Effects of cash transfers on household resilience to climate shocks in the arid and semi arid counties of northern Kenya

ABSTRACT

Climatic events and other natural-related disasters experienced in the arid and semi-arid lands of northern Kenya negatively affect the pastoral livelihoods of the communities. Addressing vulnerability to climate shocks among pastoral communities of Kenya’s Arid and Semi-lands presents a persistent challenge. Cash transfer programmes have increasingly grown as one mode of building household resilience. Understanding the role of cash transfer interventions on household resilience to climate shocks is key to policy programming. This paper aimed at determining the effects of cash transfers on household resilience to climate shocks. The paper evaluated the Hunger Safety Net Program, which is one of the largest unconditional cash transfer programs in Kenya. The Hunger Safety Net Program targeted poor people in northern Kenya including the counties of Turkana, Wajir, Marsabit, and Mandera. To establish the impact the paper compares households which received cash transfers with those that did not receive transfers. The panel fixed effects model was used to determine the effects of cash transfers on household resilience. The results indicate that cash transfers have positive significant effects on household resilience to climate shocks.

KEYWORDS:

• Cash transfers
• resilience
• vulnerability
• livelihoods
• fixed effects

1. Introduction

In the arid and semi-arid lands (ASALs) of northern Kenya pastoralism is the dominant livelihood activity. Pastoralism is very sensitive to shocks which induce heavy losses in livestock, exposing pastoralists to poverty. Poor households, especially those with small livestock holdings and those who lack well-established social support systems, are more vulnerable (Ouma, Obando, and Koech 2012). Vulnerability compels households to take negative coping mechanisms such as selling livelihood assets and withdrawing children from school, both of which have long-term implications for the households (Hansen et al. 2004; Jensen, Barrett, and Mude 2017).

Prior to 1990s, the focus of humanitarian organizations and donors was on relieving the hunger crisis through food aid (Kripke 2005). Since then, attention has shifted from simply easing drought-related hunger to the enhancement of household resilience (Bhalla et al. 2018; Brugh et al. 2018). Accordingly, the intervention measures have also changed. Cash transfer programmes have increasingly grown as one of the modes of building household resilience (Agrawal et al. 2020; d’Errico et al. 2020; Ulrichs, Slater, and Costella 2019; Weingärtner et al. 2019).

The term resilience has assumed significance in many disciplines and applications in widely varying contexts (Alexander 2013). Accordingly, the term has attracted different conceptual definitions depending on academic orientation and context (Weichselgartner and Kelman 2014). Unifying themes, and different approaches to promoting the concept of ‘resilience’ are increasingly structuring debates around climate change, social protection, sustainable development, and humanitarian responses to emergencies (Alinovi et al. 2010; Cissé and Barrett 2018; Mekuyie, Jordaan, and Melka 2018; Ngoma, Finn, and Kabisa 2021; Pain and Levine 2012; Sagara 2018; Thulstrup 2015; Tumusiime 2015; Ulrichs, Slater, and Costella 2019; Weingärtner et al. 2019; Yiridomoh et al. 2021).

The concept captures the fundamental abilities of individuals, communities or states, and their associated institutions, to withstand and recover from shocks (OECD 2013). The resilience of a household depends on its ability to anticipate and manage its exposure to negative livelihood shocks without resorting to negative coping strategies of permanently diminishing its productive assets (Jensen, Barrett, and Mude 2017; Schipper, Lisa, and Langston 2015). There is a clear link between resilience and livelihood in that the latter must incorporate mechanisms for adapting and bouncing back when challenges arise (Twigg and Calderone 2019). Livelihood strategies are defined as programmatic initiatives that increase an individual’s income-generating capacity by expanding their assets base by providing cash transfers, infrastructure, support services, market expansion activities, and training (Twigg and Calderone 2019).

This paper determined the effects of cash transfers in building household resilience to climate-related shocks in the ASALs of northern Kenya. The paper evaluated the Hunger Safety Net Program (HSNP), which is one of the largest cash transfer programs in Kenya. The rest of the paper is organized as follows: section 2 provides evidence of past studies on cash transfers; section 3 provides a description of the study area and a detailed methodology for computing the resilience index; section 4 provides a discussion of the results; finally, our conclusions and policy recommendations are presented in section 5.

2. Literature review

It is widely accepted that to avoid hardships and losses associated with climate shocks, strong emphasis needs to be placed on building resilience (Levine et al. 2012; Smith and Frankenberger 2018). The major question among researchers, donor communities and policymakers is how to facilitate vulnerable communities to prepare for, cope with and build resilience to shock (Bowen et al. 2020). Over the years, social protection interventions have been viewed as a promising strategy (Agrawal et al. 2020; Jorgensen and Siegel 2019; Weingärtner et al. 2019). Among these, cash transfers aim at smoothing incomes and consumptions, and reducing vulnerability to poverty (Jorgensen and Siegel 2019; Agrawal et al. 2020). A growing body of literature suggests that cash transfers are consistent with the improvement of households’ and communities’ ability to build resilience against adverse shocks. Valli, Peterman, and Hidrobo (2019) have shown that economic transfers can foster social cohesion between displaced and host communities. This has implications for pastoral communities of Kenya’s ASALs who have to move with their livestock in search of food and feed to avoid the ravages of climate shocks. Evidence also suggests that cash transfers strengthen social support among friends, families and community members (de Milliano et al. 2021). Furthermore, cash transfers are likely to increase households’ probability of enrolling and retaining children in schools (Baird et al. 2013; d’Errico et al. 2020; Mostert and Castello 2020; Kilburn et al. 2017), and reduce child labour (Churchill et al. 2021; Jayawardana, Baryshnikova, and Thien Anh Pham 2021). In turn, such positive educational impacts improve households’ capacities to break the intergenerational transmission of poverty. However, Iqbal, Farooq, and Ul Haq Padda (2021) suggest that desired impacts take time to materialize; using panel data, they observed that women's empowerment was not realized until the 5th to 8th repeated receipt of cash support.

In most developing countries, cash transfers meant to protect vulnerable households from weather-related shocks are still poorly understood (Kohlitz, Chong, and Willetts 2019). In Kenya studies demonstrate progress has been made on cash transfer programs; Kilburn et al. (2016) provide evidence on mental health, Haushofer and Shapiro (2016) on household consumption, Song and Imai (2019) on poverty reduction and Dietrich and Schmerzeck (2019) on nutrition improvement. The contribution of cash transfers towards building household resilience is given little attention in empirical works, especially in the context of ASALs where climate shocks are common (Ulrichs, Slater, and Costella 2019). Addressing this gap can have important implications for the design and programming of social protection systems.

2.1 Nature of present and future problems of climate crisis/shocks

Climate change-induced climate variability may eventually result in climatic shocks that cause environmental degradation. People's livelihoods, health, agricultural production, labour productivity, and socioeconomic well-being are affected indirectly by environmental degradation (Masuda et al. 2019; Thakur and Bajagain 2019; Nastis, Michailidis, and Chatzitheodoridis 2012). Globally People’s lives and livelihoods are affected by climate change, especially those living in extreme poverty, and action is needed to address climate effects (Sanson, Burke, and Van Hoorn 2018). Climate change has both direct and indirect effects on humanity across the world, including agricultural production, resource availability, water and the prevalence of diseases (Sanson, Burke, and Van Hoorn 2018). Regular occurrence of extreme climate-related stocks such as droughts and floods have increased vulnerability among agriculturalists who rely on climate circumstances, reducing long-term growth and hence affecting a large number of people (Sewando, Mutabazi, and Mdoe 2016). The Northern region of Kenya is associated with a higher risk of drought imposing heavy social and ecological burdens on people and ecosystems, such as overexploitation of fragile ecosystems and unsuitability for agricultural uses (Jensen, Barrett, and Mude 2017). The communities in the arid and semi-arid counties in northern Kenya are highly vulnerable to natural and human-made calamities such as drought, floods, and conflict. The key contributing factor to the high vulnerability of communities affected by disasters in the arid and semi-arid counties is their low ability to engage in other sources of livelihoods apart from pastoralism (UNDP 2018). Frequent occurrence of drought in the ASALs is a daily threat with negative significant impacts on pastoral livelihoods and increased vulnerabilities (Mureithi 2018; USAID 2018).

In developing countries, people are considered vulnerable to climate change due to social, economic, and environmental conditions. Climate change has negative impacts on households due to their low ability to cope with and adapt to climate shocks (Cutter et al. 2009; Nelson et al. 2010). Vulnerability to climate change and climate-related hazards is not a result of poverty only however, it can be a result of discrimination, disability, gender and inequalities in income and lack of access to resources. This means that these groups have fewer resources to help prepare for, absorb, cope and recover from adverse climatic effects hence they are more vulnerable (Pörtner et al. 2022).

The Intergovernmental Panel on Climate Change (IPCC) projects that the occurrence and intensity of climate change-induced shocks are increasing globally (IPCC 2014). Effects of extreme climate-related disasters would increase extra stress on human health, food security and water resources, where the rural poor are extremely vulnerable and adversely affected (IPCC 2014; McCarthy et al. 2001). Future climate change impacts are projected to worsen poverty and intensify inequalities within and between nations, these impacts are said to increase significantly by 2030 (Hallegatte and Rozenberg 2017; Olsson, Galaz, and Boonstra 2014; Roy et al. 2018). Climate change is projected to have compounding impacts on livestock, including negative impacts on fodder availability and quality, availability of drinking water, direct heat stress and the prevalence of livestock diseases (Godde et al. 2021; Nardone et al. 2010; Rojas-Downing et al. 2017). Key hazards, exposure, and susceptibility as a result of future climate change are difficult to measure and are based on information from the past as well as potential future vulnerabilities and livelihood issues (Pörtner et al. 2022).

In Africa, households use a variety of coping and adaptation strategies options to lessen the negative effects of future climate change risk, including agricultural and livelihood diversification, savings to smooth consumption, income diversification, selling productive assets or using formal or informal safety nets (Gao and Mills 2018; Thierfelder et al. 2017; Thornton et al. 2018). International donor organizations and national governments are collaborating in the Sahel to establish shock-responsive social protection systems to increase coping mechanisms for climate extremes and unpredictability, as well as account for shifting climate threats (Ben Mohamed 2011; Bowen et al. 2020; Kendon et al. 2019; Sissoko et al. 2011).

2.2 Conceptual framework

Cash transfers are part of a larger social protection plan aimed at addressing both current needs and building a better long-term social protection system. Social protection involves the provision of income or consumption transfers to the poor to protect the vulnerable from livelihood risks and improve the social status and rights of the disadvantaged (Devereux and Sabates-Wheeler 2004). The approach to social protection is that cash transfers can help disadvantaged households manage risk and invest in human capital and physical assets to enhance resilience (Browne 2013). In theory, cash transfers enable the beneficiaries to make their own decisions on critical needs and expenditures leading to satisfaction among beneficiaries (Rumble 2007). Cash transfer programs mediate growth facilitating access to credit, providing more continuity and security, and help overcome cost restrictions which can influence the household decision. Cash transfer programs intercede development encouraging access to credit, providing more certainty and security in consumption, and conquering cost limitations which can impact the family unit choice (Browne 2013). When cash transfers are provided to households, it is expected that households will change their behaviour such as eliminating the frequent use of negative coping mechanisms. These mechanisms include borrowing food from households, reducing the number of meals eaten daily, and eating food of lower quality (Maxwell, Caldwell, and Langworthy 2008).

3. Materials and methods

3.1 Study area

The Hunger Safety Net Program (HSNP) targeted poor people in ASALs of northern Kenya including the counties of Turkana, Wajir, Marsabit, and Mandera. The ASALs are characterised by frequent vulnerability to food insecurity, mainly associated with low and erratic rainfall coupled with high rates of evapotranspiration. Average annual rainfall, in the ASALs, ranges between 150 and 850 mm, with evapotranspiration rates double the amount of precipitation (Mati et al. 2006). Moreover, in the last two decades, the region has experienced extreme climatic conditions, which have affected the environment and livelihoods of the communities (Ibrahim and Abdulla 2015). The HSNP intended to support vulnerable households, in the target areas, by increasing their capacities to meet immediate essential needs, as well as encourage them to accumulate and retain assets. The program designers expected that the program would also have positive impacts on broader aspects of household well-being (Merttens et al. 2013).

The programme was implemented in two phases. Phase 1 started in 2009 and ended in 2012 and reached 60,000 households every two months with KES 2,150 per household (Merttens et al. 2013). The households collected the cash, from pay points of their respective convenience, using a biometric smartcard. Phase 2, funded by DFID and the government of Kenya, began in 2013 and ended in 2018. It aimed to reach the poorest 100,000 households with a monthly cash transfer of KES 2700. It also served another 180,000 households with periodic emergency transfers (Merttens et al. 2018).

3.2 Data

The HSNP panel data were obtained from the World Bank data catalogue (World Bank 2020). The data were available in four waves and covered responses from the community and household levels. The first wave captured baseline data collected in 2009. The second and third covered midline and endline surveys conducted in 2010 and 2012, respectively. The fourth, for data collected in 2016 was not used in this study because there was no proper household-to-household link with the other three waves. This study was based on the randomized design of the HSNP Forty-eight program sub-locations were selected from the pool of all HSNP sub-locations. The selection of the sub-location was by probability proportional to size (PPS). From the selected sub-locations each pair were randomly assigned between treatment and control at a public lottery event. In the selection of beneficiary households, three types of targeting mechanisms were implemented simultaneously within the treatment sub-locations: community-based targeting (The community is instructed to select those households that are most food insecure. Up to half of the community’s households are to be selected this way), dependency ratio targeting (All households in which a certain percentage of the members are older than 55, younger than 18, disabled or chronically ill are eligible), and a social pension approach (All members in the community over the age of 54 years were eligible to receive transfers). A simple random sampling was followed to select the treatment and control households. At the beginning of the HSNP, a total of 5108 households were selected. In the midline and endline surveys, a decision by HSNP stakeholders to reduce the sample size was made (Merttens et al. 2013). The final sample size for the endline survey round was 2436 households among which 1,224 were in the treated group and 1,212 control group households. This study focused on 1,224 treatment group households and 1,212 control group households for which there were observations at both baseline and endline. The treatment group received cash transfers (beneficiaries) while the control group did not receive cash transfers (non- beneficiaries).

3.3 Empirical approach

The paper adopted the FAO (2016) Resilience Index Measurement and Analysis framework (RIMA). The framework starts from a premise that resilience is not directly measurable and is, therefore, captured through proxies. The framework distinguishes two types of proxies: one is descriptive and serves to rank or target households, while the other is inferential and assesses the determinants of resilience. Accordingly, the RIMA is a two-part procedure, descriptive and inferential analysis. The descriptive analysis stems from the conceptualisation by Alinovi, Mane, and Romano (2008, 2010) that a household’s resilience is a complex concept dependent on a combination of factors, referred to as pillars as indicated by equation 3.1. These are social safety nets (SSN); access to basic services (ABS), adaptive capacity (AC), income and food access (IFA) and Assets (A) (FAO 2016). These pillars are considered observed endogenous variables in that they can cause and be influenced by resilience (FAO 2016). This combination of factors requires construction of an index, the Resilience Capacity Index (RCI). The pillars of resilience are generated separately; different variables are aggregated to compute each pillar. (3.1) $\mathrm{R}=\mathrm{f}(\mathrm{IFA},\mathrm{ABS},\mathrm{A},\mathrm{SSN},\mathrm{AC})$(3.1) FAO (2016) noted that potential endogeneity issues could arise during the construction of the resilience index using direct techniques of resilience measurement because some of the variables utilized in the construction of the index would be used as independent variables in the regression. In this paper, indirect approaches were more suitable due to the endogeneity issue that would skew the estimations and consequently the inferences made afterwards. The paper followed the approach by Asiimwe et al. (2020); Banda et al. (2016); Muricho et al. (2019) to construct the resilience index.

The steps followed to compute the household resilience index were first identifying the variables used to generate the resilience pillars – see Appendix 1. The second step involved the generation of the resilience index using the principal component analysis method (PCA). The resilience index, therefore, is the weighted sum of the factors generated by principal component analysis (PCA) as indicated in equation 3.2(3.2) ${\mathrm{R}}_{\mathrm{i}}={\mathrm{w}}_{\mathrm{IFA}}\mathrm{IF}{\mathrm{A}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{A}}{\mathrm{A}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{ssn}}\mathrm{SS}{\mathrm{N}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{AC}}\mathrm{A}{\mathrm{C}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{ABS}}\mathrm{AB}{\mathrm{S}}_{\mathrm{i}}$(3.2) , the weights are the proportions of variance explained by each factor. (3.2) ${\mathrm{R}}_{\mathrm{i}}={\mathrm{w}}_{\mathrm{IFA}}\mathrm{IF}{\mathrm{A}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{A}}{\mathrm{A}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{ssn}}\mathrm{SS}{\mathrm{N}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{AC}}\mathrm{A}{\mathrm{C}}_{\mathrm{i}}+{\mathrm{w}}_{\mathrm{ABS}}\mathrm{AB}{\mathrm{S}}_{\mathrm{i}}$(3.2) where: ${\mathrm{R}}_{\mathrm{i}}$ is the resilience index, $\mathrm{IF}{\mathrm{A}}_{\mathrm{i}}$ represents the income and food access, ${\mathrm{A}}_{\mathrm{i}}$ assets, $\mathrm{SS}{\mathrm{N}}_{\mathrm{i}}$ is social safety nets $\mathrm{A}{\mathrm{C}}_{\mathrm{i}}$ is the adaptive capacity and $\mathrm{AB}{\mathrm{S}}_{\mathrm{i}}$ is access to basic services. ${\mathrm{w}}_{\mathrm{IFA}},{\mathrm{w}}_{\mathrm{A}},{\mathrm{w}}_{\mathrm{ssn}},{\mathrm{w}}_{\mathrm{AC}},{\mathrm{w}}_{\mathrm{ABS}}$ are the weights for the resilience pillars. After the determination of the RCI, a min–max rescaling method was adopted to ensure that the resilience index lies between zero and one. Rescaling of the resilience index serves three purposes that include, easier regression interpretation, easier setting the thresholds that are common and cross-country valid, and impact evaluation. When impact evaluation is run against the resilience score it is possible to assess whether the score has increased by x per cent (Asiimwe et al. 2020; FAO 2016; Muricho et al. 2019). The paper used the panel fixed effects equation (3.3(3.3) $\mathrm{HR}{\mathrm{I}}_{\mathrm{it}}={\beta }_{\mathrm{o}}+{\beta }_1\mathrm{time}+{\beta }_2\mathrm{transfer}+{\mathrm{Z}}_{\mathrm{it}}\lambda +{\mathrm{a}}_{\mathrm{i}}+{\mu }_{\mathrm{it}}$(3.3) ) to estimate the effects of cash transfers on household resilience. The choice between panel fixed effects and random effects was based on a Hausman test for endogeneity to find out the model which gives unbiased estimates (Wooldridge 2016). (3.3) $\mathrm{HR}{\mathrm{I}}_{\mathrm{it}}={\beta }_{\mathrm{o}}+{\beta }_1\mathrm{time}+{\beta }_2\mathrm{transfer}+{\mathrm{Z}}_{\mathrm{it}}\lambda +{\mathrm{a}}_{\mathrm{i}}+{\mu }_{\mathrm{it}}$(3.3) where: $\mathrm{HR}{\mathrm{I}}_{\mathrm{it}}$ is the household resilience at time t, ${\beta }_{\mathrm{o}}$ is a constant, ${\beta }_1$ represents the effect of going from baseline to end line, ${\beta }_2$ represents the effects of cash transfers on household resilience, time is the time dummy, the transfer is a dummy variable equal to one for the beneficiary households and zero for the non-beneficiaries, ${\mathrm{Z}}_{\mathrm{it}}\lambda$ represents a vector of household characteristics and their coefficients which affect household resilience. These factors control for other observable differences across the households that could affect household resilience. ${\mathrm{a}}_{\mathrm{i}}$, are the unobserved household effects and ${\mu }_{\mathrm{it}}$ is the idiosyncratic error (time-varying error). The underlying assumption of the household fixed effects estimator is that the unobservable does not change at the household level and can therefore be differenced away (Tiwari et al. 2016).

4. Results and discussion

4.1 Descriptive statistics

Table 1 shows that the average household size in the four counties was about 6 for both beneficiaries and non-beneficiaries. This is a bit higher compared to the country’s average household size which is 3.9 (KNBS 2019). The large household size can be attributed to high poverty levels as the poor tend to have many children. The average age of the household head was 55.11 and 53.92 years for the beneficiaries and non-beneficiaries, respectively, where 69.72% were male-headed in the non-beneficiary households and 70.1% in the beneficiary households. The results in Table 1 indicate that majority of the households received food aid and only 5.15% of the beneficiary households participated in cash-for-work programs. This can be an indicator of food aid dependency. The major source of livelihood in the arid and semi-arid counties is pastoralism, 74.35% and 83.75% owned livestock in the beneficiary and non-beneficiary households, respectively. In pastoral areas land ownership in most cases, it is communal, from the results we find that only 15.28% and 11.8% owned land in the beneficiary and non-beneficiary households, respectively, these can be the few agro-pastoralist. The majority of the household heads had basic education with few advancing to a higher level of education. On average 94.85% of the beneficiary household heads had gone up to the primary level, 4.08% had gone up to the secondary level and 1.06% had gone to the tertiary. The low levels of education in the region may affect how households absorb or adapt to a given shock. The low levels of education can be attributed to marginalization, poverty and lack of great exposure to the importance of education. Only 6.78% and 7.18% of the house head had a disability in the beneficiary and non-beneficiary households, respectively.

Table 1. Demographic characteristics of the household head.

Variable	Beneficiaries	Non-beneficiaries
Percentage of the male-headed household	70.10	69.72
Percentage of households that received food aid	77.12	85.15
Percentage of households that participated in cash for work	5.15	3.88
Percentage of households that owned Livestock	74.35	83.75
Percentage of households that owned Land	15.28	11.80
Percentage of household heads with primary education	94.85	96.86
Percentage of household heads with secondary education	4.08	2.31
Percentage of household heads with tertiary education	1.06	0.83
Percentage of household heads with Disability	6.78	7.18
Mean household age	55.11	53.92
Mean household size	5.95	5.87

Table 2 shows variables used to generate the resilience index using the principal component analysis (PCA). PCA is about reducing the dimensionality of the data set consisting of a large number of interrelated variables while retaining the variation present in the data set as much as possible (Gambo Boukary, Diaw, and Wünscher 2016). The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.712 above the recommended 0.5 showing that the variables used in the PCA were adequate and conclusions can be drawn from the index constructed (Banda et al. 2016; Gambo Boukary, Diaw, and Wünscher 2016; Muricho et al. 2019; Wooldridge 2016).

Table 2. Factor loadings of variables used in principal component analysis.

Variable	Factor score
Number of livestock owned	0.0056
Land ownership	−0.0358
Food aid received	0.0607
Cash for work	0.0458
Other cash transfers	−0.1530
Education level	0.1880
Dependency ratio	−0.0138
Savings	−0.0404
Health expenditure	0.3409
Education expenditure	0.4982
Market access	0.0190
Log total expenditure	0.9733
Average food consumption	0.9515
Log food expenditure	0.9276
Chi-square	33001.763
Degrees of freedom	91
p-value	0.000
Kaiser-Meyer-Olkin (KMO)	0.712

The decision rule by Kaiser (1960) is common in most statistical packages and it is widely used to decide on factor retention (Henson and Kyle Roberts 2006). Bartlett's score of sphericity had a P value of 0.000 at a 5% level of significance indicating that the variables used were inter-correlated and that the correlations did not result from a sampling error (Asiimwe et al. 2020; Muricho et al. 2019). The principal components with eigenvalues greater or equal to one were retained and their factor loadings are reported (Kaiser 1960). In the discussion below the first-factor score was reported since it was more relevant in determining resilience. From Table 2 Access to basic services was defined by three variables that are; health expenditure, access to the market and education expenditure. Access to the market was measured in terms of minutes taken to reach the nearest market. Access to basic services is important in building household resilience. The factor scores for education expenditure, market access and health expenditure were positive. Access to basic services can be hindered by geographical location which increases the cost of accessing these services hence making households more vulnerable (Alinovi, Mane, and Romano 2008; Stifel and Minten 2008). Poor access to public services affects the capacity of the household to manage risks and respond to a crisis (Alinovi, Mane, and Romano 2008). Livestock is a major source of livelihood in pastoral communities, the results indicate that the variable number of livestock owned was positive, an implication that households with large livestock herds are more likely to be resilient. Land ownership had a negative factor score (−0.0358), this can be explained by the fact that in most pastoral communities land is communally owned and this may make households prevalent to conflicts over the communal resource and hence vulnerable to shocks (Tesso, Emana, and Ketema 2012). The Social safety net pillar was estimated by variables such as whether the household received food aid, household participated in cash for work and whether the household received other cash transfers. The factor scores for food aid cash for work were 0.0607 and 0.0458, respectively. Social safety nets are made to protect households from effects of persistent drought by supporting their livelihoods and helping them meet their immediate needs (Shiferaw et al. 2014). Social safety nets act as insurance mechanisms before the occurrence of a shock, or can be activated after a shock has taken place (Brück, d’Errico, and Pietrelli 2019). Adaptive capacity provides the household with the capacity to absorb the shock (Bowen et al. 2020). Adaptive capacity was determined by dependency ratio, education level, average consumption and savings, their factor scores were −0.0138,0.1880, 0.9515, and −0.0404, respectively. Households with better education are likely to make better decisions regarding building a broad asset base, livelihood diversification and ways to manage shocks (Asiimwe et al. 2020). The implication of the negative factor score of the dependency ratio is that a household with a high dependency ratio regularly bears a relatively high financial burden, which could limit their capacity to make savings for security against livelihood shocks (Amadu, Armah, and Aheto 2021; Muricho et al. 2019). The income and food access pillar was explained by food expenditure and total household expenditure, their values were in log form to minimize the possible outliers present. Total household expenditure was used as a proxy of total household income, the results indicate that total household expenditure had the highest positive factor score of 0.9733. Households with wide range of the sources of income are likely to be more resilient, income enables households access basic services and build a broad asset base to cushion them against shocks (Asiimwe et al. 2020; Muricho et al. 2019).

4.2. Cash transfers effect on household resilience

The resilience index was scaled to range between 0 and 1 for easier interpretation (Asiimwe et al. 2020; FAO 2016; Gambo Boukary, Diaw, and Wünscher 2016; Muricho et al. 2019). The results in Table 3 indicate that the average resilience index was 0.36 which is below the average. The households who received cash transfers on average had a higher resilience index of 0.37 compared to the non-beneficiaries’ 0.35. Cash transfer acts as a tool to help households from falling into further destitution. Households living in Wajir county on average had the highest resilience index than those living in other counties, and households in Turkana had the least resilience index. Male-headed households had a higher resilience index compared to the female headed households. The t test p values in Table 3 indicate that resilience index was significantly different by gender, beneficiary category and across the counties.

Table 3. Resilience capacity index summary statistics in the four counties.

Variable	Mean	p-values
Resilience index (overall)	0.36
Mandera	0.38	0.00
Turkana	0.31	0.00
Wajir	0.41	0.00
Marsabit	0.33	0.00
Non-beneficiaries	0.35	0.00
Beneficiaries	0.37	0.00
Male headed household	0.37	0.00
Female-headed household	0.33	0.00

The results for the fixed effects model are shown in Table 4. The choice between random effects and fixed effects was based on the Hausman test. After conducting the Hausman test the p-value was 0.000 which was very significant at 5% and this was a conclusion that the fixed effects model was more efficient. The R-squared reported in Table 4 indicates that 97% of the variations were explained by the model hence a good fit. The results in Table 4 indicate that cash transfers have a positive effect on household resilience at a 5% level of significance consistent with d’Errico et al. (2020) results. In the households which received cash transfers, their resilience was positively influenced that is, if cash transfers are increased by one unit household resilience will increase by 2% on average. The level of education had a positive effect on household resilience and was significant at 5%, this is an implication that households with at least a member with a higher level of education are more resilient. Household heads with higher education levels are expected to have improved decision-making capabilities and increased access to investment opportunities (Asiimwe et al. 2020). The results are consistent with Banda et al. 2016; Keil et al. (2008); Tesso, Emana, and Ketema (2012) but they contradict Asiimwe et al. (2020) results which found a negative relationship between education level and household resilience. The results indicate that access to credit had a positive effect on resilience. This implies that households which could not access credit are less resilient. Access to credit facilities during the time of crisis is not easy as financial institutions are not willing to give loans hence households explore other exploiting options which might interfere with their resilience capability (Tesso, Emana, and Ketema 2012). The average household size was 6 higher than the national household size as indicated in Table 1. The results show that there was a significant positive relationship between household size and household resilience. The results are consistent with Banda et al. (2016) and Keil et al. (2008) but contradict Kasie et al. (2017) study on household resilience to food insecurity which found a negative relationship between household size and resilience. Large households are likely to be more resilient than small households. This paper agrees with Banda et al. (2016) that large households are more likely to have diversified sources of income as compared to smaller households and hence more resilient to shocks such as droughts. The log of total expenditure had positive significant effects on household resilience, these results are consistent with Asiimwe et al. (2020); Muricho et al. (2019); Opiyo, Wasonga, and Nyangito (2014). Log total expenditure is a proxy for total income since households tend find it hard to disclose their income. Households with high spending are likely to have high levels of income. The coefficient of log total exp shows that as household income increases by one unit household resilience will increase by 16.85. The results indicate that male headed households were more resilient than the female headed households. The results contradicts Muricho et al. (2019) who found that female headed households were more resilient than the male headed households. However, the results are consistent with Opiyo, Wasonga, and Nyangito (2014); Tesso, Emana, and Ketema (2012) results who found that female-headed households were less resilient as a result of bias in resource allocation and decision-making. These discrepancies are a result of different methodologies used by authors. Table 4 shows that the effect of other transfers was negative meaning they were less resilient than those who didn’t receive. These other transfers are cash received from relatives and friends. The negative relationship can be attributed to the nature of these transfers since they not regular and predictable, hence households may use them to smooth consumption rather than investing in productive activities.

Table 4. Panel fixed effects model results for effects of cash transfers on household resilience.

Resilience	Coef.	Std. Err.	p-value
Cash transfer	0.226	0.05	0.00
Location	−0.022	0.025	0.375
Year	0.026	0.018	0.157
Household age	0.001	0.002	0.467
Household size	0.116	0.012	0.00
Gender of the household	0.38	0.053	0.00
Credit access	0.369	0.041	0.00
Education level	1.085	0.087	0.00
Other transfers	−0.677	0.087	0.00
Log total exp expenditure	16.849	0.043	0.00
Constant	−168.761	36.321	0.00
R-squared	0.974	N	6039.000
F-test	14712.517	Prob > F	0.000

N represents the number of observations.

5. Conclusions and recommendations

This study used a panel fixed effects estimator to determine the effects of cash transfers on household resilience. From the study, a resilience index was generated using principal component analysis. The resilience index was determined by five pillars that is assets, adaptive capacity, income and food access, social safety nets, and access to basic services (d’Errico et al. 2020; FAO 2016; Gambo Boukary, Diaw, and Wünscher 2016; Mekuyie, Jordaan, and Melka 2018). The resilience index was then rescaled to range between zero and one for easy interpretation in regression analysis. Cash transfers are important in cushioning households against shocks. The results indicated that cash transfers, access to credit, gender, size of the household and income proxied by household expenditure had a significant positive effect on household resilience. The beneficiaries were more resilient than the non-beneficiaries and this can be attributed to the cash transfers received. From the results, we can conclude that regular and predictable cash transfers can help poor and vulnerable households from resulting in negative coping strategies and falling further into destitution. The study was done in an area which experienced frequent droughts during the time the transfers were given. Cash transfers, therefore, can be used as a tool to help households make adjustments to respond to climate-related shocks. The study showed that large households were more resilient than small households. This study recommends that the amount of cash transfers should be given depending on the size of the households and poverty levels to achieve better results. Cash transfers provided to improve household resilience to climate shocks should have some requirements that are geared towards improving some of the pillars of resilience such as asset accumulation, and investing in productive activities this enables the household to graduate from cash transfer dependency and become more resilient. The study further recommends that cash transfers should be accompanied by modifications of policies geared towards improving climate resilience in the ASALs. For future studies, researchers need to examine intra-household decisions making on budget allocation and how this can influence household resilience in the context of cash transfer beneficiaries.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by African Economic Research Consortium (AERC): [Grant Number AE/TG/19-02(Award 1310)].

Appendix

Table A1. Variables used to determine the resilience index.

Pillar	Variable Type	Measurement
Access to basic services (ABS)
Education expenditure	Continuous variable	Measured as total household education expenditure
Health expenditure	Continuous variable	Measured as total household health expenditure
Access to the nearest market	Continuous variable	Measured as the time in minutes taken to arrive at the nearest market centre.
Social safety net (SSN)
Cash for work	Dummy variable	Equals 1 if the household receives cash for work and 0 otherwise
Food aid	Dummy variable	Equals 1 if the household received food aid and 0 otherwise.
(Other cash transfers from relatives, friends)	Dummy variable	Equals 1 if the household receives a cash transfer and 0 otherwise;
Adaptive capacity (AC)
Education level	Discrete variable	Measured by the highest education grade the household head has attained.
Average daily consumption	Continuous variable	Average daily consumption is measured by taking the average food expenditure in seven days.
Dependency ratio	Continuous variable	The ratio between the total number of household members below 15 years or above 65 to the number of household members between 15 and 65 years.
Savings	Dummy variable	Equals 1 if the household saves and zero otherwise
Assets (A)
Land ownership	Dummy variable	Equals 1 if the household owns the land and zero otherwise
Livestock owned	Continuous variable	Measured by the number of livestock owned by the household.
Income and food access
Food expenditure	Continuous variable	Total monthly household food consumption expenditure
Total household expenditure	Continuous variable	Total monthly household consumption expenditure