The rural livelihoods framework describes how households in rural or agricultural economies must rely on a diverse set of activities to survive, such as farming, herding or off-farm employment (de Sherbinin et al., 2008; Ellis, 1998, 2000). Household members mobilise their available assets to engage in these activities. These assets can take many forms, including natural capital (local environment, e.g., water, forest, soil), social capital (e.g., networks, groups), human capital (e.g., education, local knowledge, people, workers), physical capital (e.g., land, livestock) and financial capital (e.g., cash, available credit). Households typically have a combination of these assets at their disposal, which may be ever-shifting. Investment in one area may require the depletion or rebalancing of others. In West Africa and other rural areas, the main forms of capital households rely on are weather-dependent (Goedde et al., 2019). In such a system, weather pattern change can disrupt natural capital, requiring households to re-balance or prioritise other types of capital to support themselves and to plan for the future. These efforts to stabilise household resources can take different forms, some of which may involve generating human capital (e.g., fertility responses), and may also be informed by cultural expectations about children, family and gender roles.

Across West Africa, child labour serves as a means for increased household production and to instruct children in the livelihood skills they need to support themselves in the future (Busquet et al., 2021). Thus, performing labour is considered a vital part of a child’s social integration into society (Adonteng-Kissi, 2018; Busquet et al., 2021; Twum-Danso, 2009). While childrearing is costly in the short term, especially as the costs of supporting children rise, children participate in household livelihood tasks from an early age. Parents gradually scale up their children’s livelihood tasks based on age and ability (Busquet et al., 2021). By age five, children can usually assist adults in tending crops (Bass, 2004). By age 12 for girls and by age 14 for boys, children are considered equal to adults in terms of their ability to work (ibid). Over the long term, children are expected to provide financial security and to care for their parents in old age (Stecklov, 1997).

Beyond having productive value, children bring social and affective benefits to their parents. Marriage and childbearing are nearly universal in sub-Saharan Africa, and especially in the Sahel (Shapiro and Gebreselassie, 2014). Kinship relations are strong in West Africa; marriage, or the bringing together of kin groups, and childbearing, or the continuation of one’s lineage, are important milestones (Bawah et al., 1999; Caldwell and Caldwell, 1987; Smith, 2020). Children provide socio-cultural benefits to their parents by ensuring the “competent masculinity” of the father (Smith, 2020, p. 107) or by solidifying the social position and relationships of the mother (Bledsoe and Pison, 1994; Hollos et al., 2009). Young people in West Africa typically feel social pressure to have at least a few children; a “few” may be at least four children in some parts of the region (Smith, 2004, 2020).

A person’s prospective fertility ideal – i.e., their ideal number of children before they have begun childbearing – is likely shaped in part by how they view their future, and generally reflects the prevailing social norms about family and children. Voluntary childlessness in this region is rare, at less than 0.02 per cent (Verkroost and Monden, 2022). Thus, given the productive, affective and social benefits of having children in this region, young people in West Africa are likely to consider having at least a few children ideal. The exact number tends to reflect the number of children they expect to need to ensure household survival and to support them in their old age, as well as how many children they can support. In the context of climate change, ideal fertility also tends to be reflective of anticipated coping strategies (Brauner-Otto, 2014). Changes in household resources may change perceptions about children and their associated costs and benefits (Thiede, 2022). Children often contribute to the household by completing tasks that may become more time-consuming after climate change impacts the land. For this reason, children’s labour may be particularly valuable during periods of natural capital resource strain (Beegle et al., 2006). Children’s labour frees adults, and particularly women, from doing such work (Cain, 1977). Thus, changing weather patterns may influence how young people envision their family size, suggesting that they will need a larger number of children than they would under average conditions.

However, it is also plausible that environmental and household resource stress will reduce the demand for children and will therefore encourage a smaller ideal family size. If the household cannot support as many children, or if the degradation of environmental conditions makes resources scarce, then, ceteris paribus, the value of children’s labour in collecting those scarce resources would decline (Biddlecom et al., 2005). Recent studies conducted in Nigeria and Malawi have suggested that economic uncertainty or environmental shocks can encourage young adults to delay parenthood (Smith, 2020) or postpone further childbearing (Grant and Curtis, 2022). When they have fewer or lower quality natural capital resources at their disposal, young adults might anticipate having fewer resources to rear children in the short and the long term and may therefore see having a smaller family as more advantageous. Eissler et al. (2019) found that in West Africa, the ideal family size was predicted to be lower when people were experiencing negative precipitation anomalies.

A number of scholars have explored the concept of ideal family size; its usefulness in predicting behaviour, such as contraceptive use or subsequent birth rates (Coombs, 1974; Westoff et al., 1957); and, more recently, its dynamism (Hayford, 2009; Trinitapoli and Yeatman, 2018; Yeatman et al., 2013; Yeatman and Sennott, 2014). The literature uses various terms to describe family size preferences, such as expected, desired or ideal family size. These concepts may have slightly different semantic meanings, but are all “tapping a common psychological orientation” (Ní Bhrolcháin and Beaujouan, 2019, p. 31) related to the future number of children. Fertility preferences tend to vary over the life course (Hayford, 2009), reflecting both personal and broader, community-level circumstances or experiences (Schoen et al., 1999; Sennott and Yeatman, 2012; Trinitapoli and Yeatman, 2011, 2018; Yeatman et al., 2013). People’s preferences are particularly flexible at younger ages (Iacovou and Tavares, 2011; Yeatman et al., 2013), and tend to change as family formation becomes within reach (Schoen et al., 1999; Yeatman et al., 2013). Nonetheless, people’s fertility preferences may be predictive of their behaviour in the short term (Hayford, 2009). Moreover, in this region, family size ideals are strongly correlated with total fertility rates at the population level (Bongaarts and Casterline, 2013). The prospective ideal number of children of young adults is likely to capture preferences that drive their behaviour in the near future, which coincide with major life course transitions such as marriage (DHS StatCompiler, 2023a) and, quickly thereafter, first childbearing (DHS StatCompiler, 2023b). While young adults may lack exact information about or experience with the actual costs and benefits of childrearing, there is evidence that the perceived costs of having children have a large impact on their desired fertility (Mason, 1997).

The relationship between the local environment and fertility is not universal. Depending on the context, poor environmental conditions or a negative change in access to natural resources may, for example, yield an increase in the preferred family size (Biddlecom et al., 2005; Brauner-Otto and Axinn, 2017; Eissler et al., 2019), the desire to have another child in the future (Sasson and Weinreb, 2017) or a lower likelihood of using contraception (Brauner-Otto, 2014). In the rural livelihood framework, households that invest in generating human capital by having (many) children typically do not have other types of capital at their disposal to insulate them from poor environmental conditions or some kind of negative change. In such contexts, it is often the poorest households that focus their efforts on childbearing (Sasson and Weinreb, 2017). When the land is degraded and resources are constrained, additional labour may be needed to generate sufficient crops, firewood, water or fodder. Given that land degradation may increase the demand for children who can provide labour, the demand for children should rise when environmental strain increases (Bhandari and Ghimire, 2013; Biddlecom et al., 2005; Brauner-Otto, 2014; Brauner-Otto and Axinn, 2017; Dasgupta, 1995; de Sherbinin et al., 2008; Filmer and Pritchett, 2002; Lutz and Scherbov, 2000). Childbearing, at least in these contexts, is seen as advantageous, as children’s labour can help households exploit the remaining natural capital resources.

In other contexts, a negative change in environmental conditions can encourage lower actual fertility (e.g., pregnancies or births) (Alam and Pörtner, 2018), reduce the desire to have a child (Brooks et al., 2023; Eissler et al., 2019) or reduce the ideal family size (Eissler et al., 2019). In Kenya, a poor growing season lowered the desire for children in the short term (Brooks et al., 2023); and in Tanzania, crop losses are associated with higher contraception use (Alam and Pörtner, 2018). In some contexts, this negative affect is concentrated in poorer or less endowed households. In Indonesia, for instance, temperature shocks have been linked to reduced fertility intentions among farming households and less educated women (Sellers and Gray, 2019). Thus, fertility responses are varied and context-dependent.

The effects of environmental conditions vary across regions (Eissler et al., 2019) and country boundaries (Brooks et al., 2023) within sub-Saharan Africa. There are other geographical sources of variation that may also be salient. West Africa is a climatically diverse region. Thus, considering the region as a whole obscures potential variation in the natural and physical capital resources available. Moving from north to south, the climate shifts from arid to equatorial (Köppen and Geiger, 2020). See Figure 1 for a map of the countries included in this study and the delineations of each climate zone within the countries’ boundaries.

The arid climate zone is characterised by hot, dry weather and desert and steppe landscapes. Temperatures are high all year round in these areas, with average peaks reaching 40 degrees Celsius (Tappan et al., 2016; Steel et al., 2023). The desert receives only minimal rainfall, under 200mm annually, which falls in a concentrated period lasting under two months. The steppe gets between 500 and 600mm of rain annually over a five-month period (USAID, 2018). By the end of the 21st century, an increase in the average temperature of up to six degrees Celsius is expected in this zone (Niang et al., 2014), which will lead to more erratic rainfall patterns (Lee et al., 2023), including both negative and positive changes in rainfall amounts (Niang et al., 2014). In addition to recurrent drought and expanding desertification, the arid climate zone suffers from poor soil and short growing seasons (Kpadonou et al., 2017). The staple crops here are produced for subsistence with little input and are less vulnerable to small weather deviations (OECD/SWAC, 2009).

The equatorial climate zone is nearest to the equator and includes tropical rainforests and savanna landscapes. Within this climate zone there are three sub-climate zones with distinct seasonal rainfall patterns. The varying wet seasons receive from 700 mm to over 1800 mm of rain annually (Steel et al., 2023; Tappan et al., 2016). Average temperatures are warm in the equatorial climate zone, never dropping below 18 degrees Celsius (Tappan et al., 2016). In this zone, increases in temperature are likely to be between two and four degrees Celsius by the end of the 21st century (Niang et al., 2014), which will increase the incidence of floods and droughts (Lee et al., 2023). The sustained periods of rain and the fertile soil in this climate zone allow for significant and diverse crop cultivation. However, moving south within the zone, the crops that are grown become increasingly water-intensive (OECD/SWAC, 2009) and irrigation is extremely limited (World Bank, 2020). In addition to staple food crops, cash and export crops are grown in this zone (OECD/SWAC, 2014).

These climate-specific conditions suggest that individuals living in different climate zones may interpret weather conditions and changes differently and may thus have different ideals related to family size. This association is not commonly considered in environmental and fertility research. In one notable exception, fertility timing was found to vary in response to rainfall patterns only in the dry climates of rural Mexico, and not in the humid ones (Simon, 2017).

Household resources, as they relate to the way individuals or households make their living, may also modify the relationship between weather change and fertility. Ownership of livestock or of agricultural land implies some degree of reliance on these assets for livelihood pursuits (animal husbandry or crop cultivation), and may therefore influence notions of family size based on the household labour needs associated with these different livelihoods.

There are clear empirical differences by livelihood in fertility outcomes, such as in the total fertility rate (Grace and Nagle, 2015) or the desire or intention to have a child (Sellers and Gray, 2019). For example, in Indonesia, delays in the onset of the monsoon season, which can affect rice planting, were found to be associated with an increase in the intention to have another child among non-agricultural households, but with a decrease in this intention among farming households (Sellers and Gray, 2019). This notable contrast amid similar environmental conditions is a reason to suspect that the households’ agricultural pursuits and/or the physical capital resources at their disposal are important factors in their fertility intentions. In an example from Mali based on livelihood zones,¹ the total fertility rate and the desire for another child was found to decline across the crop cultivation, agropastoralist (crop cultivation and animal herding) and pastoralist (animal herding) livelihood zones (Grace and Nagle, 2015). Although not everyone living in a given zone is engaged in that particular practice, this overall pattern suggests that livelihood type matters for fertility and its psychosocial antecedents.

Animals provide eggs, meat or dairy products, or they are beasts of burden. Animal husbandry can take several forms, which vary by season, animal type, herd size or the amount of time, labour or money available to be invested in their care; with the most intensive forms (pastoralism and transhumance) involving the seasonal movement of herds and people (Turner and Hiernaux, 2008). Young men are preferred for herding labour, which is often hired out (ibid). The fertility of pastoralists has historically been lower than that of settled cultivators (Henin, 1968, 1969; Meir, 1987; Borgerhoff Mulder, 1992; Swift, 1977). This is likely because livestock ownership requires less labour and small populations are therefore advantageous to pastoralists’ survival (Swift, 1977). As a type of physical capital, livestock are movable and easily sold, unlike land, which is fixed (Loker, 1993; Moll, 2005; Borgerhoff Mulder, 1992). While the movability of livestock may make pastoralists somewhat more resilient to local climate change than cultivators who are fixed in place, they are vulnerable to the weather changes, desertification and land use changes that restrict available pastureland over a broader area (Bekele et al., 2022).

Subsistence farming often produces low yields, with most households growing staple crops that require basic inputs on small amounts of land. Thus, food insecurity is high among subsistence farmers (Giller et al., 2021). As crop cultivation is labour-intensive and responds to increased inputs (e.g., more workers or fertiliser), cultivators tend to have higher fertility (Grace and Nagle, 2015). However, the land security hypothesis (Stokes and Schutjer, 1984) argues that actually owning the land that they farm creates economic security for cultivators, which, in turn, lowers their demand for children (Carr et al., 2006; Stokes and Schutjer, 1984), because their need for financial support in old age is somewhat reduced (Bhandari and Ghimire, 2013).

For young men and women living in sub-Saharan Africa today, conceptualising one’s future family size is somewhat fraught because of the opposing forces of “traditional” cultural pronatalism and the “modern” realities of increasingly expensive investments in children, which may feel out of reach in uncertain times (Agadjanian, 2005; Smith, 2004, 2020). Children are highly valued and bring social benefits to their parents (Agadjanian, 2005; Bledsoe, 1980; Bledsoe and Pison, 1994; Hollos et al., 2009; Smith, 2004, 2020), but the parental responsibilities of men and women differ, and may produce distinct ideals related to family size. Men feel more financial responsibility for children (Manuh, 1997; Smith, 2020), and this pressure may encourage the postponement of childbearing or a smaller desired family size among men (Smith, 2004, 2020). However, the desire to meet social expectations and to generate “wealth in people” may keep the ideal family size high for men. Women are the primary caregivers for young children, which places the burden and the reality of birthing and caring for large families on them (and their bodies) (Agadjanian, 2005). Nonetheless, childbearing confers social benefits and status on women, solidifying their marital relationship and their status with their in-laws (Bledsoe and Pison, 1994; Hollos et al., 2009).

These push and pull factors can also stem from environmental circumstances. Disruptions to natural capital, and thus to household resources, will affect men and women in different ways. Men/boys and women/girls interact with the environment in gendered ways (Nightingale, 2006), given their different social positions (Marter-Kenyon et al., 2022; Parpart et al., 2000; van Dijk and Bose, 2016). Moreover, these differences in environmental interaction patterns will vary depending on how environmental stressors influence individuals’ notions of the future, such as their prospective ideal family size and the composition of that ideal family.

Men and women have different levels of access to natural and other resources related to the gendered division of household and agricultural labour (Alesina et al., 2013; Meinzen-Dick et al., 2014) and to gendered differences in power and access to information and opportunity (Marter-Kenyon et al., 2022; Jin et al., 2015; Lambrou and Piana, 2006; Masika, 2002; Silvestri et al., 2012; Terry, 2009). Generally, land tenure rights, inheritance systems and financial capital availability tend to favour men (Fletschner and Kenney, 2014; Quisumbing et al., 2001), giving them more access to and decision-making power over land and physical capital. Thus, men are more likely than women to determine what is grown and what technologies or practices are used (Meijer et al., 2015). In some contexts in which both men and women engage in crop cultivation, they grow different crops. “Men’s crops” are generally more commercially valued, while “women’s crops” are typically used for subsistence, and are more vulnerable to environmental changes (Carr, 2008; Carr and Thompson, 2014; Orr, Homann Kee-Tui, et al., 2016; Orr, Tsusaka, et al., 2016). When working on the same fields, women tend to do more menial tasks like weeding (Orr, Homann Kee-Tui, et al., 2016; Quisumbing et al., 2015) or serve as subordinate labourers (Quisumbing et al., 2015). Women’s interactions with the environment are often related to household subsistence, such as gathering resources to care for the household and the children (Guèye, 2000; Marter-Kenyon et al., 2022). Men and women also occupy and rely on different spaces. For many of their household gathering tasks, women use commonly held land (Agarwal, 1994), which is likely to degrade faster since the users lack individual control over the land and the number of users is often high. Because these tasks are considered “supplemental”, the degradation of these resources is less of a concern (even to women) because it affects women’s time use and not men’s income (Quinn et al., 2003). However, since women’s tasks are needed for household survival, women may still be burdened by environmental change (Bohra-Mishra and Massey, 2011).

Men and women also differ in their environmental risk perceptions and priorities. Due to gendered patterns of engagement with environmental resources, men and women tend to observe different changes and prioritise different actions to mitigate adverse changes (Carr, 2008). In low-income agricultural populations in West Africa, men are more likely than women to perceive risks associated with environmental change (Cullen et al., 2018; Kristjanson et al., 2015) and to express concern about climate change (Badmos et al., 2018), likely due to the gendered division of labour and of income-generating tasks (Quinn et al., 2003) and to differential access to information about the environment (Badmos et al., 2018). Women, by contrast, are more aware than men of scarcity regarding household livelihoods and are more concerned about small-scale adaptation (Briggs et al., 2003).

These differences in environmental interactions may influence the formation of family ideals, including the ideal gender composition of families. Sub-Saharan Africans are not widely known to show sex preference for children. The health or mortality consequences of sex preference in its extreme form that are observed in some Asian contexts are not evident in sub-Saharan Africa (Basu and De Jong, 2010; Jayachandran and Pande, 2017; Kashyap and Behrman, 2020). However, several studies have found evidence of son preference in sub-Saharan African countries (Bongaarts, 2013), specifically among men (Campbell, 1991; Campbell and Campbell, 1997; Mwageni et al., 2001). But there is also evidence that while son preference exists, a relative balance in offspring is considered desirable in these countries (Mwageni et al., 2001). Drivers of son preference may stem from landholding and inheritance systems that favour males (Mwageni et al., 2001), as well as a general dominance of males in decision-making (Andriano et al., 2021). Male children and their potential labour contributions could be seen as more valuable. The wage-earning ability of sons may also mean greater old-age financial security for parents over the long term, further generating a preference for sons.

To meet the research goals of this paper, I integrated two datasets: the women’s and the men’s Demographic and Health Surveys from Burkina Faso, Mali, Nigeria and Senegal, which were fielded from 2008 through 2018;² and monthly precipitation and temperature data from Climatic Research Unit Time Series data version 4.05.

The Demographic and Health Surveys (DHS) are nationally representative household surveys completed using multistage cluster sampling. The clusters represent a village in rural areas. The DHS provide individual-level and household-level data on women aged 15–49 and men aged 15–64. The surveys are conducted about every four to five years in each country. The surveys included in this sample have geocoded data representing the cluster location. The DHS data were accessed through the Integrated Public Use Microdata Series (IPUMS) DHS dataset, which provides harmonised DHS data across time and geography (Boyle et al., 2022). I have created 10-kilometre “buffers” around each cluster location to determine the weather conditions prevailing around each cluster.³

The weather data come from Climatic Research Unit Time Series data version 4.05 (CRUTS4.05), produced by the Climatic Research Unit (CRU) at the University of East Anglia (University of East Anglia Climatic Research Unit et al., 2021). These data are remotely sensed, gridded rasters with a 0.5-degree spatial resolution (about 55 km² grid in West Africa). I use a monthly measure of near-surface temperature in degrees Celsius and monthly precipitation in centimetres for each 10-km cluster buffer to create baseline measures and measures of change.

The first step of the analysis is to establish whether or not there is a relationship between weather pattern change and the prospective ideal fertility of young, childless men and women in Sahelian West Africa, and if the effects of different types of weather vary. In a second set of models, I stratify the sample by climate zone and explore variation by physical capital. In a third set of models, I stratify the sample by gender of the respondent and examine variation by physical capital. In other gender-stratified models, I also predict the ideal number of boys and the ideal number of girls, and investigate the variation in the relationship between these outcomes and weather pattern change by physical capital.

I largely employ Poisson regression models predicting the ideal number of children because the outcome is a count variable, and the mean and the variance of the outcome variable are approximately equal. However, the men’s stratified models predicting the ideal number of boys and the ideal number of girls are negative binomial regression models because the mean is over-dispersed, and is therefore not suitable for Poisson regression. In addition to the controls, all models include birth year fixed effects to account for changing trends across birth cohorts. To account for multiple men and women residing in each DHS cluster, I cluster the standard errors.

Table 1 describes the sample of rural, young, childless men and women surveyed by the Demographic and Health Surveys programme in Senegal, Burkina Faso, Mali and Nigeria. In total, the average ideal number of children for this sample is 5.7. When broken down by the sex of the child, the average ideal number of girls is 2.3 and the average ideal number of boys is 2.9. Women make up the majority of the sample, at 71 per cent. The average male respondent is 23.3 years old and the average female respondent is 17.5 years old. A small portion (17 per cent) of the respondents are the household head or are married to the household head. Over half of the sample, 58 per cent, have a primary-level education or more. About one-quarter of the respondents report that their current occupation is in the agricultural sector. Over half of the sample, 56 per cent, live in the equatorial climate zone. In terms of household physical capital, 80 per cent of the respondents live in a household that owns agricultural land, and three-quarters live in a household that owns livestock.

The men and women in this sample were surveyed in 3,724 clusters. In the 10 kilometres surrounding each cluster, the average monthly rainfall was 9.2cm and the average temperature was 27.7 degrees Celsius during the 10-year baseline period. The range of change in temperature was large, between −2.25 and +1.25 standard deviation units, which is equal to −3.4 to 2.5 degrees Celsius. The range of change in precipitation was more limited, between −0.38 and +0.24 standard deviations units, which translates to −3.3 to 2.6 cm of rain.

The first goal of the analysis is to establish whether or not there is a relationship between changing weather patterns and the ideal number of children for young, childless men and women in Sahelian West Africa, and to examine the differences by weather type. Model 1 (Table 2) shows the results of the Poisson model predicting the ideal number of children, which indicate that while there is a relationship between weather pattern change and ideal fertility among this sample of young, childless adults, it is confined to the change in temperature. For each one standard deviation unit increase in temperature, the increase in the log of the expected count of the ideal number of children is 0.069 (p < .01). This main model also shows that being in the equatorial climate zone has a small, negative effect on the ideal number of children (−0.045 p < .001). Both physical capital measures have positive effects on the outcome: household landownership has an effect of 0.047 (p < .001) and household livestock ownership has an effect of 0.024 (p < .001). Gender is also an important predictor, as being a woman has a negative effect of −0.186 (p < .001). These results confirm that climate, physical capital and gender are salient for the formation of the ideal family size.

The top panel of Figure 3 provides the predicted count of the ideal number of children for this model; at about a z-score of zero (average temperature conditions), the predicted ideal number of children is 5.6. Where the change in temperature is negative, reaching a value of −2.25 standard deviation units, the ideal count is 4.8 children. On the opposite end, where the change in temperature is positive, reaching a z-score of 1.25, the ideal count is 6.12 children. In simpler terms, recent hotter temperatures are associated with a higher ideal number of children. The change in precipitation is not a significant predictor of the ideal number of children. In the bottom panel of Figure 3, the relatively flat line shows that the minimal changes in average precipitation do not affect the ideal number of children, which stays steady at about 5.6 children.