probit and Stochastic Frontier (SF) model was done. ... productivity calculated of 7.14 (8.45 and 7.01 for adopters and non-adopters ...... Adoption is the full use of an innovation as the best course .... Transfer of information reduces uncertainty about the performance of a new technology ...... Free press, New York, USA. pp.10-.
!!
!" #$ % &' ( #% &&( ( #)( ( # % # * ) +,,-
! "#$$
! "#$$
%&% # &
!"#$ % "
! & ' & ( )*! &
" " *)+,-!
. $ ! / " 01
" . 2,)3 % "
! & ' & ( " " 2,)3
ADOPTION OF IMPROVED MAIZE VARIETIES AND TECHNICAL EFFICIENCY AMONG SMALLHOLDER FARMERS IN KABAROLE DISTRICT
BY RODGERS MUTYEBERE REG. NO: 2011/HD02/4164U B.VOCATIONAL STUDIES IN AGRIC./EDUC. (KYAMBOGO UNIVERSITY)
A THESIS SUBMITTED TO THE DIRECTORATE OF RESEACRH AND GRADUATE TRAINING IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE AWARD OF MASTER OF SCIENCE DEGREE IN AGRICULTURAL ECONOMICS OF MAKERERE UNIVERSITY
MARCH 2016
DECLARATION I, Rodgers Mutyebere, hereby declare that this thesis is my own and has never been submitted for any award in any university. Where other sources of information are used, they have been duly acknowledged.
...................................................
.........................................
Name and Signature
Date
This thesis has been submitted for examination with our approval as internal supervisors.
................................................................
..........................................
)LUVWVXSHUYLVRU¶V6LJQDWXUH'DWH Dr. Jackline Bonabana-Wabbi
«««««««««« 6HFRQGVXSHUYLVRU¶V6LJQDWXUH'DWH Dr. Ateenyi Twaha
ii
DEDICATION To my wife Monica Komucunguzi Akiiki and children Raphel Mwesigwa Araali, Ruth Komukama Akiiki and Rachel Komujuni Atwooki.
iii
ACKNOWLEDGEMENT Special thanks to my academic supervisors: Dr. Jackline Bonabana-Wabbi and Dr. Ateenyi Twaha who tirelessly guided me to produce this thesis. I would like to extend my sincere thanks to the Embassy of Belgium for the financial assistance, especially in my second year of study. I would also like to acknowledge the efforts of those who played particular roles in data collection, entry and guiding me during econometric analyses. Special thanks to Mr. Ronald Kabbiri, Ms. Sheila Alice Nanyanzi, Mr. Edwin Akugizibwe and Dr. Moses Muhumuza all of Mountains of the Moon University whose guidance helped me a lot throughout this research. Finally, I would like to thank smallholder maize farmers in Kabarole District who dedicated their precious time to give lengthy information about maize production during the survey.
iv
ABSTRACT Due to low maize productivity and dwindling resources such as land and capital, studies about adoption of improved agricultural technologies and Technical Efficiency (TE) are necessary. As such, this study examined adoption of improved maize varieties and Technical Efficiency among smallholder farmers in Kabarole District. The specific objectives of the study were: to examine the socio-economic factors determining adoption of improved maize varieties, to estimate and compare TE for adopters and non-adopters of improved maize varieties, to analyze the determinants of TE among maize farmers and to estimate and compare output response to various inputs improved and local maize, thereby determining the multi-factor productivity. Primary data were collected from 160 respondents taken from four sub-counties in the district. Maximum Likelihood Estimation (MLE) of parameters in the probit and Stochastic Frontier (SF) model was done. Input elasticities of inputs were obtained by estimating the Cobb-Douglas model. Formal education, income, farm sizes and land ownership were found to be correlated with adoption of improved maize. Maize farmers in the sample area operated below their production frontiers (45.6 percent mean TE) indicating inefficiencies. However, adopters operated at a higher frontier (50.3 percent TE) than nonadopters (44.2 percent TE). Sources of inefficiency established were: income, credit and family size, formal education, male-headed households, off-gate market and land ownership LQFUHDVHG IDUPHUV¶ 7( ,QSXWV VWXGLHG ZHUH PRUH HODVWLF IRU LPSURYHG PDL]H 0XOWL-factor productivity calculated of 7.14 (8.45 and 7.01 for adopters and non-adopters respectively) indicates lack of optimal use of all inputs. The study indicates the need to adopt all inputs as a package if they are to obtain potential yield. In terms of policy, this study recommends that making improved inputs affordable and enKDQFLQJ VPDOOKROGHU IDUPHUV¶ LQFRPH ZRXOG LPSURYHIDUPHUV¶DGRSWLRQDQG7(
v
LIST OF ACRONYMS APP
Average Physical Product
C-D
Cobb-Douglas
DSIP
Development Strategy Investment Plan
GDP
Gross Domestic Product
MAAIF
Ministry of Agriculture, Animal Industry and Fisheries
MDIs
Micro-Deposit Institutions
MLE
Maximum Likelihood Estimation
MPP
Marginal Physical Product
Mt
Metric tons
NARO
National Agricultural Research Organization
NAADS
National Agricultural Advisory Services
OLS
Ordinary Least Squares
OPVs
Open Pollinated Varieties
SACCOs
Saving and Credit Cooperative Organizations
SF
Stochastic Frontier
SPSS
Statistical Package for Social Scientists
SSA
Sub-Saharan Africa
t/ha
tons per hectare
TE
Technical Efficiency
UBOS
Uganda Bureau of Statistics
UNHS
Uganda National Housing Survey
VIF
Variance Inflation Factor
WFP
World Food Programme
vi
TABLE OF CONTENTS Contents
Page
DECLARATION ......................................................................................................................... ii DEDICATION ............................................................................................................................ iii ACKNOWLEDGEMENT .......................................................................................................... iv ABSTRACT ................................................................................................................................. v LIST OF ACRONYMS .............................................................................................................. vi LIST OF TABLES ....................................................................................................................... x LIST OF FIGURES .................................................................................................................... xi 1.0 CHAPTER ONE: INTRODUCTION ................................................................................ 1 1.1 Background to the study ........................................................................................................ 1 1.2 Problem Statement ................................................................................................................. 7 1.3 Objectives of the study........................................................................................................... 8 1.3.1 Specific objectives .............................................................................................................. 8 1.4 Hypotheses ............................................................................................................................. 8 1.5 Justification of the study ........................................................................................................ 9 2.0 CHAPTER TWO: LITERATURE REVIEW ................................................................. 11 2.1 The Maize subsector in Uganda ........................................................................................... 11 2.2 Adoption of improved technologies..................................................................................... 13 2.2.1 The theory of adoption of improved technologies ............................................................ 14 2.2.2 Theoretical framework for measuring adoption of improved technologies ...................... 18 2.2.3 Factors affecting adoption of improved technologies ....................................................... 20 2.3 Efficiency in farming ........................................................................................................... 26 vii
2.3.1 Technical Efficiency ......................................................................................................... 27 2.3.2 Theoretical framework for measuring Technical Efficiency ............................................ 28 2.3.3 Factors affecting Technical Efficiency ............................................................................. 30 2.4 Maize output response to inputs .......................................................................................... 35 2.4.1 Theoretical framework for measuring output response to inputs ..................................... 36 2.4.2 Fertilizer use and maize output ......................................................................................... 38 2.4.3 Plot size and maize output ................................................................................................ 39 2.4.4 Herbicide use and maize output ........................................................................................ 40 2.4.5 Labour use and maize output ............................................................................................ 40 2.4.6 Maize seed and output ...................................................................................................... 41 3.0 CHAPTER THREE: METHODOLOGY........................................................................ 43 3.1 Empirical models ................................................................................................................. 43 3.1.1. Examining factors that affect adoption of improved maize varieties .............................. 43 3.1.2. Estimating Technical Efficiency ...................................................................................... 44 3.1.3. Estimating maize output response to inputs..................................................................... 46 3.1.4 Description of variables and apriori expectation .............................................................. 47 3.2 Data analysis ........................................................................................................................ 51 3.3 Study area............................................................................................................................. 52 3.4 Sampling procedure and sample size ................................................................................... 53 3.5 Data Collection .................................................................................................................... 55 3.6 Data Validity and Reliability ............................................................................................... 55 4.0 CHAPTER FOUR: RESULTS AND DISCUSSION ...................................................... 57 4.1 General descriptive analysis ................................................................................................ 57 viii
4.2.2 Multivariate analysis of the socio-economic factor affecting adoption ............................ 61 4.3 Technical Efficiency among smallholder maize farmers..................................................... 63 4.3.1 Technical Efficiency for adopters and non-adopters ........................................................ 63 4.3.2 Results of the Stochastic Frontier and inefficiency models .............................................. 65 4.3.3 Factors determining the level of Technical Efficiency ..................................................... 67 4.4 Response of maize output to inputs for adopters and non-adopters (C-D model) ............... 69 5.0 CHAPTER FIVE: SUMMARY, CONCLUSIONS AND RECOMMENDATIONS ... 72 5.1 Summary .............................................................................................................................. 72 5.2 Conclusions .......................................................................................................................... 73 5.3 Recommendations ................................................................................................................ 74 5.4 Suggestions for further research .......................................................................................... 75 REFERENCES ......................................................................................................................... 76 APPENDICES .......................................................................................................................... 83 Appendix A: Map of Uganda showing location of Kabarole District ....................................... 83 Appendix B: Kabarole District map .......................................................................................... 84 $SSHQGL[& )DUPHUV¶TXHVWLRQQDLUH ....................................................................................... 85 Appendix D: Detailed Results-Tables ....................................................................................... 89
ix
LIST OF TABLES Page
Tables
Table 1: Decomposition of GDP (%) for Uganda ..................................................................... 1 Table 2: Export value of selected crops in 2009 in Uganda ...................................................... 3 Table 3: Maize types and their attributes ................................................................................... 6 Table 4: Description of variables in the probit and Technical inefficiency models showing apriori expectation ................................................................................................................... 49 Table 5: Socio-economic and input factors for adopters and non-adopters............................. 58 Table 6: Results of Maximum Likelihood Estimation for the probit model............................ 62 Table 7: Difference in Technical Efficiency between adopters and non-adopters .................. 65 Table 8: Results of Maximum Likelihood Estimation of the Stochastic Frontier and inefficiency models .................................................................................................................. 66 Table 9: Input elasticity for all inputs, seed and labour type ................................................... 69
x
LIST OF FIGURES Page
Figures
Figure 1: Adopter categories based on innovativeness ............................................................ 17 Figure 2: Multi-stage sampling for the study area ................................................................... 54 Figure 4: Percentage Technical Efficiency for adopters and non-adopters ............................. 64
xi
1.0 CHAPTER ONE: INTRODUCTION 1.1 Background to the study In Africa, the agricultural sector employs about 60 percent of the labour force. The sector is among the significant contributors of Gross Domestic Product (GDP) for most countries in Sub-Saharan Africa (SSA) (Muzari et al., 2012). According to Alila and Atieno (2006), in Kenya in 2005, the sector contributed about 25 percent, in Tanzania in 2010 the sector contributed about 28 percent (Mashindano et al., 2011)
and in Uganda in 2011/12, it
contributed 22.5 percent to GDP (UBOS, 2013). +RZHYHUFRPSDUHGWRLQGXVWU\DQGVHUYLFHVHFWRUVDJULFXOWXUDOVHFWRU¶VFRQWULEXWLRQWR*'3 KDV GHFOLQHG LQ WKH UHFHQW \HDUV 7DEOH 7KLV LV SDUWO\ GXH WR WKH VHFWRU¶V VORZ JUowth estimated at 2 percent in 2010 (Bategeka et al., 2013). Compared to the growth of other sectors such as indusWU\¶V SHUFHQW DQG VHUYLFHV¶ SHUFHQW DJULFXOWXUH ODJJHG EHKLQG between 2002 and 2009, registering only 1.7 percent growth per annum (MAAIF, 2011). Table 1: Decomposition of GDP (%) for Uganda Sub-sector Agriculture (Value-added) Industry (Value-added) Service, etc (Value added)
2005 22 27 50
2006 21 27 52
2007 20 28 52
2008 19 29 52
2009 20 29 51
2010 21 28 51
Source: UBOS (2011) The slow growth of the agricultural sector is a concern to all stakeholders, including policymakers, Non-Governmental Organizations (NGOs) and scientists. This is because in Uganda agriculture employs up to 73 percent of the rural population. The subsector also accounts for about 50 percent of merchandise export and therefore a very important foreign exchange earner (Bategeka et al., 2013).
1
The agricuOWXUDOVHFWRU¶VVORZJURZWKWUDQVODWHVLQWRORZIRRGSURGXFWLRQ8QOLNHPRVWSDUWV RI66$8JDQGD¶VIRRGSURGXFWLRQKDVUHPDLQHGVWDJQDQW SHUFHQWSHUDQQXP FRPSDUHG to world production levels estimated at 3.9 percent (MAAIF, 20110). Surprisingly, the FRXQWU\¶VSRSXODWLRQJURZWKLVRQHRIWKHZRUOG¶VKLJKHVWHVWLPDWHGDWSHUFHQWSHUDQQXP (UBOS, 2013). This mirrors increased food insecurity which may lead to Malthusian Population trap if immediate interventions are not explored. Food security is among the current contentious challenges facing the world, especially in SSA and Asian countries. Hunger is number one on the list of the world's top ten health risks killing more people every year than AIDS, malaria and tuberculosis combined (FAO, 2012). According to WFP (2013), food security is the situation where people have all-time access to sufficient, safe and nutritious food for a healthy and active life. However, majority of the rural population in Uganda, especially children, women and elderly, consume insufficient food quantities that are nutritionally inadequate due to deficiencies in major nutrients such as proteins (Bouagnimbeck and Ssebunya, 2011). Therefore, there is need to improve agricultural production in Uganda through increasing production and productivity of major staple food crops such as maize (Zea mays) to ensure food security and improved livelihoods of the rural population. While production examines total output (Kg) obtained from a given crop, productivity is defined as the ratio of the value of total farm output to the value of total inputs used in production (Saweda, 2011). Whatever increase in agricultural production in Uganda , it has been due to area expansion rather than adoption of agricultural technologies and efficiency among farmers (Sserunkuuma, 2005). However, as production resources such as land become scarce, there is need to explore not just adoption of improved technologies but also efficiency in the use of improved inputs and output elasticities. 2
Maize (Zea mays) LV RQH RI 8JDQGD¶V PDMRU QRQ-traditional export and food security crops (Agona and Muyinza, 2005). The crop has main export markets in the neighbouring countries such as Kenya and Southern Sudan. For example, in 2009, maize export was estimated to have generated over USD 16.0 million from 2,355 Metric tons (Mt), being the largest export earner among non-traditional export crops and second largest to sugar in overall export crops as shown in Table 2 (MAAIF, 2011).
The crop also had more informal export value
estimated at USD 45.83 million in 2010 (FAO, 2012). Table 2: Export value of selected crops in 2009 in Uganda Crop
Export value (Million USD)
Maize
16,002
Legumes
3,160
Sugar and sugar confectionary
38,504
Tobacco
336
Coffee
412
Source: MAAIF (2011) Maize derives its importance as a major export crop because it is demanded by many countries for various uses. The crop can easily be processed into flour for making a paste locally known as Posho, can also be eaten as green or roasted, processed to make animal feeds such as silage and maize bran. The crop has a lot of unexploited potential uses, for example, making cooking oil, bio-fuel (which is an alternative to gasoline) and starch syrup (Agona and Muyinza, 2005). Exploiting these potential uses, would solve most problems faced by the rural people such as inadequate capital to purchase cooking oil, starch syrup and use of bio-fuel would conserve the environment by replacing wood fuel. $ERXWSHUFHQWRI8JDQGD¶VUXUDOSRSXODWLRQJURZVPDL]H 2EDDet al., 2005). The crop is among the five priority crops (including banana, rice, cassava and millet) targeted countrywide under the Development Strategy Investment Plan (DSIP) of the Ministry of 3
Agriculture, Animal Industry and Fisheries (MAAIF) (MAAIF, 2010). The crop is also the PRVWWUDGHGXQGHUWKH:DUHKRXVH5HFHLSW6\VWHP :56 VLQFHWKHV\VWHP¶VLQFHSWLRQLQWKH country in 2006 (Okoboi, 2011). Despite the importance, maize productivity is low and production is mainly at subsistence level, characterized by low adoption of improved technologies, reliance on natural soil fertility and rainfall. These and other factors have led to low maize productivity estimated at 1.94 tons per hectare (t/ha) per season compared to the SRWHQWLDO\LHOGRIWKDDQGWKDWRIWKHZRUOG¶VPDMRUFRUQSURGXFLQJFRXQWULHVOLNH8QLWHG States of America (USA) estimated at 8.2 t/ha (Obaa et al., 2005). Adoption of improved technologies such as improved maize varieties together with efficiency in their production could be determined by a number of factors some of which may be socioHFRQRPLF VXFK DV IDPLO\ VL]H IDUPHUV¶ DJH DQG H[SHULHQFH LQFRPH HGXFDWLRQ ODQG ownership and access to credit, extension and market. The rate and intensity of adoption may also depend on the superiority of variety-specific attributes of an improved over the local variety. However, other factors may be external like adaptability to local environmental conditions such as soil and climate. In Uganda, various improved maize varieties are being grown replacing traditional types. According to MAAIF (2012), National Agricultural Research Organization (NARO) had, by 2011, produced and disseminated about twelve (12) improved maize varieties grouped into Open Pollinated Varieties (OPV) and Hybrid. The OPVs include Longe 1, Longe 3, Longe 4, Longe 5 (Nnalongo or High Quality Protein maize) and Longe 6 (Ssalongo). Hybrid varieties include Longe 2H, Longe 5H, Longe 9H, Longe 10H, Longe 7H-IR (Kayongo) and Kawanda Composite varieties (Okoboi, 2011). An improved maize variety is a maize population that has been scientifically bred and conforms to the International Union for the Protection of New Plant Varieties (IUPV) criteria 4
of being distinct, uniform and stable (Bellon et al., 2006). However, as used in this study, an improved maize variety refers to Longe series and/ or Hybrid maize which was purchased from a certified seed company and planted by farmers for the first time. Local maize, on the other hand, is the one selected from previously grown and home-saved seed regardless of whether it was saved from Longe series (OPVs) or Hybrid. The farmers that planted an improved maize variety are referred to as adopters and those that planted local maize are referred to as non-adopters. Improved maize varieties are generally high yielding and therefore superior to local maize grown in Uganda. This is perhaps due to higher response to other inputs such as fertilizers and herbicides by improved compared to local maize. According to Bouagnimbeck and Ssebunya (2011), improved varieties can also relatively withstand the harsh local environmental conditions such as drought and foliar diseases and contain higher proportion of protein and calories than local maize (Mignouna et al., 2010). In addition, the varieties are early maturing, which is a likely remedy to food insecurity facing the country and other parts of SSA. Improved OPV and Hybrid varieties differ in regard to characteristics such as yield potential, maturity period and adaptability to specific growing conditions like drought, pests and diseases (Bouagnimbeck and Ssebunya, 2011). Hybrid varieties have a higher yielding potential than OPVs (Muzari et al., 2012). However, Hybrid varieties are expensive because farmers have to buy new seed every planting season, if they are to obtain reasonable yield (MAAIF, 2012). On the other hand, seeds from an OPV can be replanted up to three seasons without major drop in yield meaning that the farmer can buy seeds once in every three seasons. However, for this research, both Hybrid and OPV planted twice are considered local. According to Beinempaka et al. (1990), Hybrid maize is produced by crossing inbred 5
varieties using controlled pollination, while OPVs (Longe series) are selected from crosses made between varieties of diverse genetic composition using open pollination. Therefore, compared to Hybrids, OPVs are not pure lines but crosses. Attributes for each category of maize type are as shown in Table 3. Table 3: Maize types and their attributes Attributes
Desirable
Maize type Local
OPVs (Longe series)
Hybrids
-The farmer can plant home-
-Suitable for all areas except
-Suitable
saved seeds
highlands
including highlands
-Seeds are adapted to local
-Relatively resistant to foliar
-Early maturity (100-120
environment
diseases
days)
-Varieties are adapted to low
-Contains
nutrient
protein
-Storage pest resistant
cooking habits
-Early maturity (105-115
-Lodging resistant
-Locally available
days)
-Easy to shell
-Can withstand local pests and
-Seeds can be recycled for
-Highest yield
disease pressure
three seasons
-Large Kernels
-Low yielding
-Mature at different times
-Seeds are very expensive
-Less resistant to new pest and
-Not resistant to storage
to buy
disease problems
pests
-Not tasty
-Mature at different times
-Small kernels
-Highest
-Not resistant to storage pests
-High nutrient needs
requirement
-Not locally available
-Not locally available
levels
and
farmer
high
quality
for
all
areas
-Drought resistant
-Mature at the same time Undesirable
nutrient
-Cannot grow in highlands
Source: Bouagnimbeck and Ssebunya (2011) and MAAIF (2012) Because of the aERYHDGYDQWDJHV1$52¶VHIIRUWVWREUHHGDQGGLVVHPLQDWHLPSURYHGPDL]H varieties are in a positive direction towards solving the problem of food insecurity in Uganda. +RZHYHU DGRSWLRQ RI WKHVH YDULHWLHV SHU VH LV QRW D FRPSOHWH UHPHG\ ,QFUHDVLQJ IDUPHU¶s efficiency in their production is equally important if productivity is to be realized fully. The farm is efficient if it is not possible to produce more output without using more units of inputs. Studies on efficiency are necessary in developing countries such as Uganda where 6
production resources are becoming too scarce to support the needs of increasing population. Consequently, most parts of the country will be hit by severe food insecurity (UBOS, 2013). Thus, the need to improve efficiency in the use of few available improved technologies as opportunities for developing and adopting better technologies significantly decline. Adoption and efficiency are alternatives to area expansion for productivity growth. However, the two have received little research attention in the country. 1.2 Problem Statement In spite of the remarkable efforts by the government of Uganda to increase food production, productivity of major staple crops such as maize is still lower than potential yields (Kasirye, 2013). There is need for a multi-faceted approach to increase productivity, not only through increased acreage but also adoption of improved varieties and Technical Efficiency among smallholder farmers. The question of adoption of improved maize varieties and TE among smallholder maize farmers in Kabarole District is four-fold: what socio-HFRQRPLFIDFWRUVGHWHUPLQHWKHIDUPHUV¶ decision to adopt improved maize; for those who adopt (adopters), are they more Technically Efficient than those who do not (non-adopters); what factors determine the level of TE; finally, is output response to input quantities for improved maize different from that of local maize? This empirical study was set to suggest answers to those questions which can help to enhance adoption of improved maize and Technical Efficiency. The study will also suggest approaches to help smallholder farmers to produce at or close to their production frontiers. Studies conducted in Uganda, including that done by Obaa et al. (2005), Hyuha (2006), Asiimwe (2007), Kibirige (2008), Mugisha and Diiro (2010) and Okoboi (2010) assessed adoption of improved maize , TE and output response to inputs separately yet the three go hand in hand. In addition, these studies were conducted in other parts of the country, yet 7
VRFLDO V\VWHPV ZKLFK DIIHFW LQGLYLGXDOV¶ LQQRYDWLYHQHVV DUH KHWHURJHQHRXV DQG G\QDPLF (Sahin, 2006). No empirical research compared Technical Efficiency and input elasticity for adopters and non-adopters of improved maize, thus motivating this study in Kabarole District, which is one of the major maize growing districts in Uganda. 1.3 Objectives of the study The general objective of this study was to assess adoption of improved maize varieties and Technical Efficiency among smallholder farmers in Kabarole District. 1.3.1 Specific objectives The specific objectives of the study were: 1. To determine the socio-economic factors that affect adoption of improved maize varieties (Longe series and Hybrid) in Kabarole District 2. To estimate and compare Technical Efficiency for adopters and non-adopters of improved maize varieties in Kabarole District 3. To analyze determinants of Technical Efficiency among maize farmers in Kabarole District 4. To estimate and compare maize output response to input quantities, namely; nitrogenous fertilizer, seeds, labour, plot size and herbicides for improved and local maize in Kabarole District and thereby determine the multi-factor productivity in maize crop. 1.4 Hypotheses 1. )DUPHUV¶HGXFDWLRQSRVLWLYHO\LQIOXHQFHV DGRSWLRQZKLOHIDPLO\VL]HQHJDWLYHO\DIIHFWV adoption of improved maize varieties. 8
2. Adopters of improved maize varieties in Kabarole District are more technically efficient than non-adopters. 3. Off-gate market for maize positively influences the of smallholder maize farmers while EHLQJPDOHQHJDWLYHO\LQIOXHQFHVIDUPHUV¶7( 4. Multi-factor productivity is higher for improved than local maize. 1.5 Justification of the study Maize is a major food security and cash crop for most people in rural areas of East Africa. In the DSIP by MAAIF, maize was the only selected cereal among the ten strategic crops in the 5-year plan (Okoboi, 2010). However, over the past decade, there has been low maize productivity, which could be attributed to a number of factors such as small-scale production, ORZ DGRSWLRQ RI LPSURYHG WHFKQRORJLHV DQG IDUPHUV¶ WHFKQLFDO Lnefficiency among others (Obaa et al., 2005). This justifies a research on output response to inputs as well as determinants of adoption of improved maize and Technical Efficiency as important factors GHWHUPLQLQJ WKH FURS¶V SURGXFWLYLW\ 0HUH DGRSWLRQ RI LPSURYHG WHFKQRORJLHV LV QRW sufficient, but if combined with efficiency, productivity growth will be assured. While many other factors determine adoption and TE, it is important to examine socio-economic factors such as education, family size, household income, sex, land ownership and institutional factors like access to extension services, maize markets and credit as indicated by Danilo (2004). Findings on yield response, adoption of improved maize varieties and Technical Efficiency among smallholder farmers in Kabarole District will be used as a case study and should be important to stakeholders mainly researchers, local governments, extension workers, policyPDNHUVDQGPDL]HIDUPHUV5HVHDUFKHUVZLOOEHDEOHWRH[SODLQGLVSDULWLHVEHWZHHQIDUPHUV¶ 9
actual and potential yield. Policy makers will understand constraints to adoption of improved maize and efficiency of smallholder farmers and address them. Farmers will be able to address socio-economic constraints such as low education that hinder them from producing at their production frontiers. Finally, the study should help extension workers and local governments to adopt practical approaches for information dissemination about improved YDULHWLHVDQGHQVXUHWKDWWKLVLQIRUPDWLRQGLIIXVHVWRUXUDODUHDVFRQVLGHULQJIDUPHUV¶VRFLReconomic constraints.
10
2.0 CHAPTER TWO: LITERATURE REVIEW 2.1 The Maize subsector in Uganda Maize is one of the major crops shouldering the agricultural sector. The crop is grown throughout the country. However, there are differences in the production of the crop by region. As indicated in the 2005/2006 Uganda National Housing Survey (UNHS), Eastern UHJLRQDFFRXQWHGIRUSHUFHQWRIWKHFRXQWU\¶VWRWDOFURSSURGXFWLRQ 8%26 $QGLQ the 2008/2009 UNHS by UBOS (2010), Eastern Uganda produced 1.1 million Mt (46.9 percent) followed by Western Uganda (21.1 percent), Central Uganda (19.1 percent) and Northern Uganda was the last with 12.9 percent. Between the year 1995 and 2009, there was a substantial increase in maize production from 537,000 Mt to 2,362,000 Mt as indicated in the 2008/09 UNHS (UBOS, 2010). Increased maize production implies that per capita crop consumption is expected to rise and Uganda is likely to be food secure. This is a positive JHVWXUHHVSHFLDOO\DV8JDQGD¶VSRSXODWLRQJURZWK DWSHUFHQWSHUDQQXP RXWSDFHVIRRG growth (1.7 percent per annum) as indicated by Okoboi and Barungi (2012). Maize production in Uganda varies from district to another. In the 1999/2000 UNHS by UBOS (2007) indicates that Iganga District produced 303,000 Mt, followed by Mubende (171,000 Mt), Soroti (138,000 Mt) and Kabarole District (125,000 Mt). By virtue of its location, Kabarole District has potential maize market in the neighbouring districts and Democratic Republic of Congo (DRC), but at current productivity (1.86 t/ha), the district cannot satisfy the available market. Hence the need for studies on approaches specific to the GLVWULFWIRUHQVXULQJLQFUHDVHLQWKHFURS¶VSURGXFWLYLW\ The annual domestic market for maize is estimated to range from 350,000-400,000 Mt per year. Kampala alone accounts for 50 percent of the formal internal maize market (UBOS, 11
2011). The difference between annual production and domestic consumption indicates that PRVWRI8JDQGD¶VPDL]HLVH[SRUWHG>PDL]HH[SRUWVHVWLPDWHGWRUDQJHEHWZHHQDQG 250,000 Mt per year as observed FAO (2012)]. However, potential export demand is higher WKDQ8JDQGD¶VDFWXDOPDL]HH[SRUWVGXHWRHPHUJLQJPDUNHWVLQWKH(DVW$IULFDQ&RPPXQLW\ (EAC), Southern Sudan, and Democratic Republic of Congo (DRC). Uganda has the potential to increase maize output beyond the current production levels. However, as noted by Sserunkuuma (2005), increase in maize production by smallholder farmers has been mainly due to area expansion. Through adoption of improved varieties and Technical Efficiency, productivity is expected to increase further in order to fill the existing export deficit indicated above. &RPSDUHG WR RWKHU FRXQWULHV¶ DYHUDJH IDUP VL]H VXFK DV WKDW RI *KDQD HVWLPDWHG DW ha/household/season (Akramov, 2012) 8JDQGD¶V DYHUDJH PDL]H SORW VL]H LV VWLOO VPDOO estimated at only 0.34 ha/household/season (Okoboi, 2011). This is partly blamed on the use of rudimentary technologies for example hand hoe and planting local seeds (Obaa et al., 2005). Poor technologies are blamed on the collapse of produce cooperatives which used to provide hoes, fertilizers and improved seeds alongside credit. Planting local maize remains a critical concern as it is low yielding, late maturing and less resistant to pests and diseases (MAAIF, 2012). Adoption of improved maize seed is hypothesized to mitigate this problem. While adoption of improved maize is considered as a remedy to problems of growing local maize, farmers should be guided on appropriate choice of the improved variety to ensure economic profitability and acceptance. A study by Bouagnimbeck and Ssebunya (2011) has VKRZQWKDWWKHUHLV\LHOGYDULDELOLW\ZLWKLQLPSURYHGYDULHWLHV8JDQGD¶VSRWHQWLDO\LHOGIURP Hybrid is estimated at 8t/ha/season compared to 5 t/ha for Longe series (MAAIF, 2012), but WKLVGHSHQGVRQIDUPHUV¶PDQDJHULDODELOLWLHVDQGDJUR-ecological zone. 12
Okoboi and Barungi (2012) showed that only 32 percent of sampled maize farmers in the 2008/09 census grew improved maize varieties. While Eastern Uganda had the highest adoption rate (45.7 percent), Western had lowest (18.2 percent). However, the study only covered Bushenyi, Kabale, Kisoro and Kanungu districts in Western Uganda but not Kabarole or any neighbouring districts. Earlier, Okoboi (2011) generated the analysis from the 2005/06 national census
and also indicated Western Uganda as having the lowest
DGRSWLRQ RI LPSURYHG PDL]H YDULHWLHV SHUFHQW FRPSDUHG WR WKH FRXQWU\¶V DYHUDJH RI 10.07 percent. 2.2 Adoption of improved technologies According to Mignouna et al. (2010), adoption of an innovation per se is not enough but it is important that people continuously make the adoption decision. Accepting a given innovation also may not necessarily lead to a sustainable increase in productivity. There is need for adaptation (the ability of the farmers to continue using the technology) rather than just adoption of the innovation. But this will depend on its superiority to existing ones, adaptation to local environment and socio-economic factors such as price of the technology, sex of the farmer, education and access to credit, extension and market (Bellon et al., 2006). For example, improved technologies may be available but not accessible due to high cost. But what matters most is the initial acceptance of an improved maize variety, which depends on the phenotypic (such as grain size, maturity period, taste and resistance to pest and diseases) rather than genotypic characteristics such as colour (Obaa et al., 2005). Subsequent adoption ZLOOGHSHQGRQWKHIDUPHUV¶FKRLFHEDVHGRQWKHPD[LPL]DWLRQRIH[SHFWHGXWLOLW\ Due to uncertainty associated with an innovation, Mignouna et al.(2010) emphasize the need for predictability of the outcome if a farmer is to adopt the technology. Lack of predictability of the outcomes is perhaps one of the reasons for low adoption of most agricultural 13
technologies such as improved maize varieties in Uganda. Farmers fear that the new idea may destabilize their socio-economic system (Obaa et al., 2005). Farmers take time to learn about the new technology before adoption (Kasirye, 2013). However, the length of time varies with the type of technology, person and place. This is because some farmers in developing countries like Uganda are risk-averse thus, tend to first observe if the new and untried technology edges the conventional practices (BonabanaWabbi and Taylor, 2008). Another reason for the time lag between first knowledge of the technology and its adoption is because productivity-improving technology should be a bundle of innovations rather than a single technical or managerial intervention (Germano et al., 2006). For example, adoption of an improved maize variety can lead to significant increase in crop productivity only if other improved technologies such as new planting methods, fertilizer and herbicide application are adopted. The package nature of technology makes the rate (time taken) and intensity (the level) of technology adoption difficult to evaluate (Kafle, 2010). This implies that smallholder farmers may adopt an improved technology but if not as a full technological package, the effect of the adopted technology may be minimal. 2.2.1 The theory of adoption of improved technologies Over the past years, there has been an extensive body of literature on the economic theory of WHFKQRORJ\DGRSWLRQ+RZHYHUGLIIXVLRQRILQQRYDWLRQV¶WKHRU\E\5RJHUV LVWKHPRVW popular adoption theoretical framework (Sahin, 2006). According to Rogers (2003), technology
and
innovation
are
synonymous.
Hycenth
et
al.
(2010)
define
technology/innovation as any new knowledge, equipment, methods, product, idea or practice introduced into and utilized in an economic or social system. 14
The innovation might have been introduced in social system long time ago but members did not have enough information about it. Therefore, there is need for appropriate channels for its diffusion. Diffusion of an innovation is the process by which it is communicated through certain channels over time among members of the social system (Rogers et al., 2007; Hycenth et al. 2010). Thus, if communication about an innovation is not effective or withdrawn at an early stage, adoption may be negatively affected. Effective and consistent communication about a technology plays a significant role in augmenting its adoption. Communication should be consistent (should not be withdrawn) to help potential adopters in the decision-making process. According to Rogers (2003), an individual goes through the chronological stages in the adoption process of knowledge (where the socio-economic characteristics, communication channel and personality factors are important) and persuasion (by the technology attributes such as relative advantage, compatibility, complexity, trialability and observability). Other stages are: decision which is characterized by adoption or rejection, implementation (characterized by continued adoption or discontinuation) and finally, confirmation which refers to post-adoption assessment of the technology where farmers seek for supportive messages about the decision. When an innovation is introduced in a social or economic system, there are two expected decisions: adoption or rejection. Adoption is the full use of an innovation as the best course of action available and rejection is the decision not to adopt an innovation (Sahin, 2006). As already noted, adoption of a technology is characterized by a lot of uncertainty because potential adopters are not sure about its profitability and performance in the local environment which are the main determinants. The uncertainty can be reduced through providing information about the innovation using the right communication channel such as media and extension services. Margriet et al. (2001) pointed out other two barriers to 15
adoption as inability and unwillingness to adopt. Inability to adopt may be due to high cost of obtaining information; technology may be too expensive, excessive labour requirements and inadequate managerial skills. Unwillingness to adopt may be because information about the technology conflicts with the current production system, ignorance or belief in traditional practices. Technology-specific attributes also determine the rate of adoption (Sibiya et al., 2013). For example, the aspect of relative advantage of the new technology compared to existing ones is important. Other characteristics are: compatibility which is the degree to which an innovation is consistent with the existing values, past experience and need of potential adopters. Trialability, which is the degree to which an innovation may be experimented, observability which deals with the degree to which the results of an innovation are visible to others, finally, complexity which is the degree to which an innovation is perceived as relatively difficult to understand and use (Sahin, 2006). Of all the attributes mentioned, only the last has a negative correlation to adoption of innovation. Adoption decision is not necessarily a binary process because intensity may change over time (Heike, 2012). Earlier adopters may discontinue and non-adopters may later adopt as the social system learns more about the technology and the socio-economic status changes. Because the decision involves substantial cost while benefits accrue over time, some technologies may be abandoned again 7KHUHIRUH DQ LQGLYLGXDO¶V DVVHVVPHQW RI WKH QHZ technology is subjective and dynamic (Margriet et al., 2001). This means adopting an innovation does not rule out rejecting it at a later stage and vice versa. Rogers (2003) identified two major adopter-categories based on the innovativeness. These are earlier adopters which include innovators, early adopters, early majority and later adopters that include the late majority and laggards as shown by the normal distribution curve in 16
Figure 1. The difference between earlier and later adopters is the socio-economic status, personality related factors and communication which are usually positively related to innovativeness (Sahin, 2006). In addition, cost and social status motivation aspects of an innovation are very important. Earlier adopters are more status motivated for adoption than the later adopters who perceive status as less significant.
Source: Sahin (2006) Figure 1: Adopter categories based on innovativeness In the study by Hycenth et al. (2010), innovators comprise of 2.5 percent and are individuals who are venturesome, educated and gatekeepers ready to bring the innovation in from outside the social system. Early adopters, who comprise 13.5 percent, are likely to be leaders, popular and educated. People come to this category for advice about the innovation. The early majority have deliberate and many social informal contacts. According to Sahin (2006), the late majority and laggards wait until their peers adopt the innovation. They are first skeptical about the innovation but adopt later due to economic necessity. However, the laggards only adopt after seeing the results of the innovation. Only adopter category of individual is illustrated above but not non-adopters.
17
2.2.2 Theoretical framework for measuring adoption of improved technologies Binary Choice Models (BCM), as indicated by Greene (2002) are used to analyze outcomes which are binary (dichotomous). In this case the respondent either adopts improved maize variety (Y=1) or does not (Y=0). A set of factors such as variety-specific attributes such as grain size, test, resistance and adaptability as well as socio-economic factors such as land ownership, sex, farm size, education, family size, access to credit, extension and market, which are gathered in x VKDSHWKHIDUPHU¶VGHFLVLRQVXFKWKDW Prob(Y=1|x) = F(xȕ DQG Prob(Y=0|x) = 1- F(xȕ where, Prob is the probability, Y is binary dependent variable; ȕ is a vector of parameters that reflect the impact of change in x on the probability estimated and x are explanatory variables. Since F(xȕ ( \_x), one can construct a Linear Probability Model (LPM) in that y= E(y |x) +[y- E(y |x)] = x¶ȕᖡ
(2)
and ᖡ is the error term. The major disadvantages of LPM is that estimated probabilities for binary responses cannot be constrained to lie between 0 and 1 and the model is estimated by using Ordinary Least Squares (OLS) where the error term is prone to heteroscedasticity (Soderbom, 2009). To overcome the above disadvantages of LPM, two non-linear models have been suggested. The common models are the logistic distribution (Logit) and the normal standard distribution, yielding the probit models. In the probit model, Probሺܻ ൌ ͳȁሻ ൌ ʣሺݔᇱ ߚሻ and Prob (Y=0|x)ൌ ͳ െ ʣሺݔᇱ ߚሻ.
(3) 18
where, ʣሺǤ ሻ is the standard normal distribution function. For a logistic cumulative distribution function; the logit model is obtained where; ᇲ
Probሺܻ ൌ ͳȁሻ ൌ
ೣ ഁ
ᇲ
ଵା ೣ ഁ
Prob ሺܻ ൌ Ͳȁሻ ൌ1-
ൌ ߉ሺݔǯߚሻ and
ᇲ
ೣ ഁ
ᇲ
ଵା ೣ ഁ
=1- ߉ሺݔǯߚሻ
(4)
where, ߉ሺǤ ሻ is the logistic cumulative distribution function. The strength of the probit and logit is that the two models can constrain the estimated probabilities to lie between 0 and 1 and are estimated using Maximum Likelihood Estimation (MLE) which gives unbiased and efficient estimates compared to OLS. Through iterations, MLE determines the direction and size of the probit estimates until a final log-likelihood is obtained. It also establishes a function that maximizes the probability of explanatory variables to predict an explained variable (Gujarati, 2004). For both the probit and logit models, ݈݅݉௫ఉ՜ஶ ሺܻ ൌ ͳȁሻ ൌ ͳ and ݈݅݉௫ఉ՜ିஶ ሺܻ ൌ Ͳȁሻ ൌ Ͳ
(5)
Although parameter estimates obtained by both the probit and the logit models differ due to the fact that the two distributions have different scales, it requires very large sample sizes for WKHWZRPRGHOVWREHVLJQLILFDQWO\GLIIHUHQW7KHUHIRUHLWLVXSRQRQH¶VFKRLFHWRXVHHLWKHU the probit or the logit model. However, the probit model was adopted for this study because parameters in the probit model are easier to interpret by considering a change in the probability (of an explained variable) per unit change in the explanatory variable.
19
2.2.3 Factors affecting adoption of improved technologies Factors affecting adoption of improved technologies could be economic, farmer-specific, institutional and technology-specific. Economic factors include expected benefits, income, price of implements, costs involved, profit and opportunity cost associated with the technology. Farmer-VSHFLILF FKDUDFWHULVWLFV LQFOXGH IDUPHUV¶ HGXFDWLRQ JHQGHU DQG IDPLO\ size. Institutional factors deal with the extent to which institutions impact on adoption of technologies by smallholder farmers (access-related factors) and include access to extension services, market and credit. Technology-specific factors refer to factor-input characteristics and include adaptability of the technology to local environment, consumptive traits such as taste as well as resistance to pests and diseases (these are beyond the scope of this research. Readers interested can get information from Okoboi (2011). Socio-economic variables among others are the most important determinants of technology DGRSWLRQ7KLVLVEHFDXVHPRVWDGRSWLRQOLWHUDWXUHKDVEHHQRQVRFLHW\LQGLYLGXDO¶VDWWLWXGH personality or socio-economic characteristics of the social system. Adoption of a technology can succeed if it respects culture, norms and values of the people receiving it (Mugisha and Diiro, 2010). And, it is bound to be rejected if it is radical and contradicts with aspects of social system. The influence of farmer-specific, institutional and economic factors on adoption of technologies is examined below: 2.2.3.1 Farmer-specific factors Most smallholder farms (up to 90 percent) in Africa depend on family labour for activities such as planting, weeding and harvesting (Gianessi and Williams, 2011). While various studies have established mixed relationships between family size and adoption of improved technologies, Angu (2004) noted that a larger family size is an advantage as it is a source of 20
family labour for opening up bigger farms that enhance adoption of improved technologies such as improved maize varieties. Large family sizes are also associated with more off-farm income and access to adoption enhancing factors such as credit, extension and education (Hyuha, 2006). Where land and labour markets exist, family size can positively or negatively affect adoption of improved varieties arguing that family size would have a positive effect on adoption if all members participate in maize production and a negative effect if not participating due to off-farm activities (Cheryl et al., 2003 and Kafle, 2010). However, Kudi et al. (2011) obtained a negative and significant relationship between family size and adoption of improved technologies. Education of household members is very important in ensuring adoption of improved technologies and efficiency among smallholder farmers (Kudi et al., 2011). The question is whose education matters among the household members. Most studies, including Mignouna et al. (2010) and Angu (2004), found out that education of household head matters most because he/she is at the top of decision making process for various farm activities. Their analysis further established that education of adopters of new maize variety was higher than for non-adopters. Educated farmers comprehend technical information about the new technology as well as extension leaflets and handbooks (Cheryl et al., 2003 and Kibaara, 2005). Hyuha (2006) established that education of all household members is important in adoption of improved tecKQRORJLHV ,I DOO KRXVHKROG PHPEHUV¶ HGXFDWLRQ LV KLJK LW LV associated with access to information, credit and other production technologies as well as offfarm income which augment adoption. Adoption decision for agricultural technologies also depends on the gender of the household head. Gender refers to the role attached to biological sex (whether male or female). Mignouna et al. (2010) established that 74 percent of households in Africa are male-headed. This 21
implies that there is more access to production resources such as land and information compared to female-headed households. Land and other resources can be used as collateral for acquisition of credit. Male farmers also have higher education and off-farm income, which are incentives to increased adoption. Saweda (2011) posted similar findings about gender and adoption of new technologies and argued that male farmers had higher access to extension services. This is because most extension workers are male and therefore malebiased (Germano et al., 2006). The challenge is that while men have relative advantage over women in terms of accessibility to resources, the later contribute the highest labour force for maize productivity (Cheryl, et al., 2003; Gianessi and Williams, 2011). In many adoption studies including that done by Kafle (2010), age is negatively associated with adoption of improved technologies. The negative association is due to the fact that older farmers tend to stick to their old technology because they are risk-averse. Kasirye (2013) also argued that farmers tend to withdraw from the technology after long use hence a negative association. On the other hand, Mignouna et al. (2010) established a positive and significant relationship between age and adoption of the new Striga resistant maize seed. They explained that older farmers accumulated more experience from maize cultivation in Striga infested areas over years and could differentiate risk involved between the past technologies and the new maize introduced to control the parasite. 2.2.3.2 Institutional factors Low extension visits is one of the reasons for low impact of agricultural research and inability by scientists to access, process, disseminate and exchange scientific information. Transfer of information reduces uncertainty about the performance of a new technology and PDNHVIDUPHUV¶DVVHVVPHQWREMHFWLYHUDWKHUWKDQVXEMHFWLYHKHQFHHQKDQFLQJDGRSWLRQ(Kafle, 2010). In Uganda, the National Agricultural Advisory Services (NAADS) programme was 22
mandated to offer extension services and market information to farmers, replacing the WUDGLWLRQDO H[WHQVLRQ V\VWHP +RZHYHU WKH SURJUDPPH¶V ZRUN ZDV MHRSDUGL]HG E\ SROLWLFDO influence, underfunding, corruption and influence peddling in the purchase of inputs (MAAIF, 2012; UBOS, 2010). Cheryl et al. (2003) and Tiamiyu et al. (2009) obtained a positive and statistically significant effect of extension services on adoption. They argued that farmers who planted improved varieties were likely to have more extension contact because many extension offices also supply improved inputs. The role of extension services is not worth compromising because it is thought to counter-balance the effect of low formal education on adoption of improved technologies. Access to markets is very important in enhancing adoption of improved technologies and it is two-fold: input market where input technologies such as improved maize are purchased and output market where maize grain is sold. Lack of enough output market information is responsible for low farm-gate prices in Uganda (Okoboi and Barungi, 2012). There is limited LQIRUPDWLRQDERXWDYDLODELOLW\RIEHWWHUSULFHVIRUIDUPHUV¶SURGXFHHOVHZKHUHKHQFHIDUPHUV sell their produce at loZSULFHV7KHVH /RZRXWSXWSULFHV DUHUHVSRQVLEOHIRUIDUPHUV¶ORZ motivation to adopt improved technology (Mignouna et al., 2010). Types of maize markets based on location are mainly categorized into farm-gate and off-gate markets. Farm-gate market is the type of market where up-country buyers buy the farm produce at the farm and therefore reduce transport costs. Off-gate markets are markets such as village assembly markets where farming communities assemble their produce on a specific day and sell their produce to traders who further sell it to other areas. These markets can be daily or weekly and operate at low capital but lack storage or warehouse facilities. However, access to ZDUHKRXVHV DFWV DV D SULFH LQFHQWLYH DQG LPSURYHV IDUPHUV¶ DFFHVV WR LPSURYed maize. Market information therefore, attracts high output and low input prices which encourages large-scale production and hence adoption of improved technologies (Germano et al., 2006). 23
But this is determined by other socio-HFRQRPLF IDFWRUV VXFK DV IDUPHUV¶ DFFHVV WR FUHGLW income and land ownership. &UHGLWLVLPSRUWDQWLQUDLVLQJWKHIDUPHU¶VFDSLWDODQGSXUFKDVLQJSRZHUIRULPSURYHGPDL]H varieties. A positive correlation between credit and adoption of improved technologies was statistically significant in only two out of twenty two estimation sites in Ethiopia (Cheryl et al., 2003). However, access and use of credit on farm activities are separate issues. Some farmers acquire credit but do not use it for agricultural production. Lack of access and use of credit is a major institutional constraint to adoption of improved technologies. This is because 8JDQGD¶V UXUDO ILQDQFLDO V\VWHPV DUH XQGHYHORSHG IUDJPHQWHG DQG QRW LQWHJUDWHG LQWR WKH formal financial sector (Agona and Muyinza, 2005). The problem of lack of access to credit worsened after the collapse of cooperative unions which were replaced by commercial banks. Most commercial banks are not rural-oriented, and have few loan products meant for smallscale farming. The few available products are directed towards commercial agriculture and agro-trading, charge high interest rates and are characterized by short repayment periods (Bategeka et al., 2013). Land ownership determines whether farmers adopt improved technologies or not. It (land ownership) refers to the rights and institutions that govern access and use of land (Eze et al., 2011). In a study by Chirwa (2008), about 57 percent landowners in Malawi planted improved varieties. This is perhaps because they save money that would be for rent to purchase improved seed whose cost is usually high. According to Bategeka et al. ( 2013), renting land is one of the constraints to adoption of improved technologies in Uganda, EHFDXVH LW UHGXFHV IDUPHU¶V LQFRPH 6RPHWLPHV HYHQ ODQG RZQHUV GR QRW XVH LW EXW LQVWHDG rent it out to users who are landless and condemned to the status of tenants. The former have
24
no incentive to increase land productivity because they cannot mortgage it to acquire credit. Therefore, they do not care about improving its quality through fertilizer application. 2.2.3.3 Economic factors Income is required to purchase agricultural technologies such as improved maize and to pay for other factors of production such as labour that enhance adoption. Income can be categorized into on- and off-farm. Income is known as on-farm if it is obtained from the sale of farm produce and off-farm if it is obtained from other activities away from the farm. OffIDUPLQFRPHZDVIRXQGSRVLWLYHO\UHODWHGZLWKWHFKQRORJ\DGRSWLRQGXHWRHQKDQFHGIDUPHUV¶ ability to purchase inputs (Cheryl et al., 2003). On the other hand, on-farm income indicates the level of farm profit. It is expected that wealthier farmers will have much capital to plough back into the production process (Tiamiyu et al., 2009). However, it is difficult to estimate income at farm level due to the irregular nature of farm sales and off-farm work (Kafle, 2010). Households with more income in general, are associated with higher adoption of improved technologies. For example, Mignouna et al. (2010) noted that adopters of a new maize variety in Kenya had higher income than non-adopters, due to differences in purchasing power. The price of the technology also determines its adoption intensity. If it is expensive, smallholder farmers shun it because they cannot raise the required money (Saweda, 2011). Therefore, adoption of improved technologies will be enhanced if the cost is subsidized or if IDUPHUV¶ LQFRPHV DUH HQKDQFHG ,PSURYHG LQSXWV VXFK DV IHUWLOL]HUV DUe sensitive to price changes as well as the prices of the crop to which it is used (Okoboi and Barungi, 2012). Therefore, the demand of most agricultural inputs is derived demand because it is influenced by demand and price of the product.
25
2.3 Efficiency in farming The farm is Economically Efficient (EE) if it is not possible to produce more output without taking resources away from the production of another output. The concept of EE is critical in developing countries such as Uganda as production resources are outstripped by increasing human population. As resources become scarce, food growth rate becomes sluggish. Therefore, population growth outpaces food production (UBOS, 2013). There is need to improve efficiency in the use of few available resources as opportunities for developing and adopting better technologies reduce. Efficiency is an alternative to area expansion for productivity growth. However, it has received little attention in Uganda and increase in output of most crops has been due to increased acreage other than adoption of improved technologies (Sserunkuuma, 2005). When two approaches (area expansion and efficiency) are combined, they result into a much higher increase in productivity. Farrell (1957) pioneered the work of studying efficiency in the use of production resources and categorized it into three: 1. Technical Efficiency (TE) or Production Efficiency which is an engineering concept that refers to the physical ratio of output to factor input. The greater the ratio, the greater the magnitude of TE. It can also be defined in terms of the ratio of observed to corresponding frontier output. 2. Allocative Efficiency (AE) or Price Efficiency occurs at a point where Marginal Physical Product (MPP) is equal to the Marginal Factor Cost (MFC) and deals with the extent to which farmers make efficient decisions by using inputs up to the level at which their marginal contribution to production value is equal to factor costs. 3. Economic Efficiency (EE) on the other hand is a combination of TE and AE and occurs at a point where both TE and AE have been obtained. 26
Achieving TE or AE but not both is a necessary but not a sufficient condition to ensuring EE. This means that farmers can obtain inputs at relatively low prices but if not optimally combined during production, productivity growth may not be achieved. Therefore, EE is a necessary and sufficient condition for productivity growth (Ohajianya and Onwerem, 2012). This study compares TE of adopters and non-adopters of improved maize in Kabarole District. 2.3.1 Technical Efficiency A farmer is said to be technically efficient if it is impossible to raise farm output without increasing the use of at least one input. Technical Efficiency (TE) is therefore defined in terms of the ratio of observed output (Yi) to the corresponding frontier output ( ) כgiven the level of inputs used (Farrell, 1957). With TE, inputs like fertilizers, seeds, labour and herbicides are put to optimum use to increase farm success by producing maximum output for a given set of inputs (Farrell, 1957). Technical Efficiency emphasizes the use of obtainable technologies to achieve maximum output. It can be measured by using input-output quantity without introducing their prices and ranges between 0 and 1. If TE is equal to 1, the farm is said to be fully efficient, otherwise, it means the farm is technically inefficient in resource utilization (Khai and Yabe, 2011). Different scholars have defined technical inefficiency in different ways: Hyuha (2006) defines it as the amount by which the level of production of the farm is less than the frontier output. Technical inefficiency, defined by Rakipova et al. (2003), is a measure of the distance the farm is off its production function under variable returns-to-scale. Technical Efficiency on the other hand, relates to the degree to which a farmer produces the maximum possible output from a given input bundle or uses the minimum possible input to produce a given
27
output, known as output- or input-oriented TE measures respectively (Giang and Lewis, 2013). 2.3.2 Theoretical framework for measuring Technical Efficiency Two approaches are used to estimate TE: the econometric (parametric) approaches and the non-parametric approaches such as Data Envelopment Approach (DEA) (Asiimwe, 2007). The latter uses the mathematical programming methods that provide the frontier only for the most efficient farm (Khai and Yabe, 2011). While parametric methods are good at estimating single product and multiple input type of production, DEA (non-parametric) is used for multiple input and multiple output type of production (Rakipova et al., 2003). DEA is widely used in literature including Rakipova et al. (2003) and Giang and Lewis (2013). As pointed out by Rakipova et al. (2003), when using DEA in estimating TE, a farmer is DVVXPHGWRPD[LPL]HRXWSXWJLYHQWKHOLPLWHGUHVRXUFHVDQGVXEMHFWWRDOORWKHUIDUP¶VRXWSXW minus weighted inputs being less or equal to zero. They argue that when using this approach, the most efficient farmer will have TE scores equal to one while other less efficient farmers fall within the envelop defined by the empirical frontier and efficiency scores less than one are obtained. A farmer that did not produce any output would have an efficiency score of zero. With DEA, in practice, a farmer may obtain efficiency scores of one. However, this does not necessarily indicate that he or she has obtained the highest level of TE for a given production function. However, TE scores of one indicate that, in a given sample, no other farm is more technically efficient using the same inputs (Rakipova et al., 2003 and Khai and Yabe, 2011). Compared to parametric methods, DEA lacks modern statistical techniques for estimation thus, making parametric methods superior in estimating efficiency (Kibirige, 2008). And,
28
since maize production is an example of single output but multiple input enterprises, this study adopted the parametric (econometric) methods for measuring TE. Parametric (econometric) methods are further categorized into two: deterministic and stochastic methods. Deterministic methods handle all data of the farm but have a setback of attributing all observed variations from the production frontier to inefficiency of the farmer. However, others variations could be due to measurement errors and other factors beyond the farm such as weather, pests and diseases and luck. Most literature on productivity, including that of Danilo (2004), Njeru (2010) and Giang and Lewis (2013) emphasize the inclusion of a second set of factors in the analysis of TE, which DUHQHLWKHULQSXWVQRURXWSXW7KHVHIDFWRUVH[RJHQRXVO\LQIOXHQFHIDUPHUV¶REVHUYHGRXWSXW and characterize his or her socio-economic environment. Danilo (2004) argues that specific factors such as household size, market, credit, extension, gender, age, income and land ownership, among others, have been listed as usual candidates under socio-economic factors. Including these factors in the Stochastic Frontier (SF) models allows the association of variation in output with other causes not only those in technological and external domains. Therefore, SF models account for the measurement errors and other stochastic noise as separate from farm-specific factors that prevent the farm from producing at or close to its production frontier. Therefore, SF models provide more detailed explanation for observed IDUPV¶7(ZKLFKPDNHVWKHPVXSHULRUWRGHWHUPLQLVWLFPHWKRGVDQGIRUWKLVUHDVRQWKH\ZHUH adopted for this study. Other efficiency studies that used SF model include that done by Thiam et al., (2001), Hyuha (2006), Kibirige (2008) and Ohajianya and Onwerem (2012). Technical Efficiency (TE) of the farm is denoted by output oriented TE as follows: ܻ ൌ ܻ כǤ ܶܧ And,
29
ܶܧ ൌ כ
ൌ
ൌ
ݕ ݁ݔሺݔ Ǣ ߚሻ ௫ሺ௫ Ǣఉିఓ ሻ ௫ሺ௫ Ǣఉሻ
ൌ ݁ݔሺെߤ ሻ
(6)
where ܻ݅ כൌ ݂ሺ ݅ݔǢ ߚሻǡthe frontier production function of the ith farm, Yi is the observed output and µ i is a one-sided non-negative error term and the larger the µ i, the less technically efficient the farmer, i is the households, xi is the vector of inputs used by household i and ȕis a vector for the parameters estimated. If TEi=1, then household i is efficient otherwise, TEi is less than one and this indicates technical inefficiency. 2.3.3 Factors affecting Technical Efficiency Most farmers in Uganda are subsistence due to technical inefficiencies in utilization of resources such as land, labour, fertilizers and herbicides. Technical inefficiency is also found to be attributed to a number of socio-economic factors such as low education (Hyuha, 2006 and Asiimwe, 2007) poor access to credit, extension services, land and market (Kibirige, 2008 and Sibiko et al., 2013) as expounded on below: Saweda (2011) reviewed literature on agricultural productivity, social capital and food security in Nigeria. That study revealed that being male, having more formal education, owning land, accessing credit and extension services augmented the rate of adoption of LPSURYHGWHFKQRORJLHVDQGWKHUHIRUHFRQWULEXWHGWRKLJKHU7(LQ1LJHULD¶VDJULFXOWXUH
30
Mignouna et al. (2010) used the Tobit model to estimate factors influencing adoption of IRM1 and production efficiency in Western Kenya. Their study found out that TE of maize farmers ranged between 25 and 100 percent. This indicates that some farmers were technically efficient in adoption and utilization of the new technology while others were not. The study further indicated that age, education, maize production gap, risk aversion, contact to extension agents, membership to a social group and accessibility to IRM seeds were VLJQLILFDQWYDULDEOHVLQVKDSLQJIDUPHUV¶GHFLVLRQRQDGRSWLRQDQGHIILFLHQF\LQ the use of the technology. Kibaara (2005) VWXGLHG 7( LQ .HQ\D¶V PDL]H SURGXFWLRQ XVLQJ WKH 6WRFKDVWLF )URQWLHU 6) approach. The study indicated 49 percent as the mean TE and observed that adoption of Hybrid maize increased the level of TE by 36 percent (6.14 bags). The study further revealed that access to markets and extension, fertilizer use and having more years of formal education increased TE in maize production. However, as established by that study gender was not a major determinant of TE and it was done in Kenya which has different socio-economic and agro-ecological conditions from those in Uganda. In Central Gujarat, high TE of 72.78 was obtained by rice farmers (Narala and Zala, 2010). The study indicated that rice output can be increased by nearly 27 percent if the existing WHFKQRORJ\DQGUHVRXUFHVZHUHWREHHIILFLHQWO\XWLOL]HG)DUPHUV¶H[SHULHQFHLQULFHIDUPLQJ VL]HRIWKHDUHDXQGHUULFHHGXFDWLRQDQGGLVWDQFHIURPIDUPHUV¶UHVLGHQWVWRPDUNHWZHUHWKH main determinants of TE. Studies on TE in agricultural sector are increasing due to high demand for efficiency in the utilization of available resources especially by resource constrained households. Okoboi 1
IRM is Imazapyr Resistant Maize that was developed in Kenya in response to the problem of severe outbreak of striga weed in maize fields
31
(2010) studied imprRYHGLQSXWVDQGSURGXFWLYLW\LQ8JDQGD¶VPDL]HVXE-sector, by estimating the SF of the Cobb-Douglas (C-D) production function. The results showed that only2 fertilizer and traction power had positive and significant effects on maize yield. The study concluded that increase in area cultivation was among the main factors sustaining maize productivity in Uganda. According to Sibiko et al. (2013) farm size is positively related to TE. This is because when a farmer increases the scale of production, he/she also improves the intensity of management to reduce farm risk. Technical Efficiency and productivity can be increased by increasing acreage. Rakipova et al. (2003) earlier obtained a positive relationship between farm size and TE. They argued that due to economies of size, farmers would be technically efficient. However, Asiimwe (2007) noted that opening more land for agriculture is not environmentally friendly because of deforestation and other forms of environmental degradation suggesting that alternative approaches of increasing productivity such as efficiency on small available plots should be explored. Land ownership also enhances TE (Rahman and Umar, 2009). Sibiko et al. (2013) observed that a one hectare increase in the use of own land to grow beans increased TE by 1.5 percent due to conservation through bush fallowing to allow land to regain fertility especially if the farmer owns a larger plot size. Saweda (2011) also argues that a favourable land ownership system enhances adoption of improved technologies and hence TE among farmers. However, where soils are originally fertile, land ownership may not significantly determine TE (Asiimwe, 2007). Most studies about the effect of extension services on TE reveal a positive correlation between the two variables (Ogundari, 2009; Njeru, 2010 and Saweda, 2011). Hyuha (2006) 2
Other inputs studied by Okoboi (2010) were seeds, area cultivated, herbicides/fungicide, labour and manure.
32
examined profit efficiency among rice producers in Eastern and Northern Uganda. Technical Efficiency levels of 75, 70 and 65 percent were obtained for Pallisa, Lira and Tororo districts respectively. The analysis further indicated limited access to extension services, low education and lack of experience in rice production as the main causes of inefficiency. ([WHQVLRQVHUYLFHVLPSURYHIDUPHUV¶PDQDJHULDODELOLW\DQGDFFHVVWRSURGXFWLYLW\HQKDQFLQJ opportunities such as credit, improved technologies and market. Sibiko et al. (2013) established that farmers who accessed extension services had a higher TE by 6.4 percent than those who did not due to insights received about good management practices. Group action promotes other socio-economic factors such as accessibility to credit, market DQG H[WHQVLRQ ZKLFK DIIHFWV PHPEHUV¶ 7( Kibirige (2008) analyzed the impact of Agricultural Productivity Enhancement Programme (APEP) on TE and AE of maize farmers in Masindi District. The analysis indicated that mean TE for APEP farmers was higher (67 percent) than non-APEP farmers (49 percent). Therefore, membership to the programme would reduce inefficiency among maize farmers. 6H[ RI WKH KRXVHKROG KHDG DIIHFWV WKH IDUP¶V 7( 6WXGLHV LQFOXGLQJ 5DKPDQ DQG 8PDU (2009), Ogundari (2009) and Saweda (2011) have posted a positive and significant relationship between male-headed households and TE. However, where labour markets exist, male heads migrate to towns to look for paid labour and leave the farm management role to women, who may be engaged in other domestic chores such as cooking and washing, that reduce farm time hence inefficiency. )DUPHUV¶ DJH LV OLQNHG WR H[SHULHQFH DQG WKHUHIRUH LQFUHDVHV WKH OHYHO RI 7( +RZHYHU Sibiko et al. (2013) IRXQGRXWWKDWDQLQFUHDVHLQIDUPHUV¶DJHE\RQH\HDUUHGXFHGHIILFLHQF\ by 0.2 percent. The reason suggested, is that older farmers are more reluctant to take up new technologies and being old does not mean more experience. This contradicts results earlier 33
posted by Saweda (2011), where age was found to be positively correlated to TE due to accumulated experience. Family size is related to availability of family labour for various farm activities. It replaces hired labour which is expensive for resource constrained smallholder farms in SSA (Ogundari, 2009). Hired labour is associated with low TE because when workers are paid money without strict supervision, they do not do appropriate and timely work. Most studies including that by Obwona (2006) and Kibirige (2014) established a positive and significant relationship between family size and TE. Larger family sizes offer cheap labour to the farm especially at peak cropping times such as weeding and harvesting. However, an increase in family size would reduce TE especially if all family members do not participate in farm activities (Rahman and Umar, 2009). Market is another important factor determining the level of TE in farming. As earlier stated in the hypotheses of this study, off-gate maize market increases TE and maize productivity. According to Kibirige (2014), selling maize at home had a negative and significant influence on TE due to low farm-gate prices obtained by the farmer compared to the prevailing prices. However, the effect of an off-farm market on TE depends on distance from the farm. As the latter increases, the level of TE reduces (Salasya et al., 2006). Sibiko et al. (2013) observed that a one kilometer increase in distance reduces TE by 0.8 percent due to high cost of WUDQVSRUWZKLFKUHGXFHVIDUPHUV¶SURILWPDUJLQ Although household income is also a major determinant of TE, Rahman and Umar (2009) did not find the effect significant. Money is required for various activities such as purchasing seeds, fertilizer, herbicides and hiring labour. The effect needs to be separated; while farmincome increases the level of TE, off-farm income negatively affects TE due to less time spared for the farm because the farmer has multiple jobs (Obwona, 2006). However, 34
Rakipova et al. (2003) obtained contradicting results and argued that if high off-farm income is obtained, management and timeliness in farm decisions also improve hence TE. Most studies including Rakipova et al. (2003), Hyuha (2006), Obwona (2006) Asiimwe (2007), Ogundari (2009) and Saweda (2011) indicate a positive relationship between formal education and TE. Hyuha (2006) noted that education of all household members is important in improving TE. However, Kibirige (2014) found out that only that of the household head and the spouse had a positive and significant relationship on TE. The study explained that most of the spouses sampled were women and therefore more involved in the decisionmaking process for the farm. Rahman and Umar (2009) established a positive but not statistically significant effect of the variable on TE. Asiimwe (2007) recommended increasing IDUPHUV¶HGXFation to encourage group formation, access to input credit and subsidies which reduces inefficiency. Access to credit increases farm income to finance farm activities. This can be obtained from financial institutions such as commercial banks and micro-finance institutions. In spite of a positive relationship, access to credit did not significantly affect TE (Rahman and Umar, 2009). Obwona (2006) noted that when farmers obtain input credit and properly utilize it, their TE improves. This is because it alleviates the problem of lack of capital for resource constrained families. However, all the studies on TE done as reviewed above, did not examine the differences in TE for adopters and non-adopters of improved maize. 2.4 Maize output response to inputs Adoption of improved maize could enhance response output response to input quantities of fertilizers, herbicides, plot size and labour. Therefore, input constraints can also hinder an improved variety from achieving its yield potential hence affecting its sustainability and also DIIHFWLQJIDUPHUV¶HIILFLHQF\ (Salasya et al., 2006). For example, the type, time and method 35
of application of a fertilizer determine the amount of output obtained from a given crop like maize. The cost of a given input also determines quantities of input used. Kibaara (2005) showed that small-scale farmers incurred high costs in terms of hiring labour, fertilizer and seeds while large-scale farmers only incurred high cost of land preparation due to mechanization. 2.4.1 Theoretical framework for measuring output response to inputs Output response to input factors, known as total productivity is measured in terms of output per unit inputs. This can be obtained by estimating the Cobb-Douglas (C-D) production function. The C-D production function was adopted in this study and has widely been used in agricultural studies, including that by Kibaara (2005), Okoboi (2010) and Ohajianya and Onwerem (2012). The production function describes the technical relationship that transforms inputs into outputs. It is part of micro-economic theory that deals with a given set of inputs (Derbatin, 2012) ,W LV DVVXPHG WKDW WKH IDUPHUV¶ PDLQ REMHFWLYH LV WR PD[LPL]H SURILW E\ HLWKHU increasing output, y or by minimizing the cost of inputs, xi (e.g. cost of cultivating the land, cost of nitrogenous fertilizer, seeds, herbicides and paying labour) used to produce output yi. Cobb-Douglas (C-D) production function is one of the functional forms that can be used to estimate output response to inputs. Other functional forms include transcendental, quadratic and translog. If the C-D production function contains two inputs, x1 and x2, its functional form is: ߚ
ߚ
y =ʹ ʹݔ ͳ ͳݔܣ
(7)
where y is yield obtained, x1 and x2 are inputs, A HPERGLHV WKH PDQDJHU¶V VNLOO DQG RWKHU factors, while ȕ are parameters representing elasticity. 36
From Equation 7, Marginal Physical Product (MPP) and Average Physical Product (APP) can be obtained as shown below: ఉ ିଵ ఉమ ݔଶ
ܲܲܯ௫భ ൌ ߚଵ ݔܣଵ భ
(8)
ఉ ିଵ ఉమ ݔଶ
ܲܲܣ௫భ ൌ ݔܣଵ భ
ఉ
(9)
ఉ ିଵ
ܲܲܯ௫మ ൌ ߚଶ ݔܣଵ భ ݔଶ మ ఉ
(10)
ఉ ିଵ
ܲܲܣ௫మ ൌ ݔܣଵ భ ݔଶ మ
(11)
ܲܲܯ௫భ Τܲܲܣ௫భ ൌ ߳ଵ ൌ ߚଵ
(12)
ܲܲܯ௫మ Τܲܲܣ௫మ ൌ ߳ଶ ൌ ߚଶ
(13)
Where ߳ͳ and ߳ʹ represent elasticities for inputs x1 and x2 respectively. MPP refers to additional output that can be produced by using one additional unit of input ߲ݕ
and is calculated as ݆ݔܲܲܯൌ ߲ ݔ. This is known as first order condition of the production ݆
function. The second order condition is obtained by differentiating MPP, which is less than zero, indicating that when quantities of an input are increased indefinitely, while holding other factors constant, it will result in diminishing marginal productivity, where each additional unit of input results into lower output. This is calculated asǣ
డெೣೕ డ௫ೕ
ൌ
డమ ௬
డ௫ೕ మ
. MPP is
obtained as a product of ݆ݔܲܲܣand elasticity (݆߳ ሻ. APP on the other hand measures the level RIIDUPHUV¶HIILFLHQF\ZKLFKGHSHQGVRQWKHOHYHORIDOOLQSXWVXVHGGXULQJSURGXFWLRQ,WFDQ also be obtained as follows: ݆ݔܲܲܣൌ
ܱݐݑݐݑ ݆ݔ
ݕ
ൌ ( ݔDerbatin, 2012). ݆
37
The term returns-to-scale parameter also known as the function coefficient refers to how output, yi responds to the given levels of inputs, xi and it is obtained by summing up the ratios of MPP and APP for each input. Therefore, ߝ ൌ ߳ଵ ߳ଶ ൌ ଵ ߚଶ ൌ ܲܲܯ௫భ Τܲܲܣ௫భ ܲܲܯ௫మ Τܲܲܣ௫మ
(14)
where ߝ is elasticity and j stands for input quantities, and x for input type If the production function is homogeneous of degree n, and all inputs are represented in the production function, then the parameter representing the returns-to-scale coefficient is the degree of homogeneity. For a multiplicative power production function with j inputs, the degree of homogeneity and returns to scale are determined by summing up the j respective ȕ coefficients which are elasticities of production for individual inputs. And, even if the production function is not homogeneous, returns-to-scale can still be determined by summing up the respective ratios of MPP and APP (Derbatin, 2012). Returns-to-scale can be categorized into three: increasing, decreasing and constant. If σ ߚ݆ ൌ ͳ, then the production function exhibits constant returns-to-scale, if σ ߚ݆ ͳ the
production function shows increasing returns-to-scale, while if, σ ߚ݆ ൏ ͳ, the production function shows decreasing returns-to-scale coefficients. 2.4.2 Fertilizer use and maize output Due to complementarities in technology adoption, high output from improved varieties can only be realized if another improved input such as nitrogen fertilizer is applied (Bouagnimbeck and Ssebunya, 2011). Okoboi (2010) generated similar results about 38
adoption of improved maize varieties and fertilizer use. The study observed that planting improved maize seeds without applying fertilizers had slightly lower output compared to that obtained when fertilizers are applied, but lower profit margin due to the higher marginal cost of fertilizer compared to the marginal revenue obtained because fertilizers are expensive. Compared to improved varieties, local maize is less responsive to fertilizer use. For example, improved maize varieties gave 323 kg/ha more than local maize when fertilizer was applied to both maize types (Holden, 2009). Okoboi and Barungi (2012) noted that use of inorganic fertilizer could increase output by 4060 percent. While, Kibaara (2005) estimated that a one percent increase in use of inorganic fertilizer would increase maize output by 17 percent. However, this depends on whether farmers apply the right fertilizer, to the right crop, at the right time and by the right method. For example, Kibaara (2005) further observed that maize farmers in Kenya who also grew tea, applied NPK fertilizer meant for tea on maize crop yet it is not right for maize. Thus, farmers realized low output response. The right fertilizers for maize crop include DAP, CAN and urea (Beinempaka et al., 1990). Most studies, including that of Kasirye (2013), separate inorganic from organic fertilizers because use of the later by smallholder farmers is difficult to estimate. 2.4.3 Plot size and maize output Plot size was found to be negatively related to maize productivity whereby, an increase in the area under maize by one hectare would reduce maize output by 41 percent (Okoboi, 2010). 7KLVLVSRVVLEO\EHFDXVHDVWKHDFUHDJHLQFUHDVHVIDUPHUV¶DWWHQWLRQDQGPDQDJHPHQWSHUXQLW reduce. Other studies including that of Kimhi (2003) attribute the negative relationship to imperfect land and labour markets, particularly lack of strict supervision of family labour
39
when producing on large scale. The inverse relationship can also be explained by higher land conservation efforts on smaller than larger plots. 2.4.4 Herbicide use and maize output Maize output increases considerably with herbicide treatment due to increase in per ear grain number, seed weight, length and diameter as well as increase in number of the kernel rows (Larbi et al., 2013). Maize output was significantly higher on plots that were sprayed with herbicides such as Atrazine and 2, 4-D (Calliherbe and bextra) than those hand weeded at early stages; possibly because the latter is labour intensive, time consuming and weeds reappear immediately after it is applied. However, herbicide treatment did not significantly increase the number of cobs per plant as the character is genetically controlled rather than by environment (Salarzai, 2001). There are variations in the effect of different types of herbicides on maize output. Higher grains per cob and therefore higher output (5.15 t/ha) were recorded on plots treated with Gramaxone (post-emergence) compared to Stomp 330 E (pre-emergence) which had only output of 3.94 t/ha (Salarzai, 2001). However, all plots treated with herbicides significantly gave higher maize grain compared to the control plots (Naveed et al., 2008). They further observed that output response to herbicide application depended on the amount of herbicide applied. Application of full doze of foramsulfuron + isoxadifen-ethyl (1125 ai/ha) produced 4.46 t/ha. With the reduced doze of the same herbicide, yield significantly reduced by 3 percent. 2.4.5 Labour use and maize output Labour (both family and hired labour) is required to carry out a number of farm activities such as land preparation, planting, weeding, fertilizer application and harvesting. Kibirige 40
(2008) established a 0.01 unit increase in maize output for any additional man-day used. Kibirige (2014) observed that up to 7 percent increase in maize output could be realized with one man-day increase (labour). However, in both studies, the effect was not significant. In Kenya a higher output response to labour of up to 42.2 percent was reported by Kibaara (2005). All studies considered labour in general but did not examine maize response to different forms of labour such as family, hired and social labour. 8JDQGD¶V HVWLPDWHG ODERXU SURGXFWLYLW\ RQ PDL]H FURS LV NJPDQ-day. Eastern Uganda had the highest labour productivity of 7.04 kg/man-day, followed by Central Uganda (5.23 kg/man-day), Western Uganda (5.00 kg/man-day) and Northern Uganda (4.61 kg/man-day) (Okoboi, 2010). Considering total labour in man-days, the study indicated Western Uganda as having more man-days (145.81 man-GD\V FRPSDUHG WR (DVWHUQ 8JDQGD¶V PDQdays (Okoboi, 2010). Therefore, Western Uganda has the lowest labour productivity. 2.4.6 Maize seed and output The response of maize output to the amount of seed depends on a number of factors; the most important ones being variety of the seed (whether improved or local), pest and diseases management and whether sorted or not (Mignouna et al., 2010). Local maize seeds may not give high output due to their inability to respond to inputs such as fertilizers. Bouagnimbeck and Ssebunya (2011) noted that farmers who plant Hybrid maize seed and apply fertilizers obtain higher output compared to those who plant home-sourced seeds and apply fertilizers. This is due to complementarities of the two improved technologies where the use of one input enhances productivity of the other. The recommended mean seed rate per acre is 10 kg (Agona and Muyinza, 2005). However, this depends on whether it is pure or mixed stand (Beinempaka et al., (1990). It also depends 41
of the ecological conditions, method of planting and grain size. Mugisha and Diiro (2010) established that mean maize output obtained from improved seed as 2941.5 kg/ha per season which was significantly higher than those obtained from local seeds (1694 kg/ha per season). However, the study conducted in Eastern and Central Uganda, does not show the differences in output obtained within different improved maize varieties (i.e. Longe series and Hybrid). Kibaara (2005) found out that maize output had highest response to seed compared to fertilizer and labour. One additional kg of seeds increased maize output by 52kgs. However, the study was done in Kenya which has different socio-economic and agro-ecological conditions from those of Uganda and Kabarole District in particular.
42
3.0 CHAPTER THREE: METHODOLOGY 3.1 Empirical models Empirical models used in the study are explained by objective as shown below. Analysis of socio-economic factors affecting adoption of improved maize varieties was accomplished using descriptive statistics and probit model as explained in 3.1.1. The Stochastic Frontier (SF) was specified to determine the level of TE among maize farmers as shown in 3.1.2. Maximum Likelihood Estimation (MLE) was used to estimate the parameters in the SF model in order to determine the effect of each of the socio-HFRQRPLFIDFWRUVRQIDUPHUV¶7( (3.1.2.1). Section 3.1.3 shows the Cobb-Douglas (C-D) production function model that was used to estimate maize output response to input factors namely; nitrogen fertilizers, seeds, labour, plot size and herbicides. Lastly, description of variables in the probit, inefficiency models and C-D production function as well as apriori expectation is shown in 3.1.4. 3.1.1. Examining factors that affect adoption of improved maize varieties A probit model, specified to analyze the socio-HFRQRPLF IDFWRUV GHWHUPLQLQJ WKH IDUPHUV¶ decision to adopt improved maize varieties in Kabarole District is expressed as a regression model below: ݕ ൌ ݅ݔԢߚ ߝ݅ ; and ᖡi̱ሺͲǡ ߪ ଶ ሻ
(15)
The assumption is that a farmer chooses to plant an improved maize variety depending on whether the utility difference of adopting the variety exceeds a certain threshold (zero in this case) such that;
ݕ ൌ ൜
ͳ݂݅ כ ݕ Ͳ Ͳ݂݅ כ ݕൌ Ͳ
(16)
43
Where, yi is a dummy representing a binary dependent variable for adoption. If ݅ݕൌ ͳ, (the farmers adopted Longe series or Hybrid maize in the September-January (2012/13) planting season) ݅ݕൌ Ͳ(otherwise), כݕis unobserved variable showing utility difference, ݅ݔare independent variables which include X1= sex, SEX (dummy), X2= education, SCH (years), X3= income, INCM (UGX), X4= family size, FMLY (people), X5= credit access, CRDT (dummy), X6= off-gate market, MKT (dummy), X7= land ownership, OWN (dummy) and X8= extension access, EXT (dummy). ȕ are parameters estimated using MLE, in which after obtaining the coefficients, using the post-HVWLPDWLRQ 67$7$ FRPPDQG µPI[¶ PDUJLQDO effects were obtained (Appendix Table D3), ߝ݅ is the error term representing other factors and assumed to be independent of xi, and i is the household. 3.1.2. Estimating Technical Efficiency The SF model specified to estimate the level of TE among adopters and non-adopters of improved maize in Kabarole District takes the form in Equation 17. ݅ݕൌ ݂൫ ݆݅ݔǡ ߙ൯Ǥ ݁ᖡ
i=1, 2,... 160 households
(17)
M «LQSXWV
where yi is the scalar for maize output (kg) produced by the ith household from either improved varieties or local maize; Xji is the vector for input j quantities used by the ith household; e LV WKH H[SRQHQWLDO YDOXH Į are parameters for inputs estimated and ᖡ is the error term which is composite as seen below:
ᖡ= ݅ݒ ߤ݅
(18)
vi denotes the measurement (random) error, representing natural factors such as weather, pests and diseases, luck and other factors outside the farm, on which household i has no control. It is two-sided representing the usual statistical noise found in any relationship and is 44
distributed as ̱ܰ ݅ݒሺͲǡ ߪʹ ݒሻ. µ i is a non-negative random and one-sided variable representing farm-specific factors that are associated with technical inefficiency, which are normally distributed and independent of other µ i¶s [ߤ݅ ̱ܰሺͲǡ ߪʹߤ ሻሿ and ranges between 0 for efficiency and 1 for inefficiency. µ i¶s are also independent of vi¶V The j inputs in the above SF model include X1i«;5i which are input quantities as shown below: X1= size of plot, PLT (ha) used to grow maize X2= quantity of seeds, SEED (kg) planted X3= labour, LBR (man-days) used in maize production X4= quantity of nitrogen fertilizer, FERT (kg) applied X5= quantity of herbicide, HERB (litres) 3.1.2.1 Estimating determinants of Technical Efficiency The farm overall TE is obtained by combining Equations 17 and 18 as seen below: ݅ݕൌ ݂൫ ݆݅ݔǡ ߙ൯Ǥ ݁ሺݒߤሻ
(19)
The inefficiency model was specified for efficiency analysis, where µ i in equation 19 is written as a function below: µ i = f (Zkiȕk N «.8 socio-economic factors
(20)
where, Zk is the vector of socio-economic factors (Z1= sex, SEX (dummy), Z2= education, SCH (years), Z3= income, INCM (UGX), Z4= family size, FMLY (people), Z5= credit access, CRDT (dummy), Z6= market, MKT (dummy), Z7= land ownership, OWN (dummy) and Z8= extension access, EXT (dummy), ȕk are parameters estimated. 45
The general f (.) and the ᖡ in equation 19 can be expressed as the S.F model as in equation 21. ݈݊ ݅ݕൌ ߙͲ σͷ݆ൌͳ ߙ݆ ݈݆݊ܺ݅ ൣ ݆݅ݒ ൫ߚͲ σͺ݇ൌͳ ߚ݇ ܼ݇݅ ൯൧
(21)
where ln is natural logarithm, yi is observed maize output, i is the household, Z represents socio-economic factors determining TE of the farm, ߙͲ and ߚͲ are intercepts, while ߙ݆ ߚ݇ are parameters estimated for inputs and socio-economic factors respectively. Estimation of parameters (Į, ȕ and Ȟ) in the C-D production function model was done in a two-step process. After obtaining the TE from SF model using inputs j, then TE for individual farmers were regressed on socio-economic factors k IRU WKH LQHIILFLHQF\ PRGHO *DPPD Ȗ REWDined in variance parameters of the inefficiency model measures technical inefficiency and lies between zero and one. When it is zero, it means there are no inefficiency effects in the model and when it is one, there are farmer-specific inefficiencies that are not random. 3.1.3. Estimating maize output response to inputs The study applied a C-D production function because it (C-D production function) is preferred when three or more independent variables are involved. In a C-D model, all inputs and outputs can easily be expressed in a log form and interpreted using most data analysis software (Derbatin, 2012). It is also less affected by problems such as degrees of freedom and multicollinearity (Gujarati, 2004). The C-D production function used to estimate maize output response to inputs is specified as shown in equation 22: ܱܷܷܶܲܶ ൌ ܣൣܲܶܮఉଵ ܴܶܧܨఉଶ ܵܦܧܧఉଷ ܴܤܮఉସ ܤܴܧܪఉହ ൧
46
(22)
The production function can be linearized as shown below: lnOUTPUTi OQ$ȕ1lnPLTiȕ2lnFERTiȕ3lnSEEDiȕ4lnLBRiȕ5lnHERBi
(23)
where ln is the natural logarithm and i is the household. A is the coefficient parameter that HPERGLHV WKH PDQDJHUV¶ VNLOO DQG RWKHU IDFWRUV DIIHFWLQJ WKH FRPELQDWLRQ RI LQSXWV GXULQJ production and ȕ are coefficients representing elasticity estimated. 3.1.4 Description of variables and apriori expectation 3.1.4.2 Variables in the probit and Technical inefficiency models Sex (SEX) of the household head is very important in determining adoption of improved maize varieties and TE of the farm. Being male is positively related to accessibility to production enhancing factors such as extension, formal education, market, credit and improved varieties. Most studies, including that by Cheryl et al. (2003) and Saweda (2011) revealed that being male increases access and ownership of production resources such as land, that can be mortgaged to acquire credit. Compared to women, men also provide more potential labour for farm activities. Hyuha (2006) found out that women can only offer 75 percent RIWKHPHQ¶VSRWHQWLDOODERXUIRUFHDQG5DKPDQDQG8PDU REWDLQHGDSRVLWLYH and statistically significant relationship between male-headed households and TE. Therefore, being male is expected to increase adoption and reduce inefficiency. Formal education (SCH) of members is associated with improved labour quality through skill building and high off-farm income which may increase adoption of new technologies and TE. Angu (2004) and Cheryl et al. (2003) observed a positive and statistically significant relationship between education and adoption of maize and wheat technologies. Ohajianya and Onwerem (2012) argue that education of all household members is important. However, Okoboi (2010) noted that only education of the household head of up to seven (7) years is
47
necessary in reducing inefficiency. Further increase in the number of years of formal education may increase inefficiency, as most educated farmers seek off-farm employment in the formal sector, thereby, leaving farm-work to less educated and elderly household members. Therefore, this variable is expected to increases adoption, but may or may not increase TE. ,QFRPH ,1&0 LVDPHDVXUHRIDJLYHQKRXVHKROG¶VZHDOWK%RWKRQ- and off-farm income is expected to facilitate adoption of production inputs as pointed out by Cheryl et al. (2003), Angu (2004) and Kudi et al. (2011), which reduces inefficiency. However, off-farm income is associated with activities away from the farm which reduces farm time and therefore may increase or reduce efficiency. Family size (FMLY) covered household members, including those who were residing in the home four (4) months prior to the study. Kudi et al. (2011) obtained a negative and statistically significant result about family size and adoption and explained that, depending on LQGLYLGXDO¶V FRQWULEXWLRQ ODUJH IDPLOLHV PD\ UHGXFH KRXVHKROG¶V LQFRPH EHFDXVH RI KLJK consumption expenditure. Large families may provide labour for timely farm activities as pointed out by Tiamiyu et al. (2009) and therefore reduce inefficiency. However, in this study, the variable is expected to have negative influence on both adoption and TE. Adoption of improved maize and efficiency in maize production is determined by the amount of capital owned by the household. Credit (CRDT) enhances capital availability used to purchase improved maize and pay farm activities (Cheryl et al., 2003). Therefore, access to credit is expected to enhance adoption and reduce inefficiency. Access to other markets (MKT) for maize away from the farm is an incentive to selling at higher prices compared to farm-gate priceV7KLVZRXOGUDLVHWKHIDUP¶VSURILWDQGLQFRPHIRU increased adoption. Through off-gate markets, maize farmers are able to meet and share 48
farming experience about improved maize varieties and agronomic practices hence increasing adoption and TE. Land owners (OWN) save the money that would be used for rent to purchase improved seed hence increasing adoption. However, renters (land rented in) produce on large scale and practice good management in order to achieve a return on investment to offset the cost of renting. Therefore, it is expected that renters are more efficient than landowners. Access to extension services refers to contact between extension worker and the farmer to share experience on agriculture. Access to extension services (EXT) is a necessary condition for diffusion of research information about new technologies among farmers hence augmenting adoption. The more the number of extension visits, the more the experience shared and the less inefficient the farmer becomes. Table 4: Description of variables in the probit and Technical inefficiency models showing apriori expectation Variable category
Farmer-specific Education (SCH) Sex (SEX) Family size(FMLY) Institutional Credit (CRDT) Extension (EXT) Market (MKT) Land ownership (OWN) Economic Income (INCM)
Description of variables
Expected sign for probit model
Expected Sign for Technical inefficiency model
Formal education level of the respondent in years Sex of the respondent (dummy: 1 if male and 0 otherwise). Family size (number of people in household)
࢜ࢋ
േ࢜ࢋ
࢜ࢋ
െ࢜ࢋ
െ࢜ࢋ
࢜ࢋ
Access to agricultural credit (dummy: 1 for access and 0 otherwise) Extension visits (dummy: 1 if yes and 0 otherwise) Maize market (dummy: 1= off-gate market, 0 otherwise) Land ownership for the maize plot (dummy: 1 for rented and 0 otherwise)
࢜ࢋ
െ࢜ࢋ
࢜ࢋ
െ࢜ࢋ
࢜ࢋ
െ࢜ࢋ
െ࢜ࢋ
െ࢜ࢋ
Total Household income (UGX) including onand off-farm
࢜ࢋ
േ࢜ࢋ
49
Note: +means the variable increases adoption (probit model) but it means increase in inefficiency (Technical inefficiency model); while ± indicates otherwise. 3.1.4.2 Variables in the Cobb-Douglas production function Maize output (OUTPUT) is the dependent variable for the total maize grain harvested the farmers during the September 2012-January 2013 planting season. It is expected to increase with increase in the use of the quantities of inputs for both adopters and non-adopters. Plot size (PLT) is the proportion of land (ha) a household allocated to maize (whether improved or local) during the September 2012-January 2013 growing season. As shown in literature, as the plot size increases, output per unit area decreases due to divided attention which results into poor management. However, in this study, maize output is expected to be positively correlated with plot size because, compared to small-scale farmers, large scale producers plant improved varieties and practice good management to avert risks involved with the new technology. Nitrogen fertilizer (FERT) (Kg) mainly urea, CAN and DAP are commonly applied to boost maize productivity. Most studies, including that of Hyuha (2006), have confirmed low fertilizer use on most farms in Uganda. Therefore, there is loss of soil fertility which results into low output. A positive relationship between FERT and maize output is expected. Seed quantity (SEED) refers to the amount of maize seeds (kg) planted by a household in a given plot. It is expected that output is positively correlated with a given quantity of seeds planted up to a level [e.g. 10 kg/ha as established by Agona and Muyinza (2005)], beyond which, any further increase3 reduces output.
3
It is expected that when more and more seed quantities are planted in a fixed plot size, e.g.
ௗ మ ୳୲୮୳୲ ௗୗୈమ
, it results into less
output than the previous one indicating a negative marginal productivity slope. This is beyond this study coverage.
50
Labour (LBR) measured in man-days is included in the model because it is one of the primary factors of production. It is measured in man-days and disaggregated into family, social and hired habour. It is expected that maize response is higher for hired than family and social labour. Most studies, including that by Okoboi (2010) aggregate herbicides, fungicides and insecticides into agro-chemicals and study it as one variable. However, herbicides (HERB) (litres) applied by the household is isolated in this study, as the most important factor in controlling weed growth in the early growth of maize. Herbicide use is expected to improve maize output. 3.2 Data analysis STATA/SE 11.2 and SPSS 16.0 software were used for analysis. STATA SE 11.2 was used because it is simpler to use on most econometric analyses such as probit, C-D production function and SF models as well as diagnostic tests such as heteroscedasticity and multicollinearity. On the other hand, SPSS 16.0 was used for generating descriptive statistics. In descriptive statistics, basic features such as means, percentages and standard deviation of the data were described. As pointed out by Asiimwe (2007), use of descriptive analysis is limited by lack of ability to show the relationship between two or more variables and does not show the effect of an independent variable on the response of a dependent variable. Therefore, descriptive statistics were used alongside multivariate analysis. In this study, probit and Stochastic Frontier models that apply MLE were used for adoption and TE analysis respectively. Maximum Likelihood Estimation (MLE) makes use of the specific distribution of the error term and is more efficient than OLS (Gujarati, 2004). The advantage with such multivariate approaches (probit and stochastic frontier models) is that 51
they show the magnitude and direction of association between two or more variables (Asiimwe, 2007). 3.3 Study area The study was conducted in Kabarole District located in Western Uganda (Appendix A). It borders with districts of Bundibugyo in the north, Kasese in the east and Kamwenge and Kyenjojo in the south (UCC, 2010). The district has a total area of 8,318.2 sq. km and a population of 359,180 persons (50.1 percent males and 49.9 percent females). The population of people residing in urban areas is about 11.3 percent. Average household size is 5.08 persons, indicating a high potential labour supply and market for maize (UBOS, 2013). Administratively, Kabarole District is made up of three counties: Burahya, Bunyangabu and Fort portal municipality. Fort portal municipality is sub-divided into West Division, South Division and East Division. Bunyangabu County is comprised of the sub-counties of Kibiito, Rwimi, Kabonero, Kisomoro and Buheesi. While Burahya County is comprised of Rutete, Kasenda, Mugusu, Karambi, Bukuku, Busoro, Hakibaale and Kichwamba sub-counties (Appendix B) (UCC, 2010). The district lies in the banana-coffee system, characterized by deep black, well-drained, fertile volcanic soils with high organic matter to support maize productivity (FAO, 2012). It also lies at an altitude of 3,556 metres above sea level and receives reliable bimodal rainfall varying between 750-1000mm annually, with the first rains coming between February and May and second rains starting in August and ending in November. All these geographical factors create favourable conditions for high maize productivity. However, maize productivity in Kabarole District (and Western Uganda in general), estimated at 1.86t/ha, is slightly lower than the national average of 1.94 t/ha as established by (Okoboi, 2010). 52
The district was selected because it is the fourth largest producer of maize in Uganda after Iganga, Mubende and Soroti (MAAIF, 2011), the only one in Western Uganda. The major economic activity in the district is agriculture, dominated by maize and banana production. The latter is severely affected by the Banana Bacterial Wilt (BBW) disease, leaving maize as the only potential livelihood support crop (NAADS, 2012). The district has diversified agricultural enterprises, namely cattle, pigs, goats, sheep and poultry which provide rich manure to increase land productivity for smallholder maize farms (Okoboi, 2010). 3.4 Sampling procedure and sample size Multi-stage sampling technique was used as shown in Figure 2. In the first stage, purposive sampling was used to select sub-counties of Rwimi and Kibiito (Bunyangabu County), Rutete and Kasenda (Burahya County) (Appendix B). This was because these are among the major maize producing sub-counties in the district (NAADS, 2012). In the second stage, four villages were randomly selected from each sub-county. From every village, ten households (and one farmer from each household) were randomly selected, having obtained the list containing maize growing households from respective Local Council (LC) 1 chairpersons. Thus, a total of 160 maize farmers were selected as follows: 4 sub-counties, 4 villages per sub-county, 10 farmers per village and therefore, 40 farmers per sub-county.
53
3.5 Data Collection A pre-tested questionnaire (Appendix C) was used for the survey to collect primary data. It covered information about the type of maize planted (improved or local), socio-economic factors for a household which included education, family size, household income, sex, land ownership, access to extension services, credit and maize markets. Production information included plot size for maize. Other data included type of labour used, nitrogen fertilizer use, herbicide use, variety of maize planted and source, maize yield, income obtained from onand off-farm. 3.6 Data Validity and Reliability Validity is the ability of an instrument to measure what it purports to. It also refers to the degree to which the results can be generalized to a wider population, cases or situations. It is the ability of an instrument to fairly and comprehensively cover the domain or items it is meant for. On the other hand, reliability refers to dependability, consistence and replicability of an instrument in different groups of respondents over time. The concept of reliability is concerned with precision and accuracy of instruments (Louis et al., 2009). Reliability is a necessary but not a sufficient condition for validity, while validity may be a sufficient but not a necessary condition for reliability. Data validity and reliability for this study was achieved by pre-testing the questionnaire on April 12th and 13th, 2013 using two selected farmers from each sub-county. Redundant and misplaced questions were dropped, incorrect ones were corrected, and those that were left out due to oversight were included to ensure appropriate final survey questionnaire. Training of enumerators on appropriate data collection techniques was done. Four graduates (2 for social sciences and 2 for agriculture) were enlisted to conduct the survey. These graduates were well conversant with local languages spoken in the area. Actual data collection was done between April 15th and 30th, 2013. The researcher participated in data 55
collection as well, to oversee the field activities. The completed questionnaires were revised for any possible errors before data entry. Data entered were cross-examined to identify possible entry errors by running the frequencies in SPSS 16.0 and a variable with less than 10 observations was only used for descriptive but not econometric analyses. Data were then transferred to STATA/SE 11.2 to test for heteroscedasticity and multicollinearity in order to ensure unbiased estimates and for detailed statistical analysis in 3.6 below. According to Gujarati (2004), heteroscedasticity test is very important in cross-sectional data (the type used in this study) because of heterogeneous nature of households in rural settings. Breusch-Pagan test is one of the methods used to detect the presence of heteroscedasticity (Appendix Table D1). In the results, a low chi-square value of 0.02 as obtained which is less than 0.05, in this case, according to Greene (2002), it leads to the acceptance of the null hypothesis (homoscedasticity), indicating no heteroscedasticity problem in the data. When there is a perfect linear relationship among explanatory variables, the model cannot be uniquely computed. The term collinearity implies that two variables are nearly perfect linear combinations of one another. A variable whose Variance Inflation Factor (VIF) value is greater than 10 indicates that there is multicollinearity and should be dropped from the model (Greene, 2002). Variance Inflation Factor (VIF) method was preferred to correlation coefficient because the later does not give conclusive results (Gujarati, 2004). Results of the multicollinearity test shown in Appendix Table D2, indicates that extension variable (EXT) was dropped from the econometric analysis because it was found collinear with formal education. The variable also had fewer observations than 10.
56
4.0 CHAPTER FOUR: RESULTS AND DISCUSSION This chapter is comprised of four sections: Section 4.1 presents the general descriptive analysis of the socio-economic and input factors (land, nitrogen fertilizers, seeds, labour and herbicides). Section 4.2 gives the multivariate (probit model) estimations of the socioeconomic factors determining adoption of improved maize. Technical Efficiency (TE) for adopters and non-adopters and a discussion on socio-economic factors determining the level of TE from the SF model are presented in section 4.3. Lastly, estimation of input elasticities from the C-D production function model is presented in section 4.4. Each of the sections 4.2, 4.3 and 4.4 is organized into two sub-parts, corresponding to adopters and non-adopters. 4.1 General descriptive analysis Table 5 shows that, adopters had a comparative advantage in terms of socio-economic factors and inputs compared to non-adopters of improved maize varieties. Apart from income, there was no statistically significant difference in the means of other socio-economic variables between adopters and non-adopters. However, there was significant difference in means of herbicide, seed and plot size between adopters and non-adopters.
57
Table 5: Socio-economic and input factors for adopters and non-adopters Variable
Continuous Socioeconomic variables Education (years) Income (UGX) Family size (people) Non-continuous socio-economic variables Sex (1=male) Market (1=off-gate) Land ownership (1= rented) Extension (1=yes) Credit (1=yes) Input variables Fertilizer (kg) Herbicide (ltrs) Labour (man-days) Seed (kg) Plot (ha)
Adopters n=(35) Mean
Std. dev.
7.23 380,000 5.49
3.47 380,734.68 2.10
Non-adopters (n=125) Mean Std. dev.
5.72 187,000 5.05
3.055 181,920.4 0.04
Total (N=160) Mean
Std. Dev.
6.05 230,000.0 5.15
3.22 250,882.5 2.05
Sig
0.161 0.001 0.388
n(%)
n(%)
N(%)
22(62.9) 03 (8.6) 22(62.9)
72(57.6) 18(14.4) 81(64.8)
94 (58.8) 21 (13.1) 103(64.4)
0.521 0.367 0.290
03 (8.6) 06(17.1) Mean 25.06 3.00 37.91 16.29 2.03
02 (1.6) 14(11.2) Mean 0.69 2.50 36.31 10.38 1.26
05 (3.1) 20 (12.5) Mean 10.44 2.81 36.77 11.98 1.44
0.136 0.356
Std. dev. 35.25 1.05 21.77 11.20 1.21
Std. dev. 1.05 2.07 18.76 6.83 0.73
Std. Dev. 22.13 1.47 19.32 9.15 0.91
0.678 0.060 0.122 0.055 0.005
Note: Due to few observations (n
