ANALYSIS OF FACTORS AFFECTING TECHNICAL
EFFICIENCY OF SMALLHOLDER MAIZE FARMERS IN
ETHIOPIA
BY SORSIE GUTEMA DEME
Submitted in partial fulfilment of the requirement for the degree
M.SC. (AGRICULTURAL ECONOMICS)
In the
SUPERVISOR(S):MS N. MATTHEWS FACULTY OF NATURAL AND AGRICULTURAL
SCIENCES
MR J. HENNING DEPARTMENT OF AGRICULTURAL ECONOMICS
UNIVERSITY OF THE FREE STATE
APRIL 2014 BLOEMFONTEIN
DECLARATION
I, Sorsie Gutema Deme, hereby declare that this dissertation submitted for the degree of
Master of Science in Agricultural Economics, at the University of the Free State, is my own
independent work and has not previously been submitted by me for a degree at this or any
other University, and that all material contained herein has been duly acknowledged. I further
cede copyright of the dissertation in favor of the University of the Free State.
_________________________
Sorsie Gutema Deme
UFS, Bloemfontein
April 2014
ii
DEDICATION
I dedicate this work to my family.
iii
ACKNOWLEDGEMENTS
I wish to express my sincere gratitude to the following persons and institutions that made this
study a success. My gratitude first goes to N. Matthews and J. Henning, my supervisors, for
your guidance and constructive suggestions throughout the study. I gratefully acknowledge your
patience and assistance without which this dissertation would have not been completed in the due
time.
I also want to gratefully acknowledge the staff members of the Department of Agricultural
Economics: Prof. B.J. Willemse, Prof. B. Grove, Dr. D. Strydom, Dr. H. Jordaan, Dr Geyer
and Abiodun Ogundeji for creating a student friendly academic environment and assisting me
when required. In addition, I would like to acknowledge Ms Louise Hoffman, Ms Chrizna van
der Merwe and Ms Ina Combrinck that are friendly and supportive staff members of the
Department.
I appreciate my fellow students and friends namely Mrs Sylvia Israel-Akinbo, Dr. Mussie,
Zerizghy, Ms Amah Ampong, Mr Toba Fadeyi and Mrs Elizabeth Girmay for their invaluable
help during my study. Thank you for being there for me.
I am thankful towards the Central Statistical Agency of Ethiopia, Ethiopian Development
Research Institute and African Capacity Building Foundation for granting me a joint
scholarship fund to further my studies. I also wish to thank the Central Statistical Agency of
Ethiopia for providing the data used in this dissertation, with special thanks to Mr Solomon
Gizaw for his invaluable assistance in compiling the data in the format I want.
I would like to express my utmost thanks to my dear husband Oljira Kuma for the motivation
and encouragement you have been giving me. Thank you for your care, commitment and
understanding, especially for taking absolute responsibility for our children during my study.
My son, Aboma Oljira and my daughter, Elili Oljira, you are my greatest blessings always. I
am very grateful for having such wonderful children in life which inspires and motivates me
even in the middle of hardships.
iv
Finally, I would like to praise almighty God for His unfailing love and the wonderful
opportunities I received from his hand.
v
ABSTRACT
Agriculture is the dominant sector of the Ethiopian economy which typically consists of
smallholder rain fed farming systems. Low production and productivity characterises
Ethiopian agriculture resulting in the country being unable to meet the increasing food demand
of its population. As a result, the country continuously faces food insecurity and to some
extent relies on food aid and food imports. The key to growth of agricultural production in
Ethiopia lies in increasing the productivity and efficiency of smallholder farmers. The
Ethiopian government has given substantial policy emphasis to increased productivity of
smallholder crop farmers through the Agricultural Development Led Industrialization (ADLI)
strategy. The ADLI strategy emphasises on increasing the adoption and intensification of yield
enhancing inputs such as fertilisers and improved seeds to boost crop productivity, especially
maize which is the principal crop. In response to the efforts of the development strategy,
substantial improvements in the adoption and utilisation of the yield enhancing inputs have
been observed in maize production; however the maize yield is not showing expected
improvements. The low levels of maize productivity might be the result of technical
inefficiencies existing in smallholder production. Information about the technical efficiency of
smallholder maize farmers at farm level is important for improvements in productivity.
However in Ethiopia this information is limited making an empirical study of the technical
efficiency necessary. The research investigated the factors affecting the technical efficiency of
smallholder maize farmers in Ethiopia with the aim of generating reliable information about
the level of technical efficiency and the factors affecting technical inefficiency of smallholder
maize production. Stochastic Frontier Analysis technique was employed and the data for the
research was secondary data obtained from the Central Statistical Agency of Ethiopia
consisting of 438 observations.
From the empirical estimation, it is found that nitrogen is an important input that can increase
maize productivity significantly. Seed and labour inputs are found statistically insignificant in
explaining maize production. The estimated value of , which is a parameter used to indicate
the proportion of total variance that is attributed to technical inefficiency is 0.99 and
significant. The value of  revealed that about 99% of the random variation in output of maize
production is attributed to the technical inefficiency component which indicates the
vi
importance of examining technical inefficiencies in maize production. The estimated mean
technical efficiency score of the sample is 77% with the minimum and maximum efficiency
scores of 3 to 96%, respectively. The mean technical efficiency implies that on average, the
sampled maize farmers are able to obtain 77% of their potential output using the current
production inputs. The finding suggested the presence of considerable levels of technical
inefficiency that contributed to decreased maize productivity. The farmers have the potential
to increase their maize production by about 23% by using their existing resources and
technology more efficiently. While examining the determinants of technical efficiency, age,
gender, household size, oxen, extension, irrigation, credit, seed type and soil protection were
found to be important factors affecting the technical efficiency of the sampled maize farmers.
The study revealed the possibility of improving the current low maize productivity by
removing the technical inefficiencies. The current level of low technical efficiency can be
addressed through increasing farmers’ access to rural credit and extension services, promoting
soil and land conservation practices and by promoting small-scale irrigation schemes.
Key Words: productivity, smallholder farmers, maize, technical efficiency, factors affecting
technical efficiency, stochastic frontier analysis
vii
TABLE OF CONTENTS
TITLE PAGE............................................................................................................................i
DECLARATION......................................................................................................................ii
DEDICATION..........................................................................................................................iii
ACKNOWLEDGEMENTS....................................................................................................iv
ABSTRACT.............................................................................................................................vi
TABLE OF CONTENTS......................................................................................................viii
LIST OF TABLES..................................................................................................................xii
LIST OF FIGURES...............................................................................................................xiii
LIST OF ACRONYMS.........................................................................................................xiv
CHAPTER ONE:INTRODUCTION................................................................ 1
1.1 BACKGROUND AND MOTIVATION ................................................................1
1.2 PROBLEM STATEMENT.....................................................................................2
1.3 RESEARCH OBJECTIVES ...................................................................................5
1.4 CHAPTER OUTLINE............................................................................................5
CHAPTER TWO:LITERATURE REVIEW .................................................. 6
2.1 INTRODUCTION ..................................................................................................6
2.2 THE CONCEPT OF PRODUCTIVITY.................................................................6
2.2.1 IMPORTANCE OF PRODUCTIVITY ....................................................................7
2.3 THE CONCEPT OF EFFICIENCY IN PRODUCTION .......................................8
2.3.1 TECHNICAL EFFICIENCY ...................................................................................8
2.3.2 ALLOCATIVE EFFICIENCY ...............................................................................10
2.4 MEASURING TECHNICAL EFFICIENCY.......................................................11
2.4.1 DATA ENVELOPMENT ANALYSIS .....................................................................12
2.4.2 STOCHASTIC FRONTIER ANALYSIS .................................................................13
2.5 THE PRODUCTION FRONTIER AND THE TECHNICAL INEFFICIENCY
MODEL................................................................................................................15
2.5.1 REVIEW OF INPUTS DEFINING PRODUCTION FRONTIER .........................15
viii
2.5.2 FACTORS AFFECTING TECHNICAL INEFFICIENCY.....................................16
2.5.2.1 Age, Gender and Education of Household Heads .................................................17
2.5.2.2 Household Size, Farm Size, Land Tenure and Oxen.............................................18
2.5.2.3 Extension, Irrigation, Credit and Off-farm Income...............................................21
2.5.2.4 Seed Type, Organic Fertiliser and Soil Protection ................................................23
2.6 CONCLUSION AND IMPLICATIONS FOR THE RESEARCH ......................23
CHAPTER THREE: THE STUDY AREA PROFILE AND NATURE OF
THE DATA EMPLOYED ............................................................................... 25
3.1 LOCATION AND TOPOGRAPHY ....................................................................25
3.2 AGRO-ECOLOGY...............................................................................................26
3.2.1 TIGRAY .................................................................................................................29
3.2.2 AMHARA ..............................................................................................................29
3.2.3 BENISHANGUL GUMUZ ....................................................................................30
3.2.4 SOUTHERN NATIONS NATIONALITIES AND PEOPLES.................................30
3.2.5 OROMIA ...............................................................................................................30
3.2.6 HARARI ................................................................................................................31
3.2.7 SOMALI ................................................................................................................31
3.2.8 DIRE DAWA .........................................................................................................31
3.3 AGRICULTURE ..................................................................................................32
3.3.1 FARMING SYSTEM AND RURAL LAND USE IN ETHIOPIA......................33
3.3.2 CROP PRODUCTION..........................................................................................34
3.3.2.1 Maize Production ..................................................................................................36
3.3.3 SOIL CONSERVATION AND LAND MANAGEMENT ........................................38
3.3.4 SUMMARY ............................................................................................................39
3.4 TYPE AND SOURCE OF DATA........................................................................40
3.5 CHARACTERISATION OF RESPONDENTS...................................................41
3.5.1 MAIZE YIELD.......................................................................................................41
3.5.2 SEED AND FERTILISER .....................................................................................42
3.5.3 FARM SIZE AND LABOUR .................................................................................45
3.5.4 SOCIO-ECONOMIC VARIABLES .......................................................................46
3.5.4.1 Age and Household Size .......................................................................................46
3.5.4.2 Gender and Education ...........................................................................................47
ix
3.5.4.3 Ownership of Resources (Tenure and Oxen) ........................................................49
3.5.4.4 Credit and Off-farm Income..................................................................................50
3.5.5 FARM MANAGEMENT PRACTISES ...................................................................53
3.5.5.1 Extension Services ................................................................................................53
3.5.5.2 Irrigation ................................................................................................................54
3.5.5.3 Soil Protection .......................................................................................................55
3.6 SUMMARY..........................................................................................................56
CHAPTER FOUR:METHODOLOGY.......................................................... 57
4.1 JUSTIFICATION OF THE STOCHASTIC FRONTIER ANALYSIS MODEL 57
4.2 ESTIMATING TECHNICAL EFFICIENCY USING STOCHASTIC FRONTIER
ANALYSIS ..........................................................................................................59
4.3 EMPIRICAL ESTIMATION OF TECHNICAL EFFICIENCY..........................61
4.3.1 CHOICE OF FUNCTIONAL FORMS..................................................................61
4.3.1.1 Production Frontier Model Specification .............................................................62
4.3.1.2 Variables Defining Production Frontier ................................................................62
4.3.2 TECHNICAL INEFFICIENCY MODEL SPECIFICATION.................................64
4.3.2.1 Variables Defining Technical Inefficiency Model ................................................65
4.3.2.2 Multicollinearity Problem .....................................................................................70
4.4 PRINCIPAL COMPONENT REGRESSION.........................................................71
4.4.1 THEORETICAL BASIS OF PCR ..........................................................................71
4.4.1.1 Estimation of Principal Components.....................................................................72
4.4.1.2 Regression with Principal Components.................................................................73
4.4.1.3 Determining the Significances of the standardised Variables Using Significant PCs
75
4.4.2 APPLICATION OF PCR ......................................................................................76
4.4.2.1 Estimation of Principal Components.....................................................................77
4.4.2.2 Determining Significance of the Principal Components .......................................78
4.5 HYPOTHESIS TESTING ....................................................................................79
x
CHAPTER FIVE:RESULTS AND DISCUSSION ....................................... 83
5.1 ANALYSIS OF PARAMETERS OF THE STOCHASTIC PRODUCTION
FRONTIER ..........................................................................................................83
5.2 TECHNICAL EFFICIENCY ESTIMATIONS ....................................................85
5.3 ANALYSIS OF TECHNICAL INEFFICIENCY MODEL PARAMETERS......86
5.3.1 SOCIO-ECONOMIC VARIABLES .......................................................................88
5.3.1.1 Age and Gender.....................................................................................................88
5.3.1.2 Household Size and Oxen .....................................................................................88
5.3.1.3 Credit and Off-farm Income .................................................................................89
5.3.2 FARM MANAGEMENT PRACTISES ...................................................................90
5.3.2.1 Extension and Irrigation ........................................................................................90
5.3.2.2 Seed Type and Soil Protection ..............................................................................91
CHAPTER SIX:SUMMARY, CONCLUSION AND IMPLICATIONS.... 93
6.1 SUMMARY..........................................................................................................93
6.1.1 BACKGROUND AND MOTIVATION ..................................................................93
6.1.2 PROBLEM STATEMENT AND RESEARCH OBJECTIVES ................................94
6.1.3 LITERATURE REVIEW ........................................................................................95
6.1.4 STUDY AREA PROFILE AND DATA ..................................................................97
6.1.5 METHODOLOGY.................................................................................................98
6.1.6 RESULTS AND DISCUSSION..............................................................................99
6.1.6.1 Technical Efficiency Estimation ...........................................................................99
6.1.6.2 Factors Affecting Technical Inefficiency............................................................100
6.2 CONCLUSION AND POLICY IMPLICATION...............................................102
6.2.1 POLICY IMPLICATIONS ...................................................................................103
6.2.3 IMPLICATIONS FOR FURTHER RESEARCH .................................................104
REFERENCES……………………………………………………….…….…………………………………………………104
APPENDIX A……………………………………………………….…….…………………………………………………...120
xi
LIST OF TABLES
Table 3.1: Total area cultivated (hectares) and total production (tons) of cereals,
pulses and oilseeds for smallholder farming in 2011/12 Meher season,
Ethiopia.............................................................................................................35
Table 3.2: Number of sampled households considered from the selected regions and
respective percentages in the secondary data set used………………….....40
Table 3.3: Summary statistics of maize yield of the sampled respondents during
2011/12 Meher season, Ethiopia.....................................................................41
Table 3.4: Summary statistics of seed and fertiliser use of respondents......................42
Table 3.5: Summary statistics of farm size and labour use of respondents ………....46
Table 3.6: Summary statistics of age of farm household heads and family size..........47
Table 4.1: Definitions of variables included in the production function, measurement
units and expected signs..................................................................................62
Table 4.2: Description of variables included in the technical inefficiency model,
measurement units and expected signs..........................................................66
Table 4.3: Principal Components estimated, eigenvalues and cumulative percentage
of variances explained....................................................................................76
Table 4.4: Maximum Likelihood Estimates of the retained PCs……………………..77
Table 4.5: Generalised Log likelihood tests of hypotheses …………………………..79
Table 5.1: Maximum Likelihood estimates of stochastic production frontier
parameters………………………………………………………………..…..82
Table 5.2: Maximum Likelihood estimation results of technical inefficiency model
parameters........................................................................................................86
xii
LIST OF FIGURES
Figure 2.1: Graphical representation of technical and allocative efficiency using two
inputs (X1& X2) and one output (Yi) .............................................................9
Figure 3.1: Graphical representation of the nine administrative regions and the two
city administrations of Ethiopia....................................................................26
Figure 3.2: Traditional Agro-ecological Zones of Ethiopia.............................................27
Figure 3.3: Graphical representation of smallholder farmers’ maize area cultivated,
total production and yield during 2003/04 to 2011/12 Meher season,
Ethiopia...........................................................................................................37
Figure 3.4: Distribution of respondents by type of seed used.........................................43
Figure 3.5: Distribution of respondents based on fertiliser use......................................44
Figure 3.6: Gender distribution of the household heads.................................................47
Figure 3.7: Distribution of respondents based on education of household heads.........48
Figure 3.8: Tenure distribution of the respondents.........................................................49
Figure 3.9: Oxen ownership distribution of the respondents.........................................50
Figure 3.10: Credit-access based distribution of the respondents……………………....51
Figure 3.11: Off-farm income participation based distribution of the respondents…...52
Figure 3.12: Extension service based distribution of respondents....................................53
Figure 3.13: Distribution of respondents based on irrigation use....................................54
Figure 3.14: Distribution of respondents based on soil protection...................................55
Figure 5.1: Cumulative Probability Distribution of Technical Efficiency Scores for
sample of smallholder maize farmers in Ethiopia………………………....84
xiii
LIST OF ACRONYMS
ADLI Agricultural Development Led Industrialisation
AE Allocative Efficiency
AEZs Agro-Ecological Zones
AISE Agricultural Input Supply Enterprise
CSA Central Statistical Agency
DAP Di Ammonium Phosphate
DEA Data Envelopment Analysis
DMU Decision Making Unit
ESE Ethiopian Seed Enterprise
FAO Food and Agricultural Organization
FARA Forum for Agricultural Research in Africa
FDRE Federal Democratic Republic of Ethiopia
GDP Gross Domestic Product
HDI Human Development Index
IFAD International Fund for Agricultural Development
IFPRI International Food Policy Research Institute
MoA Ministry of Agriculture
MoARD Ministry of Agriculture and Rural Development
MoE Ministry of Education
MoFED Ministry of Finance and Economic Development
MOI Ministry of Information
MFIs Micro-finance Institutions
MLE Maximum Likelihood Estimation
NMA National Metrological Agency
N Nitrogen
xiv
OECD Organization for Economic Cooperation and Development
OLS Ordinary Least Square
P Phosphorus
PCR Principal Component Regression
PCs Principal Components
RATES Regional Agricultural Trade Expansion Support
RUSACCOs Rural Saving and Credit Cooperatives
SFA Stochastic Frontier Analysis
SNNP Southern Nations Nationalities and Peoples
SPSS Statistical Package for Social Science
TE Technical Efficiency
UNDP United Nation Development Organisation
xv
CHAPTER ONE: INTRODUCTION
1.1 BACKGROUND AND MOTIVATION
Food insecurity and poverty are widespread and persistent in Sub-Saharan Africa with
approximately two-thirds of the population depending on agriculture for their livelihood
(Mupanda, 2009). Low agricultural productivity and an increased population are the main
problems that contribute to increased food insecurity and poverty in Sub-Saharan Africa in
general and in Ethiopia in particular (Mupanda, 2009; Geta, Bogale, Kassa & Elias, 2010).
The Ethiopian economy is highly dependent on agriculture which contributes approximately
43% of the Gross Domestic Product (GDP), 85% of employment and 90% of export earnings
(MoARD, 2010).Ethiopian agriculture is predominantly rainfed, smallholder farming which is
undertaken on areas averaging less than two hectares. About 12 million smallholder farmers
are engaged in agriculture, from which about 95% of agricultural GDP is earned (MoARD,
2010), whereas large and medium-scale commercial farms contribute five percent to
Ethiopia’s Agricultural GDP (CSA, 2011c).
Maize (Zea mays) is one of the most important food crops produced by smallholder farmers in
Ethiopia (CSA, 2012a). Maize is used as a staple food for human consumption, animal feed, a
source of raw materials for numerous industrial products and an important trade commodity
(Nigussie, Tanner & Twumasi-Afriyi, 2002;FARA, 2009). Maize is the most produced crop in
the country accounting for 28% of total grain produced during Meher season (September to
February) of 2011/12 (CSA, 2012a). Other staple cereal crops grown in Ethiopia are teff
(Eragrostis tef), sorghum (Sorghum vulgare) and wheat (Triticum aestivum) which make up
about 16%, 18% and 13% respectively of the total grain production in Meher production
season of 2011/2012 (CSA, 2012a). More than nine million smallholder farmers were
involved in maize cultivation on about two million hectares of land during the 2011/12 Meher
season (CSA, 2012a). Total production was about six million tons with an average yield of
about 2.95 ton per hectare (ton/ha) during the same season (CSA, 2012a). Due to the relative
role of maize in total grain production and due to the participation of a large number of
smallholder farmers in maize production, maize is a priority crop contributing to the country’s
1
national food security. Nevertheless, low production and low productivity characterises
Ethiopian agriculture, specifically maize production (World Bank, 2006; MoARD, 2009).
The country is naturally endowed with abundant arable land of about 51.3 million hectares and
numerous river basins that hold great potential for irrigation, making the country suitable for
agricultural development (MoARD, 2010). Despite having abundant resources for agricultural
development, there is a growing food shortage in the country attributed to the poor
performance of the agricultural sector (Alene & Hassan, 2003) which is evident in lower
standards of living of rural farming households. Rural areas of Ethiopia have the largest
concentration of absolute poverty, illiteracy and infant mortality (Diao, 2010) with the country
facing food insecurity and relying on food aid and to some extent food imports (Adenew,
2003:1; Diao, 2010).
In an agriculture dependent poor economy, it would be expected that growth in agricultural
production, especially in crop growth, would contribute more in reducing poverty than strong
macro-economic growth (Boccanfuso & Kabore, 2004). Thus, the key to growth in
agricultural production in Ethiopia lies in increasing productivity and efficiency of smallholder
farmers (Owour, 2000). Substantial policy emphasis is given to the agricultural sector in
Ethiopia because of the importance of agriculture in poverty alleviation, improving food
security and in promoting overall economic development (Spielman, Kelemwork & Alemu,
2011). The government of Ethiopia adopted the Agricultural Development Led
Industrialization (ADLI) strategy in 1994 as its economic development strategy. The main
goal of the strategy was to attain fast and broad-based development of the agricultural sector
and to promote the overall economy through the linkage effects of agriculture to other sectors
of the economy (Diao, 2010). Under ADLI, greater emphasis is given to increasing the
productivity of smallholder crop farmers through intensification of yield enhancing
technological inputs such as fertilisers and improved seeds along with better extension
services and farm management practices (Diao, 2010).
1.2 PROBLEM STATEMENT
Despite having abundant agricultural resource potential and following a consistent agricultural
policy to boost agricultural productivity, the expected productivity increment was not
achieved. The level of rural poverty is high and about 39% of the Ethiopian population still
2
live below the poverty line measured by the percentage of the population living on less than
the equivalent of US$1.25 per day (UNDP, 2013). Nearly 44% of the Ethiopian population are
undernourished (CSA, 2011b) and Ethiopia’s inability to feed its population remains a
dilemma that triggers broad economic and sociological debates.
The agricultural development strategy of Ethiopia gave due emphasis to increasing agricultural
productivity of smallholder farmers through the increased use of technological inputs in cereal
crop production (IFPRI, 2010b). In response to the agricultural development strategy, there is
an indication of substantial improvements in the adoption and use of chemical fertilisers,
improved seeds and other related inputs in Ethiopia, particularly in maize production; however
maize yield has not shown substantial improvement (Mulat, 1999; Arega & Zeller, 2005). One
of the reasons for low maize productivity could lie in the technical inefficiency existing in
smallholder production (Gebreselassie, 2006). From the policy perspective, insufficient
attention was given to obtaining information about the production efficiency of the
smallholder farmers. This is mainly attributed to farmers’ inability to select appropriate
technologies even though farmers are able to use the technologies efficiently when the
technologies are chosen for them (Kalirajan, 1991). However, information about farm level
technical efficiency of smallholder farmers is equally important in improving the productivity
of the smallholder farmers (Alene & Hassan, 2003).
The prevailing empirical studies of the Ethiopian smallholder’s technical efficiency indicate
the existence of technical inefficiencies. Among the studies, Fesessu (2008) examined the
extent of technical efficiency and factors affecting technical efficiency of coffee production in
Southern and South-western Ethiopia. Fesessu (2008) found an average technical efficiency of
71% where age, membership in farming associations, farming experience with other crops,
family size and extension services are the determinants that decrease technical inefficiency.
Altitude and coffee farming experience are variables that increase technical inefficiency
among the coffee producers (Fesessu, 2008). Similarly, Derege (2010) examined the level of
technical efficiency and the main determinants of coffee production in Jimma zone of
Ethiopia. Derege (2010) obtained an average technical efficiency of 72%. From the study
Derege (2010) found that education, distance from the market, family pressure and poor soil
fertility tends to increase technical inefficiency while proximity to a source of off-farm
income, cereal crop production, gender and good soil fertility decreases technical inefficiency.
3
Furthermore, Alene and Hassan (2003) examined the determinants of farm level technical
efficiency among the adopters of improved maize production technology in Western Ethiopia
and obtained an average technical efficiency of 76%. The study indicated that farm size,
education, access to credit, timely availability of modern inputs, extension, plot quality, tenure
and age are factors that decrease technical inefficiency while distance from the market
increases technical inefficiency (Alene & Hassan, 2003). Similarly, Geta et al. (2010)
analysed the productivity and efficiency of smallholder maize producers in Southern Ethiopia
and found an average technical efficiency of 40%. According to Geta et al. (2010), agro-
ecology, oxen holding, farm size and the use of improved seed are important factors that
decrease technical inefficiency among the farmers. Arega and Zeller (2005) estimated
technical efficiency of multiple crop production including maize, wheat and barley in Eastern
Ethiopia. Arega and Zeller (2005) obtained an average farm level technical efficiency of 79%.
The study indicated that extension services, education, credit and input supply systems are the
main determinants that decrease technical inefficiency among the farmers (Arega & Zeller,
2005). Bachewe (2009) explored the sources of inefficiency and growth in agricultural output
in subsistence agriculture in Ethiopia and obtained an average farm level technical efficiency
of 40%. According to Bachewe (2009), availability of sufficient productive labour and
increased educational levels of the farmers are factors that decrease technical inefficiency.
From the literature reviewed, it was found that information on farm level technical efficiency
on maize production in Ethiopia is limited. No empirical studies were found that investigated
maize farmers’ technical efficiency incorporating most of the maize producing regions in
Ethiopia. The few studies undertaken on maize production efficiency by Alene and Hassan
(2003) and Geta et al. (2010) are at zone levels which cannot give an indication of efficiency
status at national level. Other studies were undertaken on products such as coffee and multiple
grain crop production technical efficiency. Bachewe (2009) used a single index real value of
output for multiple subsistence crops including maize where a technical efficiency estimate for
maize production cannot be separately analysed. Since maize is a priority crop in terms of total
production and due to its contribution to national food security, a comprehensive analysis of
smallholder maize farmers’ technical efficiency is important.
Despite the important role maize has in the livelihood of Ethiopia, its productivity is low
compared to the potential level (Alene & Hassan, 2003). According to Schneider and
Anderson (2010) and IFPRI (2010b), there is a large maize yield difference between the
4
potential yield and the actual yield estimates in Ethiopia. IFPRI (2010b) noted that maize
production has a potential average yield of 4.7 ton/ha where the actual national maize yield
estimate is about 2.95 ton/ha during 2011/12 Meher season (CSA, 2012a). Given the persistent
food security issues facing Ethiopia, there is a need to improve maize productivity. One way
of improving farm productivity is through improving farmers’ technical efficiency. Technical
efficiency and productivity improvements are possible if farm level technical efficiency and its
determinants are identified.
1.3 RESEARCH OBJECTIVES
The main objective of the study is to identify factors affecting technical efficiency of
smallholder maize farmers for selected regions in Ethiopia. The main objective will be reached
through the completion of the following sub-objectives.
i. Estimating technical efficiency of smallholder maize farmers using Stochastic Frontier
Analysis (SFA) which estimates a production frontier against which the farmers’ actual
production is evaluated to quantify their technical efficiency.
ii. Identifying and analysing the socio-economic and farm management factors that affect
technical inefficiency of smallholder maize farmers in order to better understand the
constraints that prevent farmers from producing the maximum potential output.
1.4 CHAPTER OUTLINE
The study is organised into the remaining five chapters. Chapter two provides an overview of
the relevant literature on productivity and efficiency in production. The concepts of
productivity and efficiency in production, measurement of technical efficiency and variables
used in Stochastic Frontier Analysis in crop production are reviewed. Chapter three discusses
the study area profile and nature of data used in the study. In Chapter four, the methodological
framework applied in order to achieve the sub-objectives is discussed. Chapter five provides a
presentation and discussion of the results. Finally, Chapter six provides a summary, conclusion
and implications.
5
CHAPTER TWO: LITERATURE REVIEW
2.1 INTRODUCTION
When discussing the economic performance of producers, it is usual for them to be described
as being more or less “efficient,” or more or less “productive” (Fried, Lovell & Schmidt,
2008:7). Due to performance variation, not all producers are equally successful in utilising
their inputs to achieve potential yield, given the technology at their disposal (Kumbhakar &
Lovell, 2000:3). Through efficient utilisation of the available resources, farm households can
produce maximum possible output under the given technology and favourable operating
conditions. Identifying the extent of efficiency is thus important since it can lead to significant
resource savings which in turn can have an important effect on policy formulation and farm
management (Bravo-Ureta & Rieger, 1991). Efficiency and productivity are interrelated
concepts of production. Although these concepts are related, they do not have the same
meaning. Therefore, the concepts and relationships of productivity and efficiency should be
clearly defined. The next section will discuss the concepts of productivity and efficiency in
production, measurement of technical efficiency, review of the production frontier and
technical inefficiency models.
2.2 THE CONCEPT OF PRODUCTIVITY
Productivity is a natural measurement of performance that measures the level of the physical
output produced from the quantity of input(s) used. Productivity is measured by the ratio of
output produced to input(s) used (Latruffe, 2010).Estimating productivity is easy if a producer
uses a single input in production. When more than one input is employed, productivity
measurement is complex and a method of aggregating the inputs into a single index of inputs
is required (Fried et al., 2008: 522). Productivity measurement is a relative concept and can be
measured by comparing one year’s performance with the previous year’s performance or
relative to other producers (Coelli, Prasada, O’Donnell & Battese, 2005:2). Larger ratios of
productivity (output/input) measures are associated with better performance (Coelli et al.,
2005:2). Productivity variation is the difference between output growth and input growth.
Variation in productivity across producers or across time is attributed to differences in
6
production technology, scale of operation, differences in operating efficiency and differences
in the operating environment (Fried et al., 2008:8).
2.2.1 IMPORTANCE OF PRODUCTIVITY
Productivity improvement is one means of improving output in production. Agricultural
output can be increased either by increasing productivity of inputs or through expansion of
farm size in production (FAO & OECD, 2012). Through expansion of farm size, farmers can
be more productive by exploiting economies of scale, which arises from the differential access
to credit, adoption of more capital intensive technologies, better access to capital, willingness
to take risks and personal and political influence (Andrew, 1999). However, there is an
argument that smaller farms are more productive than larger farms. For example, Dyer (1996)
argued that small sized farms are more productive because they are poorer and are driven to
labour intensification through self-exploitation. Given these arguments, increasing output by
expanding farm size is not a sustainable way of poverty reduction because an increase in
production will take place within an environment characterised by a scarcity of arable land
resource (FAO, 2011). As a result, expansion of farmlands could be due to the use of marginal
lands that are not suitable for farming (FAO & OECD, 2012).
Increasing agricultural output by increasing productivity through more intensive use of land is
very important as it does not require utilisation of additional land for cultivation (FAO &
OECD, 2012).Increased agricultural productivity and the resulting increase in output
contribute to poverty reduction and to broader economic development (Mellor, 1999). The
primary effects of increased agricultural productivity include contributions to ensuring food
security and poverty reduction by increasing food availability. The increase in output
contributes to the decrease in food prices and increase of farm and off-farm employment
which consequently improves the rural economic environment (Adenew, 2003: 2; Clunies-
Ross, Forsyth & Huq, 2009:459). Improvement of the rural economic environment in turn
decreases urban poverty by slowing down rural-urban migration and urban unemployment
(Mellor, 1999; Thirtle et al., 2001; Clunies-Ross et al., 2009:458).
In addition to the poverty reduction role, agricultural productivity can also contribute to the
overall development of an economy (Kuznet, 1965: 239). An increase in agricultural
productivity contributes to the growth of other sectors of the economy (Adelman & Morris,
7
1988). According to Kuznet (1965: 239), agricultural productivity enhances economic
development through four contributions namely: product contribution; factor contribution;
market contribution and foreign exchange contribution. Product contribution means more
output will be available for the economy through increased productivity while factor
contribution refers to the release of excess factors such as labour and capital from agriculture
to other emerging sectors. Market contribution refers to the increasing demand arising from
the agricultural sector for products of the other sectors and the increased supply of food and
raw materials by agriculture to the other sectors which develops a market. Foreign exchange
contribution refers to the role of agriculture in international trade (Kuznet, 1965: 239).
Increased agricultural productivity increases export earnings and determines the
competitiveness of countries in global trade, mainly for countries whose export is dominated
by agricultural commodities (Cluines-Ross et al., 2009: 457). This is specifically true in
Ethiopia where about 90% of the country’s export is dependent on agricultural commodities
(MoARD, 2010). Given the importance of agricultural productivity in an economy, one of the
methods of increasing productivity is through increasing productive efficiency which is
discussed in the following section.
2.3 THE CONCEPT OF EFFICIENCY IN PRODUCTION
Production efficiency is the degree of success producers achieve by allocating the inputs at
their disposal and the outputs they produce in an effort to meet certain objectives (Kumbhakar
& Lovell, 2000:15). Production efficiency is an important factor for productivity growth
especially in developing countries where resources are scarce, as production can be increased
through improving efficiency in production without the use of additional inputs (Alene &
Hassan, 2003). Efficiency in production can be seen in terms of technical efficiency, allocative
efficiency or a combination of technical and allocative efficiency which is called economic
efficiency (Fried et al., 2008: 20) or overall efficiency (Farrell, 1957). Given the concept of
efficiency, the following section is a discussion of technical efficiency.
2.3.1 TECHNICAL EFFICIENCY
Technical efficiency (TE) is the ratio of the actual production to an optimal level of production
(Greene, 1993). Technical efficiency is defined as the ability to minimise input usage in the
production of a given output vector or the ability to obtain maximum output from the given
8
input vector (Kumbhakar & Lovell, 2000:17). A technically efficient producer could produce
the same level of output using lessor inputs or could produce more output using the same level
of inputs. However, not all producers are technically efficient (Fried et al., 2008:20). The
concept of technical efficiency can be better explained graphically using a simple example
involving a producer using two factors of production (X1& X2) to produce a single output (Y).
Figure 2.1 provides a graphical representation of productive efficiency under a two input and
one output technology set represented by an isoquant SS’. The knowledge of a unit isoquant of
fully technically efficient producers represented by SS’ in Figure 2.1 permits the measurement
of technical efficiency.
Figure 2.1: Graphical representation of technical and allocative efficiency using two
inputs (X1& X2) and one output (Yi)
Source: Farrell (1957)
From Figure 2.1, isoquant SS’ represents production of output level Y with different levels of
input X1 and X2. Suppose a given producer uses an input combination of the two factors
defined by point P to produce a unit of output. Under this framework, every input combination
along the isoquant SS’ is considered technically efficient while any input combination above
9
and to the right of the isoquant SS’ such as point P defines a technically inefficient input
combination. This is because, at point P, the input package used is greater than the minimum
input necessary to produce a unit of output (Murillo-Zamoralo, 2004). However, point Q
represents a technically efficient input combination because it lies on the efficiency isoquant
(SS’). Thus the technical efficiency (TE) of the producer at P is defined as:
= ( . )
The value of TE is bounded between zero and one. TE takes a value of one for a perfectly
efficient producer and moves towards zero for inefficient producers (Farrell, 1957). Given the
theoretical explanation of technical efficiency, estimation and analysis of technical efficiency
in crop production assists to determine the scope of raising productivity of inefficient
producers. Knowledge of the level of technical efficiency contributes significantly to
realisation of national policy goals such as achieving food security, poverty alleviations and
growth and development through improving performance of inefficient producers (Uaiene,
2008; Mupanda, 2009). The cost association with the inputs and outputs is also important in
production and is considered when calculating Allocative Efficiency.
2.3.2 ALLOCATIVE EFFICIENCY
Allocative efficiency (AE) is the ability of a producer to use inputs in optimal proportions
given their respective prices (Coelli et al., 2005: 5). Allocative efficiency is achieved when a
producer operates at the least-cost combination of inputs to produce a specified level of output
(Kumbhakar & Lovell, 2000: 15). If information of input prices are known and a particular
behavioural objective such as cost minimisation is assumed, allocative efficiency of a producer
can be derived. Suppose in Figure 2.1 the producer uses the two inputs (X1 and X2) given their
respective prices P1 and P2 to produce a certain amount of output (Y). It is assumed that a line
segment AA’ is an isocost line with a slope equal to the ratio of the prices of the two inputs.
With these assumptions, the only points that minimise input costs are allocatively efficient.
The optimal input selection for the cost minimising producer is at the point where the isocost
line AA' is tangent to isoquant SS’ which is at point Q’. Thus, Q’ is the point of optimal input
combination where the producer is both technically and allocatively efficient. Point Q is where
the producer is technically efficient but allocatively inefficient, so it is not the optimal input
10
combination point. Further, if the producer is to change the proportions of input combinations
until they are the same as those represented at Q’, the cost can be reduced by a factor of
OR/OQ as long as factor prices remain the same. Therefore, allocative efficiency (AE) that
characterises the producer at point P is given by the ratio:
= ( . )
The distance RQ from Figure 2.1 represents the reduction in production costs that would occur
if production were to occur at the allocatively and technically efficient point Q’ instead of at a
technically efficient, but allocatively inefficient point Q (Coelli et al., 2005:53). The measure
of AE takes a value of one for an allocative efficient producer and becomes closer to zero as
the producer become less allocative efficient.
When producers are both technically and allocatively efficient, they are economically
efficient. Given both technical efficiency and allocative efficiency in Figure 2.1, the producer
at point Q’ is economically efficient. Economic efficiency can also be computed as the product
of technical efficiency and allocative efficiency (Farrell, 1957). Since economic efficiency is a
combination of technical and allocative efficiency, economic inefficiencies will arise from
technical and/or allocative inefficiencies (Bravo-Ureta & Pinheiro, 1997). Given the
distinction of various concepts of productivity and efficiency in production, the following
section discusses the measurement techniques of technical efficiency.
2.4 MEASURING TECHNICAL EFFICIENCY
Technical efficiency estimation involves a comparison of actual performance with optimal
performance located on the relevant frontier. Since the true frontier is unknown, an empirical
approximation is needed (Fried et al., 2008:33). The measurement of a farm specific technical
efficiency is based upon deviations of the farm’s actual performance from the efficiency
frontier (Kumbhakar & Lovell, 2000:3). If a producer’s actual production point lies on the
efficiency frontier (assuming that a production frontier is the same as a production possibility
curve), the farm is technically efficient and if it lies below the frontier, it implies the presence
of technical inefficiency (Pascoe & Mardle, 2003).
11
There are two main approaches of estimating technical efficiency among producers, namely a
parametric or a non-parametric approach. The difference between the two approaches is that
the former approach specifies a particular functional form based on econometric techniques
while the latter is based on mathematical programming (Sarafidis, 2002). In empirical work,
the two most popular techniques of efficiency measurement are Data Envelopment Analyses
(DEA) and Stochastic Frontier Analysis (SFA).The two techniques are different in their
treatment of random noise and for flexibility in the structure of production technology
(Porcelli, 2009). In the following sub-sections, the distinctions between the two efficiency
measurement techniques are discussed.
2.4.1 DATA ENVELOPMENT ANALYSIS
Data Envelopment Analysis (DEA) is a non-parametric technique of efficiency measurement
that is based on mathematical programming (Sarafidis, 2002). Development of DEA was
influenced by the early works of Debreu (1951), Koopmans (1951) and Farrell (1957). DEA
was first introduced by Charnes, Cooper and Rhodes(1978) and is widely employed in
management sciences mainly in operational researches (Kumbhakar & Lovell, 2000:7).
Technical efficiency estimation using DEA involves the use of a linear programming method
to construct a non-parametric frontier over the sample data. Efficiency is estimated using the
distances of each observation relative to the frontier (Coelli et al., 2005:162). In DEA, relative
technical efficiency of each decision-making unit (DMU) is measured by using a ratio of
weighted sum of output to a weighted sum of input. The weights for both outputs and inputs
are selected in a manner that calculates efficiency measures for each DMU subject to the
constraint that no other DMU can have relative efficiency scores greater than unity (Charnes et
al., 1994). DEA establishes the basis to identify the level of potential improvement for the
inefficient producers relative to the efficient producers (Solwati, 2001). Given the background
of DEA efficiency measurement technique, DEA has certain strengths and weaknesses that
will influence the decision to utilise the method.
Efficiency estimations using DEA does not require specification of a functional form and there
is no imposition of statistical assumptions about the distribution of error terms. These
properties free the model from specification bias. In addition, the model can accommodate
efficiency estimation involving multiple outputs more easily and provides an indication of the
scale of operation for individual DMUs in the sample (Sarafidis, 2002). However, DEA has
12
weaknesses that limit its suitability and appropriateness. The principal limitation of DEA is
that the model does not make provision for statistical noise and all deviations from the frontier
are considered as inefficiency. As a result, efficiency estimates obtained by using DEA can be
biased and unreliable in studies where the data has the influence of statistical noise (Pascoe &
Mardle, 2003). In addition, assessment of goodness of fit of a DEA model is difficult as there
is no proper definition of goodness of fit that enables model comparisons and the standard
criteria cannot be used for assessment of the goodness of fit of DEA model (Sarafidis, 2002).
Furthermore, efficiency estimation using DEA is sensitive to outlying observations which can
provide misleading information and do not allow hypothesis testing (Sarafidis, 2002; Fried et
al., 2008). There is an alternative technique of efficiency measurement called SFA.
2.4.2 STOCHASTIC FRONTIER ANALYSIS
Stochastic Frontier Analysis (SFA) is a parametric technique of efficiency measurement that is
based on econometric estimation (Sarafidis, 2002). The literature that influenced the
development of SFA was the theoretical literature on productive efficiency which began in the
1950s with the work of Koopmans (1951), Debreu (1951), Shephard (1953) and Farrell
(1957). Aigner, Lovell, & Schmidt (1977) and Meeusen and van den Broeck (1977)
introduced the SFA model simultaneously. The SFA model allows for technical inefficiency
and a symmetric random noise error terms (Kumbhakar & Lovell, 2000: 8). The SFA model
handles the effects of inefficiency and random noise on output separately through the
introduction of a composed error term. The composed error term consists of a symmetric
disturbance term (Vi) and a non-negative inefficiency term (Ui) (Kumbhakar & Lovell, 2000:
8). The primary motivation of introducing the symmetric disturbance term into the efficiency
estimation is due to the fact that deviations of actual observations from the frontier
observations might not entirely be under the control of the producers (Fried et al., 2008: 114).
Thus, specification of the SFA model permits output to be specified as a function of
controllable factors of production, random noise and technical inefficiency (Kumbhakar &
Lovell, 2000).
SFA uses Maximum Likelihood Estimation (MLE) to estimate a frontier function in a given
sample, which is the method first used by Greene (1980) and Stevenson (1980). MLE can
estimate the production function parameters (β) and the technical inefficiency model
parameters (δ), simultaneously (Porcelli, 2009). MLE of an unknown parameter is defined to
13
be the value of the parameter that maximises the probability of randomly drawing a particular
sample of observations (Coelli et al., 2005:217). By employing specified distributional
assumptions, it is possible to derive the likelihood function, which can be maximised with
respect to all SFA parameters to be estimated (Fried et al., 2008:36).
In the procedure of efficiency estimation, SFA considers separate assumptions regarding the
distributions of the random noise and inefficiency variables that potentially lead to more
reliable efficiency estimations (Kumbhakar & Lovell, 2000). The symmetric disturbance term
is assumed to be identically, independently and normally distributed with zero mean and
constant variance or Vi~ iidN (0, σ2V) throughout. The inefficiency variable (Ui), however, has
developed into different distributional assumptions. In the original development of SFA, half-
normal and exponential distribution were considered (Aigner et al., 1977). These assumptions
were developed into more flexible general distributions such as gamma-distribution (Greene,
1980), truncated normal distribution (Stevenson, 1980) and the four-parameter Pearson family
distributions (Lee, 1983). The half-normal and exponential distributions assume mode at zero,
implying the highest proportion of the producers examined are perfectly efficient. Truncated
normal and gamma distributions however, allow wider ranges of distributional shapes
including non-zero means. The truncated normal distribution implies that the one sided error
term (Ui) is obtained by truncating at zero with the possibility of a non-zero mean that also
generalises the half-normal distribution (Fried et al., 2008: 130). Some empirical analyses
suggest that the use of the truncated normal model has less difficulty in estimation, unlike
gamma which has complex procedures to follow (Ritter & Simar, 1997; Fried et al., 2008).
Like all other models, SFA has strengths and weaknesses. The main strength of the SFA is that
the effects of random noise on output can be separated from the effects of technical
inefficiency (Fried et al., 2008). This is an important property of SFA, especially when the
data undertaken has the influence of random effects (Sarafidis, 2002). The SFA model permits
hypothesis testing as to the functional form of the frontier and the significance of individual
explanatory variables (Sarafidis, 2002). The main weaknesses of SFA are the requirements of
specification of the functional forms and formulation of distributional assumptions about the
error terms (Henderson & Kingwel, 2002).
Given the strengths and weaknesses of the two techniques, the SFA is preferred over DEA in
certain circumstances. When random influences and statistical noises are perceived to
14
influence the data and when the omitted variables may influence the final results, SFA is
preferred. Moreover, when hypothesis testing is important and measurement of goodness of fit
of the estimated model is required, SFA model is more appropriate (Sarafidis, 2002).
2.5 THE PRODUCTION FRONTIER AND THE TECHNICAL
INEFFICIENCY MODEL
The estimation of technical efficiency and examining the determinants of technical
inefficiency using the SFA requires that a production frontier and a technical inefficiency
model are estimated. The production frontier is used to estimate the level of technical
efficiency whereas the inefficiency model is used to identify the potential determinants of
technical inefficiency. The inputs that define a production frontier and factors affecting
technical inefficiency in crop production are discussed further to determine the factors that can
determine technical efficiency of Ethiopian maize farmers.
2.5.1 REVIEW OF INPUTS DEFINING PRODUCTION FRONTIER
Estimation of technical efficiency using SFA involves estimating the unknown production
frontier (Coelli et al., 2005).Farm inputs such as seed, fertiliser and labour are the primary
inputs used in smallholder crop production based on the empirical literature (Alene & Hassan,
2003; Arega& Zeller, 2005; Gebreselassie, 2006; Bachewe, 2009). Application of appropriate
seed and fertiliser can increase production considerably (Gebreselassie, 2006). In crop
production, farmers use improved or traditional seed as production inputs (Morris et al., 1999).
Despite the productivity differences between improved and traditional seed, seed is a
conventional input in crop production. From the empirical studies, Geta et al. (2010) and
Idiong (2007) found that seed quantity has a positive influence on cereal production. Another
input that is applied in crop production is fertiliser.
Fertiliser is an important input in crop production. Low soil fertility is one of the biophysical
constraints affecting smallholder production (Sanchez & Roland., 1997; Ayalew & Dejene,
2011). Through application of fertiliser, soil fertility and land productivity can be improved.
Application of either organic or chemical fertiliser or integrated use of both fertilisers is
expected to increase production (IFPRI, 2010a; Ayalew & Dejene, 2011). Chemical fertiliser
is a yield enhancing input in crop production and its application increases productivity
15
considerably, especially if used with improved seeds and irrigation (Gebreselassie,
2006).Chemical fertilisers enhance the uptake of important nutrients such as nitrogen and
phosphorus and their concentration in plant tissues when applied (Abdulahi et al.,
2006).Smallholder farmers in Ethiopia are constrained from using chemical fertilisers and
from applying the recommended amount due to the high cost of the chemical fertilisers in the
country (Ayalew & Dejene, 2011). However, the application of fertiliser requires maintaining
the levels recommended by agricultural scientists together with the timing and method of
application for better productivity (Ayalew & Dejene, 2011). Among the empirical studies,
Alene and Hassan (2003) and Geta et al. (2010) found an increasing effect of fertiliser on
maize production.
Similar to fertiliser, labour is an important input required in crop production. Smallholder
farming activities such as land preparation, planting, fertiliser application, weeding and
harvesting require adequate labour throughout the production process. Availability of adequate
labour enhances production return by enabling the households to undertake the farming
activities properly (Geta et al., 2010). Sources of agricultural labour include family labour and
hired labour. Alene and Hassan (2003), Fesessu (2008) and Bachewe (2009) found that crop
production responds to labour use positively. Given the review of variables defining the
production function, the following section is a review of the factors affecting technical
inefficiency in crop production.
2.5.2 FACTORS AFFECTING TECHNICAL INEFFICIENCY
In the technical inefficiency model, the dependent variable is the index of technical
inefficiency (Ui) and the independent variables are variables used to explain the technical
inefficiency of the producers (Kumbhakar & Lovell, 2000:261). Variables that increase
technical inefficiency have positive parameter estimates and vice versa (Bachewe, 2009).
Identifying and examining the determinants of technical inefficiencies in production can reveal
options for technical efficiency improvement (Bachewe, 2009). There are many socio-
economic and farm management factors that affect technical inefficiency of smallholder crop
farmers. Based on literature, the common factors are age, gender, education, household size,
farm size, land tenure, ownership of oxen, access to extension services, irrigation, access to
credit, off-farm income, seed type, organic fertiliser and soil protection (Alene & Hassan,
2003; Bachewe, 2009; Derege, 2010; Geta et al., 2010). These variables will be discussed by
16
explaining their effects on inefficiencies in farming as cited by previous research and the
possible influence on the current research.
2.5.2.1 Age, Gender and Education of Household Heads
Technical efficiency variations across smallholder farmers can be as a result of differences in
age, gender or education of the household heads. Age of a household head can have a
decreasing effect on technical inefficiency, meaning that when the age of a household head
increases, technical inefficiency in production decreases. Among the empirical studies, Ayele
et al. (2006), Fesessu (2008) and Maseatile (2011) found that age has a decreasing effect on
technical inefficiency. Possible reasons for such a relationship can be due to the fact that with
increased age, farmers perform better through having better resources at their disposal and can
be better aware of mechanisms for risk coping from life experience. In contrast, Makombe et
al. (2011) found an increasing effect of age on smallholder farmers’ technical inefficiency.
The increasing effect of age on farmers’ technical inefficiency could be due to the fact that
older household heads will become more conservative towards acceptance of new ideas,
technologies and practices which can result in an increase in technical inefficiency (Gbegeh &
Akubuilo, 2013).Other possible reasons could be that when the age of household heads
increases, the households may not be able to accomplish the usual farm activity due to old age.
In other words, younger farm household heads can have a better education, capacity to work,
ability to gather information, new ideas and practices that can decrease the farms’ technical
inefficiency (Bravo-Uteta & Pinheiro, 1997; Gbegeh & Akubuilo, 2013). From literature, the
potential effect of age on technical inefficiency is mixed, meaning that the age of a household
head can have either an increasing or a decreasing effect on technical inefficiency. Similar to
age, gender of household heads can also influence the level of technical inefficiency.
Gender of household heads provides indications of technical inefficiency variations among the
farm households (Bachewe, 2009). According to the empirical findings of Solis et al. (2008)
and Bachewe (2009), male-headed households are less technically inefficient than female-
headed households. A possible reason for gender based technical inefficiency variation could
be that male-headed households have better access to land, credit, technological inputs and
other supportive services than their female counterparts (AWM, 2009; OXFAM, 2012).
Another possible reason could be that female household heads undertake farming activities in
addition to their normal homemaker role which can increase their farm technical inefficiency.
17
Derege (2010) found that female-headed coffee farming households are less technically
inefficient than male-headed households. According to Derege (2010), a possible reason could
be that the female household heads made an increased effort towards follow up and
supervision of the farm work for better production than the male household heads. Literature
indicates that the gender of household heads can have either an increasing or a decreasing
effect on farm technical inefficiency. Another variable related to farm household heads is
education.
Education increases farmers’ ability to obtain, process and use information relevant to
agricultural practices that can decrease farm technical inefficiency (Bachewe, 2009).
Accordingly, education may enhance farm productivity by increasing the ability of the farmers
to adjust to risk and adopt new innovations (Weir, 1999). According to Admassie and Asfaw
(1997), Ayele et al. (2006) and Derege (2010), more years of schooling decreases technical
inefficiency in production. The possible reason for this relationship could be that farmers with
better educational levels tend to be more efficient since they can respond more readily when
using new technologies and can therefore produce closer to a technology frontier (Derege,
2010). In contrast, Mkhabela (2005) and Belloumi and Matoussi (2006) found an increasing
effect of education on farm technical inefficiency. The findings indicate the possibility that the
increased level of education of a farm household head increases the level of technical
inefficiency. A possible explanation for the result is that with increased years of schooling,
farmers can have alternative job opportunities to choose from so that their devotion to the farm
work will decrease. Based on the above findings, education of a household head can have
either an increasing or a decreasing effect on technical inefficiency. Next, the effects of
household size, farm size, land tenure and oxen on technical inefficiency of smallholder
farmers will be reviewed.
2.5.2.2 Household Size, Farm Size, Land Tenure and Oxen
Household size refers to the number of household members living in each farm household
(CSA, 2012a). A farm household can be either a person living alone or a group of people
(related or unrelated in either kinship or marriage) who live together in the sense that they
have common housing arrangements or they are supported by a common budget (CSA,
2012a). The size of a farm household can affect farm technical inefficiency either positively or
negatively. Fesessu (2008) found a decreasing effect of household size on smallholder
18
farmers’ technical inefficiency. This confirms the importance of larger household sizes in
decreasing technical inefficiency. On the other hand, Derege (2010), Maseatile (2011),
Baruwa and Oke (2012) found an increasing effect of household size on farm technical
inefficiency. The increasing effect of household size on technical inefficiency has an
implication that although a large household size enhances the availability of family labour, it
may not guarantee an increased efficiency and it can rather lead to inefficiency. Since
smallholder farmers cultivate smaller farmlands, increased household size can result in
underutilisation of household labour and increased inefficiency. In other words, when larger
households derive their livelihood from farm activities only, much of the household labour
could be used on the smaller farms unnecessarily where the task could be accomplished by
using less labour (Maseatile, 2011; Baruwa & Oke, 2012).
Farm size is also an important variable commonly included in empirical efficiency analysis of
smallholder production (Uaiene, 2008). Smallness or largeness of a farm can affect farm level
technical inefficiency. There are different economic arguments about the size of farmlands and
associated productivity (Masterson, 2007).Bravo-Uteta and Pinheiro (1997), Huang and
Kalirajan (1997), Andrew (1999) and Khaile (2012) found a decreasing effect of farm size on
technical inefficiency. This supports the notion that larger farms have an efficiency advantage
over smaller farms. These findings indicate that expansion of farm size contributes to better
productivity through encouraging adoption of more capital-intensive technologies, better
access to capital due to economies in transaction costs, willingness to take risks and personal
and political influence (Andrew, 1999). Related to this argument, the average size of
farmlands of the Central highlands of Ethiopia has fallen from 0.5 hectare in 1960’s to about
0.2 hectare by 2008 (Diao, 2010). The smallness of the farms is usually seen as a constraint to
productivity of smallholder producers (Diao, 2010). The decline in farm size in the area is
attributed mainly to the fact that farmlands are divided among family descendants and
therefore farm size decreases overtime resulting in smaller and fragmented farms (Diao,
2010).Nonetheless, there are researchers who argue that smaller farms are more productive
than the larger farms. For example in Egyptian crop farming, Dyer (1996) found an inverse
relationship between farm size and productivity. Dyer (1996) argued that small sized peasant
farms are more productive than large farms because the farmers are poorer and are driven to
labour intensification through self-exploitation. Similarly, Ellis (1993) noted that small farms
produce more output per hectare mainly by using family labour which is easy to manage
compared to large farms where hired labour needs more supervision and management costs.
19
Supporting this argument, Parikh et al. (1995) and Masterson (2007) found an increasing
effect of farm size on technical inefficiency confirming that smaller farms are more technically
efficient than larger farms. However, Bardhan (1973) suggested that farm size and
productivity relationships cannot be concluded without considering other important factors of
production besides land that affect productivity. Given the different views about the effects of
farm size on productivity, the empirical findings suggest that farm size has a mixed effect on
technical inefficiency, meaning that some studies found a decreasing effect while others found
an increasing effect of farm size on technical inefficiency.
Land tenure is another important variable which refers to the type of ownership of farmlands,
whether privately owned, rented or share cropping (Uaiene, 2008). One of the key issues
related to land tenure is the degree to which the tenure arrangement encourages sustainable
farm practices. It is generally believed that privately owned farms provide necessary
incentives for farmers to better manage the lands and to make necessary investments that lead
to an improvement of productivity of the lands (Nega et al., 2003). From the empirical
findings, Gavian and Ehui (1999), Alene and Hassan (2003) and Binam et al. (2003) found a
decreasing effect of privately owned farms on farm technical inefficiency. This implies that
privately owned farms are more technically efficient. In contrast, Corppenstedt and Abbi
(1996) and Rahman and Umar (2009) found that privately owned farms are more technically
inefficient than rented or share-cropping farms. Although tenure has a mixture of decreasing
and increasing effects, most of the literature support the fact that private tenure has a
decreasing effect on farm technical inefficiency. Similar to tenure, ownership of an adequate
number of oxen can also cause variations in technical inefficiency among farm households.
Ownership of oxen is a variable of interest in technical efficiency analysis of smallholder
farmers in Ethiopia. This is because oxen are the main source of draft power used for farming
activities in smallholder crop production in the country (Geta et al., 2010). Ownership of
adequate oxen augments labour input and enhances productivity by reducing the time needed
to accomplish farm operations such as land preparation and sowing. Farmers need at least one
pair of oxen to be able to prepare their land well and timely. From the empirical studies,
Gebreegziabher et al. (2004) and Geta et al. (2010) found that ownership of oxen is an
important variable that has a decreasing effect on smallholder farmers’ technical inefficiency.
Therefore, ownership of an adequate number of oxen affects technical inefficiency negatively.
20
Information availability and funding is also important in the efficiency of farming and these
factors are discussed in their importance in determining technical inefficiency.
2.5.2.3 Extension, Irrigation, Credit and Off-farm Income
Extension service refers to the advice and training provided by farm extension workers to the
farmers about farming operations (CSA, 2012a). Farmers who receive extension services,
advice and assistance can have better information which can decrease their farm’s technical
inefficiency (Fesessu, 2008). From the empirical studies, Alene and Hassan (2003), Obwona
(2006), Fesessu (2008) and Maseatile (2011) found a decreasing effect of extension on farm
technical inefficiency. The findings indicate that access to extension services increase
productivity. Binam et al. (2004) found an increasing effect of extension services on farm
technical inefficiency. According to Binam et al. (2004), the possible explanation for such a
relationship was attributed to the weak performance of information delivery systems inherent
in public operated extension services, present in most developing countries. Even though,
there is a condition where extension service increases technical inefficiency, most of the
studies supported the decreasing effect of extension on farm technical inefficiency. Possible
information from extension services would include the use of resources in production, such as
water application. Irrigation is an important source of crop water that is seldom used in
Ethiopian smallholder production to improve crop yields.
Irrigation of crops also has the ability to increase technical efficiency of farmers through
improving crop yields. Irrigation is a means of providing sufficient water for plant growth to
prevent water stress that can possibly reduce productivity of the inputs used. Sufficient water
is a requirement to attain the potential level of production (Haise & Hagan, 1967). Irrigated
agriculture has an important role in reaching the broader development goals of achieving food
security, poverty alleviation and improved quality of life (Haise & Hagan, 1967). Irrigation is
important in Ethiopian agriculture as the country is vulnerable to weather and climate changes
because of high dependence on rainfall (Hordofa et al., 2008). Through the use of irrigation,
farmers can reduce their production risks associated with inadequate rainfall (Hordofa et al.,
2008). Supporting the importance of irrigation, Makombe et al. (2011) found that farmers that
have access to irrigation are more technically efficient than farmers without access to
irrigation. Therefore, irrigation has a decreasing effect on farm technical inefficiency. The use
21
of irrigation and other factors in production is limited without funding. Access to funding,
such as credit, is therefore important for increased technical efficiency.
Credit can affect the technical inefficiency of farmers by affecting their decisions regarding
agricultural financing (Uaiene, 2008). In Ethiopia, the formal financial sectors are not well
developed to provide credit services to the poor rural farm households (Yehuala, 2008).Lack
of access to formal credit is frequently described as a key problem for smallholder farmers of
Ethiopia (Croppenstedt et al., 2003).Constraints of rural credit affect productivity and
efficiency of resource lacking farmers by limiting them from financing for both short-term and
long-term farm investments (IFPRI, 2010a). Among the empirical studies, Obwona (2000),
Binam et al. (2004), Gebreegziabher et al. (2004) and Maseatile (2011) found a decreasing
effect of credit on technical inefficiency. The findings confirm the importance of credit in
decreasing technical inefficiency. Another source of smallholder farmers’ agricultural
financing is off-farm income, which can affect farmers’ technical inefficiency.
In rural areas of developing countries, off-farm income participation is an alternative source of
income for farm households to support the economic well-being of the household and to
finance the farming operations (Beyene, 2008).Off-farm income provides farmers with
potential capital for purchasing productivity enhancing inputs such as improved seed and
fertiliser in addition to supporting the consumption needs of the farmers (Gebreegziabher et
al., 2004). As a result, off-farm income can decrease farm technical inefficiency. In line with
this idea, Gebreegziabher et al. (2004), Haji (2006) and Maseatile (2011) found a decreasing
effect of off-farm income on farm technical inefficiency. The pursuit of off-farm income
activities by farmers can also increase farm level technical inefficiency in such a way that the
increased participation of the farmers in off-farm income activities reduces the amount of
household labour available for the farming activities (McNally, 2002; Goodwin & Mishra,
2004; Geta et al., 2010). The finding supports the idea that increased off-farm income
opportunities reduce farm resources and farmers’ efforts that could otherwise be used for
farming activities. From the empirical literature, it is observed that off-farm income can have
either a decreasing or an increasing effect on technical inefficiency. This depends on whether
the off-farm income obtained contributes to improving the performance of the farm sector or if
off-farm income switches the resources from farm to the off-farm sector. Funding
opportunities such as credit and off-farm income is important to finance inputs such as seeds
and fertilisers and to participate in soil protection practises.
22
2.5.2.4 Seed Type, Organic Fertiliser and Soil Protection
Improved seed is expected to provide higher returns than traditional seeds in production. This
is due to the fact that the characteristic of improved seed is systematically altered in ways that
bring higher productivity (Morris et al., 1999). From literature, Geta et al. (2010) and
Maseatile (2011) found a decreasing effect of improved seeds on technical inefficiency. The
findings confirmed the relative importance of improved seeds in decreasing technical
inefficiency hence increasing productivity. In addition to the use of improved seed, application
of organic fertiliser can influence technical inefficiency of smallholder farmers.
Organic fertiliser application improves the fertility of soils thereby decreasing technical
inefficiency in crop production (IFPRI, 2010a; Ayalew & Dejene, 2011). Smallholder farmers
commonly apply organic fertilisers such as manure and compost in crop production in order to
improve soil fertility (Gilleret al., 2006). However, organic fertiliser has a lower nutrient
concentration compared to chemical fertilisers and release nutrients slowly, therefore larger
quantities of organic fertilisers need to be applied (Emiru, 2004). According to Gruhn et al.
(2000), application of both organic and chemical fertilisers together is the best alternative for a
balanced and efficient plant growth and for greater productivity (Gruhn et al., 2000). As a
result, the application of organic fertiliser has a decreasing effect on technical inefficiency.
The last variable reviewed is soil protection practises of the smallholder farmers. Smallholder
farmers protect their farmlands from erosion through practises such as terracing, planting trees
and contour ploughing which are expected to contribute to soil conservation and improved
farm productivity (IFPRI, 2010a). Geta et al. (2010) found that integrated soil fertility
management practises decrease technical inefficiency significantly. Although the organic
fertiliser application and soil protection practises are important variables in productivity
improvement, they were not investigated in previous technical efficiency studies.
2.6 CONCLUSION AND IMPLICATIONS FOR THE RESEARCH
Technical efficiency indicates how efficient producers are in the way they use their limited
resources in production. Therefore, information about the level of technical efficiency is
important to improve crop production. Technical efficiency analysis in Ethiopian smallholder
maize production is important to improve the food insecurity problems of the country.
23
Productivity of maize can be improved by having adequate information about farm technical
efficiency and the associated constraints. From the literature, the distinction of DEA and SFA
in measuring technical efficiency is identified. The principal difference between SFA and
DEA efficiency estimation is that SFA, unlike DEA, can distinguish between the effects of
random shocks and inefficiency separately. Estimation of technical efficiency and
investigation of the determinants of technical inefficiency by using SFA requires estimation of
a production frontier and technical inefficiency models. The former is used to estimate the
level of technical inefficiency while the latter is used to examine the determinants of technical
inefficiency. The inputs defining the production frontier and factors affecting technical
inefficiency models were reviewed thoroughly.
Based on literature, the inputs defining the production function are seed, fertiliser and labour.
The common factors affecting technical inefficiency in crop production are age, gender,
education, household size, farm size, tenure, oxen, extension, irrigation, credit, off-farm
income, seed type, organic fertiliser and soil protection. Previous studies indicate that private
tenure, ownership of adequate oxen, access to extension, irrigation, credit, off-farm income,
improved seed, organic fertiliser and soil protection would decrease technical inefficiency.
The variables age, gender, education, household size, farm size and off-farm income would
either decrease or increase technical inefficiency. The review of the variables provides a
framework in determining the variables to be incorporated in the study and provides
information about the possible effects of these variables on technical inefficiency.
24
CHAPTER THREE: THE STUDY AREA PROFILE AND
NATURE OF THE DATA EMPLOYED
The objective of this Chapter is to provide an overview of the study area in terms of location,
topography and agriculture. The chapter discusses the data used and characteristics of the farm
households studied.
3.1 LOCATION AND TOPOGRAPHY
Ethiopia, officially known as the Federal Democratic Republic of Ethiopia, is located in the
North-Eastern part of the horn of Africa with a total surface area of 1.12 million square
kilometres (Mengistu, 2006). Ethiopia is bordered by Eritrea to the North, Djibouti and
Somalia to the East, Kenya to the South and the Republic of Sudan and the Republic of
Southern Sudan to the West (Mengistu, 2006). At present, Ethiopia is structured into a
federation of nine ethnic based administrative regions and two centrally chartered city
administrations (Sori, 2009). Figure 3.1 is a graphical representation of the administrative
regions and city administrations of Ethiopia.
As indicated in Figure 3.1, the administrative regions are: Tigray, Afar, Amhara, Oromia,
Somali, Benishangul Gumuz, Southern Nations Nationalities and Peoples (SNNP), Gambela
and Harari. The two chartered city administrations are Addis Ababa and Dire Dawa. Among
the regions, the study area includes Tigray, Amhara, Oromia, Somali, Benishangul Gumuz,
SNNP, Harari and Dire Dawa regions. Dire Dawa city administration is included because there
are rural agricultural areas that were surveyed. Although the study was planned to cover all the
regions in the country, Afar and Gambela regions were not included because the secondary
data used for the study did not provide adequate information relevant to the study for the two
regions. Therefore, the selection of the study area is based on the data obtained for each region
from the secondary data source.
25
Figure 3.1: Graphical representation of the nine administrative regions and the two
city administrations of Ethiopia
Source: (Oromia Region Map, 2013)
Ethiopia has an extremely varied topography which consists of high mountains, deep gorges
with rivers, rolling plains and dissected plateaus divided by the great East African Rift Valley
(FAO, 1984). The altitude ranges from 110m below sea level at the Danakil Depression to the
highest peak at mount Ras Degen which is 4 600m above sea level (FAO, 2005). The diversity
of the country’s topography determines the wide variations in agro-ecology, climates, soils,
vegetation, and settlement patterns (Camberlin & Philippon, 2001; FAO, 2005; CSA, 2009b).
In the following section, the agro-ecological classifications, the related climatic conditions and
soil types of the study area are discussed.
3.2 AGRO-ECOLOGY
Agro-ecological zones (AEZs) are areas where predominant physical conditions guide
relatively homogenous agricultural land use options (CSA, 2006b). Elevation is the basis for
26
the traditional agro-ecological division of Ethiopia. The six traditional agro-ecological zones
of Ethiopia are: Bereha, Kolla, Weina-Dega, Dega, Wurch and Kur. Figure 3.2 indicates the
traditional AEZs of Ethiopia.
Figure 3.2: Traditional Agro-ecological Zones of Ethiopia
Source: (CSA, 2006b)
Bereha refers to hot lowlands of less than 500m above sea level and crop production in this
region is limited due to a shortage of rainfall. Kolla refers to lowlands between 500 to 1 500m
above sea level. Sorghum, finger millet, sesame, cowpeas and groundnuts are crops
predominantly grown in Kolla AEZ. Generally, Bereha and Kolla are AEZs of the low land
areas of Ethiopia and are not suitable for maize production (CSA, 2006b). Woina Dega refers
to highlands between 1 500 and 2 300m above sea level and it is the most suitable AEZ for
crop production, particularly for maize and teff. Dega refers to cold highlands between 2 300m
and 3 200m above sea level. Barley, wheat, oilseeds and pulses are commonly cultivated crops
in Dega AEZ (CSA, 2006b). Woina Dega and Dega are AEZs where most of the population of
27
the country live and where agricultural activities are predominantly practised (Chamberlin &
Schmidt, 2011). Wurch refers to highlands between 3 200m and 3 700m above sea level and it
is conducive for barley production. Kur refers to highland areas of altitude above 3 700m
above sea level and is primarily used for grazing animals and is not suitable for crop
production (CSA, 2006b).
The general climatic elements such as rainfall, temperature, humidity, sunshine and wind are
affected by geographic locations and altitudes (Mengistu, 2006). The average temperature and
distribution of rainfall vary across regions in the country. The average annual temperature
varies from below 10ºC in the cool highlands to above 35ºC in the hot lowlands (FAO, 2005).
Areas having elevations above 1 500m receive substantially more rainfall than the lowlands.
The average annual rainfall for the country is about 848mm, varying from about 2 000mm
over some areas in South-West Ethiopia to less than 100 mm over lowlands in Afar region
(Tadesse, 2000). The general feature of Ethiopian rainfall is that the rainfalls are often
followed by storms, with very high rainfall intensity and extreme spatial and temporal
variability. There is also very high annual and intra-seasonal drought prevalence in the country
(Tadesse, 2000).
There are diverse soil types in Ethiopia. The MoA (2000) identified 19 soil types in the
country based on the chemical and physical properties derived from parent geological
materials that are modified by weathering and other transformative processes. The six
dominant soil types with their respective distribution in the country are: Leptosols (29.8%),
Nitosols (12.5%), Vertisols (10%), Cambisols (9.4%), Calcisols (9.3%) and Luvisols (7.8%)
(CSA, 2006b). Leptosols have limited agricultural potentials due to the shallowness of the
soils. Vertisols also have limited agricultural potential due to water logging nature (drainage
problem) of the soil, despite having good chemical properties. Cambisols, Calcisols, Nitosols
and Luvisols have relatively good physical and chemical properties for crop production (CSA,
2006b). Given the national summary of agro-ecology, climate and soil types of the country, in
the following section the general overview of agro-ecology, climate, and soil distribution in
the regions of the study are discussed with respect to the relative location of the regions from
the north to west, south, central and East Ethiopia.
28
3.2.1 TIGRAY
Tigray region is situated in Northern Ethiopia and has a diverse topography including peak
highlands, mid-lands and lowlands, which together create diversified agro-ecological zones
(Gebrehiwot, 2008). The dominant agro-ecological zones of Tigray region are Kolla, Weina-
Dega and Dega in decreasing order (CSA, 2006b). The wide range of variation in altitude
influences temperature and climatic conditions of the region. Tigray region has a semi-arid
climate characterised by a long dry season (NMA, 2007). The main rainy season is between
June and September. Rainfall distribution in the region is characterised by high temporal and
spatial variability with annual precipitation ranging from 500 to 1 000mm (NMA, 2007). The
region is mostly prone to drought which causes catastrophic food shortage and periodic famine
(Gebrehiwot, 2008). The main soil types available in Tigray region are Cambisols, Vertisols,
Nitosols and Fluvisols (Hadguet al., 2013). Like other parts of the country, Tigray has high
soil degradation problems in that the soils are deficient in Nitrogen (N) and Phosphorus (P)
(Nedasa, 1999). Although application of N and P is necessary for production, most of the
farmers of the region are not using chemical fertiliser (Hadgu et al., 2013).
3.2.2 AMHARA
Amhara region is found in Northern Ethiopia with diverse altitude ranging from 500 to 4 620m
above sea level (Alemu et al., 2009). The region is predominantly characterised by Kolla,
Weina Dega and Dega agro-ecological zones. The average annual rainfall of Amhara region
varies from 300 to over 2 000mm. The region receives rainfall during the months of June to
September (BoFED, 2011). Drought and land degradation are main challenges to crop
production in Amhara region. The region is inherently prone to food shortage and famine
related to rainfall variability (Bewket, 2009). The dominant soil types of Amhara region are
Vertisols, Leptosols, Luvisols, Alisols and Nitosols (CSA, 2006b). According to Bewket
(2009), most of the soils of Amhara region are deficient of N, P and organic nutrients. The
factors that aggravated soil degradation in the region are ruggedness of the topography,
expansion of cultivation into steep lands owing to increasing population pressure, intense
grazing pressure, and torrential rains that cause soil erosion (Bewket, 2009).
29
3.2.3 BENISHANGUL GUMUZ
Benishangul Gumuz region is located in the North-western part of Ethiopia. The region has a
diverse altitude that ranges from 580 to 2 731m above sea level. The dominant AEZs in the
region are Kolla, Weina Dega and Dega in a decreasing order (CSA, 2006b). According to the
Benishangul Gumuz Food Security Strategy (BGRFSS) (2004) report, average annual rainfall
of the region ranges from 800 to 2 000mm. Predominant soil types of the region are Nitosols
and Vertisols (CSA, 2006b). Generally, the region is characterised by erratic rainfall, soil
degradation, weak infrastructure development of roads and markets and heavy prevalence of
crop pests and diseases that constrain agricultural productivity of the region (BGRFSS, 2004).
3.2.4 SOUTHERN NATIONS NATIONALITIES AND PEOPLES
The Southern Nations Nationalities and Peoples (SNNP) region is located in the Southern and
South-western part of Ethiopia. The predominant agro-ecological zones of the SNNP region
are Kolla, Weina Dega, Dega and Bereha (CSA, 2006b). The region has diverse climatic
conditions ranging from a hot arid and semi-arid climate to tropical humid climate. Average
annual rainfall of SNNP ranges from 400mm to 2 200mm while temperature ranges from 10oC
to 27oC. There is a wide variety of soil distribution in the SNNP region and predominant soils
include Nitosols, Cambisols, Vertisols, Phaeozems, Luvisols and Andisols (CSA, 2006b).
Agriculture is the dominant economic activity of the SNNP where, maize, sorghum, teff,
wheat, coffee and root crops are the main crops produced (BoARD, 2006).
3.2.5 OROMIA
Oromia region extends from Central to Eastern, Southern and Western parts of the country
sharing borders with all other regions except Tigray (CSA, 2006b). Western and Central
Oromia have predominantly Weina Dega and Dega AEZs while the Eastern and Southern
Oromia consist of Kolla AEZ (CSA, 2006b). Annual rainfall of Oromia region ranges from
400 to 2 400mm in which Western Oromia gets the highest rainfall while the lowlands of
Eastern and South-Eastern Oromia has the lowest rainfall in the region (PCDP, 2010).Average
annual temperature of the region varies from less than 10oC in the highlands to over 30oC in
the lowlands (PCDP, 2010). Oromia has relatively fertile soils of volcanic origin, although
there are acidic, basic, ferrogenous, sodic or saline soil types of low agricultural potential
(PCDP, 2010). Dominant soil types of the region are Nitosols, Vertisols, Cambisols, Leptosols
30
and Luvisols (CSA, 2006b). Because of the relatively suitable soil types and higher rainfall,
Oromia is the leading crop producing region in Ethiopia (Oromia Investment Commission,
2012). Accordingly, Oromia region accounts for 49% of major food crops produced in the
country (CSA, 2007). Nevertheless, the soils in the highlands of Oromia have been subjected
to degradation due to erosion, over grazing, deforestation and inappropriate farm management
(PCDP, 2010).
3.2.6 HARARI
Harari is the smallest region which is located in the South-eastern highlands of Ethiopia within
altitudinal range of 1 300 to 2 300m above sea level, surrounded by Oromia region in all
directions (CSA, 2006b). The major agro-ecological zones of Harari region are Weina Dega
and Kolla. Average annual rainfall of Harari varies between 850mm and 870mm and the main
soil types of the region are Luvisols, Vertisols, Cambisols and Acrisols (HPRS, 2011). Maize,
teff and sorghum are the main crops produced in the Harari region (Abesha, 2009).
3.2.7 SOMALI
Somali region is found in the Eastern and South-eastern part of Ethiopia with altitudes ranging
from 200m to 1 800m (DPPB, 2004). Agro-ecology of Somali region is mostly Kolla and
Bereha (CSA, 2006b). Average annual rainfall of Somali ranges from 150mm to 1 000mm
while average temperature ranges from 19oC to 40oC (DPPB, 2004). Although mixed farming
is practised, including maize production, pastoralism is the most prevalent livelihood of the
Somali region (DPPB, 2004). The main soil types of the Somali region are Calcisols,
Leptosols and Gypsisols (CSA, 2006b). Most parts of the region have sandy soils, rocky and
hilly landscapes which are not suitable for crop production (DPPB, 2004). Nevertheless,
sorghum and maize are grown in some parts of the Somali region. There are many problems
confronting the socio-economic conditions of the Somali region. These are drought, flood,
conflict, environmental degradation, crop and livestock diseases, lack of water and poor
infrastructure (SRSS, 2004).
3.2.8 DIRE DAWA
Dire Dawa city administration is found in South-eastern Ethiopia, neighboured by Somali
region in the North, East and West as well as by Oromia region in the South (CSA, 2006b).
31
Dire Dawa administration includes Dire Dawa city and the surrounding rural areas of the
region with altitudes between 960m to 2 450m above sea level. Kolla and Weina Dega are the
main AEZs of the region (CSA, 2006b). Mean annual rainfall varies from 550mm to 850mm.
Dire Dawa has large distributions of Leptosols soil type (CSA, 2006b). Mixed farming
systems are the common farming practise in which sorghum and maize are the main cereal
produced along with livestock rearing (DDAEPA, 2011). Given the regional discussion of the
agro-ecology, climate, soil and cropping potential of respective areas in the study, the
following section provides an overview of agriculture in Ethiopia.
3.3 AGRICULTURE
Ethiopia is an agrarian country where about 43% of GDP and 85% of total employment is
agriculture based (MoARD, 2010). Within agriculture, about 60% of agricultural GDP is
derived from crop production whereas livestock and other agriculture accounts for 27% and
13% of agricultural GDP, respectively (Gebre-Selassie & Bekele, 2013). Because of the
greater contribution of agriculture to the national economy, the government of Ethiopia has
adopted the Agricultural Development Led Industrialisation (ADLI) strategy as a national
development strategy (Diao, 2010).The main goal of ADLI was to attain fast and broad-based
development within the agricultural sector and to stimulate the overall development of the
economy through the linkage effects of agriculture to other sectors (Diao, 2010).
Implementation of ADLI focuses on increasing agricultural productivity of smallholder
farmers through increasing the use of modern farm inputs (example: fertiliser, improved seeds,
pesticides, herbicides) along with better farm management practises (Dercon & Hills, 2009).
However, the agricultural productivity of Ethiopia remains low and the country is unable to
match the food demand of the ever increasing population (IFAD, 2008). Low agricultural
productivity of the country is commonly attributed to limited access of the smallholder farmers
to agricultural inputs, financial services, improved production technologies, irrigation,
agricultural markets, the prevailing poor soil and land management practices (IFAD, 2008).
Given the general overview of Ethiopian agriculture, the farming systems and the rural land
use of the country is summarised as follows.
32
3.3.1 FARMING SYSTEM AND RURAL LAND USE IN ETHIOPIA
Farming in Ethiopia is generally rainfed farming. According to Awulachew et al. (2007), only
about five percent of the total annual agricultural production uses irrigation while the
remaining 95% is based on rainfall. The country is naturally endowed with many irrigable
river-basins and lakes that can be used to irrigate about 3.7 million ha (Awulachew et al.,
2007). The main river basins are Mereb, Tekeze, Awash, Denakil, Abbay, Ayisha, Baro-
Akobo, Omo-Gibe, Wabi-Shebele, Genale-Dawa and Ogaden while most of the lakes are
found in the Rift Valley basin of Ethiopia (Awulachew et al., 2007). Despite having good
potential for irrigation development, the operational irrigation schemes of Ethiopia covers
nearly two percent of the total agricultural land which contributes to the nearly five percent of
agricultural production (Awulachew, 2007). The existing irrigation scheme of Ethiopia is
classified as small-scale if the area covered by irrigation is less than 200ha; medium-scale if
the area is between 200 to 3 000ha and large-scale if the irrigation covers an area larger than
3000ha (Awulachew et al., 2007). Only small-scale irrigation schemes are used for grain crop
production, while medium and large-scale irrigations are used for production of crops such as
cotton, sugarcane, vegetables and fruit in the country (Hordofa et al., 2008). Of the prevailing
small-scale irrigation scheme, 17 to 22% is found in Oromia, Amhara, SNNP and Tigray
regions (Hordofa et al., 2008).
The small-scale irrigation schemes include the modern schemes and traditional schemes.
Modern schemes usually have fixed or improved water control or diversion structures, and
water-users’ associations that have laws while the traditional scheme is developed and
managed by community tradition (Awulachew et al., 2007). The medium and large scale
schemes however, are mostly public schemes owned and managed by the government and in
certain cases by large communities (Awulachew et al., 2007). Some of the reasons for low
levels of irrigation development of the country are a lack of capital to invest in irrigation
development and lack of appropriate water resources’ development strategies for long periods
(Adenew, 2003). Because of the extreme dependence of the country on rainfall, climatic
changes and variability of rainfall influences the livelihood of the country (Awulachew et al.,
2007).
Rural land use in the country is categorised into: crop areas (82%), fallow land (4%), grazing
land (9%), wood land (1%) and other land use (4%) (CSA, 2012b). Crop areas are the parts of
33
rural land that is under annual or perennial crop production, while fallow land refers to rural
land that is kept idle for at least one agricultural season and with a maximum period of five
years. Fallowing of rural land is used to protect farmlands from exhaustion of the important
mineral nutrients caused by continuous cultivation (CSA, 2012b). Grazing land refers to part
of rural land that is used for growing herbaceous forage, while woodland is rural land that is
under tracts of timber which has a value as wood, timber and other wood products (CSA,
2012b).Other rural land use includes areas occupied by the farmers’ houses, gardens, barns,
wells and ponds (CSA, 2012b). The classification indicates that most of the rural land of
Ethiopia is already occupied by crop production. In the following section, crop production and
soil protection practises of the country are assessed.
3.3.2 CROP PRODUCTION
Crop production in Ethiopia consists of cereals, pulses, oilseeds, vegetables, root crops, fruits,
coffee, enset, khat, hops, sugarcane, cotton and tobacco production (CSA, 2012a). Crop
farming is generally categorised into smallholder farming and large-scale commercial farming
based on the acreage cultivated. If crop farming is operated on areas less than 25.2ha, it is
considered as smallholder farming whereas if farming is operated on areas greater than 25.2ha,
it is considered as commercial farming (CSA, 2009a). Commercial farms are not widely
spread in Ethiopia and the contribution of these farms to total agricultural output is less than
four percent of the total agricultural production (CSA, 2009a; Alemayehu et al., 2011).
Smallholder farming is mostly mixed subsistence farming in which crop production is
undertaken side by side with livestock farming. Smallholder farming accounts for 96% of the
total area under agriculture and nearly 95% of the total agricultural GDP (Alemayehu et al.,
2011; Gebre-Selassie & Bekele, 2013).
Given the dominance of the smallholder farming in Ethiopian crop production sector, there are
two distinct cropping seasons for temporary crops in Ethiopia, Meher and Belg (CSA, 2012a).
The classification is based on harvesting time of temporary crops. Belg includes crops
harvested during March to August while Meher includes crops harvested between September
and February. Meher is the main cropping season that accounts for 92% of total crop
cultivated area and 97% of total crop production while Belg season accounts for only eight
percent of total crop cultivated area and three percent of total crop production (Alemayehu et
al., 2011). In this study, Meher season’s crop production performance of smallholder farmers
34
is evaluated because Meher is the main grain crop production season which is also true for
maize production. Given the distinctions of Meher and Belg seasons, a summary of total area
cultivated and total production for cereals, pulses and oilseeds for the year 2011/12 Meher
season is given in Table 3.1.
Table 3.1: Total area cultivated (hectares) and total production (tons) of cereals,
pulses and oilseeds for smallholder farming in 2011/12 Meher season,
Ethiopia
Crop Area Production
Hectares % Tons %
Cereals 9 588 923.71 79.34 18 809 961.70 86.06
Pulses 1 616 809.37 13.38 2 316 201.24 10.60
Oilseeds 880 870.81 7.28 730 880.03 3.34
Total 12 086 603.89 100 21 857 042.97 100
Source: CSA (2012a)
As is indicated in Table 3.1, during the year of 2011/12 Meher season, the total estimate of
grain crops (cereals, pulses and oilseeds) cultivated areas were about 12.08 million hectares
whereas total production was about 21.86 million tons. Cereal was cultivated over an area of
about 9.59 million hectare which accounts for 79.34% of total grain crops area cultivated. The
total cereal production was about 18.8 million tons which accounts for about 86.06% of total
grain production. The five main cereal crops grown in Ethiopia are maize, teff, wheat sorghum
and barley which accounts for the three-quarters of total area cultivated and 29% of
agricultural GDP (Alemayehu et al., 2011). Cereals are the most produced crops in the
country.
Pulses and oil seeds are the second and third highest produced crops, after cereals based on
acreage. Pulses are cultivated on areas of more than 1.6 million hectares which accounts for
about 13.8% of the total grain crop area. Total pulse production is about 2.3 million tons
which accounts for 10.6% of total grain production. The three dominant pulse crops produced
are faba-beans, haricot beans and chickpeas. Likewise, oilseeds are cultivated from an area of
about 880 000 hectares which is equal to 7.29% of total grain area cultivated. The total
quantity of oilseeds produced is about 730 000 tons which accounts for 3.34% of the total
grain production. The main oil seeds produced in the country are niger-seed, sesame and
35
linseed (CSA, 2012a). Of all grain crops, maize is widely produced in Ethiopia and accounts
for 17% of total grain area cultivated and 28% of total grain production. Therefore, maize
production is discussed in the next section.
3.3.2.1 Maize Production
Maize is produced in almost all regions of the country, but the three largest maize producers
are Oromia (61%), Amhara (20%) and SNNP (12%) (Schneider & Anderson, 2010).
Smallholder farmers are the predominant producers of maize during both the Meher and Belg
seasons (RATES, 2003). Meher season accounts for about 78% of total maize area cultivated
and 90.5% of total maize production annually while Belg accounts for 22% of total maize area
cultivated and 9.5% of total maize production (Alemayehu et al., 2011). This study will assess
smallholder farmers’ maize production performance in Meher season only.
The CSA (2012a) report indicated that in 2011/12 Meher season, maize is cultivated from an
area of more than two million hectares, which accounted for 17% of total grain crop cultivated
area, by more than six million smallholder farmers. The total quantity of maize produced on
this area was about six million tons (28% of total grain production) with an average maize
yield of about 2.95 ton/ha (CSA, 2012a). According to the annual reports of CSA during
2003/04 to 2011/12 Meher season, the total area cultivated, total maize production and yield
were following increasing trends (Alemayehu et al., 2011). Figure 3.3 provides the general
increasing trend of the maize area cultivated, total production and yield over the period of
2003/04 to 2011/12 Meher season.
36
7
6
5
4
3
2
1
0
Year
         Production in million tons      Yield in tons/ha Area in million hectares
Figure 3.3: Graphical representation of smallholder farmers’ maize area cultivated,
total production and yield during 2003/04 to 2011/12 Meher season,
Ethiopia
Source: CSA (2003/04 to 2011/12)
From Figure 3.3, total maize area cultivated, yield and total production showed an increasing
trend for the given period. The level of maize area cultivated increased from about 1.34
million hectares to more than two million hectares between 2003/04 and 2011/12 (49%
increase). Total maize production also increased from about 2.5 million tons to about 6.1
million tons (144% increase). The increase in production is partly due to the increased
cultivation of additional lands and due to increased yield overtime. The increase in yield is the
bigger contributor to the increase in production relative to the area. The average maize yield
increased from about 1.86 ton/ha in 2003/04 to about 2.95 ton/ha in 2011/12 (58% increase).
The increase in yield is mainly attributed to the increased use of modern farm inputs such as
fertiliser and improved seeds as well as better utilisation of farm management practises
(Alemayehu et al., 2011). Even though, maize yield was increasing, the observed yield level is
still lower than the potential yield level of the country (Alemayehu et al., 2011). According to
Alemayehu et al. (2011), there are various constraints in Ethiopian maize production such as
underutilisation of modern farm of inputs (i.e, fertiliser, improved seeds and pesticide),
insufficient use of irrigation, high soil degradation due to erosion, inadequate agricultural
37
Production (million tons), area (milion ha),
yield (ton/ha)
research and extension services and constraints in market development (i.e, input supply,
storage problem and price volatility).
The increase in production that is attributed to an increase in area cultivated can be as a result
of increased cultivation of marginal and less productive lands that are potentially not suitable
for crop production (Alemayehu et al., 2011). This is because the more productive highland
areas of Ethiopia are already under cultivation and any expansion of areas cultivated can be at
the expense of reduction in forests and grazing lands (Alemayehu et al., 2011). Increasing
production by further increasing the area cultivated is not a sustainable means of production
growth due to the facts that land is a limited resource and due to the negative environmental
implications(Alemayehu et al., 2011). Sustainable growth of crop production requires
increasing the productivity of the farms already under operation rather than expanding areas
cultivated to marginal and unproductive areas (Alemayehu et al., 2011). Soil conservation and
land management practises can contribute to enhanced productivity of the operational
farmlands and can also ensure sustainability of the land use. Given the overview of
smallholder maize production, in the following section, soil conservation and land
management practise of the country is discussed.
3.3.3 SOIL CONSERVATION AND LAND MANAGEMENT
Soil conservation and land management practise enhances land productivity. For example,
Benin (2006) and Pender and Gebremedhin (2006) found that soil protection through terracing
increased crop yield. Similarly, Zikhali (2008) found that contour ridges have a positive
impact on land productivity while Kassie et al. (2011) found that minimum tillage has a strong
positive influence on crop productivity. Soil conservation practises such as crop rotation,
fallowing and inter-cropping are very limited in the highlands and potentially agricultural
areas of the country due to small and fragmented farm sizes which are not suitable for such
practises (IFPRI, 2010a).
As a result, Ethiopia became one of the countries that have the highest rate of soil nutrient
depletion in sub-Saharan Africa (IFPRI, 2010a). There is severe topsoil degradation, organic
matter depletion and depletion of soil physical, macro and micro-nutrients as well as salinity
and acidity problems (IFPRI, 2010a).According to Hurni (1993), soil erosion causes on
average, a loss of about 4.2 tons of soil per hectare per year in Ethiopia which contributes to
38
the decrease in productivity. As a result, most of the soils are depleted and became deficient in
important soil nutrients including N and P (Mengistu, 2006).
Land degradation in the country is exacerbated by factors such as deforestation, erosion,
population pressure, overgrazing and inadequate planning of land use (MoARD, 2010).
Furthermore, Pender and Gebremedhin (2006) noted that factors such as cultivation of
marginal lands, soil erosion, continuous mono-cropping, climatic changes and high population
growth are factors that contributed to the increased land degradation in Ethiopia. The
cumulative impact has resulted in the continued dependence of the country on food aid
(Pender & Gebremedhin, 2006).
3.3.4 SUMMARY
Ethiopia has a diverse topography which determines the agro-ecology, climate and farming
systems in the country. Agriculture is the dominant sector of the Ethiopian economy and is
characterised by smallholder rainfed farming systems. Crop production is the largest
contributor of the agricultural GDP and there are two cropping seasons, Meher and Belg.
Among the crops produced in the Ethiopia, maize is an important food crop where Oromiya,
Amhara and SNNP are the largest maize producing regions. Maize is produced during both
seasons but, Meher is the main maize production season in terms of total area, total production
and yield.
Historical data of maize production in Ethiopia during 2003/04 to 2011/12 Meher season,
indicated that maize production increased over time, along with the maize area cultivated and
yield. Therefore, the increase in maize production is the result of the increase in area cultivated
and yield. The increase in output through increasing the area cultivated is not a sustainable
way of increasing productivity because an increase in production will take place within an
environment characterised by a scarcity of arable lands. Rather, increasing productivity from
the operational farms is important for sustainable crop production. The prevailing low level of
soil conservation practise in the country also resulted in the depletion of important soil
nutrients including phosphorous and nitrogen which in turn decreases the level of productivity.
By having adequate information about the state of technical efficiency of the smallholder
maize farmers and by examining the determinants, productivity can be improved.
39
3.4 TYPE AND SOURCE OF DATA
The data used in this study is secondary data obtained from the Central Statistical Agency of
Ethiopia. The data formed part of the Ethiopian Rural Socio-economic Survey (ERSS). The
Central Statistical Agency of Ethiopia, in collaboration with the World Bank, conducted a
living standard and agricultural survey in 2011/12. The purpose of the survey was to obtain
comprehensive agricultural, welfare and socio-economic information on rural and small town
households in the country. Under the ERSS, the CSA used a two-stage stratified cluster
sampling procedure to select the agricultural households, first by selecting enumeration areas
(EAs) and then the households surveyed. The CSA employed a questionnaire based on
personal interviews to collect the data. For the current study, only farm households that are
involved in maize productions were considered consisting of 438 observations. The sampled
farm households considered from the corresponding regions and the associated percentage are
provided in Table 3.2.
Table 3.2: Number of sampled households considered from the selected regions and
respective percentages in the secondary data set used
Regions Number of households Percentage of households in
chosen from each region the sample from respective
regions
Tigray 64 15
Amhara 107 24
Benishangul Gumuz 27 6
SNNP 96 22
Oromia 103 24
Harari 21 5
Somali 9 2
Dire Dawa 11 2
Total 438 100
Table 3.2 indicates the number of sampled households considered from the selected regions
and respective percentages in the sample. In order to ensure a sufficient sample size in the
most populous regions, the CSA has set larger quotas for Amhara, Oromia, SNNP, and Tigray
regions. Therefore, the percentage of the sample for the respective regions include: Oromia
(24%), Amhara (24%), SNNP (22%), Tigray (15%), Benishangul Gumuz (six percent) and
Harari (five percent), Somali (two percent) and Dire Dawa (two percent). Based on the data
40
used in the study, the respondents can be characterised on their production, socio-economic
and farm management information.
3.5 CHARACTERISATION OF RESPONDENTS
The purpose of this section is to discuss the characteristics of the respondents. The first section
discusses maize yield and the production inputs used by the respondents. The second section is
used to discuss the socio-economic and farm management practices of the respondents.
3.5.1 MAIZE YIELD
Maize yield is the quantity of maize produced in tons per hectare. Maize yield estimation is
calculated based on the maize crop cutting from an area of 4m2 for respective farmers in the
sample (CSA, 2013). The summary statistics of maize yield of the respondents in the sample is
provided in Table 3.3.
Table 3.3: Summary Statistics of maize yield of the sampled respondents during
2011/12 Meher season
Variable Unit Average Std. Deviation Minimum Maximum
Maize Yield ton/ha 2.61 1.50 0.00 8.30
From Table 3.3, the average maize yield estimate is 2.61 ton/ha. The yield level was highly
variable among the respondents with minimum and maximum yields of zero and 8.3 ton/ha,
respectively and a standard deviation of 1.5 ton/ha. The minimum crop yields of zero ton/ha
was not expected. However, upon investigation of the maize yield reported, it was found that
farmers were not able to harvest a crop due to damage to the crop. Crop damage could be due
to frost, drought, flood, insects or other natural calamities (CSA, 2013). The inputs used in
maize production include several different variables including seed, fertiliser and labour. The
input variables used to estimate the production function is discussed in the following section.
41
3.5.2 SEED AND FERTILISER
Seeds and fertilisers are the primary inputs in crop production (Ayele & Bosire, 2011).
Summary statistics of total seed and chemical fertilisers used are provided in Table 3.4.
Farmers had the choice to use either improved seeds or traditional seeds. Similarly, the
respondents used different combinations of fertiliser, including chemical, organic,
combinations of chemical and organic or neither. Information regarding the quantity of the
organic fertiliser applied was not available therefore the quantity of organic fertiliser used will
not be discussed.
Table 3.4: Summary statistics of seed and fertiliser use of respondents
Variables Unit Average Std. Deviation Minimum Maximum
Seed (total) Kg/ha 72 207 1 3 473
Improved seed Kg/ha 45 59 1 447
Traditional seed Kg/ha 77 226 1 3473
Phosphorus Kg/ha 15 58 0 999
Nitrogen Kg/ha 20 63 0 988
From Table 3.4, the average quantity of total seed used in maize production for the sample is
nearly 72 kg/ha with a standard deviation of 207 kg/ha. The standard deviation indicates a
very wide variation in seed use among respondents. It is observed that some of the farmers
used improved seed while others used traditional seed in maize production. The average
quantity of improved seed used is 45 kg/ha while that of traditional seed use is 77 kg/ha. This
indicates that the farmers applied on average greater quantities of traditional seeds than
improved seeds per hectare. The standard deviation is 59 kg/ha for improved seed and 226
kg/ha for traditional seed which also indicates wider dispersion in the quantity of traditional
seed used over that of improved seed. The percentage of respondents using improved seeds
compared to those using traditional seeds is provided in Figure 3.4.
42
17%
83%
Improved seed Traditional seed
Figure 3.4: Distribution of respondents by type of seed used
Figure 3.4 indicates that only 17% of the respondents used improved seed while 83% used
traditional seed. This indicates that the use of improved seed is very low compared to
traditional seed. Based on the type of seed used, there is a yield difference among the
households in the sample. The average maize yield of the households that applied improved
seed is 3.38ton/ha while the average of the farmers that used traditional seed is 2.47ton/ha.
The maize yield obtained from improved seed is greater than that of traditional seed, which is
expected. According to the literature, less improved seed is used in Ethiopia as there is
insufficient supply of improved seed relative to demand, limited choice in the varieties
available in the market, lack of rural credit, low level of extension services, high costs of
improved seeds and limited competition among seed suppliers (Alemu et al., 2009).
According to Spielman et al. (2011), 60% of maize seed supply is controlled by the public
sector, primarily the Ethiopian Seed Enterprise (ESE).Due to the low seed production, the
supply of improved seed in Ethiopia usually falls short of the demand (Spielman et al., 2011).
In addition, seed is distributed after the appropriate planting time (Sahlu & Kahsay, 2002 &
DSA, 2006).Other problems experienced in the seed supply chain include poor quality seeds
such as poorly cleaned seeds, broken seeds and low germination rates (DSA, 2006).To
improve the nutrient levels of soils in the ground in order to increase productivity, farmers
may apply chemical and organic fertilisers in maize production.
43
The quantities of Phosphorus and Nitrogen nutrients used that are presented in Table 3.4 are
calculated based on DAP and Urea fertilisers used by each household. On average, farmers in
the sample applied 15kg/ha of Phosphorus with a standard deviation of 58 kg/ha while the
average application of Nitrogen is 20kg/ha with a standard deviation of 63 kg/ha. Integrated
application of organic fertiliser (farm yard manure, compost, green manure and transferred
biomass of leguminous trees) and chemical fertiliser improves soil fertility and enhances crop
yield (Ayalew & Dejene, 2011). Therefore, distribution of the respondents with regard to
application of chemical and organic fertiliser is important. Figure 3.5 provides graphical
representation of the distribution of respondents based on fertiliser use.
11%
36%
17%
36%
Chemical and organic Chemical but no organic
No chemical, but organic No chemical and no  organic
Figure 3.5: Distribution of respondents based on fertiliser use
Figure 3.5 indicates four categories of respondents observed which include farmers that used
only chemical fertiliser, farmers that used both chemical and organic fertilisers, farmers that
used only organic fertiliser and farmers that used neither of the two fertilisers. The percentage
of respondents that used a combination of chemical and organic fertiliser is 11%, while the
percentage of respondents that used only chemical fertiliser is 17%. Another 36% of the
respondents used only organic fertiliser while the remaining 36% used neither chemical nor
organic fertiliser.
44
Despite the fact that the use of chemical fertiliser increases maize productivity considerably,
most of the respondents did not use chemical fertiliser. There are various reasons for the low
adoption and utilisation of chemical fertiliser in Ethiopia. According to Fufa and Hassan
(2006) and Tesfaye et al. (2011), the primary reasons for low utilisation of chemical fertilisers
are farmers’ expectations of rainfall, high price of fertiliser, limited risk management services,
lack of rural credit, small farm sizes, lack of transportation infrastructure and unavailability of
fertiliser in the markets at the right time.
According to Fufa and Hassan (2006), farmers’ expectations about rainfall conditions
influence the use of chemical fertiliser. Whenever the expected rainfall is less than normal,
farmers are unwilling to use chemical fertiliser (Fufa & Hassan, 2006). The high price of
fertiliser is associated with high transaction costs incurred by the suppliers (related to
transportation, storage and handling costs), as well as taxes and profit earned by the sellers
(Jayne et al., 2003). Unavailability of fertiliser at the right time for the farmers is related to the
international procurement and shipping to import fertiliser (Tesfaye et al., 2011).Currently, the
Agricultural Input Supply Enterprises (AISE) and the cooperative unions are the sole
importers and distributors of chemical fertilisers in the wholesale and retail market in Ethiopia
(Spielman et al., 2011). Absence of competitive private traders as an alternative supply source
is also a problem causing lower adoption and utilisation of fertiliser (Spielman et al., 2011). In
addition, low levels of education amongst the farmers and insufficient information provided by
extension services contributes to the limited adoption and utilisation of chemical fertilisers
(Zerfu & Larson, 2011; Spielman et al., 2011). The cost and utilisation of seed and fertiliser is
affected by the farm size and availability of labour. These attributes are also important and
discussed in the following section.
3.5.3 FARM SIZE AND LABOUR
Farm size refers to the size of the maize farm lands measured in hectares (ha). Similarly,
labour refers to the quantity of labour used in man-days per/ha during the entire Meher season
for maize production. Labour is an aggregation of household and hired labour with the
assumption that each person works on average eight hours per day on maize farms. Table 3.5
provides a summary of farm size and labour use of the respondents.
45
Table 3.5: Summary statistics of farm size and labour use of respondents
Variables Unit Average Std. Deviation Minimum Maximum
Farm size ha 0.13 0.18 0.001 2.36
Labour man days/ha 392.33 1 185.07 1.19 14 549.28
From Table 3.5, the average farm size is 0.13 ha with a standard deviation 0.18 ha. The
minimum and maximum farm sizes are 0.001 ha and 2.36 ha, respectively. Although the
average farm size of the sample is very small, there is strong variability of the sizes cultivated.
The average farm size of the sample (0.13 ha) is smaller than the national average farm size of
smallholder farmers in Ethiopia which is 0.96 ha (CSA, 2012b). The average amount of labour
used is 392 man days/ha with a standard deviation of 1 185 man days/ha. There is also wide
variation of labour used among the respondents. This can be due to the fact that the farmers
are cultivating very small farm sizes and when the amount of labour used is converted on a per
hectare basis, the value of labour is magnified. Apart from physical inputs used in the
production process, several other factors can also influence the efficiency of production such
as age, gender of the household heads and several other socio-economic variables.
3.5.4 SOCIO-ECONOMIC VARIABLES
The socio-economic variables in the study include age, gender, education, household size,
tenure, ownership of adequate oxen, credit and off-farm income. The descriptive statistics of
the variables are discussed subsequently starting with age and household size.
3.5.4.1 Age and Household Size
Age of a household head is measured in years while household size refers to the number of
household members living in each farm household which can be either a person living alone or
a group of people who live together sharing a common house or supported by a common
budget (CSA, 2012a).Table 3.6 provides information on the summary of age and household
size of the respondents.
46
Table 3.6: Summary statistics of age of farm household heads and family size
Variables Unit Average Std. Deviation Minimum Maximum
Age of household head years 44 15 18 93
Family size count 5 2 1 13
The average age of the head of the farm household is 44 years. The minimum and maximum
ages are 18 and 93, respectively. Similarly, average household size of the respondents is five.
The minimum household size is one while the maximum is 13. The household size of the
sample is also nearly equal to the average rural household size in Ethiopia which is 4.9 (CSA,
2011b). Household size commonly influences agricultural performance through the
contribution to family labour (Fesessu, 2008; Baruwa & Oke, 2012).
3.5.4.2 Gender and Education
The distribution of the respondents based on gender of the household heads is represented
graphically in Figure 3.6.
15%
85%
Male Female
Figure 3.6: Gender distribution of the household heads
From Figure 3.6, the gender distribution revealed that 85% of the respondents are male-headed
households while 15% are female-headed households. In other words, the sample contains 371
47
male-headed and 67 female-headed households. The gender of farm households can be a cause
for technical efficiency variation among the respondents (Bachewe, 2009). Gender is an
important issue in Ethiopia because of the prevalence of gender inequality in the country. Female-
headed households can be inefficient relative to male-headed households as female-headed
households have less opportunities to access land, credit and technological inputs (OXFAM, 2012;
AWM, 2009).As education can influence decision making and other managerial capabilities on
a farm, the following section is a discussion of the education level of the respondents.
Education of household heads may enhance farm productivity by affecting households’
decisions to adopt and use technological innovations better (Weir, 1999). Based on education,
the sampled households are classified into two categories: those having household heads with
formal education of grade three or less and those with grade four or above. The classification
is in accordance to the current education policy of Ethiopia, where completion of formal
education of grade four is considered as a basic education threshold that can influence
agricultural decision making of the farmers (MoE, 2008).Education distribution of the sample
is indicated in Figure 3.7.
24%
76%
Grade three or less Grade four and above
Figure 3.7: Distribution of respondents based on education of the household heads
From Figure 3.7, about 76% of the respondents have household heads with an educational
level of grade three or less while 24% of the respondents have an education of grade four or
above. The category is based on the idea that farmers that have completed grade four could
48
develop basic knowledge that is required to handle the agricultural decision makings better
than the farmers not having the educational level. Due to variations in the education of the
household heads, the extent of technical efficiency of the farmers can vary.
3.5.4.3 Ownership of Resources (Tenure and Oxen)
Tenure refers to the form of farm land ownership of the respondents which is categorised into
private ownership and other forms of land ownership that can be renting or share cropping.
Figure 3.8 provides the distribution of respondents based on tenure type.
13%
87%
Private tenure Other tenure
Figure 3.8: Tenure distribution of the respondents
As indicated in Figure 3.8, about 87% of the respondents own the land they farm on while
13% of the respondents are operating on rented or share cropping land. Private ownership of
farm lands gives the farmers greater incentives for practising conservation and investing in the
farm (Alene & Hassan, 2003; Uaine, 2008).
Similar to tenure, ownership of oxen of the respondents during the production season can
cause performance variation. Ownership of adequate oxen enables the farmers’ timely
operation of farming activities. Distribution of the respondents based on ownership of
sufficient oxen (having at least two oxen) is provided graphically in Figure 3.9.
49
38%
62%
Have two or more oxen Less than two  oxen
Figure 3.9: Oxen ownership distribution of the respondents
As indicated in Figure 3.9, about 62% of the respondents did not have an adequate number of
oxen, while 38% of the respondents had an adequate number of oxen. Farmers that had
adequate oxen may have had an advantage over the farmers that did not have adequate oxen
for timely preparation of their farmlands (Gebreegziabher et al., 2004; Geta et al., 2010).
3.5.4.4 Credit and Off-farm Income
Access to rural credit and off-farm income activities are important socio-economic variables
that can influence smallholder farmers’ ability to purchase inputs and can also improve the
livelihood of the households. Figure 3.10 provides the distribution of the respondents with
regard to credit access.
50
29%
71%
Accessed credit No credit
Figure 3.10: Credit access based distribution of the respondents
As indicated in Figure 3.10, only about 29% of the respondents received credit while the
remaining 71% did not receive credit. The sources of finance in Ethiopia are state and private
owned banks, Micro Finance Institutions (MFIs), rural saving and credit cooperatives
(RUSACCOs), and informal money lenders that include traders, neighbours, friends and
family (Ameha, 2011).The existing banks and their branches are concentrated in urban areas
and provide credit to traders, wealthy individuals and government projects. In addition, the
banks set high interest rates and request collateral which limits credit to mainly large land
owners (Getahun, 2001). Therefore, most of the needy smallholder farmers are excluded from
credit services provided by banks in Ethiopia (Ameha, 2011). The predominant suppliers of
credit for smallholder farmers are the MFIs and RUSACCOs. Ameha (2011) noted that MFIs
have achieved remarkable growth in terms of credit outreach and service delivery to rural farm
households.
Many factors constrain rural credit in Ethiopia. The literature indicates that some of the
constraints are: limited capacity of credit outreach to rural areas; absence of financial
institutions and the associated high travel costs; shortage of loanable funds and low saving
habit of the clients (Yehuala, 2008; IFAD, 2011). Yehuala (2008) noted that the presence of
complicated credit application procedures, requirement of security, minimum loan amounts,
terms of repayment, restriction of credit for specific purposes are some of the constraints
51
farmers face. Furthermore, Ali and Deininger (2012) found that political and social networks
are key determinants of credit access in the rural credit service system of the country. Given
the credit system and the constraints in Ethiopia, farmers’ access to credit service can
potentially cause performance variation among the respondents (Binam et al., 2004).
Similar to credit, some of the farmers have participated in off-farm income activities as a
supplementary source of income. The distribution of respondents with regard to off-farm
income participation is presented graphically using Figure 3.11.
46%
54%
Have off-farm income No off-farm income
Figure 3.11: Off-farm income participation based distribution of the respondents
Figure 3.11 indicates that 54% of the respondents have received off-farm income whereas
46% of the respondents received none. Since production and productivity of the smallholder
farmers is low, the households’ farm income may not be sufficient for the household’s
consumption, to purchase inputs and make necessary investments on their farms (Beyene,
2008). As a result Ethiopian households commonly participate in off-farm income activities to
get additional income during slack periods (Tesfaye et al., 2011). The off-farm income
participation of the households consider if any of the household members have participated in
off-farm income activities. Farm households that received off-farm income could be able to
purchase inputs and necessary investments in the farming sector which could potentially
improve their farm productivity (Beyene, 2008). The pursuit of off-farm income could also
52
lead to a decrease in farm productivity if the farmers failed to manage the farms properly due
to their involvement in other activities. Although farm households commonly participate in the
off-farm income activities during the slack periods, the secondary data used do not provide
information whether the households participated in any off-farm income activities during slack
periods. The shortfall of information on the off-farm activities of the household can hold a
shortcoming for the research. It is expected that household who participate in off-farm activity
can show increased technical efficiency in production (Haji, 2006).
3.5.5 FARM MANAGEMENT PRACTISES
This section discusses farm management practises of the respondents, including the
distribution of respondents with regard to access to extension services, irrigation use and soil
protection activities.
3.5.5.1 Extension Services
Extension service refers to the advice and training provided by agricultural extension experts
regarding better farm management. Figure 3.12 provides the distribution of respondents with
respect to extension.
21%
79%
Have extension service No extension service
Figure 3.12: Extension service based distribution of the respondents
53
Figure 3.12 indicates that 79% of the households have no access to extension services whereas
21% of the respondents have access to extension services during the production season. The
low level of extension service use by the respondents can be due to the low outreach of
extension services in the country (Spielman et al., 2011). Nonetheless, there are some
improvements in the extension service system of the country in recent years which include
expansion of training and deployment of extension agents, shifting the responsibility of input
distribution from extension agents to co-operatives and the provision of agro-ecological zone
specific extension services (Spielman et al., 2011).However, rigorous monitoring and impact
evaluation of the extension service delivery system is lacking in the country (Spielman et al.,
2011).
3.5.5.2 Irrigation
Irrigation is a means of supplying sufficient water for better crop production which is specific
to the smallholder farmers studied. Through the use of irrigation, farmers can improve crop
production and intensification, thereby sustaining and improving their livelihood and food
security (Hordofa et al., 2008).Distribution of the sampled households based on the use of
irrigation is presented in Figure 3.13.
4%
96%
Irrigated Not irrigated
Figure 3.13: Distribution of respondents based on irrigation use
54
From Figure 3.13, it is observed that about four percent of the respondents used irrigation
whereas the remaining 96% used none. This suggests that the use of irrigation in the
smallholder farmers studied is very limited. The overall existing irrigation schemes in the
country(including the commercial farms) cover about two percent of the total agricultural land
(Hordofa et al., 2008).Only small scale irrigation and water harvesting schemes are used for
smallholder grain crop production (Hordofa et al., 2008). Ground water irrigation and
agricultural water management schemes are not integrated into the agricultural development
policy and strategy of the country (Awulachew et al., 2007).
3.5.5.3 Soil Protection
Soil protection refers to soil conservation practises by the respondents to protect maize
farmlands from erosion by using different methods such as contour ploughing, terracing,
planting trees, inter-cropping. Soil conservation practises are very important for sustainable
productivity improvement (IFPRI, 2010a).Figure 3.14 provides the distribution of respondents
based on their practising of soil protection.
5%
95%
Soil protection No soil protection
Figure 3.14: Distribution of respondents based on soil protection
Figure 3.14 indicates that only five percent of the respondents practised soil conservation to
avoid erosion, while 95% of the respondents did not. This distribution shows that most of the
55
maize farms are not protected from erosion and the soils and soil nutrients are likely to be
easily eroded.
3.6 SUMMARY
Characterisation of the respondents shows that there are low levels of maize yield and
utilisation of improved seeds and chemical fertilisers. Farm specific socio-economic variables
of the data indicated that the average farm size is very small. Most of the households in the
sample are male-headed and most of the household heads have an educational level of grade
three or less. Tenure distribution of the sample implied that most of the households are
operating on their private farmlands while the oxen distribution shows that most of the
households do not have an adequate number of oxen. The respondents’ credit and off-farm
income distribution indicated that the majority of the households in the sample were not
receiving rural credit whereas nearly half of the households received additional off-farm
income. Farm management practises of the sample indicated that there is little use was made
of extension services (due to low supply), irrigation or soil conservation.
Given the low levels of output and input use, weak socio-economic status, and poor farm
management practises, measurement of technical efficiency and examining the determinants
based on the current state of input and technology is important to have sufficient information
about the scope of performance improvement of the smallholder maize farmers.
56
CHAPTER FOUR: METHODOLOGY
Chapter four discusses the methodology used to achieve the objective of the study. The
chapter is divided into five sections. The first section is a justification of the SFA model. The
second section discusses the procedures followed to estimate technical efficiency and to
analyse the factors affecting technical inefficiency. The third section discusses the empirical
application of the SFA model, while the fourth section discusses the Principal Component
Regression technique used in the SFA estimation. The final section discusses the hypothesis
tests used to evaluate the estimation procedure.
4.1 JUSTIFICATION OF THE STOCHASTIC FRONTIER ANALYSIS
MODEL
SFA was developed nearly simultaneously by Aigner et al. (1977) and Meeusen and van den
Broeck (1977) who were motivated by the idea that deviation from the production frontier
might not entirely be under the control of the producer. The SFA model has the ability to
separately identify the effects of statistical noise from that of technical inefficiency through a
composed error term(εi). The composed error term consists of a random noise component (Vi)
that captures the effects of statistical noise and an inefficiency component (Ui) that captures
the effects of technical inefficiency (Kumbhakar & Lovell, 2000).The use of SFA has found
wide acceptance in literature because of its consistency with theory, versatility and relative
ease of estimation (Battese & Coelli, 1992; 1996). The SFA is commonly applied in
agricultural efficiency measurements, especially in developing countries where data are
heavily influenced by statistical noise (Sarafidis, 2002).
Efficiency analysis using SFA requires separate assumptions about the distributions of the two
error components that leads to a more accurate efficiency estimation. The independently
distributed noise component follows a normal distribution with zero mean and constant
variance. The inefficiency component however, has developed from the original half-normal
and exponential distributions of Aigner et al. (1977) to truncated normal and gamma
distributions by Stevenson (1980) and Green (1990), respectively. The truncated normal
distribution generalises the half-normal distribution through assuming Ui to be truncated at
zero and to have a non-zero mean (Fried et al., 2008: 130). As a result, either mean or mode of
57
the technical inefficiency error term is used to estimate technical efficiency of each producer
(Kumbhakar & Lovell, 2000: 85). Distributional assumptions to estimate some frontier models
are automated in software packages (Coelli et al., 2005:252).For example, FRONTIER
estimates half-normal and truncated normal models whereas LIMDEP estimates exponential
and gamma models (Coelli et al., 2005: 252).For this study, a truncated normal distribution is
assumed as the truncated normal distribution provides a more flexible representation of the
pattern of efficiency in the data (Kumbhakar & Lovell, 2000).
SFA estimates a production efficiency frontier by using the actual data. The efficiency scores
are obtained from the deviations of the observed production from the relative production
frontier (Sarafidis, 2002).After the introduction of SFA, the model was used to estimate
technical efficiency (Battesse & Coelli, 1993). However, the studies were unable to explain
technical inefficiency because the model could not analyse the sources of efficiency variations
among the producers (Battesse & Coelli, 1993). In the pursuit of finding explanations of the
technical inefficiency model, two approaches have been developed to explain technical
inefficiency (Pitt & Lee, 1981; Kumbhakar et al., 1991; Battesse & Coelli, 1995).Early studies
adopted a two-stage approach where, efficiencies are estimated first and then the estimated
efficiencies are regressed against a vector of explanatory variables in the second stage
(Kumbhakar & Lovell, 2000). The two-stage approach suffers from inconsistency of
assumptions regarding the independence of distribution of the inefficiency error terms
(Kumbhakar et al., 1991; Binam et al., 2004; Fried et al., 2008). The inconsistency of
assumptions implies that in the first step, SFA model estimates technical efficiency by
assuming that the inefficiency error term is independent of the influence of the vector of farm
specific explanatory variables. In the second step, the relationships between the inefficiency
error term and the farm specific explanatory variables is assumed as if there is a linear
relationship (Schmidt, 2011).
More recent studies including Kumbhakar et al. (1991), Reifschneider and Stevenson (1991)
and Battesse and Coelli (1995) adopted a single-stage approach. In a single-stage approach,
technical efficiency estimation and the relationships of technical inefficiency error term and
the vectors of explanatory variables are estimated in a single step. Unlike the two-stage
approach, the single-stage approach allows simultaneous estimation of the production frontier
and the technical inefficiency model parameters (Coelli & Battese, 1995). In this study, a
single-stage SFA model is used. The choice of the single-stage approach over the two-stage is
58
due to the short-comings of the two-stage approach arising from inconsistency of assumptions
about the distribution of the inefficiency error terms. The estimation of technical efficiency
and its determinants required the use of FRONTIER 4.1 (Coelli, 1996).
4.2 ESTIMATING TECHNICAL EFFICIENCY USING STOCHASTIC
FRONTIER ANALYSIS
Assuming that producers are producing a single output using multiple inputs, SFA provides
the relative frontier against which production performance is evaluated (Kumbhakar & Lovell,
2000: 63). The specification of SFA incorporates the effects of random noise and technical
inefficiency Ui, on output(Kumbhakar & Lovell, 2000).The stochastic production frontier
specification incorporates the difference of the technical inefficiency variable from that of the
symmetric random variable that affect output (Bachewe, 2009). A stochastic production
frontier given by Battesse and Coelli (1995) for a cross-sectional data takes the form:
= ( + − ) ( . )
where, Yi is the scalar of output for producer i; Xi is a vector of input variables used by
producer i; β is a vector of unknown parameters to be estimated; Vi is a symmetrically
distributed error term which is assumed to be identically and independently distributed with
mean zero and unknown variance, that is Vi ~ iidN(0, σ2V); Ui is a non-negative random
variable associated with technical inefficiency in production. Ui is assumed to be
independently distributed and is obtained by truncation (at zero) of the normal distribution
with a mean,(Ziδ) and variance, (σ2U) (Battesse & Coelli, 1995).The truncation(at
zero)distribution of Ui implies that the deviation of actual production from a frontier
production is assumed to take a zero or positive value. Producers with zero deviations are
efficient producers that lie on the efficiency frontier while those with positive deviations lie
below the efficiency frontier and are inefficient (indicating shortfalls from the frontier output,
for the given level of inputs and technology) (Bachewe, 2009).Zi is a vector of variables that
influence technical inefficiency of the producers and δ is a vector of technical inefficiency
parameters to be estimated (Battesse &Coelli, 1995).Given the stochastic production frontier
of producer i by Equation 4.1, the associated technical efficiency (TE) can be estimated.
59
TE measures the output of producer i relative to the output that could be produced by a fully
efficient producer using the same input vector (Coelli et al., 2005). TE of producer i is
estimated by the ratio of actual output to the relative frontier output as specified by Coelli et
al. (2005):
= = ( (−+ )) ( . )
where, = ( + − ) is the actu(al ou+tput )which is obtained in the presence of thetechnical in efficiency effects whereas is the corresponding frontier output
under condition of random shocks (Coelli et al., 2005).When dividing the actual−output by thefrontier output, after cancelling the similar terms, the remaining value is exp( ) , which
represents technical efficiency. The value of TE ranges between zero and one (0≤TE≤1). TE
takes a value of one when the producers are technically efficient, becomes closer to zero when
the producers are less technically efficient and becomes zero when producers are fully
inefficient (Bachewe, 2009).
Given the stochastic production frontier specification by Equation 4.1, the potential
determinants of technical efficiency can be identified. The technical inefficiency term Ui is
assumed to be a linear function of some explanatory variables, Zi and a parameter of the
technical inefficiency model to be estimated, δ. According to Battesse and Coelli (1995), the
technical inefficiency effect, Ui in the stochastic frontier model displayed in Equation 4.1 takes
the form:
= + ( . )
where, is a random variable defined by normal distribution with a mean of zero and
unknown variance ( ). This assumption is consistent with Ui being a non-negative
truncation of normal distribution with a mean, (Ziδ)and variance, (σ2U) such that the point of
truncation is at (-Ziδ) that means Ziδ (Battesse &Coelli, 1995).
60
The empirical estimation of the technical efficiency required the use of principal component
analysis to eliminate the structure in the variables used to explain technical inefficiency.
During the estimation of the production frontier and the technical inefficiency model a
significant level of mulitcollinearity was present. The data structure was therefore reduced
with the use of principal component analysis before the production frontier and technical
inefficiency model was solved using Frontier. The regression results along with the
eigenvector information were used to estimate the significance of the individual variables that
was retained in the principal components. The next section will discuss the estimation of
technical efficiency
The rest of the Chapter is used to discuss the procedures used to estimate the production
function and the factors of technical inefficiency. The first section will discuss the choice of
the functional form the estimation and the variables used to determine the production frontier
and the inefficiency model. The last section of the Chapter is dedicated to explaining the
Principal Component regression procedure used to overcome multicollinearity followed by the
some hypothesis tests.
4.3 EMPIRICAL ESTIMATION OF TECHNICAL EFFICIENCY
4.3.1 CHOICE OF FUNCTIONAL FORMS
The production frontier given by Equation 4.1 can be specified using different functional
forms such as linear, quadratic, Cobb-Douglas, translog and Leontief (Coelli et al., 2005: 211).
Functional forms are commonly determined by requirements such as flexibility of the
functional forms, the general conformity and adequacy of the models in explaining a given
data and on theoretical bases hypothesised to adopt a specific functional form (Griffin et al.,
1987). The uses of Cobb-Douglas and translog functional forms dominate applications of
production frontier literature (Fried et al., 2008: 98). Given the alternatives, different
functional forms were compared using the data and various model evaluation criteria such as
R-squared, Log likelihood, Aikaike info criterion and Schwarz criterion. The results of
alternative production functions fitted are provided in Appendix A. Based on the criteria, the
Cobb-Douglas production function was found to fit the data best and was chosen for the final
SFA specification.
61
4.3.1.1 Production Frontier Model Specification
Assuming a Cobb-Douglas production function, Equation 4.1 can be re-written as:
= + + − ( . )
where, lnYi represents logarithm of output (maize yield) measured in ton/ha for i observations
(i = 1, 2, 3, ... ,438). lnX refers to logarithm of inputs involved in the production function
including seed (kg), nitrogen (kg) and labour (man days) used in maize production. refers to
parameters of the production function to be estimated corresponding to the inputs used,
whereas Viand Ui are the error terms representing random effects and technical inefficiency
effects, respectively.
4.3.1.2 Variables Defining Production Frontier
In the estimation of the production frontier, the explanatory variables considered are seed,
nitrogen and labour. All the variables are measured on a per hectare basis. The variables
included in the production function, the measurement units and expected signs are given in
Table 4.1. The choice of the input variables is made on the basis that these inputs are
conventional inputs used in maize production by smallholder farmers in the country
(Gebreselassie, 2006; Spielman et al, 2011). Other conventional inputs such as pesticides and
herbicides were not included because the secondary data set used for the study does not
contain these variables.
Table 4.1: Definitions of variables included in the production function, measurement
units and expected signs
Variables Measurement unit Expected signs
X1= Seed Quantity of total seed  used (kg/ha) +
X2= Nitrogen Quantity of Nitrogen  used (kg/ha) +
X3= labour Quantity of labour used (man days/ha) +
62
Seed refers to the quantity of seed used in maize production measured in kg/ha. According to
Geta et al. (2010) and Idiong (2007), increased application of seed quantity results in
increased crop production. Based on these findings, the quantity of maize seed applied is
expected to increase maize production and the expected sign is positive. Nitrogen is another
important input applied in maize production which is also measured in kg/ha. Increased
application of Nitrogen increases maize productivity through enhancing the uptake and
concentration of Nitrogen and Phosphorus nutrients in maize tissues which are important for
plant growth (Abdulahi et al., 2006; IFPRI, 2010a).Among the empirical studies, Maseatile
(2011), Alene and Hassan (2003) and Geta et al. (2010) found that the increased application of
Nitrogen results in an increase in maize production. Based on these findings, the expected sign
of Nitrogen is positive. Another input considered is labour which refers to the total labour used
in maize production measured in man days/ha, during Meher season of 2011/12. Labour is an
aggregation of household labour and hired labour used in maize production by assuming that
each person works on average, eight hours per day on a maize farm. Among the empirical
studies, Fufa and Hassan (2003) and Bachewe (2009) found that with the use of additional
labour, crop production increases. This is due to farming activities such as land preparation,
planting, weeding and harvesting requiring availability of adequate labour for timely operation
of the activities (Geta et al., 2010). Based on the results, labour is expected to have an
increasing effect on production and the expected sign is positive.
It is important to note that phosphorus was also another input considered in maize production,
but it was excluded from the analysis because the variable was highly correlated with
Nitrogen. Producers use DAP as a source of fertiliser which contains both Nitrogen and
Phosphorus in fixed proportions. Correlation between Nitrogen and Phosphorus was therefore
expected. However, when the two variables are included together for estimation of the
production function, they resulted in biased parameter estimates. Although applications of
both nutrients are required to increase maize productivity, the application rate indicates that
more quantity of Nitrogen is required than Phosphorus in maize production due to the nature
of maize in nutrient utilisation (Plessis, 2003; Onasanya et al., 2009). Nitrogen is a vital
nutrient for plant and yield growth in maize production because it is a major component of
substances such as amino-acids and nucleic-acids which form the living plant tissue for growth
(Haynes, 1986: 24).It is also an integral component for many other compounds essential for
plant growth processes including chlorophyll and many enzymes (Onasanya et al., 2009). In
addition, application of Nitrogen enhances the uptake and concentrations of Nitrogen and
63
Phosphorus nutrients in the maize tissues more than Phosphorus (Abdulahi et al., 2006). By
considering the above reasons, Nitrogen was retained for analysis.
4.3.2 TECHNICAL INEFFICIENCY MODEL SPECIFICATION
In order to identify and analyse the determinants of technical efficiency, a technical
inefficiency model was employed. In the technical inefficiency model, the dependent variable
is the technical inefficiency variable (Ui) and the explanatory variables are the factors that are
hypothesised to affect technical inefficiency (Zi). A positive sign of a coefficient of a technical
inefficiency model parameter implies that the variable considered has an increasing effect on
technical inefficiency and vice versa. The implication of the relationship is that variables that
increase technical inefficiency will decrease technical efficiency. Given the distinction of
technical inefficiency and efficiency, the next section provides an empirical specification of
the technical inefficiency model.
The empirical specification of the technical inefficiency model of Battesse and Coelli(1995) in
Equation 4.3 takes the form:
= + + ( . )
where, Ui is the technical inefficiency variable derived from Equation 4.1 and is assumed to be
a function of farm specific socio-economic and farm management explanatory variables. The
farm specific explanatory variables that are proposed as determinants of technical inefficiency
are age, gender, education, household size, farm size, tenure, oxen, extension, irrigation,
credit, off-farm income, seed type, organic fertiliser and soil protection. The variable irefers
to a random variable which is assumed to be obtained by normal distribution with zero mean
and unknown variance confirming that Ui is non-negative and truncated at zero with the point
of truncation given by (-Ziδ) i.e., Zi.δ. The empirically specified production frontier and
technical inefficiency models are then estimated using a single-stage SFA approach. The
Maximum Likelihood estimates of all the parameters of the production frontier and the
technical inefficiency models, defined by Equations 4.4 and 4.5 are estimated simultaneously
using the FRONTIER 4.1 program. The Likelihood function is expressed in terms of the
64
variance parameterisation given by sigma square ( ) and gamma ( )upon which test for the
presence of technical inefficiency in an estimated model is based (Battesse & Coelli, 1993).
refers to the total model variance consisting of a va)riance due to random effects ( ) and avariance due to technical inefficiency effects ( which is parameterised as (Battesse &
Coelli, 1995):
= + ( . )
where, σ and σ represent variances accounted by random effects (Vi) and technical
inefficiency effects (Ui) in the model, respectively. The estimates ofvariance parameter are
mainly used to identify the parameter which represents the proportion of total model
variance that accounts for technical inefficiency. The parameterisation of given by Battesse
and Coelli (1995) takes the form:
= ( . )
The value of ranges between zero and one (Baruwa & Oke, 2012).When is zero, it
indicates that technical inefficiency effects are absent in the estimated SFA model and all
variation from the frontier is due to random noise which suggests the appropriateness of the
use of OLS than SFA technique for analysis. When is closer to one, the model indicates that
most of the variation of output from the frontier is accounted by technical inefficiency, which
suggests the presence of technical inefficiency in the model and confirms appropriateness of
SFA technique (Baruwa & Oke, 2012).
4.3.2.1 Variables Defining Technical Inefficiency Model
Fourteen farm specific explanatory variables are included in the technical inefficiency model.
Table 4.2 provides the variables with their respective measurement units and expected signs.
Age, gender and education of the household heads are variables to be discussed first. The
reason for considering age, gender and education of the household heads is due to the
assumption that household heads are responsible for farm households’ production decisions
which result in differences in the technical efficiency levels (Pattanyak et. al., 2003).Age of a
65
household head is measured in years and is included in the technical inefficiency model to
identify whether differences in age among the household heads contribute to differences in
technical inefficiency. Based on past empirical studies, age has a mixed effect on technical
inefficiency. In some studies, age has a decreasing effect (Fesessu, 2008; Derege, 2010) while
in others, it has an increasing effect on technical inefficiency (Makombe et al., 2011; Gbegeh
& Akubuilo, 2013). Thus, the expected sign of age of a household head can be either positive
or negative.
Similarly, gender of a household head is included in the model to determine if gender has a
significant impact on farm level technical inefficiency. Gender is a dummy variable that is
assigned a value of “1” for male-headed households and “0” for female-headed households.
Literature indicates that gender of household heads can have a negative or a positive effect on
farm technical inefficiency. For example, Bachewe (2009) found a negative effect of gender
on technical inefficiency which implies that male-headed households are less technically
inefficient. This is attributed to the fact that male-household heads have better access to
resources such as land, credit and technological inputs than female-headed households. In
addition, female household heads undertake normal household homemaker activities as well as
the usual farm work (AWM, 2009; OXFAM, 2012). Derege (2010) found a positive effect of
gender on farm technical inefficiency indicating that the female-headed households are more
technically efficient than the male-headed ones. Derege (2010) explained that the female-
headed household farms were more efficient due to the increased follow-ups and closer
supervision the female household heads made to the farms. Based on the findings, the
expected sign of gender can be either positive or negative.
66
Table 4.2: Description of variables included in the technical inefficiency model,
measurement units and expected signs
Expected
Variables Measurement unit
signs
Z1=Age Age of household heads in years -/+
Dummy variable = 1 if the household head is male, 0
Z2=Gender
otherwise -/+
Dummy variable = 1 if the household head has educational
Z3=Education
of grade four or above, 0 otherwise -/+
Z4=Household size Number of people living in a farm household (count) -/+
Z5=Farm size Farm size in hectare -/+
Dummy variable = 1 if the farm land is privately owned, 0
Z6 =Tenure
otherwise -
Dummy variable = 1 if the respondent has two or more
Z7=Oxen
oxen, 0 otherwise. -
Dummy variable = 1 if the farm is under extension service, 0
Z8=Extension
otherwise -
Z9=Irrigation
Dummy variable = 1 if the farm is irrigated, 0 otherwise -
Dummy variable = 1 if the respondents has received credit, 0
Z10=Credit
otherwise -
Dummy variable = 1 if the respondent received off-farm
Z11=Off-farm income
income, 0 otherwise -/+
Z12=Seed type Dummy variable = 1 if improved seed is used, 0 otherwise -
Z13=Organic fertiliser Dummy variable = 1 if organic fertiliser is used, 0 otherwise -
Dummy variable =1 if the respondent used soil protection
Z14=Soil protection
from erosion, 0 otherwise -
Education of a household head is included as a dummy variable where a value of “1” was
assigned to household heads that have completed education of grade four or above and “0” for
household heads with an education level of grade three or less. The classification is in
accordance to empirical literature by Appleton and Arsene (1996) that education of at least
grade four is needed to affect crop production. Most of the sampled farmers did not acquire
any formal education that this classification can provide the required information whether
education is a determinant of technical inefficiency in the sample. Empirical literature
indicates that education has a mixed effect on technical inefficiency. Education can increase
farmers’ ability to obtain, process and use information relevant to agricultural practices that
67
can decrease farm technical inefficiency (Jamison & Lau, 1982; Phillips, 1994; Weir, 1999).In
contrast, increased education can make farmers more inefficient in farming by shifting
educated labour into off-farm employment (Alemu et al., 2009). Based on the findings, the
expected sign of education is either positive or negative.
Another important variable used in technical inefficiency analysis is household size, which is
included as a continuous variable. The variable is used to determine if household size is an
important factor affecting technical inefficiency among the respondents. From literature,
household size can have either a decreasing or an increasing effect on technical inefficiency.
For example, Fesessu (2008) found a decreasing effect of family size on technical inefficiency
indicating that large households are important sources of family labour. While Derege (2010)
found an increasing effect of household size on farm technical inefficiency where large
household sizes increase farm technical inefficiency. Therefore, the expected sign of
household size can be either positive or negative. Farm size is another continuous variable of
the model which is measured in hectare. There are different economic arguments about farm
size and the associated technical inefficiency(Masterson, 2007).Some studies found an
increasing effect of farm size on technical inefficiency(Parikh et al., 1995; Masterson,
2007)while others found a decreasing effect of farm size on technical inefficiency (Bravo-
Uteta & Pinheiro, 1997, Huang & Kalirajan, 1997; Khaile, 2012). As a result, the expected
sign of farm size is either positive or negative. A closely related variable to farm size is farm
tenure.
Tenure is a dummy variable that refers to the type of farmland ownership and has a value of
“1” for private ownership and “0” for other form of ownership. The variable is included to
determine if type of tenure affects farm technical inefficiency among the respondents. Past
studies indicate that private ownership of land encourages sustainable farm practises by
creating an incentive to manage and invest in the lands (Alene & Hassan, 2003; Nega et al.,
2003). This indicates that private ownership of land decreases technical inefficiency. Based on
literature, private ownership of land is hypothesised to decrease farm technical inefficiency
and the expected sign is negative. Similar to tenure, the ownership of oxen is a dummy
variable that is assigned a value of “1” for farmers that own adequate numbers of oxen
(minimum of one pair of oxen) and “0” for farmers that did not own oxen during the
production season.
68
Interest in the oxen variable is due to the fact that oxen are the main source of draft power
used in smallholder farming in Ethiopia (Geta et al., 2010). Ownership of oxen enhances
productivity by augmenting labour input to accomplish farm operations such as land
preparation, sowing and fertiliser application timely. Gebreegziabher et al. (2004) and Geta et
al. (2010)found that ownership of oxen decreases technical inefficiency. Based on the
findings, ownership of adequate oxen is hypothesised to decrease farm technical inefficiency
and the expected sign is negative.
Access to extension, irrigation, credit and off-farm income is included as dummy variables in
the technical inefficiency model. Farmers that have access to extension services are assigned a
value of “1” and those without the service are assigned a value of “0”. Extension is expected
to have a negative effect on farm technical inefficiency because farmers that received
extension services would obtain advice and training about better farming operations (Alene &
Hassan, 2003; Fesessu, 2008). Irrigation can also be a source of variation for farm level
technical inefficiency. Irrigation is included as a dummy variable taking a value of “1” for
farmers that used irrigation and “0” for farmers that did not. Irrigation is a means of providing
sufficient water for plant growth to prevent water stress that can possibly reduce productivity
of the inputs used (Haise & Hagan, 1967). Therefore, irrigation is expected to have a
decreasing effect on farm technical inefficiency and the expected sign is negative. Similarly,
farmers that received credit are assigned a value of “1” and farmers that did not receive credit
are assigned a value of “0” to determine whether access to credit affects technical inefficiency
of the respondents. Obwona (2000) and Binam et al. (2004)indicated that access to credit
enabled farmers to finance the farming operations and is expected to have a decreasing effect
on technical inefficiency. Based on the findings, access to credit is expected to decrease farm
technical inefficiency and the expected sign is negative. In the same manner, off-farm income
is included as a dummy variable assigned by a value of “1” for maize farmers that received
off-farm income and “0” for maize farmers that did not receive off-farm income. Literature
indicates that off-farm income has a mixed effect on technical inefficiency. For example,
Gebreegriabher et al. (2004) and Haji (2008) found a decreasing effect of off-farm income on
technical inefficiency whereas McNally (2002), Goodwin and Mishra (2004) and Geta et al.
(2010) found an increasing effect of off-farm income on technical inefficiency. Therefore, the
expected sign of off-farm income is either positive or negative.
69
The final variables to be discussed in the technical inefficiency model are seed type, organic
fertiliser and soil protection. The type of seed used in maize production is included as a
dummy variable having a value of “1” for farmers that used improved seed and “0” for
farmers that used traditional seed. Farmers that used improved seed are expected to be more
productive than those using traditional seeds. Therefore, improved seed is expected to have a
negative effect on farm technical inefficiency and the expected sign is negative. Similarly,
application of organic fertiliser is included as a dummy variable with a value of “1”
representing respondents that applied organic fertiliser and “0” for respondents that did not use
organic fertiliser. Utilisation of organic fertiliser is expected to improve nutrient content of the
soils which in turn increases productivity hence decreasing technical inefficiency. From the
empirical findings, Tadesse and Abdissa (1996), Gruhn et al. (2000) and Ayalew and Dejene
(2011) found that application of organic fertiliser increased maize yields. Based on the
findings, the use of organic fertiliser decreases technical inefficiency and the expected sign is
negative.
The final variable of the model is soil protection from erosion which is a dummy variable
assigned a value of “1” for respondents that used soil protection and “0” for respondents that
did not. The variable is included to determine the effect of soil protection from erosion on
farm technical inefficiency. Soil erosion is a serious problem that causes loss of soil and
important soil nutrients and thereby increases technical inefficiency (Hurni, 1993). Utilisation
of soil protection from erosion is expected to increase productivity sustainably (IFPRI, 2010a).
From empirical studies, Geta et al. (2010) found that integrated soil management practices
decreased technical inefficiency. Based on the finding, soil protection is expected to decrease
technical inefficiency and the expected sign is negative.
4.3.2.2 Multicollinearity Problem
Given the description of the variables included in the SFA model, the final data prepared for
the SFA analysis consists of 438 observations. The first attempt to estimate the SFA was
unsuccessful due to the presence of high multicollinearity among the technical inefficiency
model explanatory variables. Multicollinearity refers to linear correlations among the
explanatory variables in the inefficiency model that can lead to biased parameter estimates
(Gujarati, 2003). Principal Component Regression (PCR) is one of the multivariate statistical
techniques that can solve problems of multicollinearity (Jouan-Rimbaud et al., 1995). The
70
PCR technique is used to solve the problem of multicollinearity detected among the
explanatory variables of the technical inefficiency model. The idea of PCR is to transform the
originally correlated explanatory variables into a new set of uncorrelated variable called
Principal Components (PCs) (Kline, 1994). The following section is a discussion of the
Principal Component Regression.
4.4 PRINCIPAL COMPONENT REGRESSION
In this section, the procedures and application of the Principal Component Regression (PCR)
are discussed. First, the theoretical basis of PCR is discussed followed by the empirical
application. When applying a PCR the first step is to standardise the original explanatory
variables that is used to fit the technical inefficiency model. The eigenvalues and eigenvectors
of the correlation coefficient matrix will also be obtained. This is followed by the estimation
of the PCs which is accomplished by multiplying the standardised variables with eigenvector
matrix of the correlation coefficient. The estimated PCs are then incorporated in a SFA
regression as the explanatory variables for the technical inefficiency model. The the
significance of the PCs are determined based on the SFA regression results. The significant
PCs are used to determine the significances of the standardised variables. Finally by using the
coefficients and significances of the standardised variables, the coefficients of the original
variables are determined. In the empirical application of the PCR, the same procedures are
pursued to determine significances of the originally correlated variables so that the problem of
multicollinearity is solved.
4.4.1 THEORETICAL BASIS OF PCR
Through the process of PCR, the dimensionality of the original data is reduced without loss of
information (Motsoari, 2012). The process starts by standardising the original explanatory
variables. Fo
= (
llow−ing F)ekedulegn et al. (2002), the variables are standardised as: ( . )
w̅ here, refers to the ith standardised variable, Zi refers to the original explanatory variables.and Szi represent the mean and standard deviation of the original explanatory variables,
respectively. After standardising the variables, the next procedure is estimation of PCs.
71
4.4.1.1 Estimation of Principal Components
Formulation of eigenvalues and eigenvector of the correlation coefficient matrix of the
correlated variables is also required to estimate PCs. Eigenvalues indicate the amount of
variances explained by each PC while eigenvectors are the weights used in a linear
transformation when computing PC scores (Kline, 1994). While estimating the PCs,
components of all the variables involved in the PCR are ranked in order of their importance
based on the eigenvalues of each PC in a decreasing order (Khaile, 2012). The procedure then
involves retaining PCs that have eigenvalues of greater than one whereas the PCs with
eigenvalues of less than one are excluded from further analysis (Khaile, 2012).
Given a correlation coefficient matrix (C) of the correlated variables that has a k x k
dimension, assuming 1,2,…,j be the eigenvalues and V=[V1, V2, …,Vj] be a matrix of
eigenvectors of the correlation coefficient matrix. Estimation PCs involve calculating the
eige−nvalues =by solving a determinant equation given by Draper and Smith (1981) as: ( . )
and calculating the associated eigenvectors (Vj) of the correlation matrix by solving a
determinant equation given by Fekedulegn et al. (2002) as:
−  = ( . )
The computed eigenvectors (V) are orthogonal to one another and provides a solution which
satisfies the condition that the product of an orthogonal vector with its transpose equals 1
(V Ti Vi = 1) and given two vectors Vi and Vj, the≠y are orthogonal if and only if the innerproducts of the two vectors is zero (Vi’Vj=0) for i j (Fekedulegn et al., 2002).The computed
eigenvectors are arranged to form a matrix of eigenvectors as:
V= VVV⋮ V
V
V⋮
………⋮ V
V
V⋮ (4.11)
72
Once the eigenvalues and eigenvectors of the correlation matrix are formulated, the PCs are
estimated by multiplying the standardised explanatory variables given by Equation 4.8 with
the matrix of eigenvectors given in Equation 4.11 (Fekedulegn et al., 2002). Therefore, the
estimation of the PCs represented as P is given by multiplying the matrix of the standardised
variables with the matrix of eigenvector based on Fekedulegn et al. (2002) as:
= ( . )
where, Zs refers to a matrix of standardisedexplanatory variables and V is a matrix of
eigenvectors (Fekedulegn et al., 2002). Equation 4.12 can be re-written in matrix form as:
PP⋮ PP⋮ …P P ……⋮ PP
P⋮ ……= ⋮ ⋮ …⋮ ⋮ V
V⋮ VVV V⋮
…
……⋮ V
V
x V⋮ (4.13)
The estimated PCs contain the same number of information as the original variables except
that the PCs are uncorrelated and can be ranked by the magnitude of their eigenvalues (Draper
& Smith, 1981). Starting with n number of correlated original explanatory variables, the same
n number of PCs can be obtained through estimation. However, only PCs that can account for
the majority of variances among the variables are retained (Kline, 1994). Therefore, only the
PCs that are associated with eigenvalues of greater than one will be retained (Loehlin, 1992).
4.4.1.2 Regression with Principal Components
The significance of the retained PCs will be estimated by SFA regression by involving the PCs
instead of the originally correlated explanatory variables in the technical inefficiency model of
the SFA (Magingxa et al., 2006).The technical inefficiency model of SFA that involves the
retained PCs can be specified as:
= + + ( . )
73
where, P represents a matrix of the PCs obtained by Equation 4.12; represents amatrix of
coefficients associated with l number of retained PCs (Magingxa et al., 2006). Standard errors
of the coefficients of the PCs are computed from the corresponding variances of the
coefficients of the estim)ated PCs (Magingxa et al., 2006). Variances of the estimatedcoefficients of the PCs ( is formulatedbased on Fekedulegn et al. (2002) as:
( ) = ( ) = ( , …, ) ( . )
where, is variance of residuals obtained from the regression given in Equation 4.14.
Standard errors of coefficients of the retained PCs are given by:
( ) = ( . , . , … . ) ( . )
where, std. ( ) represent estimates of standard errors for respective coefficients. With the
use of the estimated coefficients and standard errors and the associated t-ratios, the
significances of the PCs are determined (Fekedulegn et al., 2002). The significant PCs are
used to determine significances of the standardised variables while the insignificant PCs are
eliminated (Magingxa et al., 2006). After the elimination of the insignificant PCs, Equation
4.14 is adjusted to the number of significant PCs (Khaile, 2012). Following Khaile (2012),
given l number of retained PCs, assuming elimination of r number of insignificant PCs,
Equation 4.14 is re-written as:
= + + ( . )
where, and are used in Equation 4.17 to differentiate the intercept and residual terms
pertaining to regression with the significant PCs from the ones used in retained PCs (Equation
4.14) of the SFA technical inefficiency model (Khaile, 2012). The next procedure is
transformation of the PCs to the standardised explanatory variables and determining their
significances (Fekedulegn et al., 2002).
74
4.4.1.3 Determining the Significances of the standardised Variables
Using Significant PCs
By using the significant PCs, the coefficients and standard errors of the standardised variables
will be determined. The estimated coefficients of the significant PCs will be used to determine
the coefficients of the standardised variables (Maseatile, 2011). Similarly, standard errors of
the PCs are used to estimate the associated variances of the PCs to be used in computing the
variances of the standardised variables (Maseatile, 2011). The coefficients of the PCs are
transformed back to the standardised explanatory variables as: , =Vl-r l-r which is written
in matrix form as (Magingxa et al., 2006):
⋮,, ……, = ⋮ ⋮ …⋮ ⋮ x ⋮ (4.18)
where, Vl-r is an eigenvector matrix of the correlation coefficient; l-r is a vector of coefficients
of the significant PCs estimated using Equation 4.17 (Khaile, 2012). Once the coefficients of
the standardised variables are determined, the associated variances of the coefficients will be
determined by using the PC variance estimators. Based on the procedure of Magingxa et al.
(2006), PC estimator of variances of the standardised variables is given as:
= ( . )
where represents the square of eigenvector elements obtained by squaring the eigenvector
elements given by Equation 4.11. refers to a matrix of squared standard error elements of
the coefficients of the significant PCs given in Equation 4.17 (Khaile, 2011). Furthermore, the
standard errors of the PC estimators of the standardised variables are obtained by the square
root of the variance estimators of the standardised variables which is given as:
= / ( . )
75
The estimated coefficients and the standard errors of the standardised variables are used to
estimate the coefficients and standard errors associated with the original variables (Maseatile,
2011). Following the proce,dures of Fekedulegn et al. (2002), the estimated coefficients of thestandardised variables, are transformed back to the original unstandardised coefficients
as:
, = , , = , , … ( . )
and
, = , − − , …− , ( . )
Where Sai refers to standard deviations of the unstandardised variables and b
s , bs so,pc 1,pc, b 2,pc,…
bsk,pc represent coefficients of the standardised variables. The coefficients of unstandardised
variables can 1be computed when standard deviations (Sai) of the unstandardised variables arecalculated by (Khaile, 2012).
4.4.2 APPLICATION OF PCR
Given the theory of PCR discussed above, this section discusses the application of the PCR in
SFA to solve the problem of multicollinearity observed among the correlated technical
inefficiency model variables. During the first estimation of the SFA it was noted that
correlation between variables in the technical inefficiency model result in biased results. The
14 variables for the technical inefficiency model were therefore standardised and before a
Principal Component Analysis were used to identify uncorrelated components that can be
included in the SFA. The PC’s were included in the SFA regression as the explanatory
variables in the second stage model. The use of Frontier 4.1 allows the researcher the
opportunity to estimate the production function and the technical inefficiency model as
simultaneous equations to overcome a two-stage estimation procedure. The significant PC’s
were identified based on the SFA regression results. From the regression output, it is found
that all the retained PCs are statistically significant and the PCs are then used to estimate the
significances of the standardised variable. The coefficients of the standardised variables are
76
finally transformed back to the original unstandardised variables. The procedures pursued are
discussed in the next sub-sections.
4.4.2.1 Estimation of Principal Components
Estimation of PCs is based on the formulation of eigenvector matrix of the correlation
coefficient of the variables and the standardised variables. Therefore, the 14 technical
inefficiency model variables are standardised based on the formula given by Equation 4.8. The
eigenvalues and eigenvectors of the correlation coefficient matrices are obtained by using
Statistical Package for Social Science (SPSS). The PCs are then estimated by multiplying the
standardised variables with the eigenvector matrix of the correlation coefficient. Only five PCs
with eigenvalues of greater than one were retained as they account for the majority of the
variances among the correlated variables. Table 4.3 provides the five retained PCs with their
respective eigenvalues and percentage of variances explained.
Table 4.3: Principal Components estimated, eigenvalues and cumulative percentage
of variances explained
Principal Components Eigenvalues Percentage of variances
explained
PC1 1.78 13.93
PC2 1.51 13.23
PC3 1.29 12.78
PC4 1.18 11.89
PC5 1.06 10.07
Total 6.82 61.90
As indicated in Table 4.3, each PCs has an eigenvalue of greater than one. The result indicates
that the five PCs explained about 61.9% of the variation within the originally correlated
variables. The result fulfilled the goal of determining a reduced number of PCs that can
explain most of the total variance. Once the important PCs are estimated, the next step is
regression with the five retained PCs to determine the significances of the PCs.
77
4.4.2.2 Determining Significance of the Principal Components
Significances of the retained PCs were determined by the SFA regression in which the PCs
were incorporated in the technical inefficiency model. The results of the parameters of the PCs
and respective statistics are given in Table 4.4.
Table 4.4: Maximum Likelihood Estimates of the retained PCs
Variables Parameter Coefficients of Std-errors t- ratios Probability
the PCs
Constant α -9.2944 3.3246 -2.7956 0.005***o
PC1 α -0.9789 0.3491 -2.8040 0.005***1
PC2 α2 -0.9667 0.3388 -2.8532 0.005***
PC3 α3 -0.5474 0.1916 -2.8571 0.005***
PC4 α4 -0.2221 0.1246 -1.7836 0.075*
PC5 α5 0.8766 0.3023 2.8993 0.004***
*** refers to significant at 1%, * refers to significant at 10%.
The SFA estimation result given in Table 4.4 indicates that all the five PCs are statistically
significant. Four of the five retained PCs are significant at a one percent level while PC4 is
significant at a 10% level. Since all PCs are significant, the estimated parameters of the PCs
are used to determine the significances of the correlated explanatory variables. The
coefficients of the PCs are used to estimate the coefficients of the standardised explanatory
variables. This is accomplished by multiplying the coefficients of the PCs with the
eigenvectors of the correlation coefficient matrix.
The standard errors of the PCs are used to estimate variances of the standardised variables.
The procedures of variance estimation include first estimating the variance matrix of the PCs
by squaring the standard errors of the PCs. Then, variances of the standardised variables are
estimated by multiplying the variance matrix of the PCs with the squared eigenvector matrix.
The standard errors of the standardised variables are then obtained from the square root of the
variance matrix of the standardised variables. After the coefficients and the standard errors of
the standardised variables are determined, the corresponding t-ratios are obtained by dividing
the estimated coefficients to their respective standard errors. The significances of the variables
78
are then determined by using the t-distribution of the variables. The coefficients of the
estimated standardised variables are transformed back to the original unstandardised
coefficients by dividing the standardised variables by standard deviations of the
unstandardised variables. Finally, the MLE result of all the variables of the technical
inefficiency model after transformation in to the original variables are presented and discussed
in Chapter five.
4.5 HYPOTHESIS TESTING
SFA allows various hypotheses tests including tests for the presence of technical inefficiency
and tests to determine whether the technical inefficiency effects are stochastic (Belloumi &
Matoussi, 2006). For hypothesis testing, the generalised Likelihood Ratio is used which is
defined by:
= − { [ ( )] − [ ( )]}( . )
Where Ln [L (H0)] and Ln [L (H1)] are the values of Log Likelihood function for the frontier
model under the null hypothesis and alternative hypothesis, respectively (Battesse & Coelli,
1995). The generalised Likelihood Ratio ( )hasapproximately chi-square or a mixture of chi-
square distribution with degrees of freedom determined by the number of parameters assumed
to be equal to zero in the null hypothesis, H0 (Battesse & Coelli, 1995). The decision to accept
or reject the null hypothesis is based on the comparison of λ against the chi-square critical
value. If λ is greater than the critical value, the null hypothesis is rejected, but if λ is less than
the critical value, the null hypothesis is accepted (Battesse & Coelli, 1995).
Two hypotheses were tested, one is to test for the presence of technical inefficiency effects in
the SFA model and the other is to determine if the technical inefficiency effects are stochastic.
The hypotheses tested and the final decisions are presented in Table 4.5.
The first hypothesis is used to determine the presence of technical inefficiency effects in the
SFA model. Information about the presence of technical inefficiency effects in an estimated
model is important to determine whether the use of the SFA with the technical inefficiency
model is appropriate.
79
Table 4.5: Generalised Log likelihood tests of hypotheses
Null Hypothesis (H0) Log likelihood LR test Critical Decision
function Value withstatistic () (2 0.05)
 = δ0 = δ1 = δ2 = … δ8 -74.59 297.97 17.67* Reject H0
 = 0 -114.69 217.76 5.14* Reject H0
*The critical values for both hypotheses involving γ = 0 are obtained from Kodde and Palm (1986), with degrees
of freedom equal to 10 for the first hypothesis and degrees of freedom equal to 2 for the second hypothesis based
on Belloumi and Matoussi (2006)
In doing so, a null hypothesis was imposed with a restriction implying that technical
in)efficiency effects are absent from the SFA model. The restrictions involve both the gamma( parameter and technical inefficiency model parameters are jointly zero (Alene & Hassan,
2003). The null hypothesis can be written out as given by Alene and Hassan (2003):
:  = = = = = ⋯ = = ( . )
The decision rule to accept or reject the null hypothesis is determined by comparing the
calculated  against the chi-square critical value. The calculated value of  is 297.97 whereas
the critical value is 17.67. Since the value of is greater than the critical value, the null
hypothesis is rejected and the alternative hypothesis is accepted. The rejection of the null
hypothesis confirms the presence of technical inefficiency effects and the appropriateness of
using the SFA with the technical inefficiency model (Alene & Hassan, 2003).
The second hypothesis test was used to determine whether the technical inefficiency effects
are stochastic. Information about the stochastic nature of the technical inefficiency effects is
important to determine whether the SFA is an appropriate model of efficiency analysis. To test
the second hypothesis, a null hypothesis that states the technical inefficiency effects are not
stochastic is imposed with restriction that  is equal to zero (Hassan & Ahmad, 2005). The null
hypothesis is written out following Hassan and Ahmad (2005) as:
:  = ( . )
If the gamma parameter is equal to zero, then the variance of the technical inefficiency effects
is zero so that the SFA model is reduced to the traditional mean response function (Hassan &
80
Ahmad, 2005). The decision rule to accept or reject the null hypothesis is by comparing the
calculated value of against the chi-square critical value following the same procedure as the
first hypothesis testing. The calculated value of is 217.76 while the critical value is 5.14.
Since the value of  is greater than the critical value, the null hypothesis stating that technical
inefficiency effects are not stochastic is rejected. The rejection of the null hypothesis implies
that technical inefficiency effects are stochastic which supports the appropriateness of the use
of the SFA model. In conclusion, the hypothesis tests indicated the presence of technical
inefficiency effects among the maize farmers and the appropriateness of fitting the SFA model
which includes the technical inefficiency model.
4.6 SUMMARY
The chapter discusses the procedures used to achieve the objective of the study. The study
employs a SFA model which requires the use of FRONTIER to estimate technical efficiency
and to analyse the determinants of technical inefficiency (Coelli, 1996). Assuming that
producers are producing a single output using multiple inputs, SFA provides the relative
frontier against which production performance is evaluated (Kumbhakar & Lovell, 2000: 63).
Different functional forms such as linear, Cobb-Douglas and translog can be used to specify
the production frontier (Coelli et al., 2005: 211). The functional forms are compared using the
data and model evaluation criteria such as R-squared, Log likelihood, Aikaike info criterion
and Schwarz criterion. Based on the evaluation, Cobb-Douglas production function is found to
fit the data well and is chosen for the final SFA specification.
In the production function estimated, the dependent variable is maize yield (ton/ha), and the
explanatory variables are seed (kg), nitrogen (kg) and labour (man days) used in maize
production. In the technical inefficiency model, the farm specific explanatory variables are
age, gender, education, household size, farm size, tenure, oxen, extension, irrigation, credit,
off-farm income, seed type, organic fertiliser and soil protection. The first attempt to estimate
the SFA was unsuccessful due to the presence of severe multicollinearity among the technical
inefficiency model explanatory variables. Principal Component Regression (PCR) is a
multivariate statistical techniques used to solve the problem of multicollinearity. The PCR
transforms the correlated technical inefficiency model explanatory variables into a new set of
uncorrelated Principal Components (PCs) (Kline, 1994). The PCs are then used in SFA
regression and their significances are determined. Five PCs are extracted that explained about
81
61.9% of the variation within the originally correlated variables. All the five PCs are
statistically significant and are used to determine the significances of the correlated
explanatory variables.
SFA allows various hypotheses tests by using the generalised Likelihood Ratio against the
Chi-square distribution (Battesse & Coelli, 1995). Two hypotheses were tested where the first
one is the test for the presence of technical inefficiency effects in the estimated model and the
other is the test to determine if the technical inefficiency effects are stochastic. The hypotheses
tests confirmed the presence of technical inefficiency effects and the technical inefficiency
effects are stochastic supporting the appropriateness of using of the SFA model.
82
CHAPTER FIVE: RESULTS AND DISCUSSION
Chapter five presents and discusses the empirical results of the SFA. The empirical results are
obtained from the FRONTIER software used to fit a production frontier (Coelli, 1996). The
chapter is organised into three main sections. In the first section, the estimated parameters of
production function are presented and discussed. The second section presents and discusses
the technical efficiency estimation results. In the third section, the results of the technical
inefficiency model are presented and discussed. Although a single stage SFA approach is used
to generate the results of the production function and the technical inefficiency models, for
ease of discussion, the two sets of variables are discussed in separate sections.
5.1 ANALYSIS OF PARAMETERS OF THE STOCHASTIC
PRODUCTION FRONTIER
This section is used to present and discuss the estimated parameters of the Cobb-Douglas
stochastic production frontier. The discussion includes the estimated statistics of input
parameters used in maize production including seed, nitrogen and labour as well as the
diagnostic statistics indicating model variance and gamma. Table 5.1 presents the Maximum
Likelihood estimates of the parameters of the Cobb-Douglas production frontier.
Table 5.1: Maximum Likelihood estimates of stochastic production frontier
parameters
Variable Parameter Coefficient Std. error t-ratio Probability
Constant βo 0.6912 0.0472 14.6544 0.0000***
Ln (Seed) β1 0.0166 0.0324 0.5122 0.6088
Ln (Nitrogen) β 0.0234 0.0054 4.3122 0.0000***2
Ln (Labour) β3 -0.0278 0.0197 -1.4170 0.1572
Diagnostic Statistics
Sigma-Square 2=2u +2v 2.7664 0.8573 3.2267 0.0013***
Gamma =2 2u/ 0.9931 0.0024 409.1748 0.0000***
LR function -74.59
LR test 297.97
Mean Technical Efficiency 0.77
*** represents significant at 1%.
83
The estimated coefficients of the production frontier parameters are positive for seed and
nitrogen, but negative for labour. The positive coefficients of seed and nitrogen indicate that
increasing the use of seed and nitrogen increases maize production. The increasing effect of
seed and nitrogen on maize production is as expected because production of maize is
dependent on the quantities of these inputs used. However, only the coefficient of nitrogen is
found statistically significant at a one percent level in explaining maize production while seed
is found insignificant at all levels. The increasing effect of nitrogen on maize production is in
line with the findings of Alene and Hassan (2003), Rahman and Umar (2009) and Geta et al.
(2010) in which application of fertiliser increases production significantly. Unexpectedly, the
negative sign of the labour coefficient indicates that the use of additional labour will decrease
maize production, but the relationship is statistically insignificant.
Along with the stochastic production frontier parameters, variance parameters of the model are
also estimated and given in Table 5.1. Sigma square (2) represents the variance of the
composed error term which consists of the variance due to random effects (2v) and the
variance due to technical inefficiency effects (2u) (Battese & Corra, 1977). The estimated
value of 2 is 2.06 and is significant at a one percent level, which confirms the correctness of
the specified distributional assumptions of the error terms. The other variance parameter
estimated is gamma ().The  parameter is used to show the proportion of total variance that is
attributed to technical inefficiency in the estimated model. The value of  is 0.99 and is
significant at a one percent level. The magnitude of  implies that 99% of the random variation
in output of maize production is attributed to the technical inefficiency component which
indicates the importance of capturing technical inefficiency in production. The significance of
 indicates that technical inefficiency effects are significant in determining the level and
variability of maize production. From the estimated value of , it is observed that only about
one percent of random variation in maize production is attributed to random shocks that are
out of the control of the farmers.
The estimated production frontier leaves much to be desired with very few explanatory
variables and low level of significance for the explanatory variables. The data used for the
estimation of the SFA is from secondary data obtained from the Central Statistical Agency of
Ethiopia. The data formed part of the Ethiopian Rural Socio-economic Survey (ERSS). The
84
data available to the study was therefore limited and hence limited the researcher in
explanatory variable that could be included in the production function.
5.2 TECHNICAL EFFICIENCY ESTIMATIONS
This section presents and discusses the SFA estimation of technical efficiency scores. As is
indicated in Table 5.1, the estimated mean technical efficiency score for the sample is 77%.
The mean technical efficiency of 77% indicates that on average, the sampled maize farmers
are able to obtain only 77% of potential output from the given mix of production inputs. The
finding suggests the presence of considerable level of technical inefficiency among the studied
households. The mean technical efficiency estimated provides an indication that the sampled
maize farmers have a potential of increasing their output by about 23% by using the existing
resources and technology more efficiently. The cumulative probability distribution of the
estimated technical efficiency scores of the sample is summarised in Figure 5.1.
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Technical Efficiency Scores
Figure 5.1: Cumulative probability distribution of technical efficiency scores for
sample of smallholder maize farmers in Ethiopia
Figure 5.1 indicates that the estimated technical efficiency scores range from 3 to 96%. The
cumulative probability distribution of the efficiency scores indicates that the lower 10% of the
85
Cummulative Probability
sample is associated with technical efficiency of less than 45%. About 50% of the sample has
technical efficiency scores of less than 70%. It is also observed that nearly 81% of the sample
has mean technical efficiency scores of less than 80%. However, the final 19% of the sample
has technical efficiency scores ranging from 80 to 96% which is the highest efficiency
category. More than half of the sampled farmers are operating with a technical efficiency score
of less than the sample mean.
The estimated mean technical efficiency is closely comparable to the technical efficiency
estimates of Alene and Hassan (2003) where the reported mean technical efficiency of
sampled maize farmers was 76% in Western Ethiopia. The other closely comparable finding
was that of Arega and Zeller (2005) where the mean technical efficiency score reported was
79% for a sample of maize, wheat and barley farmers in Eastern Ethiopia.
5.3 ANALYSIS OF TECHNICAL INEFFICIENCY MODEL
PARAMETERS
The hypothesis testing confirmed that technical inefficiency effects are present in the SFA
model. Given the presence of technical inefficiency, there is a need to improve technical
efficiency of the farmers by examining the determinants of technical inefficiency. The aim of
this section is therefore to present and discuss the technical inefficiency model parameters of
the smallholder maize farmers.
The estimated coefficients of the technical inefficiency model parameters are of particular
interest in examining technical inefficiency model parameters. In the technical inefficiency
model, the dependent variable is a technical inefficiency variable (Ui) and the explanatory
variables are farm specific socio-economic and farm management explanatory variables (Zi).A
variable that has a positive parameter estimate will have an increasing effect on farm technical
inefficiency and vice versa. The implication is that the variable that has an increasing effect on
technical inefficiency will have a decreasing effect on technical efficiency and vice versa.
Table 5.2 presents a summary of the SFA results of parameters of the technical inefficiency
model and respective statistics. Of the total technical inefficiency model explanatory variables,
the estimated coefficients of age, gender, household size, oxen, extension, irrigation, credit,
seed type and soil protection are negative and statistically significant. These variables were
86
found as having decreasing effects on technical inefficiency hence increasing effects on
technical efficiency of the maize farmers. The estimated coefficient of off-farm income is
positive and statistically significant indicating that off-farm income increases technical
inefficiency of the maize farmers. The remaining variables: education, tenure, farm size and
organic fertiliser are not statistically significant determinants of technical inefficiency. The
variables that are found significant in determining technical inefficiency are discussed as
socio-economic and farm management variables.
Table 5.2: Maximum Likelihood estimation results of technical inefficiency model
variables1
Variable Parameter Coefficient Std-errors t-ratios Probability
Constant αo -9.2944 3.3246 -2.7956 0.0054***
Age δ1 -0.5988 0.15407 -3.88689 0.0001***
Gender δ2 -0.7318 0.22999 -3.18200 0.0016***
Education δ3 0.1266 0.14430 0.87770 0.3806
Household size δ4 -0.75049 0.26590 -2.82241 0.0049***
Farm size δ5 -0.11697 0.88783 -0.13175 0.8952
Tenure δ6 -0.69203 0.14291 -4.84257 1.7872
Oxen δ7 -0.62774 0.25054 -2.50554 0.0126**
Extension δ8 -0.90109 0.29507 -3.05382 0.0024***
Irrigation δ9 -0.43334 0.26620 -1.62789 0.0030***
Credit δ10 -0.47581 0.14696 -3.23771 0.0013***
Off-farm income δ11 0.79963 0.31460 2.54177 0.0114**
Seed type δ12 -0.72495 0.28263 -2.56504 0.0111**
Organic fertiliser δ13 -0.20670 0.12732 -1.62352 0.1052
Soil protection δ14 -1.62213 0.79844 -2.03164 0.0428**
*** refers to significant at 1%, ** refers to significant at 5%
1 The results presented in Table 5.2 are the individual significance of the explanatory variables
included in the inefficiency model estimated within the SFA/PCR framework.
87
5.3.1 SOCIO-ECONOMIC VARIABLES
5.3.1.1 Age and Gender
From Table 5.2, the estimated coefficient of age is negative and highly significant. The
relationship implies that whenever age of household heads increases, farm technical
inefficiency will decrease. A possible reason for such a relationship can be due to the fact that
with increased age, farmers can have better resources at their disposal and can be better aware
of mechanisms for risk coping that is developed from life experience, which can potentially
decrease their technical inefficiency (Fesessu, 2008).The finding is consistent with the results
obtained by Fesessu (2008), Maseatile (2011) and Khaile (2012).
Similarly, the estimated coefficient of gender is negative and highly significant on affecting
technical inefficiency, suggesting that male headed households are more technically efficient
than female-headed households. A possible explanation for the finding is that male headed
households have better access to farm resources such as land, credit, technological inputs and
other supportive services than their female counterparts in Ethiopia, which makes the male
headed households more efficient (FARA, 2009; AWM, 2009). Another possible explanation
is that female household heads are responsible for other household activities (such as child
bearing and care, cooking and cleaning) that compete with the time and effort allocated to the
farm work, making them less efficient (Solis et al., 2008). The result is consistent with the
findings reported by Solis et al. (2008), Bachewe (2009) and Maseatile (2011). Analysis of
age and gender of household heads revealed that farm technical inefficiency will decrease
whenever the age of household heads increase and male headed households are less
technically inefficient than female headed households.
5.3.1.2 Household Size and Oxen
The estimated coefficient for household size is negative and highly significant. The finding
suggests that whenever household sizes increase, technical inefficiency will decrease. A
possible reason for a decreasing effect of household size on technical inefficiency is mainly
attributed to the fact that large households contribute to family labour to assist in the farming
activities such as sowing, weeding and harvesting. This is particularly true whenever most of
the household members contribute to the farming activities. The result is consistent with the
findings of Bachewe (2009), Geta et al. (2010) and Rahman and Umar (2009).
88
Another important variable is ownership of an adequate number of oxen which was included
to determine whether ownership of oxen is a source of technical inefficiency variation. The
estimated coefficient of oxen is negative and significant at a five percent level. The finding
suggests that ownership of oxen decreases technical inefficiency. The result is as expected and
confirmed the importance of ownership of an adequate number of oxen to augment labour
input and enhance productivity by reducing the time needed to accomplish farm operations
such as land preparation and sowing. The finding is consistent with the finding of Geta et al.
(2010) where ownership of oxen is found to be an important determinant of technical
inefficiency in production. The empirical analysis of household size and oxen suggested that
both variables have decreasing effects on technical inefficiency.
5.3.1.3 Credit and Off-farm Income
The estimated coefficient of credit is negative and highly significant which indicates the
importance of credit in decreasing farm technical inefficiency. Access to credit can decrease
technical inefficiency of the farmers through enabling them to make necessary short and long-
term farm investments (IFPRI, 2010b). The finding is as expected and is in line with the
findings of Alene and Hassan (2003), Obwona (2006) and Maseatile (2011). Despite the
important role credit has in smallholder maize production, it was observed that about 71% of
the farm households did not receive credit. The low level of credit participation of the farm
households could be mainly due to low level of credit coverage, which can also be attached to
the lack of loanable funds by the suppliers of rural credit. Other problems associated with
credit are complicated credit application procedures that the farmers undergo to obtain credit
and the requirements of high collateral in the country (Ameha, 2011).Off-farm income is a
dummy variable that is proposed as a determinant of technical inefficiency. The estimated
coefficient of off-farm income is positive and statistically significant at a five percent level.
The result suggests that with an increase in pursuit of off-farm income, technical inefficiency
will increase. The possible explanations for an increasing effect of off-farm income on maize
farm technical inefficiency could be that the off-farm income received might not be used for
financing the farming activities and the farmers might spent much of their time working off
the farm and failing to manage the maize farms properly. The finding supports the fact that
increased off-farm income opportunities will reduce farm resources and farmers’ effort that
could otherwise be used for farming (McNally, 2002; Goodwin & Mishra, 2004).The result is
consistent with the findings reported by Alene and Hassan (2003), Obwona (2006) and
89
Baruwa and Oke (2012). The analysis of credit and off-farm income revealed that credit
decreases farm technical inefficiency whereas off-farm income increases farm technical
inefficiency in the sample.
5.3.2 FARM MANAGEMENT PRACTISES
The estimated coefficients of the farm management variables of the technical inefficiency
model are discussed in this section. The variables discussed include extension, irrigation, seed
types, organic fertiliser and soil protection.
5.3.2.1 Extension and Irrigation
Extension is a dummy variable that was hypothesised to influence technical inefficiency of
maize production negatively. The empirical result indicates that extension has a negative and
highly significant coefficient. Extension therefore decreases technical inefficiency as expected.
The possible reason for the result is that farmers who obtained extension services can have
better information about farm management practises and better agricultural technologies. The
finding is supported by Obwona (2006) and Bachewe (2009) who acknowledged the
importance of extension services as a key policy instrument to improve agricultural
productivity. Despite the fact that most of the farmers studied (79%) do not receive extension
services, the study revealed that extension services are an important determinant of technical
inefficiency.
Irrigation is another important dummy variable proposed to affect technical inefficiency
negatively. The estimated coefficient of irrigation is negative and significant at a five percent
level. The negative coefficient of irrigation suggests that the use of irrigation decreases
technical inefficiency and the relationship is expected. Although most of the farmers (96%) in
the sample did not use irrigation, the empirical findings confirmed that the use of irrigation
contributes to a reduction of technical inefficiency significantly. From the distribution of the
estimated technical efficiency, it was observed that the irrigated farms were much more
efficient than the non-irrigated farms. The average technical efficiency score specific to the
irrigated farms is 85% which is much greater than the average technical efficiency for the non-
irrigated farms, which is 77%. The findings for irrigation is in line with the results reported by
Makombe et al. (2011) and Khai and Yabe (2011) that supported the importance of irrigation
90
in increasing crop productivity of smallholder farmers. The low usage of irrigation in Ethiopia
is mainly due to a lack of development of irrigation infrastructure. The empirical analysis of
extension and irrigation services indicated that both variables have decreasing effects on farm
technical inefficiency.
5.3.2.2 Seed Type and Soil Protection
Given the fact that smallholder maize farmers use either improved seed or traditional seed, the
variable of seed type is included as a dummy variable to determine if the use of improved seed
decreases technical inefficiency. The empirical results indicate that improved seed has a
negative and significant coefficient at a five percent level. The finding is in line with the initial
hypothesis and the use of improved seed indeed has a decreasing effect on technical
inefficiency. The finding is consistent with the empirical studies of Geta et al. (2010) and
Maseatile (2011) where the use of improved seed increases technical efficiency of maize
production.
Soil protection is the final dummy variable which was hypothesised to decrease technical
inefficiency. The estimated coefficient of soil protection is negative and statistically
significant at a five percent level. The finding is in line with the initial hypothesis and
confirms that soil protection can significantly increase technical efficiency in maize
production in the country. The average technical efficiency for the farmers that used soil
protection is nearly 83% which is greater than the average technical efficiency of the total
sample (77%). The result of soil protection is consistent with the finding reported by Geta et
al. (2010) where soil and land management practises decrease technical inefficiency
significantly. The findings of seed type and soil protection confirmed that the two variables are
important factors that have decreasing effects on technical inefficiency.
5.4 SUMMARY
Chapter five presented the empirical findings of the study. The chapter contained three main
sections which include presentation and discussion of the estimated parameters of production
function, the estimated results of the technical efficiency and the technical inefficiency model
parameters in Ethiopian maize production. The inputs defining the production frontier in the
study are seed, nitrogen and labour. The estimated coefficients are positive for seed and
91
nitrogen, and negative for labour, though only nitrogen is found statistically significant. The
positive coefficients of seed and nitrogen indicate that increasing the use of seed and nitrogen
increases maize production. The negative sign of the labour coefficient indicates that the use
of additional labour decreases maize production, but the relationship is statistically
insignificant and the finding is unexpected. Along with the stochastic production frontier
parameters, variance parameters of the model are also estimated. Sigma square (2) represents
the variance of the composed error term (Battese & Corra, 1977). The estimated value of 2 is
2.06 and is highly significant, confirming the correctness of the specified distributional
assumptions of the error terms. The other variance parameter estimated is gamma (), which
shows the proportion of the total variance that is attributed to technical inefficiency in the
model. The value of  is 0.99 and is highly significant implying that 99% of the random
variation in output of maize production is attributed to the technical inefficiency. The
significance of  indicates that technical inefficiency effects are significant in determining the
level and variability of maize production.
The estimated results of the technical efficiency indicate that the mean technical efficiency
score is 77% for the sample. The mean technical efficiency of 77% implies that the sampled
maize farmers are able to obtain only 77% of potential output from the given mix of
production inputs, on average. The finding suggests the presence of considerable level of
technical inefficiency that the sampled maize farmers have a potential of increasing their
output by increasing their technical efficiency by about 23%. The estimated mean technical
efficiency is closely comparable to the technical efficiency estimates of Alene and Hassan
(2003) and Arega and Zeller (2005).Given the presence of technical inefficiency in the study,
maize production, and the study analysed the determinants of technical inefficiency.
Among the technical inefficiency model explanatory variables involved in the study, the
estimated coefficients of age, gender, household size, oxen, extension, irrigation, credit, seed
type and soil protection are negative and statistically significant. The variables were found
having decreasing effects on technical inefficiency hence increasing effects on technical
efficiency of the maize farmers. The estimated coefficient of off-farm income is positive and
statistically significant indicating that off-farm income increases technical inefficiency of the
maize farmers. The remaining variables: education, tenure, farm size and organic fertiliser are
not statistically significant determinants of technical inefficiency.
92
CHAPTER SIX: SUMMARY, CONCLUSION AND
IMPLICATIONS
Chapter six provides a summary, conclusion and the implications of the research. In the first
part of the chapter a summary is presented of the introduction, literature review, study area and
data, methodology and results. This is followed by the conclusion with regard to achievement
of the objectives, possible policy implications and implications for further research.
6.1 SUMMARY
6.1.1 BACKGROUND AND MOTIVATION
Low agricultural productivity and an increasing population are the main problems in the Sub-
Saharan Africa and in Ethiopia in particular that increased food insecurity and poverty (Geta et
al., 2010). Ethiopian economy is highly dependent on the agricultural sector where agriculture
accounts for 43% of GDP, 85% of employment and 90% of export earnings (MoARD, 2010).
Agriculture in Ethiopia is predominantly rainfed smallholder farming from which about 95%
of the agricultural GDP is derived (MoARD, 2010).
Maize is one of the most important food crops produced by smallholder farmers in the country
and accounts for 28% of total grain crops produced during the 2011/12 Meher season (CSA,
2012a). Due to the relative role of maize in total grain production and the participation of a
large number of smallholder farmers in its production, maize is a priority crop that can
contribute significantly to national food security of the country. Nevertheless, low production
and productivity characterises Ethiopian agriculture, specifically in maize production
(MoARD, 2009).
The country is endowed with abundant natural resources that hold great potential for
agricultural development (MoARD, 2010). Despite having good resource potential there is a
growing food shortage in the country due to the poor performance of the agricultural sector.
Poor performance of the agricultural sector is evident in the rural farming population’s lower
standard of living, seen in high levels of poverty, illiteracy and infant mortality (Diao, 2010).
93
As a result, the country continuously faces food insecurity and to some extent relies on food
aid and food imports (Adenew, 2003; Diao, 2010).
In an agriculturally dependent poor economy, it is expected that growth in agricultural
production, especially in crop growth, would contribute more to reduce poverty than stronger
macro-economic growth (Boccanfuso & Kabore, 2004). The key to growth in agricultural
production in Ethiopia lies in increasing the productivity and efficiency of smallholder farmers
(Owour, 2000). Due to the importance of agricultural productivity, the Ethiopian government
has given substantial policy emphasis to increasing productivity of smallholder farmers
through the Agricultural Development Led Industrialization (ADLI) strategy (Diao, 2010).
The ADLI strategy emphasises the role of adoption and intensification of yield enhancing
technological inputs such as fertilisers and improved seeds to increase crop productivity (Diao,
2010).
6.1.2 PROBLEM STATEMENT AND RESEARCH OBJECTIVES
Despite the fact that Ethiopia has abundant agricultural resource potential and is following a
consistent agricultural policy to boost productivity, the country could not achieve the expected
productivity increment resulting in food insecurity and poverty remaining at high levels (CSA,
2011b). In response to the agricultural development strategy pursued, studies by Mulat (1999)
and Arega and Zeller (2005) indicated substantial improvements in the adoption and use of
fertiliser, improved seeds and related inputs, but maize yields have not shown a substantial
improvement. One of the reasons for low maize productivity could be due to the presence of
technical inefficiencies in production. Therefore, information about farm level technical
efficiencies of smallholder farmers is important in improving the productivity of the
smallholder maize farmers (Alene & Hassan, 2003).
In Ethiopia, empirical studies on the technical efficiency of smallholder maize production are
limited and the few existing studies indicate the presence of technical inefficiencies. Among
the studies, Alene and Hassan (2003) have examined the technical efficiencies associated with
maize production in Western Ethiopia where the estimated efficiency is 76%. Similarly, Geta
et al. (2010) has examined the technical efficiency of maize production in Southern Ethiopia
and obtained an efficiency estimate of 40%. The technical efficiency estimates of maize
production provides good indications of efficiency in the zones studied, but they could not
94
provide a comprehensive estimate of the technical efficiency of the country. Furthermore,
Arega and Zeller (2005) and Bachewe (2009) explored technical efficiencies of multiple grain
crop production by using a single aggregate index value and obtained technical efficiencies of
79% and 40%, respectively. These studies also could not provide separate technical efficiency
estimations for maize production because of the use of a single aggregate index value for the
crops. From the literature, it was found that there was no comprehensive study of technical
efficiency of smallholder maize production that involved the main maize producing regions of
Ethiopia namely: Amhara, Oromia and SNNP regions.
Therefore, the main objective of the study is to identify factors affecting the technical
efficiency of smallholder maize farmers in selected regions of Ethiopia. The main objective
will be reached through the completion of the following sub-objectives:
i. Estimating technical efficiency of smallholder maize farmers using Stochastic
Frontier Analysis (SFA) which estimates a production frontier against which the
farmers’ actual production is evaluated to quantify their technical efficiency.
ii. Identifying and analysing the socio-economic and farm management factors that
affect the technical inefficiency of smallholder maize farmers using the SFA in
order to better understand the constraints that prevent farmers from producing the
maximum potential output.
6.1.3 LITERATURE REVIEW
When discussing the economic performance of producers, the two commonly used concepts
are productivity and efficiency (Fried et al., 2008). Productivity is a measure of physical
output produced from the quantity of inputs used (Latruffe, 2010). Variations in productivity
across producers or across time are attributed to differences in the following: production
technology, scale of operation, operating efficiency and operating environment (Fried et al.,
2008:8).Productivity is one means of improving output growth as output can be increased
either by increasing productivity of inputs or through expansion of farm sizes in production
(FAO & OECD, 2012). Growth in output that is achieved by increasing agricultural
productivity is a sustainable means of output growth without the need to expand farm sizes
95
(FAO & OECD, 2012). Increased agricultural productivity and the resulting increase in output
contribute to poverty reduction and to broader economic development (Mellor, 1999).
Efficiency, on the other hand, is the degree of success producers achieve in allocating the
inputs at their disposal and the outputs they produce in an effort to meet certain objectives
(Kumbhakar & Lovell, 2000). Not all producers are equally efficient in utilising their inputs to
achieve a potential yield due to performance variations among the producers. Efficiency can
be examined in terms of technical, allocative or economic efficiency (Fried et al., 2008).
Technical efficiency is a ratio of the actual production to optimal production (Green, 1993),
while allocative efficiency is the ability of a producer to use inputs in an optimal proportion,
given their respective prices (Coelli et al., 2005). Economic efficiency is a combination of
technical efficiency and allocative efficiency (Farrell, 1957).
Measuring and analysing technical efficiency in crop production assists in determining the
scope of raising the productivity of inefficient producers (Uaiene, 2008). Knowledge of the
level of technical efficiency contributes to the realisation of policy goals such as achieving
food security, poverty alleviation and growth and development through the improving
performance of inefficient producers (Mupanda, 2009). Technical efficiency can be measured
using either a parametric or non-parametric approach. The former uses an econometric
technique of estimation while the latter uses mathematical programming (Sarafidis, 2002).
Developments in the technique of efficiency measurement are influenced by the early works of
Debreu (1951), Koopmans (1951), Shephard (1953) and Farrell (1957).
In empirical work, the two most popular techniques of efficiency measurement are DEA and
SFA. DEA is a non-parametric technique of efficiency measurement based on mathematical
programming and it was first introduced by Charnes et al. (1978). DEA is widely applied in
literature mainly in operational research. However, DEA’s main limitation is that there is no
provision for statistical noise and all the deviations from optimal frontier are considered as
inefficiencies (Fried et al., 2008). SFA is an alternative efficiency measurement technique
which was first introduced by Aigner et al. (1977) and Meeusen and van den Broeck (1977).
SFA is a parametric technique of efficiency measurement that is based on econometric
estimation which requires specification of functional forms to estimate production frontiers.
The SFA model measures efficiency by accommodating statistical noise in the estimation of
96
technical efficiency (Kumbhakar & Lovell, 2000). For the estimation of technical efficiency
and examining the determinants of technical inefficiency using SFA, there are two approaches;
a two-stage and a single-stage approach (Kumbhakar & Lovell, 2000).In early studies, the
two-stage approach was applied where first technical efficiency is estimated, and then the
estimated technical efficiency is regressed against the determinants of technical inefficiency.
The two-stage estimation is based on the OLS method and suffers from inconsistency of
assumptions about the distribution of the inefficiency error terms (Kumbhakar & Lovell,
2000). More recent studies employed a single-stage approach in which both the production
frontier and the technical inefficiency model equations are estimated simultaneously
(Kumbhakar et al., 1991; Battesse & Coelli, 1995).
The SFA model was preferred to DEA because of the ability of the model to separately
accommodate the effects of statistical noise from that of technical inefficiency (Sarafidis,
2002). SFA is more applicable in crop efficiency analysis because of the probability that the
data could be affected by statistical noises arising from random shocks such as changes in
drought, frost, flood, etc. (Hordofa et al., 2008). A single-stage SFA model was used to
measure technical efficiency and to identify the determinants of technical inefficiency in the
maize production studied. FRONTIER version 4.1 software that was developed by Coelli
(1996) is used for the simultaneous estimation of production frontier and technical inefficiency
model parameters.
6.1.4 STUDY AREA PROFILE AND DATA
Ethiopia is located in the north eastern part of the horn of Africa with a total surface area of
1.12 million square kilometres (Mengistu, 2006). The country is structured into a federation of
nine ethnic based administrative regions and two centrally chartered city administrations (Sori,
2009). The nine regions are Tigray, Afar, Amhara, Oromia, Somali, Benishanful-Gumuz,
Southern Nations, Nationalities and Peoples (SNNP), Gambela and Harari while the two city
administrations are Addis Ababa and Dire Dawa. The study area included Tigray, Amhara,
Oromia, Somali, Benishangul-Gumuz, SNNP, Harari and Dire Dawa. Afar and Gambela
regions were not included because the information relevant to the study was not adequately
obtained for these two regions. Due to the presence of extremely varied topography, there is
wide variation in agro-ecology, climate, soil, vegetation and settlement patterns among the
study areas in the country (Camberiln & Philippon, 2001).Maize is produced in almost all
97
regions of Ethiopia, but the three largest producers are Oromia (61%), Amhara (20%) and
SNNP (12%) (Schneider & Anderson, 2010).
Crop production in Ethiopia is rainfed smallholder farming systems (CSA, 2012a). There are
two cropping seasons for temporary crops, Meher and Belg. Meher is the main cropping
season that accounts for 97% of total crop production (Alemayehu et al., 2012). Maize is
produced in both seasons, but Meher accounts for 90.5% while Belg accounts for 9.5% of the
total maize production annually (Alemayehu et al., 2012).
The data used for the study is a secondary data set obtained from the Central Statistical
Agency of Ethiopia for maize production performed during Meher season of 2011/12. The
data was collected as part of the Rural Socio-economic Survey of the country. In the process
of sample selection, the CSA set larger quotas for the most populous regions. Based on the
quotas, the percentage of the sample composition of each region in the study was as follows:
Oromia (24%), Amhara (24%), SNNP (22%), Tigray (15%), Benishangul-Gumuz (six
percent), Harari (five percent), Somali (two percent) and Dire Dawa (two percent). The total
number of observations used is 438. From the descriptive statistics of the data, it was observed
that the sampled households are characterised by lower input usage, weak socio-economic
conditions and poor farm management practises which can potentially influence productivity
of maize production.
6.1.5 METHODOLOGY
Chapter four described the methodological procedures used to achieve the sub-objectives. SFA
was used to estimate the level of technical efficiency and to identify and analyse factors
affecting the technical inefficiency of smallholder maize farmers for the selected regions of
Ethiopia.
A Cobb-Douglas production model was used to relate seed use, fertiliser and labour to maize
production in the SFA specification. The linear technical inefficiency model consisted of
explanatory variables that are hypothesised to affect technical inefficiency. The choices of the
variables were determined by the literature reviewed and type of data available (as it is
secondary data). The technical inefficiency model’s explanatory variables were: age, gender,
98
education, household size, farm size, tenure, oxen, extension, irrigation, credit, off-farm
income, seed type, organic fertiliser and soil protection.
The first attempt to estimate the SFA was unsuccessful due to the presence of high
multicollinearity among the technical inefficiency model explanatory variables.
Multicollinearity refers to linear correlations among explanatory variables in a multiple
regression model that can lead to biased parameter estimates. In order to solve the problem of
multicollinearity, Principal Component Regression (PCR) was used. Through PCR, the
correlated explanatory variables were transformed into a new set of uncorrelated Principal
Components (PCs), which are further used to estimate the parameters and significances of the
correlated original technical inefficiency model variables(Kline, 1994). The extraction of the
Only five PCs that were associated with eigenvalues greater than one are extracted.
Significances of the five PCs were determined by incorporating the five PCs in SFA
regression. All the five PCs were found significant and the PCs were used to determine the
coefficients and significances of the correlated technical inefficiency model variables.
Hypotheses testing were conducted to test for the presence of technical inefficiency effects in
the estimated SFA model and to examine whether the technical inefficiency effects are
stochastic. For hypothesis testing, the generalised Log Likelihood statistic (λ) was used
against the chi-square and the critical values obtained with the degrees of freedom determined
by the number of parameter restrictions imposed under the null hypothesis. The hypotheses
tests confirmed the presence of technical inefficiency effects in the model and the inefficiency
effects are stochastic such that SFA, with the technical inefficiency model, was an appropriate
model for efficiency measurement.
6.1.6 RESULTS AND DISCUSSION
6.1.6.1 Technical Efficiency Estimation
The Cobb-Douglas production function indicated that maize yield has a positive response to
seed and Nitrogen, but negative to labour. The positive coefficients of seed and nitrogen
indicate that increasing the use of seed and Nitrogen increases maize production. However,
only Nitrogen is statistically significant in explaining maize yield. The significance of
Nitrogen implied that increasing the use of Nitrogen application will increase maize
99
production significantly. The sign of labour coefficient is unexpectedly negative implying that
the use of additional labour will decrease maize production, but the relationship is statistically
insignificant.
Variance (2) of the estimated SFA model is 2.06 and highly significant. The significance of
the variance indicated the correctness of the specified distributional assumption. The estimated
value of  is 0.99 and is significant at a one percent level. The value of implies that about
99% of the random variation in output of maize production is attributed to technical
inefficiency and indicates the importance of examining technical inefficiency in production.
The estimated mean technical efficiency score for the sample is 77% indicating that on
average, the sampled maize farmers are able to obtain only 77% of potential output from the
given mix of production inputs and available resources. The finding suggests the presence of a
considerable level of technical inefficiency among the studied households. The mean technical
efficiency estimated gives an indication that the sampled maize farmers have the potential for
increasing their output by about 23% by using the existing resources and technology more
efficiently. Given the observed technical inefficiency in maize production, it is possible to
improve production by improving the level of technical efficiency. In order to improve the
current state of technical efficiency, it is necessary to examine factors affecting technical
inefficiency of maize production.
6.1.6.2 Factors Affecting Technical Inefficiency
While examining factors affecting technical inefficiency, a positive sign for the estimated
coefficient indicates that the associated variable increases technical inefficiency and vice
versa. The implication is that a variable that increases technical inefficiency will decrease
technical efficiency and vice versa. The estimated coefficients of age, gender, household size,
oxen, extension, irrigation, credit, seed type and soil protection are negative and significant in
that they have a decreasing effect on technical inefficiency. The estimated coefficient of off-
farm income is positive and significant with an increasing effect on maize farm technical
inefficiency. Education, tenure, farm size and organic fertiliser are not statistically significant
determinants of technical inefficiency.
The coefficient of age is negative and highly significant indicating that whenever age of
household heads increase, farm technical inefficiency will decrease. The finding supports the
100
possibility of having better resources and experience with increased age that can potentially
decrease inefficiency (Fesessu, 2008). Similarly, the coefficient of gender is negative and
significant suggesting that male-headed households are more technically efficient than female-
headed households. The finding supports the fact that male-headed households could have
better access to farm resources such as land and credit than their female counterparts, causing
efficiency differences between male and female-headed households. Female household heads’
responsibilities for domestic household activities also make male headed farm households
more efficient relative to female headed ones (FARA, 2009; AWM, 2009).
The coefficient for household size is negative and significant indicating that whenever
household size increases, technical inefficiency will decrease which could be attributed to the
contribution of large households to family labour. Similarly, the coefficient of oxen is negative
and significant indicating that ownership of adequate oxen decreases maize farm technical
inefficiency. The finding is as expected and is consistent with the finding of Geta et al. (2010).
Another important variable estimated is credit which has a negative and significant coefficient.
The finding confirms the importance of credit in decreasing farm technical inefficiency by
increasing the ability of farmers to overcome the financial constraints to buy inputs and make
necessary investments. The result is in line with the findings of Alene and Hassan (2003) and
Obwona (2006). Off-farm income is another important determinant of the smallholder maize
farmers’ technical inefficiency. The coefficient of off-farm income is positive and significant
indicating that the pursuit of off-farm income will increase farm technical inefficiency. This is
due to the possible reason that participation in off farm income decreases farm labour and
effort which could potentially lead to inefficiency in the farm work. The finding is similar to
the findings of Alene and Hassan (2003) and Baruwa and Oke (2012).
Extension is an important determinant of technical inefficiency with a negative and significant
coefficient. Extension therefore decreases technical inefficiency and the result was expected.
The finding is consistent with findings of Obwona (2006) and Bachewe (2009).Another
important variable is irrigation. The coefficient of irrigation is negative and statistically
significant suggesting that the use of irrigation decreases technical inefficiency. The average
technical efficiency score of the irrigated farms were much more efficient than the total
sample farms where the average TE was 85% for the irrigated farms compared to 77% of the
101
total sample mean TE score. The finding of irrigation is in line with the results reported by
Khai and Yabe (2011) and Makombe et al. (2011).
Other important variables are seed type and soil protection practises of the farmers which were
found to have negative and significant coefficients. The negative sign of seed type indicates
that the use of improved seed decreases technical inefficiency among the maize producers.
The finding is consistent with the empirical studies of Geta et al. (2010) and Maseatile (2011).
Similarly the negative coefficient of soil protection confirms the importance of soil protection
in decreasing technical inefficiency in maize production. The average technical efficiency of
the farmers that used soil protection is nearly 83% which is greater than the mean technical
efficiency of the total sample (77%). The result of soil protection is consistent with the finding
reported by Geta et al. (2010).
6.2 CONCLUSION AND POLICY IMPLICATION
Due to low production and productivity of food crops, Ethiopia is unable to supply the
growing food demand of its population. Despite the fact that there are improvements in
adoption and utilisation of yield enhancing inputs in maize production, the level of yield did
not show the expected productivity increment. In an attempt to investigate the problems
causing low maize productivity of smallholder farmers in Ethiopia, the potential determinants
of technical efficiency were analysed by studying a sample of farmers selected from most of
the maize producing regions of the country.
The results from the study indicated that there is a considerable level of technical inefficiency
among the maize farmers that contributed to lowered productivity. The results show that most
of the variation in maize production is due to technical inefficiency. The results furthermore
indicated that it is possible to improve the current productivity by increasing technical
efficiency. The current level of low production efficiency can be addressed through improving
the access of farmers to rural credit and extension services, by promoting soil and land
conservation practises and by promoting small-scale irrigation schemes. Based on the findings,
some policy implications are drawn.
102
6.2.1 POLICY IMPLICATIONS
Access to rural credit is an important factor that can contribute to an improvement in
productivity of the smallholder farmers. Credit is an important source of capital for the poor
farmers to be able to purchase yield enhancing technological inputs, own an adequate number
of oxen for farming and to make land improvements that can potentially increase farm
productivity. Micro-finance institutions (MFIs) and rural saving and credit cooperatives
(RUSCCOPs) are the primary suppliers of rural credit in Ethiopia (Ameha, 2011). From the
literature, the main constraints prevailing in the rural credit service of the country are the
limited capacity of credit outreach, shortage of loanable funds, complicated credit application
procedures that the farmers undergo to obtain credit and the requirements of high collateral
(Yehuala, 2008; IFAD, 2011). Developing the loanable fund of the credit suppliers could
improve credit outreach and improve credit access to the farm households. This can be
addressed partly by promoting the mobilisation of savings from farm households through
better deposit rates which encourage the farmers to save. In addition, promoting the
coordination of MFIs to the commercial banks to acquire loanable funds at affordable interest
rates and with minimal collateral can help to overcome the problem of loanable fund
shortages. Government can promote and facilitate these efforts through the collaborative work
of the Ministry of Agriculture and Rural Development (MoARD) and the Ministry of Finance
and Economic Development (MoFED).
Extension services have important contributions in improving the current productivity levels
through enhancing the management capacity of the farmers. Extension service coverage needs
to be expanded so that all smallholder farmers become beneficiaries of the service. In addition
to increasing the coverage of extension services, there should be a coordinated system of
monitoring and evaluation of service delivery. To overcome the prevailing high levels of soil
and land degradation farmers, in conjunction with extension services, need to implement soil
conservation practises such as terracing, inter-cropping, planting trees and contour ploughing.
In an attempt to increase maize productivity irrigation has very important role, given the
unreliable and erratic rainfall of the country. The role of low cost irrigation and water
harvesting schemes should not be overlooked. The application of yield enhancing inputs will
work better if used together with sufficient amounts of water required for the crop production.
Promotion of infield rainwater harvesting, low cost water harvest irrigation schemes such as
103
constructing of earth dams, river diversions and hand pumps can greatly contribute to
increased maize productivity and food security in the country. Besides irrigation, the use of
drought resistant seeds should be promoted through the integrated work with seed suppliers
and extension services.
6.2.3 IMPLICATIONS FOR FURTHER RESEARCH
All the various factors that can influence the technical efficiency of the smallholder maize
farmers were not exhaustively explored due to data limitations. Future studies that can address
the shortcomings through incorporating all the possible factors in the production frontier and
the technical inefficiency model are highly encouraged.
The scope of study is comprehensive through the inclusion of the major maize producing
areas. However, regional variations in maize production were not accounted for. The
possibility of regional variations could be a possible reason for the high proportion of variance
due to technical inefficiencies to the total variance in the estimated model. Future studies that
could account for regional variations in maize production would be more informative.
The study has some data limitations. As secondary cross-sectional data is employed, in-depth
investigation about some of the variables, especially credit, off-farm income and the
household head’s decision-making role was impossible. For example, information regarding
credit indicates only that some farmers had received credit and others had not. The data did not
provide further information as to whether all the farmers had applied for credit. Similarly, off-
farm income does not show if the farmers used the off-farm income for the betterment of the
maize production or for other purposes. In addition different variable combinations, such as
combinations of improved seeds versus traditional seeds, improved seed with chemical
fertilisers or without chemical fertilisers were not accounted for. Therefore, future studies that
use primary data can provide better insights into the variables involved in a technical
efficiency analysis of maize farmers. Panel data based studies can also provide a clear picture
of the production efficiency over time.
104
REFERENCES
Abdulahi, H., Sarobol, E., Vichukit, V. & Suwannarat, C. (2006).Uptake and Response of
Maize (Zea mays L.)Varieties to Nitrogen and Phosphorus Fertilisers in Semi-arid Areas
of Eastern Ethiopia. Kasetsart Journal of Natural Science, 40, 604-615.
Abesha, N. (2009). Domestic Energy Consumption and Deforestation in Harari Region:
Assessment of Students’ Awareness and Views. Addis Ababa University, Department of
Geography and Environment Education. Ethiopia.
Adelman, I. & Morris, C. (1988).Comparative Patterns of Economic Development, 1850-
1914. Baltimore, United States: John Hopkins University Press.
Adenew, B. (2003). The Food Security Role of Agriculture in Ethiopia. Paper prepared for the
International Conference. Rome, Italy, 20-22 October 2003.
Admassie, A. & Asfaw, A. (1997). The Impact of Education on Allocative and Technical
Efficiency of Farmers: The case of Ethiopian Smallholders. Working Paper Presented at
the Seventh Annual Conference on the Ethiopian Economy in Nazret. Addis Ababa
University, Ethiopia.
Agricultural Water Management (AWM).(2009). Agricultural Water Management National
Situation Analysis. http://awm-solutions.iwmi.org/ethiopia_update.aspx. Accessed [20
September 2013].
Aigner, D.J., Lovell, C.A.K. & Schmidt, P. (1977).Formulation and Estimation of Stochastic
Frontier Production Function Models. Journal of Econometrics, 6(1):21-37.
Alemayehu, S.T., Dorosh, P. & Sinafikeh, A. (2011).Crop Production in Ethiopia: Regional
Patterns and Trends: Ethiopia Strategy Support Program (ESSPII).Summary of ESSP II
Working Paper 16. Addis Ababa. Ethiopia.
Alemu, A.B., Nuppenau, E.A. & Bolland, H. (2009). Technical Efficiency Across Agro-
ecological Zones in Ethiopia. Agricultural Journal, 4(4):202-207.
Alene, A.D. & Hassan, R.M. (2003).The Determinants of Farm-level Technical Efficiency
among Adopters of Improved Maize Production Technology in Western Ethiopia.
Agrekon, 42(1):1-14.
Ali, D.A. & Deininger, K. (2012).Causes and Implications of Credit Rationing in Rural
Ethiopia: The Importance of Spatial Variation. Policy Research Working Paper 6096.
Addis Ababa. Ethiopia.
105
Ameha, W. (2011).Meeting the Financial Needs of Smallholder Farmers in Ethiopia. In
Djurfelt, G. Aryeetey, E. & Isinka, A.C.(eds).African Smallholder Food Crops, Markets
and Policy. UK: CPI Antony Rowe: 156-165.
Andrew, D. (1999). Farm Size and Productivity in Malawian Smallholder Agriculture. Journal
of Development Studies, 35(5):141-161.
Appleton, S. & Arsene, B. (1996). Education and Agricultural Productivity: Evidence from
Uganda. Journal of International Development, 8, 415-444.
Arega, A.D. & Zeller, M. (2005). Technology Adoption and Farmer Efficiency in Multiple
Crops Production in Eastern Ethiopia: A comparison of Parametric and Non-parametric
Distance Functions. Agricultural Economics Review, 6(1):6-17.
Awulachew, S.B., Yilma, A.D., Loulseged, M., Loiskandl, W., Ayana, M. & Alamirew, T.
(2007).Water Resources and Irrigation Development in Ethiopia. IWMI Working Paper
No 123. Addis Ababa. Ethiopia.
Ayalew, T. & Dejene, A. (2011).Integrated Application of Compost and Inorganic Fertilisers
for Production of Potato (Solanum tuberosum l.) at Angacha and Kokate in Southern
Ethiopia. Journal of Biology, Agriculture and Healthcare, 1(2):15-17.
Ayele, G., Bekele, M. & Zekaria, S. (2006). Productivity and Efficiency of Agricultural
Extension Package in Ethiopia. Ethiopian Development Research Institute Research
Report 5. Addis Ababa. Ethiopia.
Ayele, S. & Bosire, C. (2011).Farmers’ Use of Improved Agricultural Inputs and Practises:
Review and Synthesis of Research in Ethiopia. Addis Ababa. Ethiopia.
Bachewe, F.N. (2009). The State of Subsistence Agriculture in Ethiopia: Sources of Output
Growth and Agricultural Inefficiency. PhD Thesis. Faculty of Graduate School.
University of Minosota, USA.
Bardhan, P.K. (1973). Size, Productivity and Returns to Scale: An Analysis of Farm-Level
Data in Indian Agriculture. Journal of Political Economy, 81(6):1370-1386.
Baruwa, O.I. & Oke, J.T.O. (2012).Analysis of the Technical Efficiency of Smallholder
Cocoyam Farms in Ondo State of Nigeria. Tropicultura, 30(1):36-40.
Battese, G.E. & Coelli, T.J. (1992). Frontier Production Functions, Technical Efficiency and
Panel Data: With Application of Paddy Farmers in India. The Journal of Productivity
Analysis, 3, 153-169.
Battese, G.E. & Coelli, T.J. (1993).A Stochastic Frontier Production Function Incorporating a
Model for Technical Inefficiency Effects. Working Paper in Econometrics and Applied
Statistics No 69. Department of Econometrics. University of New England, Armidale.
106
Battese, G.E. & Coelli, T.J. (1995).A Model for Technical Inefficiency Effects in a Stochastic
Frontier Production Function for Panel Data. Empirical Economics, 20, 325-332.
Battese, G.E. & Coelli, T.J. (1996).Identification of Factors Which Influence the Technical
Inefficiency of Indian Farmers. Australian Journal of Agricultural Economics, 40, 103-
128.
Battese, G.E. & Corra, G.S. (1977).Estimation of a Production Function Model with
Applications to the Pastoral Zone of Eastern Australia. Journal of Agricultural
Economics, 21(3):169-179.
Belloumi, M. & Matoussi, M.S. (2006).A Stochastic Frontier Approach for Measuring
Technical Efficiencies of Date Farms in Southern Tunisia. Agricultural and Resource
Economics Review, 35(2):285-298.
Benin, S. (2006).Policies and Programs Affecting Land Management Practices, Input Use and
Productivity in the Highlands of Amhara Region of Ethiopia. International Food Policy
Research Institute, Washington D.C. USA.
Benishangul Gumuz Region Food Security Strategy
(BGRFSS).(2004).http://www.cangoethiopia.org/assets/docs/Benishangul-Gumuz
strategy pdf. Accessed[15 June 2013].
Bewket, W. (2009).Rainfall Variability and Crop Production in Ethiopia: Case Study in the
Amhara Region. Proceedings of the 16th International Conference of Ethiopian
Studies. Trondheim, 2009 p823-836.
Beyene, A.D. (2008). Determinants of Off-farm Participation Decision of Farm Households in
Ethiopia. Agrekon, 47(1):150-161.
Binam, J.N., Sylla, K., Diarra, I. & Nyambi, G. (2003). Factors Affecting Technical Efficiency
among Coffee Farmers in Cote d Ivoire: Evidence from the Center West Region. African
Development Review, 15(1):66-76.
Binam, J.N., Tonye, J., Wandji, N., Nyambi, G. & Akoa, M. (2004).Factors Affecting
Technical Efficiency among Smallholder Farmers in Slush and Burn Agriculture Zones
of Cameroon. J.N. Food Policy, 29, 531-545.
Boccanafuso, D. & Kabore, T.S. (2004).Macroeconomic Growth, Sectoral Quality of Growth
and Poverty in Developing Countries: Measure and Application to Burkina Faso.
Working Papers No 04-07. Sherbrook. Canada.
Bravo-Ureta, B.E. & Pinheiro, A.E. (1997). Technical, Economic and Allocative Efficiency in
Peasant Farming: Evidence from the Dominican Republic. The Developing Economies,
35(1):48-67.
107
Bravo-Ureta, B.E. & Rieger, I. (1991).Dairy Farm Efficiency Measurement Using Stochastic
Frontiers and Neoclassical Duality. American Journal of Agricultural Economics,
73(2):421-428.
Bureau of Agriculture and Rural Development (BoARD). (2006). Southern Nation Nationality
and People Regional State Agriculture and Rural Development Bureau Annual
Report.Awassa. Ethiopia.
Bureau of Finance and Economic Development (BoFED).(2011). Amhara National Regional
State Geography and Climate. http://www.amharabofed.gov.et. Accessed [1 June 2013].
Camberlin. P. & Philippon, N. (2001). The East African March-May Rainy Season Associated
Atmosphere Dynamics and Predictability over the 1968-97. Am. Meterol. Soc., 15, 1002-
1019.
Central Statistical Agency (CSA). (2004). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2005). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2006a). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA).(2006b). Atlas of the Ethiopian Rural Economy Prepared by
a Collaborative Work of CSA, IFPRI & EDRI. Addis Ababa. Ethiopia.
Central Statistical Agency (CSA). (2007). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2008). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2009a). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA).(2009b). Innovations in Data Dissemination at the Central
Statistical Agency of Ethiopia. IASSIST Quarterly Fall. Addis Ababa. Ethiopia.
108
Central Statistical Agency (CSA). (2010). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2011a). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA).(2011b). Ethiopian Demographic and Health Survey. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA).(2011c). Large and Medium-scale Commercial Farms
Sample Survey.(2011). Statistical Bulletin 505. Addis Ababa. Ethiopia.
Central Statistical Agency (CSA). (2012a). Agricultural Sample Survey Report on Area and
Production of Major Food Crops: Private Peasant Holdings. Meher Season. Addis
Ababa. Ethiopia.
Central Statistical Agency (CSA). (2012b).Agricultural Sample Survey Report on Land
Utilisation: Private Peasant Holdings. Meher Season. Addis Ababa. Ethiopia.
Central Statistical Agency (CSA).(2013). Ethiopian Rural Socio-economic Survey Report.
Addis Ababa. Ethiopia.
Chamberlin, J. & Schmidt, E. (2011).Ethiopian Agriculture: A Dynamic Geographic
Perspective, Ethiopia Strategy Support Program II. ESSP II working Paper 017. Addis
Ababa. Ethiopia.
Charnes, A., Cooper, W.W., Lewin, A.Y. & Seiford, L.M. (1994).Data Envelopment Analysis:
Theory, Methodology and Applications. Boston: Kluwer Academic Publishers.
Charnes, A., Cooper, W.W. & Rhodes, E. (1978).Measuring the Efficiency of Decision-
Making Units. European Journal of Operational Research, 2(6):429-444.
Clunies-Ross, A., Forsyth, D. & Huq, M. (2009).Development Economics. University of
Strathclyde, London: McGraw-Hill Higher Education.
Coelli, T.J. (1996). A Guide to Frontier Version 4.1: A Computer Program for Stochastic
Production and Cost Function Estimates. University of New England, Armidale,
Australia. CEPCA Working Paper, 7(96):1-33.
Coelli, T.J., Prasada, R.D., O’Donnell, C.J. & Battese, G.E. (2005).An Introduction to
Efficiency and Productivity Analysis. 2nded. New York: Springer Science + Business
Media.
109
Corppenstedt, A. & Abbi, M. (1996).An Analysis of the Extent and Causes of the Technical
Efficiency of Farmers Growing Cereals in Ethiopia. Ethiopian Journal of Economics,
5(1):39-62.
Croppenstedt, A., Demeke, M. & Meschi, M.M. (2003).Technology Adoption in the Presence
of Constraints. The Case of Fertiliser Demand in Ethiopia. Review of Development
Economics, 7(1):58-70.
Debreu, G. (1951). The Coefficient of Resource Utilisation. Econometrica, 19, 273-292.
Dercon, S. & Hill, R.V. (2009).Growth from Agriculture in Ethiopia: Identifying Key
Constraints. Paper Prepared as Part of the Study on Agriculture and Growth in Ethiopia.
Addis Ababa. Ethiopia.
Derege, E.A. (2010). Analysis of Factors Affecting Technical Efficiency of Coffee Producers
in Jimma Zone: A Stochastic Frontier Analyses. M.Sc (Economic Policy Analysis)
Thesis. Department of Economics. Addis Ababa University, Ethiopia.
Development Studies Associates (DSA). (2006). Study on Improving the Efficiency of Input
Markets. Ministry of Agriculture and Rural Development, Addis Ababa. Ethiopia.
Diao, X. (2010).Economic Importance of Agriculture for Sustainable Development and
Poverty Reduction: The Case Study of Ethiopia. A paper presented to the Working Party
on Agricultural Policy and Markets. Paris, France, 17-20 May 2010.
Dire Dawa Administration Program of Adaptation to Climate Change (DDAPA).(2011).
www.epa.gov.et/Download/Climate/RegionalClimateChangeAdaptationProgrammes/Dir
eDawaAdminstrationClimateChangeAdaptation Program.pdf. Accessed [15 June 2013].
Disaster Prevention and Preparedness Bureau (DPPB).(2004). Somali Regional State
Summary.http://www.SomaliRegionalStateSummary.SomaliEthiopia.Accessed [20 June
2013].
Draper, N.R. & Smith, H. (1981).Applied Regression Analysis. 2ndEdition. New York: John
Wiley and Sons.
Dyer, G. (1996). Output Per Acre and Size of Holding: The Logic of Peasant Agriculture
underSemi-Feudalism. Journal of Peasant Studies, 24, 103-131.
Ellis, F.(1993).Peasant Economics. Second Edition. Cambridge: Cambridge University Press.
Emiru, S. (2004).Organic Farming in the Context of Food Security in Ethiopia: “Novelty or
Necessity”. Challenges and Prospects of Food Security in Ethiopia. Proceedings of the
Food Security Conference. UNCC, Addis Ababa, Ethiopia, 13-15 August 2003.
Farrell, M.J. (1957). Measurement of Productive Efficiency. Journal of the Royal Statical
Society Series A, 120(3):253-290.
110
Fekedulegn, B.D., Colbert, J.J., Hicks, J.R. & Schuckers, M.E. (2002). Coping with
Multicollinearity: An Example on Application of Principal Components Regression in
Dendroecology. Research Paper NE-721.US Department of Agriculture Forest Service.
North-eastern Research Station, USA.
Fesessu, H. (2008). Technical Efficiency of Ethiopian Coffee Production: An Empirical Study.
PhD Thesis. Department of Economics. Faculty of the Graduate School. Howard
University, Washington D.C.
Food and Agricultural Organization (FAO) (1984).Geo-morphology and Soils: Assistance to
Land Use-Planning Project. Field Document 2. Addis Ababa. Ethiopia.
Food and Agricultural Organization (FAO).(2005). AQUASTAT-FAO’s Information System on
Water and Agriculture.http://www.fao.org/nr/water/aquastat/water_use/index.stm.
Accessed [10 July 2013].
Food and Agricultural Organisation (FAO). (2011).Save and Grow: A policy Maker's Guide to
the Sustainable Intensification of Smallholder Crop Production. Rome, Italy.
Food and Agricultural Organization (FAO) & Organization for Economic Cooperation and
Development (OECD).(2012). Sustainable Agricultural Productivity Growth and
Bridging the Gap for Small-Family Farms. Interagency Report to the Mexican G20
Presidency. Mexico, 12 June 2012.
Forum for Agricultural Research in Africa (FARA). (2009). Patterns of Change in Maize
Production in Africa: Implications for Maize Policy Development. http://www.fara-
africa.org/media/uploads/library/docs/policy_briefs/patterns_of_change_in_maize_prod
uction_in_africa.pdf. Accessed [30 June 2013].
Fried, H.O., Lovell, C.A.K. & Schmidt, S.S. (2008).The Measurement of Productive Efficiency
and Productivity Growth. Oxford: Oxford University Press.
Fufa, B. & Hassan, R.M. (2003).Stochastic Maize Production Technology and Production
Risk Analysis in Dadar District of East Ethiopia. Agrekon, 42(2):116-128.
Fufa, B. & Hassan, R.M. (2006). Determinants of Fertiliser Use on Maize in Eastern Ethiopia:
A weighted Endogenous Sampling Analysis of the Extent and Intensity of Adoption.
Agrekon, 45(1):38-49.
Gavian, S. & Ehui, S. (1999). Measuring the Production Efficiency of Alternative Land
Tenure Contracts in a Mixed Crop-livestock System in Ethiopia. Agricultural
Economics, 20, 37-49.
111
Gbegeh, B.D. & Akubuilo, C.J.C. (2013).Socioeconomic Determinants of Adoption of Yam
minisett by Farmers in Rivers State of Nigeria. Wudpecker Journal of Agricultural
Research, 2(1):033-038.
Gebreegziabher, Z., Oskam, A. & Woldehanna, T. (2004).Technical Efficiency of Peasant
Farmers in Northern Ethiopia: A Stochastic Frontier Approach. Proceedings of the
Second International Conference on the Ethiopian Economy Volume 2. Addis Ababa.
Ethiopia.
Gebrehiwot, T. (2008).Rural Food Security in Tigray, Ethiopia: Policy Impact Evaluation.
M.Sc. Thesis. International Institute for Geo-information Science and Earth Observation.
Enschede, the Netherlands.
Gebre-Selassie, A. & Bekele, T. (2013).A Review of Ethiopian Agriculture: Roles, Policy and
Small-scale Farming Systems: Country Analysis. Addis Ababa, Ethiopia.
Gebreselassie, S. (2006).Intensification of Smallholder Agriculture in Ethiopia: Options and
Scenarios. Paper prepared for the Future Agricultures Consortium Meeting at the
Institute of Development Studies. Addis Ababa, Ethiopia, 20-22 March 2006.
Geta, E., Bogale, A., Kassa, B. & Elias, E. (2010).Productivity and Efficiency Analysis of
Smallholder Maize Producers in Southern Ethiopia. J Hum Ecol, 41(1):67-75.
Getahun, H. (2001). Assault on Rural Poverty: A case of Ethiopia. USA: University Press of
America.
Giller, K., Misiko, M. & Tittonell, P. (2006).Managing Organic Resources for Soil
Management: Magazine on Low External Input and Sustainable Agriculture. LEISA-
LEUSDEN, 22(4):16.
Goodwin, B. & Mishra, A. (2004).Farming Efficiency and the Determinants of Multiple Job
Holding by Farm Operators. American Journal of Agricultural Economics, 86(3):722-
729.
Greene, W.H. (1980). Maximum Likelihood Estimation of Econometric Frontier Functions.
Journal of Econometrics, 13(1):27-56.
Greene, W.H. (1990). A Gamma-Distributed Stochastic Frontier Model. Journal of
Econometrics, 46(1/2):141-164.
Greene, W.H. (1993). Econometric Analysis. Second Edition. Prentice Hall: Englewood Cliffs.
Griffin, R.C., Montgomery, J.M. & Edward, M.R. (1987).Selecting Functional Form in
Production Function Analysis. Western Journal of Agricultural Economics, 12(2):216-
227.
112
Gruhn, P., Goletti, F. & Yudelman, M. (2000).Integrated Nutrient Management, Soil Fertility,
and Sustainable Agriculture: Current Issues and Future Challenges. International Food
Policy Research Institute, Washington D.C. USA.
Gujarati, D.N. (2003). Basic Econometrics. Fourth Edition. New Delhi: Tata McGraw-Hill
Publishing Company Limited.
Hadgu, F., Gebrekidan, H., Kibret, K. & Yitaferu, B. (2013).Effects of Nitrogen, Phosphorus
and Soil Water Content on Growth and Nutrient Uptake of Maize (Zea mays L.)Grown
on Fluvisols in Tigray region of Ethiopia. Journal of Biodiversity and Environmental
Sciences, 3(5):68-85.
Haise, H.A. & Hagan, R.M. (1967).Soil, Plant and Evaporative Measurement as Criteria for
Scheduling Irrigation. In Hagan, R.M., Haise, H.R. & Edminster, T.W. (eds). Irrigation
of Agricultural lands. Monog. Am. Soc. Agron. Medison: 577-604.
Haji, J. (2008).Economic Efficiency and Marketing Performance of vegetable production in
the Eastern and Central parts of Ethiopia. Unpublished PhD Thesis. Department of
Economics. Swedish University of Agricultural Sciences, Uppsala.
Harari Peoples Regional State (HPRS). (2011). Programme of Plan on Adaptation to Climate
Change.www.epa.gov.et/Download/Climate/Regional Climate Change Adaptation
Programmes/Harri Regional State Climate Change Adaptation Program.pdf. Accessed
[25 May 2013].
Hassan, S. & Ahmad, B. (2005). Stochastic Frontier Production Function: Application and
Hypothesis Testing. International Journal of Agriculture and Biology, 7(3):427-430.
Haynes, R.J. (1986). Mineral Nitrogen in the Plant-Soil System. London: Academic Press.
Henderson, B. & Kingwell, R. (2002).Technical and Allocative Efficiency of Broadacre
Farmers. Paper Presented to the 46th Annual Conference of the Australian Agricultural
and Resource Economics Society. Western Australia, 13-15 February, 2002.
Hordofa, T., Menkir, M., Bekele, B. & Erkossa, T. (2008).Irrigation and Rain-fed Crop
Production System in Ethiopia. Institute of Agricultural Research and International
Water Management Institute (IWMI)Sub-regional Office for the Nile Basin and East
Africa. Addis Ababa. Ethiopia.
Huang, Y. & Kalirajan, K.P. (1997). Potential of China’s Grain Production: Evidence from
Household Data. Agricultural Economics, 17(2-3):191-199.
Hurni, H. (1993). Land Degradation, Famines and Resource Scenarios in Ethiopia. In
Pimental, D. (eds). World Soil Erosion and Conservation. Cambridge: Cambridge
University Press: 27-62.
113
Idiong, I.C. (2007). Estimation of Farm Level Technical Efficiency in Small Swamp Rice
Production in Cross River State of Nigeria: A Stochastic Frontier Approach. World
Journal of Agricultural Science, 3(5):653-658.
International Food Policy Research Institute (IFPRI). (2010a). Fertiliser and Soil Fertility
Potential in Ethiopia: Constraints and Opportunities for Enhancing the System. Soil
Fertility Potentials. Addis Ababa. Ethiopia.
International Food Policy Research Institute (IFPRI). (2010b). Maize Value Chain Potential in
Ethiopia: Constraints and Opportunities to Enhancing the System. Working Paper on
Maize Diagnostics. Addis Ababa. Ethiopia.
International Fund for Agricultural Development (IFAD). (2008). Enabling Poor Rural People
to Overcome Poverty. Federal Democratic Republic of Ethiopia, Country Strategic
Opportunities Programme. Note to Executive Board Directors for Review. Rome. Italy.
International Fund for Agricultural Development (IFAD).(2011).Rural Financial
Intermediation Programme Interim Evaluation. Report No. 2373-ET. Addis Ababa.
Ethiopia.
Jamison, D.T. & Lau, L.J. (1982).Farmers Education and Farm Efficiency. Baltimore: John
Hopkins University Press.
Jayne, T.S., Govereh, J., Wanzala, M. & Demeke, M. (2003). Fertiliser Market Development:
a comparative analysis of Ethiopia, Kenya and Zambia. Food Policy, 28, 293-316.
Jouan-Rimbaud, D., Walczak, B., Massart, D.L., Last, I.R. & Prebble, K.A. (1995).
Comparison of Multivariate Methods Based on Latent Vectors and Methods Based on
Wavelength Selection for the Analysis of Near-infrared Spectroscopic Data. Analytica
Chimica Acta, 304(3):285-295.
Kalirajan, K. (1991). The Importance of Efficient Use in the Adoption of Technology: A
Micro Panel Data Analysis. Journal of Productive Analysis, 2, 113-126.
Kassie, M., Zikhali, P., Pender, J. & Kohlin, G. (2011).Sustainable Agricultural Practices and
Agricultural Productivity in Ethiopia: Does Agro-ecology Matter? Discussion Paper
series 11-05. Addis Ababa. Ethiopia.
Khai, H.V. & Yabe, M. (2011).Technical Efficiency Analysis of Maize Production in
Vietnam. Journal of ISSAAS, 17(1):135-146.
Khaile, P.M.E. (2011). Factors Affecting Technical Efficiency of Small-scale Raisin Producers
in Eksteenskuil. M.Sc (Agric. Economics) Thesis. Department of Agricultural
Economics. University of the Free State, Bloemfontein.
Kline, P. (1994). An Easy Guide to Factor Analysis. London: Routledge Kolb.
114
Kodde, D.A. & Palm, F.C. (1986).Wald Criteria for Jointly Testing Equality and Inequality
Restrictions. Econometrica, 54(5):1243-1248.
Koopmans, T.C. (1951). Analysis of Production as an Efficient Combination of
Activities. Activity Analysis of Production and Allocation, 13, 33-37.
Kumbhakar, S.C., Ghosh, S. & McGuckin, J.T. (1991).A Generalised Production Frontier
Approach for Estimating Determinants of Inefficiency in the US Dairy Farms. Journal
of Business and Economics Statistics, 9(3):279-286.
Kumbhakar, S.C. & Lovell, C.A.K. (2000).Stochastic Frontier Analysis. Cambridge:
Cambridge University Press.
Kuznet, S. (1965).Economic Growth and Structure. The American Economic Review,
63(3):247-258.
Latruffe, L. (2010). Competitiveness, Productivity and Efficiency in the Agricultural and Agri-
Food Sectors. OECD Food, Agriculture and Fisheries Papers No 30. OECD. Rennes.
France.
Lee, L.F. (1983). A Test for Distributional Assumptions for the Stochastic Frontier Functions.
Journal of Econometrics, 22(3):245-264.
Loehlin, J.C. (1992). Latent Variable Models. An Introduction to Factor, Path, and Structural
Analysis. Hillsdale: NJ Lawrence Erlbaum Associates.
Magingxa, L.L., Alemu, Z.G. & van Schalkwyk, H.D. (2006). Paper Prepared for
Presentation at the International Association of Agricultural Economists Conference.
Gold Coast, Australia, 12-18 August 2006.
Makombe, G., Namara, R., Hagos, F., Awulachew, S.B., Ayana, M. & Bossio, D. (2011).A
Comparative Analysis of the Technical Efficiency of Rain-fed and Smallholder Irrigation
in Ethiopia. International Water Management Institute. IWMI Working Paper 143.
Addis Ababa. Ethiopia.
Maseatile, M.S. (2011). Productivity of Small-Scale Maize Farmers in Lesotho. M.Sc (Agric.
Economics) Thesis. Department of Agricultural Economics. University of the Free State,
Bloemfontein.
Masterson, T. (2007).Productivity, Technical Efficiency and Farm Size in Paraguayan
Agriculture. Working Paper 490. Paraguay.
McNally, S. (2002).“Are Other Gainful Activities on Farms Good for the Environment?”.
Journal of Environmental Management, 66, 57-65.
115
Meeusen, W. & van den Broeck, J. (1977).Efficiency Estimation from Cobb-Douglas
Production Functions with Composed Error. International Economic Review, 18(2):435-
444.
Mellor, J. (1999). Faster, More Equitable Growth: The Relation between Growth in
Agriculture and Poverty Reduction. Agricultural Policy Development Project Research
Report No 4. USAID. Washington D.C.
Mengistu, A. (2006). Country Pasture/Forage Profile: Ethiopia. Office of Knowledge
Exchange, Research and extension. FAO. Rome. Italy.
Ministry of Agriculture (MoA).(2000). Agro-ecological Zonation of Ethiopia. Addis Ababa,
Ethiopia.
Ministry of Agriculture and Rural Development (MoARD).(2009). Agricultural Investment
Potential of Ethiopia. Addis Ababa. Ethiopia.
Ministry of Agriculture and Rural Development (MoARD).(2010). Ethiopia’s Agricultural
Sector Policy and Investment Framework Draft Final Report. Addis Ababa. Ethiopia.
Ministry of Education (MoE).(2008). National Report on the Development and State of the Art
of Adult Learning and Education. Addis Ababa. Ethiopia.
Ministry of Water Resources (MoWR).(2005). Irrigation (Potential) in Ethiopia. Ministry of
Water Resources, Addis Ababa. Ethiopia.
Mkhabela, T. (2005).Technical Efficiency in a Vegetable Based Mixed-cropping Sector in
Tugela Ferry, Msinga District, KwaZulu-Natal. Agrekon, 44(2):187-2004.
Morris, M.L., Risopoulos, J. & Beck, D. (1999).Genetic Change in Farmer-Recycled Maize
Seed: A Review of the Evidence. CIMMYT Economics Working Paper No 99-07. D.F.
Mexico.
Motsoari, C. (2012). Access to Credit and Agricultural Production in Lesotho. M.Sc. (Agric.
Economics) Thesis. Department of Agricultural Economics. University of the Free State,
Bloemfontein.
Mulat, D. (1999). The Challenge of Increasing Food Production in Ethiopia. Proceedings of
the Eighth Annual Conference on the Ethiopian economy. Nazareth, Ethiopia, 30
October - 1 November 1998, p49-72.
Mupanda, K. (2009). Market Access, Marketing Behaviour and Technical Efficiency among
Farming Households in Mozambique. M.Sc. (Agric. Economics) Thesis. Graduate
School. Purdue University, West Lafayette, Indiana.
Murillo-Zamoralo, H.L. (2004). Economic Efficiency and Frontier Techniques. Journal of
Economic Surveys, 18(1):33-77.
116
National Metrological Agency (NMA). (2007). Annual Monthly Rainfall in Tigray, 1997-
2006. Addis Ababa, Ethiopia.
Nedasa, B. (1999). Soil and Nutrient Management in Sub-Saharan Africa in Support of the
Soil Fertility Initiative. Proceedings of the Expert Consultation. Lusaka. Zambia.
Nega, B. (2003). Development Options for Ethiopia: Rural, Urban or Balanced?. In: Degefe,
B. & Nega, B. (eds).The Role of Urbanisation in the Socio-Economic Development
Process. Ethiopian Economic Association. Addis Ababa. Ethiopia.
Nigussie, M., Tanner, D. & Twumasi-Afriyie, S. (2002). Enhancing the Contribution of Maize
to Food Security in Ethiopia. Proceedings of the Second National Maize Workshop.
Addis Ababa, Ethiopia, 12-16 November 2001, p10-12.
Obwona, M. (2006).Determinants of Technical Efficiency Differentials amongst Small and
Medium-scale Farmers in Uganda: A case of Tobacco Growers. AERC Research Paper
152. Nairobi. Kenya.
Onasanya, R.O., Aiyelari, O.P., Onasanya, A., Oikeh, S., Nwilene, F.E. &. Oyelakin, O.O.
(2009). Growth and Yield Response of Maize (Zea mays L.) to Different Rates of
Nitrogen and Phosphorus Fertilisers in Southern Nigeria. World Journal of Agricultural
Sciences, 5(4):400-407.
Oromia Investment Commission.(2012).http://www.oromiainvest.gov.et/. Accessed [1 July
2013]
Oromia Region Map. (2013).http://www.bing.com/images/Oromia region map. Accessed [5
July 2013].
Owuor, J. (2000). Determinants of Agricultural Productivity in Kenya: Kenya Agricultural
Marketing and Policy Analysis Project. Tegemeo Kenya Agricultural Research Institute.
Michigan State University.
OXFAM.(2012). Country Profile Ethiopia. http://www.oxfam.ca/our-work/publications
publications-and-reports/ethiopia-country-profile. Accessed [10 May 2013].
Parikh, A., Ali, F. & Shah, M.K. (1995).Measurement of Economic efficiency in Pakistani
Agriculture. American Journal of Agricultural Economics, 77(3):675-685.
Pascoe, S. & Mardle, S. (2003). Efficiency Analysis in EU Fisheries: Stochastic Production
Frontiers and Data Envelopment Analysis. 2ndEdition. CEMARE: University of
Portsmouth.
Pastoral Community Development Program (PCDP). (2010). General Profile of the Oromiya
National Regional State. http://www.pcdp.gov.et/OromiyaRegionalProfile.pdf. Accessed
[20 May 2013].
117
Pattanayak, S.K., Mercer, D.E., Sills, E.O., Yang, J.C. & Cassingham, K. (2003). Taking
Stock of Agroforestry Adoption Studies. Agroforestry Systems, 57,173-186.
Pender, J. & Gebremedhin, B. (2006).Land Management, Crop Production and Household
Income in the Highlands of Tigray in Northern Ethiopia: An Econometric Analysis.
International Food Policy Research Institute. Washington D.C.
Pitt, M.M. & Lee, L.F. (1981). Measurement and Sources of Technical Inefficiency in the
Indonesian Weaving Industry. Journal of Development Economics, 9, 43-64.
Phillips, J.M. (1994). Farmer Education and Farmer Efficiency: A Meta-analysis. Economic
Development and Cultural Change, 43, 149-165.
Plessis, J. (2003). Maize Production. Department of Agriculture, Republic of South
Africa.www.nda.agric.za/publications. Accessed [10 May 2013].
Porcelli, F. (2009). Measurement of Technical Efficiency: A Brief Survey on Parametric and
Non-parametric Techniques.ww2.warwick.ac.uk/fac/soc/economics/staff/phd_students/
porcelli/porcelli_dea_sfm.pdf. Accessed [20 May 2013].
Rahman, S.A. & Umar, H.S. (2009).Measurement of Technical Efficiency and Its
Determinants in Crop Production in Lafia Local Government Area of Nasarawa State.
Journal of Tropical Agriculture, Food, Environment and Extension, 8(2):90-96.
Reifschneider, D. & Stevenson, R. (1991). Systematic Departures from Frontier: A
Framework for the Analysis of Firm Inefficiency. International Economic Review,
32(3):715-723.
Regional Agricultural Trade Expansion Support Programme (RATES). (2003). Maize Market
Assessment and Baseline Study for Ethiopia. Center for Regional Agricultural Trade
Expansion Support. Nairobi. Kenya.
Ritter, C. & Simar, L. (1997).Another look at the American Electrical Utility Data. Working
Paper. CORE and Institut de Statistique, Universite Catholique de Louvian, Louvian-la-
Neuve. Belgium.
Sahlu, Y. & Kahsay, M. (2002).Maize Seed Production and Distribution in Ethiopia.
Proceedings of the Second National Maize Workshop. Addis Ababa, Ethiopia, 12-16
November 2001, p160-165.
Sanchez, P.A. & Roland, J.B. (1997).Replenishing Soil Fertility in Africa. Proceedings of an
International Symposium. Soil Science Society of America, American Society of
Agronomy Madison. Wisconsin. USA.
Sarafidis, V. (2002).An Assessment of Comparative Efficiency Measurement Techniques.
Europe Economics Occasional Paper 2. London.
118
Schmidt, P. (1986). Frontier Production Functions. Econometric Reviews, 4, 289-328.
Schmidt, P. (2011). One-step and Two-step Estimation in SFA Models. J Prod Anal, 36, 201-
203.
Schneider, K. & Anderson, L. (2010).Yield Gap and Productivity Potential in Ethiopian
Agriculture: Staple grains and pulses. Evans School Policy Analysis and Research brief
No 98. University of Washington. USA.
Shephard, R.W. (1953). Cost and Production Functions. Princeton: Princeton University
Press.
Solis, D., Bravo-Ureta, B.E. & Quiroga, R.E. (2008).Technical Efficiency among Peasant
Farmers Participating in Natural Resource Management Programmes in Central
America. Journal of Agricultural Economics, 60(1):202-219.
Solwati, T. (2001).Establishing the Practical Frontier in Data Envelopment Analysis. PhD
Thesis. Department of Mechanical and Industrial Engineering. Faculty of Applied
Science and Engineering. University of Toronto, Canada.
Somali Regional State Summary (SRSS).(2004). Food Security Monitoring and Early
Warning.www.dppc.gov.et/Livelihoods/Somali/Downloadable/SNRSRegional
Summary.pdf. Accessed [2 July 2013].
Sori, A.T. (2009).Profile of Religious Experience in Ethiopia. Seminar Paper Presented at
Religious Pluralism and Public Presence in the US Summer Institute. Department of
Religious Studies. University of California. Santa Barbara.
Spielman, D.J., Kelemwork, D. & Alemu, D. (2011).Seed, Fertiliser and Agricultural
Extension in Ethiopia. ESSP II Working Paper No 20. International Food Policy
Research Institute. Addis Ababa. Ethiopia.
Stevenson, R.E. (1980). Likelihood Functions for Generalised Stochastic Frontier Estimation.
Journal of Econometrics, 13(1):57-66.
Tadesse, T. (2000).Drought and its predictability in Ethiopia. Drought: A Global Assessment.
Routledge, 1, 135-142.
Tadesse, Y. & Abidissa, G. (1996).Effect of Compost and NP Fertilisers on Growth and Yield
of Maize and Pepper. Proceedings of the Third Conference of Ethiopian Society of Soil
Science (ESSS). Ethiopian Science and Technology Commission, Addis Ababa,
Ethiopia, 28-29 February 1996.
Tesfaye, A., Githiri, M., Derera, J. & Debele, T. (2011).Subsistence Farmers’ Experience and
Perception about the Soil and Fertiliser Use in Western Ethiopia. Ethiopian Journal of
Applied Science and Technology, 2(2):61-74.
119
Thirtle, C., Irz, X., Lin, L., Mckenzie-hill, V. & Wiggins, S. (2001).Relationship Between
Changes in Agricultural Productivity and the Incidence of Poverty in Developing
Countries. Report No 7946. Department for International Development. London.
Uaiene, R.N. (2008). Determinants of Agricultural Technical Efficiency and Technology
Adoption in Mozambique. PhD Thesis. Department of Agricultural Economics. Purdue
University, West Lafayette, Indiana.
United Nation Development Organisation (UNDP). (2013). Human Development Report: The
Rise of the South. hdr.undp.org/en/media/HDR2013_EN_Statistics.pdf. Accessed [15
June 2005].
Weir, S. (1999).The Effects of Education on Farmer Productivity in Rural Ethiopia. CSAE
Working Paper Series 99-7. Oxford. Centre for the Study of African Economies.
World Bank. (2006). Ethiopia: Managing Water Resources to Maximise Sustainable Growth.
The World Bank Agriculture and Rural Development Department Report. Washington
DC. USA.
Yehuala, S. (2008).Determinants of Smallholder Farmers Access to Formal Credit. The case
of Metema Woreda, North Gonder. M.Sc. Thesis. Department of Rural Development and
Agricultural Extension. Haramaya University, Ethiopia.
Zikhali, P. (2008). Fast Track Land Reform and Agricultural Productivity in Zimbabwe.
Working Papers in Economics No 322. University of Gothenburg. Sweden.
Zerfu, D. & Larson, D. (2011).Incomplete Markets and Fertiliser Use: Evidence from
Ethiopia. Policy Research Working Paper 5235. The World Bank Development
Research Group. Ethiopia.
120
APPENDIX A ALTERNATIVE PRODUCTION FUNCTIONS
FITTED
Linear Production Function
Dependent Variable: YIELD
Method: Least Squares
Date: 04/03/14   Time: 11:33
Sample: 1 438
Included observations: 438
Variable Coefficient Std. Error t-Statistic Prob.
C 2.561004 0.078195 32.75169 0.0000
SEED -0.000752 0.000426 -1.763653 0.0785
N 0.004402 0.001172 3.756726 0.0002
LABOUR 5.14E-05 7.24E-05 0.709852 0.4782
R-squared 0.033048 Mean dependent var 2.614772
Adjusted R-squared 0.026364 S.D. dependent var 1.504574
S.E. of regression 1.484608 Akaike info criterion 3.637270
Sum squared resid 956.5630 Schwarz criterion 3.674550
Log likelihood -792.5620 F-statistic 4.944380
Durbin-Watson stat 1.602707 Prob(F-statistic) 0.002190
Cobb Douglas Production Function
Dependent Variable: LOGYIELD
Method: Least Squares
Date: 04/03/14   Time: 11:07
Sample: 1 438
Included observations: 438
Variable Coefficient Std. Error t-Statistic Prob.
C 0.328004 0.086793 3.779157 0.0002
LOGSEED 0.043920 0.049188 0.892903 0.3724
LOGN 0.023953 0.009388 2.551498 0.0111
LOGLABOUR -0.019654 0.036290 -0.541570 0.5884
R-squared 0.159190 Mean dependent var 0.313842
Adjusted R-squared 0.091170 S.D. dependent var 0.406840
S.E. of regression 0.404981 Akaike info criterion 1.039136
Sum squared resid 71.18011 Schwarz criterion 1.076417
Log likelihood -223.5709 F-statistic 2.340194
Durbin-Watson stat 1.801217 Prob(F-statistic) 0.072782
121
Translog Production Function
Dependent Variable: LOGYIELD
Method: Least Squares
Date: 04/03/14   Time: 11:14
Sample: 1 438
Included observations: 438
Variable Coefficient Std. Error t-Statistic Prob.
C 0.617984 0.225018 2.746380 0.0063
LOGSEED 0.263064 0.188058 1.398847 0.1626
LOGN 0.091842 0.054552 1.683589 0.0930
LOGLABOUR -0.453198 0.179556 -2.523990 0.0120
LOGSEED^2 -0.136399 0.073819 -1.847735 0.0653
LOGN^2 0.010341 0.018482 0.559506 0.5761
LOGLABOUR^2 0.059509 0.049594 1.199930 0.2308
LOGSEED*LOGN -0.002389 0.024812 -0.096269 0.9234
LOGSEED*LOGLABOUR 0.090859 0.097687 0.930098 0.3528
LOGN*LOGLABOUR -0.023058 0.018641 -1.236961 0.2168
R-squared 0.033936 Mean dependent var 0.313842
Adjusted R-squared 0.013621 S.D. dependent var 0.406840
S.E. of regression 0.404059 Akaike info criterion 1.048056
Sum squared resid 69.87692 Schwarz criterion 1.141257
Log likelihood -219.5242 F-statistic 1.670528
Durbin-Watson stat 1.820984 Prob(F-statistic) 0.093806
122