Introduction

The fragile economy and agriculture-based livelihoods of many developing countries are highly natural resource dependent. The rapid and dynamic change of the world environment puts pressure on the lives of the general public in such countries in innumerable ways. This susceptibility of an economy and the livelihoods to such environmental perils manifested itself in multifarious ways, including economic shocks, natural disasters, and conflict-related losses which cause an annual economic loss of more than USD 250 billion, according to the 2017 Sustainable Development Goals (SDGs)(Gershon et al., 2020). Climate change manifests in African countries as both “gradual decline in the length of rainy seasons, frequent droughts, floods, heatwaves and dust storms” and also extreme weather shocks such as floods, droughts, rainfall variability, rainfall precipitation, high temperatures, and more (Asmare, 2018). Such climate impacts have severe effects on the welfare of household (Zanhouo & Acma, n.d.;Thanh et al., 2020;Shiferaw et al.,2014))

 

In the face of these pressures, apart from the usual traditional government responses and insufficient initiatives, there are no strong institutions or preventive actions that would help to minimize the impacts of climate shock on the life of the poor or to help them cope with these changes (Federman et al., 2014). Agriculture is the dominant sector for the economy of many developing countries in general and Sub-Saharan Africa (SSA) in particular. Ethiopia, the focus of this paper, has such an agriculture based economy, with the sector remaining the main activity and contributing the lions share in GDP, employment creation and exports (Solomon et al., 2021). Rainfed agriculture represents the largest proportions, meaning that good agriculture seasons are conditioned by good rainfall and ideal temperature. Due to this, climate change poses a threat to the agricultural sector and agriculture-based livelihoods (Birthal & Hazrana, 2019). Rainfall variability, driven at least in part by climate change, is among the dominant factors affecting the productivity and production of agriculture and the livelihood of the poor and their economies (Birthal & Hazrana, 2019).

 

Climate change has particular effects on developing countries with greater dependence on rainfall agriculture with small landholdings, low agricultural productivity, slow economic growth and persistent poverty (Lottering et al., 2021). Drought leading to decreased crop yields and livestock production accounts for about 25% of the natural disaster of the continent (Shiferaw et al., 2014). Such drought has been increasing in frequency, intensity and duration with climate change. The situation is especially critical as dozens of millions of people as still food-insecure and children are at risk of acute malnutrition (Mare et al., 2018). Climate shock both directly and indirectly negatively impacts agricultural production, causing loss of life, deterioration of health, loss of livelihoods and the like. These all work towards the reduced effects of the household welfare (Zimmerman & Carter, 2003).

 

The effects are not evenly distributed. Climate change impacts are mediated through “several factors such as population, technology, policy, social behavior, land use patterns, water use, economic development, and diversity of economic base and cultural composition” (Naumann et al., 2014). Further, Amartya Sen also argued that drought leads to famine and loss of livelihoods to the extent that there is capability failure which in turn depend on market access and people’s social, economic and political entitlements (Singer, 1982; Koo & Martin, 2021; Kinda, 2016).The effect of climate change is higher in threatening the life of the rural poor households than the urban and non-farm households. Drought regions are more prone to the climate change shocks than other areas (Solomon et al., 2021). Deforestation, soil erosion and land degradation all contribute to the food security status and welfare status of households. Thus, the issue of climate change needs critical intervention on ways to minimize its impact so that the burden of poverty will be lessened and the welfare aspects of the households will be improved.

 

Unfortunately, different studies come up with inconclusive results. The effect of climate change is significant with some of them and inconclusive with others. Household characteristic variables play an important role in this regard. Thus, the present study aims to comprehend the heterogeneous effects of the climate shock on household welfare using the socioeconomic survey data from Ethiopia’s Central Statistical Authority (CSA), wave 4 (2018/19). The study builds by integrating the dominant aspects of climate vulnerability shocks typically drought, rainfall variability and temperature variations. The novel contribution of this study is providing a comprehensive assessment the effects of climate shock by developing the augmented variable by the interaction of the climate shock variable with the household characteristic variable. The study used instrumental variable approach.

 

Methodology

Other scholars have handled the inquiry hypothesis in distinct manners: the portfolio asset bifurcation (Zimmerman & Carter, 2003); poverty trap hypothesis (Barrett et al., 2016); the poverty dynamics breaks out the household and individual level that is disaggregate the impact of shocks by asset holding, the individual level effects and the permanent consequences of shocks (Hoddinott, 2006); the livelihood diversification approaches (Ellis, 2000). This study used an instrumental variable method to capture the heterogeneous effects of the economy wide impact of climate change on household welfare. Along with the key indicators of climate change such as drought, flood, rainfall variability, temperature and the like, they study also took into account the effect of other control variables influencing the household welfare.

 

Data from climate shocks and the welfare effects were taken from wave 4 (2018/19) of the Ethiopian socioeconomic survey by the Central Statistical Authority. This is publicly available data. The empirical strategy follows the works of (Vijay, 2000) based on the permanent income and full insurance models of consumption dynamics. According to this approach the mitigation of adverse impacts of shocks and consumption smoothing is done by resorting to different coping mechanisms. The two major shocks were identified based on the cross-sectional variations of the idiosyncratic shocks which are unique to households, and covariate shocks where many households experience the shock together. Major shocks considered here are: temperature variation, rainfall variability, drought, and market shocks (price volatility).

 

The following hypotheses need to be evaluated when taking into account such viewpoints on how shocks affect the wellbeing of households:

​

H1: The effects of climate shocks matter in influencing the decomposed household welfare.

​

H2: The heterogonous effects of climate shocks matter in influencing the household welfare.

​

H3: The market shock matters in influencing the household welfare.

​

The empirical analysis is commenced by providing the conceptual framework of the study that frames the econometric model to be applied. The major shocks are identified and the direction of movement from each shock to the multiple well-being effects of each shock is also set.

​

​

​

​

​

​

​

​

​

​

​

​

 

The conceptual framework of the study

​

The econometric model used to estimate the effects, the heterogenous impacts of shocks on dependent variable (household welfare outcome) follows the following specification:

​

​

​

 

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

 

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

 

Estimation Strategy

The instrumental variable approach (IV estimation), which is used in such models to get around the endogeneity issue, is used to estimate the stated econometric model (Wooldrigde et al., 2010). The IV coefficient is consistent and greater than the OLS estimates even though the dependent variable is measured using a continuous scale and the ordinary least square (OLS) method is used to produce estimates (Faradiba, 2021). From the above multiple regression model:

​

​

​

The OLS assumption that there should be no correlation between the explanatory variables and the error term is violated when the explanatory variables and error term are correlated, hence it cannot be applied for estimation. Such issue has been described as an endogeneity problem that the OLS approach cannot handle. Breaking the link between the two gives the solution for OLS and with the following condition, the instrumental variable approach is employed to deal with the case.

​

​

​

The key contribution of this paper is examining the heterogenous impact of climate shocks along with the effect of shocks on the household welfare. Thus, the basic model is augmented by interacting the climate shock variable with household characteristic variable and the extent to which π is different from zero demonstrates the degree of the heterogeneity of shocks across the household characteristics.

​

Result and Discussion

The estimated impacts of climate shocks on household wellbeing are covered in this subsection using the welfare indicator of household food consumption expenditures. Three significant topics are covered in this section. These are: Summary of descriptive statistics, the OLS and IV regression results, as well as the post estimation results.

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

Regression Analysis

Following is a presentation of the regression results for the heterogenous impacts of climatic shocks on household welfare. The relationship between the climate shock variables and household welfare, as measured by household food spending, is negative and substantial, as shown in table 3 below. According to OLS and 2SLS (IV) estimations, a 1 degree increase in the mean temperature expressed causes a 0.095 and 0.062 percent decrease in family welfare other things remaining the same. The same is true for household welfare, which decreases by 0.051 and 0.046 percent, respectively, with a 1% increase in rainfall variability other things keeping constant. This demonstrates how negatively the climatic shock variables impact the wellbeing of the household. This signifies that the households’ capacity to meet their food consumption demand is compromised by climate-related shocks. Variability in temperature and rainfall are deemed unfavorable weather conditions that impair production, earnings, and spending on food consumption. Thus, the channel through which   the climate shock uses to influence the wellbeing of the household matters. In this case, the fall in agricultural production causes a decrease in household income, which further causes a decrease in the amount spent on food, worsening household welfare. Thus, the climate variability impact is very eminent and there is a negative effects of climate variability on the household welfare This finding is consistent with (Tefera et al., 2022; Weldearegay & Tedla, 2018; Mumuni, 2022; WU et al., 2021; Mekonnen et al., 2021).

 

The economic shocks, rise in price of food and rise in price of inputs, are closely related to the well-being of the poor. We therefore measure the welfare effects of the price change from producer and consumer sides. From the producer side, an increase in price of inputs decreases agricultural production and leads to market instabilities. This, further escalates the price of output and increase in consumption expenditure to be at the same level of consumption as that of before. Thus, increase in prices of food and prices of input worsens the welfare of the household. An increase in price of food also leads to consumption switching to less quality and cheaper, less organic food items which are not good for health and hence affects well-being negatively(Magrini et al., 2015). Price shocks have detrimental effects on household welfare and poverty (Anna,2015); (Adekunle et al., 2020); (Tukae & Xiaohua, 2016); (Etang Ndip & Touray, 2019).

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

Household characteristics matter for the ability to adapt to climate shocks. As there are few other sources of income generation in rural areas, land is the primary source of agricultural output there. Therefore, access to land cause an increase in farm yields, which raises household income and increases spending on food. Land ownership therefore results in significant increases in production and welfare (Mogues, 2011). Up to a point, a household’s ageing results in a rise in production due to increased efficiency and experience in agricultural practices. This ultimately caused a rise in income and food expenditures. In comparison, both OLS and 2SLS (IV) estimations demonstrate that square age has a diminishing impact on the welfare of the household, demonstrating that productivity declines over time and that retirement occurs(Oshimi et al., 2016). Further, household size has positive impact on the welfare of households showing that family size and economic welfare are positively related(Espenshade et al., 1983). Marital status contributes negatively to the welfare of the household as per the empirical results of the study. Another interesting result is the heterogenous effects of shock on the household welfare. This heterogeneity is captured by the interaction of the climate shock (drought) with the natural capital (land ownership) and it is statistically significant at conventional levels. This shows that there is a degree of heterogeneity in the effects of the climate shocks among the household characteristics(Etang Ndip & Touray, 2019). Access to credit has positive and significant impact on the welfare of households.

​

​

APPENDIX: Postestimation tests for IV-based estimations

A series of postestimation tests can be used while implementing IV-based estimations. These tests are: Relevance condition, Test for endogeneity, test for overidentifying restrictions, test for weak instruments,

 

Test for Endogeneity: The Durbin-Wu-Hausman (DWH) test

 

In a regression model, the DWH test finds endogenous regressors (predictors). The OLS will fail if endogenous variables are used in the model since they are dependent on other factors. When there is a correlation between the predictor variable and the error term, an alternate strategy is the instrumental variables estimator.

​

​

​

​

 

From the above endogeneity test the null hypothesis of variables are exogenous is accepted and the instrumental variables used in the IV regression analysis are the proper ones as the p values are not significant. We cannot conclusively reject the null hypothesis of exogeneity. Weak Instrument Test (Relevance criterion) The caliber of the instrumental variables taken into account affects how reliable the endogeneity test is. This establishes the instrument’s capacity for explanation. The chosen instrument needs to be relevant and exogenous (Ullah et al., 2020) and this is done using the Stok -Yogo test shown below. As shown below the instruments are strongly influencing the endogenous variable at conventional level of significance. This shows the explanatory power of the instrument. This is the relevance criterion conditions of the instruments where                                   

As shown below the two IV variables are individually significant (t-test) and jointly significant from the F-test.

​

​

​

​

​

​

​

​

​

​

​

​

​

Further, from the first step regression given below R-squared (R2 ) and the adjusted R-squared ( ) shows the normal coefficient of determination obtained by regressing the endogenous variable on the exogenous and instruments. But, for the weak instrument test R2 and are not that much relevant. In place, the Shea’s partial R2 explaining how much the instruments explain the endogenous variable matters a lot. Since the value is almost around the normal R2 with minor differences, it is a good signal that the instruments are not weak. This can further be confirmed from the F-statistic. F statistic with two instruments is highly significant. The rule for weak instrument test is comparing the F- statistics with the 2SLS and Likelihood Information Maximum Likelihood (LIML) and how much bias we are going to tolerate (10% or 20%, or 25% or 30%). At all levels of bias, the F-statistic is greater than the 2SLS bias and the LIML bias. And, hence we can conclude that the IVs are the strong instruments of the endogenous variable.

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

Test of the overidentifying restriction test

 

The over identification test is done after the regression in run with IV method and this test is applied to check the exogeneity conditions of the instruments. It is the test of checking whether the instruments are correctly identified and uncorrelated with the disturbance term. The results of the test based on the chi-square of the Sargan’s and Basman’s tests are not significant (see p value below) implying that the instruments are valid and the model is correctly specified. This shows that the null hypothesis of the exogeneity is not rejected and the IVs are truly exogenous ones.

​

​

​

​

​

​

​

​