1 Introduction

Catastrophic events become disasters when they disrupt human life and cause high human, social and economic costs in terms of loss of life, exacerbation of pre-existing inequalities and destruction of infrastructure, among other consequences. In this sense, recent studies focus on analysing the social aspects of disasters taking into account the role played by gender in daily life and the impact this has on the population. The first works looking at gender and disasters began to appear in the 1990s and highlighted the fact that the ways in which men and women were affected by such phenomena were as much related to gender (Anderson 1994; Enarson and Morrow 1998; Enarson and Scanlon 1999; Fothergill 1999) as to socioeconomic status (Kumar-Range 2001). From then on, numerous studies have dealt with demonstrating the differential impact of these phenomena on the health and mortality of men and women (Oxfam International 2005; Akerkar 2007; Llorente-Marrón et al. 2021) and increased violence against women (Enarson 1999; Fisher 2010; Seager 2014), as well as the influence of gender roles during and in the aftermath of a disaster (Fothergill 1999; Cocina Díaz et al. 2022; Dema Moreno et al. 2022).

Regarding the socio-economic impact of disasters of natural origin. Authors such as Moser (1993), Enarson et al. (2006) and Bradshaw and Fordham (2013) have highlighted that the intersection of gender and socioeconomic level conditions the impact and the lived experience of the disaster since women usually have access to fewer economic resources than men. This gender difference is, in turn, the result of the sexual division of labour. This reflects, on the one hand, the fact that the reproductive roles that women usually undertake are unpaid and, on the other, that in productive terms, women’s salaries are lower than those of men (Moser 1993; Enarson et al. 2006; Bradshaw and Fordham 2013). At the same time, the literature indicates that women’s limited access to economic resources prior to a disaster makes it more difficult for them to recover from its effects in the aftermath (Fordham and Ketteridge 1998; Fothergill et al. 1999; Austin and McKinney 2016). In this sense, studies such as those conducted by Dufka (1988) and Arenas Ferriz (2001) revealed that the development of recovery plans does not generally take into account the informal economy, this being the sphere within which many women habitually work. An example of this is the fact that after the earthquakes in Mexico in 1985 and El Salvador in 2001, women at the lower end of the socioeconomic scale had to deal not only with the destruction of their homes, but also the loss of their livelihood: for many of them, their home was not only the place where they lived, but also their work space as it was where they carried out informal work, principally related to caring for dependents, preparing food and washing clothes (Dufka 1988; Arenas Ferriz 2001). More recently, Llorente-Marrón et al. (2020a) analysed the socioeconomic impact of the 2010 earthquake in Haiti on gender relations, observing that gender inequality not only persisted, but also widened after the disaster.

Another factor which merits attention is whether women live in rural or urban areas. To this end, numerous studies have demonstrated that the vulnerability of women increases after a disaster, especially in rural areas (Gray 1993; Afriyie et al. 2017; De Silva and Kawasaki 2018; Hamidazada et al. 2019; Lebni et al. 2020; Llorente-Marrón et al. 2020b; Lin et al. 2021). Women’s limited access to education, as well as the low average household income and gender roles which, in certain contexts, prohibit women from communicating with men outside their family all increase the impact of disasters of natural origin on women living in rural areas (Wakefield 2005; Hamidazada et al. 2019; Lin et al. 2021). In the same way, the poor coordination of disaster risk reduction agencies and the fact that, after the disaster, the health needs of rural women are often neglected, limit their recovery and their long-term health (Hamidazada et al. 2019; Lebni et al. 2020).

Finally, it is worth pointing out that most research analysing the impact of disasters on the living conditions of women focus on a single territory and/or a single disaster of natural origin, and do not engage in any comparative analysis of the socioeconomic impact of the same disaster in rural versus urban areas. The use of a comparative perspective in research allows for a more comprehensive approach to social reality than the analysis of a single area and/or single disaster event. Moreover, numerous research studies have shown that the differential impact of disasters on men and women is due to gender inequalities that exist before the event (Austin and McKinney 2016; Aheeyar et al. 2019; Petraroli and Baars 2022; Ngu et al. 2023). In this sense, conducting comparative studies allows us to demonstrate that this differential impact is not something incidental that only occurs in one place or disaster, but something that is the result of structural gender inequalities and the extent to which it is similar or not in different areas and/or types of disaster.

For this reason, the current study attempts to address this gap by analysing and comparing the effects that the earthquakes which hit the Dominican Republic in 2003, Honduras in 2007 and 2009 and Haiti in 2010 had on the gender relations in urban compared to rural areas. In this way we attempt to identify whether, despite the specific characteristics particular to each of the countries selected, there were similarities in the way the three countries experienced the respective earthquakes in terms of the impact on gender relations, and whether there are differences in how urban and rural areas experience the impact of such disasters of natural origin. Comparative analysis can also identify patterns and resilience factors that are specific to the region. This allows governments and organisations to develop more informed policies and direct resources more efficiently towards risk prevention and management. The selection of the Caribbean and, specifically, the Dominican Republic, Honduras and Haiti as the research area relates to the fact that most research into the analysis of disasters of natural origin are carried out on data from Asia. In this article, therefore, we aim to shed light on a territory –the Caribbean–which is much less explored. The study of the socioeconomic impacts of disasters of natural origin in Latin America and the Caribbean is crucial for the development of effective adaptation and mitigation strategies and can be used to educate and raise public awareness of the importance of disaster preparedness.

2 Study area

The region of the Caribbean is situated to the east of Central America and north of South America and comprises over 700 islands and islets and a total of 24 countries –including the Dominican Republic, Honduras and Haiti– which all have coastlines bathed by the Caribbean Sea. A large number of these territories are situated on the Caribbean Plate which collides with the North American Plate, the South American Plate, the Nazca Plate and the Cocos Plate. These collisions translate not only into the frequent seismic activity that the region experiences, but also, occasionally, tsunamis (LACGEO 2023).

The first of the earthquakes selected for study in the current work occurred on 22 September 2003 in the Dominican Republic. Its epicentre was located 3 km from Puerto Plata and it measured 7 on the MercalliFootnote 1 scale (Fig. 1). The earthquake resulted in 3 deaths, 15 people being injured and the destruction of hundreds of buildings (USGS 2023a).

Fig. 1
figure 1

Intensity of the earthquake in the Dominican Republic 2003. Source USGS (2023a)

The second and third earthquake studied here both occurred in Honduras. The first was on 15 September 2007 and had its epicentre in the province of Yoro (Fig. 2). It measured 5 on the Mercalli scale and resulted in the deaths of 18 people, with a further 1,800 being affected to varying degrees (USGS 2023c). Almost two years later, on 28 May 2009, another earthquake hit the country. On this occasion the epicentre was in the Caribbean Sea and the event measured 8 on the Mercalli scale (Fig. 3) and caused the deaths of 7 people, injured 40 others and affected more than 130 buildings (USGS 2023d).

Fig. 2
figure 2

Intensity of the earthquake in Honduras 2007. Source USGS (2023c)

Fig. 3
figure 3

Intensity of the earthquake in Honduras 2009. Source USGS (2023d)

The fourth and final earthquake studied in this work took place on 12 January 2010 in Haiti, its epicentre being 25 km from Port-au-Prince (Fig. 4). It measured 9 on the Mercalli scale and was one of the most destructive natural events to hit the country in recent decades, with a death toll of 316,000 people, 300,000 others injured, 1.3 million displaced and 97,294 houses destroyed (USGS 2023e).

Fig. 4
figure 4

Intensity of the earthquake in Haiti 2010. Source USGS (2023e)

While the intensity of these events obviously has plays its part in terms of their impact, the literature also highlights the fact that the socioeconomic situation as well as the gender inequality present in each of the countries involved must also be taken into account in order to more fully comprehend the repercussions of each earthquake on the affected population (Anderson 1994; Enarson and Morrow 1998; Enarson and Scanlon 1999; Fothergill 1999; Kumar-Range 2001). In this sense, Haiti is the country in Latin America and the Caribbean with the lowest Human Development Index (HDI) and, according to the Gender Inequality Index (GII), it also has the greatest gender inequality (UNDP 2023). The latest data from the HDI indicates that, in 2021, Haiti achieved a rating of 0.535, its historic lowest rating being in 2010, probably as a result of the earthquake (UNDP 2023). The HDI of Honduras in 2021 was 0.621, and in 2007 and 2009 –when the two earthquakes under analysis here occurred– the country’s rating was, respectively, 0.568 and 0.594 (UNDP 2023). Meanwhile, the Dominican Republic achieved an HDI rating of 0.767 in 2021 –higher than the world average of 0.732– although in 2003 –when the earthquake occurred– it was 0.660 (UNDP 2023). In terms of their level of gender inequality, the Dominican Republic and Honduras are both on a par with the world average, occupying positions 80 and 137, respectively in the global/world ranking of GII (UNDP 2023).

Furthermore, several studies suggest that the educational level of the population conditions not only the impact of the disaster, but also the way of coping with it (Afriyie et al. 2017; Petraroli and Baars 2022). In this regard, it should be noted that, according to HDI and GII data (UNDP 2023), literacy levels for men and for women in the three countries at the time the respective earthquakes occurred differed. In the case of the Dominican Republic, the female literacy rate in 2002 –the closest date to the earthquake for which there is data– was 87.18%, while it was 86.81% for men. In Honduras, at the time of the 2007 earthquake, literacy levels were 83.45% for women and 83.75% for men, while data for 2010 –that closest to the 2009 earthquake– had risen to 84.73% for women and 84.79% for men. As for Haiti, the data for 2006 –closest date to the 2010 earthquake– shows a literacy rate for women of 44.60% and for men of 63.09%. If we consider the most recent data available –for 2016– it is clear that these differences between countries have been maintained: in the Dominican Republic, women’s literacy was 93.7%; 88.93% in Honduras and 58.30% in Haiti, with the corresponding male literacy rate being 93.79%, 89.05% and 65.28%.

3 Methodology

To carry out the present study we used as our data source the Demographic Health Survey (DHS) developed by the United States Agency for International Development (USAID). The technique selected to measure the impact of the earthquakes in the Dominican Republic (2003), Honduras (2007 and 2009) and Haiti (2010) on gender relations was the Differences in Differences (DID) technique. This method compares the differences that are produced over time between a population affected by a specific event –that is to say, in this case, provinces that were impacted by the earthquake– and another population that was not affected, which acts as a control group (Gertler et al. 2017).

Since our objective here is to analyse the effects of the earthquakes selected on gender relations and, specifically, to ascertain whether the conditions of female-led households improved or became worse after the event, the differences that will be analysed are those that are produced in the “Wealth Factor” (WF). The WF corresponds to a proxy developed in collaboration with the World Bank which attempts to approximate the assets of a family, taking into account not only their income, but also their access to services such as electricity and water, among others, and whether or not the household has a bathroom and/or domestic appliances (Filmer and Pritchett 2001). To analyse the changes in the WF we firstly calculated the differences in this variable between a group of people that were affected by the respective event in each country, and a group who were not. We then compared the differences between the two moments in time, that is, before and after the earthquake (Fig. 5).

Fig. 5
figure 5

Differences in Differences estimator (DID). Source Produced by the authors based on data from Gertler et al. 2017

Table 1 (below) shows the year of the WF survey data prior to and after the earthquakes studied in each country. In the analysis we took into account only those households where there was a female head of household or where a male head of household had a female partnerFootnote 2. In this way we were able to ensure that we only considered one questionnaire per household and avoided replication by considering, for example, daughters, nieces, sisters or grandmothers. Comparing data from the surveys conducted prior to and following the earthquake allowed us to construct the time variable t. As such, the women surveyed in the pre-disaster waves constituted the control group (t = 0), whilst survey data relating to the years after the disaster comprised the postdisaster group (t = 1).

Table 1 Pre- and postdisaster surveys in the countries analysed

Table 2 shows all the provinces affected (R = 1) for each of the earthquakes analysed. To determine which areas were impacted, we used information provided in the International Disaster Database (EM-DAT) developed by the Centre for Research on the Epidemiology of Disasters.

Table 2 Provinces affected by the earthquake analysed in each country studied

As the aim of the present study is to quantitively analyse the effects of earthquakes on gender relations, we considered the dichotomous variable “sex of head of household” (SH), which was given a value of 1 if the head of household was male and 0 if it was a woman. Alongside this we took into account other variables that measured characteristics prior to the earthquake and which, given that they can vary over time, might explain any variations found in WF. Firstly, we considered the educational level, treating it as two dichotomous variables, namely, “Secondary education” whereby a woman was given a value of 1 if she had completed secondary education and of 0 if she had not, and “Higher education” where a value of 1 indicated that a woman had a higher education qualification and a value of 0 that she did not. In addition, we consider as covariates the age and the squared age of the head of household since together they indicate the effect of life cycle on wealth (Llorente-Marrón et al. 2020a). Finally, we also included variables related to family such as “Number of children under 5” and “Number of household members”, as well as to employment status, which we treated as a dichotomy where a value of 1 was given if the women surveyed replied yes to the variable of “Currently working”, and a value of 0 otherwise.

Furthermore, to be able to carry out comparisons of the effects of the earthquakes on gender relations in rural and urban areas, we divided the survey data for each country into two subsets. Doing this enabled us to carry out two DID estimations, one for urban areas and one for rural areas, which are expressed by the following model:

$${WF}_{i}={\beta }_{0j}+ {\beta }_{1j}{t}_{ij}+ {\beta }_{2j}{R}_{ij}+ {\beta }_{3j}{SH}_{ij}+ {\beta }_{4j}\left({t}_{ij}*{R}_{ij}*{SH}_{ij}\right)+\dots + {\beta }_{kj}{X}_{kij}+ {u}_{i}$$

where \({X}_{k}\) refers to each of the k covariables taken into account, \(j\) denotes the territory analysed (the Dominican Republic \(j=DR\); Honduras \(j=Ho\); and Haiti \(j=Ha\)) and u is the random disturbance term. It should be noted that the robustness of the model is not compromised as the value of each of the k covariables considered was not affected directly by the earthquakes (Stock and Watson 2015).

3.1 Estimation and results

3.1.1 Analysis of the model assumptions

Prior to estimating the model, we contrasted the parallel trends assumption necessary to be able to apply the DID method (Athey and Imbens 2006; Wing et al. 2018). Through this assumption, we were attempting to demonstrate that, in the absence of contamination, the wealth factor would follow a similar pattern over time in both the control group and the affected group. That is to say, if there had not been an earthquake, the evolution of the wealth factor would have been similar for people affected by the earthquake and those who were not. If this condition were not to be met, the results of the DID estimation would not be valid as the effect of the earthquake would be confounded with differences that could be accounted for by changes in the tendency in the WF over the same time period.

Although we cannot directly observe the behaviour of the group of people affected by the earthquake in a context where the disaster is absent, we can, however, get an indication by examining their behaviour in time periods preceding the earthquake. In the case of the Dominican Republic and Haiti, there are survey data prior to 2002 and 2005/2006, respectively, –which can be used to carry out the minimum-quadratic estimation:

$${WF}_{ij}={\beta }_{0}+ {\beta }_{1}{t}_{ij}+ {\beta }_{2}{R}_{ij}+ {\beta }_{3}{SH}_{ij}+ {\beta }_{4}\left({t}_{ij}*{R}_{ij}*{SH}_{ij}\right)+ {u}_{i}$$

This estimation enables the evaluation of whether the slope of the straight line is statistically the same for the control group and the affected group. Tables 3 and 4 show the results of the estimations for the Dominican Republic and for Haiti in terms of, respectively, urban and rural areas. For the Dominican Republic, the variable t has a value of 1 if the data refers to the year 1990, and 0 otherwise. For the regressions corresponding to Haiti, a value of 1 corresponds to data from the year 2000, and a 0 value indicates the data is not from this year. Since the interaction in terms of time, region and sex of the head of household is not significant for either the urban data (ρ – value DR = 0.3190; ρ – value Ha = 0.7440) or that for rural areas (ρ – value DR = 0.3779; ρ – value Ha = 0.4566), we can state that the parallel trends assumption is met in both countries.

Table 3 Validation of the parallel trends assumption for urban areas in the Dominican Republic and Haiti
Table 4 Validation of the parallel trends assumption for rural areas in the Dominican Republic and Haiti

In the case of Honduras, no previous data were available prior to that collected in 2005/2006, and therefore we carried out two “placebo” tests (Gertler et al. 2017; Furquim et al. 2020). The first was conducted with the control group and the results are shown in Table 5 (urban areas) and Table 6 (rural areas). From the group formed by the provinces not impacted by the earthquakes, provinces were randomly selected to form the “false” affected group (namely the provinces of Colón, La Paz, Lempira, Olancho and Valle (D = 1). The results obtained demonstrate that the interaction between time, region and sex of head of household was not statistically significant (ρ – value urban areas = 0.6164; ρ – value rural areas = 0.8298). This means that, among the areas not affected by the earthquake, the parallel trends in the evolution of the wealth factor in the predisaster period were maintained in the postdisaster period.

Table 5 Validation of parallel trends assumption in urban areas in Honduras for the control group in the placebo test
Table 6 Validation of parallel trends assumption in rural areas in Honduras for the control group in the placebo test

The second placebo test was carried out on the affected group, that is to say, data from the provinces impacted by the earthquakes. The results are shown in Table 7 (urban areas) and Table 8 (rural areas), and we randomly selected as the control group the provinces of Atlántida and Cortés (D = 0) for the calculations. The results demonstrate that the interaction between time, area and sex of head of household was not statistically significant (ρ – value urban areas = 0.2534; ρ – value rural areas = 0.9070), thus supporting the null hypothesis of parallel trends.

Table 7 Validation of parallel trends assumption in urban areas in Honduras for the affected group in the placebo test
Table 8 Validation of parallel trends assumption in rural areas in Honduras for the affected group in the placebo test

4 Results

The results of the DID estimation based on data from the DHS are shown in Table 9 (urban areas) and Table 10 (rural areas). The F-statistic test indicates that the model proposed in this work to analyse the socioeconomic impact of the earthquakes that hit the Dominican Republic, Honduras and Haiti is statistically significant for p-value < 0.01, both in urban areas (F-StatisticDR = 617.7776, F-StatisticHo = 344.3428, F-StatisticHa = 161.1067) and in rural areas (F-StatisticDR = 379.9297, F-StatisticHo = 66.9632, F-StatisticHa = 268.2927).

The variable t incorporates the fixed time effects of the earthquakes analysed, and it is negative for all the urban areas analysed (\({\beta }_{1DR}\)= -0.102755, \({\beta }_{1Ho}\)= -0.218548, \({\beta }_{1Ha}\)= -0.341643), and the effect was statistically significant\(\left(p-value< 0.0001\right)\). This indicates that after the disaster there was a generalized drop in the wealth factor of families living in urban areas. In rural areas, the regression coefficient for t was negative in the Dominican Republic and Honduras (\({\beta }_{1DR}\)= -0.518463, ρ – value = 0.0000; \({\beta }_{1Ho}\)= -0.065117, ρ – value = 0.0138), while for Haiti, it was positive (\({\beta }_{1Ha}\)= 0.13834, ρ – value = 0.0532). These results demonstrate that, after the disaster, families in rural areas of the Dominican Republic and Honduras experienced a drop in their wealth factor but families in rural in Haiti actually rose after the disaster.

The fixed effects of territory corresponding to the variable “R” show some differences between countries. In urban areas in the Dominican Republic and Honduras, the effects are positive, albeit not statistically significant with respect to the Dominican Republic. Living in the provinces affected by the earthquake in Honduras positively affected the income of families. In Haiti, the fixed effects of province of residence are negative (\({\beta }_{2Ha}\)= -0.197027, ρ – value = 0.053), demonstrating that living in an urban area affected by the earthquake reduced families’ wealth factor. For their part, in the rural areas in Honduras and Haiti these effects were positive and statistically significant (\({\beta }_{2Ho}\)= 0.528355, ρ – value = 0.0000; \({\beta }_{2Ha}\)= 0.200915, ρ – value = 0.0041), while in the Dominican Republic they were not statistically significant (ρ – value = 0.1891).

The fixed effects that the sex of head of household had on family income are reflected in the coefficient of the variable “SH”. In urban areas, they are positive and statistically significant \(\left(\alpha <0.1\right)\) for all the countries analysed (\({\beta }_{3DR}\)= 0.020168, ρ – value = 0.0646; \({\beta }_{3Ho}\)= 0.055402, ρ – value = 0.0027; \({\beta }_{3Ha}\) = 0.047988, ρ – value = 0.0000), which indicates that having a male head of household positively influences a family’s wealth factor. In rural areas, however, the effects are negative in each of the three countries (\({\beta }_{3DR}\)= -0.095767, ρ – value = 0.0532; \({\beta }_{3Ho}\)= -0.099534, ρ – value = 0.0001; \({\beta }_{3Ha}\)= -0.269611, ρ – value = 0.0000), which demonstrates that, in rural areas in these countries having a male head of household has a negative impact on a family’s wealth factor.

The interaction between the variables t, R and SH encompasses the predisaster and postdisaster differential in terms of their impact on household wealth as a function of the sex of the head of household. It is positive in all the regions studied in both urban areas (\({\beta }_{4DR}\)= 0.156418, ρ – value = 0.0000; \({\beta }_{4Ho}\)= 0.123683, ρ – value = 0.0000; \({\beta }_{4Ha}\) = 0.062065, ρ – value = 0.0447) and rural areas (\({\beta }_{4DR}\) = -0.095767, ρ – value = 0.0532; \({\beta }_{4Ho}\) = -0.099534, ρ – value = 0.0001; \({\beta }_{4Ha}\)= -0.269611, ρ – value = 0.0000). These results point to the earthquakes having had a generalized negative impact on household income, and that the effect intensified when the head of the household was a woman. Effectively, the household income of women-led households dropped with respect to that of households headed by men, and the gap between them widened.

In our analyses we took into account a number of circumstances that could influence the income level of a family. To this end, the statistically significant positive effect \(\left(\alpha <0.01\right)\), in urban areas with respect to the variables “Secondary education” (urban areas: \({\beta }_{5DR}\) = 0.526892, \({\beta }_{5Ho}\) = 0.675417, \({\beta }_{5Ha}\)= 0.310639; rural areas: \({\beta }_{5DR}\) DR = 0.637235, \({\beta }_{5Ho}\) = 0.876681, \({\beta }_{5Ha}\) = 0.778678) and “Higher education” (urban areas: \({\beta }_{6DR}\)= 0.846420, \({\beta }_{6Ho}\) = 1.042829, \({\beta }_{6Ha}\) = 0.785035; rural areas: \({\beta }_{6DR}\) = 0.837181, \({\beta }_{6Ho}\)= 1.39853, \({\beta }_{6Ha}\) = 0.90881), is a reflection of the fact that a family’s wealth increases with the higher educational level of the head of the household. In addition, age has a positive influence \(\alpha <0.01\)on household wealth in the urban areas analysed for all countries (\({\beta }_{7DR}\)= 0.035899, \({\beta }_{7Ho}\) = 0.03503, \({\beta }_{7Ha}\)= 0.025141), as it also did (\(\alpha <0.05)\) in rural areas (\({\beta }_{7DR}\) = 0.045302, \({\beta }_{7Ho}\)= 0.028399, \({\beta }_{7Ha}\) = 0.022179). For its part, data for the variable “Age2” indicate that once the maximum level has been reached, its effect on family income reduces year on year \(\left(\alpha <0.01\right)\), in both urban areas (\({\beta }_{8DR}\) = -0.000297, \({\beta }_{8Ho}\)= -0.000345, \({\beta }_{8Ha}\) = -0.000326), and in rural areas (\({\beta }_{8DR}\) = -0.000502, \({\beta }_{8Ho}\) = -0.000309, \({\beta }_{8Ha}\) = -0.00036).

The impact of family structure on wealth factor is demonstrated through the variables “Number of children under 5” and “Number of family members”. Having children under the age of 5 has a significant negative effect on family wealth factor, both urban areas (\({\beta }_{9DR}\) = -0.109555, \({\beta }_{9Ho}\) = -0.074360, \({\beta }_{9Ha}\) = -0.137418) and rural areas (c\({\beta }_{9DR}\) = -0.148339, \({\beta }_{9Ho}\) = -0.080566, \({\beta }_{9Ha}\) = -0.051152). In contrast, the effect of the size of the household differs between countries and area (urban/rural) studied. While in urban areas in all three countries the effect of number of family members on household wealth is significant \(\left(\alpha <0.01\right)\), in the Dominican Republic and Haiti this effect is positive (\({\beta }_{10DR}\) = 0.030982, \({\beta }_{10Ha}\) = 0.047688), but it is negative in Honduras (\({\beta }_{10Ho}\) = -0.018198). In rural areas, the variable is also statistically significant for each country\(\left(\alpha <0.05\right)\), and once again the sign of the effect differs between countries: in the Dominican Republic the effect the size of a household has on its wealth is positive (\({\beta }_{10DR}\) = 0.023455), but is negative in Honduras and Haiti (\({\beta }_{10Ho}\)= -0.028258, \({\beta }_{10Ha}\) = -0.035162).

Finally, the “currently working” variable captures the fixed effects on household income of whether or not the women surveyed are in paid employment. The data show that in urban areas of Honduras and Haiti, the effect was positive and statistically significant (\({\beta }_{11Ho}\)= 0.107394, \({\beta }_{11Ha}\)= 0.071542). The effect for the Dominican Republicwas not statistically significant (ρ-value = 0.5697). With respect to rural areas, the effect was positive and statistically significant (\(\alpha <0.01 )\)for the Dominican Republic and Honduras (\({\beta }_{11RD}\)= 0.116638, \({\beta }_{11Ho}\)= 0.121942), but was not statistically significant for Haiti (ρ-value = 0.2164).

Table 9 Results of the estimation of the effect of the earthquakes in the Dominican Republic (2003), Honduras (2007 and 2009) and Haiti (2010) on gender relations in urban areas
Table 10 Results of the estimation of the effect of the earthquakes in the Dominican Republic (2003), Honduras (2007 and 2009) and Haiti (2010) on gender relations in rural areas

5 Discussion

The results of our study reveal, in line with earlier scientific evidence, that the earthquakes that occurred in the Dominican Republic (2003), Honduras (2007 and 2010) and Haiti (2010) have had a negative impact on the wealth factor of households headed by women, in urban areas as well as rural areas (Gray 1993; Enarson et al. 2006; Zottarelli 2008; Horton 2012; Bradshaw and Fordham 2013; Llorente-Marrón et al. 2020a; Lebni et al. 2020). Furthermore, it is of particular note that, in rural areas, the sign of the effect that the disaster had on household wealth with respect to the gender of the head of the household changed from the predisaster period to the postdisaster. Although before the earthquake having a male head of household had a negative effect on household income, following the disaster it became families headed by women that suffered drops in their income. This change may be due, as other studies have indicated, to the fact that women’s vulnerability increases following a disaster, particularly in rural areas (Gray 1993; Afriyie et al. 2017; De Silva and Kawasaki 2018; Hamidazada et al.2019; Lebni et al. 2020; Llorente-Marrón et al. 2020b; Lin et al. 2021).

Some works have suggested that the greatest impact of a disaster on female-headed households is due to two aspects. On the one hand, the intervention and recovery process often exclude women from fair compensation and support. On the other, the reproductive and community roles that women play are usually ignored through directing the bulk of post-disaster recovery efforts to supporting productive activities. (International Recovery Platform 2010). In addition, in the development of such recovery plans the informal economy is generally not taken into account, this traditionally being a work environment in which many women find themselves, particularly in rural areas. As such, many women are unable to return to activities they worked in prior to the disaster and are obliged to rely financially on a partner or family member or take up precarious employment (Arenas Ferriz 2001; Bradshaw and Fordham 2013).

Furthermore, another line of investigation has found that age and educational level have a positive effect on family income (Koopmans et al. 1989; Cao et al. 1996; Fan 2003; Cuaresma 2010). It is worth highlighting that educational level is also a determining factor with respect to preparedness for a disaster, in both the emergency phase and the evacuation phase, as well as in terms of postdisaster adaptation (Afriyie et al. 2017; Chowdhury et al. 2021; Petraroli and Baars 2022). In this sense, it should be taken into account that disaster preparation on the part of women and girls may be affected by the fact that in many territories their access to education, particularly secondary and higher education, is limited and that in certain disaster situations girls are obliged to abandon their education before even finishing primary school (Wakefield 2005; Tanner et al. 2022).

As far as family structure is concerned, our results demonstrate that having children under the age of 5 has negative repercussions on family income in all the territories and areas (rural/urban) studied. Having a child is linked to loss of earning power for women, and can make it more difficult for them to enter the job market or oblige them to go back to work part-time or give up their job completely in order to dedicate their time to the care or education of their children (Mincer and Polachek 1974; Becker 1985; Gafni and Siniver 2015; Campos-Vazquez et al. 2022; Dominguez-Folgueras et al. 2022).

In contrast, the impact of the number of household members on family income varied between countries, and also, on occasion, between rural and urban areas. In the case of the Dominican Republic, an additional family member has a positive effect in both rural and urban areas, while in Honduras, the effect is negative in both areas. These divergences may be the result of differences in per capita income in the two countries at the time of the respective earthquakes. Whilst the active population at the time of the disaster in the three countries was around 50%, GDP per capita –in current $US– was $2,399.5 in the Dominican Republic in 2003, but in Honduras it was only around $1,559.89 in 2007 and $1,762.35 in 2009 (World Bank 2023). With respect to Haiti, the impact of an additional family member on household income is positive in urban areas, but negative in rural areas. The productive structure of this country, which had a per capita GDP –in current $US– in 2010 of $1,204.86 and where there is great polarization between the primary and tertiary sectors and the employment rate of children aged 7 to 14 is approximately 40%, may explain the positive effect found in urban areas (World Bank 2023). In rural areas, the negative effect might be a consequence of the housing conditions in these areas, characterised by limited access to basic resources such as drinking water and toilet facilities (Ministry of Public Health and Population 2013), making many homes inadequate for housing large numbers of people. Adding an extra member to the household under such circumstances may involve additional costs and therefore impact negatively on household income.

6 Conclusions

In this article we have analysed the socioeconomic impact of the earthquakes that occurred in the Dominican Republic (2003), Honduras (2007 and 2009) and Haiti (2010) with respect to gender relations, comparing urban and rural areas. Our results are in line with various previous works and demonstrate that whilst these disasters affect the population as a whole, families headed by a woman are affected more severely, particularly in rural areas. As such, gender inequalities existing prior to the earthquake do not disappear and are in fact heightened after the disaster.

The dependent variable “Wealth Factor” employed in this work is comprehensive in character in that it not only takes into account family income, but also their living conditions, such as access to electricity, drinking water and domestic appliances. Its use permits a broader picture to be seen in terms of the realities of the families affected by the earthquakes. For this reason, the results allow us to draw the conclusion that the greater impact felt by female-headed households as a result of these disasters is not only economic, but also affects their living conditions. In addition, we observed how variables such as educational level, having young children, number of household members and age also influence household wealth.

To date, only a few studies have carried out a joint analysis of disasters of natural origin occurring in different countries (Neumayer and Plümper 2007; Austin and McKinney 2016; Eastin 2018; Aheeyar et al. 2019; Austin et al. 2021; Newnham et al. 2022; Sohrabizadeh and Parkinson 2022; Lecoutere et al. 2023; Nguyen and Nguyen 2023). In the case of the Dominican Republic, Honduras and Haiti, we have been able to establish that, despite the socioeconomic specificities of each of the three countries, the trends with respect to the impact of the earthquakes are similar, and the greatest differences are linked to internal inequality between urban and rural areas. Furthermore, our analysis has enabled us to observe that the gender inequalities are not situational in nature, rather, in the three studied countries, they are systemic and structural.

The findings of this study deepen our understanding of the impacts of disasters of natural origin. Moreover, they have serious political implications which should be taken into account when developing disaster risk reduction policies since they demonstrate that the gender inequalities that are known to exist in the postdisaster period are no more than a reflection of those that existed prior to the catastrophic event. In this sense, public policies should be developed that address the situation prior to the disaster and focus on improving the material living conditions of women, encouraging girls’ and women’s access to education at all levels, promoting women’s incorporation in the job market and protect the positions and remuneration of women who become mothers or are responsible for young children. Likewise, policies should be enacted that incentivise women becoming owners of good quality homes. Home ownership is not only a symbolic element of family and community empowerment, but also a condition that reduces women’s vulnerability in disaster scenarios (Fernández Saavedra et al. 2023).

The research conducted in this paper has allowed us to contribute to the existing literature on gender and disasters through a comparative analysis of earthquakes occurring in different countries. However, there is a limitation to the findings that needs to be addressed. While earthquakes are the main object of analysis in this article, it should be noted that other catastrophic events could have occurred in each country within the two DHS data collection periods. These disasters, which have not been included, may have had some impact on gender inequalities that are not reflected in this publication.

Finally, we cannot end this article without highlighting the fact that the sources of data that are disaggregated by sex which have allowed us to reveal the gendered impact of disasters of natural origin are limited in number. This lack of data conditions not only the production of knowledge about gendered effects, but also the development of appropriate policies to combat existing gender inequalities. It is therefore essential that public administrations and national and international bodies and organisations set up data collection procedures that involve disaggregation not only by sex, but also by age and socioeconomic level, as well as other relevant sociodemographic characteristics in order to develop research and public policies that are more tailored to the realities of people’s everyday lived experience.