In Kenya livestock sub-sector is an integral part of the agricultural sector contributing about 4% of the national Gross Domestic Product (GDP) mainly from the production of milk, meat, eggs, hides, skins and wool (KNBS, 2018). The bulk of the livestock are found in arid and semi-arid lands (ASALs), comprising 84% of Kenya’s total land mass. These areas are characterized by low, unreliable and poorly distributed rainfall, supports a quarter of the country’s  total  human  population  of  40.5 million (Ojigo and Dabom, 2013; World Bank, 2010) as well as 60% of the livestock population and most of the country’s wildlife (Ngugi and Nyariki, 2005). Most of Kenya’s small-scale farmers occupy mainly this region, pursuing traditional livestock production with traditional technologies. These farmers are unlikely to meet the growing demand for food from an increasing population (Leisinger and Schmitt, 1995; GoK, 2012).

Pastoralism  is   the   dominant   production   system  in Kenya’s arid and semi-arid lands (ASALs), but over time, it has been confronted by limited access to better farm technologies, requisite skills and market services (Otieno et al., 2012).  Further, weak linkages between research-extension service providers and farmers have hampered adoption of technologies by pastoralists. Until recently, pastoral areas were viewed as net consumers of national wealth, offering poor prospects of return on investment. As a result, productivity and growth have remained relatively low; despite the fact that the sub-sector is expected to play an important role in the development of these areas (Mugunieri and Omiti, 2007; Oluoch-Kosura, 2010).

Over the years the government was the sole provider of animal health services in the country. Empirical evidence however show high-potential areas and market-oriented livestock systems were better served (Oruko and Ndung’u, 2009), while marginal ASALs lacked adequate access to animal health services (Oruko et al., 2000). In mid-1980s, the country implemented the Structural Adjustment Programs (SAPs) which were characterized by market liberalization in the veterinary sector resulting in the gradual reduction of government involvement in the provision of AI services (Richter et al., 1990; De Haan and Bekure, 1991). Liberalization of animal health services started in 1992 with liberation of the pricing policy of milk and milk products followed by privatization and reduced involvement of government in animal breeding and artificial insemination (AI) services; input and veterinary drugs supplies; animal health care and dipping services; de-regulation of the processing and marketing of milk (Mudavadi et al., 2001). To bridge the gap, private sector providers were promoted as an alternative to state provision and as a means to reduce the government financial burden and improve efficiency of AI delivery (Tambi et al., 1999). Privatization of AI services increasingly became a necessity as government funding to veterinary dwindled, with the transfer of activities, functions, responsibilities and property from the public to the private providers.

In the context of AI services, privatization is viewed as a process of refocusing public sector by decentralising responsibilities, not merely as a means of reducing government expenditure, but also as an approach to increase its adoption by farmers in marginal areas, which require knowledge about the current and future demand, disease epidemiology, changing livestock systems, and socioeconomic changes (Kebede et al., 2014). The structural reforms led to reduction of the government and financial burden in the delivery of AI services but witnessed the reduction of farmers demanding the service. Consistent with the reforms, as much as 95% inseminations are now conducted by private AI service providers and cooperatives (Makoni et al., 2014). However, progress has remained slow and livestock production continue experiencing ineffective extension services and low adoption of AI.  This  is  attributed  by  a myriad of factors among them low demand for AI, vastness of the area, harsh terrain and hostile environment, poor road network infrastructure, which exacerbate the problems posed by the long distances between frontline personnel and pastoralists (GoK, 2010). 

The demand response, influenced by the farmers’ attitude towards the AI, determines the involvement and efficient delivery of the service in ASALs areas (Tambi et al., 1999). Scholars have argued that the problem of technology adoption by farmers are not only associated with the technology per se but also by socio-economic disparities and environmental challenges (Croppenstedt et al., 2011). From an economic perspective, the benefits of adoption of AI should create sufficient motivation to farmers to adopt the technology since the economic nature of the AI is expected to drive farmers who enjoyed these services to be willing to pay for them (Kartamulia et al., 1995). It has been shown that AI adoption involves decision on investment, transaction and opportunity costs (Ferraro and Simpson, 2002), and its benefits should at least compensate farmers for the associated costs.

Successfully participation of private sector in the AI delivery require preparedness of all actors to engage in open processes and foster the self-confidence and local leadership necessary for their own lessons and capacities to bring about desired outcomes (Kebede et al., 2014). As argued by Rivera et al. (2009), privatization of AI services will depend on farmers’ willing to pay for these services and where extension services have previously been provided free of charge, assessment should be made to understand commercial demand for agricultural information.

So far studies have shown that farmers often make decisions regarding uptake of new or improved agricultural technology with enhanced efficiency in delivery, and its adoption depends on awareness about the technology and willingness to pay for it. Establishing the opinion of stakeholders is thus crucial before introduction of a technology since it shapes the direction of their adoption and diffusion (Kimenju and De Groote, 2008). Promoting AI requires determination of the “price” which will not lead to inefficiencies and ineffective outcomes (OECD, 2010; Wunder, 2007). Quantification of these costs is often constrained by lack of information on the factors that a farmer includes in the decision-making process as well as information asymmetries that allows providers to overestimate the opportunity cost of AI delivery. Thus, contingency valuation methods are increasingly being used to estimate the willingness to pay (WTP) on the side of the buyer. Studies evaluating farmer’s willingness to pay for AI services among pastoral farmers in Kenya are very rare, thus a knowledge gap. It’s on this basis that the study sought to understand pastoralist’s willingness to pay for AI and empirically determine farmer’s socio-economic characteristics which make  them  more  or  less  likely  towards  paying  for  AI services. This information is very important for both County governments and private service providers participating in the provision of AI services in arid and semi-arid lands (ASALs).

 

 

Study area

Data for this study was collected between November 2016 and January 2017 from Narok and Kajiado counties located in southern Kenya (Figure 1). Narok County lies between latitudes 0° 50´ and 1°50´ South and longitude 35°28´ and 36°25´ East covering an area of 17,933 km². In 2012, the population of the county was 979,770 and 169,220 households. The county is home to the famous Maasai Mara Game Reserve, one of the most popular tourism destinations in Kenya.  The rangelands surrounding the Maasai Mara National Reserve can be divided into three range units based on bio-geographic and climatic differences. The western unit consists mainly of grasslands and comprises the Maasai Mara National Reserve. The Loita Plains stretch out in the North eastern part of the study area and are covered by dwarf shrub and whistling thorn (Acacia drepanolobium) grasslands. The eastern area, with the Siana Hills and Plains, supports Croton dichogamus bush and several other woody species interspersed with grasslands (Stelfox et al., 1986).

 

 

The dominant vegetation in the county includes forest land in the Mau area and grasslands and shrubs in the lowland areas of Suswa, in Narok North, Osupuko and Loita divisions in Narok South as well as the Mara sections in Transmara. These areas are suitable for livestock  rearing  and  irrigation. Rainfalls  amounts  are influenced by the passage of inter tropical convergence zones giving rise to bi-modal rainfall pattern. Long rains are experienced between the months of February and June while the short rains are experienced between August and November. Rainfall ranges from 2,500 mm in wet season to 500 mm during the dry season. In 2017, the population of Narok county was projected to be 1,239,320 (Narok County Government, 2013).  

Kajiado county on the other hand is bounded between latitudes 10° 0’ and 30° 0’ South and longitudes 36° 5’ and 37° 5’ East with an area of 21,900 km². In 2012, the population of the county was 804,796 distributed in 173,464 households (Kajiado County Government, 2013). The main   physical features in the County are plains, valleys and occasional volcanic hills ranging from an altitude of 500 m above sea level at Lake Magadi to 2500 m above sea level in Ngong Hills. The county is divided into three different areas namely; Rift Valley, Athi Kapiti plains and Central Broken Ground. Vegetation type in the county is determined by altitude, soil type and rainfall. The county has a bi-modal rainfall pattern, with the short rains fall between October and December while the long rains fall between March and May. The rainfall amount ranges from as low as 300 mm in the Amboseli basin to as high as 1250 mm in the Ngong hills and the slopes of Mt. Kilimanjaro. Temperatures vary both with altitude and season (Amwata, 2013; Bobadoye et al., 2014). 

The two counties are inhabited by the Maasai community who are mainly pastoralists, that is, at least 50% of their livelihoods depend on domestic livestock (Swift, 1988). Pastoralists differ from livestock rangers by their practice of taking herds to pasture and water, rather than having fodder grown or brought to them although purely nomadic in the past, many pastoralists are less mobile today (Fratkin and Roth, 2005). Pastoralism is the main source of livelihood to majority of rural households in the both counties. The most common  livestock  kept  are  dairy  and  beef  cattle, goat and sheep, with milk, meat, hide and skin, wool and mutton as the main products. Most families move between sedentary and mobile activities, while the larger part of a family, mainly women, children and elderly, have settled down pursuing small scale subsistence farming, some family members (often young men) still take the herds to pastures and water.

Sampling procedure

Multistage sampling technique was used. In the first stage, Keiyan, Kilgoris and Lolgorian divisions of Narok County and Namanga, Mashuru, Ngong and central divisions of Kajiado County were purposively selected because of their large concentrations of Sahiwal cattle populations. Moreover, these are high ranching zones suitable for Sahiwal production. In the second stage, pastoralist populations in these areas were divided into two strata based on their production systems, that is, Agro-pastoralists and Nomadic pastoralists using stratified random sampling technique. Third stage involved acquisition of lists of both nomadic and agro-pastoralists from District Livestock Development Officers (DLPO’s) where systematic random sampling technique was applied to each list to obtain 205 agro-pastoralist and 179 nomadic pastoralists households for interview.

This sample size was calculated using the proportion sample size determination formula as given by Mugenda and Mugenda (1999).

                                      

where n is the desired sample size of livestock farmers in Narok and Kajiado Counties, z is the standard normal deviate at the required confidence level, p is the proportion in the target population estimated to have characteristics of interest, q is 1- p, and d is the level of statistical significance set.

Analytical framework

Following the analytical framework of Hanemann et al. (1991), WTP for AI services by respondents was estimated using open-ended questions asking the respondents to declare the maximum amount they would be willing to pay, or close-ended, asking the respondents if they would be willing to pay a specific amount or not (dichotomous choice). In the current study, a closed-ended question approach was adopted given that most of the pastoralists were aware of the AI but could not arbitrarily attach a true value to the service. Moreover this approach is easier and more realistic since questions correspond more to a real market situation. On the other hand, the open-ended format is appropriate when the farmer is well informed about the new technology or product and its characteristics. However, literature indicates that such an approach would be misleading if the respondent lacks appropriate information and incentives to comprehensively determine the values to attach if a market were to exist (Boyle, 2017). 

The use of contingent valuation (CV) methods to estimate farmers’ valuation of non-market goods or new technologies as developed by social economists is not common, but it is widely used in environmental studies, wildlife conservation and natural resource economics (Hanemann et al., 1991). The technique is appropriate in imploring producers’ WTP for a product that is not yet on the market, such as AI. Applicability of this approach demands that the researcher crafts a hypothetical market for non-market good, requests a set of subjects to operate in that market, and records the outcomes. The values generated through this hypothetical market are treated as estimates of the value of the non-market good or service (Mitchell and Carson, 1989).

In many transactions, farmers are offered a technology at a given price such that after considering his ability to buy, the decision is then reached on whether to buy or not.  Estimating WTP using single-bounded method, the individual only responds to one bid which is incentive-compatible; it is in the respondent’s strategic interest to say ‘‘yes” if his WTP is greater or equal to the price asked, and ‘‘no” otherwise (Mitchell and Carson, 1989). Utility maximization implies that a farmer will then only answer ‘‘yes” to the offered bid if his maximum WTP is greater than the bid. However, the single-bounded method requires a large sample size and is statistically inefficient (Hanemann et al., 1991). In order to ensure efficiency of the estimates, double bounded method was adapted by offering the respondent a second bid, higher or lower depending on the first response. This approach includes more information about the respondents WTP and, therefore, provides more efficient estimates and tighter confidence intervals (ibid). Table 1 presents the definition of variables included in the model used.

 

 

The respondent was asked if he/she was willing to pay an amount B_i, for the provision of AI services on his farm per animal. If the farmer answers no then it can be assumed that 0 ≤ WTP <B_i, if he answers yes then B_i ≤ WTP < ∞. More explicitly, the respondents will fall within one of the following categories: The farmer answers yes to the first question and no to the second question, then B_i^u >B_i thus it can be inferred that B_i ≤ WTP < B_i^u. The individual answers yes to the first question and yes to the second question, then B_i^u ≤ WTP < ∞. The individual answers no to the first question and yes to the second question, then B_i^d <B_i, thus conclude that B_i^d ≤ WTP < B_i,. The individual answers no the first and second questions, then we have 0 < WTP < B_i^d.

Adapting the modelling framework of Hanemann et al. (1991), the likelihoods of these outcomes are π^yy, πⁿⁿ,  π^yn, π^ny, respectively. Under the assumption of utility-maximizing farmer, the formulas for these likelihoods are as shown below. In the first case where the respondent accepts the initial and second higher bid, we have

                                                        

In the second case where the respondent rejects the initial bid and second lower bid, we have 

                                                                                

Third case is where the respondent accepts the initial bid and rejects the second bid, we have  B_i^u>B_{i ;}

                          

The last case is where the respondents rejects the initial bid and accepts the second bid, we have

                 

Computing the mean willingness to pay, a logistic curve was specified, fitted on the data and estimated. The log-likelihood function was then defined as follows and estimated:

                                                                                                                                                                                                             

where and are binary-valued  indicator variables.

The final step was to specify and estimate a WTP regression model to determine factors influencing WTP. The regression method allows inclusion of other factors in the analysis, in particular socioeconomic characteristics of the respondents to explain the bidding behaviour. However, WTP can be computed with or without inclusion of covariates in the modelling strategy as illustrated by Lopez-Feldman (2012).

 

 

Farmers’ awareness and willingness to accept AI services

Prior to establishing the extent to which farmers would be willing to pay for AI, it is imperative that we determine if they are aware of this technology. To illustrate this, descriptive analysis was carried out and the results presented in Figure 1. It was evident that 89.9 and 90.3% of farmers in Kajiado and Narok Counties respectively knew about the existence of AI services. This provides much needed background upon which sensitization needs to be built in order to achieve maximum diffusion of the technology. However, awareness of AI does not guarantee its uptake as noted by Chinese consumers towards biotech rice. Lin et al. (2006) found that consumers who were aware of biotech foods were less inclined to purchase biotech rice than those who had no or little awareness. Moreover, the impact of the awareness variable was not statistically significant in the case of biotech soybean oil (ibid).

Despite high levels of awareness, most farmers still use the natural service method as the major breeding method. This could be attributed to the fact that accessibility of the services is still a major challenge due to infrastructural constraints, faith in the technique and the communities’ preference for bull service as reported by Janssen-Tapken et al. (2006). Inadequacy in number of AI technicians and accompanying inputs would hamper farmers’ access to and adoption AI technology (Kassa and Wuletaw, 2018). Despite AI as a technology being technically beneficial, lack of appropriate delivery system, its adoption and effectiveness in pastoral areas is declining. The consequence of this fall is undesirable because the genetic potential and productivity of the dairy herd in ASALs is bound to decline very rapidly.The inadequate incentives for both public and private breeders to practice in ASALs have been a major hindrance to the spread of AI services across from the highlands to the ASALs.

The long distances that must be covered by a service provider between one household to another in ASALs and to the nearest markets and the cost incurred outweigh the revenues that are likely to be generated from such business. This therefore necessitates deliberate government intervention in deploying public AI service providers and facilitates their movements within these areas (ibid).

The decision to pay for a particular technology depends solely on the prior response on the willingness to accept it. This underscores the importance of qualitative studies on perceptions of both producers and consumers of services and goods before introducing them in the market. The question of amount is only relevant if the farmer is willing to accept AI otherwise a hypothetical scenario has to be created to entice him to reveal his willingness to accept (Boyle, 2017). This is based on the assumption that there are underlying constraints to access AI (accessibility, cost and success rate) such that if they are addressed then they may be willing to value the technology. Figure 2 indicates that among farmers who were aware of AI (89.9% in Kajiado and 90.3% in Narok), 38.8% from Kajiado were willing to accept and adopt compared to their counterparts (23.3%) in Narok County. This implies that farmers have reservations about the adoption of AI despite wide spread knowledge about AI. Current study findings are inconsistent with the findings of Dehinenet et al. (2014) who found awareness of diary technologies through livestock training to have increased farmer’s probability of adopting and owning the improved technologies.

 

 

Monetary valuation for artificial insemination in pastoral areas

To ensure sustainability of the technology in pastoral areas, farmers were presented with different bids to establish amount they were willing to pay for AI. On average, 51.7% of the sampled farmers in Kajiado were willing to pay the initial bid proposed to them. Table 2 illustrates farmers bidding behavior with respect to different bids that were given.

 

 

The results also indicates that as the bid increases from KES 600 to KES 3000, the number of farmers affirming their ability to incur that cost declines. This is rational of farmers because as the cost of a new technology increases, given their cost outlay, they pursue a minimization objective and keep their production goals intact. The second bid is contingent on the response and amount indicated by the farmer in the initial bidding (Hanemann et al., 1991; Boyle, 2017). It is evident from Table 3 that farmers were willing to pay a second bid 48.9 and 50.5% in Kajiado and Narok counties respectively. The second bid offered was either a discount to the first bid offered for those farmers who declined  to  pay  initial  bid or a premium on the initial bid for the farmer who were willing to pay initial bid as the true price for getting AI. The bidding behaviour of farmers towards the second bid was similar such that as the amount increases, then few are willing to incur such cost as can be seen when bid rises from KES 400 to KES 3600 as shown in Table 3.

 

 

Table 4 shows results of a double bounded contingent valuation approach without including covariates. The results revealed an average of KES 1881.25 as the mean WTP for AI by pastoralists in ASALs of Kajiado and Narok Counties. This reflects a premium of 25.4% placed on AI by pastoralists with reference to the base price of KES 1500 offered for exotic breeds in Kenyan highland.

 

 

However, the bidding decision by the farmer is informed by various factors including his awareness towards AI, access to credit facilities to finance new technologies, herd size, household size, age, education levels, access to extension services, and his off-farm income.  It’s worth noting that in expressing the amount they are willing to pay for the provision of the service, there is implied price comparison between the cost of the bid and the cost of acquiring the bull. Sahiwal bull at market price at that time was KES 120,000 if bought from KALRO – Naivasha and KES 80,000 if bought from the local markets. However, the survey revealed that most Sahiwal farmers interested in replacing the bull or acquiring an extra one would prefer getting it from KALRO. Inclusion of factors influencing the bidding behavior of the farmer, the Mean WTP for AI services reduces to KES 1853.19. This reflects a deviation of KES of 353.19 (23.5% of base price of KES 1500). As indicated in Table 5, awareness, herd size and access to extension  had  significant  positive  influence  while  farm income had significant negative effect on farmer’s bidding process.

 

 

Farmer’s ability to purchase new technologies depends on his/her disposable income given existing production cost outlay. In this study, off-farm income had a positive significant effect in establishing farmer’s WTP for AI services in pastoral areas. This could be attributed to the fact that pastoralists with extra income have the ability to buy more productive technologies to increase their output. These results confirm findings of Kimenju and De Groote (2008) who found consumers with higher income to have high WTP for fortified maize. Availability of off-farm income has a positive effect on technology adoption with little necessity to seek credit from lending facilities for most farmers in rural areas (Mwangi and Kariuki, 2015; Mmbando and Lloyd, 2017). This implies that farmers with off-farm income have higher propensity for new technologies.  

 

 

 

 

 

 

 

To ensure sustainability of the adopted technology, it is imperative that the beneficiaries be willing to financially and materially support its existence. In the current study, most farmers showed their willingness to accept AI technology  despite   challenges   in  accessing   service providers. Existence of enabling market environment will motivate private service providers to operate in Kenyan pastoral areas. It was established that most farmers were WTP an average of KES 1853.2 for AI per cow. This reflects a premium of 23.6% placed on AI by pastoralists with reference to the base price of KES 1500 offered for exotic breeds in Kenyan highland. It is therefore recommended that both county governments and non-governmental organizations organize field days for pastoralists so that relevant information about AI is disseminated and free trials done on lead farmers. Moreover, government should consider ensuring high quality semen is distributed to pastoralist at subsidized rate till they gain confidence in the technology. This is because adoption of AI has the potential in easing the demand of the Sahiwal bull from an already limited supply.

 

The authors have not declared any conflict of interests.