1. Introduction
The cropping sector plays a central role in Burkina Faso, employing a significant portion of the labour force and providing livelihoods and income for the majority of the population. However, this sector is primarily rainfed and exposed to irregular precipitation, leading to interannual yield fluctuations that prevent farmers from ensuring stable production. Female farmers are particularly vulnerable to this risk. Compared to male farmers, they often use minimal technical equipment, have limited access to water resources or agricultural inputs and cultivate smaller, lower-quality plots, typically located in areas that are more exposed to climatic events (Sepahvand, Reference Sepahvand2019; Diendéré and Ouédraogo, Reference Diendéré and Ouédraogo2023). Aware of this challenge, Burkinabe authorities have progressively integrated gender mainstreaming into their rural development strategies and adaptation to climate change plans. However, these efforts struggle to produce significant results and certainly deserve further development, particularly given the likelihood of future changes in rainfall patterns expected with climate change in this country (Sawadogo et al., Reference Sawadogo, Neya, Semde, Awouhidia Korahiré, Combasséré, Traoré, Ouédraogo, Diasso, Abiodun, Bliefernicht and Kunstmann2024).
In this context, this study aims to examine the economy-wide gendered effects of the interannual rainfall variability experienced by the cropping sector in Burkina Faso. From this perspective, it first contributes to the economic literature that addresses the effects of climate-related events through a gender lens (see, e.g., Dev and Manalo (Reference Dev and Manalo2023) for a review). However, beyond being a simple gender study case for a climate-sensitive developing country, it secondly contributes to the literature by adopting a macroeconomic modelling perspective, rather than the commonly used microeconomic empirical approaches. To this end, it uses a macroeconomic computable general equilibrium (CGE) model, which, following studies that incorporate gender dimensions into CGE modelling (see, e.g., Fontana (Reference Fontana, Rai and Waylen2014) for a review), differentiates cropping activities and the workforce by sex, while also incorporating household home production.
As a first step, the annual rainfall pattern observed in Burkina Faso over the past decade is simulated. The results highlight its economy-wide effects and confirm the greater sensitivity of female cropping activities to rainfall variability. They also indicate the differential impacts on female and male workers, revealing the interactions between the non-market sphere of households and the market sphere of the economy during a rainfall shock. As a second step, the model is used to assess the effectiveness of coping strategies, such as promoting water management systems and encouraging the use of more water-stress-resistant crop varieties. The results indicate that these strategies could mitigate the effects of rainfall variability in the country and that targeted measures to support female agriculture could effectively reduce its specific vulnerability.
The study is structured as follows: the section 2 reviews the literature that assesses climate-related risks in developing countries through a gender lens. Section 3 presents the key features of the CGE model developed specifically for the study, details its empirical foundations and provides insight into the main characteristics of the Burkinabe economy. Sections 4 and 5 present and discuss the simulation results for the rainfall variability scenario and the coping strategy scenarios, respectively. Finally, section 6 concludes.
2. Literature review
In recent decades, within the context of climate change, the socio-economic impacts of climate-related events have received growing attention in economic studies on developing countries. One key insight from this literature is that the concept of ‘vulnerability’ occupies a central role (Preston et al., Reference Preston, Yuen and Westaway2011). From this perspective, many studies aim to identify the most vulnerable groups, i.e., individuals or communities not only at risk due to their exposure to climate events but also because of their marginalization within the social organization and their lack of access to resources. Various vulnerability criteria have been examined across studies, including wealth, education, race, age, class and health status. However, recent studies, in line with the UN’s Sustainable Development Goals (SDGs), particularly SDG5 and SDG13, increasingly highlight the specific vulnerability of women, thereby underscoring that climatic shocks are not gender-neutral (Alston, Reference Alston2014; Bob and Babugura, Reference Bob and Babugura2014; Jost et al., Reference Jost, Kyazze, Naab, Neelormi, Kinyangi, Zougmore, Aggarwal, Bhatta, Chaudhury, Tapio-Bistrom, Nelson and Kristjanson2016; Eastin, Reference Eastin2018; Nyasimi et al., Reference Nyasimi, Ayanlade, Mungai, Derkyi, Jegede, Leal Filho, Manolas, Azul, Azeiteiro and McGhie2018; Rao et al., Reference Rao, Lawson, Raditloaneng, Solomon and Angula2019; Dev and Manalo, Reference Dev and Manalo2023). Some of these studies have focused on Burkina Faso (Romero et al., Reference Romero, Belemvire and Saulière2011; Laeticia, Reference Laeticia2012; Sarr et al., Reference Sarr, Atta, Ly, Salack, Sangar, Ourback, Subsol and George2015; Sepahvand, Reference Sepahvand2019; McOmber, Reference McOmber2020; Dickin et al., Reference Dickin, Segnestam and Sou Dakoure2021; Diendéré and Ouédraogo, Reference Diendéré and Ouédraogo2023), revealing that women are more vulnerable to climate events.
While most analyses assessing the effects of climate-related events rely on microeconomic and empirical approaches, some studies use CGE simulation models (see, e.g., Wei and Aaheim (Reference Wei and Aaheim2023) for a review). Grounded in Walrasian theory, this modelling does not limit the effects of a given shock to a single sector but provides a comprehensive analysis by capturing its overall impacts through the linkages between prices, income, supply and demand in the economy. However, in the standard framework, CGE models lack gender dimensions and therefore require adaptations to be applied with a gender lens (Fontana, Reference Fontana, Rai and Waylen2014).
The first adaptation explored in the literature builds on the pioneering work of Arndt and Tarp (Reference Arndt and Tarp2000) and involves disaggregating labour, economic activities or households based on sex. Applied to the assessment of the effects of climate-related events, this approach has, for instance, been used in Latin America (Andersen et al., Reference Andersen, Breisinger, Mason-D'Croz, Jemio, Ringler, Robertson, Verner and Wiebelt2014; Estrades and Laborde, Reference Estrades and Laborde2014) and South Africa (Chitiga et al., Reference Chitiga, Maisonnave, Mabugu and Henseler2019). Recent studies have focused on Burkina Faso to evaluate the potential consequences of long-term climate change (Sawadogo and Fofana, Reference Sawadogo and Fofana2021) or severe drought events (Sawadogo, Reference Sawadogo2022).
The second adaptation aligns with the pioneering work of Fontana and Wood (Reference Fontana and Wood2000) and involves considering not only the market sector of the economy but also the non-market household sector. Home production, which refers to the services produced and consumed by households themselves (such as meal preparation, water chores and childcare), mobilizes a large proportion of the labour force in developing countries, predominantly women (OECD, 2019). Accounting for it therefore enhances the ability of CGE models to address gender issues, notably by enabling the endogenization of specific labour supplies for women and men based on time-allocation decisions within households. To our knowledge, although this approach has been used for various purposes since the 2000s, only recent studies by Escalante and Maisonnave (Reference Escalante and Maisonnave2022a, Reference Escalante and Maisonnave2022b) have adopted it to assess the gendered effects of climate shocks in Bolivia.
Building on previous literature, this study uses a gender-aware CGE model to assess the gendered effects of the interannual rainfall variability experienced by the cropping sector in Burkina Faso. Similar to Souratié et al. (Reference Souratié, Koinda, Decaluwé and Samandoulougou2019), Sawadogo and Fofana (Reference Sawadogo and Fofana2021) and Sawadogo (Reference Sawadogo2022), this model differentiates cropping activities and the labour force based on sex. Additionally, similar to Escalante and Maisonnave (Reference Escalante and Maisonnave2022a, Reference Escalante and Maisonnave2022b), it includes the home production activity of households.
3. Gender-aware CGE model
3.1. Key features of the model
Consistent with the macroeconomic focus of the study, the model adopts a relatively aggregated structure by encompassing four non-cropping activities (industry, marketed services, non-marketed services and non-cropping primary activities), four representative agents (households, government, firms and rest of the world) and three production factors (capital, female labour and male labour). However, to better align with the specific objectives of the analysis, particular emphasis is placed on disaggregating the cropping sector (as mentioned later). On this basis, the model's logic primarily follows that of the PEP-1-t dynamic model. Since the latter relies on well-established CGE modelling assumptions, as detailed by Decaluwé et al. (Reference Decaluwé, Lemelin, Robichaud and Maisonnave2013), the following paragraphs highlight only the main features that distinguish our model from this standard framework.
The first distinguishing feature is the integration of gender-related aspects into both cropping activities and the workforce. For the cropping sector, nine main crop types cultivated in the country were first identified (corn, rice, millet and sorghum, fonio, tubers, pulses, oilseeds, fruits and vegetables, and other crops) to account for the varying impacts of rainfall variability across these crops (section 4). Each of these nine categories was then further divided based on the landowner's gender, allowing for a distinction between female and male activities. Regarding the workforce, it is assumed that female and male workers belong to distinct labour markets and are imperfect substitutes in the production processes of economic activities. These processes are represented by a two-level nested CES function. At the top level, capital is combined with a composite labour factor. At the bottom level, this composite factor consists of female and male labour, with their relative demands determined by firms’ profit-maximization behaviour, depending on the elasticity of substitution and respective wages. At this point, it should be noted that, similar to Löfgren et al. (Reference Lofgren, Harris and Robinson2002), while maintaining the standard assumption of perfect mobility for female and male labour, each activity is assumed to pay activity-specific wages to each type of worker, reflecting exogenous activity-specific distortions from equilibrium wages.
The second distinguishing feature of the model builds on Fontana and Wood (Reference Fontana and Wood2000) by incorporating household home production activity alongside marketed activities. For simplicity, this home production activity is assumed to result in a single good and to rely exclusively on labour, without capital or other inputs. Its process is modelled using a CES function, where the respective use of women and men workers is driven by their degree of substitutability and their opportunity costs, which are the respective average wage they can expect in the labour market. It is further assumed that the home-produced good is entirely consumed by the household itself. Compared to the standard model, this consumption now enters the household utility function, alongside the consumption of market goods and the ‘consumption’ of leisure time for both female and male workers. Using the common Stone–Geary specification for this utility function, the household's maximization programme results in linear expenditure system (LES) demand functions for the four types of goods (see, e.g., Fofana et al. (Reference Fofana, Cockburn and Decaluwé2003) for details). Finally, it is assumed that after accounting for time spent on basic needs (such as sleeping, eating and personal care) and minimum leisure, female and male workers have a fixed amount of available time, which is allocated exclusively to paid work, leisure and housework. Within this framework, the overall labour supply to the marketed economy for each type of worker is not exogenous, as in the standard model. Instead, it is endogenously determined (and expressed in hours) based on time-allocation decisions within households. In the labour market, this supply meets the demand from economic activities, with a flexible wage ensuring full-employment equilibrium.
With this extended framework, the general equilibrium of the economy for a given period is characterized by the system of prices, capital rents and wages, as well as workers’ time allocations that simultaneously balance all markets in the economy and ensure the equality of supply and demand for home-produced goods within households. Similar to the PEP-1-t model, the between-period modelling specifications are set with a dynamic recursive approach. Exogenous volume variables and some key parameters grow at a constant rate for each period, and an accumulation rule is applied to the capital stock of economic activities through an investment function, which is determined in the preceding period (Decaluwé et al., Reference Decaluwé, Lemelin, Robichaud and Maisonnave2013).
3.2. Data
3.2.1. Gendered social accounting matrix
Since the aim of this study is to examine the effects of rainfall variability recorded in Burkina Faso during the 2013–2022 period (section 4), we defined the model's initial equilibrium using an aggregated version of the Ministry of Agriculture's 2013 social accounting matrix (SAM) as a starting point. Compared to other SAMs developed for the Burkinabe economy (e.g., the 2019 SAM by the United Nations Development Programme), one key advantage of this 2013 SAM is its explicit focus on the agricultural sector. Following the methodology of Souratié et al. (Reference Souratié, Koinda, Decaluwé and Samandoulougou2019), we engendered this initial SAM by incorporating gender-based distinctions in cropping activities and labour, using microeconomic data from the 2013 Permanent Agricultural Survey and the 2014 Continuous Multisectoral Survey. Additionally, since the study requires an assessment of the time women and men dedicate to work, housework and leisure, as well as the volume of home production, we also referred to INSD/OCDE (2020). Based on the 2018 Time-Use Survey, it indicates that women spend, on average, 5.9 times as much time on housework as men. Conversely, men spend, on average, 1.4 times as much time working in the economy. Finally, consistent with the model specifications, we accounted for wage differences across activities and worker types by referring to Donald et al. (Reference Donald, Tonmoy-Islam and Robakowski-Van Stralen2024), which reports an average gender labour income gap of 18 per cent in favour of men in the country, and to Sepahvand (Reference Sepahvand2019), which indicates that female cropping activities are 17 per cent less productive than male activities.
Most of the model's parameters are calibrated using this gendered SAM, following standard procedures in CGE modelling. For the specific cases of LES demand functions and labour supply functions, we used the De Melo and Tarr approach, as detailed in Fofana et al. (Reference Fofana, Cockburn and Decaluwé2003). In the absence of available data, we adopted the same value as in that study for the share of minimum subsistence requirement for home commodity consumption (30 per cent) and set the elasticity of labour supply with respect to income at −0.1. For the household home production function, we set the elasticity of substitution between female and male labour at 0.5 to reflect the rigidity of gender roles within households.
3.2.2. Characteristics of the Burkinabe economy
Key macroeconomic data of the Burkinabe economy can be extracted from the SAM (see online appendix 1). Regarding households, the data show that they derive the majority of their income from primary sources (33 per cent from male labour, 19 per cent from female labour and 40 per cent from capital) and that the structure of their consumption expenditure underscores the importance of cropping products (20 per cent), which primarily meet their food needs. Regarding economic activities, the data confirm that this economy is dominated by the primary sector, which accounts for 46 per cent of the GDP and 80 per cent of export revenues. Cropping activities play a significant role within this sector, generating 22 per cent of the national GDP and 25 per cent of export revenues. Being particularly labour-intensive compared to other activities, they contribute to 40 per cent of total male labour income and 63 per cent of total female labour income. Male activities are overrepresented in this cropping sector, generating nearly three-quarters of its value added and factor income. Among both female and male activities, oilseeds and millet and sorghum activities hold the largest shares of value added (33 and 24 per cent for male activities, respectively, and 27 and 24 per cent for female activities, respectively). The data also show that cropping products are primarily destined for domestic markets, except oilseeds, the country's main cash crop, whose exports account for 22 per cent of national export revenues. At this point, it is worth noting that other primary products (mainly mining products) remain the country's main exports, representing 55 per cent of total export revenues.
4. Simulation of the effects of rainfall variability in Burkina Faso
4.1. Definition of the scenario
Defining a scenario of rainfall variability (ScVar) to be simulated with the model requires specifying the impacts of precipitation on the cropping sector. At this level, consistent with most CGE studies evaluating climate effects on the agricultural sector, rainfall is considered an implicit production factor for each cropping activity. In this study, it is assumed that when it deviates from its mean level, it either positively or negatively affects the total factor productivity of the activity (Equation (1)). Accordingly, the scale parameter (
$B_{j,t}^{VA}$) of its production function is treated as an exogenous variable, varying over time due to exogenous rainfall shocks (
$Rai{n_t}$), with an elasticity parameter (
${\varepsilon _j}$) that was initially estimated (see online appendix 2). At this point, it should be noted that, for each cropping activity, positive rainfall deviations (‘good’ years) are assumed to benefit all types of land, whether irrigated or non-irrigated. However, negative rainfall deviations (‘bad’ years) are assumed to affect only the non-irrigated portion of the total land used, specific to each crop (see online appendix 2). In other words, for simplicity, it is assumed that rainfall deficits do not impact water reserves for irrigation or that irrigation systems are designed to function independently of direct rainfall:
\begin{equation}B_{j,t}^{VA} = \left( {1 + {\varepsilon _j}.\frac{{Rai{n_t} - Rai{n_{mean}}}}{{Rai{n_{mean}}}}} \right).B_{j,to}^{VA}\,\,\,\,\,\,\,\,\forall j\, \in cropping\,activities\end{equation}In the dynamic perspective of the model, defining the ScVar scenario also requires specifying a sequence of time periods characterized by annual deviations of rainfall from their mean level. For this purpose, we used precipitation data observed in the country from 2003 to 2022 to define the mean level over this period and selected the last 10 years (2013–2022) as the reference period for the simulation (figure 1).
Figure 1. Rainfall deviations from the mean level in Burkina Faso (2003–2022).
4.2. Simulation results
Tables 1 and 2 shows the results from the simulation of the ScVar scenario, covering the whole 10-year period as well as two sub-periods corresponding to ‘good’ and ‘bad’ years. As is common in dynamic CGE modelling, these results are presented from a counterfactual perspective and compared to a ‘Business as Usual’ scenario (ScBAU). The latter is defined as a balanced growth path for the Burkinabe economy, where exogenous volume variables and key parameters of the model evolve according to the annual population growth rate of 3 per cent observed in the country, and rainfall is supposed to be at its mean level for each period. On these bases, the average (
$A{v_{dev}}$) or maximum deviations (
$Ma{x_{dev}}$) are calculated from the relative differences (in percentage) found each year between the values obtained in the ScVar and ScBAU scenarios. All other things being equal, they therefore allow for distinguishing the effects of rainfall variations over the period compared to the ScBAU scenario. The indicator ∆CV is the absolute difference in the coefficient of variation observed between ScVar and ScBAU and thus can be considered as the part of variability attributable to rainfall in the simulation.
Table 1. Selected economy-wide effects of rainfall variability in Burkina Faso over the 10-year simulation period
Source: Authors, based on the simulation results.
a Average relative deviation from ScBAU over the period.
b Average absolute deviation from ScBAU over the period.
Table 2. Selected gendered effects of rainfall variability in Burkina Faso over the 10-year simulation period
Source: Authors, based on the simulation results.
a Average relative deviation from ScBAU over the period.
b Average absolute deviation from ScBAU over the period.
c Gender wage gap = 
${w_{Male}} - {w_{Female}}$; gender participation gap = 
$L_{Male}^S - L_{Female}^S$.
4.2.1. Economy-wide effects of rainfall variability
Selected economy-wide effects of rainfall variability are presented in table 1. According to the general equilibrium logic of the model, these effects result from the interplay of several mechanisms, including changes in market balances in the economy, which lead to variations in prices, wages and incomes for agents, as well as shifts in time allocation for workers within households.
Bearing this logic in mind, the first insight from the simulation is that cropping activities are logically particularly affected by rainfall shocks. In the scenario, rainfall variations act as exogenous shocks that impact their productivity either negatively (‘bad’ years) or positively (‘good’ years). For a given year, all other things being equal, this direct shock compels farmers to adjust their demand for labour. In bad years, the decline in productivity necessitates employing additional workers to maintain production levels and meet product demand. In good years, the opposite effects are at play. Such effects are observed in the simulation. Total labour use in the cropping sector deviates on average by +0.71 per cent during bad years and by −0.82 per cent during good years, compared to the ScBAU scenario. However, in bad years, increasing labour fails to fully compensate for the productivity decline, leading to an annual negative average production deviation of −1.32 per cent. In contrast, during good years, production sees a positive deviation of +1.72 per cent. In this context, a significant portion of rainfall shocks is transmitted through price changes, which deviate on average by +2.53 per cent during bad years and −3.03 per cent during good years.
The second insight from the simulation is that rainfall shocks affecting cropping activities also transmit to other sectors, although this indirect effect is logically lower. For example, in bad years, non-cropping activities experience average deviations from the ScBAU scenario in production and prices of –0.22 and –0.16 per cent, respectively. In good years, these deviations are +0.27 and +0.09 per cent, respectively. At this level, multiple intersectoral transmission channels come into play. These include, for instance, labour mobility across sectors, backward and forward linkages through intermediate consumption (e.g., with the food industry) and changes in the structure of final demand for products, driven by variations in agents’ income and relative prices. In this context, the entire economy is affected by rainfall variability. During bad years, GDP and the price index deviate on average from the ScBAU scenario by −0.48 and +0.45 per cent, respectively, thereby confirming the results obtained in other CGE studies focusing on the adverse impacts of climate events.
Good years logically contribute to opposite deviations, with GDP and the price index deviating by +0.60 and −0.61 per cent, respectively. At this point, it is worth noting that, over the 10-year period, the positive effects of good years (which are more frequent in the ScVar scenario) offset the negative effects of bad years, resulting in a net positive outcome for the economy. Rainfall fluctuations also logically amplify the variability in most of the indicators. This effect is particularly pronounced for crop prices, confirming that a significant portion of a shock is transmitted through this channel. Figure 2 shows, for instance, period by period, the annual deviations observed between the ScVar and ScBAU scenarios for prices and production of cropping activities.
Figure 2. Effects of rainfall variability on cropping activities over the 10-year simulation period (annual deviation between the ScVar and ScBAU).
The third insight from the simulation is that household well-being logically deteriorates in bad years and improves during good ones. During bad years, households’ real income deviates on average negatively by −0.32 per cent from the ScBAU scenario. Conversely, during good years, this deviation is +0.41 per cent. Household consumption of market goods, leisure and domestic goods is also affected, deviating negatively during bad years by −0.37, −0.21 and −0.26 per cent, respectively, and positively during good years by +0.51, +0.24 and +0.29 per cent, respectively. Additional food security indicators complete this well-being analysis. Food availability is approximated by the per capita volume of crop products sold on the national market. Food access is approximated by households’ real crop consumption. Food vulnerability is approximated by the ratio of the volume of imported crops to households’ real crop consumption. On this basis, results show that bad years logically degrade households’ food security, with average deviations from the ScBAU of −1.90, −2.44 and +3.72 per cent for availability, access and vulnerability, respectively. In good years, these deviations are +2.42, +3.13 and −4.66 per cent, respectively.
4.2.2. Gendered effects of rainfall variability
Given the specific focus of the study, table 2 details the gendered effects of rainfall variability. The first insight from these results is that, by changing labour demand across various activities, the overall labour supplied by each type of worker to the marketed economy and, therefore, the equilibrium wage that balances each segment of the labour market, rainfall shocks affect the wage gap and participation gap initially observed between men and women in the labour market. Compared to the ScBAU scenario, the gap between the average wage of male and female workers increases by +1.13 per cent on average during bad years, while it decreases by −1.28 per cent on average during good years. In contrast, the gap between the overall volume of labour provided by male and female workers in the labour market decreases by −5.55 per cent on average during bad years, while it increases by +6.22 per cent on average during good years.
The second insight is that rainfall shocks have differential impacts on cropping activities, with female activities being more sensitive than male activities. During bad years, female and male crop productions deviate on average negatively from the ScBAU scenario by −1.40 and −1.29 per cent, respectively. In contrast, during good years, these deviations are +1.82 and +1.69 per cent, respectively. At the same time, prices deviate positively by +2.70 per cent for female activities and +2.47 per cent for male activities. During good years, these deviations are −3.22 and −2.96 per cent, respectively. Differences are also observed in the use of labour. During bad years, compared to the ScBAU scenario, female activities increase their demand for women and men workers by +0.85 and +0.65 per cent, respectively. For male activities, the increases are +0.75 and +0.62 per cent, respectively. In good years, the deviations are −0.97 and −0.77 per cent for female activities and −0.84 and −0.73 per cent for male activities. At this level, the respective demands for female or male labour in each type of activity are determined by the nature of production processes, the degree of substitutability between labour types and their relative wages.
The third insight is that rainfall shocks have a significant and differential impact on men's and women's allocations of time within households. For instance, in bad years, the time women devote to paid labour deviates from the ScBAU scenario by +0.63 per cent, whereas in good years it deviates by −0.71 per cent. At the same time, men's time devoted to paid labour deviates by +0.20 per cent in bad years and −0.22 per cent in good years. Opposite effects are observed for the time devoted to housework and leisure. For housework, women experience deviations of −0.23 per cent in bad years and +0.26 per cent in good years, while men experience corresponding deviations of −0.41 per cent and +0.45 per cent. For leisure, the deviations are −0.12 and +0.14 per cent for women, and −0.31 and +0.34 per cent for men in bad and good years, respectively.
Once again, these observed effects result from the interaction of numerous, and sometimes contradictory, mechanisms. For instance, all other things being equal, as nominal wages represent an opportunity cost for leisure time and housework time, increases (decreases) generate positive (negative) substitution effects on paid labour supplies for the economy. These effects may differ for women and men, depending on the nature of their respective supply functions. However, all other things being equal, as nominal wages are also a component of households’ income, increases (decreases) also generate negative (positive) income effects on labour supplies, through the increase (decrease) in the consumption of normal goods, including leisure and home-produced goods. In addition, all other things being equal, changes in the prices of market products also indirectly influence the allocation of workers’ time within the household. For example, since leisure, domestic goods and market goods are net substitutes in the household utility function, increases (decreases) in prices contribute to an increase (decrease) in the demand for leisure and domestic goods, which in turn leads to a decrease (increase) in workers’ labour supplies. Finally, the respective allocations of time for each type of workers are also indirectly determined by the characteristics of the home production process within households, which reflect gender inequalities and role rigidities, such as the higher use of female labour and a low level of substitutability between female and male workers.
5. Options for coping with rainfall variability
5.1. The ‘gender-climate nexus’ in Burkinabe agricultural policy
Although the agricultural sector's share of total public spending declined over the last decade from 11.5 per cent in 2011 to 9.8 per cent in 2020 (World Bank, 2023), this sector remains a priority for the government, as it is recognized as a key lever for growth, food security and poverty reduction. From this perspective, the current Burkinabe agricultural policy is defined in the 2016–2025 Rural Development Strategy (2016–2025 RDS), which outlines two objectives particularly relevant to our analysis (GBF, 2015).
The first one complements those of the National Adaptation Plan to Climate Change (MEEA, 2024) and focuses on promoting climate-smart practices to enhance agricultural productivity and strengthen farmers’ resilience to climate change and variability. Several initiatives have been launched in this regard in the country at both local and national levels. They include, for instance, the development of drought-resistant seed varieties, the expansion of water management systems and the promotion of soil fertility improvement techniques (such as zaï techniques and stone cordons).
The second objective complements those of the National Gender Policy (MPF, 2019) and focuses on reducing gender inequalities in rural areas by promoting the empowerment of female farmers. In this respect, various initiatives aim to secure their land ownership rights, prioritize them in the allocation of newly irrigated land, enhance their access to inputs such as seeds and fertilizers, promote innovative technologies on the plots they cultivate and develop tailored financial schemes through microfinance or dedicated funds. Other initiatives focus on facilitating their participation in agricultural value chains and markets by promoting women's cooperatives, supporting the processing and marketing of their agricultural products and strengthening their capacities in agricultural techniques and farm management.
Over the past decade, as the intersection of climate and gender policies has been increasingly recognized (FAO, 2018), these two objectives have been progressively integrated into a unified approach. From this perspective, female farmers are viewed not only as a group to be empowered due to their particular vulnerability but also as key actors possessing critical knowledge and skills that can help develop context-specific adaptation solutions to adverse climate conditions. As an illustration of this ‘Gender-Climate nexus’, the country recently committed to the Gender Transformative Mechanism initiative led by the International Fund for Agricultural Development to explicitly strengthen both women's resilience to climate change and their role in adaptation efforts (IFAD, 2024).
5.2. Water-management strategy
Currently, only 140,000 hectares (ha) of land are water-managed in Burkina Faso, while significant potential remains for developing irrigable lands (235,000 ha) and lowlands (500,000 ha), along with ample reserves of easily mobilizable water (DGESS/MAAH, 2021). In this context, the first strategy that we consider for coping with rainfall variability aligns with a key objective of the 2016–2025 RDS, which is to improve agricultural water management systems in the country.
At this point, it should be noted that, in defining the scenario for this strategy, we chose not to account for its financial aspects. Indeed, at the microscale, the characteristics of each water management project can vary widely (Evans et al., Reference Evans, Giordano and Clayton2012) depending on its level of implementation (partial or full), its scale (ranging from large-scale infrastructure such as dams, reservoirs and water transport systems to small-scale solutions) and the type of lands where it is deployed (e.g., lowlands, wetlands and floodplains). They also depend on various technical aspects regarding water collection (e.g., wells and boreholes), extraction (e.g., manual or motorized pumps) and distribution (e.g., gravity-fed systems, drip irrigation and Californian systems). As a result, as illustrated by the World Bank (2023), the cost per hectare of water management projects in the country varies widely, ranging from 0.7 million to 10 million FCFA (African Financial Community Franc). In this context, given the macroscale focus of the study, considering the financial aspects of a nationwide water management strategy is challenging without specific information on the locations of the projects or the technical solutions adopted. Keeping this point in mind, the Water Management scenario (ScWM) is defined based on the following hypotheses.
First, it is assumed that, over the 10-year simulation period, newly water-managed lands progressively replace lands that are initially non-irrigated. To adopt realistic objectives, the 2016–2025 RDS, which proposes an annual expansion of 2,500 ha of irrigated land in the country, is taken as a reference. This expansion is considered only for cropping activities that already benefit from irrigation (corn, rice, fruits and vegetables and other crops) and, for simplicity, is supposed to be distributed among these activities in proportion to the non-irrigated land area they use in the initial period.
Second, for each of these activities, two alternative hypotheses regarding the distribution of the newly water-managed lands between male and female activities are considered. The first hypothesis (HWM1) aligns with the 2016–2025 RDS, which proposes allocating 70 per cent of these lands to male activities and 30 per cent to female activities. The second hypothesis (HWM2) assumes an equal division between the two groups. Third, the same sequence of rainfall variations as in the ScVar scenario is considered over the 10-year period. However, their impacts differ from those in the previous simulation, as the increase in the share of water-managed lands in the total land used is assumed to generate two distinct effects on the selected cropping activities. On the one hand, by enabling production outside the rainy season, it contributes to higher average yields, regardless of rainfall conditions (good year or bad year). At this level, based on AQUASTAT data, the yield difference between irrigated and non-irrigated land is conservatively set at +10 per cent. On the other hand, since only non-irrigated land is assumed to be affected during years of rainfall deficit, the expansion of water-managed lands helps mitigate the intensity of ‘bad year’ shocks when they occur.
The first four columns of table 3 present selected results from the simulation of the ScWM scenario. They are expressed as the percentage change in average rainfall effects compared to the ScVar scenario. One key insight from these results is that the strategy appears to be effective for coping with rainfall variability, regardless of how the newly water-managed lands are allocated between female and male cropping activities. For example, under the HWM1 hypothesis, since improving water management is expected to increase yields, regardless of rainfall conditions, the increase in production that was previously observed during good years without the strategy would be higher by +13.6 per cent for female activities and +25.1 per cent for male activities. Price declines would also be higher by +13.2 and +28.3 per cent, respectively. By additionally decreasing the share of lands affected by rainfall deficit, the strategy would also help mitigate the adverse effects of bad years. For instance, under the HWM1 hypothesis, the decrease in female and male production previously observed without the strategy would be lessened by −15.7 and −29.4 per cent, respectively, and the increases in prices would be reduced by −16.1 and −34.3 per cent, respectively. The second insight of the simulation is that a more favourable redistribution of lands for female cropping activities would logically contribute to a greater reduction in their specific vulnerability. For instance, under the HWM2 hypothesis, the negative effects of bad years on their production and prices would decrease by −19.3 and −22.0 per cent, respectively, compared to −15.7 and −16.1 per cent under the HWM1 hypothesis.
Table 3. Changes in gendered effects of rainfall variability in Burkina Faso in the presence of coping strategies over the 10-year simulation period (reduction or augmentation of the effect compared to the same scenario without the strategy)
Source: Authors, based on the simulation results.
5.3. Improved seeds strategy
The use of short-cycle crop varieties, known for their tolerance to water stress situations, is often suggested as a possible adaptation strategy in response to rainfall deficits (CGIAR, Reference Walker and Alwang2015). Following the major drought of the 1970s, Burkina Faso has long been engaged in the production of such varieties and in their diffusion through, for instance, awareness campaigns for farmers and subsidy programmes. However, despite these efforts, the national average adoption rate of improved seeds by farmers remains relatively low in the country (except for cotton, where it reaches nearly 95 per cent) and has even declined, from 19.3 per cent in 2014 to 15.4 per cent in 2020 (MAAH, 2021). Several factors can explain this situation (Holtzman et al., Reference Holtzman, Kaboré, Tassembedo and Adomayakpor2013; Sawadogo-Compaoré et al., Reference Sawadogo-Compaoré, Compaoré and Yila2022). These include economic barriers limiting farmers’ access to improved seeds, such as high prices despite subsidies, irregular or insufficient availability of seeds and storage difficulties. They also include institutional factors, such as the inefficient organization of the seed sector, which involves numerous actors (the state, research institutes, private producers and distributors, farmer communities and international partners) and suffers from a lack of coordination. Additionally, they include socio-cultural factors, such as the resistance of farmers who are accustomed to using traditional seeds they have selected over decades and may be reluctant to adopt improved seeds produced by the agroindustry.
In this context, the second strategy that we consider for coping with rainfall variability aligns with a second key objective of the 2016–2025 RDS, which is to increase farmers’ use of short-cycle crop varieties, known for their tolerance to water stress situations. From this perspective, the improved seeds scenario (ScIS) assumes an increase in the national adoption rate of improved seeds for cropping activities (except oilseeds) over the 10-year simulation period. Similar to the previous scenario, two alternative hypotheses are considered. In the first one (HIS1), male and female activities are supposed to progressively achieve a rate of adoption of improved seeds of 100 and 70 per cent, respectively, at the end of the 10-year period. In the second one (HIS2), these rates are 100 per cent for both activities. For each activity, it is assumed that, while the strategy is neutral in good years, it helps reduce the negative impact of bad years on yields, in proportion to the percentage of farmers using improved seeds and the level of protection they provide. On this last point, CGIAR (Reference Walker and Alwang2015) notes that they might reduce the effects of low rainfall by 30 to 50 per cent under ideal production conditions, and by less than 10 per cent under the most adverse conditions. Taking a conservative approach, we consider a 20 per cent reduction for all crops. Finally, it is worth noting that, due to the variety of actors and measures involved in implementing such a strategy, as well as the lack of specific data documented in the academic literature or government reports to precisely quantify its actual costs, its financial aspects are, once again, not considered in the scenario.
On these bases, the last two columns of table 3 show the change in the gendered effects observed in the ScVar scenario during bad years as a result of the strategy. They are of a similar nature to those obtained for the water management strategy, with, however, relatively weaker protective effects.
6. Conclusion
By using a CGE model that differentiates cropping activities and the workforce by sex and includes the home production of households, this study investigates the impacts of rainfall variability in Burkina Faso from both a macroeconomic perspective and a gender lens. As emphasized by other studies on similar climate-sensitive countries, it confirms the greater sensitivity of female agriculture to rainfall variability, highlights the differential impacts on female and male workers in the labour market and within the household sphere and underscores the opportunity for implementing coping strategies in the country. However, while these findings highlight the importance of integrating gender dimensions in assessing the effects of climate-related events and designing adaptation policies, they must be interpreted with caution due to potential avenues for improvement.
One line of research could involve examining the modelling specifications adopted in the study. The CGE model we use follows the standard framework fairly closely, with the notable exception of incorporating household home production. Representative agents exhibit maximization behaviour in both the market and home sectors, workers are perfectly mobile, prices and wages adjust instantaneously to ensure balance across different markets and so forth. Nevertheless, the reality in Burkina Faso is likely more complex, and other elements that depart from the standard framework, for instance, rigidities in intersectoral migrations, unemployment, delays or constraints in farmers’ reactions, should also be considered to better reflect this complexity.
A second line of research could involve a deeper analysis of the situation of Burkinabe women. Although such an analysis is severely limited by data availability, disaggregating households by the gender of the household head could provide a more nuanced understanding of how gender dynamics within these households interact with rainfall variability and how vulnerability differs among them. In the same spirit, coupling the CGE model with a microsimulation model could also be relevant for moving beyond the representative agent hypothesis that underpins CGE modelling. This would enable the capture of decision-making processes at the individual level, contributing to a more dynamic and interconnected representation of the economy from the macro level to the micro level.
A third line of research could explore regional differences within the country. At this level, a multi-regional model disaggregated by climatic zones (Sahelian in the North, Sudanese-Sahelian in the Centre and Sudanian in the South) would certainly be relevant. By accounting for differences in rainfall patterns, agricultural practices and economic structures, this could enhance the understanding of how these different regions are affected by rainfall variability and how they may respond differently to various policy interventions. This could help identify region-specific vulnerabilities and strengths, thereby supporting more tailored and context-specific policy recommendations.
Finally, the adaptation strategies also warrant further investigation. On the one hand, there are certainly other strategies that could be considered. On the other hand, although the funding aspects of these strategies are difficult to evaluate precisely due to the wide range of measures they encompass, they should still be addressed. This could enable a cost–benefit analysis of the policy, which is a key issue in the context of the government's limited budgetary room to manoeuvre. Furthermore, public investments in agriculture in general, or in female activities in particular, certainly generate macroeconomic crowding-out effects on private investments, the consequences of which should also be examined in a general equilibrium framework.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1355770X25000051.
Funding statement
This study received technical and financial assistance from the Partnership for Economic Policies (PEP) (www.pep-net.org), funded by the UK Department for International Development (DFID) (UK Aid), and the Government of Canada through the International Development Research Center.
Competing interests
The authors declare none.
Open access
