Factors That Influence Technical Efficiency of Sorghum Production: A Case of Small Holder Sorghum Producers in Lower Eastern Kenya

,


Introduction
Grain sorghum, Sorghum bicolor (L.)Moench, is the h most important cereal crop grown in the world [ ]. Probably because of its versatility and diversity [ ], sorghum is mainly growninthearidandsemiaridlands(ASALs)ofAfricaand Asia for rural food security.e future of sorghum enterprise is linked to the contributions of sorghum to food security, income growth, and alleviation of poverty.
is is more relevant in developing countries in the African continent than in other developed nations.
I nK e n y aso r gh u mi satr a d i ti o n alc r o p ,wh i c hi sgr o w n in many parts of the country especially in the ASALs o ft h ec o u n t r y .
ec r o pl o s tf a v o u rw i t hf a r m e r sw h e n maize became the preferred crop and staple food a er its introduction by the European settlers.However, due to the desire to stabilize food security in the country there is now renewed interest in promoting drought-tolerant crops such as sorghum, which are known to be well adapted to harsh environments [ ].
A lot of research on sorghum breeding has been going on in Sub-Saharan Africa (SSA).Stable, high-yielding sorghum varieties (HYSVs) have recently been developed [ ].In Kenya, for example, the initiatives to promote sorghum production are mostly concentrated in the ASALs.Eastern Kenya is characterized by increasingly frequent drought occurrences, sometimes extending for two-to-three years in a stretch.Over the past two decades, there have been repeated Advances in Agriculture maize crop failures in many parts of eastern Kenya especially because of droughts [ ]. Sorghum promotion in this region is undertaken as a government strategy to enable the people to meet household food security needs and increase rural income [ , , ].
e area under sorghum production in Kenya has been increasing from , ha in to , ha in , but the national average yield per hectare has been decreasing from .metric tons (MTs) per hectare to .MTs/ha over t h es a m ep e r i o d[ ].Several public e orts supplemented by those of nongovernmental organizations (NGOs) and other stakeholders like International Sorghum and Millet (INTSORMIL) program and International Crop Research Institute for the Semi-Arid Tropics (ICRISAT) have, for instance, provided interventions for harnessing opportunities for productivity enhancement (HOPE), targeted at improving productivity and marketing of sorghum.ese interventions have included breeding, distribution of improved HYSVs that are pest and disease tolerant, and promotion of resource conserving management practices.In spite of all these e orts, there has been marked variability in production from the expected potential yields and the actual yields.e expected potential yield, for instance, for the Gadam sorghum variety is -.MTs/ha −1 but farmers have only realized production of up to .MTs/ha −1 so far [ , ]. Variability in production is a function of di erences in scales of operation, production technologies, operating environment, and operating e ciency [ ]. Production increases depend mainly on the e cient use of available appropriate technologies but not necessarily on adoption rates of new technologies [ ]. Chimai [ ] noted that, for the small-holder farmers, variation in production due to di erences in e ciency may be a ected by various factors, which include regional and farm speci c socioeconomic factors.For purposes of policy implications in e ciency analysis, it is very important to identify factors that in uence e ciency. is study, therefore, aims at identifying the farm and farmer characteristics that in uence levels of technical e ciency and estimating the marginal e ects of these factors among the small holder sorghum producers in lower eastern Kenya.
Various studies have measured technical e ciency and its determinants among di erent types of farmers and countries, which provide useful information for this study.However, e ciency in these studies is relative and tends to be speci c to farmers' groups and country under study.A number of empirical studies have attempted to investigate the relationship between technical e ciency and various socioeconomic variables and demographic factors such as levels of formal education, age, family size, access to credit, extension services, and experience [ -].However, technical ine ciency may arise primarily due to managerial incompetence and, therefore, e ciency di erences could be explained in the context of the management characteristics such as training, experience, and motivation [ , , ].Other factors identi ed include membership in agricultural associations, land ownership, value of household assets, use of fertilizers, and tillage methods adopted [ , , , ].While some of the factors identi ed in studies can provide a general idea of what in uences e ciency, generalization may not be possible because each country and agricultural product has unique characteristics.
ere are many approaches used in the identi cation of these factors, which may vary to some extent with the methodology employed.e most commonly followed procedureinmostofthea pproachesiswha tisusuallyreferred to as the two-step procedure.In the rst step, the e ciency or ine ciency score is estimated.Secondly, the estimated score is taken as a dependent variable and is then regressed against a number of other explanatory variables that are hypothesized to a ect e ciency levels [ ].
e various methods used in regression include ordinary least square (OLS) and Tobit regression models.
To achieve the objective of this study, the two-step procedure was employed where technical e ciency results from Chepng' etich's et al e population of interest for this research comprised sorghum growing households (HHs) in selected districts in the lower eastern Kenya, at least the HHs that grew sorghum in the -cropping season.A sample size of farm households, and HHs in Makindu and Machakos districts, respectively, was determined proportionately using the total population of the districts.A multistage sampling procedure was employed.First sorghum farmers Advances in Agriculture were selected using purposive sampling method with the help of extension o cers in the two districts and then the selected farm households were subjected to systematic simple random sampling where every th sorghum farmer was selected to achieve the required sample size.
Data was collected from sorghum farmers between June and August by use of pretested semistructured questionnaires administered by trained enumerators.Information on demographic, institutional, physical, and socioeconomic factors, yields, and inputs used to grow sorghum by each HH in -cropping season was collected.
. .e Tobit Model.An econometric analysis (censored regression model) based on the two-limit Tobit model was used to identify farm and farmers characteristics a ecting smallholder e ciency in sorghum production.
is was undertaken with the help of STATA so ware program.e technical e ciency scores generated from DEA model as described by Chepng' etich et al. [ ] were then regressed on the selected farm and farmer characteristics variables in order to identify their in uence on technical e ciency.Technical e ciency scores range between and ; hence the two-limit Tobit reg ression mo del [ ] was used as shown below: where refers to the th decision making unit (DMU); is the e ciency scores of the th DMU; * is the latent e ciency; are parameters to be estimated; is an error term that is independently and normally distributed with mean zero and common variance of 2 ( NI ; 2 ); and are host of farm and farmers characteristics variables.us, the Tobit model used in this study is speci ed as follows:

Results and Discussion
. .Descriptive Analysis . . .Farmer Socioeconomic Characteristics/Pro les.Majority ofthehouseholdssurveyedintheregionweremale-headed with less than a quarter of the total sampled population being female-headed.Speci cally, about % of the households in Machakos and .% in Makindu were male-headed.Most of the sampled household heads had low levels of formal education.Almost half ( %) of all the farmers captured in the survey terminated their formal education at primary level, % of the HH heads (HHHs) having not attended a n yf o rmalschoola tall.I tisno tedtha tu ptoabo u t %o f the sampled HHs had less than years of sorghum farming experience, while % had more than years of the same farming experience.e number of years of sorghum farming experience di ers across the region.For instance, a high number of the sampled farmers ( .%) had less than years of the farming experience in Machakos district, while less than % had more than ve-year experience in the same district.In Makindu, % of the sample had more than ve years of the sorghum farming experience.Intermsofage,mostofthesampledHHHswererelatively old as indicated by the mean age in each category.e youngest sampled HHHs were and years old, while theoldestHHHwas and yearsoldinM achakosand Makindu districts, respectively, with an average age of years in both districts as shown in Table .A m o n ga l lt h e surveyed HHHs, the young HHHs in the -years age bracket were only % and middle-aged HHHs between a n d y ea r so l dc o n s ti t u t ed % ,wh il eth em a jo ri tyo fth e sampled HHHs ( %) were in the age bracket of -years.Only % of the HHHs were above years old.
e HHs with smallest household sizes had two persons, while those with largest HH sizes had persons, with an average of six members per household as summarized in Table .About % of the total sampled HHs had at most persons, while approximately % of the household surveyed had six persons per household in the region.Household assets ranged from KES , to KES . million with a mean of about KES , .Eighty percent of the households had assets value falling below KES million.es m a l l e s ts i z eo f land used for sorghum production was .ha, while the largest was .ha, with an average land size of .ha dedicated for sorghum production (Table It is observed that most of the households used improved varieties of sorghum seeds like the Gadam, Seredo, and Serens varieties.e farmers used seed rates ranging from .kg/ha to kg/ha with a mean seed rate of .kg/ha as summarized in Table .Twenty-six percent of the farmers used the recommended seed rate of kg/ha.Twenty-ve percent used seed rate of kg/ha which is half the recommended seed rate, while % used double the recommended rate of kg/ha.
Methods of land preparation varied among the surveyed HHs, but the most common method practised was use of ox-plough which up to % of the respondents used.
irty percent of the surveyed households did not prepare their land before planting, whereas not more than % of the HHs used hand digging as a land preparation method before planting.In more speci c terms, majority of the households in Makindu district never prepared their land before planting as was observed that % of all the sampled HHs in Makindu district did not actually plough their farms beforeplanting.However,about %oftheMakindusampled HHs who ploughed their farms before planting did so using the ox-ploughing method.Although none of the sampled HHs in Makindu used hand digging as a land preparation method, approximately % of the sampled HHs prepared their sorghum planting land using hand digging in Machakos district.
Most o en some farmers spread manure in their farms beforeplanting.Halfofthesampledhouseholdsusedmanure, while the remaining half did not.e labour used in the sorghum production was mainly provided by family memb e r s .O n l y %o ft h es a m p l eu s e dh i r e dl a b o u ri no n eo r more of the activities undertaken during the sorghum production processes.Hired labour was engaged mainly during land preparation, especially by those households who used ox-ploughing method.More labour was used in sorghum farms, especially during bird scaring where up to hours per day could be spent in the farm during that period in which the whole bird scaring operation lasted for about three-to-four weeks.
. . .Economic Activities.Households in lower eastern r e g i o no fK e n y ae n g a g e di nv a r i o u se c o n o m i ca c t i v i t i e sa s sources of earning incomes.
ese activities ranged from crop farming, livestock keeping, charcoal burning, small businesses, and formal employment to casual labour employment.Most of the households depended on farming as their only source of income.Fi y-four percent depended on farming as a sole income earner to meet the household daily requirements, while the rest of the sampled population were either employed, having small businesses, or involved in casual labour employment in addition to farming.
Most of households ( %) earned less than KES , per month as o -farm income.irty percent of them did not access any o -farm income.Most of crops grown in the region were maize, green grams, pigeon peas, beans, sorghum, sweet potatoes, and vegetables such as kales, spinach, and tomatoes.Livestock kept by the farmers were mainly goats, chicken, and bees with a few farmers having cattle especially oxen to provide animal power for ploughing and pulling ox-cart for carrying water and other farm produce.More than half of the households surveyed did not get any income from their crops and livestock because they produced mainly for subsistence and only sold the surplus, which was hardly there.Only a few earned more than KES , per month from livestock or per season in case of crops (Figure ).

. . . Transport and Communication.
e market proximity f r o mt h ef a r m sw a so na v e r a g e k ma w a yw h e r et h e sa m p ledfa rmer ss pen ta na v era g eo fKES ino ner et urn journey.e return distances travelled by the sampled HHs t om a r k e t sr a n g e df r o mh a l fak i l o m e t e rt oa b o u t k m .e most commonly used available means of transport was motorcycles.However, a motor vehicle was available once or twice a week only during the open market days for those HHs living in the most interior regions.

. . . Access to Agricultural Credits and Extension Services.
While it is known that access to credits for sorghum production and marketing is important, it was observed that only a few sampled HHs ( %) accessed necessary agricultural credits.A majority of the sampled HHs ( %) did not access any credits for sorghum production and marketing.
It was also observed that a reasonable proportion of sampled households ( %) received production advice from extension services.e most common advice given was on agronomic practices and the best varieties to grow in the ASALs environment.ese pieces of advice were received by a majority of the sampled respondents ( %).Only % of the respondents were given information on availability of good markets.In addition, households belonging to farmers clubs or association were %.
. .Factors In uencing Technical E ciency.Many variables were found to in uence technical e ciency of sorghum production in the lower eastern Kenya.Out of the variables analysed, seven were found to in uence technical e ciency positively and were found to be statistically signi cant at % level as shown in Table .esevariablesincludeeduca tion levels of household heads in terms of years spent in formal schooling, number of years of experience in sorghum farming, HH membership in farmer clubs or associations, size of land planted with sorghum, hired labour, use of manure, and production advice on sorghum production. is implies that increasing use of these factors in the sorghum production processes would improve technical e ciency of sorghum production.Only one variable, the size of HHs, was found to in uence technical e ciency negatively and was found to be statistically signi cant at % level.
Experience in sorghum farming was found to be positive and signi cant at % level.is implies that as years pass with continuous sorghum farming, farmers tend to increase their capacity to do better in sorghum farming; hence, they become more technically e cient.Over time, the farmers are better placed to acquire more knowledge and skills necessary for choosing appropriate new farm technologies.ese ndings areinlinewiththoseofGuletal.[] and Padilla-Fernandez and Nuthall [ ]  Membership in farmer clubs or associations in uenced positively technical e ciency. is suggests that households that belonged to farmer associations or clubs or related organizations were more likely to bene t from better access to inputs such as improved sorghum varieties and information on improved farming practices.Similar results were also realised by Kariuki et  Being a member of an organisation provides an avenue for information and technology transfer by extension agents and o en leads to sharing of information even among members themselves. is enables the household heads to make appropriate decisions, which in the long run enhance productivity and e ciency.Moreover, in some farmer associations, certain farming practices and new technologies are undertaken together as a group.
ese practices enhance learning and pooling of labour resources.
e area planted with sorghum had a signi cant direct relationship on technical e ciency only in Makindu district.is indicates that e ciency of sorghum production increases with size of land under sorghum, but this relationship was not signi cant in Machakos district.Notably, most of thehouseholdsinMakinduplantedsorghumonlargerfarms as compared with those in Machakos hence engaging hired labour in the production of sorghum.Hired labour is highly associated with e ciency because of its high productivity per unit labour.Similar positive results of farm size and technical e ciency have been observed by Chiona  Hired labour had a positive and statistically signi cant relationship with technical e ciency at % level.Use of hired labour proved more e cient than family labour in the production of sorghum.e productivity per unit of hired labour was high. is could be attributed to the fact that hired labour acts as an incentive for the households to be more e cient in supervisory role because of the cost incurred.Comparable results were reported by Elibariki and Shuji [ ] and Chimai [ ] Interestingly, the size of the household was negative and signi cant at % level in explaining the technical e ciency, implying that, as the household size expanded, the technical e ciency of sorghum production decreased.e researchers argued that household size increased the labour available, hence increase in the technical e ciency.But, according to Nchare [ ], it is observed that although householdsizeincreasesthelabouravailableforproduction, it is usually associated with production ine ciency hence technical ine ciency.is can be explained by the abundance of that available labour at farm level, which will lower the productivity per unit labour.
Other variables such as male-headed households, number of dependants, household assets, use of improved seed varieties, seed rate, household o -farm incomes, income Advances in Agriculture Advances in Agriculture from other farm activities, and use of credits were positive for technical e ciency but not necessarily signi cant at % level.Variables such as age of the household head and land preparation methods were actually negative for technical e ciency and were not signi cant at % level.Despite being expected to in uence technical e ciency negatively, age of the HHH was found to vary between the districts.In Makindu, age of the HHH had the expected positive sign although it was not signi cant.Amaza et al.
[ ] argued that older farmers were likely to be less e cient than younger ones. is is because younger farmers were likely to be more progressive and more willing to adopt new agronomical practices hence higher e ciency in production.On the other hand, Elibariki and Shuji [ ]f o u n da g eo f the household head to have a positive in uence on technical e ciency. is argument was based on the assertion that as farmers grow old they gain more experience in the production of various agricultural practices; hence they become more e cient.
Assets possessed by the household had the expected positive sign though not signi cant.According to Chimai [ ] assets are taken to indicate the household wealth status.In regard to smallholder farmers, assets are expected to in uence technical e ciency positively. is is because assets act as shock absorbers, especially when sold o in times of need.eincomefromtheassetscouldbeusedtopurchase inputs and to hire various production practices.
Cr edi tusewasf o undtoha vetheexpectedposi tivesign t h o u g hn o ts i g n i c a n ta t %l e v e l .A sa r g u e db yN c h a r e [ ] and Amaza et al. [ ] credit was expected to reduce the nancial di culties the farmers usually face especially atthebeginningoftheproductionprocess.ecreditcould enable the farmers to have the capital to purchase inputs and to have resources to prepare their land on time before planting.Padilla-Fernandez and Nuthall [ ]a r g u e dt h a t credit to the farmers may act as an instrumental motivation to produce more e ciently apart from being able to purchase the required inputs for production.
Seed rate was expected to in uence technical e ciency negatively but on the contrary it had a positive sign though not signi cant at % level.is could be attributed to most of the households found using low seed rate compared with the recommended rates.Overall, household sites were tested and those households residing in Makindu were positive but not signi cant. is means that belonging to Makindu was not enough to signi cantly in uence a farmer to attain higher levels of e ciency.
In Machakos district, education of the household head, experience in sorghum farming, membership in farmers associa tio ns,man ur euse,andfarmersr eceivingp r od uctio n advice were all positive and displayed % signi cant levels.Other variables such as age of the household head, number of dependants, assets possessed by the households, seed rate of sorghum used, use of improved seed varieties, size of land planted, use of credits, and amount of household o -farm income were all positive for technical e ciency although they did not display signi cant di erences at % level.Variables such as male-headed households, household size, land preparation method, and income from other farm activities were negative for technical e ciency and were not signi cant at % level.e di erences in the two districts occurred in variables such as size of land planted with sorghum, which was positive for technical e ciency in Machakos district, while negative in Makindu district though not signi cant at % level in both districts.Household size was negative for technical e ciency and signi cant at % level in Makindu, while in though negative for technical e ciency it was n o ts i g n i c a n ta t %l e v e l .O t h e rv a r i a b l e st h a td i s p l a y e d di erences were male-headed households and income from other farm activities, which were both positive for technical e ciency in Makindu district but negative for technical e ciency in Machakos, though they were all not signi cant at % level in both districts.
e results from the Tobit model were also subjected to postestimation test using marginal e ect analysis in order to estimate the trivial change from each factor that in uences technical e ciency (TE).Quanti cation of the marginal e ects of these variables is important in order to estimate the change that will occur with respecttoachangeinoneunitofthatvariable.Asshownin Table ,thevariablesthatweresigni cantat %levelinthe Tobit model have the highest change in production quantity per hectare (Kg/ha). is implies that a change in one unit of thevariableinquestionwouldcauseabiggerchangeinterms of Kg/ha of sorghum harvested.For instance if hired labour is improved by one unit it will increase sorghum output by approximately Kg/ha and Kg/ha in Makindu and Machakos, respectively.

Conclusion and Recommendation
In this study factors in uencing technical e ciency of sorghum production were investigated in a sample of sorghum producing households in lower eastern Kenya using both descriptive statistics and Tobit models.
e empirical ndings show that some variables such as the formal education levels of the HHHs, years of sorghum farming experience, membership in farmer associations, land sizes planted with sorghum, hired labour, use of manure, and farmers receiving production advice all displayed positive technical e ciency regimes at % statistical signi cance level.i si m p l i e st h a ta ni n c r e a s eo rad e c r e a s ei na n yo ft h e s e variables would cause an increase or a decrease in technical e ciency of sorghum production, respectively.Although negative, the household size was found statistically signi cant a t %l e v e li m p l y i n gt h a ta ni n c r e a s eo ft h eh o u s e h o l ds i z e would decrease the technical e ciency and vice versa.
Sorghum farmers are not fully technically e cient and therefore there is room for e ciency improvement, which can be undertaken by addressing important variables that either positively or negatively in uence levels of technical e ciency in the lower eastern Kenya through policy formulation or review.Formal education is an important positively in uencing factor of technical e ciency in sorghum production.
us appropriate policy formulation and/or review should be Advances in Agriculture . . .Farm Characteristics.Most of the land parcels used for sorghum production were individually owned.Only % o ft h es a m p l e dh o u s e h o l d so w n e dl a n dc o m m u n a l l y ,w h i l e %o w n e dl a n dt h r o u g hl e a s e h o l d s .
income from livestock and other crops in the lower eastern Kenya.Source: survey results [ ].
but are contrary to those of Ajewole and Folayan [ ]. Household heads with more years of formal schooling werefoundtobemoree cientthantheircounterpartswith lessyearsofformalschooling.eseresultsareinagreement with the ndings of Amaza et al. [ ], Nyagaka et al. [ ], Chirwa [ ], Shehu et al. [ ] ,A j e w o l ea n dF o l a y a n[ ], Elibariki and Shuji [ ], Mussa et al. [ ], Idiong [ ], and Njeru [ ].All these studies have argued that producers with high formal education levels (≥ years) are able to detect a n dr e d u c ei n e c i e n c yi np r o d u c t i o n .F o r m a l l ye d u c a t e d farmers are generally better placed to receive, interpret, and respond to new information.Education increases chances to adopt and respond rapidly to the use of improved appropriate technologies such as water harvesting and soil conservation technologies and the agronomic practices.Less educated farmers are less receptive to improved farming techniques andprovidepoorsupervisionontheirfarms[ , ].
[ ], Chirwa [ ], Elibariki and Shuji [ ], and Gul et al. [ ], but on the contrary Chimai [ ]a n dJ a v e de ta l .[]f o u n dt h a ts i z e of land planted with sorghum had a negative in uence on technical e ciency.
. Production advice given to the sorghum producing households had a direct and signi cant relationship with technical e ciency.Production advice increased technical e ciency. is corroborates with the work of Chiona [ ], Nchare [ ], Amaza et al. [ ], Ajewole and Folayan [ ], Javed et al. [ ], and Wakili [ ]. rough such production advice, farmers were able to get rst-hand information on new agricultural innovations and techniques that would ensure increased sorghum production in the study area.
Similar results were obtained by Chimai [ ], Nchare [ ], Yusuf and Malomo [ ], and Mussa et al. [ ]w h oa l lf o u n d o u tt h a tH Hs i z eh a dan e g a t i v ei n u e n c eo nt e c h n i c a l e ciency.On the contrary, Ajewole and Folayan [ ], Shehu et al. [ ], and Wakili [ ]f o u n dt h a th o u s e h o l ds i z ew a s positive and signi cant in explaining technical e ciency.

T:
Tobit model results showing farm and farmer characteristics that in uence technical e ciency in lower eastern Kenya.signi cance at %, HHs: households, and HHH: household head.
. [ ] research were regressed against a number of explanatory variables hypothesized with the help of the Tobit regression model.Since technical e ciency scores lie between and , the dependent variable in the regression model did not have normal distribution.issuggests that the OLS regression was not appropriate and estimation with OLS would have led to biased parameters estimates.etechnical e ciency scores were continuous; hence, Probit and Logit models could not be used either because they are only used when the dependent variable takes two values Agrcredit is agricultural credit, Othrincm is other income, Srgmfarmsize is sorghum farm size, Ndependents is number of dependents, Srgmseed is sorghum seed rate used, Manure is manure use, Improvseed is improved seed, Clubmbr is club member, and Expr is experience in sorghum farming.
age + 3 H/edu + 4 Prodadvice + 5 Adptill + 6 Hlabortill + 7 O ncm + 8 Asset + 9 Agrcredit + 10 Othrincm + 11 Srgmfarmsize + 12 Ndependents + 13 Srgmseed + 14 Manure + 15 Improvseed + 16 Clubmbr + 17 Expr + , ( ) where E score is e ciency score, Malehd is male-headed household, H/age is age of household head, H/edu is formal education of household head; Prodadvice is production advice, Adptill is land preparation method, Hlabortill is hired labour, O ncm is o -farm income, Asset is assets, : Summary of descriptive statistics of selected farmer and farm characteristics in the lower eastern Kenya.c e n to ft h es a m p l e dH H su s e dl e s st h a n h ao fl a n d to grow sorghum out of which majority of them (about %) used .ha to grow sorghum.Land sizes planted with sorghum were conspicuously di erent among households in the two districts.As observed in Table ,t h ea v e r a g e land sizes planted with sorghum were .ha and .ha in Machakos and Makindu districts, respectively.Majority of thehouseholdsinMakinduplantedsorghumon .haland sizes, as opposed to majority of households in Machakos who planted sorghum on .ha land sizes.
). Ninety-seven Advances in Agriculture T Notes: a person day = hrs work per day for an adult person; HH: household and Dist.: district.T: Summary of descriptive statistics of selected farm characteristics in the lower eastern Kenya.Notes: a person day = hrs work per day for an adult person; HH: household and Dist.: district.