*3.3. Model Specifications*

A logit model is a statistical approach used when the dependent variable (DV) is nominal with two or more categories commonly grouped as dummy variables (Table 3). On the other hand, the independent (predictor) variables (IVs) in a logit model can be qualitative (nominal or ordinal level) and interval/ratio level variables. This study used a logit model to identify factors that control the likelihood of households' exposure to climate change-induced impacts. Climate change-induced impacts considered in this study include the likelihoods to exposure to drought, harvest loss, flooding, and hunger. The exposure to hunger is examined vis-à-vis eighteen socio-economic and institutional s (Table 2). The eighteen dependent variables were selected among other socioeconomic and institutional

variables after all the variables were tested to have a relationship with the four dependent variables considered for this study. The logit model showed that the selected eighteen variables have some explanatory power on the dependent variables considered.


**Table 2.** Variables description.

1. Productive Safety Net Program is a governmen<sup>t</sup> program being implemented in the study area to support poor and vulnerable people. 2. Are groups of 5 people organized by governmen<sup>t</sup> and used for political, economic, and social purposes. 3. Drought resistant varieties are plant seeds that are supplied by the governmen<sup>t</sup> and can grow in moisture stress conditions. 4. Urea and NPS are the two types of chemical fertilizers supplied to farmers by the governmen<sup>t</sup> (usually in loan) NPS refers to Nitrogen-Phosphoric fertilizer containing Sulphur. 5. Improved animal fodder is a plant species (supplied by the government) that is used as a forage to feed animals during a shortage of pasture. 6. HHS are members of a family living in a house.



\* Correlation is significant at the 0.05 level (2-tailed).

Logistic regression involves a linear function:

$$Z = b\_0 + b\_1 X\_1 + b\_2 X\_2 + \dots + b\_n X\_n \tag{2}$$

where Z is a dependent variable and Xi ... Xn are explanatory variables.

The logistic regression function is thus, given by:

$$p = \frac{\varepsilon^z}{1 + \varepsilon^z} \tag{3}$$

where *p* is the probability that an event occurs; and the odds in favor of the occurrence are related according to *p* = Odds 1−odds= *p* 1−*p*.

$$\log\left(\frac{p}{1-p}\right) = \beta\_0 + \beta\_1 X \tag{4}$$

Thus, odds is defined for an event with probability *p* and the logit is the log of the odds, i.e.; Model - *p*(*x*) = probability of the event occurring at (Y=(*x*))

$$p(y/\mathbf{x}) = \frac{e^{\beta\_0 + \beta\_1 X\_1 + \beta\_2 X\_2 + \dots + \beta\_k X\_k}}{1 + e^{\beta\_0 + \beta\_1 X\_1 + \beta\_2 X\_2 + \dots + \beta\_k X\_k}}\tag{5}$$

A smaller sample size with a large number of predictors creates problems with the analysis, specifically when there are categorical predictors with limited cases in each category. Descriptive statistics were run for each of the predictors to solve the problem. The result showed that all categorical variables have a required number of cases set as a minimum standard [44]. A multicollinearity test was done to check for high intercorrelations among predictor (independent) variables. The test result suggested that the maximum r-value of the independent variables is 0.7, which is within an acceptable limit [44]. The presence of outliers or cases that would not well be explained by the model was also checked to make sure the model fits the data well. The model tried to find out factors that control the HHs likelihood of exposure to climate-induced impacts such as the exposure to the e ffects of drought, harvest loss, flooding, and hunger in the study area. R<sup>2</sup> and adjusted R<sup>2</sup> were also used to discuss the amount of variation explained by the independent variables. The former supposes that every independent variable in the model explains the variation in the dependent variable by explaining the variation in percent as if all independent variables in the model a ffect the dependent variable. On the other hand, the adjusted R<sup>2</sup> gives the percentage of variation explained by independent variables that in reality, a ffect the dependent variable. In this context, it is also called marginal e ffect after logit.

#### **4. Results and Discussion**

#### *4.1. RF Condition and Drought Incidents*

Studies on RF over Ethiopia revealed high variability and drought incidents [32,35,45]. Similarly, the analyses of RF data have shown declining RF trends and incidents of drought during the years under consideration (1983–2014). Annual and growing season (*belg* and *kiremt*) RF amounts had shown declining trends in the study area. The *belg* season (February, March, April, and May) RF accounts for 40% of the annual RF in the study area. It is an important season not only to grow short-growing season crops, but also it provides water for livestock and helps the growth of pasture. Though the current study does not check daily RF data of the growing seasons, the experts and focus group discussion participants reported that problems related to its onset, duration, and o ffset have serious consequence on food security in the study area. However, despite its huge role for livelihood security, it has shown a significant (*p* ≤ 0.05) declining trend in the last three decades (Figure 2a, Table 3). The *kiremt* season (June, July, August, and September) is the main rain growing season in the study area which accounts for 41% of the total annual RF. *Kiremt* RF has also shown a declining trend over the years under consideration (Figure 2b), though the trend is not statistically significant (Table 3). The trend analysis of annual RF data has also revealed a declining trend, though the result is statistically insignificant (Table 3). The annual RF had varied from its average amount, which amounts to 1100 mm. The year 2009 was the driest year among the years under consideration with the annual amount of 796.37 mm. The years 1984, 1999, 2000, 2002, 2003, 2004, and 2009 were the drier with annual RF of 950.44 mm, 870.1 mm, 934.02 mm, 960.09 mm, 994.71 mm, and 925.08 mm, respectively.

The years 1983 and 1996 were the wettest years with annual RF amounts of 1420.42 mm and 1368.41 mm, respectively (Figure 3). On the other hand, the analysis of minimum, maximum, and mean monthly temperature had shown increasing trends, all of which are statistically significant

**Figure 2.** Rainfall trend in the study area (1983–2014). Source: Own computation from Ethiopian National Meteorological Agency data.

Source: Own computation from Ethiopian National Meteorological Agency data.

> **Figure 3.** Annual RF standard anomalies in the study area (1983–2014).

The RF Standard Anomalies (RSA) helps to estimate the extent of drought based on the RF data [46] (Table 4).


**Table 4.** Drought severity index based on standard rainfall anomalies (SRA).

Source: Janowiak et al. 1986.

The analysis of annual RSA pointed out seven droughts incidents of different extent during the three decades under consideration. The year 2009 with annual SRA value of −2.11 was the driest year in the last three decades, which coincides with the 2009/10 extreme drought all over the country. The years 1999 and 2012 were also years of severe drought in the study area with SRA vales of −1.6 and −1.45, respectively. The other four moderate drought years were in 1984 (SRA = −1.04) 2000 (SRA = −1.16) 2002 (SRA = −0.98), and 2004 (SRA = −1.22) (Figure 3).

#### *4.2. Types of Climate Change-Induced Impacts*

Drought and flooding are the most common climate change variability-induced risks in Ethiopia [2,39,47]. The findings of this study nevertheless revealed other climate change-induced impacts that are recurrent in the study area. The common ones include climate-induced seasonal epidemics, drought, harvest loss, flooding, hunger, migration, and school dropout [48].

#### 4.2.1. Climate Change and Variability Related Epidemics

Climate change-related epidemics globally claim over 150,000 lives per year [26]. Studies on the health impact of anthropogenic climate change show that climate change is affecting human health. Diseases such as cardiovascular mortality, respiratory illnesses, water-related contagious diseases, and malnutrition are directly or indirectly associated with climate change [24,26–50]. Malaria, cholera, and animal diseases were the major climate change and variability induced epidemics reported by the HHs in the study area. Of the total respondents asked about the incidence of malaria in their locality, 46% (*n* = 183) said that they had suffered from a malaria epidemic in their family. Malaria has been the most prevalent disease in the study area and its prevalence increases in the wet season from May to November. During wet seasons, the availability of moisture creates a conducive environment for mosquitos to breed and spread, such as those places left open to harvest rainwater.

On the other hand, thirty-five percent (*n* = 138) of the households reported cholera incidences. Water supply coverage in the study area has been marginal because of poor hydro-geological conditions and limited resources to improve water supply coverage. Water scarcity forces people to use untreated water for domestic consumptions. Key informants in Boricha district said that such incidences usually happen during water-scarce seasons where families resort to using unsafe water for domestic consumptions. Diarrheal diseases incidents are very common during the dry season (from December to March) affecting health and household labor availability in the area. Twenty-one percent (*n* = 85) of the study participants also reported the prevalence of animal diseases. The problem is more acute for agro-pastoral communities living in the western parts of the study area, especially in Loka Abaya district. Some HHS also reported the prevalence of animal diseases, such as trypanosomiasis. It affects animal farming in the study area. About 21% (*n* = 81) of respondents reported that they had experienced climate change-related livestock diseases. Trypanosomiasis is one of the tropical livestock diseases, which has widely spread all over the SSA [51,52]. The World Health Organization (WHO) has also pointed out that climate change induced animal diseases are affecting cattle header Maasai pastoralist communities living in Tanzania and Kenya [53].
