Modelling Determinants of Farmers’ Choice of Adaptation Strategies to Climate Variability and Extreme Events in Kitui County, Kenya

The study was carried out to assess determinants of farmers’ choice of specific adaptation strategies to climate variability and extreme events in selected agro-ecological zones in Kitui County. Descriptive survey design was used. The study area was stratified into four study sites with respect to four different agro-ecological zones and a total of 341 households selected to constitute the sample size. Multivariate probit regression model was run in Stata version 12 to determine the influence of different socio-economic characteristics on farmers’ choice of specific adaptation strategies. The model results indicated that age, gender, farming experience, membership to farmers’ organization, education level, access to extension services and proximity to market had a significant varying influence on farmers’ choice of several adaptation strategies. The study established that different socio-economic characteristics had a different influence on the farmers’ choice of specific adaptation strategies. The study therefore recommends that climate variability adaptation policies, programs and projects by governmental and non-governmental development agencies should target specific socio-economic characteristics that are relevant to the adaptation strategies in question.


INTRODUCTION
Changes in the climate system in recent decades have caused significant impacts on the natural and human systems on all continents and across the oceans (IPCC, 2014). According to IPCC (2007), variability in temperature and rainfall patterns have been predicted to cause significant effects on global agriculture, due to extreme weather events such as droughts and floods and changes in patterns of pests and diseases. The effects of changing temperature and rainfall patterns are more pronounced in developing countries owing to their geographic exposure, low income, greater reliance on rain-fed agriculture and other climate sensitive sectors coupled with its weak capacity to adapt to the changing climate (Belloumi, 2014; Thomas et al.,2005;Slingo et al., 2005).
The IPCC (2007) report estimated that Africa will be the most vulnerable continent to the progressive changes in climate globally, due to its low adaptive capacity resulting from the multiple stresses of poor infrastructure, poverty and governance. Kenya is one of the most vulnerable countries to climate variability and extreme events in Africa due to its low adaptive capacity and dependence on climate-sensitive sectors such as agriculture and fisheries as the key drivers of its economy (FAO, 2011; Since agriculture is the mainstay of most rural communities in Kenya (Republic of Kenya,2005) negative developments in agriculture would adversely affect the rest of the livelihoods that are depended on agricultural production. The cumulative effects of climate variability and extreme events in Kenya therefore pose a significant threat towards the attainment of the country's Vision 2030 (Parry et al.,2012) as well as the implementation of the Sustainable Development Goals (UNECA, 2018; UNCCS,2017). Implementation of adaptation strategies will therefore be paramount to cushion communities from the effects of climate variability and extreme events and promote sustainable livelihoods in the advent of a changing climate (Akinnagbe and Irohibe,2014;Schipper et al.,2008;IPCC,2007).
According to Maddison (2006), the ability and decision to adopt a particular adaptation strategy is determined by several institutional and socio-economic factors. In the face of climatic variability, farmers may opt to adopt several strategies instead of relying on a single strategy to exploit complementarities or substitutability among alternatives (Ojo and Baiyegunhi,2018). The current study therefore sought to examine the determinants of farmers' choice of specific adaptation strategies in different agroecological zones in Kitui County.

Profile of the study area
The study was carried out along a transect line (in a buffer zone of 5km radius on both sides of the line) in semihumid, transitional semi-humid to semi-arid, semi-arid and arid zones in Kitui County. The study sites are shown in Fig. 1.

Study Design and Sampling Techniques
Descriptive survey design was used. The target population for the study was the agro-pastoral farmers in the study area. The unit of study was the household while the respondents comprised of the head of the households.Stratified sampling method was used to classify the study sites with reference to four different agro-ecological zones in Kitui County. One sub-location in each agro-ecological zone was randomly selected along a transect line (in a buffer zone of 5km radius on both sides of the line). Systematic random sampling method was used to identify respondents in the selected sub-location.
The sample size for the study was determined by calculating 10% of the number of households in each of the four sub-locations. According to Mugenda and Mugenda (2003), a sample size of 10% provides an adequate representation of the target population in descriptive research. The total sample size for the study was 341 households with 39,160, 38 and 104 households from the arid, semi-arid, transitional semi-arid to semihumid and the arid zones, respectively.

Data Collection and Analysis
Primary data was collected through administration of questionnaires to 341 respondents. Interviews with key informants were also conducted. Multivariate Probit (MVP) regression model was run in Stata version 12 to assess the determinants of farmers' choice of different adaptation strategies in the study area.
The MVP decision model is guided by the random utility theoretical model which describes a choice decision in which an individual has a set of alternative adaptation strategies from which to choose (McFadden, 1978). The model assumes that each adaptation option has distinct attributes that influence a farmer's choice over another alternative and is based on the notion that the utility is derived by choosing several alternatives.
The utility random model is described below as applied by Feleke et al. (2016).
Assuming that Uj is the expected utility that a farmer will gain from using adaptation strategy j whereas Uk is the expected utility for not choosing adaptation strategy j but rather k.
The linear random utility model of adapting to climate variability by choosing jth adaptation strategy (Uj) can be expressed as a function of explanatory variables Xi as shown below: The linear random utility model for ith farmer who does not use jth adaptation strategy but rather kth adaptation strategy is given by: Where xi is a vector of explanatory variables βj and βk are vectors of parameters for choosing jth and kth adaptation strategy respectively, μj and μk are error terms for choosing jth and kth adaptation strategy, respectively. According to Gujarati (2006), the error terms in the above equations are assumed to be normally independently and identically distributed.
If a farmer chooses to adopt jth adaptation strategy to climate variability, then the expected utility that the farmer gets is greater than the expected utility for not using that strategy and according to Falco is a matrix of unknown regression coefficient, εi is a vector of residual error distributed as multivariate normal distribution with zero means and unitary variance; where Ʃ is the variance-covariance matrix.
The off-diagonal elements in the correlation matrix, = represent the unobserved correlation between the stochastic components of k th and J th options (Cappellari and Jenkins,2003).
The relationship between Zij and Yij is: The likelihood of the observed discrete data is then obtained by integrating over the latent variables: Where, Ai1 is the interval (0, ∞) if Yij=1 and the interval (-∞, 0) otherwise and is the probability density function of the standard normal distribution.
Since the coefficient estimates from MVP regression show the direction of influence rather than the magnitude (Mullahy,2017), to interpret the effects of explanatory variables on the probabilities, marginal effects were derived as follows: where, δij-denotes the marginal effect of the explanatory variable on the probability that alternative j is chosen. For the purpose of this model, selected adaptation strategies to climate variability and extreme events adopted by farmers were used as the dependent variables while farmers' socio-economic characteristics were used as the explanatory variables for the model as described in Table 1 and Table 2, respectively. The pairwise correlation coefficients (Rho) shown in Table  3 also indicated a positive correlation between the pairs most of which are highly significant implying that the sets    The marginal effects presented in Table 4 were used to quantify the influence of explanatory variables on the dependent variables in the model.   The multivariate probit regression results indicated that age of the household head had a positive but insignificant influence on the adoption of use of manure. There was however a negative influence of age of the household head on the adoption of crop diversification, planting drought resilient crops, planting hybrid crop varieties, soil conservation techniques, agroforestry and use of pesticides. The negative influence of age of the household head was significant on the adoption of crop diversification, soil conservation techniques and agroforestry. Marginal effects results showed that a unit increase in age reduced the probability of adopting crop diversification, soil conservation techniques and agroforestry by a factor of 0.01, 0.08 and 0.01, respectively.
Gender of the household head had a positive influence on the adoption of crop diversification, drought resilient, soil conservation techniques, use of manure and agroforestry which was significant on the adoption of crop diversification, use of manure and agroforestry. The marginal effects results indicated that male headed households were 12%, 7% and 9% more likely to adopt crop diversification, use of manure and agroforestry, respectively than their female counterparts. In regards to the adoption of use of fertilizers and pesticides, gender of the household head had a significant negative influence with marginal effects of 0.15 and 0.11, respectively. This implies that female headed households were 15% and 11% more likely to use fertilizers and pesticides, respectively than their male counterparts. The results indicated a positive but insignificant influence of household size on the adoption of drought resilient crop varieties. A negative influence of household size was however noted on the adoption of crop diversification, hybrid crop varieties, soil conservation techniques, agroforestry and use of pesticides and fertilizers with the influence being significant only on the adoption of soil conservation techniques with a marginal effect of 0.02. This implies that a unit increase in household size reduced the probability of adopting soil conservation techniques by 2%.
Membership in a farmers' organization had a negative insignificant influence on the adoption of drought resilient crops, hybrid crop varieties, manure and fertilizers. There was however a positive influence on the adoption of crop diversification, soil conservation techniques, pesticides and agroforestry. The significant negative influence of membership in a farmers' organization had a marginal effects 0.13 implying that membership to farmers' organization decreased the probability of adopting hybrid crop varieties by 13%. On the other hand, membership in a farmers' organization significantly increased the probability of adopting soil conservation techniques by 17%.
The results indicated that farming experience had a positive influence on all the adaption strategies except for adoption of use of manure, which was negative but insignificant. The positive influence of farming experience was significant on the adoption of crop diversification and soil conservation techniques with marginal effects of 0.01 on each, implying that a unit increase in farming experience increased the probability of adopting crop diversification and soil conservation techniques by 1%.  to credit facilities were 0.09 more likely to adopt drought resilient crops than those without access to credit facilities.
In regards to access to extension services, a positive influence was noted on the adoption of all the adaption strategies except for use of hybrid crop varieties. The results indicated a significant positive influence on the adoption of drought resilient crops, soil conservation techniques, use of fertilizers and agroforestry. The marginal effects results indicated that access to extension services increased the farmers' probability of adopting the drought resilient crops, soil conservation techniques, use of fertilizers and agroforestry by 17%, 10%, 9% and 16 %, respectively. Access to weather information however had an insignificant negative influence on the adoption of all the adaptation strategies except for the use of fertilizers and agroforestry.
The results further indicated that distance to market had a negative influence on the adoption of all the adaption strategies except for use of manure implying that ease of access to the market increased farmers probability of adopting the different adaption strategies. The negative influence of distance to market was significant on adoption of crop diversification, use of hybrid crop varieties, soil conservation techniques, fertilizers, agroforestry and pesticides whose marginal effects implied that a unit increase in distance to the market reduced farmers' probability of adopting crop diversification, use of hybrid crop varieties, soil conservation techniques, fertilizers, agroforestry and pesticides by a factor of 0.01, 0.05, 0.02, 0.01, 0.01 and 0.04, respectively.
The influence of land size was positive on the adoption of crop diversification, drought resilient crops and soil conservation techniques. There was however a negative influence of land size on the adoption of hybrid crop varieties, use of fertilizers, manure, agroforestry and pesticides. A significant influence of land size was noted on the adoption of drought resilient crops and use of fertilizers implying that an increase unit in land size increased the probability of adopting drought resilient crops while decreasing that of adopting use of fertilizers by 1%.

IV. DISCUSSION
The ability and decision to adopt a particular adaptation strategy is determined by several socio-economic factors (Maddison,2006). Results from the present study indicated that different socio-economic characteristics of farmers had a different influence on the farmers' choice of specific adaptation strategies to climate variability and extremes.
The results showed that there was a significant negative influence of age on the adoption of crop diversification, soil conservation techniques and agroforestry which implies that younger farmers in the study area were more likely to adopt the adaptation strategies as compared to older farmers. This could be because younger farmers are innovative and likely to try new technologies and methods to improve agricultural productivity. Conversely, in most cases older farmers are often not aware of recent innovations in agriculture and/or are reluctant to try new methods. Similar findings where there was a significant negative influence of age on the adoption of mixed cropping and improved crop varieties were reported in other studies (Ojo and Baiyegunhi,2018; Ali and Erenstein, 2016).
With regards to gender of the household head, female headed households were more likely to use fertilizers and pesticides compared to their male counterparts. This could be attributed to the fact that female headed households have less access to resources such as land and therefore resort to invest in use of fertilizers and pesticides to boost their agricultural productivity in their small pieces of land.On the other hand, the results indicated that male headed households were more likely to adopt crop diversification, use of manure and agroforestry as opposed to female households. This could be because womenheaded households are usually constrained by family labor since they are culturally assigned responsibility in domestic activities and also have less access to resources and information compared to male headed households which limit their ability to carry out labor-intensive activities. The easiness with which male headed households adapt to climate change compared to female headed ones was also highlighted by Tenge De Graffe and Heller (2004)   Contrary to the expectation, access to weather information had an insignificant negative influence on the adoption of all the adaptation strategies except for the use of fertilizers and agroforestry. Similar studies noted that access to weather information increases farmers awareness on climatic changes which is essential in making informed decisions on preparedness to reduce agricultural losses that might occur from climate variability and extreme events thereby increasing the probability of farmers' to adopt different adaption strategies ( The results further indicated that distance to market had a negative influence on the adoption of all the adaption strategies except for use of manure implying that ease of access to the market increased farmers probability of adopting the different adaption strategies. Proximity to market facilitates farmers' access to information and agricultural inputs such as hybrid crop varieties, fertilizers and pesticides as well as a market for selling agricultural outputs increasing the likelihood of adopting different adaptation strategies. The results are inconsonance with findings by Marie et al. (2020)who noted that farmers with access to market were 0.34 times more likely to adopt climate change adaptation strategies than those without.Further,the results corroborate findings by Belay et al. (2017) who found a positive and significant effect of distance to market on farmer input intensity and crop diversification.
Lastly, land size increased the probability of adopting drought resilient cropswhile reducing the probability of adopting use of fertilizers. The mixed effect of land size on adoption of the different strategies could be because a large farm size allows farmers space to practice crop diversification and also discourage adoption of high cost strategies. The results of the study are in consonance with findings from Žurovec and Vedeld (2019). A positive and significant relationship between land size and farmers' adoption of a combination of several adaptation strategies such as agroforestry, perennial plantation, crop-livestock diversification and improved varieties was also reported by Fadina and Barjolle (2018).

V. CONCLUSION AND RECOMMENDATIONS
The study established that age, gender, farming experience, membership to farmers' organization, education level of the household head, access to extension services and proximity to market had a significant varying influence on farmers' choice of different adaptation strategies. The study therefore recommends that policies, programs and projects by governmental and nongovernmental development agencies aimed at helping farmers adapt to climate variability and extreme events should target specific socio-economic characteristics that are relevant to the adaptation strategies in question.