1 Introduction

Climate change poses various threats to different countries, and its threat to food security, economic stability, and social disruption in African economies is enormous (African Development Bank 2022). Essentially, the detrimental impact of climate change in Africa becomes a major issue as about 85% of the continent’s inhabitants rely on rainfall for agricultural and other related activities for their livelihood (Moyo 2016). Recent research findings indicate that smallholder farmers are directly impacted by climate change. For instance, smallholder farmers are affected by shorter growing seasons, less reliable precipitation patterns, increasing land loss due to aridity and desertification, and poor soil quality (Brown and Funk 2008; Morton 2007). In Sub-Saharan Africa (SSA), poor and unsustainable agricultural practices are used by farmers to mitigate against the effects of climate change (Grabowski et al. 2016). However, Fisher et al. (2015) and Kassie et al. (2015) note that such practices often result in decreased crop yields, a higher rate of food insecurity, and poverty. Thus, effective adaptation measures are an integral part of any climate change policy response, as they can help reduce vulnerability (Osman-Elasha et al. 2006).

Accordingly, the Climate-Smart Agriculture (CSA) was introduced by the Food and Agricultural Organization (FAO 2013) at the 2010 Hague Conference on Agriculture, Food Security, and Climate Change as a potential means of decreasing GHG emissions, adapting and building resilience to climate change, and increasing agricultural productivity and incomes. Climate resilience continues to be at the core of CSA practices, and if African governments do not cooperate, the principle of “leaving no one behind” may be compromised. Currently, only Southern and North Africa exhibit the most resilience to climate shock, as observed in the African Development Bank (2022) report in Fig. 1. North Africa has a mean climate resilience score of 63.5, while Southern Africa has a mean score of 43.6. Regional and national differences remain incredibly wide. The negative consequences of climate-related disasters have persisted in West Africa (mean score of 25.6 climate resilience index), East Africa (mean score of 25.0 climate resilience index), and Central Africa (mean score of 20.6 climate resilience index). When compared to the Democratic Republic of the Congo, Chad, Central African Republic, South Sudan, Somalia, and Libya, Mauritius, Cabo Verde, Tunisia, Morocco, Botswana, and South Africa had the six highest country climate resilience index scores. This suggests a connection between climate-related disasters and economic progress.

Fig. 1
figure 1

Source: Adapted from AfDB (2022)

Status of Climate Resilience Index (CRI) among African countries, average score (2010–2019).

Full size image

To adapt to climate change by 2050, the African Development Bank (2022) report estimates that the continent will need to spend at least $50 billion every year. Sub-Saharan rural farmers’ perspective of climate change is a major factor influencing the adoption of adaption techniques (Juana et al. 2013). So, it’s a good question for researchers and policymakers to rethink whether and how farmers’ views on climate change influence their adoption of CSA practices. Therefore, this study investigates two important questions: (1) What factors drive rural farmers’ understanding of and their response to the perceived effects of climate change? and (2) What accounts for the adoption of CSA methods among rural farmers? We answer these questions using data from the Ethiopian Socioeconomic Survey (ESS) and the Nigerian General Household Survey (NGHS) of 2018–2019.

Our investigation focuses mainly on rural smallholder farmers in Ethiopia and Nigeria who have low perception of and are prone to climate change. An empirical concern in studying what factors influence farmers’ perceptions of and responses to changes in climate is the selection bias that stems from the non-random selection of farmers. This means that farmers who are cognizant of climate change may likely take proactive steps to adapt to those conditions compared to those without such perception, which creates a self-selection bias. To address this issue, we employ a two-stage Heckman selection framework to identify the elements impacting farmers’ perspectives and subsequent adaptation choices. Furthermore, we employ both the multinomial and binary logistic regressions in assessing the determinants of specific adaptation measures such as water management, chemical management, land management, and organic fertilizer management. Our empirical work not only reveals some fascinating insights into what influences rural farmers’ decisions and actions regarding adapting to climate change, but also reveals variations in the importance of our explanatory variables in explaining perceptions and adaptation to climate change between Ethiopian and Nigerian rural farmers. We find that in both Ethiopia and Nigeria, household income and plot size influence farmers’ adaptations to climate change. However, farmers with bigger plots who run the risk of massive production loss are more likely to adopt measures of coping with climate change. We show that in both Ethiopia and Nigeria, rural farmers’ adaptation decisions are heavily influenced by agricultural extension programs and community social networks.

Overall, our research contributes significantly to the growing research on how rural farmers understand and respond to climate change. This is especially true in terms of the need to scale up accessibility and decrease the cost of appropriate technologies. First, we are unaware of any comparative study looking into the issue of climate change between Nigeria and Ethiopia. This is important given the strategic location and economic structures of both countries and the significance of the agricultural sector to both countries’ socioeconomic sustenance. Agriculture has been the mainstay in Ethiopia and Nigeria and has garnered several government efforts in developing the sector in both countries.

Second, we argue strongly for the dissemination of climate-related information in both Ethiopia and Nigeria. There is a current information gap between government institutions charged with getting weather reports and farmers out in the country. The latter is affected by a lack of accurate climate data, which results in a poor harvest and low levels of productivity. For this reason, it is essential to have access to information on the wide adoption and promotion of CSA practices as an efficient means of combating climate change. Lastly, we differ from most existing studies in our choice of data. The extant literature has mainly focused on self-administered surveys that often limit their coverage, see (Fentie and Beyene 2019; Mashi et al. 2022; Onyeneke et al. 2021; Saguye 2017). This may be as a result of challenges in generating relevant climate-related variables from the existing integrated surveys on agriculture. Thus, we differ by attempting to explore the existing surveys for Ethiopia and Nigeria, thereby generating a fairly large sample of over 2,500 per country. This makes our study more representative of the Ethiopian and Nigerian rural farmers and provides a justification for a broader policy recommendation. Policymakers in Sub-Saharan Africa, especially in Ethiopia and Nigeria, need a clear understanding on the determinants driving uptake of CSA practice to implement evidence-based policies that would allow for the rapid scaling up of mechanisms for adapting to climate change.

Following this, the rest of the paper is presented as follows. Section 2 reviews both theoretical and empirical literature. The data and methods are presented in Sect. 3. We present our empirical findings in Sect. 4. Our study finishes with policy suggestions in Sect. 5.

2 Theoretical and empirical review

Theoretically, farmers’ views on climate change and adaptation are shaped by their own experiences, the knowledge they have access to (Reilly and Schimmelpfennig 1999; Weber 2010), the risks they are willing to take, and the potential repercussions of those decisions (Fuster 2002; Kahneman et al. 1990). The adoption of Climate-Smart Agriculture technologies is embedded within the broader technology adoption frameworks. Hence, our study is influenced by Rogers’ (1962) seminal work on the adoption model of agricultural innovations and Schultz’s (1975) farmers’ behavior toward the usage of new technology. The Food and Agricultural Organization (2013) identifies “climate-smart agriculture" as the practices that strengthen farmer’s resilience, increase productivity and income, and mitigate climate change by decreasing emissions of greenhouse gases. Thus, climate mitigation can lessen losses, boost outputs, and maximize profits; accordingly, the way that conventional farmers see climate change in an adoption model is crucial for CSA (Deressa et al. 2009; Gbetibouo 2009; Maddison 2007). The transition from traditional farming methods to technology adoption and adaptation in Africa has raised the fundamental question of why farmers need to embrace new technologies to fight climate change.

On the empirical front, several studies have confirmed the theoretical postulations that socioeconomic factors, environmental conditions, institutional settings, and community capital are robust contributors to climate change perception and adaptation among farmers. For instance, Teklewold et al. (2019) show that socioeconomic, institutional, and environmental factors influence CSA practice uptake. Similarly, Feder et al. (1985) conclude that several exogenous variables, including household characteristics, socioeconomic and physical conditions, play a role in farmers’ adaptation decisions. The adoption of climate smart agricultural practices is also linked to improved productivity and welfare. Recent evidence from Meshesha et al. (2022) finds that when farmers adopt CSA innovations like agroforestry, compost, soil and water conservation, intercropping, improved variety, crop residue management, crop rotation, and row planting, farm greenhouse gas emissions and agricultural vulnerability to climate change decrease while crop productivity increases. Hence, recent research (Anuga et al. 2019; Jellason et al. 2021; Kangogo et al. 2021; Martey et al. 2020; Mashi et al. 2022; Obi and Maya 2021; Zakari et al. 2019) suggests that CSA practices should be expanded beyond the conventional method used by rural farmers in Sub-Saharan Africa to promote resilience against climate change.

A strand of empirical studies examining farmers’ perception and adaptation to climate change in Ethiopia has been largely region specific. For instance, to assess farmers’ perceptions of climate variability and use of CSA practices, Saguye (2017) conducted interviews with 138 households in the Geze Gofa district of Southern Ethiopia. He finds that 88.7% of farmers believed that temperature increased and that about 90% recognized that rainfall drastically decreased, resulting in frequent droughts. Fentie and Beyene (2019) analyze the responsiveness of rural households’ welfare to climate-smart agricultural techniques such as row planting in Ethiopia’s Amhara region. They find that row planting technology has a considerable and positive effect on both individual consumption and agricultural output per hectare. Finally, they show that the marginal advantage of row planting grows as farmers employ climate-smart planting technologies. By focusing on 734 farm households in Dabus watershed, Northwest Ethiopia, Asrat and Simane (2018) submit that education, age, experiences, income, extension service, and climate information among others affect climate change perception and adaptation. Furthermore, the authors find that farmers in both the wetlands and drylands are affected by both common and distinct factors. These findings largely echo the evidence from Deressa et al. (2009) on the relevance of gender, age, credit, climate information, wealth, extension service, and community integration in fostering adaptation to climate change in the Nile-basin of Ethiopia.

Next, empirical evidence from several studies on Nigeria is also limited by their region-based focus in attempting to understand the determinants of CSA adoption in the country. For instance, Mashi et al. (2022) interviewed 491 Nigerian farming households in Abuja in an attempt to identify the factors that affect farmers’ familiarity with climate-smart agricultural technologies. They showed that farmers with more education, more secure income and resources, more experience adapting to climate change, and a better understanding of the local ecosystem were the most knowledgeable about CSA adaptation methods. In their cross-sectional analysis of 347 rice farmers’ responses to changes in climate, Onyeneke et al. (2021) investigate factors such more rainfall, shorter wet periods, more frequent floods, higher average temperatures, and the use of CSA practices. Difficulties in accessing inputs, exorbitant labor costs, a lack of understanding about the weather, and insufficient extension services were identified as barriers to the widespread implementation of CSA practices among farmers. To examine how land ownership status and agricultural policies influence the practice of sustainable agriculture, Adesida et al. (2021) studied 288 farmers in Southwestern Nigeria. The authors show empirical evidence indicating that land ownership encourages the adoption of mulching techniques. Additionally, farmers who participate in agricultural programs such as the Farmers’ Field School and the World Bank’s Fadama scheme were more likely to adopt crop diversification techniques. In Anambra, a Southeastern state in Nigeria, Anugwa et al. (2022) assessed 96 rice farmers’ willingness to pay for CSA technologies. Interestingly, 70% of the farmers (several of whom are poor) do not adopt CSA, shifting the cost of CSA technologies provision to the government. However, farmers who were willing to pay for CSA find $115 annually reasonable. Further examination, however, reveals that access to credit raises the probability of adopting water-smart, energy-smart, and knowledge-smart CSA technologies. The findings of the study thereby suggest that farmers’ income may play a crucial role in influencing climate change adaptation.

Despite the preceding evidence of the determinants of climate smart agriculture adoption among farmers, several factors have been identified to pose significant hindrance to the widespread adaptation to climate change. Chief among such barriers is institution, which plays a part in the CSA practices of rural farmers and may offer opportunities for successful expansion of CSA activities (Totin et al. 2018). The recent systematic review and analysis of studies on CSA implementation and use in Africa by Barasa et al. (2021) thoroughly and persuasively makes the case that various African countries adopted the CSA to lessen the difficulties associated with low agricultural productivity. However, the authors find that many of these countries lacked climate-smart agriculture investment plans (CSAIPs). Their findings might imply that the CSAIPs’ framework is essential for a successful CSA implementation. Similar to this, Azhoni et al. (2017) find that climate change adaptation in India are hindered by institutional factors, particularly in the form of systemic issues and bureaucratic bottlenecks. In a related study, (Makate 2019a) examines smallholder farmers’ adoption of CSA innovations in Africa and finds that CSA practice adoption is low, despite farmers’ desire to accept new, creative approaches to increase productivity. To scale up CSA techniques among smallholder farmers in Africa, the author suggests that smart institutional and governmental actions are essential. This may aid in easing farmer difficulties, easing adoption restrictions, and enhancing agricultural production while addressing food in security. Furthermore, Makate (2019b) examines the relevance of regional institutions and indigenous knowledge in planning for climate change adaptation, and the author provides additional information on how successful CSA innovations can be adopted and scaled up in sub-Saharan Africa. The author asserts that involving local institutions and indigenous knowledge is essential for hastening CSA innovations adoption in smallholder farming. By incorporating local institutions and indigenous knowledge, it is possible to improve the flow of information, which in turn facilitates the mobilization of resources, the building of networks among rural farmers, and the emergence of stair-leaders in the fight against climate change (Azhoni et al. 2017; Bisaro et al. 2018; Mubaya and Mafongoya 2017; Ogunyiola et al. 2022; Tuihedur Rahman et al. 2021).

Three essential gaps are discernable from the foregoing review of the literature on climate change perception and adaptation, especially in Nigeria and Ethiopia. One, the existing studies mainly focused on specific regions within both countries, providing evidence on CSA that may suffer from external validity, thereby precluding the generalization of their findings to other regions within the sample country. Moreover, such an empirical approach has resulted in small sample sizes. Two, the within-country investigations limit the comparability of climate change perception and adaptation between countries. Accordingly, there are no studies comparing farmers’ outlook on climate change between Nigeria and Ethiopia, two of the most populated African countries where agriculture plays a key role in their development trajectories. Three, the region-specific examinations often examine farmers in urban areas, despite the large rural-driven agricultural activities in both Nigeria and Ethiopia. Addressing these gaps would require nationally representative data sets and a comparative study that will provide new insights into how rural farmers observe climate change and their coping mechanisms. Thus, to decrease climate vulnerability in Sub-Saharan Africa and ensure food security, it is essential to understand farmers’ perceptions and behaviors. This will help the continent achieve the Sustainable Development Goals. Stemming from the above review of literature, we hypothesize the following:

Hypothesis 1

Farmers with higher income are more likely to adopt CSA practices.

Hypothesis 2

Farmers with a high degree of risk are more likely to adopt CSA practices.

Our first hypothesis basically claims that farmers with higher incomes are better able to afford means of accessing information on climate change and investments in adaptation strategies. Several empirical submissions indicate that income positively explains adaptation decision by allowing farmers to bear more risks and afford the cost of various adaptation techniques (Deressa et al. 2011; Onyeneke et al. 2018; Opiyo et al. 2016). Moreover, one of the reasons for low uptake of CSA technologies is the low income of farmers. For instance, Anugwa et al. (2022) show that over 20% of Southeastern Nigerians who chose not to adopt CSA are from low income households. For the second hypothesis, our study measures a high degree of risk by the size of the farm cultivated by farmers. Effectively, farmers with large farmlands are more exposed to the risk of massive climate change-induced losses. The extant literature contends that the size of the farm could either increase the likelihood of adaptation or weaken farmers’ motivation to adapt (Teklewold et al. 2019). For instance, poor farmers with large family lands may find it expensive to bear the cost of adapting new technologies for a large field. On the other hand, wealthy farmers are likely to invest in adaptation measures that improve their yields, thus taking advantage of the large farmlands. We follow the later reasoning and add that regardless of farmers’ income, the risk of massive loss would incentivize farmers to adapt to climate change.

3 Materials and methods

3.1 Study areas

This study focuses on Nigeria and Ethiopia to understand how rural farmers in Sub-Saharan Africa conceive changes in climate and their methods of adaptation given their strategic locations and positions within the western and eastern regions of Africa, the pertinence of agricultural activities to the socioeconomic sustenance of both nations, and their exposure to the consequences of climate change. For instance, the importance of the Nigerian agricultural sector begins with its sizeable employment creation. Despite its recent decline, the sector still provides employment for about one-third (35%) of the Nigerian labour force (World Bank 2023), 70% of whom are rural dwellers (Odozi and Adeyonu 2021), indicating the sector’s importance to the sustenance of rural economy. Additionally, the sector has significantly contributed to the GDP of Nigeria, accounting for about 26.2% in 2020, second only to the service sector but higher than the manufacturing sector (National Bureau of Statistics 2021). However, despite its relevance, the sector faces several challenges related to productivity, mechanization, insecurity-induced activity decline, and more recently, climate change effects. For instance, episodes of flood and drought have been tied to agricultural production losses in Nigeria (Elum and Momodu 2017). Furthermore, the increasing humidity around the northcentral part of the country has had adverse effect on rice production (Ayinde et al. 2013). Beyond the direct effect of climate change on food production and security, it has also had a telling effect on human security and increased competition for resources, especially among farmers and herders (Alaanuloluwa Ikhuoso et al. 2020; Ani et al. 2022). In an effort to bolster the productivity and job creation of the agricultural sector, the Nigerian government has implemented several policies and programs, including the National Fadama Development Project, farmer field schools, growth enhancement support system, e-wallet platforms, the agricultural transformation agenda, and the agricultural promotion policy (Adesida et al. 2021; Federal Ministry of Agriculture and Rural Development 2016).

Similarly, the Ethiopian agricultural sector has been crucial to the growth of the country’s economy, contributing about 43% of GDP and employing 90% of rural dwellers, placing the sector as the backbone of rural livelihood and a driver of the national economy (Biru et al. 2020). However, the relevance of the agricultural sector to the sustainability of life and employment provision in Ethiopia has been persistently threatened by various impacts of civil unrest and climate change. Notably, the country has been susceptible to challenges of drought, flood, and severe variation in rainfall over the years. Boka (2017) opines that the commercialization and market integration of smallholder farmers in Ethiopian communities prone to drought and flood are often hindered by reduced farmlands, degradation, and difficulty in accessing irrigation systems. The climate change challenges further extend to the depletion of crop stocks (especially perishable produce) and livestock, making the livelihood of rural smallholders unbearable and exacerbating poverty levels. Furthermore, the climate change impacts is also witnessed in the inadequate resource availability that fosters competition and ethnic conflict among farmers and pastoralists communities (Sax et al. 2023). Similar to Nigeria, the Ethiopian government has over the years implemented measures to improve the agricultural sector, including the agricultural-led industrialization scheme, successive growth and transformation plans, agricultural growth plan, and market liberalization policies (Shikur 2020; Welteji 2018).

Despite the aforementioned effects of climate change in both Nigeria and Ethiopia, awareness among rural dwellers have been low when compared to urban dwellers. For instance, Fig. 2 shows the awareness of climate change among urban and rural dwellers in Nigeria and Ethiopia based on the Afrobarometer round 8 surveys (Afrobarometer 2020a, b). Notably, only 25% of rural respondents in Nigerians claim any awareness of climate change compared to the 39% of urban respondents. On the other hand, both rural and urban dwellers in Ethiopia fair better than Nigerians in the survey. For instance, 58% of urban respondents and 46% of those in rural areas indicate some level of awareness of climate change. Although there may be limitations to the Afrobarometer survey in terms of how climate-related questions are posed to respondents, the low perception especially among rural dwellers (majority of whom are in agriculture) raises the concerns of what determines farmers perception of climate change and what factors influence the various measures they employ to mitigate its effects.

Fig. 2
figure 2

Source: Authors’ Creation from Afrobarometer Round 8 Surveys in Nigeria and Ethiopia

Climate change awareness in Nigeria and Ethiopia (2020–2021).

Full size image

3.2 Data sources

We curated data from the nationally representative Ethiopian Socioeconomic Survey (ESS) and Nigeria’s General Household Survey (NGHS) conducted between 2018 and 2019 in Ethiopia and Nigeria alongside six other countries (Mali, Niger, Tanzania, Burkina Faso, Malawi, and Uganda) as part of the Living Standards Measurement Study—Integrated Surveys on Agriculture (LSMS-ISA) project of the World Bank (Central Statistical Agency 2021; National Bureau of Statistics 2019). In both countries, we only focus on the fourth waves of the surveys to maintain consistency and comparability of the most recent data available. For instance, the fourth wave of the ESS is not an extension of the previous waves but a new survey serving as the basis for subsequent ones (Central Statistical Agency 2021). This precludes us from using the full panel information from waves one to four in Ethiopia and as a consequence, for Nigeria too. Despite the panel information limitation, the wave 4 of the ESS contains about 6894 households across all regions of Ethiopia and across 565 enumeration areas (EA). The ESS wave 4 was designed to capture information on five broad categories, including the post-planting agriculture, post-harvest agriculture, livestock, household, and community and administered by visiting the households three times between 2018 and 2019. The first visit was from September to December of 2018, the second one between February and March of 2019, and the last enumeration between June and August of 2019 with each visit collecting information on different topics. Furthermore, the survey used a two-stage stratified probability sampling approach, consisting of a simple random sampling in the first stage (EA level) and a systematic random sampling in the second stage (household level) to select EAs and households.Footnote 1

On the other hand, the wave 4 of the Nigerian general household survey (NGHS) contains panel information linking it to previous waves. However, we are only able to use the wave 4 to be consistent with the ESS wave 4 and for appropriate comparison of farmers across both countries. The NGHS comprises 5000 households sampled across 500 enumeration areas (EAs) as the primary sampling units, using a two-stage probability sampling technique. Unlike the ESS, the NGHS was administered only twice during the survey period—the post-planting phase from July to September 2018 and the post-harvest phase from January to February 2019—using the World Banks’s Computer Assisted Person Interview (CAPI) survey software. The resulting dataset from the two phases of questionnaire administration were broadly classified into the household, community, and agricultural (post-planting, post-harvest, and livestock) information.Footnote 2 In line with our research objective of studying the rural farmers in Nigeria and Ethiopia due to their high engagement in agriculture, we restricted our data extraction to only the rural areas of both countries, yielding a sample of 2,899 and 2,862 unique households in rural Ethiopia and Nigeria, respectively.

The final survey datasets from Ethiopia and Nigeria largely contain similar information at the individual, household, and community levels, while also providing important information to assess climate change-related issues. Specifically, we extracted information on climate change perception and adaptation from several agricultural modules of the final dataset in both countries. For the climate change perception, farmers were asked questions related to crop damage or hindrance to planting such as “What was the main cause of the damage on [crop]?” and “Why did HH not harvest [crop] from [plot] during 2018/2019 agricultural season? (drought, flood, pest)". During the data extraction, we coded farmers’ observation of heavy rain, low rain, drought, flood, and pests infestation (insects, locust) as their perception of climate change effects. Thus, rural farmers who experienced changes in rainfall, flood, drought and pest infestation were coded as 1 while those who responded negatively were coded as 0.

For climate change adaptation, we extracted information on farmers’ practices for preventing crop losses and land damage from the post-planting and post-harvest modules. Farmers who responded yes to the questions “Did you use chemical fertilizers (any inorganic fertilizer) on any one of your crop field?" and "Did you use any (fungicide, herbicide, and pesticide) to prevent damage of [crop] on [field]?” were coded as 1 for practicing chemical management and 0 if otherwise. Secondly, farmers who said yes to “Have you participated in watershed activities in your community?” and “Is [field] irrigated?” were coded as 1 for practicing water management and 0 if otherwise. Third, we explore the possibility that farmers may use organic fertilizers to prevent climate-related crop damage and improve yield by coding organic fertilizer management as 1 if farmers responded positively to “Do you use any (compost, manure, any other organic fertilizer) on [field] in this agricultural season?” and 0 if otherwise. Lastly, to identify farmers adaptation to flooding and rainfall variation, we identified farmers as practicing land management if they respond positively to questions like “Is [Field] prevented from Erosion?”, “Do you exercise crop rotation on your land holding?”, “is this crop a fruit tree crop?”, and “Was the area planted with [crop] on [field] pure stand or mixed?” These questions thus form our basis for climate change perception and adaptation for the Ethiopian and Nigerian rural farmers.

3.3 Conceptual framework and empirical strategy

Adapting to climate variability centers on the deliberate decision by farmers to mitigate against the effects of changes in climatic conditions on agricultural activities and on farmers’ welfare. Thus, adaptation can be framed under the random utility framework (Makate et al. 2017) where a farmer decides to adopt a certain method or group of methods of coping with the changes in climate so as to maximize utility. In choosing the appropriate adaptation measure, the farmer compares the utility of several measures from the list of competing alternatives against the decision of not adopting any measure at all (Oyawole et al. 2020). In our context of the Ethiopian and Nigerian farmers, this expected utility of adopting a technology or method of mitigating against climate change may pertain to increased farming yield and profit, less crop loss, land preservation, and general household income and welfare upliftment. However, adaptation decision may further depend on whether the farmer has any perception regarding previous climatic changes (Waibel et al. 2018). Therefore, climate change perception becomes a central feature of the farmer’s adaptation decision. The choice of methods to adopt then becomes conditional on what shapes farmers’ perceptions.

The literature on climate change perception and adaptation alludes to the importance of other several factors that shape farmers’ perceptions and influence their decision to adopt measures of dealing with climate change. These factors include farmers’ socioeconomic features such as gender, marital status, age, and literacy (Oyawole et al. 2020); household level characteristics such as the size, income, and location (Anugwa et al. 2022); farmland or plot characteristics such as hectares (Makate et al. 2017). However, modelling adaptation decision or choice of a farmer poses a selectivity problem since farmers who understand the dangers of climate change may self-select into adaptation measures. Similar concern was raised by Amadu et al. (2020) in their model of participation in Climate-Smart Agricultural program in southern Malawi and in several other studies (Opiyo et al. 2016).

Therefore, we adopt the Heckman selection model to address the potential selection bias issue. The bias correction method of Heckman (1976) allows us to concurrently estimate the two-step decision of the farmer such that in the first step, we estimate the likelihood of perceiving climate change predicted by exogenous variables; this first stage is called the selection equation. In the second stage (the outcome equation) we estimate the decision to adopt climate change coping techniques only for farmers who report positive values for perception. This technique has been widely adopted in the climate change perception and adaptation literature (Asrat and Simane 2018; Deressa et al. 2011; Okonya et al. 2013; Opiyo et al. 2016; Waibel et al. 2018).

Accordingly, the Heckman selection framework is specified as below:

$$y_{1i}^* = x_{1i} \beta_1 + \varepsilon_{1i} ,$$
(1)

From Eq. 1, \(y_{1i}^*\) is a latent variable observed only when \(y_{1i} > 0\). Equation (1) is the outcome equation that models adaptation decision or practice conditional on perceiving changes in the climate over a given time following Eq. (2) below:

$$y_{2i}^{select} = (x_{2i} \beta_2 + \varepsilon_{2i} > 0),$$
(2)

\(y_{2i}^{select}\) is climate change perception, \(x_{1i}\) and \(x_{2i}\) are vectors of explanatory variables that affect both the perception and adaptation decision of farmers. They include observed features of the farm-holders, the household, and the community where the farmer dwells. Accordingly, \(\beta_1\) and \(\beta_2\) are parameters to be estimated in both equations. \(\varepsilon_{1i}\) and \(\varepsilon_{2i}\) are assumed to have a mean of 0 and variance of 1, hence normally distributed [N(0, 1)]. Furthermore, \(corr\left( {\varepsilon_{1i} , \varepsilon_{2i} } \right) = \rho\). When the error terms in Eqs. (1) and (2) are correlated or when \(\rho \ne 0\), applying the standard probit technique to Eq. (1) results in biased estimates (Asrat and Simane 2018). Hence, the two-step Heckman framework generates consistent estimates for the parameters in the models by correcting for selection bias. Another assumption for the consistency of the Heckman is the exclusion restriction requirement that implies that at least one explanatory variable should not be included in both the selection and outcome equations (Puhani 2000). Admittedly, it is often difficult to find a variable that meets this condition. One way to assess the independence of errors between the selection and outcome equation is through the significance of the inverse Mills ratio. Despite its simplicity, a core limitation of the Heckman model is that the sample selection bias correction only considers selection on observable covariates. However, when selection bias stems from unobservable factors, the model estimates may be weak compared instrumental variable techniques (Cameron and Trivedi 2005). Moreover, the technique has also been criticized on the basis of its emphasis on sensitivity to normality assumption (Opiyo et al. 2016).

3.4 Identification and extended adaptation assessment

In addition to the Heckman selection model on the binary perception and adaptation measures, we expand our investigation by looking at the various forms of adaption measures adopted by rural farmers in Nigeria and Ethiopia. To do this, we classify various adaptation measures into four categories, such as (1) chemical management indicating the purchase, receipt, and use of chemicals on the farmland such as fertilizers, pesticides, herbicides and fungicides; (2) land management indicating the use of any form of land and soil management such as erosion control measures, planting of trees, mixed planting, and crop rotation; (3) water management representing the use of any form of irrigation system and participating in community water shedding activities during the survey period; and (4) organic fertilizer management if the respondent uses any of the organic methods such as manure and compost.

From the four groups of adaptation measures, we created a categorical variable depending on how many adaptation measures were adopted by each plot holder. We use this approach given that several farmers reported using a mixture of the adaptation measures. The categorical variable is defined as:

0 = no adaptation measure adopted

1 = 1 adaptation measure

2 = 2 adaptation measures

3 = 3 adaptation measures

4 = all 4 adaptation measures.

The categorical variable is no way ordered; hence higher values do not show higher preference. Rather, a farmer decides to choose one or more measures according to the need of the farmland as against not adopting any measure at all. A potential limitation of our adaptation indicator is the simplistic nature of the categorization of adaptation measures. For instance, a complete categorization of water, land, chemical, and organic fertilizer management measures would require a 4 × 4 combination, since most farmers report using a mixture of various methods, thereby resulting in 15 combination of adaptation measures against the choice of not adapting. Such large combinations may complicate both the estimation of our models and the interpretation of the outcomes. Hence, we code the adaptation indicator to reflect the adoption of at least one, two, or three measures or a combination of all four measures. Despite this limitation, our adaptation indicator still provides useful information on farmers’ willingness to adapt to climate change.

To estimate the extended adaptation model with the categorical outcome variable, we use the multinomial logistic estimator following the traditional specification and the work of Waibel et al. (2018) as follows:

$$Pr\left( {y_i = j} \right) = \frac{{\exp \left( {x\beta_j } \right)}}{{\sum_{k = 1}^J \exp \left( {x\beta_k } \right)}}$$
(3)

where \(y_i\) denotes the categorical adaption variable while \(x\) represents a vector of regressors that affect adaptation decision. Here, we retain all the explanatory variables from the outcome Eq. (1) and include additional variables such as measures of climate change (rainfall, flood, drought, and pest infestation) and new community level characteristics such as community irrigation system for Ethiopia and community savings organization for Nigeria. The inclusion of the additional community indicators is guided by empirical evidence. Waibel et al. (2018) find that membership to an association positively affects water management among Vietnamese farmers. Thus, we may expect that being a member of a community savings association can influence adaptation among Nigerian farmers. On the other hand, Oyawole et al. (2020) contend that being a native in a community affects adaptation. Since rural communities are largely homogeneous with most household having historical links to the communities, we can expect that community irrigation systems may affect adaptation among Ethiopian farmers. Thus, all new variables in \(x\) are hypothesized to positively influence the adaptation of at least one climate change coping technique.

A fundamental challenge with employing the multinomial logistic (MNL) estimator is satisfying the assumption of independence of irrelevant alternatives (IIA) to achieve consistent and unbiased estimates. The IIA assumption holds that for the MNL model to be consistent, farmers’ probability of adopting one adaptation measure should be independent of the probability of adopting another measure. For instance, the probability of adopting category 1 over category 3 of our adaptation indicator should be constant even if category 2 is included or removed from the list of outcomes. Previous studies showed that the IIA assumption can be assessed using the Hausman specification test with the null hypothesis that the IIA holds (Deressa et al. 2009). Similarly, we adopt the Hausman specification test to evaluate the consistency of the MNL model for both Ethiopia and Nigeria.

3.5 Data description

We focus mainly on household members who own, use, or have control over plot(s) of land for crop production during the survey period. Thus, while this represents the head of the household in many cases, there are several other households with more than one smallholder farmer who makes most of the decisions regarding the plot. Also, we extracted variables that are as similar as possible from the socioeconomic, agricultural, and community modules of the fourth waves of the Ethiopian socioeconomic survey and the Nigerian general household survey for our empirical analysis.

3.5.1 Dependent variables

To measure farmers’ perception and adaptation choices as our selection and outcome equations’ indicators, we rely mainly on reported effect of climatic changes or conditions on crop production from the plot holder questionnaires of the ESS and GHS datasets. Accordingly, we measure climate change perception as a dummy variable = 1 if the respondent experiences any of the climatic changes such as changes in rainfall (for Ethiopia), drought (for Nigeria), flood, and pest infestation that affect agricultural activities and 0 if otherwise. While this variable provides a close measure of perception, it is conceptually different from other studies (Makate et al. 2017) who asked their respondents to describe the observed climatic changes over the years such as increase or decrease in rainfall or temperature and changes in cropping season.

Next, we measure adaptation to climate change as a dummy variable = 1 if the respondent uses any adaptation measure and 0 if no adaptation measure was reported. Adaptation measures include engaging in irrigation, erosion control, chemical fertilizers usage, organic fertilizers, crop rotation, mixed planting, tree planting and water shedding activity. Thus, we consider a farmer to have adapted to climate change if any of the listed measures is positive. We present a summary of the perception and adaptation dummies for both countries in Figs. 3 and 4 and provide the full information in Tables 9 and 10. On average, about 48% of farmers in Ethiopia experienced changes in climate as against the 16% in rural Nigeria (Fig. 3). The low perception rates almost mirror the Afrobarometer awareness indicators in Fig. 2. On the other hand, while about 75% of rural farmers engage in one adaptation measure or the other in Ethiopia, 91% of rural farmers in Nigeria practice adaptation measures. What seems interesting from Fig. 3 is that despite not reporting any climate change perception, farmers still engage in adaptation measures. This difference may stem from the imperfect measure of perception used in this study or it could be exposing a flaw in the Ethiopian Socioeconomic Survey (ESS) and the Nigerian General Household Survey (NGHS) on how climate change perception questions are framed. However, our Heckman model only selects observations for which perception is positive into the adaptation equation.

Fig. 3
figure 3

Source: Authors’ creation from ESS and NGHS (2018/2019)

Climate change perception and adaptation in Ethiopia and Nigeria.

Full size image
Fig. 4
figure 4

Source: Authors’ creation from ESS and NGHS (2018/2019)

Disaggregated measures of climate change adaptation.

Full size image

A further disaggregation of the climate change adaption measures used by rural farmers in Ethipia and Nigeria in Fig. 4 reveals that land management to mitigate against flood and erosion is crucial to rural farmers in both countries. However, rural farmers in Nigeria adopt more chemical fertilizers than organic ones, while there is no significant difference between the two forms of fertilizers for Ethiopian farmers. Lastly, Ethiopian rural farmers adopt more water management measures such as irrigation and community water shedding practices than their counterparts in Nigeria.

3.5.2 Explanatory variables selection

Our independent variables for the selection and outcome equations are mainly drawn following the extant literature on perception and adaptation and are presented in Table 1 for both countries under study. For instance, previous studies argue that men are more likely to be the custodians of farmlands, making them responsible for generally taking actions to adapt to climate change (Waibel et al. 2018). However, when multiple adaptation methods are available, Oyawole et al. (2020) find that men are more likely to utilize crop rotation strategy due to large farmlands while women opt for green manure to improve yield on small farmlands. Accordingly, we hypothesize that being a male farm holder is positively association with both perception and adaptation decision. The data in Table 1 strengthen this hypothesis such that about 68% and 82% of farm holders in Ethiopia and Nigeria, respectively, are male. Next, we include farmers’ age as a predictor of perception and adaptation. A strand of the literature notes that the outcome of age may be unknown apriori since older farmers may be either averse to adopting new technologies or open to adoption due to their years of experience (Gbetibouo 2009). However, it is more probably that the years of experience will induce farmers to adopt certain adaptation measures (Deressa et al. 2009). In our dataset, Nigerian rural farmers tend to be older than those in Ethiopia by about 7 years (Table 1).

Table 1 Explanatory variables
Full size table

Another variable that is often discussed in the literature as a determinant of adaptation is the education of farmers. Scholars contend that educated farmers may be privier to information about the changes in climate that could influence their decisions to adapt. Empirical evidence from Onyeneke et al. (2018) and Asrat and Simane (2018) indicate that education level positively influence both perception and adaptation. In our study, rural farmers have low levels of education at an average of 2.12 years and 5.78 years for Ethiopia and Nigeria, respectively. Next, despite the importance of marriage in rural communities, the marital status of farmers receives less attention in the literature as a determinant of climate change perception and adaptation. Regardless, we hypothesize a positive relationship with both perception and adaptation since in rural communities, being married extends one’s social network to include the spouse’s family thereby increasing the likelihood of accessing climate change and coping measures information. For our dataset, 80% of Nigerian rural farm holders are married compared to the 69% in Ethiopia.

Gbetibouo (2009) and Deressa et al. (2009) argue that there is ambiguity on the effect of household size on adaptation. On one hand, a large household may have the incentive to adopt certain techniques that require large manpower. On the other hand, the need to diversify household income sources may push certain members out of agriculture, thereby reducing the household size and affecting the adaptation choice. In our dataset, we include an indicator for the number of individuals per household and another indicator measuring the percentage of the household members engaged in agriculture. The latter allows us to directly assess the predictive power of having more members of the household practicing agriculture. Next, plot size is a core determinant of adaptation among farm holders (Teklewold et al. 2019). However, the size of the farm could either increase the likelihood of adaptation or weaken farmers’ motivation to adapt. For instance, poor farmers with large family lands may find it expensive to bear the cost of adapting to new technologies. On the other hand, wealthy farmers are likely to invest in adaptation measures that improve their yields, thus taking advantage of the large farmlands. Accordingly, we also hypothesize that higher household income impacts both perception and adaptation positively. This is a major working hypothesis for this investigation. It basically claims that farmers with higher incomes are better able to afford means of accessing information on climate change and investments in adaptation strategies. Several empirical submissions indicate that income positively explains adaptation decision by allowing farmers to bear more risks and afford the cost of various adaptation techniques (Deressa et al. 2011; Onyeneke et al. 2018). Another important hypothesis relates to receiving agricultural extension or advisory services. Scholars contend that extension services bring farming information, including reports on climate change and adaptation technologies, closer to farmers (Opiyo et al. 2016). Thus, we expect a positive relationship with adaptation decision since extension services provide information on various topics that would benefit the farmer.

Lastly, we include community level variables that could affect farmers’ decision to adapt to climate change. For instance, climatic conditions such as changes in temperature and rainfall at the community level are likely to influence adaptation. Similarly, community irrigation and agriculture as the most common community occupation for Ethiopia and community agricultural and savings association for Nigeria are hypothesized to positively influence perception and adaptation decisions. The intuition is that community characteristics largely influence household choices by providing social capital, insurance, and access to information (Deressa et al. 2009).

4 Empirical results and discussions

We begin by presenting the outcomes of our climate change perception and adaptation models estimated using the Heckman selection framework in Table 2 for Ethiopia and Table 3 for Nigeria, respectively.

Table 2 Heckman selection model estimates – Ethiopia
Full size table
Table 3 Heckman selection model estimates – Nigeria
Full size table

4.1 Main results

For both Ethiopia and Nigeria, the Heckman models were estimated using the twostep approach based on Eqs. 1 and 2. Specifically, both the selection and outcome equations include indicators for age, gender, marital status, education, household size, income, agricultural extension service, and the distance to the administrative headquarters. On the other hand, communal climatic conditions such as temperature and rainfall are included in the selection equation on the assumption that they provide the first evidence of changes in climate that would influence farmers’ perception. Lastly, plot size and the gender of the household head are included in the outcome equation. Such careful inclusion of control variables in the selection and outcome equations allows us to approximate the exclusion restriction requirement of the Heckman model. As discussed in the conceptual framework section, the assessment of the correlation between the unobserved components of both the perception and adaption equations (selection bias) in the Heckman model is conducted using the significance of the inverse Mills ratio. In our estimations, the coefficients of the inverse Mills ratio are insignificant for both Ethiopia (Table 2) and Nigeria (Table 3), indicating that the unobserved variables are not correlated. This points to the fact that selectivity bias is not a concern in the countries under study. This is also intuitive since in our data set there are farmers who adopted certain climate change adaptation measures for which our climate perception variable did not capture.

In the selection model (perception) for Ethiopia in Table 2, we find positive and significant coefficients for gender, age, marital status, income, agricultural extension programme participation, household size, and average annual temperature in the selection equation. For instance, being a male farm holder increases the likelihood of perception by 0.15 at 5% significance level. This agrees with Opiyo et al. (2016) for northwestern Kenya, Waibel et al. (2018) for Vietnam, and Okonya et al. (2013) for Uganda. Contrary to the above studies, years of education negatively influences perception in Ethiopia. Moreover, the results confirm the finding of Gbetibouo (2009) which indicates that participating in agricultural extension program avails farmers the opportunity to get information on climate change which shapes their perception. On the size of the household, having many household members in general increases the likelihood of perception for the rural farmers in Ethiopia. This is intuitive since rural areas often have strong social networks especially across age groups. Thus, larger households may better perceive changes in climate change resulting from different sources of information. Furthermore, annual temperature is significantly positive at 5% pointing out that changes in temperature increase the chances of perception. This supports the submission of Asrat and Simane (2018). Lastly, the farther a farmer lives from the capital, the more likelihood of perception drops. We find this interesting given that being distant from the urban areas precludes one from accessing vital information that could influence their outlook on climate change.

Turning to the outcome equation (adaptation) for Ethiopia, most of the variables turn out insignificant. The results indicate that male and married respondents have lesser likelihoods of adapting to climate change; prompting that despite perceiving changes in climate, male and married farmers have lesser likelihood of adapting compared to female and unmarried farm holders. However, we find that male headed households are more likely to adopt coping measures during climate change. This finding generally supports the idea that within rural communities, the head of the household may often makes adaptation decisions that could affect all farm holders (Waibel et al. 2018). Lastly, having a large plot of land also increases the likelihood of adaptation. Thus, having a large hectare of land may serve as an incentive for adapting to climate change since failure to do so could lead to significant loss for the farmer.

We now turn to the perception and adaptation results for Nigeria. Table 3 shows that the estimates for Nigeria are mostly insignificant for both the selection and outcome equations. Specifically, we only find age, household size, and community agricultural association to influence the likelihood of perception positively. What this means is that older farmers may tend to pay more attention to changes in the climate thereby increasing their likelihood of perception. For instance, the longer they practice agriculture, the more they are likely to observe changes in climate conditions. Also, just like the case of Ethiopia, having a large household increases the likelihood of perception stemming from information gathering by household members. On the other hand, being located in a community with an agricultural-based association increases farmers perception likelihood. This finding is crucial since membership to such association avails farmers the opportunity to benefit from information spillover. Lastly, we find that marital status and annual average rainfall reduce the likelihood of perception at 1%. On the adaptation equation for Nigeria, we find being a male farm holder increases the likelihood of adaptation. However, being a male household head reduces the likelihood of adaptation.

4.2 Extended adaptation assessment results

Next, we broaden our investigation into the determinants of farmers’ decision to adopt multiple adaptation strategies. For the extensive adaptation assessment, we use the categorical indicator of adaptation as explained in Sect. 3.4 and applied the multinomial logistic (MNL) regression technique. The marginal effect results from this empirical exercise are presented in Table 4 for Ethiopia and Table 5 for Nigeria while the MNL coefficients are presented in Tables 9 and 10 of the appendix. In addition to the baseline control variables from the Heckman model, we also control for specific climate change indicators from which our perception variable was generated such as pest infestation, flood, low and heavy rain fall (for Ethiopia), and drought (for Nigeria). Furthermore, we checked the appropriateness and consistency of the MNL model by testing the assumption of independence of irrelevant alternatives using the Hausman specification test. For both Ethiopia and Nigeria, the null hypothesis of the independence of including “At least two methods” of adaptation is not rejected. The Chi-squared (p-values) values for Ethiopia and Nigeria are 6.83 (0.3365) and 2.80 (0.9025), respectively, suggesting that the MNL is consistent and appropriate for our data set.

Table 4 Marginal effects of multinomial logistic regression estimates—Ethiopia
Full size table
Table 5 Marginal effects of multinomial logistic regression estimates—Nigeria
Full size table

Beginning with the results for Ethiopia (Table 4), we find interesting dynamics of farmers’ adaptation. For instance, we find that men have lower probabilities of adopting only one technique (column 1), while an increase in age marginally reduces the probability of adaptation. This is consistent with the argument that older farmers are more likely to be resistant to change or technology adoption (Gbetibouo 2009). Next, while a change from a single to a married farmer reduces the probability of adopting only one measure, it increases the probability of adopting at least three measures by 0.09 at the 5% level. However, educated farmers are more likely to adopt at least one technique compared to the choice of not adopting at all but tend to be averse to adopting at least two measures. This latter effect stands in contrast to our predictions, as we anticipate that more educated farmers will have greater access to knowledge about adaptation strategies. However, this finding seems to suggest that, given the information at their disposal, educated farmers could leave the sector altogether if the opportunity cost of staying in agriculture is higher.

Next, we find that higher household income, bigger plot size, and receiving agricultural advisory services from extension programmes increase the probability of adopting at least three adaptation measures but reduce the probability of adopting only one measure. These effects suggest that wealthier Ethiopian farmers with bigger plots and access to extension services are more open to climate change adaptation. Similarly, living in a community that practices collective irrigation increases the probability of adopting all adaptation measures or at least two of them. For instance, moving into such a community increases the adoption of at lease two measures by 0.09 and all four measures by 0.02, indicating the importance of communal social support and insurance. On the other hand, an increase in the distance from te administrative headquarters reduces the probability of adopting all measures by 0.07 at the 1% level. Such an effect may result from inadequate information and the difficulty in accessing markets for effective adaptation technologies.

Lastly, our measures of climate change for Ethiopia, namely pests, flood, high and low rainfalls all appear to be positively significant at the 1% level in column 4 for all four measures. This indicates that farmers who experience pest infestation due to extreme weather situations, flood, reduced rainfall or episodes of high rainfall that damage crops are likely to mitigate against future occurrence by adopting water, land, organic fertilizer, and chemical management techniques to maximize their production, preserve their plots, and improve their general welfare. This finding is crucial and indicates that rural farmers may be proactive in putting in place measures to combat the consequences of climate change. Thus, it becomes imperative for the Ethiopian government to design measures of expanding agricultural extension services to farmers and put in place programmes that will train rural farmers on modern climate-smart agricultural techniques and technologies to improve their productivity and ensure food security.

Next, we examine the marginal effect results of the multinomial logistic estimation for Nigeria in Table 5. As evident from the table, there are fewer significant variables compared to the results for Ethiopia. We, however, find that men have a higher probability of adopting all four measures (column 4) at the 10% level. On the other hand, age, marital status, ratio of farm workers in the household, and male household heads are insignificant in influencing any adaptation mix for Nigeria. Next, similar to the case of Ethiopia, household income and plot size are strongly significant in increasing the probability of adaptation from a mixture of two to four adaptation measure. The significance of family income portrays the fact that most adaptation techniques involve some form of expenditure to acquire the necessary tools for irrigation, chemicals, and other land management instruments. As noted in Waibel et al. (2018), adaptation measures can be considered forms of investment that require household income. Other important results in Table 5 include the community level variables such as savings and agricultural associations. Notably, we find that savings association only increases the probability of adopting at most two measures while agricultural associations is associated with an increase in the probability of adopting only one measure. The weakness in the community indicators may indicate a lack of cooperation among farmers in rural communities of Nigeria.

Next, we find agricultural extension programme participation to influence the adoption of all four methods but not the mixture of a few. Furthermore, distance to state capital also has a negative effect on at least three adaptation mix. This stems from the fact that most adaptation method require equipment that may not be found in the local markets. Moreover, being distant from the capital also limits the amount of information rural farmers may get regarding climate change and recent adaptation techniques. Lastly, our measure of climate change for Nigeria includes experiencing drought, flood, and pest infestation that cause damage crop production. Our estimates show that pest infestation affects the likelihood of adopting at most three adaptation techniques while flood and drought positively influence the adoption of at least two techniques among rural farmers in Nigeria. These results further buttress the fact that rural farmers tend to be cognizant of changes in climate conditions and are proactive towards forming coping mechanisms. However, support from different levels of government will further aid their productivity and improve rural life and food security.

4.3 Robustness check: logistic regression technique

To ensure the robustness of our findings especially with regards to the adaptation techniques mix by rural farmers in Ethiopia and Nigeria, we estimate binary logistic regressions with robust standard errors using each of the four classes of adaptation techniques (water management, chemical management, land management, and organic fertilizer management) as the dependent variables in separate equations. We present the marginal effect results of this exercise in Table 6 for Ethiopia and Table 7 for Nigeria and the log odds in Tables 11 and 12 of the appendix. Generally, the results for Ethiopia add certain nuances to the findings from the multinomial logit results in Table 4. Specifically, we find that being a male farm holder increases the probability of adopting water management technique such as irrigation and participating in community water shedding activities by 0.056; land management such as planting tree crops, practicing mixed planting or crop rotation, and adopting erosion prevention measures by 0.031; and organic fertilizer management such as manure and compost by 0.058. Additionally, older farmers are more averse to water management while being married increases the probability of adopting all measures except land management.

Table 6 Marginal effect of logistic regression estimates—Ethiopia
Full size table
Table 7 Marginal effects of logistic regression estimates—Nigeria
Full size table

Furthermore, household income significantly increases the probability of adopting all measures except for chemical management which is positive but insignificant. However, plot size, receiving agricultural advisory services and living in a community where irrigation is practiced have significantly positive effects on the probability of adopting all four measures. Interestingly, living far from the administrative headquarters reduces the probability of adopting chemical and organic fertilizer techniques but increases the probability of adopting water management such as irrigation that may be readily available to farmers. On the climate change indicators, only experiencing lower rainfalls and pest infestation increase the probability of adopting all four techniques. The former is intuitive since with lower rainfall, farmers will have to look for alternative sources of watering their plots. In areas with access to water bodies, irrigation systems can be practiced in keeping production going despite reduced rainfall. Lastly, for land management, all climate variables are significant except flood which was dropped from the estimation due to perfect prediction of the outcome variable. This indicates that every farmer who has experienced flood puts land management measures in place.

The results for Nigeria are less consistent compared to those of Ethiopia and are presented in Table 7. For instance, age increases the probability of adopting land management measures, while negatively affecting water management but insignificant for other measures. Conversely, educated rural farmers are less likely to implement land and organic fertilizer management techniques, but more likely to implement chemical management approaches. This dynamic may indicate that being educated exposes farmers to the benefits of chemical fertilizers at the expense of organic forms. The coefficients for household income and plot size for all adaptation strategies are positive and statistically significant, demonstrating that household income is pertinent in determining whether to invest in climate-smart agricultural techniques in Nigeria or not. However, plot size has a negative but insignificant association with water management. Although insignificant, this coefficient seems intuitive since larger plots may be more challenging to irrigate. Similar to the Ethiopian result, distance from the capital increases the probability of adopting irrigation but reduces the odds of adopting chemical management practice. Lastly, experiencing drought increases the likelihood of adopting chemical, land, and organic fertilizers management strategies but insignificant for water management. On the other hand, flood episodes increase the potential of adopting both water and land management practices while pest infestation only positively affects land management. These findings are intuitive and guide us in understanding the kind various adaptation measures rural farmers tend to adopt depending on the nature of climate change effect they experience.

5 Conclusion and policy recommendations

As it stands today, according to the Climate Policy Initiative (2021) report, only 7% of financing went into climate-related investments including protecting communities from flooding and wildfires, building agricultural systems that can withstand extreme weather, ensuring a steady supply of clean water, updating outdated infrastructure, and relocating people. Thus, they need a lot more money from corporate investments. Vertical farming and hydroponic agriculture, improved cooling and insulation systems, early warning systems for extreme weather events, coastal barriers, water desalination and wastewater treatment, and many more measures are all discussed in the literature as viable options for climate adaptation infrastructure and technologies that are affordable, proven to be effective, and have immediate impact.

Reducing carbon emissions and adapting to a changing climate are necessary, but we see adaptation as the more sustainable investment option. Using recent nationally representative survey data from the Ethiopia and Nigeria available between 2018 and 2019, we investigated the factors determining rural farmers’ adaptation options to climate change in both countries. For our empirical analyses, we used the Heckman selection framework to address potential bias in the sample due to farmers’ propensity to self-select into climate change adaptation, a multinomial logistic regression to determine how many adaptation strategies rural farmers have adopted, and a standard logistic regression to determine what factors contribute to the adoption of these strategies. Our findings yielded important outcomes worthy of mention but generally suggest that rural farmers quite adapt to mitigating against the effect of climate change. For instance, we find that household income and plot size significantly influence adaptation decisions for rural farmers in Ethiopia and Nigeria. This indicates the importance of income for investing in climate-smart agricultural practices. Similarly, farmers with bigger plots who run the risk of massive production loss tend to adopt measures of coping with climate change. These measures include water management such as irrigation systems; chemical management like inorganic fertilizers, pesticides, herbicides, and fungicides; land management like crop rotation, mixed planting, erosion control measures, and tree planting; and organic fertilizer management such as using manure and compost on the farmlands to improve productivity.

Furthermore, we find that agricultural extension programs are essential to rural farmers’ decisions on adaptation in both Ethiopia and Nigeria. Thus, providing these types of agricultural services to farmers will help them acquire the knowledge they need about climate change and the most effective and cost-efficient climate-smart technology to implement on their farms. Moreover, we find evidence in both countries that social networks within the community is crucial in influencing the adaptation decision of farmers. For instance, Nigerian rural farmers who are members of an agricultural development association are comparatively more cognizant of climate change. However, the impact on adaptability is only statistically significant when it comes to the use of chemical management techniques.

Our research provides useful policy recommendations for promoting the adoption of CSA methods in Ethiopia and Nigeria, and by extension, Sub-Saharan Africa. First, it’s crucial to stress the significance of climate-related information dissemination. Farmers’ ability to understand the implications of climate change and select effective adaptation techniques hinges on the ease with which they can access such knowledge. Our finding shows that the expansion of agricultural advisory services would go a long way in educating rural farmers and exposing them to measures that will aid productivity, profit, and general welfare despite the challenges of climate change. Moreover, agricultural extension services have been proven to work in practice. Perhaps, both the Ethiopian and Nigerian governments could learn from the agricultural extension program in Northern Ghana facilitated by the Association of Church-based Development NGOs (ACDEP). The service included disseminating information on new technologies, access to markets, skill and management practices which led to increased income for participating farmers (Danso-Abbeam et al. 2018). Similarly, the Ethiopian Rural Capacity Building Project of 2007 was found to improve rural farmers’ orientation towards commercialization and market-oriented practices (Buehren et al. 2017). Furthermore, governments could exploit the advancements in information and communication technologies in providing rural dwellers with important information on climate change. This could be through occasional broadcasts on mass media platforms, mobile notification to educated rural dwellers especially those residing far away from the cities.

Secondly, the importance of community networks cannot be overemphasized. Government efforts to facilitate rural agriculture should be embedded in community-led schemes. Since every community might have unique ways of agricultural practice, such as the case of water shedding activities in rural Ethiopia, designing extension programmes at the community level will improve collaboration and sense of diligence by farmers. We envisage that such an approach will improve rural agriculture and community development amidst climate change. Two relatively successful adaptation strategies fostered by community network and social capital include the joint forest management and watershed development programs in Andhra Pradesh, India. According to D’Silva and Pai (2003), these programs were successful in two out of three districts in part, due to tight and historical community cohesion. Hence, through communal trust and cooperation, community-led schemes for climate change adaptation in Nigeria and Ethiopia could be successful. Thirdly, our findings point to the relevance of income level in facilitating farmers’ adoption of various adaptation measures. However, since rural dwellers are relatively poor, it then becomes imperative for governments to assist farmers with low-cost funding to better position them for climate change adaptation. Examples of such financial aid include the unconditional Hunger Safety Net cash transfer program in northern Kenya. The program was implemented between 2009–2012 and 2013–2018, covering 60,000 and 100, 000 households, respectively. Matata et al. (2023) evaluated the impact of such program on household resilience during climate shocks and found a significant positive effect.

Lastly, we acknowledge some obvious limitations of our study. For instance, the ESS and NGHS datasets are lacking in unique variables that measure climate change perceptions and the specific changes in climatic conditions. This led us to use proxies that may be imperfect. Such a limitation could influence our selection equations in the Heckman selection models, thereby leading to a potential underestimation of rural farmers’ perception of climate change. A silver lining, however, is that our imperfect measure of farmers’ perception for both Ethiopia and Nigeria are consistent with the survey report by Afrobarometer (Fig. 1). Thus, this limitation has implication for subsequent nationally representative surveys to capture essential variables that are peculiar to climate change issues, especially in terms of survey questions framing. Another limitation of our study pertains to restricting our empirical investigations to Ethiopia and Nigeria. While such analysis provides useful information for understanding the distinct climate change issues between the two countries, it precludes us from making generalizations about the applicability of our findings to other countries, especially beyond Africa, without understanding the peculiar climate change issues of those countries. Moreover, Ethiopia and Nigeria are part of the eight African countries for which the Living Standard Measurement Surveys were conducted with technical support from the World Bank. A wholistic evaluation of the LSMS should include all eight countries. Moreover, our empirical analyses focused on the entire rural sample for both Ethiopia and Nigeria without delving into the within-country dynamics of climate change perception and adaptation. Our justification for limiting our analysis to the overall sample is to avoid information redundancy, since several studies have been carried out focusing on specific regions within Nigeria and Ethiopia. Nonetheless, overcoming these limitations in future research is necessary for improving our understanding of farmers’ perception and adaptation to climate change in Sub-Saharan Africa.