1. Introduction

Jobs provide the cornerstone of economic and social development by improving social well-being and by helping to boost productivity, decrease poverty and strengthen social cohesion (World Bank 2012). However, these social objectives are threatened by the increasing frequency and intensity of climate shocks (IPCC 2014). This threat has attracted increasing attention from policymakers, entrepreneurs and the scientific and international communities. Climate change is a change in the statistical average and variability of temperature, wind, humidity, cloudiness, precipitation and other quantities over a long period of time (Nordhaus 2013). It can cause climate variability, which consists of variations in climate averages and other statistics (standard deviations, extremes, etc.) across temporal and spatial scales, beyond those of individual weather events (IPCC 2014). This climate variability can lead to climate shocks, which are unpredictable weather events that result in environmental degradation.

By reducing productivity, climate shocks can affect employment (Kjellstrom, Holmer and Lemke 2009; Sudarshan et al. 2015; Zhang et al. 2018; Diallo and Atangana Ondoa 2024). Their effects can be felt in workers’ health (Kjellstrom et al. 2015; Kjellstrom et al. 2016; ILO 2019), the destruction of physical and human capital (Mueller and Quisumbing 2011) and reduced production and incomes (Dell, Jones and Olken 2012; Emran and Shilpi 2018; Adhvaryu, Kala and Nyshadham 2019; Desbureaux and Rodella 2019). The negative repercussions of these shocks on agricultural income and productivity can lead to a strategic reallocation of labour supply (Branco and Féres 2021; Colmer 2021; Josephson and Shively 2021). This reallocation may consist of an increase in working hours, the substitution of agricultural activities with non-agricultural ones, taking up a second job or migration (Rose 2001; Emran and Shilpi 2018; Minale 2018; Branco and Féres 2021). However, this depends on the capacity of workers to move between sectors and of other sectors to absorb them (Colmer 2021).

Aside from these effects on the labour market, by encouraging job creation linked to climate policies, by replacing jobs related to fossil fuels with jobs related to renewable energies and by transforming existing jobs, climate shocks can also act as a springboard for the world’s economies, particularly in Africa (ILO 2010). Countries in sub-Saharan Africa are recording a rise in natural disasters (IPCC 2014). For example, between 2000 and 2017, more than 15 per cent of natural disasters worldwide were reported in sub-Saharan Africa. For the whole African continent, climate-induced losses account for 10 to 15 per cent of gross domestic product (GDP) per capita growth for the period 1986–2015 (Baarsch et al. 2020), with recurring droughts, flooding and heatwaves (IPCC 2022). Over the past decade, the climate in Africa has been marked by extreme weather and climate events. For example, 2010 was one of the hottest years on record on the continent for the 2010–19 period. Furthermore, certain regions have experienced prolonged droughts over this period (WMO 2020a). This trend is confirmed by our data (see figure 1). The temperature spike in 2010 and the drought in 2011 were exacerbated by La Niña (WMO 2020b). This meteorological phenomenon, which results in sharp variations in temperature and more frequent droughts, began in 2010 and lasted for two years. The negative effects of these shocks could stand in the way of achieving various key Sustainable Development Goals, particularly those relating to food security and health.

Figure 1

Temperature and rainfall, 1991–2020

Source: Our own calculations based on CRU data (2021).

Given the high vulnerability of jobs and low productivity and incomes in sub-Saharan Africa (Szirmai et al. 2013; ILO 2020), the negative effects of climate shocks on employment are likely to have considerable economic and social costs. Vulnerable groups such as young people will be the hardest hit by forecast job and productivity losses. Young people account for over half the population of Africa and it is estimated that they will reach 295 million by 2035 (Filmer and Fox 2014). This demographic trend implies an increase in the number of young people entering the labour force. Young people make up 23.5 per cent of the working poor, who constitute 38.1 per cent of the labour force in sub-Saharan Africa (Yeboah and Flynn 2021).

In sub-Saharan Africa, the labour force is highly exposed to unemployment, particularly young people. In 2003, the average unemployment rate was estimated at 5.8 per cent compared with 8.9 per cent among young people (ILO 2024). Youth unemployment can be a serious problem, as it weakens human capital and prevents the accumulation of professional experience, with negative effects on future income and career opportunities (Marelli and Signorelli 2016). From a social and psychological perspective, long-term unemployment among young people can also exclude them from society, resulting in discouragement and frustration (Seghiar 2014) and giving rise to social unrest and instability. For example, unemployed young people might be ripe for recruitment by armed groups, thus aggravating the fragile state of countries already facing security crises. Accordingly, as a catalyst for economic growth, poverty reduction and peace, youth employment is at the centre of the new vision of development defined in the African Union’s Agenda 2063 and in the United Nations’ (UN) 2030 Agenda for Sustainable Development.

Our study aims to estimate the effects of climate shocks on jobs in sub-Saharan Africa. In particular, we consider the effects of the temperature peak in 2010 and the drought of 2011 on youth employment and employment in agriculture, manufacturing and services.

The existing literature has mainly concentrated on the effects of climate change on the allocation of labour by agricultural households in developing and developed countries. Among many others, contributions include Josephson and Shively (2021) on Zimbabwe; Brookes Gray, Taraz and Halliday (2023) on South Africa; Liu, Shamdasani and Taraz (2023) on India; Desbureaux and Rodella (2019) on Latin America; Branco and Féres (2021) on Brazil; and Graff Zivin and Neidell (2014) on the United States. According to this literature, labour reallocation is a coping strategy against unexpected events related to a climate shock (Rose 2001; Emran and Shilpi 2018; Minale 2018; Colmer 2021). Households generally reallocate their labour when their income and means of subsistence are affected by meteorological shocks such as rainfall anomalies (Desbureaux and Rodella 2019; Branco and Féres 2021; Josephson and Shively 2021; Brookes Gray, Taraz and Halliday 2023), temperature spikes (Graff Zivin and Neidell 2014; Colmer 2021; Liu, Shamdasani and Taraz 2023) and floods (Mueller and Quisumbing 2011). Studies also document a fall in agricultural wages due to reduced demand for labour during climate shocks (Jayachandran 2006; Emran and Shilpi 2018).

Despite this rich literature, we know very little about the effects of climate shocks on youth employment in sub-Saharan Africa. Moreover, existing studies have explored the effects of specific climate shocks, whereas transmission mechanisms differ depending on the nature of the shock. Unlike studies that consider employment at the sectoral level,¹ our study provides estimations of the effects of climate change on youth employment in general. We also examine the effects on employment of two phenomena: temperature variation and drought. This allows us to identify the operative transmission mechanisms and make targeted recommendations. Lastly, we consider the process of labour reallocation from agriculture to manufacturing and services. On this basis, we formulate the following question: what are the effects of temperature shocks and drought on employment in the countries of sub-Saharan Africa?

In order to answer this question, we combined temperature and rainfall data with data on employment for 42 countries in sub-Saharan Africa for the period 1991–2020. The results from the difference-in-differences approach indicate that temperature variation has a negative effect on youth and agricultural employment but a positive effect on employment in manufacturing and services.

The remainder of the article is organized as follows. In the second section, we discuss the mechanisms through which climate change can affect employment. The third section presents our data and descriptive statistics. The fourth section describes our econometric model. We analyse our results in the fifth section and in the sixth, we conclude and discuss some public policy implications.

2. Literature review

The effects of climate shocks on employment can be reflected in labour productivity, in investment and in sectoral reallocation.

3. Data and descriptive statistics

4. Model specification and estimation method

In order to illustrate the means by which climate shocks can affect productive jobs, we consider a Cobb-Douglas production function, presented as follows:

$Y_{\mathit{\text{it}}}\text{=}{A}_{\mathit{\text{it}}}{K}_{\mathit{\text{it}}}^{\beta }{L}_{\mathit{\text{it}}}^{1–\beta }$       (1)

where Y_it is the quantity produced in country i at time t, K_it is the stock of capital, L_it is employment, A_it is the total productivity of the factors; β and 1 – β are, respectively, the elasticities of capital and labour in relation to production.

By dividing equation (1) by L_it, we obtain:

$y_{\mathit{\text{it}}}\text{ = }{A}_{\mathit{\text{it}}}{k}_{\mathit{\text{it}}}^{\beta }$       (2)

where $y_{\mathit{\text{it}}} \text{=} \frac{Y_{\mathit{\text{it}}}}{L_{\mathit{\text{it}}}}$ is the productivity of labour and k_it is capital intensity. A climate shock can reduce production by affecting the productivity of the factors (Zhang et al. 2018; Huang et al. 2020). High temperatures impact labour productivity by causing physical discomfort, fatigue and lower cognitive functioning in workers (Hancock, Ross and Szalma 2007).

$A_{\mathit{\text{it}}} \text{=} A_0{e}^{\mu T_{\mathit{\text{it}}}}$       (3)

In equation (3), µ is a group of parameters to estimate and T_it represents the climate variables. By replacing (3) in (2) and using the logarithm, we obtain the following relationship:

$y_{\mathit{\text{it}}} \text{=} a_0\text{+} \mu T_{\mathit{\text{it}}}\text{+} \beta k_{\mathit{\text{it}}}$       (4)

Our aim is to identify the causal effect of climate shocks on jobs, estimating the effect of the treatment on the treated. More specifically, it involves comparing the jobs affected by climate shocks with the counterfactual, that is, jobs in the absence of such shocks. Since the counterfactual is not observed, it needs to be estimated by means of a comparison group. The countries affected are in the treatment group and the unaffected countries in the control group. This distinction is based on the value for temperature in 2010 or for rainfall in 2011. According to the World Meteorological Organization (WMO 2020a), the year 2010 was one of the three hottest years ever recorded on the continent. Similarly, rainfall in 2011 displayed sharp geographical contrasts. Drought affected various parts of Africa between 2010 and 2016.

The literature offers a number of tools for estimating the counterfactual, namely random assignment. This is an excellent method of evaluating the impact of a policy or programme because it generates a solid counterfactual, which is taken as the gold standard for impact evaluation (Gertler et al. 2016). This method is applicable when: (i) the number of population units eligible for a programme exceeds the number of places available in that programme; (ii) the programme has to be progressively extended to cover the eligible population. As these criteria are not met in our case, random assignment is not applicable. Contrary to this method, the difference-in-differences approach does not require precise programme assignment rules. It is applicable when information is available on the treatment and control groups both before and after the implementation of the programme (Gertler et al. 2016). In our case, we have data on both of these groups for both before and after the implementation of the treatment variable (climate shocks). We therefore use the difference-in-differences approach to estimate the effect of climate shocks on jobs. Unlike the matching and before-and-after comparison methods, this approach considers the time-invariant differences between the treatment group and the control group (Gertler et al. 2016; Glewwe and Todd 2022). However, it does not remove the time-variant differences between the two groups (Gertler et al. 2016), which can bias the estimation of the counterfactual. As a result, to provide a robust counterfactual, it must be assumed that there is no time-variant difference between the treatment and control groups. This constitutes the parallel trends hypothesis, which assumes that in the absence of treatment, the mean change in the outcome variable for the treatment countries is equal to the mean change in the outcome variable observed for controls (Mora and Reggio 2019). This implies that results reflect equivalent trends in the absence of treatment (Gertler et al. 2016).

The difference-in-differences method is widely used to evaluate the impact of public policy interventions or specific treatments on outcomes of interest (Duflo 2001; Galiani, Gertler and Schargrodsky 2005; Imbens and Wooldridge 2009; Joshi 2019; Chávez and Rodríguez-Puello 2022). It compares the treatment group (jobs affected by the climate shock) and the control group (unaffected jobs) to identify the climate shock effect. The conventional difference-in-differences estimator is obtained using standard linear regression techniques. In the simplest case, with only two periods (t ≥ 2), the treatment effect can be estimated with the parallel trend hypothesis based on a regression with one constant, the treatment indicator D_it, a mute variable for the post-treatment period (C_it), and the interaction term D_it × C_it. In this configuration, the treatment effect is identified by the parameter associated with the interaction term (δ) (see table 2). Formally, the standard difference-in-differences model can be specified as follows:

$Y_{\mathit{\text{it}}}\text{=}\alpha \text{+}\beta C_{\mathit{\text{it}}}\text{+}\gamma {\mathit{\text{T}}}_{\mathit{\text{i}}}\text{+}\delta C_{\mathit{\text{it}}}{\mathit{\text{T}}}_{\mathit{\text{i}}}\text{+}\theta X_{\mathit{\text{it}}}\text{+}{\eta }_i\text{+}{\psi }_t\text{+}{e}_{\mathit{\text{it}}}$       (5)

where Y_it is the indicator for jobs in country i, in period t; C_it is a binary variable taking the value of 1 if t is the post-treatment period; T_i is also a binary variable taking the value of 1 if country i belongs to the treatment group; δ represents the mean estimated effect of the climate shock on jobs; X_it represents all the control variables that vary both between country and over time; η_i controls for the country fixed effects; ψ_t represents the time shocks common to all the countries; e_it is the error term – the error terms are grouped at the country level; and α is the constant.

We assume that the temperature and rainfall shocks took place at the end of 2010 and 2011, respectively. Accordingly, the estimation equation is the following:

$Y_{\mathit{\text{it}}}\text{=}\alpha \text{+}{\displaystyle {\displaystyle \sum }_{t\text{=}1991}^{2020}}{\beta }_t{C}_{\mathit{\text{it}}}\text{+}\gamma T_{\mathit{\text{it}}}\text{+}{\displaystyle {\displaystyle \sum }_{t\text{=}1991}^{2020}}{\delta }_t{C}_{\mathit{\text{it}}}{T}_{\mathit{\text{it}}}\text{+}{\in }_{\mathit{\text{it}}}$       (6)

5. Presentation and analysis of results

6. Conclusions

The increasing frequency and intensity of extreme weather events leads to a loss of labour and agricultural productivity, which could have negative effects on jobs and workers’ income. Studies on sub-Saharan Africa have shown that the reallocation of labour supply is one of the main responses to climate shocks (Josephson and Shively 2021; Brookes Gray, Taraz and Halliday 2023).

By exploring the effects of climate shocks on youth employment, our study contributes to a limited literature on sub-Saharan Africa, while also strengthening our understanding of the factors that influence youth employment dynamics in particular and adult employment dynamics in general. We have shown that extreme temperatures lead to a loss of employment among young people and in agriculture. Our results indicate that the increase in temperature brings about a reallocation of labour away from agriculture and into the manufacturing and services sectors. Furthermore, we have found that drought has a positive effect on employment in desert countries, whereas the effect is negative in tropical countries.

These results enable us to make a number of recommendations for governments, enterprises and households. Governments should promote measures to strengthen resilience, such as developing infrastructure and technology, capacity-building and regulation. Investment in infrastructure and technology could promote energy transition and support structural transformation. This would be an ideal option for countries in sub-Saharan Africa, as it would enable workers who are vulnerable to high temperatures to leave the agricultural sector for more productive sectors. Capacity-building, especially among young people, could narrow the skills gap in the labour market. It could also stimulate structural change in economies and facilitate the development of new heat-resistant crops. This would contribute to improving agricultural production and reducing the risks of job loss and food insecurity.

The improvement of governance and the adoption of labour standards and rules can guarantee decent working conditions for workers and in enterprises, which could mitigate the negative effects of climate shocks. Moreover, the adoption of flexible labour market measures offers enterprises the possibility of adapting to changing economic conditions and managing their activities efficiently, which can protect jobs, especially for young people.

Enterprises and households need to adjust their practices and behaviours in order to promote the creation of new jobs and the protection of old ones. For example, investment will enable enterprises to reuse, recycle and reduce their consumption of fossil fuels, which, thanks to resultant efficiencies and productivity, can protect existing jobs and create new ones. Likewise, households could rationalize their use of resources and consumption and adopt altruistic behaviour in terms of reforestation.

We have conducted our analysis at the national level and have therefore not taken into consideration the heterogeneity among regions, which are at different stages of development. This limitation could be addressed in future research through the use of georeferenced data, which provide information at the infranational level, improving their precision and quality. Future research could also examine households’ employment decisions in the event of climate shocks, which would provide a broader understanding of the problems surrounding employment in crisis situations.

Acknowledgements

We would like to thank the African Union for its financial support for our research. We are also grateful to all those who have contributed by reading and improving the quality of this article, in particular François Nkoa.

Competing interests

The authors declare that they have no competing interests.