3.1. Macro–Micro Analysis Framework and Data Sources
To understand the economic impacts of droughts on developing economies, several studies have been performed. Most of the existing studies are partial equilibria studies (e.g., [
54,
55]) that focus on one part of the economy and ignore the indirect effects of drought. To account for the direct and indirect effects of drought events on the economy as a whole, a limited number of studies have employed computable general equilibrium modeling ([
15,
16,
42,
50,
56,
57,
58,
59,
60]).
To assess the implications of the drought on the economy of Burkina Faso as a whole, and in particular in terms of household poverty, we use a CGE model integrated with a poverty analysis model. In the macro–micro coupling, as in [
16,
50], changes in the consumption expenditure and prices of goods and services calculated in the CGE model are fed into the microsimulation module to determine the household poverty impacts of different characteristics (e.g., gender dimension of the household head) (
Figure 2). This coupling allows for a detailed analysis of the economy-wide effects of drought at the sectoral and subnational levels, with an analysis of the poverty impacts of drought.
Therefore, the 2018 SAM is used here for the calibration of the CGE model. Thus, the CGE running with the SAM is used to implement drought simulation scenarios and adaptation options. The impacts on the household production, employment, income, prices and consumption expenditure are generated. The changes in the prices of goods and services and consumption expenditure are then used as inputs to the micro model to update the household consumption expenditure from the national household living conditions survey and the national poverty line. Finally, the new poverty indicators are calculated and compared with the baseline situation.
In order to assess the impacts of the temporal trajectories of drought scenarios on the Burkinabe economy, this study uses a dynamic CGE model based on the PEP 1-t model of [
61]. The PEP-1-t CGE model is a popular model, as it is already applied for the analysis of several policies and exogenous shocks for different countries. More recently, for example, [
62] used the CGE PEP 1-t model to assess the impact of the COVID-19 pandemic state of emergency and the government’s fiscal plan on South Africa’s economy and environment. Additionally, [
63] with the CGE PEP 1-t model assessed the impact of agricultural sector reforms in Senegal. Although the model is fully described in [
61], we present the main assumptions. The CGE model is calibrated with the 2013 Burkina Faso Agricultural Social Accounting Matrix (SAM) [
64] updated for the year 2018 with 2020 World Bank development indicator data [
65]. Related to the SAM, our model has 27 production sectors, including 9 agricultural activities and 29 products including 10 agricultural products.
The core of the constructed model is based on the neoclassical general equilibrium paradigm. Producers maximize their profit under a given technology and independent prices. Therefore, each industry’s representative producers face a nested structure of production. The production is presented in a four-level process (see
Figure 3). The
Figure 3 shows the steps in the production modelling process. It shows how factors are combined to achieve the level of output in each sector. At the first level, the production function takes a Leontief form for the value added and total intermediate inputs in the fixed share. In other words, the aggregate inputs are considered to be strictly complementary, following a Leontief production function. At the second level, the value added takes a CES (constant elasticity of substitution) from comprising composite labor and composite capital. It is after this second step that our model is different to the PEP model.
Following the social accounting matrix (SAM), labor is disaggregated between agricultural salary labor, family labor and non-farm labor. To take into account the gender dimension of Burkina Faso’s economy, the different labor input categories are subdivided by gender. Thus, salaried agricultural labor and family labor are disaggregated by sex. On the other hand, non-agricultural labor is distinguished by sex and the level of qualification. To integrate this disaggregation into our modelling, we assume that at the third level composite labor is a CES function between total agricultural labor and total non-agricultural labor, although it is not relevant to substitute farm labor for non-farm labor. At the fourth level, the total agricultural labor is a CES function between agricultural labor categories (female and male family labor and female and male agricultural salary labor). At the same level, the total non-agricultural labor is a CES function combination between non-agricultural labor categories (female and male skilled labor and female and male unskilled labor).
At the composite capital demand side, the agricultural sectors are all highly agriculturally capital-intensive, and the non-agricultural sectors are also highly non-agriculturally capital-intensive. Thus, the composite capital demand is a Leontief production function between the total agricultural capital and total non-agricultural capital, where the composite capital is directly determined by the stock of the capital (proportional relationship) and there are no substitution possibilities between the total agricultural capital and total non-agricultural capital. Then, the total agricultural capital is a CES function between the agricultural capital categories (land, agricultural equipment) and the total non-agricultural capital is a CES function between the non-agricultural capital categories.
To incorporate the effects of drought into the CGE model, we modify some parameters of the standard PEP-1-t model. The first modification concerns the introduction of the potential effects of drought into agricultural activities. Thus, for a given period t, we assume that the total factor productivity parameter of the value-added function depends on an exogenous annual growth rate, reflecting a neutral technical change in line with Hicks. This annual growth rate is also a function of a random effect when drought occurs.
The model has four different types of agents. First, households are disaggregated into rural poor households, rural rich households, urban rich households and urban poor households. In this study, we grouped the households into two categories, the rural and the urban households. Each category of household receives capital and labor income and transfers from other institutions. Households pay direct taxes to the government and spend their disposable income by consuming and saving. Household consumption, as a function of prices and income, which is allocated across different commodities, is based on an LES (linear expenditure system) demand function, which is derived from the maximization of a Stone–Geary utility function. A firm’s income consists, on the one hand, of its share of capital income, and on the other hand on the transfers received from the other agents. They pay dividends to the different institutions, pay direct taxes, and save money. The revenue for the government comes from direct taxes from households and firms and indirect taxes on activities and commodities. It makes transfers to other agents, buys commodities and saves money. Finally, the rest of the global income comes from its sales on the Burkinabe market, income from labor and capital and transfers from other institutions. It buys commodities and makes transfers to domestic institutions. The difference between the rest of the global spending and income is the current account balance.
On the supply side, the domestic production is either sold on the domestic market or outside of it. It is assumed that there exists a CES function to link the domestically produced goods that are consumed at home with exported goods. A CET (constant elasticity of transformation) function determines the scope of the choices between domestic supply and exports. On the demand side, consumers can either buy commodities that are produced domestically or imported. Their choice will be influenced by the relative prices of domestically produced and imported commodities, as well as the elasticity of the substitution between the imported and domestic commodities.
In terms of the closure rules, we assume that the nominal exchange rate is the numeraire of the model. Burkina Faso is considered to be a small country, meaning it has no influence on global prices. Thus, the global prices for all commodities are fixed. Moreover, we assume that the current account balance is fixed, and this underscores that Burkina Faso cannot borrow as much as it wants from the rest of the world. The capital is mobile across activities, representing a long-term situation where the economy has time to adjust.
As mentioned before, and although our model is inspired by the PEP1-t model, it departs radically, particularly via the introduction of labor market rigidity. Thus, the assumptions of the full employment of factors, wage flexibility and the equality of labor supply and demand for work by type of work and industry are rejected for the non-agricultural labor market. In our work, we postulate that the non-agricultural labor market is not in equilibrium and that ready-made workers remain unemployed. The concept of unemployment must be used with caution in a country like Burkina Faso, but the fact remains that available workers are unemployed, and this is what we will measure as “unemployment”. Several theories have been developed to explain wage rigidity. One study [
66] provided an excellent review of some of the theories that have been implemented in CGE models. Among the developed wage rigidity theories, three categories are worth mentioning: the search and matching theory by [
67], the efficiency wages theory by [
68] and the collective wage bargaining theory by [
69]. Like [
70], we use the most common framework found in the literature, which is the wage curve introduced by [
71,
72]. We explicitly model unemployment using a wage curve, i.e., unemployment is negatively related to the real wage rate and we use −0.1 as the wage elasticity borrowed from [
73].
The dynamics are introduced through growth in the supply of the production factors. The labor supply is assumed to grow at an exogenous population grow rate. Other variables that increase with the population growth are the current account balance, minimum consumption of commodities in the equation of the LES demands, current government expenditures, public investment by category and public sector, and finally changes in inventory. The capital stock is equal to the level in the preceding period, with less depreciation plus new investment. The allocation of new private capital between categories and industries follows a modified version of the [
74] investment demand specification and varies according to the ratio of the rental rate to the user cost of that capital.
Referring to the Food and Agricultural Organization (FAO), we have four pillars of food security: food availability, food access, utilization and stability [
75]. To analyze the impacts of drought on food security, we retain two pillars, namely the availability and access to food. Therefore, the two indicators, i.e., availability and access, are calculated directly with the CGE model. The per capita food availability index is measured by the volume of production per capita in urban and rural areas. The per capita food access index is measured by per capita food consumption in rural and urban areas.
Finally, to assess the drought impacts on poverty, we combine our CGE model with the micro model using a top-down approach. Once the simulations are run with the CGE model, the changes in income and prices are transmitted to the micro module. Poverty is measured using the traditional FGT indicators used by [
76]. The change in poverty level is calculated by comparing the poverty rate before and after the occurrence of a drought. The household consumption expenditure data are taken from the 2018/2019 Survey on Household Living Conditions from the Institute of Statistics and Demography of Burkina Faso [
77].
3.3. Simulation Scenario Assumption
In this study, two simulation scenarios are implemented. First, a reference scenario is simulated by updating the constant parameters and exogenous variables from one year to the next. To do so, the population growth rate of Burkina Faso is used to build the reference scenario. In a second step, the drought scenario is simulated and compared with the reference scenario. To define the drought scenario, we refer to what has happened over the last ten years in terms of agricultural yields. The scientific literature distinguishes three types of droughts—a precipitation deficit or meteorological drought, a negative water anomaly or hydrological drought and a soil moisture deficit or agricultural drought. In this work we focus on agricultural drought. The soil moisture is a key indicator of drought conditions, since the soil fertility depends on precipitation and evapotranspiration and also on temperature, as higher temperatures result in higher evapotranspiration. According to the RCP2.6 and RCP6.0 model projections, the average annual soil moisture at one meter in height for Burkina Faso will decrease by 2.5% by 2080 [
24]. However, uncertainties exist because not all models predict the same direction of soil moisture change.
Since 1970, Burkina Faso has experienced endemic drought; however, the droughts of 1972–1974 and 1983–1984 were the most severe. According to the RCP6.0 model, the probable range of drought exposure for the national cropland area per year increased from 0.07–3.8% in 2000 to 0.04–16% in 2080 [
24]. The very likely range increased from 0.01–12.0% in 2000 to 0.01–29.0% in 2080 [
24]. According to [
79], the probability of drought occurring within a season ranges from 5% to 40%.
To define our drought scenario, we use the standard rainfall and evapotranspiration index calculated by the Global Drought Monitor for Burkina Faso (
https://spei.csic.es/map/maps.html#months=1#month=8#year=2022 (accessed on 23 March 2022)). The index is calculated per month and per year. From this data, we calculate the annual average value of the index. These values were used to construct
Figure 4.
Figure 4 shows the evolution of the standard precipitation and evapotranspiration index for Burkina Faso. This index is widely used to quantify droughts. In addition, the index indicates the probability of losing water per cultivated area, and the zero value of the index marks the median, the negative value indicates drought conditions and the positive value shows wet conditions.
Figure 4 indicates that the country has been characterized by declining soil moisture for a long time. The red horizontal lines in
Figure 4 show the threshold levels, and when the SPEI is above the upper bound 1.5 the population is facing excess precipitation, while when the SPEI is below the lower bound −1.5 the population is facing drought shocks. The histograms show the evolution of the average SPEI over the entire country during the period 1990–2021. The green (red) scatterplots indicate the maximum (minimum) values of the SPEI recorded during the period 1990–2021 in specific areas. These scatterplots are local hot spots with SPEI values well above the mean values (the histograms), indicating that Burkina Faso has experienced severe and repeated drought shocks in certain areas of the country. Indeed, if the SPEI values lower than −1 show a light drought for the corresponding year, on the other hand SPEI values between −1 and −1.5 show a moderate drought situation and SPEI values higher than −1.5 indicate a severe drought situation.
Our drought scenarios are defined based on the changes in crop yield.
Figure 1 shows that after the year 2000, Burkina Faso has experienced annual drought episodes in several localities of the country. The most severe droughts can be observed in the 2011–2012, 2013–2014 and 2018–2019 agricultural seasons. We measure the drought-related productivity shock through the proportional change in crop yields relative to the previous year.
Figure 1 shows the interannual variations in crop yields between 2009 and 2019. With respect to the annual impacts of drought in Burkina Faso, The FAO data indicate that over the last ten years, the droughts of 2011–2012 resulted in reductions in crop yields of 13% for the 2011–2012 crop year, 15.8% for the 2013–2014 crop year and 2% for the 2018–2019 crop year. To take into account the uncertainty on the impact of drought on agricultural productivity, we use a stochastic shock model. This is because it is complex to model the effects of droughts, as they can for years be intense, moderate and mild. According to
Figure 1, each year Burkina Faso observes at least one drought episode, and due to the lack of estimated information on the probability of occurrence and the number of episodes per year, we consider a stochastic shock with a uniform probability distribution with 100 iterations. Finally, we capture the maximum, average and minimum mean impacts.
To implement our drought adaptation scenarios, we consider the occurrence of an intense drought. For example, ref. [
24] analyzed the benefits and costs of four climate change adaptation measures in Burkina Faso: integrated soil fertility management, increased irrigation capacity, the adoption of improved seeds and a climate information system. They found that the most promising adaptation strategy was integrated soil fertility management, followed by increasing the irrigation capacity and the adoption of improved seeds, then finally the popularization of the climate information system. According to a World Bank study on the climate-smart investment plan for Burkina Faso, investments in water resources and irrigation are the best investments that show the highest increases in agricultural yields, followed by organic farming and water and soil conservation techniques [
25]. Thus, in this study, we focus on the adoption of improved drought-tolerant seeds, integrated soil fertility management and irrigation development.
According to the World Bank’s report on the climate-smart investment plan for the period 2018 to 2050 in Burkina Faso, an investment of
$55 million over the period in the adoption of drought-tolerant crop varieties would increase agricultural productivity by 39%; investment of the same amount in water resources and irrigation would achieve an agricultural productivity increase of 56% in 2050; finally, the adoption of integrated soil management would allow for a productivity gain of 29% in 2050 with an investment of
$55 million [
25]. In this study, these estimates are used to implement our severe drought adaptation scenario. Finally, the results analysis is based on both short-term economic effects occurring in 2022 and long-term effects occurring in 2040. Our results analysis focuses on the effects of drought shocks on economic dynamics and the macroeconomic impacts on the outputs, employment and poverty according to household-headed gender.