Introduction

Disasters occur when an individual or group is vulnerable to the impact of natural or human-induced hazards (Wisner et al. 2004; Pantuliano 2007). Disasters often lead to the breakdown of normal behaviour (Fielding 2018), as affected communities incur psychological, economic, and social consequences that often derail their sources of livelihood (Manyena 2006). The ramifications of disaster include persistent disruption of social life (Akbar and Aldrich 2018), social disorganisation, escalating social impoverishment (Pantuliano 2007; Yang et al. 2018a), and destabilisation of existing societal functions and structures (Akbar and Aldrich 2018). Nonetheless, it is important to note that individuals living in resource-scarce or conflict-trapped communities are often disproportionately vulnerable to disasters triggered by natural hazards, such as those caused entirely by weather-related events (floods), meteorological phenomena (storms), and geophysical (earthquake) hazards (Birkmann 2006) as well as the disasters caused entirely by human activities, such as violent conflicts and terrorism (Alexander 2002; Theisen et al. 2013).

While there is insufficient reliable data regarding fatalities disaggregated by gender (Fielding and Lepine 2017), a study by Neumayer and Plümper (2007) found that more women die as a result of disasters than men. Women have more relative vulnerability given the unequal gendered power relations that limit their access to and control over resources (Bradshaw 2014). The findings of previous empirical studies have supported the claim that disasters, particularly in Sub-Saharan Africa (SSA), disproportionately affect female smallholder farmers, thus widening the existing gender inequalities and exacerbating socio-economic and political risks (Tschakert 2012; Bradshaw 2014; Olaniyan and Yahaya 2016). Similarly, women are often classified as the most disadvantaged in SSA due to discriminatory policies, beliefs, and practices that in most cases limit their accessibility to livelihood resources (Fielding and Lepine 2017). As such, limited access to credit and risk-management instruments (Sewando et al. 2016), the paucity of information (Bradshaw 2014), and structural inequality and disempowerment undermine the ability of women to respond to (Terry 2009) and cope with the impact of disasters (Birkmann 2006).

The assessment of social vulnerability has in recent years become an issue of concern to the global environmental change, sustainability, and inclusive development research communities (Fekete et al. 2010; Yang et al. 2018a). While acknowledging the fact that the term vulnerability has been variedly defined among the different disciplinary orientations (Cannon 2006; Fuchs et al. 2011), the term serves as a major nexus linking the social sciences and natural sciences in the context of sustainable development, particularly on the reduction of disasters and sudden-onset hazards as well as the strategies that would ease climate change adaptation (Fekete et al. 2010; Fielding 2018). The literature focused on climate discourse defines vulnerability as the character, size, and rate of climate change, which mirrors the characteristics of natural phenomena (IPCC 2014). Vulnerability indicates the degree of loss of an individual, household, and community at risk due to the effect of hazards (Wisner et al. 2004; Adger 2006; Fell et al. 2008). In contrast, disaster discourse asserts that vulnerability is the characteristics of the exposed elements or society and their capacity to cope and recover from hazards or stressors (Wisner et al. 2004; Manyena 2006). Similarly, vulnerability determines the state of susceptibility to harm from exposure as a result of environmental and social change and/or due to the reduced capacity to adapt (Adger 2006). However, in the development context, vulnerability indicates a lack of buffers against shocks that frequently harm livelihood (Chamber 1989).

While the definition of vulnerability has been debated in the literature, a growing number of studies have argued for possible integrative discourse among the various scientific approaches to vulnerability that focus substantially on the complex interactions between the exposure, sensitivity, and adaptive capacity of a system (Fuchs et al. 2011; Yang et al. 2018b). Adger (2006) argued that exposure measures the degree to which an area is increasingly in a state of disturbance, whereas sensitivity indicates the possible impact of hazardous events on a system (Fuchs et al. 2011; Yang et al. 2018a, b). Conversely, adaptive capacity measures the ability of a system to enhance its capacity to cope, respond, and recover from hazards or stressors with little or no external assistance (Wisner et al. 2004; Manyena 2006; Fekete et al. 2010; Yang et al. 2018a, b). Thus, vulnerability is referred to as the conditions determined by physical, social, economic, and environmental factors or processes that increase the susceptibility of an individual, a community, assets, or systems to the impacts of hazards or reduce the capacity to recover from such impacts (United Nations General Assembly, henceforth UN General Assembly 2016). In this context, vulnerability analysis involves the likelihood that people will be affected by a particular hazardous event, as well as the type of livelihoods in which they engage and the impact of different hazards on them. In an attempt to reduce risk and vulnerability, a consensus was reached among the international community that perennial disasters should be monitored (Birkmann and Teichman 2010; UN General Assembly 2016) as an essential prerequisite for reducing disaster risk (Birkmann 2006), particularly in resource-scarce SSA where the capacities of people to respond to shocks is being rapidly weakened due to the limited livelihood diversification. In this way, there are important studies in the literature on vulnerability evaluation methods (Fekete et al. 2010; Ouma and Tateishi 2014; Sewando et al. 2016; Yang et al. 2018a, b) that have strongly emphasised the need to assess the vulnerability status of individuals, households, and communities.

Northern Nigeria is one of the most highly volatile and vulnerable regions of the world (Ibrahim et al. 2016). New security threats from Kiwo Haram (organised cattle rustling), Boko Haram, farmers-herders conflicts, and the Islamic State of Iraq and Syria have emerged and combined with the perennial ethno-religious crisis to increase risks and vulnerability (Olaniyan and Yahaya 2016). Furthermore, the climatic conditions have increased the region’s vulnerability to recurrent shocks triggered by extreme events such as droughts and floods (Fjelde 2015). While insufficient rainfall causes droughts due to a lack of sufficient irrigation facilities, excess precipitation causes floods that affect properties and farmlands (Birkmann and von Teichman 2010; Yang et al. 2018a). This has significantly increased the tendency of future conflicts related to climate change or water shortages, farm scarcity, and degradation (Aliero and Ibrahim 2013), particularly in the north-western sub-region where a large proportion of the population relies on rain-fed agriculture (Fjelde 2015) and any impact on the local agricultural prices increases the risk of violent events (Ibrahim and Aliero 2012).

The recurrent shocks triggered by the perennial armed banditry in remote areas has further weakened households’ coping capacities (Olaniyan and Yahaya 2016), as livelihood strategies are substantially based on livestock husbandry (Ibrahim and Aliero 2012) and crop production (Sewando et al. 2016). As such, the customary practice of mixed crop and livestock production has been completely disrupted. Rural dwellers who previously contributed approximately 3.2% of Nigeria’s GDP from cattle production alone are now struggling to cope with the vicious cycle of rustling (Olaniyan and Yahaya 2016). Thus, measuring the extent of vulnerability to recurrent shocks in remote areas of Nigeria is one of the key areas of focus of this paper.

The renewed interest in the gendered vulnerability discourse stems from the claims of disproportionate effects of disasters even within members of the same household, as it has recently been reported that women and children are more vulnerable (Gaillard et al. 2017; Fielding 2018). The gender dimensions of hazards have attracted significant attention within the literature for more than two decades (see previous studies conducted by Fothergill 1996; Fordham 1998; Enarson and Fordham 2001; Bradshaw 2014). Contrary to disasters triggered by hazards of natural origin, the human-driven disasters in SSA, particularly in northern Nigeria, have not been adequately addressed in the literature. A gendered assessment of the recurrent shocks, particularly in regard to how they affect livelihood strategies in conflict-trapped areas, is indispensable for disaster prevention, recovery, and management strategies. Several studies have documented that recurrent shocks driven by both natural hazards as well as human activities would increase the vulnerability to food price volatility, forceful migration, crop failures, climate change (Birkmann 2006; Pantuliano 2007; Yang et al. 2018b), and land grabs (Ray-Bennett 2010; Theisen et al. 2013), halting development (Akbar and Aldrich 2018) and destroying homes (Ibrahim et al. 2016) and public properties.

Certainly, disasters are becoming increasingly frequent (Altay et al. 2013; Gaillard et al. 2017), costly, and devastating (Lee et al. 2018), thus disrupting the supply chains of micro-enterprises (Kouvelis et al. 2006). Informal micro-enterprises are the backbone of livelihood strategies in the majority of developing countries around the world (Prasad et al. 2015). When disasters occur, all components of the supply chain are affected to some extent (Altay et al. 2013). As such, a key question is: What is the impact of recurrent shocks on livelihood diversification? Aside from several qualitative analyses (Fjelde 2015; Olaniyan and Yahaya 2016) and empirical studies (Ibrahim and Aliero 2012; Ibrahim et al. 2016), there has been minimal in-depth research into the effects of recurrent shocks on rural livelihoods in Nigeria. This study contributes to the debate on the disaster discourse and the livelihood diversification nexus, highlighting the effects of disaster shocks on intra- and inter-gender income disparities. This study further suggests effective disaster intervention strategies that would not only lessen the effects of disasters but also help to narrow the pre-existing inequalities.

The rest of the paper has been structured such that the next section presents a detailed research methodology. Thereafter, the results and a discussion are presented in “Results and discussion”, and the final section concludes the paper.

Methodology

Computational procedures

Cattle rustling vulnerability index

The empirical measurement of the cattle rustling vulnerability index (CRVI) depends on its conceptual definition. Drawing on Paavola’s (2008) conception of vulnerability, cattle rustling vulnerability can thus be defined as a herdsmen’s weak prevention capacity when faced with organised livestock raids. It is essential to stress that vulnerability is used in this study to denote weak adaptive capacity, in the sense that strong adaptive capacity implies a reduced vulnerability.

The multifaceted nature of disasters induced by the combination of hazards of natural origin as well as human activities underlines the importance of developing multidimensional indicators to measure the vulnerability and coping capacities of individuals in order to improve their disaster preparedness and to promote more disaster-resilient societies (Birkmann 2006). In this sense, the computational strategy of the CRVI involves systematic aggregation of the various indicators of adaptive capacity, exposure, and sensitivity to shocks driven by cattle rustling. The CRVI model can thus be expressed as:

$$ {CRVI}_i=\left({EXP}_i+{SEN}_i\right)-{ADC}_i $$
(1)

where CRVIi is the cattle rustling vulnerability index, EXPi and SENi are exposure and sensitivity to cattle raiding, respectively, that together form what is defined as “the impact” in the vulnerability model, and ADCi is the adaptive capacity. Following Sewando et al. (2016), ADC was measured with factors that determine the ability of individuals to cope with and recover from shocks driven by the raiding of livelihood assets. The first indicator of ADC is human assets, which is measured by years of education. This indicator determines the extent of life adjustment, including optimism for the future. Another important indicator of human assets is the ability to access relevant information, as this substantially determines the choice of effective cattle rustling mitigating strategies. Three questions were asked to measure the extent of the respondents’ information accessibility (see Table A of the supplementary material). The subscales of access to information were scored using a three-point Likert scale that yielded a Cronbach’s alpha value of 0.812.

Following Alexander (2002), civil protection has grown in response to the need to protect households against violent conflicts. Consequently, EXP has three sub-components that measure the nature of local security using indicators of availability of the formal security structure (such as police force, civil defence, and military), informal security network, and legal system. Four questions were asked about the respondents’ perception of formal security formation, three about the informal security network, and one about the legal system. The Cronbach’s alpha of the Likert scale used for scoring the questions was 0.785. Finally, the sensitivity had five sub-components, including herd size, number of dependents (such as siblings), and access to formal financial institutions. The Cronbach’s alpha for the measurement scales of SEN was 0.832.

Natural hazards vulnerability index

ActionAid International’s (2005) participatory vulnerability assessment (PVA) framework guided the selection of various indicators of Natural hazards vulnerability index (NHVI). The residents of hazard-prone villages were fully involved in the selection process of indicators particular to their villages. Another version of the vulnerability equation built on the ability of an individual or household to cope with and mitigate the impact of hazards (Webb and Harinarayan 1999) was then used to aggregate the multidimensional indicators of NHVI:

$$ {NHVI}_i={Hazard}_i-{Coping}_i $$
(2)

where NHVIi denotes the natural hazards vulnerability index; hazard was then defined by two proxies: environmental shocks and droughts (their measurements are provided below). Although it is a relative variable that can either be positive or negative (Sewando et al. 2016), a positive NHVI points to a higher vulnerability whereas a negative value indicates less vulnerability.

Several variables such as drought, erosion, flood, deforestation, and population density have previously been used in the literature to measure the environmental aspects of natural hazards (for example, see Alam et al. 2017; Akbar and Aldrich 2018; Yang et al. 2018b). To measure the shocks emanating from the flooding hazards, five household-centric flood assessment indicators specific to the survey area are measured using a combination of subjective and objective factors, as illustrated in Table B of the supplementary material. Flood as a rapid-onset event can hamper the resilience of the victims through the loss of valuable properties and physical structure and, in extreme situations, can lead to the death of members of the household (Cannon 2006; Fekete et al. 2010).

However, the slow-onset event variables such as erosion and drought have reduced likelihoods of inducing mortality. A range of individual-specific questions were asked to ascertain their risks and vulnerabilities. The questions about floods, erosions, and droughts were scored using a five-point Likert scale ranging from “very high” (5) to “very low” (1); a score of five (5) was rated as the most vulnerable and one (1) was the least. Cronbach’s alpha for these scales yielded 0.863, 0.821, and 0.869 for flood, erosion, and drought, respectively. However, a question about deforestation was scored using a two-point Likert scale yielding a Cronbach’s alpha of 0.912. While there are concerns about the potential subjectivity in these scales, as their respective Cronbach’s alpha values exceeded the 80% threshold, this demonstrated that the scales are effective at measuring hazards and have demonstrated uni-dimensionality and reliability.

The sustainable livelihoods framework of Scoones (1998) was instrumental in determining the indicators of coping strategies. The study carefully selected the appropriate indicators that measured the resilience, responses, and recoveries particular to the affected villages. There are four components of the coping strategy that measure the financial, human, social, and natural capital. It was hypothesised that these capitals add resilience and enhance the capacity of the affected individuals to respond and recover with little or no external assistance (Manyena 2006; Aliero and Ibrahim 2013). For instance, previous literature has suggested that higher levels of social capital can contribute to overcoming the severity of disasters (Scoones 1998; Adger 2006). In this regard, two questions were asked about financial capital. The first measured the respondents’ financial savings, and values of one (1) through five (5) were assigned, similar to the vulnerability scale. Second, access to formal financial services was scored using a two-point Likert scale, with 1 for yes and 0 otherwise. The Cronbach’s alpha for this scale was 0.815. Additionally, in line with previous literature, human capital was measured as years of formal education. However, social capital was measured as a binary variable similar to the strategy used for access to finance and yielded a Cronbach’s alpha of 0.831. Finally, natural capital was measured as the size of crops per capita in the agricultural land use (in acres).

Empirical model

The aim of the present study was to establish whether and to what extent recurrent shocks affect income diversification, serving as a proxy for livelihood diversification strategies. To this end, Eq. (3) expresses the impact of recurrent shocks on income earned through diversified economic activities, given as:

$$ {Y}_{di}={\beta}_0+{\beta}_1{CRD}_i+{\beta}_2{NHD}_i+{\gamma}_1{Z}_{\mathrm{i}}+{\mu}_i $$
(3)

where Ydi is income diversity per capita of the i-th individual, β1and β2 are the parameters of cattle rustling-driven and natural hazard-induced shocks, respectively, Zi is the vector of control variables, such as age, gender, literacy, and the ratio of income from crop and livestock production, γ1 is the vector of their respective coefficients, and μi is the white noise error term, μi~N(0, σ). The parameters in Eq. (3) were estimated via the simple ordinary least squares (OLS) regression method. It has been shown in previous empirical and theoretical literature that once the estimated residual is independently and identically identified in this method, then OLS is an unbiased estimator and can thus be used to explore variable inter-relationships at different points in the conditional income distribution (Stone and Brooks 1990; Falaris 2003; Akbar and Aldrich 2018).

As a robustness exercise, the propensity score matching (PSM) technique of Rosenbaum and Rubin (1983) was applied. The overall data were divided into cattle rustling victims versus non-cattle rustling victims on the one hand, and then natural hazard victims versus non-natural hazard victims on the other. Thereafter, four matching algorithms consisting of the nearest neighbour, radius, stratification, kernel, and local linear regression were applied to compare the estimated parameters of Eq. (3) with the PSM results.

One of the central objectives of this study is to establish whether recurrent shocks have any effects on inter- and intra-gender income disparity. To achieve this objective, decomposition analysis was performed to establish the relative contribution of between-group variance (differences of disaster effects between male and female respondents) and within-group variance (within each gender) to income diversification. If the between-group variance exceeds the within-group variance, then recurrent shocks exert a larger influence compared to other factors explaining the inter-gender disparity. The decomposition exercise follows the methodological strategy developed by Fields (2003) for treatment effect models based on the OLS estimation.

To begin the decomposition, let V0(yd) and V1(yd) be the variances in income for the female and male respondents, respectively. If pmi is the proportion of male victims, the within-group variance Vw(yd) and between-group variance Vb(yd) are given as:

$$ {V}^w\left({y}_d\right)=\left(1-{p}_{mi}\right){V}_0\left({y}_d\right)+{p}_{mi}{V}_1\left({y}_d\right) $$
(4)
$$ {V}^b\left({y}_d\right)=\left(1-{p}_{mi}\right)\ {p}_{mi}{\left({\overline{y}}_{d1}-{\overline{y}}_{d0}\right)}^2 $$
(5)

where \( {\overline{y}}_{d1} \) and \( {\overline{y}}_{d0} \) are the respective mean values of the income in group 1 (male) and group 2 (female). The coefficient of the between-groups variance measures the extent to which differences in income are driven by shocks rather than other covariates, while the within-group effect measures the extent to which demographic characteristics contribute to various income differences.

Data and descriptive statistics

Longitudinal primary data were collected over a period of 11 months (wave 1 from October to December 2014 and wave 2 from July until September 2015) via two complementary instruments: a structured questionnaire and an interview guide. The sampling procedure for the first wave was designed systematically in such a way that unbiased gender representation was ensured. In this regard, certain requirements such as cognate experience and age, among others, were waved while employing female field enumerators. Similarly, by solely focusing on individuals as opposed to households, the survey was able to cover a seemingly unbiased gender representation.

The survey areas were randomly selected within each of the identified cattle rustling trapped villages. Between 170 and 180 respondents (for a total sample size of 1750 respondents) were systematically selected in 10 villages with an estimated adult population of 250–300 in northern Nigeria (Fig. 1). Additionally, 10 sessions of focus group discussions (FGDs) were held with 100 villagers (60 women and 40 men).

Fig. 1
figure 1

Map of selected rural areas in North West States, Nigeria

Full size image

The aim of the second wave was to re-interview the same respondents who had participated in the first wave, as this would enable the re-examination of their vulnerability and the reassessment of the resilience of their frequently adopted coping strategies. In this regard, those who participated in the first wave were tracked in order to identify their movements. A handful had relocated to new communities and refused to participate in the second wave, which represented a total attrition of 133 respondents or approximately 7% of the sample. The robustness exercise across the two waves reinforced Falaris’ (2003) finding that attrition bias in longitudinal data is not a major concern.

The basic descriptive statistics of the data highlighted in Table 1 show that 53% and 47% of males and females, respectively (p > 0.05), were surveyed in the first wave. Of the overall respondents, the males had relatively more favourable statistics in terms of income (p < 0.01), literacy (p < 0.01), and herd size (p < 0.01). However, the respondents’ average income level was low, with slightly below $1 per day. The average monthly income level even decreased from $28.16 in the first wave to $17.49 in the second wave. This raises doubts in terms of the pace of recovery, as disaster recovery depends heavily on income (Aliero and Ibrahim 2013). Meanwhile, the respondents were gradually shifting away from the customary agro-pastoral mix into diversified livelihoods, such as trade (23%), off-farm jobs (12%), and civil service (7%), to recover and mitigate the impact of disasters (see Wave 1 in Table 1). However, the second wave data showed no discernible changes in livelihood diversification strategies.

Table 1 Demographic characteristics
Full size table

Furthermore, of the 1750 respondents surveyed in the first wave, 79% (81% female and 71% male in the disaggregated data) were victims of disasters triggered by hazards of natural origin and lost a significant amount of wealth, such as herding holdings, cash, and other valuable assets. The reported loss driven by natural hazards in the first wave was approximately $43 million, costing the male respondents approximately $27.57 million, which is relatively higher than the $18.55 million in loss incurred by the females. Although the data showed that there were fewer damages recorded in the second wave, a substantial amount of wealth reportedly lost was higher than the average monthly income. Of the $56.45 million in cattle rustling-driven loss in the first wave, $35.72 million was lost by the males, while $20.73 million was lost by the females (p < 0.05).

Results and discussion

Vulnerability results

A vulnerability assessment is fundamental for informing disaster-adaptation policies (Manyena 2006; Fekete et al. 2010). As noted in “Natural hazards vulnerability index”, positive vulnerability indices point to a higher vulnerability, whereas a negative value indicates less vulnerability. As highlighted in Table 2, the results from the overall data reveal positive values for both CRVI and NHVI, indicating high vulnerability to idiosyncratic shocks. It is possible that rural marginalisation is the primary reason for the results showing the alarming score in overall exposure indices. The ripple effect of these shocks could perpetuate distrust in government (Miller and Rivera 2011), which could exacerbate the alienation of rural areas.

Table 2 Vulnerability indices
Full size table

Moreover, the disaggregated results show that the female respondents had more relative vulnerability to cattle rustling than the male respondents. The CRVI for the female respondents in the first wave was 0.342, which significantly differed from 0.269 for the male respondents (p < 0.05). Women are socially excluded due to purdah customs and lack of access to information, and by implication are highly sensitive to cattle rustling. The indicators of exposure are not gender-specific and their impacts are generic. As such, a comparison between the two groups showed no significant difference (p > 0.05) across the two waves.

With regard to NHVI, the second segment of Table 2 reveals the overall values of 0.264 and 0.275 for the first and second waves, respectively. Contrary to the first wave, the gender-specific result in the second wave shows a significant difference in the index of drought-driven shock (p < 0.05). This is in contrast to the somewhat stationary indices of the coping strategy across the two waves, as the results persistently show a significant difference between the male and female groups (p < 0.01). However, vulnerability to environmental shocks was high regardless of gender, as no significant difference was found across the two waves.

Impact of recurrent shocks on livelihood diversification

The baseline results of the effect of recurrent shocks on livelihood diversification, as highlighted in Table 3, revealed consistent patterns in both the aggregated (overall) and the disaggregated (gender-specific) models. The results show a strong negative effect of hazards shock on income earned through diversified livelihood strategies. This reinforces the findings of Ray-Bennett (2010) that recurrent shocks diminish livelihood assets.

Table 3 Impact of shocks on livelihood diversification
Full size table

Importantly, the coefficients of the cattle rustling are higher in comparison to idiosyncratic shocks. In all of the estimated models, cattle rustling shocks diminished income diversification by approximately 0.5 (p < 0.05). This lends credence to the claim that cattle rustling is deepening the vicious cycle of poverty in northern Nigeria (Olaniyan and Yahaya 2016). This finding is broadly consistent with the plethora of literature that has reported a decrease in household well-being in post-disaster periods (for instance, see Facchetti 2003; Dercon et al. 2005; Tibesigwa et al. 2016; Yang et al. 2018b). Furthermore, most individuals who lost substantial amounts of wealth due to cattle rustling were found to routinely engage in jobs such as environmental resources extraction, cutting firewood, and other occupations capable of degrading the environment. This could explain why the estimated coefficients of disaster triggered by the natural hazards were higher in the second wave than in the first wave.

Moreover, although the results of the demographic variables are consistent with the a priori expectations, it appears that human capital (as measured by literacy) is the most important determinant of livelihood diversification. An increase in the rate of literacy could enhance the likelihood of participation in semi-formal and formal enterprises, which could widen opportunities beyond subsistence agrarianism.

As a robustness exercise, PSM was implemented by matching the scores of victims versus non-victims. The results highlighted in Table 4 show that the coefficients of the average treatment effect on the treatment (ATT) of various matching methods are closer to the coefficients of OLS with the same level of significance for all of the estimated models. This further reinforced the finding that affected individuals have lower income diversity, thus validating the robustness of the estimated models.

Table 4 Propensity score matching
Full size table

The implication of recurrent shocks for the variance of livelihood diversification

Table 5 shows the results of the overall decomposition of the variance of income diversification of the male and female groups, where the interest is in whether between influences (disaster-driven) overshadow the within drivers (not related to disasters) of disparities in the livelihood diversification of the groups.

Table 5. Gender-based variance of income diversification
Full size table

The decomposition results elicit several findings. First, disasters have moderate influences on inter- (between) gender income disparities, as a larger proportion of the inequalities are explained by the within factors (demographic characteristics). Second, the first wave data shows that 9% of between-group differences were accounted for by cattle rustling, while only 5.30% of such differences were driven by natural hazards rather than individual characteristics. However, these disaster-driven inequalities widened as the data from the second wave revealed an increment of income disparities as disasters intensified. The income differences driven by cattle rustling rose from 9 to 12%, while hazard-induced inequality became relatively larger, rising from 5.30 to 14%. This finding is in line with the basis for framing the humanitarian response to disasters with a substantial focus on women and children (Bankoff 2001). Indeed, many studies have expanded the categories of vulnerable classes by including the poor, the very old, the disabled (Fordham 1998; Fjelde 2015; Fielding 2018), and other similar groups that are less able to cope with disasters.

The male-female dualistic construct in disaster discourse contributes to the neglect of intra-gender disparities (Gaillard et al. 2017) in the design and implementation of disaster risk reduction (DRR) policies. Such technocratic and top-down policies ignore other important social factors, such as literacy and income levels that to a larger extent determine the individual’s capacity for disaster preparedness and disaster-resilience (Birkmann 2006; Manyena 2006; Fekete et al. 2010).

Moreover, the findings of this study identify the need for a paradigm shift in DRR strategies focusing more on the drivers of intra-gender inequality. This is consistent with the findings of previous empirical studies, which have reported that people endowed with vast human capital are more capable of bouncing back and recovering with little or no external assistance following a disaster (Manyena 2006; Akbar and Aldrich 2018) and that literacy rates in particular are major components of disaster preparedness, recovery, and resilience (Manyena 2006). Evidence has further shown that even the most vulnerable individuals were able to utilise a wide array of diverse resources, knowledge, and skills to confront disasters triggered by both hazards of natural original (Wisner et al. 2012; Gaillard et al. 2017) as well as human activities.

Conclusion

This study examined the recurrent shocks and disparities in gendered livelihood diversification in Nigerian villages. Two new indices of disasters triggered by human activities and natural hazards were constructed. The results revealed higher vulnerability of recurrent shocks. Furthermore, the OLS results show a stronger negative effect of shocks on livelihood diversification. While not challenging Western insinuations that females are relatively more vulnerable compared to males, the social construction of vulnerability should look beyond gender. In this sense, the DRR strategy should prioritise the intervention that has a direct bearing on a household’s economic, human, and social capital without solely focusing on gender. This underlines the need for a new paradigm in the DRR work culture with the stronger emphasis being placed on household-centric resilience rather than just vulnerability. As such, this study argued that rural financial development has a dedicated role to play in closing financial disparities, both within and between genders, which are widened as the vulnerability to shock increases.

With the cross-fertilisation of human ecology and political economy perspectives and the subsequent development of socio-ecological perspectives to vulnerability and resilience paradigms that served the premises for sustainable livelihood agenda, the emphasis has now shifted to adaptation (proposed through human ecology) and has been extended into the socio-economic context, which ultimately shapes human-environmental relations and expressions of vulnerability (Oliver-Smith 2013). This underlines the need for an important vulnerability evaluation capable of providing insight that would guide the monitoring of risks and vulnerabilities in remote areas. In this sense, there is an urgent need for improvement in vulnerability assessment, particularly for SSA countries such as Nigeria where over two-thirds of the population are living in rural areas and the rural economy is overwhelmingly agro-pastoral based. Linking these customary livelihood strategies with sustainability is key to building resilient livelihoods. This would reduce people’s vulnerability in the wake of projections about the future increase in the frequency and intensity of extreme events. It is left for future research to re-explore the vulnerability evaluation index system by taking into account the multidimensional and non-linear relationships between various vulnerability indicators, as this would provide further improvement to the vulnerability evaluation index for the rural region proposed in this study.