing frequency and intensity of extreme weather events, such as floods and cyclones, heat waves, cold waves, lightning, and landslides (Chowdhury et al., 2021). All forms of natural disasters negatively affect both human health and livelihood conditions (Prasad and Francescutti, 2017). In this context, women are more vulnerable than men because their livelihoods depend on natural resources threatened by frequent natural disasters. Nearly two-thirds of the female workforce in developing countries is engaged in climate-sensitive agriculture (UN WomenWatch, 2009). At the same time, women face various social, economic, and political constraints restricting their recovery from natural disasters. This limits the participation of women in designing sustainable adaptation strategies that can mitigate disaster impacts (Social Watch, 2007). There is a consistent gender disparity in crucial human development indicators, such as health, adult literacy, economic participation, and political empowerment in South Asian countries (Rajivan, 2010). On average, in South Asian countries, only 7 per cent of ministerial positions and 15 per cent of political positions in national parliaments are held by women.¹

According to the Global Gender Report 2018, Southern Asia has scored 65.8 per cent in gender parity, the second-lowest score compared to other regions. Countries like Bangladesh, Sri Lanka, and Nepal are the best performers with better female representation and gender parity; however, Bhutan and Pakistan are low-performing nations, and hence, they occupy positions at the bottom of the ranking list on gender parity at the regional and global levels (World Economic Forum, 2018). India's score in gender parity is relatively better than that of Maldives, Bhutan, and Pakistan. Nevertheless, India's ranking of 108 out of 149 countries indicates a gender-based disadvantage in achieving women's empowerment (World Economic Forum, 2018). Furthermore, women remain acutely under-represented in politics in the South Asian region.

South Asia is ranked fourth among the eight regions globally, with a gender parity score of 25.1 per cent in the political empowerment subindex (World Economic Forum, 2023). Table A1 in online appendix A shows the relative position of South Asian countries in the dimension of political empowerment. In none of the countries in the South Asian region is the share of women in Parliament higher than 33 per cent. Regarding the share of women holding ministerial positions in the past 50 years, Bangladesh is the only country in South Asia where women have occupied the highest position in politics for a larger number of years than men. There needs to be more improvement in India's position based on the political empowerment subindex, as the political empowerment score increased from 0.227 in 2006 to 0.382 in 2018, with 18.5 per cent of women holding seats in ministerial positions. Furthermore, compared to the situation in 2019, the share of women in political positions stagnated in about half of 107 countries in 2021, with India registering a steep decline in the share of women ministers from 23.1 per cent to 9.1 per cent (World Economic Forum, 2021).

In India, the percentage of women ministers increased from 2.6 per cent in 1996 to only 17.8 per cent in 2015; women were 9 per cent of all members in the State Legislative Assembly and 5 per cent of all members in the State Council in 2017 (Government of India, 2017). In addition, the share of women in senior administrative positions also remained low. For instance, in 2016, concerning the participation of women in all-India and Central Group-A services, 17 per cent of officers in the Indian Administrative Service were women (Government of India, 2017). The fact that women are often under-represented in legislative bodies and senior-level administration indicates their limited presence in the development of policies of national priorities, including those related to disaster risk reduction and climate change adaptation. One of the crucial aspects of public policy is to encourage both men and women to participate in developmental programs, including environment, conservation, and building resilience to the effects of natural disasters (Padmanaban, 2021). The greater participation of women in public and political spheres can bring change in the mitigation and adaptation measures to reduce the impacts of natural disasters (Lewis, 2016). Gaard (2015) argued that excluding women's agency as decision-makers in climate change issues can explain the increased vulnerability to natural disasters.

In this paper, we focus on an important issue that has not received enough attention in the development literature: the effect of women's political and administrative representation on total deaths from natural disasters in India. We use a panel data analysis with two-way fixed effects at the state level covering a 39-year period from 1981 to 2019 for 20 major Indian states. Our paper is the first to empirically examine the impact of women's representation in politics and senior-level public administration on natural disaster mortality (per 1,000,000 population). Using a panel estimation approach with errors adjusted for serial and spatial correlation, we produce precise estimates of the effects of female representation on disaster mortality. The number of people killed due to disasters is a better proxy of disaster strength because it is considered less arbitrary (Neumayer and Plümper, 2007). Previous studies did not estimate gender-specific mortality rates (Anbarci et al., 2005; Kahn, 2005). Hence, their findings cannot be used to create gender-responsive policies to mitigate disaster fatalities. Instead, our study examines total disaster mortality and shows female versus male disaster mortality using gender-disaggregated disaster-related deaths from 1981 to 2019. We perform a comprehensive study of women's empowerment, including their political representation and senior-level leadership positions in public administration.

We use detailed information on women legislators in the Election Commission of India from 1981 to 2019. We focus on elected women members of State Legislative Assembly (MLAs) in Indian states because the state governments are primarily responsible for meeting socioeconomic expenditures and implementing several development policies. In addition, our analysis includes women's administrative engagement as Indian Administrative Service (IAS) Officers in the state secretariat, where they have the full authority to maintain law and order situations and supervise the implementation of various policies and schemes of the government. The case of India is of particular interest as the representation of women in leadership roles is crucial to building resilience to various types of natural disasters that have become severe due to the changing climate. Our study fills the gap in the literature by examining the link between women's empowerment in politics and higher-level public administration, and natural disaster mortality in India.

The results show that increasing female political and administrative representation is associated with a reduction in total disaster mortality. However, the coefficient on female political participation is statistically insignificant at current and lagged periods, indicating no significant effects of women's political representation on lowering total disaster mortality. One of the primary reasons for such constrained impact could be the persistent underrepresentation of women in the political sphere, as we provide evidence of low political participation of females in India in the study period 1981 to 2019 (see section 2 and online appendix B). We find evidence of the lagged effects of women's administrative representation; a one standard deviation increase in women's representation in senior-level public administration lowers disaster mortality by 20.6 percentage points, which is 9.8 per cent of the sample mean. This indicates that the reduction in disaster-related deaths tends to occur with a lag rather than contemporaneously. Furthermore, results show that a one standard deviation increase in administrative representation of women results in a 24-percentage point (17 per cent of the sample mean) reduction in female disaster-related mortality and a 17.2-percentage point (6 per cent of the sample mean) fall in male disaster-related deaths. As discussed in section 2, the changes in national education policy have encouraged more women to pursue higher education, and those women have possibly joined higher-level public administration later on.

We examine two critical mechanisms through which female political and administrative representation can affect natural disaster mortality. One is related to increasing girls' educational attainment and reducing primary school dropout rates for all children. Another is related to creating employment opportunities for females in manufacturing industries. Educational attainment and an increasing share of females in manufacturing employment are mechanisms that care for women's needs and their families and help reduce their vulnerability to disasters by making them socially and economically independent.

Results suggest that women's political representation positively influences the primary educational enrolment of children by lowering school dropout rates for boys and girls. It particularly encourages girls' educational attainment by increasing their school attendance and reducing their dropout percentage. We find that women's administrative representation improves retention of students by significantly reducing girls' and overall dropout rates at the primary school level. Education is a crucial component that helps improve the disaster adaptive capacity of individuals as it provides the necessary knowledge and skills to make informed decisions regarding how to adapt and sustain livelihoods in the face of extreme weather events (Anderson, 2010). In this context, educating girls and women is one of the most important ways to ensure that families and communities are better equipped to adapt and are less vulnerable to environmental shocks (Blankespoor et al., 2010). Furthermore, results show that increasing female political participation can significantly improve employment opportunities for female workers in manufacturing. Thus, female employment in manufacturing is a critical channel that can reduce the vulnerability of females, who disproportionately suffer the adverse effects of natural disasters (Chowdhury et al., 2022).

Our results suggest that stronger women's representation in politics can lead to higher societal development outcomes by improving school retention among children and alternate employment opportunities. An important contribution of our study is to also examine the role of women in senior-level public administration and how that impacts education and female employment, the two main mechanisms that contribute to disaster preparedness and enhancing resilience to disasters. The estimates show that there is a significant reduction in total disaster mortality with a higher representation of women in public administration. These findings have potential policy significance in terms of enhancing female leadership in disaster management. In addition, the effect of women's political participation leads to a statistically insignificant decrease in total disaster mortality. This is due to the poor representation of women in State Legislative Assemblies, which translates into a weaker impact on disaster risk reduction.

The rest of the paper proceeds as follows. Section 2 describes the background of the study by highlighting the role of women in leadership and disaster risk reduction. Section 3 describes the data and descriptive statistics. Section 4 presents the empirical strategy. In section 5, we discuss the main results. Section 6 presents a discussion of the underlying mechanisms behind the results. Section 7 presents the robustness exercises. Section 8 presents some policy implications, and section 9 concludes the study.

2. Disaster risks and women in state legislation and public administration

Natural disasters generate differential effects on males and females, with females being primarily disadvantaged (Chowdhury et al., 2022). In developing countries, women are relatively more vulnerable to disasters (Sapir, 1993). For instance, women were at higher risk of death compared to men following terrible flood events in 1993 in the Sarhali district of Nepal (Pradhan et al., 2007). During the 1991 cyclone in Bangladesh, women had to bear a disproportionate burden of disaster-related losses, and ninety per cent of the victims were women (Ikeda, 1995). In the Indian context, Chowdhury et al. (2021) showed that women's socially constructed care responsibilities limit their participation in outside activities, including their work migration to the non-agricultural sector for alternate economic opportunities; therefore, women have reduced access to disaster warning information, which can increase their vulnerability to disasters. The study found that an increased allocation of public expenditure for gender budgeting and greater electoral participation of women can help to reduce gender-based vulnerability to floods.

The political representation of women in state legislatures and administrative positions has been seen to positively influence social outcomes, such as greater investments in children's education (Clots-Figueras, 2012), improving reproductive and postnatal health services (Bhalotra and Clots-Figueras, 2014), improved labor market outcomes via increased participation of women in wage employment (Priyanka, 2020), enhancing economic growth (Baskaran et al., 2024), and reducing corruption in societies (Swamy et al., 2001). These studies suggest that women's greater political representation can lead to social sector development. A detailed discussion of women's political and administrative representation in India is presented in online appendix B.

3. Data and descriptive statistics

We estimate the impact of female participation in politics and senior-level administration on natural disaster fatalities using state-level panel data for 20 major states in India in the 39 years from 1981 to 2019. To answer this research question, we combine male and female politicians and Indian administrative service data for India with gender-specific disaster mortality data. Our main variable of interest is women's representation in politics and higher-level public administration. Women's representation is proxied by two indicators: one is the number of seats occupied by women MLAs, expressed as a percentage of all contestants for MLA seats (Women's Political Participation), and the number of women in IAS Officer positions, expressed as a percentage of all women in higher education (Women's Representation in Senior-level Public Administration). A detailed description of variables and their summary statistics are presented in online appendix C.

4. Empirical strategy

We use a panel data analysis to examine the relationship between women's political empowerment and women in senior-level public administration on natural disaster mortality. We estimate the following baseline equation using a two-way fixed effects (TWFE) approach. The data span is 39 years, from 1981 to 2019, for 20 major states of India.

(1)\begin{equation}D{M_{it}}\textrm{ = }\alpha \textrm{ + }\beta \mathop \sum \limits_{s = 0}^2 {({wl} )_{i,t - s}}\textrm{ + }\gamma \mathop \sum \limits_{s = 0}^2 {(ias)_{i.t - s}} + \; \theta z_{i,t}^{\prime} + {\tau _t} + {\varphi _i} + {\varepsilon _{i,t}}\end{equation}

In equation (1), $D{M_{i,t}}$ is the ‘total (male and female) disaster mortality per million (mn) total population’ in the state ‘$i$’ affected by natural disasters (floods, cyclones, and landslides), and ‘$t$’ denotes the time.² The same equation is used with two other outcome variables, ‘female disaster mortality per mn female population’ and ‘male disaster mortality per mn male population’. Disaster mortality disaggregated by gender is a crucial indicator of disaster strength and the extent of damage caused by disasters.

The main independent variables of interest are: (i) $wl$, the measure of women's political participation, which shows the proportion of women legislators out of total contestants in each state at current and lagged periods, which refers to the proportion of women winning in the previous election. It shows the share of seats held by women in State Legislative Assemblies and the competition faced by the women candidates in contesting elections for MLAs. Contesting in elections indicates the highest form of political participation (Verma and Yadav, 1996), and (ii) $\textrm{i}as$, measured as the proportion of women IAS Officers out of the total women in higher education in each state at current and lagged periods.³

In equation (1), $z_{i,t}^{\prime}$ denotes a vector of covariates indicating socioeconomic, political, ecological, and disaster-related variables. We include state-fixed effects $({\varphi _i})$ for time-invariant state-specific effects. We also account for time-specific effects that are common across states by including year dummies $({\tau _t})$. The year-fixed effects account for common economic shocks or climatic shocks. We expect that the coefficients that capture the impacts of interest are $\beta < 0$ and $\gamma < 0$.

The socioeconomic, political, and institutional covariates include the participation of women voting in elections, women chief minister dummy, the occurrence of an election year, per capita state expenditure on health, education, and family welfare, per capita net state domestic product (per capita state income), levels of urbanization, and female-male population ratio. A few ecological control variables are precipitation, dummies indicating the intensities of all disasters, the severity of floods and cyclones, the occurrence of landslides, and the percentage of state area under forest coverage. All control variables are explained in the previous section, and their descriptive statistics are shown in in the online appendix, table A3a. ${\varepsilon _{i,t}}$ is the error term.

We have estimated specification (1) using a TWFE model with clustered and Conley standard errors.⁴ Since we use disaster-related data, serial correlation within states and spatial correlation between states need to be corrected while estimating the standard errors. In this context, we follow Couttenier and Soubeyran (2014) to tackle these two problems simultaneously in our regressions. We report two sets of standard errors for each coefficient in the analysis. One regression analysis with Conley standard errors allows for correcting serial and spatial correlation within a radius of 100 km following the procedure developed by Conley (1999). The threshold of 100 km implies that the error associated with each state is assumed to be correlated with the states that are situated within the radius of 100 km from it and that there is no spatial dependence of errors beyond 100 km. Another analysis with the clustered standard error allows for clustering at the state level and, therefore, corrects for serial correlation within a particular state (Abadie et al., 2023).

Since natural disasters may be spatially correlated, some of the explanatory variables may not be observed. In such cases, it is essential to account for dependence in the error term across observations and obtain standard errors that are consistently estimated (Driscoll and Kraay, 1998). The clustering of standard errors does not allow for spatial correlation. Therefore, we use the Conley (1999) approach that relies on the generalized method of moments technique and computes the variance-covariance matrix with spatial weights. Thus, our estimates address both spatial and serial correlation (Conley, 1999, 2008). This helps to produce spatially corrected and heteroskedastic consistent standard errors (Colella et al., 2020).

One concern in the estimation strategy could be the presence of unobserved factors that can affect both disaster mortality and women's representation in politics and higher administration. We control the risk of omitted variable bias by including several covariates, using state-fixed effects and time-fixed effects in linear regressions. The TWFE estimator includes a unit-specific time average and time-period-specific cross-sectional average, leading to double demeaning characterization for estimated coefficients (Wooldridge, 2021). Including fixed effects, the TWFE estimator helps control for the average differences across states in unobserved factors that can affect both disaster mortality and women's representation.

To examine the shape of the relationship between women's representation and disaster-related deaths in India, we plot disaster mortality against female political and administrative participation. A nonparametric method is more appropriate in this context because such a technique does not impose any prior specification on the functional form to study the relationship between disaster mortality and women in politics and public administration. We use the locally weighted scatterplot smoothing (lowess) methodology to show the plot between the variables under analysis in this study, as Echávarri and Ezcurra (2010) suggested. Figure A5 (online appendix E) shows the relationship between female representation in politics (and senior-level administration) and disaster mortality using the smoothing method. We find that the nonlinear relationship between disaster mortality and female administrative representation holds in our analysis only for the later period of our study, post-1999 (figure A5). Therefore, the nonlinear analysis is restricted to the period from 2000 to 2019.

In the subsample, ranging from 2000 to 2009, deaths from disasters show an inverted U-shaped pattern concerning the administrative representation of females. This implies that when female participation in administration was low at the initial stage, disaster mortality increased until a certain threshold point, beyond which mortality started to decline once the critical threshold level of women's administrative representation was achieved. We do not find any strong support to show the existence of a nonlinear relationship between female political participation and disaster mortality in the data from 2000 to 2019 (figure A5).

It is possible that the number of disaster-related deaths might depend on women's empowerment and additional explanatory variables that act as controls in the study. Thus, in addition to the linear model, we estimate the quadratic specification in women's administrative representation, as in Ferreira et al. (2013), where the authors estimated a specification that is quadratic in governance to show the effects of governance on fatalities due to floods. Furthermore, in the context of India, Parida et al. (2022) investigated an inverse U-shaped relationship between flood fatalities and government responsiveness. A few additional studies have also examined the nonlinear relationship between GDP per capita and disaster damages, and population killed, using fixed effect models (Kellenberg and Mobarak, 2008; Raschky, 2008; Schumacher and Strobl, 2011; Ferreira et al., 2013; Parida et al., 2022).

In light of the previous studies and considering that figure A5 reveals a quadratic relationship between females in administrative positions and natural disaster mortality, we estimate the following equation using a TWFE estimation approach. We use the same control variables as those used in the linear regression model, which are described in section 3.

(2)\begin{align}D{M_{i,t}}\textrm{ = }\alpha + {\beta _1}{({wl} )_{i,t - 1}}\textrm{ + }{\beta _2}({wl} )_{i,t - 1}^2\textrm{ + }{\gamma _1}{({ias} )_{i,t - 1}}\textrm{ + }{\gamma _2}({ias} )_{i,t - 1}^2 + \theta z_{i,t}^{\prime} \textrm{ + }{\tau _t}\textrm{ + }{\varphi _i}\textrm{ + }{\omega _{i,t}}\end{align}

Equation (2) shows the quadratic relationship; $ias_{it - 1}^2$ denotes the squared term that captures the nonlinearities. ${\gamma _1}\mathrm{\ > }\ 0$ and ${\gamma _2}\mathrm{\ < }\ 0$ reveal an inverted link between the two variables, which suggests an inverted U-shaped relationship between disaster mortality and $ia{s_{it - 1}}$. $wl_{it - 1}^2$ denotes the squared term and ${\beta _1}\mathrm{\ > }\ 0\; \,\textrm{and}\,{\beta _2}\mathrm{\ < }\ 0$ suggest an inverted U-shaped relationship between disaster mortality and $w{l_{it - 1}}$. ${\tau _t}$ denotes the year-fixed effects and ${\varphi _i}$ indicates the state-fixed effects. ${\omega _{i,t}}$ is the error term.

We did not find any issue of reverse causality from the outcome variable, ‘total number of people killed due to natural disasters per mn population.’ None of the cumulative disaster-related deaths (per mn population) in previous periods show any statistically significant relationship with women legislators and administrators in the current period (see online appendix table A2, Parts I and II).

5. Results and discussion

Using the linear model, we first investigate the effects of women's political and administrative representation on natural disaster mortality. The outcome variable is the total disaster mortality per mn total population. Table 1 presents the results from estimating equation (1) for the model with Conley standard errors (that account for serial and spatial correlation). The model includes a variety of socioeconomic, political, and exogenous disaster-specific controls.

Table 1. Two-way FE estimation: women's political and administrative representation and natural disaster mortality

Note: Conley standard errors are reported in parentheses. Natural disaster mortality includes deaths due to floods, cyclones, and landslides. In all models, we control urbanization and the female-male population ratio.

Results show that the coefficient of female political representation is negative, suggesting that a greater presence of women in politics helps reduce total deaths due to disasters. However, the coefficients with female political participation are statistically insignificant at the current and two lag periods. The constrained effects could result from a persistently low representation of women as political leaders in State Legislative Assemblies, as discussed in section 2 and online appendix B.

The effect of women's administrative representation in the current period shows a reduction in total disaster-related deaths, although the coefficient is not statistically significant. We use the first- and second-period lags in the analysis to examine the lagged effects of women's representation in senior-level administration on total disaster mortality. In the first lag period, the effect remains insignificant.

With the inclusion of the second lag period, results show that an increase in women's participation in public administration by one standard deviation results in a 20.6 percentage point fall in total disaster mortality (column 1).⁵ This is 9.8 per cent of the mean rate in our sample.⁶ We find evidence of the lagged effect of women's administrative representation on people killed as a fraction of the 1,000,000 population; the coefficient on women in IAS in years preceding the current year (year (t–2)) is negative and statistically significant. This suggests that women's administrative power and decision-making have been realized over time, leading to better human development outcomes with a decline in disaster mortality.

As discussed in section 2, the changes in national education policy might have played a major role in encouraging women to pursue more education and enter public administration in the later years (see online appendix B for a detailed discussion). This is also evident in figure A4, which shows that the share of women administrators increased particularly after 1997 (see online appendix B). Furthermore, IAS officers are professionally trained to ensure disaster management and effectively respond to save people by conducting rescue operations, evacuation, and relief measures for disaster-affected people (Government of India, 2004).⁷

We find that a one standard deviation rise in women's administrative representation (lagged two periods) leads to a 17.17 percentage point reduction in male disaster mortality, which is 6.1 per cent of the mean rate in the sample (see table 1, column C2). Similarly, with a one standard deviation increase in the administrative representation of women (in years prior to the current year), we find that female disaster mortality declines by 24.0 percentage points; this is 17.1 per cent of the mean rate in the sample (column C3). The impact of administrative representation increases over time as more educated women participate in administrative services. Table 2 describes the estimated coefficients with total and gender-disaggregated disaster mortality outcomes using clustered standard errors. The coefficients are directionally consistent with those produced using Conley standard errors, and the magnitudes are almost identical.

Table 2. Two-way FE estimation: women's political and administrative representation and natural disaster mortality

Notes: Clustered standard errors at the state level are reported in parentheses. Natural disaster mortality includes deaths due to floods, cyclones, and landslides. In all models, we control urbanization and the female-male population ratio.

Results show no statistically significant reduction in disaster mortality due to a higher share of women legislators. However, we find that administrative representation of women as more highly educated women enter public offices as administrators is associated with a statistically significant reduction in disaster mortality. One possibility is that a few women in India attained political positions in state legislatures through male political patronage and family associations, so female decision-making positions remain low (Chary, 2012), and they are unable to meaningfully use political power to influence policy outcomes in a male-dominated political territory (Kaushik, 1993). This can explain why women's political representation has not been adequate and why a greater political representation of women is still needed.

Major improvements in women's representation started after 1996 with a concentrated effort to implement policy initiatives to attain women's empowerment, as outlined in section 2. The political system has been evolving, and people's outlook is also experiencing positive changes, as evidenced by the recent data on female political and administrative participation after the mid-1990s and 2000s (see online appendix figures A2 and A4); we expect some nonlinear effects post-1999. The change in societal outlook and, therefore, greater acceptance of women candidates in politics and public administration can result in better human development outcomes through reduced deaths from natural disasters. Figure A5 shows the graphical illustration of the nonlinear relationship between disaster mortality and female political and administrative participation over the period 2000 to 2019 (see online appendix E). We find an inverted U-shaped relationship between women's administrative representation and natural disaster mortality from 2000 to 2019.

We find no evidence of such an inverted U-shaped link from 1981 to 1999 (see online appendix figure A6). Thus, in addition to the linear model, we use a quadratic specification to estimate the effects of women's representation on total and gender-specific deaths from natural disasters during the 2000–2019 period while controlling all socioeconomic, political, and institutional factors. The coefficient on female political representation is positive, and its squared coefficient is negative but not statistically significant. Although our estimated results suggest that the presence of women in politics is associated with lowering male and female deaths from natural disasters, the effects are insignificant (see table 3, column C1).

Table 3. Two-way FE estimation: nonlinear relationship between women's political and administrative representation and natural disaster mortality

Notes: Conley standard errors are reported in parentheses. Natural disaster mortality includes deaths due to floods, cyclones, and landslides. In all models, we control urbanization and the female-male population ratio. The mean of Women elected as Members of the Legislative Assembly (as % of total contestants) is equal to 0.86, and the mean of Women IAS Officers per mn women in higher education is equal to 6.9.

We find no strong evidence of a significant effect of women's political representation on disaster mortality. A critical mass of women in politics is needed to find a stronger and more effective political participation of women and their influence on achieving a significant reduction in natural disaster mortality. Further analysis is done to study the impact of the share of women in public administration on total disaster mortality per mn total population (see table 3, column C1). The coefficient associated with the linear term of women's administrative representation is positive and significant, and its quadratic term is negative and significant, showing the presence of an inverted U-shaped relationship between total disaster mortality and women's representation in public administration. Results suggest that in the initial phase, with a rise in representation of women in senior-level administration, deaths due to disasters tend at first to increase until a threshold level is reached.⁸ This relationship does not continue for an indefinite period.

Once women's administrative representation reaches a threshold level, we find that total deaths from disasters start to decline as a larger share of highly educated women enter public administration over time. Due to the gendered dimensions of disaster risk and the fact that disasters affect men and women differently, we divide the outcome variable into female and male natural disaster mortality per mn female and male populations, respectively. We find the presence of the inverted U-shaped link between male deaths due to disasters and the proportion of female administrators (column C2). The negative and significant coefficient on the squared term of women administrators indicates that male disaster mortality declines rapidly with an increased share of women in public administration. The signs on linear and squared terms of women IAS Officers show the existence of a nonlinear (inverted U-shaped) relationship between female disaster mortality and women in public administration (see table 3, column C3).

To sum up, the quadratic effect of women administrators on total disaster mortality is statistically significant, implying that women's administrative representation helps minimize disaster impacts by lowering male and female disaster mortality. In addition, we add the women chief minister dummy as a control variable to account for the presence of women in the position of Chief Minister in the state. The coefficient is negative and significant, implying that the presence of women in ministerial positions helps reduce disaster mortality.

Our findings show that increased women's administrative representation could lead to social welfare gains. Thus, if more women participate in higher-level administration and politics, the goal of disaster risk reduction and, eventually, zero disaster mortality will be more likely to be achieved.

6. Possible mechanisms

This section investigates why having more women in State Legislative Assemblies and public administration can reduce disaster mortality. In a broader sense, women's empowerment can lead to favorable policy support for men and women since our results also show a decline in total disaster mortality with a greater women's representation in state legislative assembly and public administration (see tables 1 and 2). The first welfare mechanism behind this result is analyzed through the effect of women's political and administrative representation on enhancing educational opportunities by increasing girls' enrolment in primary education as a percentage of total primary enrolment and reducing overall dropout rates in primary schools. The second underlying mechanism is studied by analyzing the effects of women's representation on enhancing manufacturing employment share for females. Therefore, we focus on two main mechanisms, education and employment opportunities, through which female political and administrative participation can help reduce vulnerability to natural disasters. We use the Hausman-Taylor (HT) (1981) strategy to estimate these effects as the HT method addresses the endogeneity problem. A detailed discussion of the HT model and the estimated equation is presented in online appendix D. In addition to the estimated results from the HT model, the results from the TWFE model are also presented in online appendix table A4. The two underlying mechanisms are also explained in detail in online appendix D.

7. Robustness exercises

A few robustness checks are carried out to validate our main results. First, we include information on mortality caused by other types of natural disasters: heat waves, cold waves, and lightning.⁹ Table 4 presents the results of regression analysis where the outcome variable is the number of total (female and male) deaths from natural disasters (cyclones, floods, landslides, heat waves, cold waves, and lightning) per mn total population.

Table 4. Robustness analysis: two-way FE estimation, women's political and administrative representation and natural disaster mortality

Notes: Conley standard errors are reported in parentheses. Natural disaster mortality includes deaths from floods, cyclones, landslides, heat waves, cold waves, and lightning. In all models, we control urbanization and the female-male population ratio.

We find a significant lagged effect of administrative representation of women on total disaster mortality. In column C1 of table 4, the coefficient on the female administrative representation in years preceding the current year (that is, year (t–2)) is statistically significant, indicating the largest effects from female representation tend to take place with lags. We find that a one standard deviation rise in women's administrative representation results in a 20.6 percentage point reduction in total disaster mortality. This is 3.6 per cent of the mean rate in the sample.¹⁰ These findings are consistent with the main results discussed in section 5 (see tables 1 and 2).

We also report the results of the estimated impact of women's representation on gender-specific disaster mortality rates (see table 4, columns C2 and C3). With greater representation of women in public administration, female disaster-related mortality declines (column C3). Male disaster-related deaths also decrease as more women enter public administration. The effects are significant only with lagged periods; an increasing impact of female empowerment is observed over time. The coefficient on women's political representation is insignificant for current and lagged periods (column C1). Our results hold with both clustering at the state level and Conley standard errors accounting for serial and spatial correlations (see table A5 in online appendix F). The impact of political and administrative participation of women on disaster mortality is robust to changing the outcome to include data on deaths from all six types of disasters.

We perform another robustness check by comparing India's high and low human development states (see table A7 in online appendix F). Following the classification of states based on the human development index (HDI) ranking, as suggested by Mukherjee et al. (2016), we segregate data into two groups of states: one representing high levels of human development and another with low achievements in human development. This helps us account for regional differences in human development and study the effects of women's political and administrative participation on disaster mortality for both states. We further study the effects of female political and administrative participation on gender-specific natural disaster mortality in high and low human development states. The outcome variable in both sample groups is disaster mortality (per 1,000,000 population). The sample is already quite small, and further disaggregation into male and female disaster mortality in two groups of states makes the results statistically insignificant.

In the case of high development states, results suggest that female political participation significantly reduces deaths from natural disasters. We find that a one standard deviation rise in women's political representation results in declining total disaster mortality by 65 percentage points (see online appendix table A7). This shows a 24.3 per cent drop in total disaster mortality relative to the sample mean. Furthermore, a one standard deviation increase in female administrative representation leads to a 4.27 percentage point fall in total disaster-related deaths, which is a 1.61 per cent decline relative to the sample mean. A more detailed analysis of the effects of female political and administrative representation on disaster mortality for high and low human development states is provided in the online appendix F, table A7.

8. Policy implications

Most previous studies have mainly focused on policy efforts to encourage greater participation of women in politics and female voter turnout in elections (Iyer and Mani, 2019). It is argued that female political empowerment can affect several development outcomes that help improve the overall well-being of people. Our study focuses on women's political and administrative participation in India. We provide estimates of the overall impact of women's empowerment on total natural disaster mortality and gender-disaggregated deaths from disasters.

We find that women's political and administrative empowerment is negatively associated with total disaster mortality and gender-specific mortality (see tables 1 and 2). This implies that improving the proportion of women legislators and administrators lowers disaster-related mortality. We do not find a significant effect on disaster-related deaths, though the coefficients have the correct sign; they are negative. The lagged effects of administrative representation of women are significant, suggesting that increased numbers of women in public administration at one point in time impact disaster mortality later in time. This is due to the persistent improvements in public policies that have led to the concentrated emphasis on increasing women's representation in government (see section 2 and online appendix B).

Furthermore, an important issue is how the improved share of females in government can result in a higher share of public spending on favorable policies for women. One such important policy response is the gender budgeting approach adopted in 2000 and institutionalized in 2005 with commitments to gender equality by making suitable budgetary allocations for women-oriented public interventions (Nair and Moolakkattu, 2018).¹¹ Gender budgeting is supposed to play a critical role in abolishing gender inequalities while improving the nation's social, educational, and economic indicators (Rudra, 2018).

Figures A7 and A8 (online appendix G) show the relationship between the participation of women in politics (and senior-level administration) and the budgeting allocation for women in the union budget statement from 2005 to 2017. Though the budgetary allocation for women-oriented programs increased in the initial years from 2008 to 2009, it declined after 2010; it remained almost constant over a certain period until 2014, after which the proportion of gender budgetary allocation out of total budgetary expenditure showed some improvements. The plausible reason is that the approach to women's empowerment in successive five-year plans downgraded women to the status of beneficiaries of the development programs, disregarded women as active agents, and failed to provide women with sufficient access to available resources (Mitra, 2019). While the gap between male and female electoral participation has decreased with increased female voter turnout, women's representation in state legislatures and government bureaucracies is much lower (Rudra, 2018).¹² Such political disadvantages to women translate into poor performance in achieving favorable policy outcomes for women.

India still needs to achieve the desired threshold level of women's participation in politics and senior-level administration to impact total disaster-related deaths significantly. We refer to this desired level as the threshold point of the involvement of women, a point beyond which an increase in the share of women in politics and higher administration can lead to a rapid decline in rates of natural disaster deaths. Our results underscore the potential for enhancing women's representation in government to reduce disaster mortality rates. The two key mechanisms through which disaster mortality can be reduced are the role of education and the increased share of women in manufacturing employment (see section 6 and online appendix D).

We find that a rise in women's political participation has not only led to improvements in girls' education but has also helped enhance the educational attainment of all children by reducing total school dropout rates (see online appendix table A4). As already highlighted, education is one of the critical mechanisms that help develop general skills that result in understanding disaster warning information and undertaking disaster preparedness measures that help reduce vulnerability to disasters (see online appendix D). Furthermore, we find that increasing female political participation can help increase women's employment share in the manufacturing sector, thus promoting alternate economic opportunities for women and less dependence on weather-sensitive agriculture. Women's financial strength improves their families' well-being and creates a more sustainable society. Financially independent individuals are more financially resilient, which enhances their resilience against the adverse effects of environmental shocks (Hussain et al., 2021).

Studies have shown that the political participation of women in India has resulted in a fall in infant mortality rates (Kumar and Prakash, 2017) and a reduction in maternal mortality rates (Bhalotra et al., 2019) and has ensured primary education for children (Clots-Figueras, 2012). However, studies have found no significant impact of gender quota in politics on tuberculosis mortality, a highly prevalent disease in low-income countries, and concluded that women's political participation is more effective in targeting women's health than targeting the general public's health (Bhalotra et al., 2019). Our paper has looked at the total mortality due to natural disasters as it has major public health consequences. Disasters directly impact the population's health by causing deaths and severe injuries (Shoaf and Rotiman, 2000). Although we find a fall in total disaster mortality with female political participation, the impact of female representation is insignificant at current and lagged periods (see tables 1, 2 and 4). There has been no large decline in total and gender-specific disaster mortality, with a sluggish improvement in the number of women legislators in state legislative assemblies of India. This creates a strong case for improving female participation in political and public spheres because better representation of women will certainly increase their influence on various policy outcomes that better address female needs (Clots-Figueras, 2011). Women have not obtained the political resources needed to ensure higher participation in the political sphere, leading to a persistent gender gap in political representation in India (Artiz Prillaman, 2019).

9. Conclusions

This paper is the first to provide empirical evidence of the effects of women's political representation in state legislative assembly and public administration on total and gender-specific natural disaster mortality in the context of a developing country. Using an extended panel of 39 years consisting of 20 major Indian states from 1981 to 2019, we estimate how women in legislation and senior-level administration minimize disaster-related deaths. The key natural disasters studied in this paper are floods, cyclones, landslides, heat waves, cold waves, and lightning. While many studies look at the substantive consequences of women in influential positions, our paper looks at a novel outcome and focuses on extreme weather events with potentially fatal consequences.

Results suggest that women's political empowerment affects disaster mortality by reducing male and female deaths due to disasters after controlling socioeconomic and political covariates; the effects are statistically insignificant for the current and lag periods. Furthermore, women's representation in public administration also helps lower the negative impacts of natural disasters by minimizing disaster mortality; the effects on disaster mortality are significant with lag periods. Therefore, we observe increasing impacts over time as more educated women enter public offices.

In terms of underlying mechanisms, we studied whether women legislators and administrators can empower females by creating jobs in the manufacturing sector, as access to employment in industries can make women financially independent and reduce their dependence on climate-sensitive agriculture. Employment opportunities help reduce women's vulnerability by making individuals economically independent and financially resilient to climate shocks (Hussain et al., 2021).

We investigated the effect of female representation on overall school attendance by reducing total primary school dropout rates for all children, encouraging girls' enrolment in primary education, and discouraging girls' dropout rates. We find a statistically significant improvement in girls' enrolment in primary education and a significant reduction in school dropout rates with a rise in female political participation. In addition, women leaders enhance boys' and girls' primary education, as evident in the significant association between women's administrative participation and reduced primary school dropout rates of children. School education plays a critical role in raising awareness regarding climate and disaster risk issues that primarily contribute to building disaster resilience (Anderson, 2010). Improving girls' education and access to alternate employment opportunities for women produces a multiplier effect on societies and leads to sustainable development (Stevens, 2010).

We have further analyzed the high and low human development Indian states ranked based on their HDI scores. When we divide the entire sample into high-HDI and low-HDI states, the effect of female political representation is only significant for the states that rank higher in human development. This suggests that the weaker involvement of women in political decision-making leads to poor human development outcomes, in this case, no significant reduction in disaster mortality in low-HDI states.

Results show that women's representation in public administration helps lower the negative impacts of disasters by significantly minimizing disaster mortality in high-HDI and low-HDI states. The plausible reason is that the Government of India has initiated an integrated approach to disaster management, in which the IAS Officers play a critical role in ensuring adequate disaster preparedness and coordinating action for disaster relief and rehabilitation measures (Government of India, 2004). Thus, females in administrative positions play a more decisive role in lowering deaths due to disasters. It is, therefore, essential to enhance women's leadership in designing effective responses to mitigate the effects of natural disasters.

The greater participation of women in senior-level public administration and women being elected to political decision-making positions will strengthen their bargaining roles at the household level, in the community, and the broader economy. Women's improved bargaining strength can lead to a more significant investment of resources to promote collaboration at all levels of disaster risk governance and increase resilience to natural disasters.