Vol. 8(2), pp. 39-51, February, 2016
DOI: 10.5897/JDAE2015.0660
Article Number: 818711256963
ISSN 2006-9774
Copyright ©2016
Author(s) retain the copyright of this article
http://www.academicjournals.org/JDAE
Journal of Development and Agricultural
Economics
Full Length Research Paper
Technical efficiency of smallholder wheat farmers: The
case of Welmera district, Central Oromia, Ethiopia
Wudineh Getahun Tiruneh
1
* and Endrias Geta
2
1
Department of Agricultural Economics, Extension and Gender Research (AEEGR), Ethiopian Institute of Agricultural
Research, Holetta Research Center, P. O. Box 2003, Addis Ababa, Ethiopia.
2
Haramaya University, P. O. Box 2003, Addis Ababa, Ethiopia.
Received 9 May, 2015; Accepted 9 December, 2015
Increasing productivity through enhancing efficiency in cereal production in general and in wheat
production in particular could be an important pace towards achieving food security. However, the
strategic conceptual and empirical analysis in the context of the efficiency, which would guide policy
makers and development practitioners in their efforts to revamp cereal productivity, is sparse. This
study was undertaken to assess the technical efficiency and factors affecting efficiency of wheat
production in Welmera district of Oromia region, Ethiopia. The primary data pertaining to farm
production, input usage, and socioeconomic and institutional factors were collected during 2012/13
cropping year through a structured questionnaire from randomly selected 180 wheat farmers. The
stochastic frontier and translog functional form with a one-step approach were employed to assess
efficiency and factors affecting efficiency in wheat production. The maximum likelihood estimates for
the inefficiency parameter depicted that most farmers in the study area were not efficient. The mean
technical efficiency was found to be 57%. Factors such as sex, age and education level of the
household head, livestock holding, group membership, farm size, fragmentation, tenure status and
investment in inorganic fertilizers affect efficiency positively and distance to all weather roads
negatively. The finding implies that there is an opportunity to improve technical efficiency among the
farmers by 43% through gender-sensitive agricultural intervention, group approach extension, and
attention to farmers’ education, scaling out of best farm practices.
Key words: Smallholder wheat farms, translog production function, technical efficiency.
INTRODUCTION
In Ethiopia, agriculture is the major option for stimulating
growth, overcoming poverty, enhancing food security and
improving distribution of income among the poor
households. It contributes about 42% to the total gross
domestic product (GDP), provides 85% of employment
opportunities, constitutes more than 80% of the nation‟s
total exports, and provides most of the foreign exchange
earnings to the economy (EPA, 2012). It also plays an
important role in providing raw materials for domestic
industries. Thus, Ethiopia‟s Growth and Transformation
*Corresponding Author. E-mail: getawudineh@gmail.com Tel: +251 911706033. Fax: 251112370377.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution
License 4.0 International License
40 J. Dev. Agric. Econ.
Plan (GTP I) set higher growth and investment targets in
agricultural sector in general and in wheat production in
particular than any of earlier Ethiopia‟s national plan and
will receive a special attention in the next five year plan
(GTP II) (MoFED, 2010). Cereal production and
marketing are the means of livelihood for millions of
smallholder households and making it the single largest
sub-sector in Ethiopian economy. Cereal accounts for
roughly 60% of rural employment, 80% of total cultivated
land, more than 40% of a typical household‟s food
expenditure, and more than 60% of total caloric intake,
represents about 30% of GDP (World Bank, 2007).
Following maize, wheat is the second most important and
productive cereal crop and its productivity shows
increasing pattern (for example increased from 18.39 to
2.1 tons per hectare in 2010/2011 and 2012/2013
cropping season, respectively (CSA, 2010, 2013).
Following South Africa, Ethiopia is the second largest
producer of wheat in sub-Saharan Africa. At a national
level about 1.63 million ha wheat was distributed with
about 4.84 million smallholder farmers (CSA, 2013).
Wheat is cultivated in the highlands of Ethiopia, mainly in
Oromia, Amhara, Southern Nations and Nationalities
Peoples (SNNP) and Tigray regions (CSA, 2013) and it is
the first most important staple crop in Welmera district.
Currently, wheat is among a few crops which have
received special attention from the Government of
Ethiopia and NGOs operating in the country. In this
regard, the government has paid attention to research
and extension of wheat technologies. Moreover, Ethiopia
has become a center of diversity in Eastern Africa for its
wheat crop (EAAPP, 2009).
Despite the importance of wheat as a food and
industrial crop and the efforts made so far to generate
and disseminate improved production technologies, its
productivity remains below its potential. The average
wheat yield was about 2.1 tons per hectare, in 2012/2013
cropping season (CSA, 2013). Ethiopia‟s current annual
wheat production of approximately 3.18 million tons is
insufficient to meet domestic needs, forcing the country to
import 30 to 50% of the annual wheat grain required.
Therefore, these facts show that Ethiopia is the net
importer of wheat to feed its growing population.
Moreover, the yield gap of over 3 tons per hectare
suggests that there is a potential for increasing production
and productivity of smallholder wheat farmers.
Some previous studies have indicated that farm
production and productivity can possibly be raised (1) by
allocating more area for production, (2) by developing
and adopting of new wheat technologies, and/or (3) by
utilizing the available resources more efficiently (Ahmed
et al., 2013; Kamruzzaman and Mohammad, 2008; Haji,
2006). Opting for the first method would mean trying to
boost output at the cost of bringing marginal areas into
cultivation. Some other authors also argued that with
limited available suitable land especially in the highlands
for cultivated area expansion, increased cereal
production and productivity will need to come from yield
upgrading (Bezabeh et al., 2014; Taffesse et al., 2012).
On the other hand, creation and introduction of new
technologies is a long term option and requires a lot of
capital for research and extension. Rather, efficient
utilization of available resources is the best way of
increasing production especially in the short run.
According to previous researches in Ethiopia, there
also exists a wide cereal yield gap among the farmers
that might be attributed to many factors such as lack of
knowledge and information on how to use new crop
technologies, poor management, biotic, climate factors
and more others (Debebe et al., 2015; Ahmed et al.,
2013; Yami et al., 2013).
Because of the scanty resources that are on ground,
recently it is getting importance to use these resources at
the optimum level which can be determined by efficiency
searches (Gebregziabher et al., 2012; Asefa, 2012; Alene
et al., 2006). Thus, increasing wheat production and
productivity among smallholder producers requires a
good knowledge of the current efficiency or inefficiency
level inherent in the sector as well as factors responsible
for this level of efficiency or inefficiency. However,
previous studies in the area of wheat production
efficiency are not extensive and crop specific, and are
also area specific (Wassie, 2014; Yami et al., 2013;
Mussa et al., 2012; Kebede and Adenew, 2011; Alene
and Zeller, 2005). These studies have been at the
household level ignoring the possible differences in bio-
physical conditions at the plot level, and also their
findings are not consistent with one another due to
various reasons like agro ecological and methodological
variations. Moreover, based on these literature reviews
and to the best of the information we have, no studies
have estimated technical efficiency of wheat farmers in
Welmera district. That is, information on the levels of farm
household technical efficiency and its determinant factors
is lacking in the study area.
Therefore, the present study is an attempt towards
assessing the technical efficiency of the farmers in the
study area and aims to bridge the prevailing information
gap on the contextual factors contributing to efficiency
differentials in the production of wheat. The objective of
the study is to measure technical efficiency of wheat
production and to identify variables affecting technical
efficiency of wheat producing farmers.
RESEARCH METHODOLOGY
Study area
The study was conducted in Welmera district of Addis Ababa Zuria
Special zone of Oromia, Regional State in Ethiopia. Welmera
district is one of the eight administrative units of the Addis Ababa
Zuria Special zone of Oromia Regional State. Geographically, the
district is located between 8°50'-9°15'N latitude and 38°25'-38°45'E
longitude and has area coverage of 66,247 ha (WORLA, 2011).
Most of its areas are high lands (Dega) and mid highlands (Weyna
Dega) with an altitude ranging from 2060 to 3380 m above sea
level. Majority of the soil is reddish-brown clayey type similar to
some other highland areas of Ethiopia (Asefa, 2012). The district is
sub-divided in to 23 rural kebele (Kebele is the lowest
administrative unit under Ethiopian condition) administrations and
one town, excluding the capital town of the district. The area is
characterized by mixed crop-livestock farming systems like other
central highlands of Ethiopia where both crop and livestock
production play a central role in the lives of the farming community.
Wheat is the first major staple crop followed by barley, tef, pulses,
oilseed, potato and other crops, respectively in the area. In
2011/2012 cropping season, about 33% of the crop land was
covered by wheat (WOA, 2012).
Sampling procedure
In order to select sample farm households, a three-stage sampling
technique was employed. In the first stage, study district was
purposively selected based on the extent of wheat production. In
the second stage, six kebeles were selected from the selected
district based on the discussion with district level agricultural
extension experts. Finally, from up-to-date list of sampling frame
(wheat growers) obtained from extension offices at each Kebele
level, 180 sample households were selected using systematic
random sampling. The sample size was determined by adopting a
sample size determination formula provided by Statistics Canada
(2010).
Data source and collection
This study used the data collected from primary sources for
2012/2013 production season. To supplement the primary data,
secondary data were collected from concerned district offices (like
Agricultural Office, Holetta Agricultural Research Center (HARC)
and Cooperative Offices) and from published and unpublished
sources. The data is cross-sectional and quantitative in nature.
Primary data contained detailed information on households‟
socioeconomic and demographic characteristics, farm charac-
teristics, inputs utilization, output produced, institutional, policy
related variables and production problems encountered were
collected from the selected farm households using structured
questionnaires filled by trained enumerators who are fluent in the
local language. Close supervision and day to day check up was
done by the researcher. The survey was conducted from July to
August, 2013.
Data analysis
To achieve the study‟s objectives, both descriptive and inferential
statistics were used. Descriptive statistics like means, standard
deviations, percentages and frequency counts were used in
describing socioeconomic characteristics of households, inputs,
output variables, frequency distribution efficiency levels and
responses on the constraints of wheat production. The stochastic
frontier production function and the inefficiency model are
simultaneously estimated with the maximum likelihood method
using the econometric software, FRONTIER 4.1 computer
programme.
Analytical framework
In this study, the stochastic frontier analysis approach was adopted
to measure the technical efficiency of wheat farms. The model was
independently proposed by Aigner et al. (1977) and Meeusen and
Tiruneh and Geta 41
Broeck (1977). The merits for this approach over Data Envelopment
Analysis (DEA) (non-parametric) is that it accounts for a composite
error term (one for statistical noise and another for technical
inefficiency effects) in the specification and estimation of the frontier
production function. For a number of reasons, the stochastic frontier
analysis (econometric) approach has generally been preferred in
the empirical application of stochastic production function model in
the developing countries‟ agriculture like Ethiopia. This might be
due to first the assumption that all deviations from the frontier arise
from inefficiency as postulated by DEA is hard to accept, given the
inherent variability of smallholder agricultural production due to
external factors like pests and weather conditions. Second, most
farms are very small and operated by family labor and hence farm
records kept rarely. The available data on wheat production are
most likely subject to measurement errors. Therefore, the stochastic
frontier production required for estimating plot level efficiency is
specified as:
) (1)
where Y
i
denotes the output for the i
th
sample farm, X
i
represents a
(1 × K) vector whose values are functions of inputs and explanatory
variables for the i
th
farm, β is a (K × 1) vector of unknown production
parameters to be estimated, V
i
s are assumed to be independent
and identically distributed random errors which have normal
distribution with mean zero and unknown variables, , that is,
and U
i
s are non-negative unobservable associated
with the technical inefficiency of production such that for a given
technology and levels of inputs, the observed output falls short of its
potential output ( or it is a one-sided error term (U ≥ 0)
efficiency component that represents the technical inefficiency of
the farm. In short, U
i
estimates the shortfall in output Y
i
of wheat
from its maximum value given by the stochastic frontier function.
In other words, the basis of a frontier function can be illustrated with
a farm using n inputs for wheat (X
1
,X
2
,….., X
n
) to produce output Y
of wheat. Efficient transformation of inputs into output is
characterized by the production function f(X
i
), which shows the
maximum output obtainable from various input vectors. The
stochastic frontier production function assumes the presence of
technical inefficiency of production. Hence, the function is defined
as:
=252 (2)
where is the error term that is composed of two elements,
and plot level data was collected from a total of n=252 wheat plots.
The stochastic frontier analysis has been used in many studies
like by Yami et al. (2013), Beshir et al. (2012), Jaime and Salazar
(2011), Tan et al. (2010), and Daniel et al. (2008) and the approach
specifies technical efficiency as the ratio of the observed output to
the frontier output, that means the technical efficiency of an
individual farmer or farm is defined as the ratio of observed output
and the corresponding frontier output, given the state of available
technology, and presented as follows:
= (3)
where F (X
i
;β).exp(v
i
-u
i
) is the observed output (Y)
and F
(X
i;
β).exp(v
i
) is the frontier output(Y
*
). Pursuing Battese and Coelli
(1995), the error term (v
i
) permits random variations in output due to
factors outside the control of the farmer like weather and diseases
as well as measurement error in the output variable, and is
assumed to be identically, independently and normally distributed
with mean zero and constant variance ( ); that is, v
i
~N(0, ).
42 J. Dev. Agric. Econ.
The u
i
is the inefficiency component of the error term and a one-
sided non-negative (u>0) random variable, is assumed to be
independently distributed as truncations at µ of the normal
distribution and variance ( ), that is, u
i
~N (µ
i
, ), but if u
i
= 0,
the assumed distribution is half-normal. The technical inefficiency
model suggested by Battese and Coelli (1995) is illustrated by:
µ
i
= Z
i
δ
i
(4)
where Z
i
is a (1 × M) vector of exogenous explanatory variables
associated with the technical inefficiency effects in the i
th
time
period, δ
i
is an (M × 1) vector of unknown parameter to be
estimated.
As mentioned earlier in the literature review, this study employed
the single stage maximum likelihood estimation method used in
estimating the technical efficiency levels and its determinants
simultaneously. This estimation procedure guarantees that the
assumption of independent distribution of the inefficiency error term
is not violated. The maximum likelihood estimation of the stochastic
frontier model yields the estimate for beta (β), sigma squared (σ
2
)
and gamma (γ), and are variance parameters; γ measures the total
variation of observed output from its frontier output. The study used
the parameterization following Battese and Coelli (1995) and is
given as,
22
2
uv
and
)(
22
2
uv
u
, where the
gamma lies between zero and one (0 ≤ γ ≤ 1). If the value is very
close to zero, then the deviations are as a result of random factors
and/or if the value is very close to 1, then the deviations are as a
result of inefficiency factors from the frontier.
Model specification
Following Aigner et al. (1977), the translog production function has
been used recently by many studies to estimate technical
inefficiency (Geta et al., 2013; Yami et al., 2013; Beshir et al.,
2012). Therefore, the translog production function stated in
Equation 6 is used for the study for its flexibility for which it places
no restriction unlike the Cobb-Douglas production function.
(Cobb-Douglas) (5)
(6)
where i=1,2,- - - n=252, and X= vector of five input variables.
Based on the aforementioned model, a stochastic frontier model
for wheat farmers is given by:
ln(output)
i
= β
0
+ β
1
ln(Area)
i
+ β
2
ln(Fert)
i
+ β
3
ln(Oxndays)
i
+
β
4
ln(seed)
i
+ β
5
ln(lab)
i
+ 1/2 β
11
ln(Area)
2
+ 1/2 β
22
ln(Fert)
2
+ 1/2
β
33
ln(Oxndays)
2
+ 1/2 β
44
ln(seed)
2
+ 1/2 β
55
ln(lab)
2
+ β
12
ln(Area)
ln(Fert) + β
13
ln(Area) ln(Oxndays) + β
14
ln(Area)ln(seed) +
β
15
ln(Area) ln(lab) + β
23
ln(Fert) ln(Oxndays) + β
24
ln(Fert) ln(seed) +
β
25
ln(Fert) ln(lab)
+ β
34
ln(Oxndays) ln(seed) + β
35
ln(Oxndays)
ln(lab) + β
45
ln(seed) ln(lab) + vi
- u
i
(7)
where output represents total yield of the i
th
plot in kilo gram (kg);
Area represents operational area of wheat of the i
th
plot in hectare
(ha); Fert represents the total amount of inorganic fertilizers used
per plot in kg; Oxndays represents the amount of oxen days used
for plowing from land preparation to planting, Seed represents the
amount of seed used per plot in kg; Lab represents the total cost of
labour per day estimated at market price, and in Ethiopia farmers
use herbicides instead of hand weeding, therefore, it is included
that the cost of herbicide per liter estimated at market price in the
total cost of labour for different farm activities, and ln represents
Natural logarithm.
The specification of inefficiency model for the target commodity
of individual producer is given as:
(8)
µ
i
= δ
0
+ δ
1
Sex + δ
2
Age + δ
3
Educ + δ
4
Fsize +δ
5
Proxwroad +
δ
6
Acredit + δ
7
Livestock + δ
8
Offrmy + δ
9
Gpmship + δ
10
Ext + δ
11
Train
+ δ
12
Frmsize + δ
13
Frgmnt + δ
14
Tenurstatus + δ
15
Costfert
(9)
where Sex is 1 if the household head is male, 0 otherwise; Age
represents the age of the household in years; Educ stands for the
education level of the household in years of formal education
completed; Fsize stands for the size of the family, is converted into
the same unit (Labour Force); Proxwroad is the distance from the
household residence to the nearest all weather road in walking
minutes; Acredit is the amount of agricultural credit received in
Ethiopian Birr (ETB; Birr is the Ethiopian currency); Livestock
represents the number of livestock owned in TLU; Offrmy is cash
income earned from off-farm activities in ETB; Gpmemship is a
dummy variable with a value =1 if the household participate in more
than one farmers group, 0 otherwise; Ext stands for the number of
extension contact (made with DAs and experts); Train stands for
the number of trainings (on new varieties, diseases and pests, crop
management) taken; Farm size stands for the total area of farm
land under operation (own land + rented in + share in) in hectare;
Frgmnt stands for land fragmentation, the number of wheat plots;
Tenurstatus is a dummy variable, with a value of 1 if the i
th
farmer
used his own farm plot, 0 otherwise, and Costfert stands for the
proportional cost of chemical fertilizer to its variable costs incurred
by the i
th
farmer per plot measured in ETB during 2012 cropping
season.
Hypotheses testing
In spite of the magnitude and significance of the variable
parameter, γ, it is also important to explain the various null
hypotheses employed in this work. Three hypotheses were tested
to scrutinize the adequacy of the specified model used in this study,
the presence of inefficiency and exogenous variables to explain
inefficiency among smallholder wheat producers. The generalized
likelihood ratio statistics was used to test the hypotheses. It is
specified as:
LR (λ) = -2 [{lnL(H
0
)}- {lnL(H
1
)}] (10)
where L(H
0
) and L(H
1
) are the values of the likelihood functions
derived from restricted (null) and unrestricted (alternative)
hypothesis. This has a chi-square distribution with degree of
freedom equal to the difference between the numbers of estimated
parameters under H
1
and H
0
. Yet, where the test involves a γ, then
the mixed chi-square distribution is used. The H
0
is rejected when
the estimated chi-square is greater than the critical.
RESULTS AND DISCUSSION
Descriptive statistics
The results of descriptive statistics for the entire variables
considered are presented in Table 2 for their mean,
minimum, maximum and standard deviation values for
continuous variables and frequencies and percents for
discrete variables. The result shows that the average
wheat productivity was 1.9 ton/ha and relatively lower
than the national average of 2.11 ton/ha for the same
cropping season (CSA, 2013). The yield was obtained by
using 153.2 kg/ha of seed, 134.46 kg/ha of fertilizers
(DAP + Urea), 17.25 oxen days/ha and 1282.9 ETB/ha of
cost of labor incurred including the cost of herbicides
(substituted for labor weed). The average size of farm
allocated for wheat was 0.68 ha from a total average of
2.5 ha. This indicates that an average household
allocated more than 27% of the farm land for wheat.
The average size of the household in labor force unit
(LFU; is a conversion factor estimated by categorizing the
age groups into nine and identifying six major farm
activities (herding and domestic chores, land preparation,
planting, weeding, harvesting and threshing, and
transporting) with key informants through FGDs, then the
key informants asked to give weight (0 to 4) to each
activity for each age group, the weight was aggregated
and divided by four times six = “1” is set equal to an able-
bodied adult equivalent) was 3.55. The conversion factor
used in estimating family members into LFU varies
according to circumstances. In the developed countries,
family size, labor power and dependency ratio has been
estimated simply by counting the number of individuals
whose age fall in defined “working-age group” or
„dependent” ranges using the standard method. Sharp
(2003) felt the standard method inadequate and used an
innovative approach to estimating the actual labor
capacity of family members based on his fieldwork
(survey) in the study of measuring destitution. This study
also felt the work of Sharp is inadequate to the context of
the study area, because it ignores the supply of labor by
elderly people who are over 60 years old and did not
consider gender differential in labor supply for the
different agricultural activities. Therefore, the study used
a (LFU)
conversion factor obtained from own informal
qualitative survey through conducting six focal group
discussions at each Kebele (Appendix).
The average livestock holding for sample households
was 7.83 TLU, earned an average off-farm income of
3961.60 ETB, the average amount of credit received by
households was 926.40 ETB, the average number of
wheat plot was one ranging from 1 to 6, and about 34%
of production expenditure was incurred for applying
fertilizers compared to its variable costs. The average
number of contact made by extension staffs with wheat
household for crop related information was 7, and wheat
growers received a one day crop specific trainings.
Membership in a farmers‟ group (MFG) indexes social
group. All of the households (100%) reported that they
are organized in one to five farmer groups and 32% of
the households reported that they belonging to more than
one farmer‟s group either in crop production and/or in
Tiruneh and Geta 43
dairy cooperatives. On average the sample households
spend about 20 min walk to reach the nearest all weather
roads.
Estimation of stochastic frontier production
Before proceeding to the analyses of technical efficiency
and its determinants, it was necessary to select the
appropriate functional form and detect the presence of
inefficiency in the production of wheat for the sample
households. In a one step modeling approach, both
Cobb-Douglas and translog frontier model can be used.
Various restrictions were imposed on the model defined
by 4 and 6. To check whether these restrictions were
valid or not, the generalized likelihood ratio tests were
used. The results of these tests of hypothesis for
parameters of the stochastic frontier and inefficiency
effects model for wheat farms in Welmera district are
presented in Table 3. The first null hypothesis tested was
that the coefficients of the interaction terms of input
variables are zero favoring the Cobb-Douglas functional
form (H
0
: β
ij
= 0). The values of the logarithm of likelihood
function for Cobb-Douglas and translog frontier model
were -107.33 and 30.25, respectively. Therefore, the
generalized likelihood ratio test is used to decide the
functional form as follows:
LR (λ) = -2 [{lnL(H
0
)}- {lnL(H
1
)}]
= -2 [-107.33 + 30.25] = 154.16
The value of the likelihood ratio statistics was found to be
154.16 and greater than the critical χ
2
value of 18.3 with
10 degree of freedom at 5% level of significance. the null
hypothesis was rejected and thus the translog functional
form is preferred to Cobb-Douglas functional form for the
data and more precise and consistent results. The
second null hypothesis which specifies technical
inefficiency effects are absent in the model (H
0
:
γ = δ
0
=
δ
1
= --- =δ
15
=0), or all wheat farmers/farms efficient in the
study area were tested against the alternative (H
1
: γ > 0
and δ
i
≠ 0 where i = 0,1, ---, 15) rejected with generalized
likelihood ratio test statistic of 95 which was larger than
2.7 critical values at 5% significance level with 1 degree
of freedom (Table 1) (Kodde and Palm, 1986) implying
that the stochastic production function had a better fit to
the data than the average production functions. In short,
H
0
: γ = 0, all wheat producers/farms are 100% efficient
and is strongly rejected. This indicates that the ex-
planatory variables specified in the model make a
significant contribution in explaining the inefficiency effect
associated with wheat production in the study sites. The
third null hypothesis, H
0
: δ
1
= --- δ
15
=0, which specifies
that the coefficients of the explanatory variables in the
efficiency model are simultaneously zero and is strongly
rejected with generalized likelihood ratio test statistics of
49.56 which was greater than 24.99 critical values