The study took place in the communes of Boukombé and Kérou, located in the Atacora department in north-west Benin. The department covers an area of 20,499 km² in the National Institute of Statistics and Economic Analysis (NISEA, 2013). The department’s economy is based on agriculture, livestock farming, fishing, gathering, crafts and trade. In addition to cotton production, maize and sorghum production are also being developed to exploit the after-effects of fertilizer use on cotton cultivation. These communes were chosen on the basis of their social experience, in addition, they are part of low-income areas affected by the effects of climate change.

Figure 2 below shows the study area.

The sample size was 240 households. The reasoned choice and simple random sampling technique were adopted. Three villages were selected per commune for the field survey. These were the villages of Dipoukomontri, Koupagou and Manta in the commune of Boukoumbé and the villages of Kangourou, Bérékossou and Firou in the commune of Kèrou. A total sample of 240 cotton-producing farm heads were interviewed, 120 from each commune. Data was collected using a monographic sheet and a modular questionnaire. Data from primary sources were supplemented by climate data for the period 2011-2017 from Météo Benin. These data were analysed using the following vulnerability indices and measures:

The Farm Household Livelihood Vulnerability Index has seven main components: i) socio-demographic profile; ii) social capital; iii) household health; iv) food; vi) livelihoods; and vii) water and natural disasters and climate variability. Each component is made up of indicators or sub-components (see Table 1 ). The Livelihood Vulnerability Index is based on Sen’s capability approach, developed by [20] and further developed by [21] and [22] . Each sub-component is measured on different levels, requiring its normalization into an index. The normalization method used to calculate the index level lative to each sub-component is that developed by [23] ; to calculate the quality-of-life index. The following formula is used to normalize the index for each sub-component.

$Indic{e}_{Sd}=\frac{S_d-S_{\min }}{S_{\max }-S_{\min }}$ (1)

Where is one of the sub-components of the 7 components for each household; $S_{\min }$ and $S_{\max }$ are the respective minimum and maximum values for each sub-component for all households.

Having obtained the normalized index for each sub-component, it is easy to calculate the index for each component. It is defined by:

$M_d=\frac{{{\displaystyle \sum }}_{i=1}^{n}Indic{e}_{Sd}}{n}$ (2)

Where $M_d$ is the index of each component, n is the number of indicators or sub-components of the main component and ${{\displaystyle \sum }}_{i=1}^{n}Indic{e}_{Sd}$ is the sum of each index obtained from each sub-component from equation (1).

One challenge in constructing a composite index lies in the choice of weights. Weights determine the intensity with which a chosen component contributes to explaining vulnerability to household living conditions. The main weighting methods proposed in the literature include equal weights, frequency-based weights, and multivariate statistical weights, such as the principal component analysis developed by [23] - [25] or multiple correspondence analysis, regression-based weights and normative weights [26] . None of these methods has been proven to be better, and most approaches to measuring multidimensional phenomena do not guarantee appropriate methods for solving the weighting problem.

Failing that, most of the various approaches take the liberty of assigning weights to each dimension in a normative manner [27] . Attention is also drawn to the trade-offs involved in using different weighting methods and the usefulness of using robustness tests to determine the impact of the specific weight value on composite indices [28] . The most commonly used approach in multidimensional poverty measurement is equal weighting [26] . Although practical, equal weighting Is far from uncontroversial [26] [29] [30] .

The new approach to measuring vulnerability to living conditions developed by [22] , contains on average at least three indicators per component. The appropriate weighting technique is frequency weighting [31] [32] . Thus, in this study, the weighting equal to the frequency weight is used ( Table 1 ). Thus, the vulnerability index is calculated from the sum of the index of each component weighted at the frequency weight from the following equation 3:

$IVM{E}_d=\frac{{{\displaystyle \sum }}_{i=1}^7{W}_{Mi}{M}_{di}}{{{\displaystyle \sum }}_{i=1}^7{W}_{Mi}}$ (3)

This index is between 0 and 0.66 when analyzing the household’s vulnerability to living conditions. An index value close to 0 indicates that the household is less vulnerable to living conditions and an index value close to 0.66 means that the household is more vulnerable to living conditions.

A new classification of the 7 components of the index of vulnerability to living conditions has been made by [33] into the following three components: i) exposure (natural disasters and climate variability); ii) adaptive capacity (socio-demographic profile, livelihood strategy, social capital) and iii) sensitivity (food, water and health). This classification is used to assess the effects of climate change on household living conditions (IPCC, 2007).

The process for calculating the exposure ( $e_d$), adaptive capacity ( $a_d$) and sensitivity ( $s_d)$) indices is identical to that for the other components defined above. Thus, the index for each component of the new classification can be defined as follows:

$C{F}_d=\frac{{{\displaystyle \sum }}_{i=1}^7{W}_{Mi}{M}_{di}}{{{\displaystyle \sum }}_{i=1}^7{W}_{Mi}}$ (4)

The following formula is used to estimate the effect of climate change on the livelihood of the household, or even the municipality.

$IVME-IPC{C}_d=\left(e_d-a_d\right)\ast s_d$ (5)

The value of the IVME-IPCCd index is between −1 and 1. For an index value close to −1, the household is less vulnerable to climate change and for an index value close to 1, the household is more vulnerable. Table 1 presents the indicators, and the weighting thresholds for the living conditions vulnerability index.

Source: Author based on literature review summary.

The results in Table 2 show that the socio-demographic profile component has the highest contribution at 0.4392. It is followed by the climate change component with 0.4159, and the health component with 0.3255. The food component has the lowest contribution at 0.1820. The livelihood vulnerability index for each farm household in the sample is 0.320. This implies that the socio-demographic profile, climate change and household health should be a policy target for reducing vulnerability to the living conditions of farm households in Benin.

Source: field survey data.

Looking at the socio-demographic profile component, female-headed households have a significantly different contribution to vulnerability to living conditions than male-headed households, at a critical threshold of 1%. This discrimination is also observed for the food component. The female-headed household is vulnerable to the socio-demographic profile, livelihoods, food, water and climate change, while the male-headed household is vulnerable to social capital and health ( Table 3 ).

Legend: ***p < 0.01, **p < 0.05, *p < 0.1; Source: field survey data.

The contributions of the socio-demographic profile, water and climate change components to vulnerability to living conditions for the Boukombé commune farm household are 0.4607, 0.2902 and 0.4371 respectively, and statistically different from zero for the Kérou farm household (0.4179, 0.2459 and 0.3948 respectively). The fact of being a farm household in the commune of Boukombé, all other things being equal, explains the difference in the relative contribution of the socio-demographic profile, water and climate change components between the study zones. In addition, the agricultural household in the commune of Kérou is only vulnerable to livelihoods ( Table 4 ).

Legend: ***p < 0.01, **p < 0.05, *p < 0, Source: field survey data.

Table 5 shows that households headed by women are more vulnerable to living conditions than those headed by men. This breakdown by zone shows that the agricultural household in the commune of Boukombé is more vulnerable to living conditions than its counterpart in the commune of Kérou. This indicates that being the female head of household and belonging to the commune of Boukombé explain the differences in vulnerability observed.

Legend: ***p < 0.01, **p < 0.05, *p < 0.1, Source: field survey data.

The contribution of the sensitivity, adaptive capacity and exposure components to the index of the effects of climate change on living conditions increases with respective values of 0.2538, 0.3296 and 0.4159 ( Table 6 ). The average effect of climate change on the living conditions of each agricultural household is estimated at 0.0209, far from −1. The farm household is vulnerable to the effects of climate change.

Source: field survey data.

Households headed by women have a high contribution to the effects of climate change on living conditions when considering the adaptive capacity component ( Table 7 ), and are significantly different from zero at the critical threshold of 1%. This means that the household’s adaptive capacity differs depending on whether the household is headed by a woman or a man.

Legend: ***p < 0.01, **p < 0.05, *p < 0.1, Source: field survey data.

The results in table 8 show a significant difference in the exposure of agricultural households in the commune of Boukombé to climatic shocks and hazards, compared to those in the commune of Kérou. Agricultural households in the commune of Boukombé are more exposed to climatic shocks and hazards than their counterparts in the commune of Kérou.

Legend: ***p < 0.01, **p < 0.05, *p < 0.1, Source: field survey data.

The average value of the index of the effects of climate change on the living conditions of each household is 0.0209 ( Table 9 ). Households headed by women (−0.0042) are less vulnerable to the effects of climate change on their living conditions than their male counterparts (0.0254). At the municipal level, agricultural households in the commune of Boukombé are more vulnerable (0.0254) to the effects of climate change than those in the commune of Kérou (0.0164). These results are validated by the difference-in-means tests, which are significant at the 1% and 5% critical thresholds ( Table 9 ).

Legend: ***p < 0.01, **p < 0.05, *p < 0.1, Source: field survey data.

The aim of this study was to analyses vulnerability to living conditions in agricultural households in the cotton-growing areas of Benin. The results show that female heads of household are less vulnerable to social capital than their male counterparts. Indeed, the interviews showed that “women belong to self-help groups more than men”. They receive support from the various groups for certain activities such as ploughing, sowing, weeding and harvesting. This “club” effect reduces individual working time and improves the productive efficiency of both the woman and the group. The “club” effect has been an essential component of human capital since the pioneering work of Becker, and of the social organization of work since the pioneering work of Smith. This dynamism of the group would be the cause of the low vulnerability of women in terms of the contribution of social capital to the index of vulnerability to living conditions. But despite this grouping effect, the female-headed household is more vulnerable to living conditions than the male-headed household. Studies by [34] on analyzing the vulnerability of populations to climate change in northern Benin have shown that young people and women are more vulnerable to climate change. According to the Analyse globale de la vulnérabilité et de la sécurité alimentaire (AGVSA), more female-headed households (12%) are food insecure than male-headed households (9%) (Institut national de la statistique appliquée et de l’économie (INSAE) and (WFP, 2017). Most female-headed households are poor households (26%). Thus, women are fragile and respond very poorly to the negative effects of climate change [34] . According to [35] , the greater vulnerability of women than men lead to marital conflicts within families.

Female-headed households are more vulnerable to food and living conditions. The female head of household represents 25%, or approximately 10.42% of the heads of household in the sample. The female head of household has “access to marginal land” and is more involved in “market gardening” and “processing sorghum into local beverages” or “shea butter”, according to the results of the interviews. Few female heads of household grow maize or sorghum, the staple foods in the study area. Consequently, vulnerability to food and living conditions could be partly explained by access to marginal land and crop choices. Despite the crucial role played by women, 70% of the world’s undernourished people are women (FAO 2008). Women are disproportionately affected by hunger because they suffer multiple discrimination in access to food, productive resources and a decent wage [36] .

The results also showed that the main components of vulnerability to living conditions are socio-demographic profile and climate change. These results are similar to those of [37] and Hien Luong in North-West Vietnam. However, the contribution of the “socio-demographic profile” component is stronger both at the level of the female household head as well as at the level of the Boukombé commune. The contribution of the “climate change” component is stronger and significantly different from zero for the commune of Boukombé than for that of Kérou. This result was confirmed by the results of our interviews, in which 100% of the stakeholders interviewed stated that “apart from sorghum, fonio, small millet and voandzou, the Boukombé farming household can no longer produce without chemical inputs”. The soil is very poor, and “a slight change in the agricultural calendar, either due to late or poorly distributed rainfall, has a negative impact on production”, and in turn on their living conditions.

The commune of Boukombé has few watercourses and a few low-lying areas, concentrated in the arrondissement of Natta and Boukombé centre, which are used for market gardening and lowland rice. As a result, off-season crops, and even home gardens, are virtually non-existent, said the people in charge of the social promotion centre and the representative of the town hall during the direct interviews. Similarly, the high proportion of female heads of household from the commune of Boukombé in the sample, 95% of whom are uneducated compared with 70% in Kérou, would explain the vulnerability of female heads of household and those from Boukombé respectively. These results are similar to those of [38] and [31] , according to whom the better educated the female head of household is, the more likely she is to adopt strategies in terms of crop selection in order to cushion the impact of climate change and survive.

On the other hand, the commune of Kérou has relatively fertile soil, crossed by the Mékrou river and the Pendjari. The commune of Kérou also benefits from the positive externalities of the state wildlife and forest reserves of “Alibori supérieur” and “Parc Pendjari”, where most farmers in the arrondissements of Firou, Kaobagou and Brignamaro are allocated land for large-scale agricultural production. It is developing more export crops and those that require water, such as roots and tubers. The soil has a better water retention capacity than that of Boukombé. These geographical advantages could explain why the commune of Kérou is less vulnerable to climate change and food shortages. The commune of Boukombé is more vulnerable (0.0254) than that of Kérou (0.0164). This result is similar to the values of the index of the effects of climate change on the living conditions of households in southern Kombo and Nimuni in the Gambia, which gave 0.023 and 0.002 respectively in the work of [39] .

The impact of climate change on household living conditions shows that the female head of household is less vulnerable (−0.0042) than the male head of household (0.0238). Although women are more vulnerable to exposure to shocks, they are less vulnerable to the impacts of climate change because their choice of crops is better adapted to the context.

The level of vulnerability can be linked to producers’ behaviors that the index does not account for. Maize and sorghum are staple crops, mainly managed by household heads in the area. Female household heads producing these are often widows or divorced, potentially explaining their lower production. Observations show that non-widowed women cultivate rice, voandzou, and fonio as income-generating activities to support household expenses. Interviews indicate that local producers use an average of 5 bags of chemical fertilizer instead of the 3 recommended by the Interprofessional Cotton Agency (ICA) due to poor soil in Atacora, Benin. This leads to high production costs, as cotton profits absorb these expenses. Benin’s lack of targeted agricultural inputs means farmers receive 6 bags from the ICA, with 3 intended for cotton and the rest for maize. However, compliance varies; some overuse while others sell portions.

Despite the presence of professional stakeholders in the agricultural sector, such as the AIC and the State through the Territorial Agricultural Development Agency hubs, there is no known database on non-agricultural households. Furthermore, this research did not include non-agricultural households due to the complexities involved in assessing the impact of climate change on them.