THE DETERMINANTS OF ADOPTION, AND IMPACT OF IMPROVED 
SORGHUM VARIETIES IN THE UPPER-WEST REGION, GHANA
BY
JOHN BAPTIST DONSAANANANG JATOE
A THESIS SUBMITTED TO THE DEPARTMENT OF AGRICULTURAL ECONOMY 
AND FARM MANAGEMENT, FACULTY OF AGRICULTURE, UNIVERSITY OF 
GHANA, LEGON, IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE 
AWARD OF THE MASTER OF PHILOSOPHY DEGREE IN AGRICULTURAL
ECONOMICS
DEPARTMENT OF AGRICULTURAL ECONOMY AND 
FARM MANAGEMENT, FACULTY OF AGRICULTURE, 
UNIVERSITY OF GHANA, LEGON.
JUNE, 2000
University of Ghana          http://ugspace.ug.edu.gh
(O  3*4832
I S
b i t e ,  C . l
University of Ghana          http://ugspace.ug.edu.gh
I do hereby declare that except for references to other peoples work which have been duly 
cited, the work presented in this thesis: “THE DETERMINANTS OF ADOPTION, AND 
IMPACT OF IMPROVED SORGHUM VARIETIES IN THE UPPER-WEST REGION, 
GHANA”, is the result o f my original work.
This work has never been presented either in whole or in part for any other degree o f this 
University or elsewhere.
DECLARATION
Dr. (Mrs.) RAMATU AL-HASSAN 
(Major Supervisor)
University of Ghana          http://ugspace.ug.edu.gh
DEDICATION
To my Parents, Na Jatoe-Moni Kangawee II and Kurimah Badingu Daana, 
for their support and guidance.
ii
University of Ghana          http://ugspace.ug.edu.gh
ACKNOWLEDGEMENT
I wish to extend my profound gratitude to Dr. (Mrs.) Ramatu Al-Hassan and Mr. K. O. Baah 
my supervisors, whose constructive criticisms and invaluable suggestions have resulted in the 
successful completion of this work. I also thank all the lecturers in the department for their 
support and contribution to my learning process.
I am greatly indebted to the Social Security Bank for awarding me a Fellowship while 
pursuing this course. I also thank the German Agency for Technical Cooperation (GTZ) and 
the Savanna Agricultural Research Institute (SARI) for supporting me through my data 
collection process. I owe a lot to Mr. Luke N. Abatania and Dr. J.B. Naab for making it 
possible for me to undertake this study at SARI. I am grateful to all the staff of SARI, 
especially the Field Officers who helped with the data collection.
I also wish to thank all my siblings, my cousins and Mr. David Q. Annang for their support, 
both morally and materially. I also thank Mr. Reindorf Darko and Mr. Adika, both o f the 
Department o f Agricultural Economy and Farm Management for their co-operation. To 
colleagues, friends and all those who have contributed in any little way towards the 
realisation of this work, I say God bless you!
To God be the glory!
iii
University of Ghana          http://ugspace.ug.edu.gh
ABSTRACT
The production of sorghum, one of the most important food crops with multiple uses in the 
savanna zone o f Ghana, has been bedevilled with several problems. The most important 
' problems are low yields, lack of improved varieties and striga infestation. These problems 
led to research efforts at the Savanna Agricultural Research Institute to produce improved 
high yielding sorghum varieties that are adapted to the socio-economic and agroecological 
conditions of northern Ghana. This study set out to describe the adoption o f improved 
sorghum varieties over the years; identify the various determinants o f adoption, and intensity 
of use; and to identify the varietal traits or characteristics that are desired by farmers. Finally, 
the study also estimates the rate of return to sorghum research.
The Probit and Tobit models are used to identify the determinants o f adoption and the 
intensity of use o f improved sorghum varieties respectively. The logistic curve is used to 
describe cumulative adoption while the economic surplus model is used to estimate the rate 
of return to sorghum research. A characteristic S-shaped diffusion curve is obtained with an 
estimated rate of diffusion of 0.3% for the period 1988-1998.
It is found that age, available family labour, non-farm income, farmers perception about the 
varieties, farm size and farm type positively influence adoption while extension visits, the 
length of the fallow period and distance to the nearest purchase point for improved seed 
affect adoption negatively. Age of farmer, available family labour, farmers’ perception about 
the varieties and farm size exert a positive influence on intensity o f use while fallow period
University of Ghana          http://ugspace.ug.edu.gh
and the distance to the nearest purchase point for improved seed exert a negative influence. 
The estimated rate o f adoption was 0.40 (40% of sampled farmers). The intensity o f use, 
measured as the proportion of sorghum area under improved varieties, was 0.56 (56.33%) 
following a positive decision. It is found that farmers’ choice o f variety to plant depends on 
yield, maturity period, market value, taste and suitability for local dishes. The rate of return 
to sorghum research is estimated at 8%. Sensitivity analysis shows that increasing the rate of 
adoption to 1% of sorghum area would raise the rate o f return to 109%.
For improved adoption and impact of improved Sorghum varieties, the study recommends 
the following measures:
(i) strengthening o f research-extension-farmer linkages,
(ii) increased farmer education about the varieties,
(iii) improvement in infrastructure and input distribution networks,
(iv) active involvement of farmers in acquisition of inputs,
(v) increased sensitivity of research to farmer resource levels,
(vi) empowering farmers to engage in non-farm income generating activities and
(vii) a more concerted effort at technology transfer.
v
University of Ghana          http://ugspace.ug.edu.gh
TABLE OF CONTENTS
Declaration
Dedication
Acknowledgement
Abstract
Table of Contents 
List o f tables and figures 
List of acronyms
CHAPTER 1: INTRODUCTION
1.1 Background
1.2 Sorghum Production and Improvement in Ghana
1.3 Problem Statement
1.4 Objectives
1.5 Relevance of study
1.6 The Study Area/Location
1.7 Organisation o f Study
CHAPTER 2: LITERATURE REVIEW
2.1 Introduction
2.2 History of Adoption Studies
2.3 Types o f adoption Studies
2.4 Determinants of Adoption
2.4.1 Socio-economic Characteristics
2.4.2 Institutional Factors
2.4.3 Technical Characteristics
2.5 Methodologies of Adoption Studies
2.6 Economic Impact Assessment
2.7 Factors Affecting the Impact of Agricultural Research
2.7.1 Research System Performance
2.7.2 Extension and Research Linkage
2.7.3 Linkages Beyond the Local System
2.7.4 Government Policies
2.7.5 Markets
2.7.6 Climate
2.8 Methodologies of Impact Studies
2.9 Findings of Rate of Return Studies
2.10 Conclusion
l
ii
iii
iv
vi
viii
ix
1
4
5
7
7
8
11
12
12
14
15
16
18
19
20
22
22
23
23
24
25
25
26
26
30
31
vi
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 3: METHODOLOGY
3.1 Introduction 33
3.2 Concept of Adoption 33
3.3 Conceptual Framework and Analytical Tools
for the Pattern of Adoption 34
3.3.1: The Logistic Curve 3 5
3.4 Theoretical Framework for Determinants of Adoption 37
3.4.1 Methodological Considerations 38
3.4.2 Empirical Multi-nomial Probit Model 42
3.5.1 Theoretical Framework and Methodological Issues for
Intensity of Adoption 44
3.6 Choice of Variables and Work Hypotheses 46
3.6.1 Socioeconomic factors 47
3.6.2 Technical Characteristics 51
3.6.3 Institutional factors 51
3.7 Data Requirements and Data Collection for the Probit and Tobit Analyses 53
3 .8 Method of Analysis for Farmers Varietal Preferences 54
3.9 Theoretical Framework and Methodological Issues for
Estimating Rate of Return on Sorghum Research 54
3.10.1 Method of Estimation 57
3.11 Data Requirements for Impact Assessment (Rate of Return Computation) 59
CHAPTER 4: EMPIRICAL RESULTS AND DISCUSSION
4.0 Introduction 64
4.1 Characteristics o f Sampled Farm-Households 64
4.2.1 The Parttern of the Adoption of Improved Sorghum Varieties in the
Upper West Region of Ghana. 66
4.2.2 Determinants o f the Adoption of Improved Varieties
of Sorghum 68
4.2.3 Determinants of Intensity o f Adoption: Tobit Model 74
4.3.4 Farmers preferred Varietal Characteristics 77
4.3.5 The Impact of Sorghum Research: Results of Rate of Return Analysis 80
CHAPTER 5: SUMMARY, CONCLUSION AND POLICY IMPLICATIONS
5.1 Summary 83
5.2 Conclusions 1 85
5.3 Policy Implications 86
5.4 Suggestions for Future Research 88
References
Appendices
89
99
University of Ghana          http://ugspace.ug.edu.gh
LTST OF TABLES AND FIGURES
Table 1.1: Sorghum Production Statistics for the Upper West Region 1989-1998 2
Table 3.1: Definitions of Variables used in the Models 48
Table 4.1: Main Characteristics of Sampled Households, UWR, Ghana, 1999/2000 65
Table 4.2: Estimates of Logistic Function, UWR, Ghana, 1999/2000. 66
Table 4.3: Estimates of the Determinants o f the Adoption o f Improved Varieties
Sorghum: Probit Model (1999/2000, UWR, Ghana) 68
Table 4.4: Estimates of the Determinants o f Intensity of Adoption:
Tobit Model (1999/2000, UWR, Ghana) 74
Table 4.5: Elasticities of the Predicted Probability of Adoption and Expected Use
Intensity o f Improved Sorghum Varieties (1999/2000, UWR Ghana) 77
Table 4.6: Farmers Preferred varietal characteristics according to percentage of
respondents (1999/2000, UWR, Ghana). 78
Table 4.7: Comparison of Improved and Traditional Varieties by Farmers 79
(1999/2000, UWR Ghana)
Figure 3.1: Producer and Consumer Surplus 55
Figure 3.2: Supply Shift with Improved Varieties 57
Figure. 4.1: Diffusion Curve for Improved Sorghum Varieties
(UWR 1999/2000) 67
viii
University of Ghana          http://ugspace.ug.edu.gh
LIST  OF ACRONYMS
CIMMYT International Maize and Wheat Improvement Center
CIP International Potato Center
GTZ German Agency for Technical Co-operation
IARCs International Agricultural Research Centres
ICRISAT International Crops Research Institute for the Semi-Arid Tropics
NAES Nyankpala Agricultural Experiment Station
NARS National Agricultural Research System
NGOs Non-governmental Organisations
SARI Savanna Agricultural Research Institute
SGA Seed Growers Association
TFP Total Factor Productivity
UWR Upper West Region
WARDA West Africa Rice Development Association
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 1 
INTRODUCTION
1.1 Background
Sorghum (sorghum bicolor (L) Moench) commonly called "guinea corn^'in'West Africa, is 
an important food crop with multiple uses. Sorghum is an important and widely cultivated 
cereal crop in the Guinea savanna zone o f Ghana. According to an Agricultural Census in 
1986, it was grown on some 181 thousand hectares. Sorghum is only exceeded by maize 
in the Northern Region and by millet in the Upper East Region in terms of area cultivated 
with cereals (MOFA, 1997). The planting can be on the flat, on ridges or on mounds 
depending on the crop mixtures. Sorghum is planted between mid to end o f May and 
early June in the Guinea savanna zone.
In Ghana, sorghum is produced in the Guinea Savannah zone, which occupies the three 
Regions o f Northern Ghana. It is also grown to a limited extent in Brong-Ahafo, Volta 
and Greater Accra Regions. Traditionally, it is grown as a rain-fed crop in mixtures with 
other cereals and legumes. It is only in the Upper West Region that sole crops of sorghum 
are common (Schipprack and Mercer-Quarshie, 1984).
Although sorghum production is most prevalent in Northern Ghana the Upper West and 
Northern Regions are the major producers of the crop. The Upper West Region ranks 
second in the production of sorghum (MOFA, 1997). In fact, the Upper West Region is in 
the heart of the area often described as the "cereal basket" o f the country, which produces 
the national supply of sorghum and millet as well as groundnut and cowpea (Terbobri et
University of Ghana          http://ugspace.ug.edu.gh
al, 1998). Table 1 shows sorghum production statistics for the region over the period 
1989-1998.
Table 1-.Sorghum Production Statistics for the Upper West Region 1989-1998
Year Output
(MT)
Area cultivated 
(Ha)
Yield
(MT/Ha)
As % o f  National Output
1989 49000 75700 0.65 22.8
1990 53942 64217 0.84 39.7
1991 58776 74400 0.79 24.3
1992 63434 73816 0.86 24.9
1993 102900 73500 1.40 31.3
1994 108944 78098 1.39 33.6
1995 109498 84200 1.30 30.4
1996 110413 88000 1.25 31.2
1997 106623 89230 1.19 32.1
1998 116461 94410 1.23 34.7
Source: Derived from PPMED statistics
A considerable amount of the sorghum produced is consumed directly as human food. 
The grain is mainly used in preparing koko', tuo zaafi2 (TZ), and maasa3 The leaves 
provide fodder for farm animals while the stalks are used in fencing, roofing, as fuel and
1 Porridge prepared from dough/fine flour.
2 Prepared by stirring and addition of flour to porridge until it is cooked and viscous; served with soup.
2
University of Ghana          http://ugspace.ug.edu.gh
for weaving baskets and mats. In addition, the ash from burnt sorghum stalks provides 
potash for soap making, a cottage industry for some rural women (Songsore and Denkabe, 
1995).
Above all, sorghum is utilised in the production of a local opaque beer known as pito. 
Pito brewing, an important cottage industry for women in Northern Ghana has assumed 
greater commercial proportions throughout the country. This can be seen in the rise in the 
number ofpito  bars in all major towns in Southern Ghana. Besides, it has been established 
that sorghum is usable and has been used in malting and brewing clear lager beer in 
Nigeria, Eastern and Southern Africa, Mexico, Germany and Belgium (Atokple, 1992), 
and hence could become an import substitute in the brewing industry in Ghana. The crop 
also enters into considerable inter-regional trade between the Upper West region and 
southern Ghana.
The production of sorghum and other crops in the Upper West Region is in the hands of 
small-scale resource-poor farmers, whose farms on the average hardly exceed 2.0 hectares 
per household. Majority of the people in this area live in rural homesteads, where the 
production o f crops and livestock is their main source of livelihood. These subsistence/ 
peasant farmers use little external input (such as improved seed and chemical fertilizers) 
and often sell the little surplus of production in excess of household consumption needs to 
supplement household income.
3Cake prepared from fermented dough.
3
University of Ghana          http://ugspace.ug.edu.gh
Despite the importance of sorghum in the Upper West Region, and the country at large, 
production o f the crop is hampered by several edaphic, physiological and biological factors 
culminating in extremely low output per unit area. Depending on the year, average yields 
range between 600 and 1400kg/ha in the Region (Table 1). These yields are rather low as 
compared to the achievable 2 Mt/Ha reported by Abatania (1999). Specific constraints 
include declining soil fertility, lack of improved varieties, and slriga infestation (Terbobri 
et al., 1998). These constraints triggered the inception of the sorghum (and millet) 
improvement programme at the Nyankpala Agricultural Experiment Station (NAES), now 
the Savannah Agricultural Research Institute (SARI), over three decades ago.
1.2 Sorghum Production and Improvement in Ghana
The most common land races of sorghum in Ghana are Guineense, Caudatum and Bilocor. 
However, the loose panicled Guinea type (also known as Belko type in Ghana) with white 
corneous grain is the most widely cultivated and most widely adapted o f all the races 
(Frolich and Buah, 1991). These races are very tall, highly photoperiod sensitive and late 
maturing. The evidence suggests that there has been considerable selection for good 
weathering and storage qualities (Harlan and de Wet, 1972). The hard and corneous seeds 
suffer little or no damage from insects even under the most primitive storage conditions. 
Mould damage is also reduced under wet conditions probably due to the open pendulous 
panicles and gaping glumes.
Farmers grow mostly the tall photoperiod sensitive cultivars and risk o f complete crop 
failure exists should the rains stop early, since these cultivars often flower very late
University of Ghana          http://ugspace.ug.edu.gh
However, these types are preferred because their long stalks are used for fencing, building 
and making baskets, hats and barns. They also have good seed quality, that is, hard, 
vitreous endosperm (Frolich and Buah, 1991). Farmers usually retain large and uniform 
panicles with large grains for use as seed stock for the next planting season. Farmers 
generally are interested in stable grain yields and consequently, cultivate a mixture o f two 
or more cultivars on the same piece o f land.
The grain yield of traditional sorghum varieties is very low. It ranges between 600 and 
1400kg/ha. However, with the use of improved varieties and better production practices, 
average yields can be pushed to between 2000 and 3000kg/ha (Frolich and Buah, 1991).
1.3 Problem Statement
Since the early 1960s, developing countries, assisted by foreign donors, have invested 
resources to strengthen their agricultural research systems. Agricultural Economists have 
often supported this strategy, arguing that technological innovations in agricultural 
production drive the development of the agricultural sector, which in turn contributes to 
the development of the general economy (Mellor, 1966; Eicher and Staatz 1984).
In Ghana, one such example is the Savanna Agricultural Research Institute (SARI), with 
support from the Ghana Government and the Federal Republic of Germany through the 
Deusche Gesellschaft fur Technische Zusammenarbeit (GTZ). An important component 
of the research mandate of SARI is the Sorghum Improvement Programme. In line with 
its mandate for research in Northern Ghana, the Savanna Agricultural Research Institute
5
University of Ghana          http://ugspace.ug.edu.gh
has focused attention on removing the constraints associated with sorghum production so 
as to step up sorghum production in the mandate area. The objective of the sorghum 
improvement programme is to develop high yielding varieties of sorghum with specific 
adaptation to different ecological conditions of the savanna. In fact, the programme is to 
identify and develop early and medium maturing varieties that have good grain quality 
(suitable for local dishes) and are resistant/tolerant to the major pests and diseases of 
sorghum.
Through SARI's sorghum improvement programme a number o f improved varieties have 
been released to farmers after being tested on-farm. Notable among such varieties are 
Naga White, Framida, Kapaala, Kadaga and Dorado (Terbobri et al, 1998). The 
superiority o f these varieties over the local varieties, in terms o f yield, maturity period and 
striga tolerance, has been proven by researchers in on-farm trials. However, the picture 
on the ground, in terms of adoption, is unknown and yet to be examined. The extent to 
which farmers have adopted these varieties is not known. The factors affecting adoption 
or non-adoption are also not known, neither are the constraints to adoption.
In addition, the ever-increasing scarcity of resources has prompted donors (and 
governments) to be much more results oriented. Priorities must be set as choices have to 
be made. Investments in agricultural research must therefore be justified in terms of 
returns to these investments. Thus it is common to have donors asking for evidence of 
people-level impacts of their investments in agricultural research. However, the benefits 
of agricultural research may be delayed, gradual and widely dispersed and hence may not
6
University of Ghana          http://ugspace.ug.edu.gh
be easy to capture. Economic impact studies provide the tools for measuring these 
people-level impacts, but no such studies have been conducted to assess returns to 
sorghum research. Thus the basic question that arises is whether or not past investments in 
sorghum research at SARI have generated sufficient returns to justify continued 
investment.
1.4 Objectives
The general objective is to evaluate the adoption and impact o f improved sorghum 
varieties introduced by the Savannah Agricultural Research Institute (SARI) and provide 
the needed information for researchers and extension organisations to guide their work. 
The specific objectives are to:
(i) describe the pattern or spread of adoption;
(ii) determine the factors that influence the adoption o f improved sorghum varieties;
(iii) estimate the rate and intensity of adoption;
(iv) compare quality of new and traditional varieties with those preferred by farmers;
(v) estimate the rate of return to research into improved sorghum.
1.5 Relevance of study
Though research efforts have made available improved varieties o f sorghum to farmers, 
the adoption of these technologies is yet on a small scale and largely unknown. Currently 
information relating specifically to the adoption of improved varieties o f sorghum in 
Ghana is not available. No attempt has been made to document the adoption diffusion 
pattern and use intensity of these improved technologies, and hence their impact. The
7
University of Ghana          http://ugspace.ug.edu.gh
provision of micro level data or information of this sort is critical for the direction and 
prioritization o f future sorghum research. It will help research and extension to be more 
effective in responding to farmers' needs. Moreover such information is necessary for 
financial support for future sorghum research as donors get more and more results 
oriented. In addition, this study may provide lessons for other crop improvement 
programmes, both here in Ghana and elsewhere, Hence the study is very germane.
1.6 The Study Area/Location
The Upper West Region, in the northwestern area o f the country with its capital at Wa, 
was carved out o f the former Upper Region in 1983. It consists o f five administrative 
districts made up of Wa, Nadawli, Lawra, Jirapa-Lambussie and Tumu. The Region is 
bounded to the east by the Upper East Region, Northern Region to the south, Burkina 
Faso to the north and Cote d'Ivoire to the west. The region covers a total land area of 
18,476 square kilometres and had an estimated population o f 439,161 persons in 1984. 
Its population density ranges between 20 and 60 persons per square kilometre. The 
economic base o f the area hinges on agriculture with over 80% of the population 
depending on farming for their livelihood (Ghana Statistical Service, 1984). It is the most 
backward region in terms of income per capita and access to basic needs. In fact the 
overall picture is one of both absolute and relative poverty (Songsore and Denkabe, 
1995).
This region lies within the Guinea Savanna belt often described as the Savanna high plains 
with an average altitude of between 180 and 300 metres above sea level. The area has a
8
University of Ghana          http://ugspace.ug.edu.gh
generally gently undulating landscape where topography is no barrier to agriculture 
(Dickson and Benneh, 1970). The area is generally well drained with the main rivers 
being the Black Volta and its many tributaries to the west and, the Kulpawn and its 
tributaries to the east. The two main soil types are the ground water lateritic soils and the 
savanna orchrosols along the Black Volta.
The most critical factor affecting agricultural activity, rainfall, is characterised by 
seasonality, variability and unreliability. The rainy season extends from April to October 
with an annual total of about 115cm (Dickson and Benneh, 1970). This situation limits 
peasants to only one harvest a year towards the end of the rainy season.
The main farming systems are bush fallow and compound farming. Bush fallowing is a 
system of rotation of crops or fields and bush. After a plot of land has been cultivated for 
a number o f farming seasons, it is abandoned and allowed to revert to secondary 
vegetation. The length o f this resting period depends on how pressing the need is for land 
for cultivation. In sparsely settled areas it may be as long as fifteen or more years while in 
densely settled areas it may be as short as three years (Dickson and Benneh, 1970). 
Compound farming is a permanent system in which the land around the individual 
compound household is manured and cultivated intensively by the members of the 
household for their staple crops. This system is most common in the densely settled areas 
of Lawra and Jirapa-Lambussie districts.
9
University of Ghana          http://ugspace.ug.edu.gh
Land is controlled through family ownership, a system that guarantees every member of a 
household access to a plot of land from what the household already enjoys use rights. 
This situation varies, but there is no sharecropping neither is there buying and selling of 
land except in the towns, for housing purposes. It should be noted that women do not 
own land.
The seasonality in rainfall often leads to soil drought during critical periods in plant 
growth thereby reducing yields drastically from year to year. Thus the region faces a 
perennial long "season of hunger" with low food stores and shortages from March to 
about July (Hunter, 1967; Songsore, 1985). During this period, those peasants who 
undertook distress sales at harvest for their cash mediated subsistence consumption often 
have to repurchase food at higher prices.
In order to turn this situation around and ensure food security, Songsore and Denkabe 
(1995) suggest a people oriented development strategy, one which is self-reliant and 
sustainable, and taps strength from within, relying on resources locally available and 
technologies adapted to the area. They argue that the problem with the region's 
agriculture is the use o f primitive tools and the lack of improved varieties and breeds to 
enhance productivity, adding that the available technologies are often technocratic, top- 
down style alien ones beyond the reach of the majority o f small farmers. The current 
study seeks to evaluate the case of improved sorghum varieties from the adaptive research 
efforts of the Savanna Agricultural Research Institute.
10
University of Ghana          http://ugspace.ug.edu.gh
The Upper West Region was chosen for this study mainly because it is the only region 
within the mandate area o f the Savanna Agricultural Research Institute where these 
improved sorghum varieties were vigorously promoted by the defunct Global 2000 
Sorghum Programme.
1.7 Organisation of Study
The study is organised into five chapters. Chapter 2 covers an exploration of literature on 
adoption and impact o f innovations. It discusses, among other things, the determinants of 
adoption and impact of agricultural technologies. Chapter 3 provides coverage o f the 
theoretical framework for the study, methods of estimation, the data requirements and 
data collection procedures used. The results of the analysis are presented and discussed in 
the fourth chapter, while chapter 5 presents the summary, conclusions and policy 
implications of the study.
l l
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 2 
LITERATURE REVIEW
2.1 Introduction
In this section literature on adoption and the impact of innovations in agriculture is 
reviewed. The chapter seeks to identify some o f the factors most likely to influence 
farmer adoption behaviour as well as those affecting the impact o f new technologies. It 
explores the methodologies employed by similar past studies and discusses some of their 
findings. The review of literature is also expected to give direction as to the choice of 
explanatory variables to include in the models and the most appropriate methods of 
analysis to use.
2.2 History of Adoption Studies
Griliches conducted one o f the earliest adoption studies on hybrid corn in the United 
States o f America (Griliches, 1957). His pioneering role reflects clearly in part o f his 
conclusion:
“this study does indicate that at least the process o f innovation, the process o f 
adapting and distributing a particular invention to different markets and its 
acceptance by entrepreneurs, is amenable to economic analysis. It is possible 
to account for a  large share of spatial and chronological differences in the use 
o f hybrid corn with the help o f economic variables”1 (Griliches, 1957, 
Econometrica 25 p522).
1 Griliches, Zvi (1957) “Hybrid Corn: An Exploration in the Economics o f Technological Change” 
Econometrica 25:501-522
University of Ghana          http://ugspace.ug.edu.gh
Since the pioneering work of Griliches, considerable attention has continued to be devoted 
to the adoption o f new agricultural technologies. To mention a few, Mansfield (1961) 
examined the diffusion of new manufacturing techniques among major industrial firms; 
Shaka and Flinn (1985) studied the adoption of modern varieties and fertilizer use on rice 
in Nepal; Akinola (1987) studied the adoption of tractor hiring scheme in Nigeria; Hailu 
(1990) studied adoption of improved seed varieties, bullock traction technology and 
agrochemicals in northern Ghana; Zinnah et al (1993) studied the adoption of improved 
mangrove swamp rice technology in West Africa; Hounkpe (1999) studied the adoption of 
narrow maize crib in Southern Benin.
Although the importance of technical innovations is well understood, the factors affecting 
the adoption o f innovations are less obvious. But one would expect comparative 
advantage considerations (dynamic as well as static) to be crucial in determining patterns 
of adoption of innovations. This was observed by Griliches (1957), Manfield (1961) and 
Nelson (1968). However, most studies o f the adoption behaviour o f individual farmers 
use static analysis that relates the degree of adoption to factors affecting it.
The tendency has been to take agriculture as a static receptacle into which we pour new 
production technology as a social input. The technology is usually regarded as 
appropriate, and farmers are viewed as static units to which technology is to be 
transferred, with the result that they will all be better off (Schuh, 1989). However, things 
do not work this way. New technology has implications for resource productivity and 
hence relative resource allocation. Also, the farmer is a social being with well-defined
13
University of Ghana          http://ugspace.ug.edu.gh
tastes and preferences. Besides, in subsistence agriculture, profit maximization may not be 
the primary objective. This realisation has led to an evolving new trend in which several 
factors, including the institutional, environmental and the technical characteristics of the 
technology itself, are being examined in adoption studies.
2.3 Types of adoption Studies
The form and focus o f an adoption study depends on the purpose o f the study. Whatever 
the purpose, however, the results often require careful analysis and presentation. 
Generally, adoption studies may seek to document the degree of adoption, especially 
where the researcher is interested in the rate o f progress of technology generation effort. 
On the other hand, adoption studies may want to analyse adoption patterns and try to 
identify/explain the determinants o f adoption.
There are two basic types of adoption studies. The first explains adoption by seeking out 
the opinions and observations of farmers. This type of adoption study tries to explain 
farmer decision-making with regard to innovation adoption through farmer decisions. It 
relies on the fact that farmers usually know what they like and what they do not like about 
a new technology and are able to express their opinions. These opinions reflect their own 
experience (CIMMYT, 1993). One of the most commonly used tools (in this qualitative 
kind o f analysis) is “decision trees” (CIMMYT, 1993). Decision trees are a method of 
representing a farmer’s choice about a new technology as series of decisions, arranged in a 
hierarchy (Gladwin, 1989).
14
University of Ghana          http://ugspace.ug.edu.gh
The second makes use of statistical comparison of the farmers’ adoption behaviour with 
various characteristics o f the farmers’ environment. Farmers may be asked to describe 
these characteristics and the researcher then tries to trace the causal links, using various 
statistical tests. These tests aid the researcher to provide a consistent, coherent, and 
logical explanation for the adoption patterns observed.
The simplest method in use here is the contingency tables (especially where the variable/ 
characteristic is nominal or categorical), using the chi-square test. Where the variable is 
continuous, one option is to compare means for adopters and non-adopters. The relevant 
test here is the t-test. Correlation and regression analyses are employed where adoption 
itself is represented as a continuous variable. In dealing with multivariate relationships, 
researchers often call upon multiple regression techniques. Two related multifactorial 
analytical techniques that are particularly useful for adoption studies are logit and probit 
analysis. Another useful technique is the Tobit model.
2.4 Determinants of Adoption
Through various national and international efforts, several new agricultural technologies 
including improved varieties of some major cereals "suitable for many of sub-Sahara 
Africa's agro-ecological zones"2 are made available. For instance, Hailu (1990) reported 
the existence of thirteen different improved varieties of four crop species among which 
were two sorghum varieties. However, the general view in the literature is that the use of 
improved seed varieties is low and often highly limited to major export crops. For
2 FAO (1986). African Agriculture: Input Supply and Incentive Policies. Annex III: Rome.
15
University of Ghana          http://ugspace.ug.edu.gh
instance, FAO (1986) reported that the use of improved seed in sub-Saharan Africa 
accounts for only 10 percent of cereal production whereas major export crops are entirely 
grown from improved seeds. Also Nweke (1988) stated that farmers are more willing to 
adopt new technologies for producing cash crops than for producing crops that provide 
for family food security. As Hailu put it,
"the mere fact that a new technology is introduced in smallholder farming by itself 
does not guarantee a widespread adoption and efficient use"3.
He argued that success in this respect depends on the fulfilment o f specific socio­
economic, technical and institutional conditions. Several studies support this view.
2.4.1 Socio-economic Characteristics
In general, household decision-making with respect to farming operations is influenced not 
only by the economic merits o f  actions but also by social factors pertinent to the 
household and those operating at the community level. For example Akinola (1987) 
observed that in a traditional African setting, the strength of family ties often compels farm 
workers to tolerate very poor working conditions and wages as a result o f their family 
loyalty. The farm operator thus assumes a financially advantageous position to spend 
more money on adoption of new technologies. In addition, such factors as income/wealth, 
farm size, relative advantage, innovation cost, and profit have been found to influence 
farmers' adoption behaviour (Kislev and Shchori-Bachrach, 1973; Rogers, 1983;
3 Hailu Z. (1990). The Adoption of Modem Fann Practices in African Agricnlture:Emmrical Evidence about the 
Tmnaets o f  Household Characteristics and Input Supply Systems in the Northern Region o f Ghana. Nyankpala 
A gricu ltu ra l R esearch  Report No.7.
16
University of Ghana          http://ugspace.ug.edu.gh
CIMMYT, 1993; Hounkpe, 1999). The innovation cost may restrict the use of certain 
technologies to only farmers with a certain level of income. Farm size may tend to be 
associated with income in most cases, and relative advantage and profit serve as 
motivation for potential adopters, which may explain their influence.
Also, with education and experience come new knowledge and skills, and an associated 
increase in confidence, as uncertainty is reduced. It has been argued that education 
enhances one’s ability to receive, decode and understand information (Schultz, 1964). 
Experience has effects similar to education as it builds confidence and dispels doubt 
(Kebede, 1992). Also, age, leadership status and gender may determine whether or not an 
individual takes decisions regarding the choice o f technology used on the farm. For 
example, Pannin (1988) noted that elders in traditional African societies are widely 
accepted as superior because experiences and decisions relating to the adoption of 
innovations are positively influenced by the age factor. However, some researchers have 
argued that younger farmers, being more adventurous and with a longer planning horizon, 
are more inclined to accept innovations than older ones (Poison and Spencer, 1992).
It is therefore not surprising that age, education, experience, leadership status and gender 
have been found to influence farmers' adoption behaviour (Schultz, 1964; Petzel, 1976; 
Mueller and Jansen, 1988; Lin 1992; CMMYT, 1993).
One factor that has not featured frequently in adoption studies is the technological level of 
the farmer. It should be noted that there exist positive interactions between certain
17
University of Ghana          http://ugspace.ug.edu.gh
technologies. Such technologies therefore complement each other in the realisation of 
their potentials. For example, the optimal effect of chemical fertilizers is reached with 
simultaneous use of modern crop varieties, as these mainly show a significantly greater 
response to fertilizer application than traditional ones (Ruthenberg, 1985). Similarly, 
Byerlee (1986) noted that adoption of certain technologies by small holders follows a 
sequential pattern and the adoption of one is expected to hasten the adoption of related 
components.
2.4.2 Institutional Factors
These are self-evident, as it does not require much to realise that the availability of inputs, 
credit and subsidies create an enabling environment and broadens the scope for a larger 
number of potential adopters. Previous adoption studies have found access to credit to 
be an important determinant o f innovation adoption (Shakya and Flinn, 1985; Akinola, 
1987). However, it is worth noting that Hailu (1990) noted that access to institutional 
credit for the majority of small farmers in northern Ghana is hardly possible. He also 
observed that increased use o f modern agricultural inputs in northern Ghana is constrained 
by deficiencies in supply and distribution systems. He stressed that input availability at the 
right time and place is a major bottleneck to adoption.
Through education, extension service conscientizes farmers, clears some doubts and may 
appeal to some farmers to give an innovation a try. Also, markets for products assure 
producers of an outlet for their produce and serve as a motivation.
18
University of Ghana          http://ugspace.ug.edu.gh
This may explain why it is widely documented that availability of inputs, credit, subsidies, 
extension service and markets for products are some of the institutional factors affecting 
the adoption behaviour of small farmers. Evidence in support of this abounds (Akibonde, 
1969; Rogers, 1983; FAO, 1986b; CIMMYT, 1993).
2.4.3 Technical Characteristics
Technology refers to a way or method of carrying out an operation/activity while 
innovation means a new approach or method. These (particularly the later) have 
implications for the available resources and the end product. Given that culture (a way of 
life o f a people) and tastes and preferences do not change overnight, the innovation itself 
or its technical characteristics are bound to influence adoption. Some important 
innovation characteristics found to be influencing adoption are compatibility with the 
existing farming system, suitability for popular uses and opportunities for new uses 
(Edwin, 1996; Hounkpe, 1999).
From the foregoing, one can say that farm level decisions about the adoption o f improved 
technologies are often governed by socio-economic and institutional factors as much as 
they are by technical factors. The current study seeks to explore the relationship between 
all these categories of factors, and farmers' behaviour with respect to the adoption of 
improved sorghum varieties and the associated technologies.
19
University of Ghana          http://ugspace.ug.edu.gh
2.5 Methodologies of Adoption Studies
Several different methods of analysis have been used in assessing the adoption of 
innovations. Most often, adoption studies have focussed on the individual (decision­
maker, particularly the household head) as the unit of observation. Again, most studies 
define adoption as being dichotomous, that is, an innovation is either adopted or not. The 
methods of analysis can generally be categorised as qualitative or quantitative, but the 
quantitative techniques dominate the literature. However, as will be seen in the discussion 
that follows, researcher perception, the purpose and the scope of the study influence the 
choice o f analytical techniques.
Francis (1976) seemed more interested in identifying early adopters in his study on 
population characteristics and development activities. He thus calculated adoption scores 
on the basis of the date that each of several new agricultural ideas and practices were first 
used by each individual. Alao (1973) assessed the level of adoption by measuring the 
percentage of farmers, out of the total number of sampled farmers, that are using the 
innovation in question. This is exactly what the logistic or cumulative normal curve, used 
to describe cumulative adoption over time, does but on a yearly basis, with the Y axis 
defined as the proportion of farmers using innovation (Edwin, 1996). However, when Y 
is defined as the proportion of area under the new practice, it gives an estimate o f use 
intensity. The method then becomes similar to the acceptability index proposed by Waugh 
(1982, cited in Engmann 1993). This index is computed as a percentage of representative 
farmers using the innovation multiplied by the percentage of land area being used for the
20
University of Ghana          http://ugspace.ug.edu.gh
new practice. Engmann (1993) modified this index to determine the intensity of adoption 
of soyabean in Ghana. She calculated it as:
Modified Acceptability Index = (N,/ Nt)* (LJ  L,) * 100 (2.1)
where Ns is the number of farmers in the sample using the new practice, Nt is the total 
number of farmers in sample, Ls is land area used for new practice by sample and Lt is the 
total land area cultivated by farmer.
In all the above cases, one thing stands out. The determinants of adoption do not seem of 
interest to the researcher. In contrast, Roe (1983) included risk preferences along with 
other factors in his adoption models. Several other adoption studies have explored the 
determinants of adoption. Generally, in many such studies models used are exponential 
functions, both univariate and multivariate. In fact, probit and logit models and their 
modified forms have been used extensively (Shakya and Flinn, 1985; Hailu, 1990; Kebede 
and Coffin, 1990; Edwin, 1996; Hounkpe, 1999).
Several studies have used the analytical techniques just described to explore the 
relationships between the adoption of improved agricultural technologies and a range of 
interdisciplinary factors. Some recent experiences include Edwin (1996) who used the 
logistic curve to examine the adoption pattern and the tobit model to analyse the adoption 
of improved rice varieties in Sierra Leone; Hounkpe (1999) used the Probit model to 
analyse the adoption of Maize narrow crib in Benin; Akinola and Young (1985) applied 
Tobit model to analyse the adoption and use of cocoa spraying chemicals in Nigeria; 
Shakya and Flinn (1985) used Probit and Tobit models to examine the adoption of modern
21
University of Ghana          http://ugspace.ug.edu.gh
varieties and fertilizer use on rice in Nepal; Akinola (1987) used probit model to analyse 
the adoption of tractor hiring service scheme in Nigeria; Hailu 1990 used the logistic 
curve, and the Probit and Tobit models to study adoption of improved seed varieties, 
bullock traction technology and agrochemicals in northern Ghana.
2.6 Economic Impact Assessment
In view o f the scarcity o f resources and the unlimited needs of man, all governments and 
donors need to justify their investments. However, the economic value o f many a public 
investment may not be obvious. It is particularly difficult to observe the impact of 
agricultural research, mainly because the benefits are diffused over several years and 
among a host o f dispersed producers and consumers. Economic impact studies provide 
the tools for measuring those benefits and comparing them with the costs o f the research. 
Approaches to impact assessment fall into three main categories: econometric approaches, 
programming methods and economic surplus methods. Although all these methods are 
used widely, the economic surplus method is the most popular because its basic techniques 
are easy to grasp, it has the least data requirements and can be applied to the broadest 
range o f situations (Masters et al., 1996).
2.7 Factors Affecting the Impact of Agricultural Research
An important aspect o f impact assessment is an analysis o f factors that influence the 
impact o f technology development and transfer i.e. factors that affect the benefit and cost 
streams of investments in agricultural research. These qualitative analyses seek to explain 
why an investment had high, low or negative returns, unlike quantitative analyses such as
22
University of Ghana          http://ugspace.ug.edu.gh
rate o f return calculations, which simply estimate the financial and/or economic returns to 
investments. Qualitative analysis may also help explain how returns are distributed. With 
these insights, the policy choice set is expanded to include policies that alter the potential 
returns o f investments.
A search through the literature reveals that factors such as research system performance, 
extension and research linkage, linkages beyond the national/local system, government 
policies, markets and climate have important influences on the impact of investments in 
agricultural research.
2.7.1 Research System Performance
In order to achieve optimum impact, the research system must be effective in generating 
improved technology. This requires the setting of appropriate priorities, effective 
scientific leadership, the availability of favourable incentives and adequate human and 
financial resources. Oehmke and Crawford (1993) noted that a combination of well 
funded National Agricultural Research System (NARS), International Agricultural 
Research Centres (IARCs) and donor efforts in countries such as Zambia, Kenya and 
Cameroon have resulted in the release of improved technologies that were adopted by 
farmers.
2.7.2 Extension and Research Linkage
Effective interaction between research and extension is necessary if the much-desired 
farm-level impact of agricultural research is to be achieved. A system that links together
23
University of Ghana          http://ugspace.ug.edu.gh
all the pieces of the development puzzle (Mazzucato and Ly, 1994) is what is required. 
There must be a well co-ordinated research and extension system. Such a system should 
provide for regular meetings and offer opportunities for interdisciplinary interaction 
among researchers and extension staff. Participation at such meetings should include staff 
of the agriculture ministry, researchers, NGOs and farmers.
For example, it was noted that in Cameroon such interaction shifted the research agenda 
from a primary focus on producing high grain yields to addressing post-harvest storage 
constraints. This system provides a feedback from the farm to researchers, and the fact 
that researchers are aware of the existence o f a network that will diffuse the product of 
their work is good motivation.
2.7.3 Linkages beyond the Local System
International Agricultural Research Centres (IARC) and other international networks can 
be very important partners in developing and sustaining an effective agricultural research 
system. By collecting, maintaining and distributing germplasm, they act as important 
catalysts for agricultural development (Sterns and Bemsten, 1994). Linkages, via donor 
projects, between the NARS and IARCs therefore enhance the technology development 
and transfer process. For example, IARCs serve as a source of alternative cultivars or 
varieties as in the case of kapaala, an ICRISAT introduced material in Ghana (NAES 
Annual Report, 1994). Besides, they provide an expanded resource base and also provide 
for human resource development through seminars and workshops.
24
University of Ghana          http://ugspace.ug.edu.gh
2.7.4 Government Policies
The line of action pursued by government can be a source o f failure o f an otherwise 
successful agricultural innovation. Policies affecting the supply and price o f agricultural 
inputs, and the market for and price of agricultural outputs clearly have effects on the 
impacts o f improved technology. Ahmed et al (1992) showed that the adoption of 
improved sorghum in Sudan was frustrated when government pricing policy changed 
adversely. Also, in Mali, Boughton and de Frahan (1994) observed that a well co­
ordinated sub-sector and mechanised farming system resulted in high rates of technology 
adoption. However, with the withdrawal o f marketing services and guaranteed prices, 
maize farmers changed their choice of technology and virtually resorted to their traditional 
practices.
2.7.5 Markets
Even in subsistence agriculture peasant farmers often have to sell farm produce to meet 
their cash-mediated subsistence requirements. It is therefore not surprising that output 
markets and input supplies (including credit) should play key roles in supporting or 
restraining adoption of productivity-increasing technology. Oehmke and Crawford (1994) 
found that lack o f effective, improved seed multiplication and distribution was a critical 
constraint in Niger and Uganda, as was lack of fertilizer in Mali. It was similarly found 
that limited output markets were constraints in Mali and Uganda.
25
University of Ghana          http://ugspace.ug.edu.gh
2.7.6 Climate
Harsh climatic conditions introduce considerable risk in farming. Variability in rainfall is 
an important risk factor in rain-fed agriculture. The more variation in the quality and 
distribution o f rainfall, the more the production of rain-fed crops becomes unstable. The 
risk that climate imposes on agricultural cultivation has direct consequences on the level of 
adoption o f new technologies. Farmers are less inclined to invest in agricultural inputs, 
which raise the potential economic losses in case o f a crop failure. Mazzucato and Ly 
(1994) noted that certain of the new varieties outproduce local varieties only when rainfall 
is adequate. Also Lowenberg-deBoer et al (1992) showed that in years o f bad rainfall, the 
cultivation practices carrying the least risk were use of local varieties with traditional 
agronomic practices. Thus climatic fluctuations increase the risk of adopting an improved 
variety.
2.8 Methodologies of Impact Studies
Studies evaluating the returns to agricultural research can be classified into two broad 
categories: ex-post and ex-ante. The ex-post studies evaluate the returns on past 
investments while ex-ante studies try to estimate future returns. Ex-post studies include a 
range o f methodologies with varying objectives and evaluation criteria. Schuh and Tollini 
(1979) summarised ex-post studies as those which measure benefits of research in terms of 
resources saved, economic surplus, production functions, impact on national income and 
changes in the nutritional status of the population.
26
University of Ghana          http://ugspace.ug.edu.gh
The production function approach as used by Thirtle (1999), assumes that the difference 
between productivity growth in a given crop and the rest o f the agricultural sector is 
attributable to crop-specific research and extension expenditures. These expenditures are 
used to explain the difference between total factor productivity (TFP) growth in the crop 
and the rest of the agricultural sector. The TFP indices are calculated and their growth 
rates obtained. The elasticity o f TFP for the crop is then employed to compute the rate of 
return. That is, the rate of return calculation is normally based on the estimated 
coefficients o f research and development in explaining TFP (Lu et al, 1979; Davis, 1981; 
Thirtle and Bottomley, 1989; Alston et al, 1995) or on the research and development 
coefficient in the estimation of the dual profit function (Jayne et al, 1994).
Evenson et al (1987) have shown that there are equivalent measures o f technological 
change based on dual relationships among production, cost and profit functions and that 
these measures are equivalent to economic accounting measures, which are based on index 
number theory.
Traditional indicators o f research output include the productivity of the research system 
and discovery of new agricultural techniques such as the number of new varieties released. 
In some cases, the percentage adoption is used to measure impact. An example is the case 
o f Xcinthomonas Campestris resistant cassava varieties in Congo where a reported 90 
percent adoption by farmers was adjudged good impact. These are not always good 
indicators o f impact. However they are important measures of progress in meeting the 
conditions necessary for impact.
27
University of Ghana          http://ugspace.ug.edu.gh
In a pioneering study on measuring the returns to investment in agricultural research, 
Schuh (1953) used the resources saved approach. He calculated how much it would cost 
to obtain the agricultural output of 1950 with the agricultural technology of 1910. He 
then compared that value with the actual cost incurred in 1950 to produce that output. 
Adesina and Zinnah (1992) estimated that rice yields increased with the introduction of 
improved rice varieties developed by the West Africa Rice Development Association 
(WARDA). They noted that these varieties resulted in aggregate increases in 1990 farm 
household incomes of US$0.4million and US$14million in Guinea and Sierra Leone, 
respectively.
Also, Nyaribo-Roberts and Ospina (1992) estimated that the development o f a dual- 
purpose goat (milk and meat) increased the incomes of adopting Kenyan farmers by as 
much as 60 percent. In Ethiopia, the complementarities between livestock and crops were 
used to develop new ploughing and crop and water management techniques. These new 
techniques increased gross returns to farming and farm labour by over 300 percent each.
Griliches (1958) used the economic surplus model to measure returns to agricultural 
research. He worked on corn in the United States of America. Then, Akino and Hayami 
(1975) refined the economic surplus approach by separating the benefits accruing to 
producers and consumers from rice breeding research in Japan.
In all these cases, the distributional effects are not addressed; neither is there a standard 
against which we could compare the measure. The rate of return measure eliminates this
28
University of Ghana          http://ugspace.ug.edu.gh
problem of lack of a yardstick. This rate is often compared with the opportunity cost of 
capital. This may be the rate at which money could be borrowed from a financial 
institution or the rate it could earn in the next best alternative investment.
Some recent examples of the application of rate of return measurement include the 
following. Lopez-Pereira et al (1991) used the Akino-Hayami approach o f disaggregated 
benefits to evaluate the impact of sorghum research in Honduras. In the said study, they 
used economic prices to compute the rate of return to sorghum research. However, Sterns 
and Bernsten (1994) used nominal estimates of costs and benefits to compute a simple rate 
o f return to sorghum and cowpea research in Cameroon. They also used the economic 
surplus model but without disaggregating the benefits. This is because in Cameroon (like 
in Ghana) the crop is in the hands o f subsistence farmers so producers are the same as the 
consumers. This is the approach adopted by the present study.
However, the rate of return measure (like other quantitative indicators) fails to capture 
certain benefits of investments in agricultural research, mainly because o f difficulties in 
quantifying them. These include improvements in the environment and sustainability of 
agricultural production; improvements in the status of women in the household; 
improvements in the human and institutional capacity for research; and improvements in 
equity (income distribution).
29
University of Ghana          http://ugspace.ug.edu.gh
2.9 Findings of Rate of Return Studies
Several rate o f return assessments find positive rates of return of an economically 
important magnitude. For instance in Ghana, returns on maize research over the period 
1962-1992 was estimated at 74 percent (World Bank, 1999). The same source estimated 
a rate of return of 34-37 percent for cotton in Senegal, whilst Abidogun estimated 42 
percent for cocoa in Nigeria. With an already established extension system in Mali, 
Boughton and de Frahan (1992) estimated a rate return of 135 percent for maize research. 
Again in Senegal, Schwartz et al (1992) estimated a rate of return of 31 percent for an 
investment in cowpea research dubbed “Operation Cowpea”. This figure rose to 92 
percent when premium was placed on cowpea harvested during the hungry season.
The above rate o f return calculations seem to support a hypothesis that rate of return 
studies focus primarily on success stories, and thus bias the available evidence in favour of 
technology development and transfer. This may not be so as for instance negative rates of 
return were registered for dates in Niger and Uganda. In Niger, Mazzucato and Ly (1994) 
estimated a positive rate only by extending the analysis to 2010 at an assumed constant 
adoption rate. Also in Cameroon, a rate o f 3 percent was estimated for cowpea while 
another as low as 1 percent was estimated for sorghum research and extension in Niger. 
However, Uganda’s negative rate of return was said to be a direct consequence o f political 
problems of the 1970s and early 1980s. The present study seeks to examine the case of 
sorghum research at SARI.
30
University of Ghana          http://ugspace.ug.edu.gh
It is worth noting that these rates of return studies were undertaken from the perspective 
o f national agricultural research systems, hence only costs associated with the national 
agricultural research organisations are included. However, most o f the activities being 
evaluated may have benefited from discussions with IARCs and other regional networks, 
access to international germplasm, and/or direct importation of improved varieties. 
Therefore, the rates are best interpreted as indicators o f return to investments in national 
agricultural research programmes if the IARCs continue to function at their current level 
o f effectiveness.
2.10: Conclusion
The literature on adoption studies indicates that there are two types of adoption studies, 
one that tries to explain adoption through farmer decisions and the other that uses 
statistical comparison of the farmers’ adoption behaviour with characteristics of the 
farmers’ environment. In either case however, researcher perception, the purpose and the 
scope o f the study influence the choice of analytical techniques. To explore the 
determinants of adoption, Probit and Tobit models have been used extensively and in 
impact studies the preferred technique is the economic surplus model.
It has been shown that most adoption studies regard the technology as appropriate, and 
farmers are viewed as static units to which technology is to be transferred. However, this 
approach is no longer justifiable as it has been noted that the success o f a new technology 
depends on the fulfillment of specific socio-economic, technical and institutional 
conditions. This study explores factors in all these categories in order to identify specific
31
University of Ghana          http://ugspace.ug.edu.gh
determinants of adoption and the intensity of use of improved Sorghum varieties. This 
will contribute to efforts at guiding policy to promote widespread adoption.
The literature also indicates that the impact of new technology depends on various factors, 
mainly external to the technology itself. The literature reports rates of return ranging from 
negative values for dates in Niger to an estimated 135 percent for maize in Mali. The 
estimated rate o f return for improved Sorghum varieties is compared with some of those 
reported.
So far there is no literature on the adoption and impact o f improved technology in the 
Upper West region. This study will contribute to filling this information gap by identifying 
the factors that influence adoption and the intensity o f adoption, and also estimating a rate 
of return to Sorghum research.
32
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 3 
METHODOLOGY
3.1 Introduction
This chapter presents a formal definition of adoption, the theoretical frameworks and 
analytical tools employed to address the study objectives. Also discussed in this chapter 
are the data requirements, data collection techniques and how estimates of certain basic 
parameters were derived for use in the estimation procedure.
3.2 The Concept of Adoption
According to Feder et al. (1985), adoption refers to the use o f a new technology or an 
innovation in long-run equilibrium when a farmer (user) has full information about the new 
technology and its potential. However, the issue may not be as simple, since most new 
technologies are often recommended in packages. For instance, the technology under 
study has three components: the (use) planting of the improved sorghum variety, the use 
o f chemical fertilizer and row planting. In such a situation adoption becomes difficult to 
define. Does adoption mean the use of all components? What combinations or levels of 
use qualify as adoption? Is any planting of improved sorghum adoption? These are a few 
o f some o f the pertinent questions at stake. For the purposes of this study however, 
adoption simply refers to the realisation o f the decision of a farmer to plant any of the 
improved sorghum varieties.
The rate of adoption is defined as the percentage of sampled farmers who planted any of 
the improved sorghum varieties in the survey year. Also, the intensity of adoption is
University of Ghana          http://ugspace.ug.edu.gh
defined as the level o f use of a given technology. In this case, the proportion of a farmer’s 
sorghum farm planted with improved varieties.
Generally, researchers and extension workers want to know the factors that affect both the 
rate and intensity o f adoption of new technologies. To the researcher, this is important in 
designing technologies that will achieve high levels of adoption. To the extension worker 
such knowledge helps in designing extension messages that respond to the identified 
factors, in order to achieve the desired results.
3.3 Conceptual Framework and Analytical Tools for the Pattern of Adoption
The first objective o f this research is to describe the pattern or spread o f adoption. The 
purpose here is to describe how the adoption of improved sorghum has changed over time 
in the Upper West Region. Diffusion is the process by which an innovation is spread 
through certain channels over time among the members of a social system. Tarde (1969) 
describes it as a process characterised by a slow advance, followed by a rapid and 
uniformly accelerated progress, followed again by progress that continues to slacken until 
it finally stops. In fact, the process seems to describe the behaviour o f agents observing a 
divine commandment that says:
“Be not the first by whom the new idea is tried, Nor the last to put the old aside.”1
1 Alexander Pope, An essay on criticism, Part II quoted by Rogers, E.M. (1983). Diffusion of Innovations. 
Macmillan, Landon; New York.
34
University of Ghana          http://ugspace.ug.edu.gh
This is based on the concept of learning and it is due to the cumulatively increasing 
influences upon an individual to adopt or reject an innovation, resulting from the 
activation of peer networks about the innovation in the system. This phenomenon is 
explained in terms the innovativeness of the individual and if the cumulative number of 
adopters is plotted, the result is an S-shaped diffusion curve (Ryan and Gross, 1943). 
The cumulative normal and logistic distributions are two of several simple S-shaped 
curves, most widely used in handling growth phenomena, diffusion and technology 
transfer (Griliches, 1957). However, the logistic curve was chosen because it is simpler to 
fit and easier to interpret.
3.3.1: The Logistic Curve
This curve documents adoption history by looking at time of initial adoptions of the new 
technology and the cumulative effect over time. Logistic curves help to determine 
whether extension needs to be strengthened and by quantifying the change in the number 
o f technology users over time, they can be used to assess impact (CIMMYT, 1993). The 
logistic curve determines pattern of adoption or diffusion by assuming that the cumulative 
proportion o f adopters follows an S-shaped (sigmoid) curve. This curve depicts a slow 
initial growth in the use of the new technology, followed by a more rapid increase and 
then a slowing down as the cumulative proportion of adoption reaches its maximum 
(which may be well below 100% of the farmers) (CIMMYT, 1993).
The most common function used to portray the curve is the logistic function. For 
technology adoption the Y-axis represents the proportion of farmers adopting or area
35
University of Ghana          http://ugspace.ug.edu.gh
applied to a technology and the X-axis represents time. The logistic curve follows a 
general form:
D Y,/dt=b Yt/k(k-Yt) (3.1)
where:
Yt = number o f individuals who have adopted the innovation at time t (cumulative 
percentage of adopters or area under improved technology at time t). 
k = the fixed population of potential adopters (upper bound on percentage adoption)
b = parameter reflecting the rate of adoption (probability of adopting upon learning
about the innovation).
Solving for Yt from the logistic differential equation generates the result:
Yt = k/(l+e'a'bt) (3.2)
where ‘a’ is the constant of integration (time when adoption begins) and ‘b’ is a constant 
related to the rate o f adoption. The equation is however transformed to be estimated by 
ordinary least square regression as:
In (Yt /k- Y,) = a + bt (3.3)
and k is estimated as the level that appears to be the upper bound o f adoption on the 
cumulative adoption curve.
The logistic curve just described is fitted to data showing the percentage of farmers using 
improved sorghum in the study area beginning from 1988. The data required for this 
analysis was collected through a survey described in section 3.7 below.
36
University of Ghana          http://ugspace.ug.edu.gh
3.4 Theoretical Framework for Identifying Determinants of Adoption
This framework applies to the achievement of second and third objectives o f this study. 
The constraints imposed on agricultural development by inelastic supply of natural 
resources can be offset by the adoption o f new technologies and technical change. In 
order to accelerate adoption of new technologies and reap maximum benefit, it is 
necessary to first identify factors that influence the adoption behaviour of small farmers. 
The adoption behavioural model is frequently used as a framework to examine variables 
associated with technology adoption. It is generally assumed that farmers' response to 
innovation adoption is individualistic. Some respond partially, others quickly, while others 
do not respond at all to new innovations.
This framework has its roots in the threshold theory of decision making in which a 
reaction occurs only after the strength of the stimuli increases beyond the farmers' reaction 
threshold (Hill and Kau, 1973). This presupposes that every individual, when faced with a 
choice has a reaction threshold determined by several influential factors. The decision to 
adopt an innovation is dichotomous between two mutually exclusive alternatives. The 
individual chooses either to adopt or not to adopt. This implies that there exists a 
'breaking point' or threshold in the dimension o f the explanatory variable(s) below which a 
stimulus elicits no observable response. Only when the strength of the stimulus reaches 
the threshold level does a reaction occur. Additional increases in the strength o f the 
stimulus results in no effect on the observed response.
Phenomena o f this sort are handled using a general model o f the form:
37
University of Ghana          http://ugspace.ug.edu.gh
Yj — BjXj +  U; 
such that
Y; = 1 if X; >X* and 0 if X; <X* (3.4)
Models such as the above, which express the dichotomous dependent variable Y as a 
linear function of the explanatory variable(s) X are called linear probability models (LPM). 
This is because E (Y/Xj), the conditional expectation of Yi given X;, can be interpreted as 
the conditional probability that the event will occur given X ; i. e., Prob (Yi = 1/X;) 
(Goldberger, 1964; Aldrich and Nelson, 1984; Judge et al., 1985).
The framework for this study focuses on the threshold decision theory as it applies to 
adoption diffusion paradigm.
3.4.1 Methodological Considerations
The fact that many methodological problems are encountered in estimating relationships in 
which the dependent variable is dichotomous is well documented in the literature. As 
already indicated the adoption of improved crop varieties by a farmer is a decision 
involving two mutually excluding alternatives, either to adopt or not to adopt. The 
decision depends on a vector of X explanatory variables and a vector of unknown 
parameters (B. A positive decision takes place only when the combined effect o f the 
explanatory variables (X'p) reaches a certain unobservable critical value X*
This means that a given farmer will adopt improved sorghum varieties only when the 
combined effect of the socio-economic and technical characteristics, and institutional
38
University of Ghana          http://ugspace.ug.edu.gh
factors is able to offset the inherent tendency of the farmer to resist change. In fact, a 
farmer adopts only when he assesses the consequences of adoption to be favourable when 
weighed against the economic, social, and technical feasibility. This critical value reflects 
the individual decision threshold and plays the role of the error term (Goldberger, 1964). 
By denoting the dependent variable as Yi that takes the value of 1 if a positive decision 
occurs and 0 otherwise, the dichotomous adoption decision model for the ith farmer is 
specified as follows:
Yi = 1 if Xj'P > Xj* and
= 0 if Xj 'P < Xj* (3.5)
i=l,2, ...Nobservations.
For this study Y; takes a value of 1 if farmer uses improved sorghum, and 0 if farmer does 
not use improved sorghum varieties. The main objective in the estimation o f such 
qualitative response models is not the reconstruction o f the decision variable Y;, but the 
estimation of the conditional probability that Yi assumes one o f the specified values (Jung 
1983). Thus the probability that a given farmer is an adopter o f improved sorghum 
Xi'P > X; *, (Y; =1) is given as:
Pi =Prob(Y =1 ) =  F(Xi 'p) (3.6)
Also, the probability that a given farmer is a non-adopter X; 'P< Xj*, (Y, =0) is given as: 
1-Pi =Prob(Yj =0) = l-F (X i *p), (3.7)
where F(Xj'P) represents a cumulative distribution function evaluated at the value of the 
argument. It is well documented in the literature that the specification of F(Xi'P) as a 
linear probability functional form and estimating the unknown parameters with the use of 
least square regression models have been found to be inefficient for the following reasons:
39
University of Ghana          http://ugspace.ug.edu.gh
(i) the disturbance term exhibits heteroscedasticity
(ii) the expected value of Y; may sometimes be greater than one or less than zero
(iii) classical tests of significance for the estimated coefficients are not applicable since the 
estimated standard errors are not consistent (Goldberger, 1964; Dagenais, 1969; Judge et 
al„ 1985).
Although Aitkin's generalized least squares can be used to correct for heteroscedasticity, 
further problems may be encountered in determining the weighting factor. Above all, the 
fundamental problem with the linear probability model is that it assumes that Prob(Yj = 1) 
= E(Y =1/X) increases linearly with X, that is, the marginal or incremental effect of X 
remains constant throughout and that makes it logically not attractive.
Several other techniques have been suggested including the logit (Berkon, 1944), gompit 
(Brennan, 1949), Zellner and Lee's (1965) mixed estimation techniques, and simultaneous 
equation systems (Hill and Kau, 1973). Also the method of discriminant analysis can 
answer some questions relating to models having dichotomous dependent variables.
However, farmers' behaviour in the adoption of improved sorghum varieties will be 
modeled using primary data and the Probit (normit) model, given the dichotomous nature 
o f adoption and the several desirable properties of the probit model. These properties 
include: (i) it constrains the expected value of Y to lie between zero and one; (ii) it 
resolves the problem of heteroscedasticity and (iii) it gives consistent estimates of the 
standard errors (Hill and Kau, 1973). Though this model assumes normal distribution of
40
University of Ghana          http://ugspace.ug.edu.gh
the sample, its use with large samples is justified even when the underlying distribution of 
the sample is not normal (Pindyck and Rubinfeld, 1991).
It is important to note however, that the parameter estimates of the probit model need to 
be interpreted with some care. In the first instance, the magnitudes o f these estimates 
depend on the unit o f measurement. They are also expressed as indices and are therefore 
not comparable with one another (Hailu, 1990; Shakya and Flinn, 1985). Furthermore, 
they do not alone, provide direct information about the effects o f changes in the 
explanatory variables on the probability of adoption. Such effects are predicted from the 
derivative o f the probability with respect to a particular explanatory variable, given by 
Maddala (1983) as
5(Pi)/5Xik = <9F(Xi’P)/0Xik = f(Xi’p)Pk (3.8)
where f(X;’p) is the value of the normal density function at X;’p and pk is the estimated 
coefficient o f the klh explanatory variable. The above derivative can be calculated at 
different levels o f the explanatory variables. However, the weighted aggregate elasticity, 
i.e the sum o f the individual elasticities weighted by their probabilities (the preferred 
choice) is used in the current study (see Hailu, 1990; Maddala, 1983). This is because the 
elasticities at the mean values of the explanatory variables tend to overestimate the actual 
response, owing to the nature of the normal density function.
As already noted, qualitative choice problems frequently concern dependent variables that 
are discrete, usually taking on only two values. There are occasions, however, in which 
the dependent variable has been constructed on the basis of an underlying continuous
41
University of Ghana          http://ugspace.ug.edu.gh
variable, for which there are a number o f observations about which we do not have 
information (Pindyck and Rubinfeld, 1991).
In the current study, farmers’ level of use of improved sorghum varieties, is a case in 
point. For the users we can obtain the proportion of an individual’s sorghum land which is 
under improved varieties. However, for those who do not use improved sorghum 
varieties, we have no measure of the proportion o f sorghum land that a given individual 
would have been willing to put under improved varieties at the time o f the survey. Thus 
the values o f the dependent variable are unobservable though the corresponding values of 
the independent variables are, hence the name censored dependent variable. The general 
model is o f the form
Yi = XiP +m
0 otherwise (3.9)
The Tobit is a combination of Probit and linear regression techniques (Maddala, 1983) 
designed to handle these special cases of qualitative choice problems.
It is therefore clear that the Tobit and Probit models share common framework, similar 
methodological concerns and estimation procedures and problems.
3.4.2 Empirical Multi-nomial Probit Model
This model enables us to estimate the probability that a farmer, having learnt about an 
innovation, will adopt the innovation given his characteristics. It thus provides a tool for 
guessing the chances that a farmer with given characteristics will adopt a technology such
42
University of Ghana          http://ugspace.ug.edu.gh
as improved sorghum, after having learned about it. In this way it helps to identify some 
o f the variables that influence farmer adoption behaviour.
The Probit model as systematized by Finney (1952) and extended by Tobin (1955) 
determines the decision variable Yi , by Y; =0 if X< X* and Y; =1 if X> X* It is clear 
that X* plays the role of a "breaking point" and thus explicitly incorporates the threshold 
concept in the model. In a multivariate case, such as the current study an aggregate 
variable A is assumed such that A is a linear combination of the explanatory variables with 
A* as threshold playing the role of the disturbance forces. The disturbance term is thus 
homoscedastic (Hill and Kau, 1973).
For the purpose of this study, the model to be estimated is given as:
where t is a random variable distributed as a standard normal deviate, that is, t is N(0,1). 
Thus, the probability of a positive decision (Yj =1) is the area under the standard normal 
curve between - °o and X/p. The larger the value of X;'p the more likely adoption is to 
take place (Pindyck and Rubinfeld, 1991).
E(Y; /  A;) =Prob(Y; =1/ A ) = F;(A;) = ^ ( P A  + P2X2+ .............+ P„Xn) (3.10)
where F is a cumulative normal distribution. This distribution is given by
(3.11)
43
University of Ghana          http://ugspace.ug.edu.gh
Since F is a normal cumulative function, whatever value A; takes will necessarily be 
transformed to the interval of zero and unity, and the difficulty of the expected value of Yi 
being greater than one or less than zero (i.e. E(Yj) >1 or E(Yi)<0) is overcome. 
Moreover, the derived functional form between A and Y is a normal sigmoid curve (Hill 
and Kau, 1973). Therefore, any linear or nonlinear relationship between X  and Y through 
A  can be accommodated. In addition, the magnitude of the effect of a relevant economic 
stimulus depends upon the status of the economic unit as well as upon the magnitude of 
the estimated coefficient Pi, i.e.,AE(Y;) = / (Aj, AAj) where AAi = PjA(Xjj). Thus the 
probability does not depend on only the explanatory variables, Xj.
The parameters in the probit model are estimated by maximum likelihood methods. The 
likelihood function L is specified as the product of the probabilities of both alternatives 
and its log is maximized with respect to unknown parameters.
L  =  n [ P ( y i= l ) f  [p(y; =  0 )]''yi
n
=  n [F (X i |3)]yi [l-F (X i P )]lyi (3 .12 )
i = 0
where n represents the number of observations in the sample. Section 3.5 below presents a 
discussion of the choice o f explanatory variables used in the models.
3.5.1. Theoretical Framework and Methodological Issues for Identifying 
Determinants of Intensity of Adoption
The Probit model provides information only with respect to the household's decision to 
adopt or not to adopt but not on the intensity of use after adoption. To measure the
44
University of Ghana          http://ugspace.ug.edu.gh
intensity of use, a model that incorporates both forms of the dependent variable 
(dichotomous and continuous) and provides information on all aspects o f the decision 
making process is required. That is, one that measures the probability of adoption and the 
intensity o f use thereafter on the basis of information on the entire sample. For such 
purposes an approach developed out of Probit and linear regression models by Tobin, 
known as the Tobit model, is found to be appropriate (Amemiya, 1984).
This model enables us to identify, in addition to estimating the probability of adopting, the 
factors which determine the extent to which a farmer having adopted improved sorghum, 
will use it. The Tobit model to be used in this study similarly estimates the probability of 
adoption and simultaneously the expected value of the continuous dependent variable Yj 
as a function of the set o f explanatory variables X; weighted by the probability that Y; >0. 
This expected value o f Y; equals zero if the estimated value of X; |3 falls below the critical 
threshold level X* (Kinsey, 1984). It is clear from the theoretical model discussed above 
that the functions for the dichotomous choice (adoption) and the intensity of use should 
have the same set of explanatory variables. Thus the Tobit model to be estimated is given 
as:
Prob(Yj =1/A;) = Fj(Aj) = Fi(p,X, + P2X2+ .............+ pnXn) (3.13)
The adoption variable in the Tobit model Yi is defined to assume a continuous character. 
An iterative maximum likelihood algorithm (White 1978) is used to estimate the models in 
order to obtain asymptotically efficient parameter estimates.
Over the entire sample the expected value of Y; is calculated as:
45
University of Ghana          http://ugspace.ug.edu.gh
E(Yi/ Xi'P) = (Xi'p) • F(Xi’p)/a + a  • f  (Xi'p)/a (3.14)
And for the sub-sample already above the limit (Yi>0) the expected value is given by 
E(Yi*) = E(Yi/Yi>0) = (Xi'p) + a  • f(X j"p/a> F(Xi'p/a) (3.15)
where F (X 'p/o) is the probability that Y; >0 calculated from the cumulative normal 
distribution, f(Xi'p/o) is the normal density function and a  is the standard error of estimate 
(Goldberger,1964; Judge et al., 1985). Equation (3.14) gives an estimate of the average 
area a sorghum farmer, selected at random from the sample is likely to have under 
improved sorghum. Also, (3.15) gives an estimate of the average area an adopter of 
improved sorghum, selected at random is likely to have under improved sorghum.
The parameters in the Tobit model, like in the probit are estimated by iterative maximum 
likelihood procedure (White, 1978). Since the maximum likelihood estimates of the 
coefficients in a large sample are approximately normally distributed, the significance of 
the individual coefficients can be tested in both models, (Probit and Tobit) by the ratio of 
the estimated coefficient and its standard error (asymptotic t-value). However, tests 
concerning the significance of all or a subset of the coefficients in the models is done by 
the log likelihood ratio test (LRT). This test is the Chi-square distributed with k degrees 
of freedom, where k is the number of parameters in the model less the constant.
3.6 Choice of Variables and Working Hypotheses for the Determinants of Adoption 
and Intensity of Adoption
In accordance with the study objectives and with direction from existing literature, 
explanatory variables have been chosen from among factors related to all the areas of
46
University of Ghana          http://ugspace.ug.edu.gh
concern. These include various socio-economic characteristics of the farmer and 
technology specific factors that influence the demand for, and the use o f improved 
sorghum varieties. They also include institutional factors that impact on the availability of, 
and market access to improved sorghum seed in the study area. The definitions and 
measures o f the explanatory variables used in the models are summarised in Table 3.1 and 
discussed below.
3.6.1 Socio-economic factors
The socio-economic factors included in the models are age of respondent, farm size, 
family labour force, non-farm income, education, the use o f complementary inputs, fallow 
period and distance to sorghum farm. As already discussed in the literature review (pp. 16- 
17) these determinants o f adoption have been proved in similar studies in the past to be 
influential in the adoption of agricultural innovations.
Whereas it is often presumed that younger farmers, being more adventurous and with a 
longer planning horizon, are more inclined to accept innovations than older ones, some 
researchers have observed otherwise. In this study youthfulness (younger age) is expected 
to have a positive effect on adoption for the same reasons expressed above.
The changes in the production process involved in the adoption of a new technology may 
bring risks resulting from imperfect information and the possibility of committing errors. 
As observed in the literature review (pp. 17), education enhances one’s ability to receive, 
decode and understand information and experience has effects similar to education as it
47
University of Ghana          http://ugspace.ug.edu.gh
Table 3.1: Definitions of Variables used in the Models
Age2
Educational Status 
Extension Contact 
Off-farm income
Farm Size 
Fallow Period 
Farm type
Farm Credit 
Fertilizer
Family Labour
Market access to 
improved seed
ADOPTION (Probit)
ADOPTION (Tobit)
Age o f household head measured by a dummy 1 if farmer between 
3 l-50years, 0 if farmer is above 50years.
Measured by a dummy l=literate 0=illiterate
Number of extension visits to farmer
Measured by a dummy 1= has off-farm income 
0 = has no off-farm income
Number of acres under sorghum
Number of years land is left to fallow
Measured by a dummy 1 if sorghum farm is around homestead (less than 
lkm), 0 otherwise.
Measured by a dummy 1 = Additional cash 0 = Own cash
Use o f complementary inputs by farmer measured by a dummy variable 1 
if  the farmer used chemical fertilizer in the survey year, 0 otherwise.
Available Family labour for farm work on household farm measured in 
persons.
Distance to nearest sales point as a proxy for market access to 
agricultural inputs measured as number of kilometres.
Measured by a dummy lif  farmer adopts and 0 otherwise.
Proportion of total sorghum area under improved sorghum
8 So defined for fear that farmers may not be able to tell their exact age
48
University of Ghana          http://ugspace.ug.edu.gh
builds confidence and dispels doubt. In view of the foregoing the experience and 
educational status of respondent, are expected to have a positive influence on adoption.
The available family labour and farm size, are two important factors often found to be key 
determinants of innovation adoption at the household level. As observed earlier on 
(pp. 16), in a traditional African setting, the strength of family ties often compels farm 
workers to tolerate very poor working conditions and wages as a result of their family 
loyalty. The farm operator thus assumes a financially advantageous position to spend 
more money on adoption of new technologies. This situation may be enforced in the study 
area where there is very limited labour market for off-farm employment and hence the 
opportunity cost of labour is very low. The two factors are therefore expected to have a 
positive effect on adoption.
Farm size, a good proxy for wealth, may have a positive influence on the adoption of new 
technologies, especially those involving the use of purchased inputs (e.g. improved seed, 
chemical fertilizers etc.). Similarly, people with off-farm income have the means to 
procure purchased inputs, which the adoption o f a new technology may require. These 
variables are therefore hypothesised to favour innovation adoption.
The use o f chemical fertilizers in the survey year has been entered as an explanatory 
variable to determine the effect of the use of complementary inputs by a farm-household 
on the adoption of improved sorghum varieties. This proxy, for the technological level of 
the farmer, is expected to have a positive effect on adoption.
49
University of Ghana          http://ugspace.ug.edu.gh
There are two types of sorghum farms in the study area, bush farms and compound farms. 
The compound farms cultivated around the homesteads are often reserved for short 
duration crops, which are harvested during the lean season, with the bush farms being 
planted to late maturing full-season varieties. Farmers were asked to indicate the distance 
to their Sorghum farms. It was noted that where farmers had more than one Sorghum 
plot, the tendency was to plant the one near the home to improved varieties. Distance to 
sorghum farm is used as a proxy for farm type, which is given a value o f 1 if the sorghum 
farm is around the homestead (less than 1km), 0 otherwise. It is therefore expected that 
the farms located near homestead and adoption o f improved sorghum will be associated 
positively, given the short maturity period of most of the available improved varieties.
As discussed earlier, one farming system in the study area is the bush fallow system in 
which a piece o f land is left to regain its fertility after having been cultivated for a number 
o f farming seasons. As a particular plot is being abandoned to fallow, the farmer works 
on another, and hence, how quickly the farmer returns to the plot in time gives an 
indication o f the level of pressure on land for cultivation. However, according to the 
induced innovation model of agricultural development, pressure on agricultural land 
provides an incentive for land-constrained farmers to adopt high yielding varieties. Longer 
fallow periods are therefore expected to have a negative effect on the improved Sorghum 
varieties.
50
University of Ghana          http://ugspace.ug.edu.gh
3.6.2 Technical Characteristics
Here the farmer’s perception of the characteristics of the improved varieties is to be 
explored. A perception index included in the models was computed from among several 
technology specific factors. These include: disease tolerance, drought tolerance, pest 
tolerance, maturity period, fertilizer requirement, varietal yield, storage or shelf life, taste 
quality, suitability for local dishes, weed suppression and marketability. Farmers were 
asked to compare (i.e. rank improved variety as 1 if better, 0 otherwise) the best known 
improved variety with their best traditional variety in terms o f these varietal 
characteristics. The index is specified as a dummy and takes on a value o f 1 if the 
cumulative score for a given characteristic is greater than 6, and 0 otherwise.
This index is expected to capture a farmer’s perception of the risks associated with the 
adoption o f improved sorghum as well as the compatibility of the improved varieties with 
the socio-cultural setting of the people in the study area. Hence the index is expected to 
have a positive relationship with adoption of improved sorghum.
3.6.3 Institutional factors
The institutional factors included in the models are, access to extension information, 
distance to the nearest purchase point for agricultural inputs as a proxy for market access, 
and farmer’s access to credit in the survey year. Although the role o f output markets is 
acknowledged no variable was included in this study to capture its effect.
51
University of Ghana          http://ugspace.ug.edu.gh
Extension organisations are expected to educate farmers on new techniques of farming 
including the available new varieties and their use. Contact with extension activities has 
therefore been included in the analysis in order to assess its role in farmer’s adoption 
decision. This is of particular interest as it may throw some light on how research and 
extension relate in the area. It has been suggested that SARI has for a long time not had a 
very strong link with the national agriculture extension system, neither has it had a strong 
extension component besides the on-farm research teams.
In fact, financial constraints are a common feature in smallholder agriculture. The 
availability of farm credit to enable small farmers to procure the necessary inputs on a 
timely basis is therefore vital. However, as observed earlier (pp. 18) access to institutional 
credit for the majority of small farmers in northern Ghana is hardly possible and increased 
use o f modern agricultural inputs in northern Ghana is constrained by deficiencies in 
supply and distribution systems. This may be particularly serious in the study area, as the 
physical infrastructure is not yet well developed. In addition, the distance between a farm 
household and the nearest purchase point affects input purchasing costs (ceteris paribus) 
and hence influences the adoption decision of the farm household. For this reason the 
distance between a farmer's village and the nearest purchase point for agricultural inputs 
(e.g. improved seed, chemical fertilizers, etc.), unlike access to credit, is hypothesised to 
have a negative effect on adoption.
52
University of Ghana          http://ugspace.ug.edu.gh
3.7 Data Requirements and Data Collection for the Probit and Tobit Analyses 
From the preceding section, it is clear that such analyses require primary data on the 
various determinants of adoption. The variables on which data was collected span all 
three categories discussed above.
The data was collected through a structured sample survey conducted in all the five 
districts o f the Upper West Region. A purposive sample of two villages from each district 
was first selected based on importance in terms of sorghum production. A random sub­
sample o f 20 farmers was then selected from the selected villages in each of these districts, 
which benefited from the dissemination of the improved varieties, giving a total of 100 
respondents. Trained enumerators interviewed the respondents using structured 
questionnaire which, covered the following areas:
- Socio-economic characteristics of farmers: age of the farmer, education, farm size, 
family labour force, the use of complementary inputs, fallow period and off-farm income. 
-Technical characteristics of varieties: disease tolerance, pests tolerance, drought 
tolerance, maturity period, fertilizer requirement, varietal yield, storage, taste quality, 
weed suppression, marketability and suitability for local dishes.
-Institutional factors: extension contact, distance to sorghum farm, distance to nearest 
purchase point for improved seed and access to farm credit. The questionnaire is found in 
Appendix 2.
53
University of Ghana          http://ugspace.ug.edu.gh
3.8 Method of Analysis for Farmers Varietal Preferences
It is important to answer the question of what influences farmers’ choice of improved seed 
so that sorghum researchers can determine what type of sorghum will be appropriate to 
the biophysical and socioeconomic circumstances of farmers in the region. Farmers were 
asked to rank the important characteristics that they consider in their choice o f sorghum 
variety to plant. This information, was collected during the survey described in section 3.6 
above, and is analysed using frequency distributions and cross-tabulations to identify the 
traits desired most by sorghum farmers and capture their varietal preferences.
3.9 Theoretical Framework and Methodological Issues for Estimating the Rate of 
Return on Sorghum Research
The main thrust o f an impact assessment analysis is to compare a situation without 
research against an alternative situation with research. In fact, this analysis should not be 
confused with a before and after comparison since this can be misleading as conditions are 
constantly changing. The present study employs the economic surplus method of impact 
assessment.
The concept of Economic Surplus provides the theoretical framework for the economic 
surplus method. In order to turn agronomic data (the value of research) into economic 
values, the surplus approach uses the concepts of 'supply', 'demand', and 'equilibrium'. 
Rooted in the Marshallian concepts of social welfare and cost, it measures the social 
returns to research in terms of changes in consumers' and producers' surpluses resulting 
from the shift in the supply curve corresponding to a shift in the production function. This
54
University of Ghana          http://ugspace.ug.edu.gh
'economic surplus' is some measure o f economic welfare. In this application 'supply' 
represents producers' production costs, while 'demand' represents consumers' consumption 
values. The interaction of these two forces results in some 'temporary equilibrium' 
quantity and price. The economic surplus measured as the sum of producer surplus and 
consumer surplus is as shown in figure 3.1 below.
Figure 3.1: Producer and Consumer Surplus
Px
where D is an inverse demand curve D = D(Q) and S is an inverse supply curve, S = S(Q); 
Cs = Consumers’ Surplus; Ps = Producers’ Surplus.
The demand and supply curves can be written as:
Pd = fd(Qd), (3.16)
Ps = £ (Qs), (3.17)
where Pd is the demand price of the product, Qd is the quantity demanded, Ps is the supply 
price o f the product, and Qs is the quantity supplied.
55
University of Ghana          http://ugspace.ug.edu.gh
It is assumed that the demand and supply functions obey their respective laws, ceteris 
paribus. Although demand and supply curves can take various shapes, it is often 
appropriate to consider them to be linear and assume parallel shifts (Masters et al., 1996). 
This simplifies the analysis without changing the end result, which is the area below the 
demand curve, between the two supply curves. Thus we have 
Ps = a + bQs, b > 0 for supply and
Pa = c + dQd, d< 0 for demand (3.18)
Finally, the levels of quantities produced must be some equilibrium (temporary though) 
between supply and demand. At each point in time, for some particular region, there is a 
single quantity that is both supplied and demanded (Qe) as well as a single price that is 
both paid by demanders and received by suppliers (Pe).
As already indicated, the effect o f research is captured as a change in economic surplus as 
innovations from research shift the supply curve to the right. This enables producers to 
supply a larger quantity at the same price or the same quantity at a lower price. Note that 
increases in the cost of production due to new technology create a further shift in the 
supply curve upward from S2 to S1. The scenario is depicted in figure 3.2 below.
56
University of Ghana          http://ugspace.ug.edu.gh
Fig 3.2: Supply shift with improved varieties
Q o  Q i  Q x
Source: Adapted from Teal and Dinwiddy, 1996,p9.
With the improved sorghum varieties, the supply curve was expected to have shifted to the 
right and price declined. Society benefits with B C i E \ F. It should be noted, 
however, that this framework assumes a static demand function. This is a limitation since 
the assumption is not realistic, at least in the face of increasing population.
3.10.1 Method of Estimation
In order to be able to make an informed opinion of the profitability o f the investment, it is 
often necessary to compare discounted costs and benefits over the period under 
consideration. This is because most research benefits do not start until after much cost has
57
University of Ghana          http://ugspace.ug.edu.gh
been incurred. Besides, during the early years following the development of a new 
technology, adoption levels may be low, and research benefits are therefore quite low. As 
adoption proceeds, the economic surplus grows exponentially, as the area between the 
supply and demand curves increases. Thus it is only after adoption is more widespread 
that significant benefits may be obtained.
Masters et al. (1996) recommend the use of the internal rate of return IRR, as the most 
important indicator for assessing the benefits of research over time. The IRR is the rate 
that results in
T
Z(B t- C()/(J+r)' = 0 (3.20)
1-0
where B, is the social benefit in year t 
C(is the research cost in year t 
/•, is the discount rate 
/is  time, t = 1,2,...T
T  is the year in which research ceases to produce returns.
This measure gives the earning power of the investment (Gittinger, 1982). The IRR can 
then be compared to any interest rate, such as the cost of borrowing funds from a bank, or 
the returns earned in other investments. If the IRR from the research investment exceeds 
these rates, then the investment was worthwhile.
The financial rate of return calculated for Sorghum research covered the period 1966- 
1998. This period was chosen for two main reasons. The first was difficulties with
58
University of Ghana          http://ugspace.ug.edu.gh
obtaining the relevant data. It should be noted that it was generally difficult to obtain the 
data necessary for a good impact assessment. Data on quantity of improved seed 
distributed to farmers could only be obtained from the Global 2000 co-ordinator, with 
very little from the Seed Growers Association and none from the national agricultural 
extension service. Secondly, the Sorghum programme released its first improved variety in 
1971 and this according to sorghum breeders was worked on for at least five years prior 
to the release.
Also, due to resource constraints and estimation problems, only the cost of sorghum 
improvement research was considered, but not the cost of technology transfer. It was 
assumed that 60 percent of the costs of the Sorghum section of SARI (which also caters 
for millet) is attributable to sorghum, based on discussions with the programme co­
ordinator. The study follows the approach adopted by ICRISAT and the International 
Potato Center (CIP) in separate cases to analyse the returns to agricultural research. Both 
cases included only their own research costs, excluding costs associated with technology 
transfer. The costs also do not include past, sunk costs (in this case costs incurred before 
1966). Thus the resulting rate should be viewed as the rate of return to sorghum 
improvement research. According to Kelly (1995) such a rate is best interpreted as an 
estimate of the economic benefit or the value of research at the margin.
3.11 Data Requirements for Impact Assessment (Rate of Return Computation)
The basic components besides observed prices and quantities, required for calculating the 
impacts of technological change on society and the distribution of the gains between
59
University of Ghana          http://ugspace.ug.edu.gh
producers and consumers are the elasticities of demand (ed) and supply (e,) and the shift 
parameter (k). Mathematically, the computations of the parameters needed for this 
exercise, following Mazzucato and Ly (1994), are as follows;
The change in production due to the improved varieties, as a proportion of total 
production is given as:
j  = (AYmt)/Y  (3.21)
where AY is the yield (kg/ha) difference between the improved varieties and the traditional 
ones; Y is average yield i.e., total production divided by total acreage (ha); and 'a,' is the 
proportion of total sorghum area under improved sorghum.
The adoption costs of the improved varieties, as a proportion of the product price, is given
c— (ACa,)/(Y »P) (3.22)
where AC is the difference per hectare between input cost on improved varieties and the 
traditional (0/ha) and P  is the annual average product price paid to producers in real terms 
(0/kg).
The net change in production costs, as a proportion of the product price, k  is given as: 
k= (j/ej- c (3.23)
where es is the elasticity of supply.
In the face of lack of per-unit cost reduction estimates this study uses the j  parameter to 
approximate k, and since technologies take time before benefits accrue the relevant per- 
unit cost reduction is given by
k, = k » a t or j  • a, (3.24)
60
University of Ghana          http://ugspace.ug.edu.gh
In the absence the required elasticity estimates this is a realistic assumption. Mazzucato 
and Ly (1994) have shown that j  is exactly equal to k  when the long-run elasticity of 
supply is equal to one.
The social gains or economic benefits from the adoption of improved sorghum varieties 
measured as the total change in economic surplus (ATS,) and captured as B + C + E  + F  
in figure 3.2 is given as:
Bt= Pi Qtk, (3.25)
Finally, the net benefit is obtained by subtracting the costs of research and this is given as: 
B„ei = B,-C, (3.26)
where C, is the total cost of research that produced the improved varieties.
Estimating the adoption of improved technologies is an essential component of an 
evaluation of the benefits of research. In the face of an acute lack of a comprehensive set 
of adoption statistics on sorghum varieties being evaluated, synthetic estimates were 
derived of adoption rates using a variety of information sources.
Seed distribution figures were obtained from MOFA seed stores, the regional Sasakawa 
Global 2000 offices and Ghana Seed Company. The statistics however, grossly 
underestimate use of improved seed. This is in view of the fact that farmers themselves 
select and retain seed of improved varieties for use in subsequent years. An estimate of 
the percentage of improved seed in use derived from farmer-retained seed was made and 
added to the quantity of seed distributed to get a more realistic estimate of total amount of
61
University of Ghana          http://ugspace.ug.edu.gh
improved seed being used in a given year. Discussions with extension personnel indicate 
that in normal years farmers could retain over 100 per cent of the distributed seed sold, for 
the following year. This implies that the retention rate is over 100 per cent since other 
farmers also obtain improved seed as gift. For example, in Niger, Mazzucato and Ly 
(1994) estimated a retention rate of 120 per cent for millet. However, a retention rate of 
100 per cent is assumed in this study to account for genetic degradation that takes place 
over time in an open-pollinated variety like sorghum. Estimates of annual area cultivated 
with sorghum over the relevant period were obtained from the PPMED of MOFA. A 
recommended 5kg/ha seed application rate for sorghum was obtained from the 
researchers.
Once the seed utilisation figure is obtained, it is then divided by the total area cultivated 
with sorghum and by the seed application rate to obtain the percentage adoption for 
improved sorghum. The relevant formula is
IS, = (ISD,-i»Ft)  + ISD, (3.27)
where
IS, = total improved seed used (kg) in year t
ISD,.i = improved seed distributed (kg) in preceding year
ISD, = improved seed distributed (kg) in current year
Ft = proportion of improved seed carried over from a given year to the next (own
seed)
62
University of Ghana          http://ugspace.ug.edu.gh
The proportion of cultivated area sown to improved sorghum is calculated as
a, = (IS, •R ') •At 1 (3.28)
where R  = seed application rate (kg/ha); AT= total area cultivated with sorghum (ha).
It is also realised that estimates of the social benefits arising from investments in research 
depend on the nature and size of the research-induced shift in the supply curve. Yield 
relativities are used to estimate the supply shift as defined earlier. Ideally, on-farm yield 
differentials between traditional varieties and new, improved varieties should be used for 
each location (district) where the improved varieties were grown to include site-specific 
differences in the yield differentials of improved varieties. However, all trials were only 
located at Wa, hence the available average yield figure of 1,35ton/Ha was used.
Output prices used in the calculation of the supply shift are the average annual market 
prices, converted to real values using price index. The price data was obtained from 
PPMED of MOFA. The cost of research was estimated as the sum of the annual salaries 
and allowances of all the staff on the sorghum improvement programme at SARI and the 
annual running cost of the sorghum section.
63
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 4 
RESULTS AND DISCUSSION
4.0: Introduction
This chapter opens with a brief discussion of the characteristics of the sampled farm- 
households based on the preliminary survey results. The study objectives are then 
addressed in turn, using the logistic function and Probit and Tobit analysis to examine 
adoption. Finally, farmers’ varietal preferences are assessed and the rate of return used to 
measure the impact of research.
4.1: Characteristics of Sampled Farm-Households
The sample of 100 respondents, turned out to be predominantly illiterate males with only 
7% being females and 17% of respondents having had either formal or non-formal 
education. The age distribution of the sample was as follows: below 31 years (5%); 31 -40 
years (38%); 41-50 years (28%) while 6% were above 60 years. The remaining 23% were 
aged between 51 and 60 years. The major reasons advanced for growing sorghum were
(i) mainly for food 40%; (ii) for food and cash 54% and (iii) mainly for cash 5%. 
Approximately 60% of respondents intercropped their sorghum, mainly with cowpea 45%, 
maize 7% and maize/cowpea in combination 4%. The major source of labour for farm 
operations as reported by the respondents was family or reciprocal labour. It was only in 
land clearing and weeding that some 11% indicated having used hired labour.
It is worth noting that 86% of the respondents did not use inorganic fertilizer, the major 
reason being that it is expensive and they could not afford it. Credit was reported to be
University of Ghana          http://ugspace.ug.edu.gh
particularly difficult to obtain especially at the beginning of the farming season (60%) and 
during weeding (21%). Only 8% of all sampled farm-households indicated they had used 
farm credit from banks, moneylenders and trader pre-financing. Thirty-eight percent 
(38%) of the respondents had sources of non-farm income, the main sources were (i) 
trading 16%; (ii) charcoal burning 6%; (iii) wage employment 5%. Some of the main 
household characteristics of the respondents are compared in Table 4.1 below.
Table 4.1: Main Characteristics of Sampled Households, UWR, Ghana, 1988 - 1999.
Variable/Units
Adopters
Mean
Non-adopters
p-value
Family labour (persons) 9.00 7.00 0.0028***
Farm size (hectares) 2.20 1.76 0.0852*
Market access (km) 10.50 17.90 0.0000***
Fallow period (yrs) 3.00 3.79 0.0527**
Extension visits (no./yr) 4.88 10.62 0.0261**
Source: Computed from survey data.
* Significance levels *** = 1%; ** = 5%; * = 10%
The statistics in Table 4.1 clearly show that the adopters’ sub-sample tend to be in better 
positions than their no-adopting counterparts. They have on the average more family 
labour, larger farms and shorter fallow periods. They were also closer to sources of farm 
inputs (purchase points). However, they had fewer visits from extension agents.
The most important traditional varieties grown are Chare/Nyoso 55%, Dawele 18% and 
LatuorlGon 11%. The major reasons why some farm-households never planted improved 
sorghum varieties as reported in the survey were that they (i) were not aware (15%), (ii) 
had no seed (13%) and (iii) preferred traditional varieties (11%).
65
University of Ghana          http://ugspace.ug.edu.gh
4.2.1: The Pattern of the Adoption of Improved Sorghum Varieties in the Upper 
West Region of Ghana.
The results of the estimated logistic function in Table 4.2 are used to fit the diffusion 
curve.
Table 4.2: Estimates of Logistic Function, UWR, Ghana, 1988 - 1999.
Variable Coefficient Standard error t-value p-value
Intercept -1.6816 0.1708 -9.8464 0.0000***
Time 0.29995 0.0232 12.9263 0.0000***
Degrees of freedom 10
Source: Computed from survey data.
* Significance levels *** = 1%
The fitted diffusion curve below describes the adoption pattern for improved sorghum 
varieties over the period 1988-1999. The curve, which approximates the expected S- 
shaped curve typical of successful diffusion, shows a gradually increasing trend in the rate 
of adoption of improved sorghum varieties. The rather painfully slow initial stages of the 
process have not featured prominently because some of the improved varieties were in use 
well before 1988. The maximum value on the curve as it tapers off is about 43%, but this 
does not give the actual proportion who planted improved varieties in 1999 since the 
curve assumes continual use after adoption. The curve in figure 4.1 depicts a normal 
diffusion process with an estimated diffusion rate of about 0.3 percent.
66
University of Ghana          http://ugspace.ug.edu.gh
Cu
m
ul
at
iv
e 
Pr
op
or
tio
n 
of 
A
do
pt
io
n
Fig. 4.1: Diffusion Curve for Improved Sorghum Varieties, UWR, 1988 - 1999
45
40
35 -
30
25
20  -
15
10 -
5
I------------------------------- T " “ i---------------- ~i------------------r~
1988  1989  1990  1991 1992  1993 1994  1995  1996  1997  1998  " 199<
Time (years)
Source: Data compiled from Field Survey
67
University of Ghana          http://ugspace.ug.edu.gh
4.2.2: Empirical Results of the Determinants of Adoption of Improved Sorghum 
Varieties
In this section the determinants are identified for the decision to adopt or not, using the 
Probit model and the intensity o f adoption using the Tobit model. The results o f the 
specified probit model are presented (Table 4.3).
Table 4.3: Estimates of the Determinants of the Adoption of Improved Sorghum 
Varieties: Probit Model (1999/2000, UWR, Ghana)
Variable Estimated
coefficient
Standard
error
Asymptotic
T-ratio
Elasticity 
of predicted 
probability
Age 1 . 4 5 6 1 0 0 . 5 4 4 2 7 2 . 6 7 5 3 0 * * * 0 . 4 8 7 6 4
Education 0 . 2 2 1 5 7 0 . 4 1 2 3 1 0 . 5 3 7 3 8 0 . 0 2 8 9 5
Extension visits - 0 . 0 2 4 8 7 0 . 0 1 4 6 5 - 1 . 6 9 7 4 0 * - 0 . 0 8 4 7 3
Credit 0 . 6 2 0 2 2 0 . 6 5 8 1 4 0 . 9 4 2 3 9 0 .  0 2 5 3 3
Non-farm income 0 . 6 7 1 0 8 0 . 4 0 4 5 1 1 . 6 5 9 0 0 * 0 . 1 0 7 2 5
Fallow - 0 . 3 1 3 9 1 0 . 1 2 3 4 0 - 2 . 5 4 3 9 0 * * - 0 . 4 7 7 8 1
ChemicalFertilizer - 0 . 4 5 2 8 2 0 . 4 1 0 7 6 - 1 .  1 0 2 4 0 - 0 . 1 0 1 6 0
Family Labour 
Distance to
0 . 2 6 0 4 3 0 . 0 7 2 1 9 3 . 6 0 7 8 0 * * * 0 . 8 4 7 1 2
nearest salespoint - 0 . 0 9 0 1 5 0 . 0 2 6 7 1 - 3 . 3 7 5 1 0 * * * - 0 . 5 8 0 1 1
Perception Index 1 . 2 4 2 9 0 0 . 5 5 2 9 0 2 . 2 4 7 9 0 * * 0 . 1 5 4 6 4
Farm Type 0 . 6 9 2 2 3 0 . 4 0 3 1 5 1 . 7 1 7 0 0 * 0 . 1 6 1 7 7
Farm Size 0 . 3 8 7 5 2 0 . 1 5 9 6 4 2 . 4 2 7 5 0 * * 0 . 3 2 7 2 7
Constant - 2 . 2 0 7 8 0 0 . 8 6 7 5 3 - 2 . 5 4 5 0 0 - 0 . 9 4 0 3 6
Likelihood-ratio statistic 65.1467***
Sample size 100
Estimated rate o f adoption 0.47
Observed rate of adoption 0.48
Mean predicted probability o f adoption 
Among adopters 0.8813
Among non-adopters 0.1364
Significance levels: *** = 1%; ** = 5%; * = 10%
On the basis of the likelihood ratio test (LRT), the probit model specified to examine the 
factors affecting the adoption of improved sorghum varieties is significant at the 1% 
significance level. This implies that the explanatory variables included in the model jointly
68
University of Ghana          http://ugspace.ug.edu.gh
influence a farmer’s decision with respect to the adoption of improved sorghum varieties. 
In addition, the model predicted 83% of the cases (adopters and non-adopters) rightly and 
could therefore be used to test the validity of the hypotheses laid down earlier.
According to the asymptotic t values of the coefficients, some o f the explanatory variables 
in the model were found not to be in conformity with a priori expectations. In the first 
instance, the extension contact variable did not show the expected sign though it was 
statistically significant at the 10% significance level. This observation contrasts sharply 
with general trends in adoption literature (see Hounkpe, 1999; Njoku, 1993; Shakya and 
Flinn, 1985). However, it is not the first of its kind as Denning (1991, cited in CIMMYT, 
1993) observed no significant statistical association between extension visits and the 
adoption o f dry-seeded rice in the Philippines. In the case of improved sorghum varieties 
in Northern Ghana, the main vehicle for the dissemination o f information between SARI 
and sorghum fanners was the contact-farmers with whom on-farm trials were conducted 
until the establishment o f a Research-Extension Liaison Committee in 1993/94. This unit 
presents research findings to the staff of the national agricultural extension network. The 
observed negative relationship raises questions about the appropriateness o f the whole 
extension method. Specifically, no attention is given to sorghum extension activities in the 
study area.
Educational status and the use of farm credit, though with the expected signs, were not 
statistically significant at the 10% significance level. This is similar to the observation by 
Hailu (1990) that, though positively related, education did not prove to be significantly
69
University of Ghana          http://ugspace.ug.edu.gh
associated with the adoption of bullock traction technology in the Northern region of 
Ghana. However, he argued that this resulted probably because the educated saw tractor 
use as a better alternative. It is interesting to note that the sampled respondents in the 
current study turned out to be predominantly illiterate, with only 7% having had any 
formal education at all. This and problems o f measurement of the variable are likely 
responsible for the observed result.
The observation on farmers’ use o f farm credit may be similarly explained. The general 
impression is that the use of credit for farming is very limited in the study area. Apart 
from cotton farmers who obtained farm credit in the form o f inputs, only 5% o f all 
sampled farm households indicated having used credit for farming in the survey year. The 
fact that it showed the expected sign could suggest validation of the hypothesis concerning 
the variable.
The chemical fertilizer variable, which captures the effect o f the use o f complementary 
inputs, neither showed the expected sign nor was it statistically significant. The non­
significance may be due to the limited use of the input, while the negative sign may be 
explained by the fact that chemical fertilizers are not normally applied to sorghum in the 
study area. This is not an isolated case as Garcia et al (1997) noted that inspite o f their 
knowledge and use of recommended technology farmers used less amounts of expensive 
inputs. In addition Coulibaly (1995) noted with concern, that in sub-Saharan Africa, 
inorganic fertilizers are mainly used on cotton and maize, and not on sorghum and millet.
70
University of Ghana          http://ugspace.ug.edu.gh
He concluded that this presents a formidable challenge to researchers, extension workers 
and policy makers.
The rest of the factors proved to be significant determinants of the adoption of improved 
sorghum varieties. Among the socio-economic factors, the age variable was found to be 
positive and significant. Hence the hypothesis that young farmers are less risk averse and 
more adventurous could be considered confirmed. This observation is similar to the 
findings o f Poison and Spencer (1992) and Manyong et al., (1998). However, this is 
contrary to the earlier observation by Pannin (1988) that decisions regarding the adoption 
of innovations are positively influenced by the age factor.
Farm size, the available family labour and non-farm income tend to exert positive 
influence on farmers’ adoption decisions. Such findings appear to be the rule rather than 
the exception. For instance, Nkonya et al (1998) found that farm size was an important 
determinant of maize varietal adoption in Northern Tanzania. Similarly, Feder et al. 
(1985) showed that the share of area allocated to modern crop varieties increases with 
farm size. Also, Poison and Spencer (1992) found farmers’ income to be an influential 
variable in adoption decisions. Besides, Akinola (1985) Pannin (1988), and Hailu (1990) 
found the available family labour to be an important determinant of adoption of chemical 
fertilizers and bullock traction technologies in the Northern region of Ghana. These 
results should be expected since the production of sorghum in the study area is in the 
hands of small-scale farmers who use little or no external inputs (including credit), with 
the farm household as the main source of farm labour.
71
University of Ghana          http://ugspace.ug.edu.gh
Farm type and the number of years that farmers fallow their farmland were found to 
significantly influence the adoption behaviour of sorghum farmers. As expected, fallow 
(indicating intensity of land use) showed a negative effect as opposed to farm type. This is 
similar to the observation by Yaron et al. (1992) that small land area, which forced 
continuous and intensive cultivation (an indication of pressure on farmland) provided an 
incentive to adopt high pay-off, input intensive innovations in the Nazareth region of 
Israel. These findings are in line with the induced innovations hypothesis of Hayami and 
Ruttan (1984) to the effect that land-constrained farmers tend to adopt high yielding input 
intensive varieties. It should be noted that compound farming in the study area is 
practiced in areas where pressure on farmland has made it almost impossible for farmers 
to practice the bush fallow system.
The distance to the nearest purchase point for improved sorghum seed, a proxy for market 
access to improved sorghum seed was significant at 1%. Coulibaly (1995) emphasized the 
importance of access to input (and output) markets as an incentive for sustained adoption 
of new technologies. Specifically, Hailu (1990) observed that the distance to supply 
centres significantly influenced farmers’ adoption of chemical fertilizers in the Northern 
region of Ghana. Hence the hypothesis that longer distances to the nearest purchase point 
exerts a negative effect on adoption of improved sorghum varieties is confirmed.
Last but not the least is the variable examining the effect o f the characteristics of the 
technology itself, the perception index. This variable was found to be significant at the 
5% and to exert a positive influence on adoption. Similar observations have been made in
72
University of Ghana          http://ugspace.ug.edu.gh
adoption studies in the past. Sperling and Loevinsohn (1993) noted that the salient factor 
in the adoption and spread of new varieties is the grower’s appreciation of the variety. In 
an earlier study, Haugerud (1988) observed that the acceptability of new potato varieties 
depended, among other things, on taste preference, the starch content, cooking time, 
blight resistance, market acceptance and compatibility with the existing farming system. 
In addition, Yapi et al. (1998) reported that the most important reasons that were 
advanced for adopting S-35, an improved sorghum variety was its early maturity, high 
yield and good taste. These are some of the characteristics factored into the perception 
index and such findings should be o f utmost interest to researchers. The discussion of the 
results of farmers’ preferred characteristics is undertaken in section 4.2.4 below.
The mean predicted probability of adoption among adopters was found to be higher 
(0.88) than it was among non-adopters (0.14). Regardless of those that did not show the 
expected signs, most of the explanatory variables as measured by the corresponding 
weighted elasticities tend to be potential areas of future policy influences. This follows 
from the interpretation of these elasticities, which is simple and direct. For example, a 
change (policy) that brings about a 10% increase in non-farm income would lead to a 1% 
increase in the probability of adoption. Similarly, any change (policy) that brings about a 
10 % increase in the available family labour, farm size or a 10% improvement in farmers’ 
perception would lead to 8%, 3% and 2% increase in the probability of adoption, 
respectively. Also, a 10% improvement in farmers’ market access to improved seed 
would lead to a 6% increase in the probability of adoption.
73
University of Ghana          http://ugspace.ug.edu.gh
It should be noted that the probabilities in themselves are not important, but the 
elasticities which measure the direction and magnitude of the effects of changes in the 
explanatory variables on the probability of adoption.
4.2.3: Determinants of Intensity of Adoption: Tobit Model
The results of the Tobit model specified to examine the rate and intensity of adoption of 
improved sorghum varieties are summarised in Table 4.4 and discussed below.
Table 4.4: Estimates of the Determinants of Intensity of Adoption: Tobit Model 
(1999/2000, UWR, Ghana)
Variable Estimated
coefficient
Standard
error
Asymptotic
T-ratio
Age 0.92167 0.35312 3.2009***
Credit 0.37059 0.46218 0.9834
Extension visits -0.01326 0.01130 -1.4393
Non-farm income 0.02931 0.25221 0.1425
Fallow -0.17074 0.07661 -2.7331***
Family Labour 
Distance to
0.06220 0.04554 1.6750*
nearest salespoint -0.02882 0.01656 -2.1346**
Perception Index 0.53019 0.34861 1.8652*
Farm Type 0.09143 0.26221 0.4276
Farm Size 0.20073 0.09543 2.5796**
Constant -1.00250 0.61314 -2.0051
Sample size 100
Estimated rate o f adoption 0.40
Observed rate o f adoption 0,48
Expected share o f area under improved varieties
entire sample E(Y) 0.2101
among adopters E(Y*) 0.5633
Mean predicted probability o f adoption
among adopters 0.5567
among non-adopters 0.2737
Significance levels: *** = 1%; ** = 5%; * = 10%
74
University of Ghana          http://ugspace.ug.edu.gh
Based on the asymptotic t-values of the parameter estimates six of the ten independent 
variables included in the model proved to be significantly associated with the intensity of 
adoption of improved sorghum varieties, and thereby confirm the hypotheses laid down.
Specifically, the results confirm that the probability o f a sorghum farmer planting 
improved varieties and the share o f sorghum area planted with improved varieties would 
be higher with improvement in his access to markets, perception about improved varieties, 
with increasing available family labour and farm size. The results confirm that younger 
farmers are more inclined to adopt and plant a higher proportion o f their sorghum area to 
improved varieties than older ones. Hailu (1990) reported findings that are very similar to 
the findings on the above variables.
The results on the fallow variable do indicate that farmers with shorter fallow periods are 
more likely to adopt and plant more o f their sorghum land to improved varieties. As 
explained earlier, this confirms the hypothesis that land-constrained farmers more often 
resort to high yielding input intensive varieties.
However, for completeness it is important to mention that non-farm income and farm 
type, though with the expected signs were not significant at the 10% level. Here again the 
extension contact variable neither showed the expected sign nor was it significantly 
associated with adoption. Also, as in the probit model the use of farm credit in the survey 
year was not a significant determinant of adoption.
75
University of Ghana          http://ugspace.ug.edu.gh
As was noted about the probit model, the parameter estimates o f the tobit model provide 
no direct information about the effect of changes in the explanatory variables on the 
probability of adoption and the intensity o f use. To predict such effects, the parameter 
estimates must be weighted by the predicted probability of adoption. The change in the 
expected share o f sorghum area under improved varieties, E ( Y )  in the entire sample as a 
result o f a change in the kUl explanatory variable is given by
0E(Y)/3Xik=F(Xi’ p>pk (4.1)
where Pk represents the coefficient of the k,h explanatory variable (Maddala, 1983). 
McDonald and Moffit, (1980) developed a useful decomposition of this change which 
helps derive the required elasticities given by
5 E ( Y ) / 5 X ik  =  F ( X j ’ p)/a • 5 E ( Y ; * ) / 3 X i k  +  S E ( Y * >  F ( X , - ’ P ) / 5 X ik (4.2)
It has been shown that by multiplying both sides of the above equation by X/E(Yi) each 
term can be converted to an elasticity measure (Kinsey, 1984). With the help of this 
procedure the elasticity of the expected value of the dependent variable, E(Y) with respect 
to the kth explanatory variable has been decomposed to obtain the elasticity of the 
predicted probability of adoption and the elasticity of intensity of use. The results are 
presented in Table 4.5.
76
University of Ghana          http://ugspace.ug.edu.gh
Table 4.5: Elasticities of the Predicted Probability of Adoption and Expected Use 
Intensity of Improved Sorghum Varieties (1999/2000, UWR, Ghana)
Variable Elasticity of
probability of 
adoption
expected sorghum 
area under improved 
varieties
Age
Fallow
Family Labour 
Distance to nearest 
purchase point 
Perception Index 
Farm Size
2 . 5 7 7 3  
- 2 . 2 9 3 2  
1 . 8 8 6 4
1 . 1 6 1 5  
- 1 . 0 3 3 4  
0 . 8 5 0 1
- 1 . 5 8 7 8  
0.  8144  
1 . 5 5 8 3
- 0 . 7 1 5 5  
0 . 3 6 7 0  
0 .  7 023
The estimated rate of adoption was 0.4 (40% of sampled farm households). The expected 
proportion of sorghum area planted with improved varieties was 0.21 (21.01%) for the 
entire sample whereas it was 0.5633 (56.33%) following a decision to adopt (i.e. among 
adopters). Judging from the reported elasticities it is clear that the explanatory variables 
have more influence on the probability of adoption than the intensity of use after adoption. 
The interpretation of these elasticities is simple and straightforward. For example, a 10% 
improvement in farmers’ market access to improved sorghum seed would lead to about 
16% increase in the probability of adoption and a 7% increase in the intensity of use.
4.2.4: Farmers preferred Varietal Characteristics
The section presents the results of farmers own ranking of the varietal traits or 
characteristics that they consider in their choice of sorghum variety to plant. The traits 
included for each farmer’s consideration were the same ones that were used to compute 
the perception index used earlier. Those traits that placed first-to-third in each
elasticities are calculated at the means of the variables
77
University of Ghana          http://ugspace.ug.edu.gh
respondent’s consideration were compiled and the results are as shown in Table 4.6. From 
the table it is clear that most farmers consider yield potential, maturity and market value in 
the choice of potential sorghum varieties. Farmers’ consideration of yield potential and 
market value in cash-mediated subsistence farming such as practiced in the study area is 
quite understandable. This reflects each farmer’s desire to meet household consumption 
needs from their own farms, and possibly have surplus to sell to meet their subsistence 
cash needs. The second trait, maturity period, deserves its place because o f the short and 
erratic nature of the rainfall pattern in the region.
Table 4.6: Farmers Preferred varietal characteristics according to percentage 
of respondents (1999/2000, UWR, Ghana).
Varietal Characteristic Percentage
Yield 87
Short maturity period 62
Market value 49
Taste 48
Suitability for local dishes 41
Source: compiled from survey data.
Finally, the place given to taste and suitability for local dishes may be explained by the fact 
that the bulk of the sorghum crop is often consumed by the farm-household.
As a follow up, farmers were also asked to compare their best traditional variety with the 
best o f the improved varieties that they know in terms of varietal traits listed in Table 4.6
78
University of Ghana          http://ugspace.ug.edu.gh
above. The results in Table 4.7 below show that a good number of the sampled farmers 
assess the improved varieties to be worse than, or at best the same as the traditional ones, 
especially in terms of their suitability for local dishes, taste and market value. The high 
non-response rate could also mean that there is very little awareness about the qualities of 
the improved varieties.
Table 4.7: Comparison of Improved and Traditional Varieties by Farmers 
(1999/2000, UWR, Ghana)
Varietal Characteristic
No Response 
/Cannot tell
Improved Variety is 
Better Same Worse
Yield 41 51 7 1
Short maturity period 42 49 6 3
Market value 45 27 9 19
Taste 44 24 17 15
Suitability for local dishes 51 12 14 23
Source: compiled from survey data.
So the low rates of adoption can be attributed to the farmers’ assessment of the improved 
varieties as not being superior to their traditional ones and/ or their lack of knowledge 
about the qualities or capabilities of the improved varieties. Farmer education on the 
capabilities of the improved varieties needs to be intensified. Above all, it is necessary to 
give these farmers’ preferences due consideration in order to come out with improved 
material with a wide acceptance.
79
University of Ghana          http://ugspace.ug.edu.gh
4.2.5: The Impact of Sorghum Research: Results of Rate of Return Analysis 
A summary of the results of the rate of return analysis is presented in appendix la. The 
estimated rate was 8 percent at 1985 prices (US Dollars). Comparing this rate with 
results from similar studies makes interesting discourse. As already discussed under 
literature review, Sterns and Bernsten (1992) reported a rate of return o f 3 percent for 
sorghum and cowpea research in Cameroon, while another as low as 1 percent was 
reported for sorghum research in Niger. Also, in Niger, Mazzucato and Ly (1992) 
estimated a negative rate of return for sorghum and cowpea research, and Laker-Ojok 
(1992) estimated a negative rate for sunflower, cowpea and soybean in Uganda. While 
the situation in Uganda was blamed on the turbulent political climate o f the late 1970s and 
early 1980s, the researchers’ in Cameroon explained that sorghum has presented a 
formidable problem to researchers throughout West and Central Africa for over 30 years. 
They concluded that low returns to sorghum research, though undesirable, may simply 
reflect long run historical trends of the performance of Sorghum research. Viewed form 
this standpoint, an 8 percent rate o f return represents an enviable achievement.
However, Lopez-Pereira et al., (1991) estimated a rate of return of 32 percent for 
sorghum in the Honduras. In the said study the real cost of capital, a proxy for the next 
best alternative investment project, was estimated to be between 15 and 20 percent. More 
importantly, according to Kelly (1995) ICRS AT ranked 110 different research areas, and 
the average rate of return for the top 20 was 39 percent. Also, Kristjanson et al., (1999) 
reported that a rate o f return estimate of 13-15 percent was obtained when CIP analysed 
returns to 15 research themes. Several other impact studies have reported even higher
80
University of Ghana          http://ugspace.ug.edu.gh
estimates for the rate of return to research. A survey o f the literature indicates that such 
rates vary widely from 21 percent for maize in Zambia (Howard et al., 1992) to 135 
percent for maize in Mali (Boughton, 1992).
From the foregoing therefore the estimated rate o f return of 8 percent, baring all data 
deficiencies, though not the worst is one o f the lowest and cannot be considered attractive. 
To test the robustness of this estimate, sensitivity analysis was conducted using three 
scenarios.
(i) Discussions with some of the sorghum researchers indicate that it is the aim o f the 
programme to get as much as 50 percent of sorghum land under improved sorghum 
varieties. Therefore the effect of changes in the adoption rate (the proportion of sorghum 
area under improved varieties) was the first scenario. An increase in the adoption rate 
from 0.1 percent to 1 percent of sorghum area gave a rate of return o f 109 percent (see 
appendix lb). This shows that the current estimated low rate of 8 percent may be due 
mainly to the ridiculously low rate o f adoption.
(ii) Atokple (1993) observed that sorghum yields can be increased to double current 
levels (to between 2 and 3t/ha) with the adoption of improved varieties, chemical 
fertilizers and improved farming practices. The second scenario looked at a 10 percent 
increase in the yield over and above the current 1.35t/ha. The result was an estimated rate 
of return of 10 percent (see appendix lc).
81
University of Ghana          http://ugspace.ug.edu.gh
(iii) The current practice by sorghum farmers is that chemical fertilizers are not applied to 
the crop. However, as noted by Sanders and Vitale (1998) there is little yield potential 
from improved sorghum cultivars unless they are combined with at least moderate levels 
of fertilization. Hence a 10 percent decline in the yield o f the improved materials (from the 
current on-farm average of 1.35t/ha) provided the third scenario. This yielded an estimate 
of 6 percent (see appendix Id). The results from the above analysis suggest that the focus 
should be on improving upon the adoption rate.
82
University of Ghana          http://ugspace.ug.edu.gh
CHAPTER 5
SUMMARY, CONCLUSION AND POLICY IMPLICATIONS
5.1 Summary
The production o f sorghum, one o f the most important food crops with multiple uses in the 
savanna zone o f Ghana, has been bedevilled with several problems. The most important 
problems are low yields, lack of improved varieties and striga infestation. These problems 
elicited research efforts at the Savanna Agricultural Research Institute to produce improved 
high yielding sorghum varieties that are adapted to the socio-economic and agroecological 
conditions o f northern Ghana. This study examines the adoption and impact o f improved 
sorghum varieties that were promoted in the Upper West region.
It was established in the review of literature that farm level decisions about the adoption of 
improved technologies are often governed by socio-economic and institutional factors as 
much as they are by technology specific factors. The Probit and Tobit models, specified to 
include factors from all three categories, are used to examine the adoption and, the rate and 
intensity of use of improved sorghum varieties. The logistic curve is used to describe 
adoption pattern while the economic surplus model provides the theoretical framework used 
to estimate the rate of return to sorghum research.
A normal adoption pattern is observed with a diffusion rate of about 0.3 percent. It is found 
that age, the available family labour, non-farm income, fallow period, farmers perception 
about the varieties, farm size, farm type and the distance to the nearest purchase point for 
improved seed are significant determinants of adoption. Education and farm credit though
University of Ghana          http://ugspace.ug.edu.gh
with the expected signs, were found not to be significant. The use o f  chemical fertilizers and 
the extension contact variable both showed the wrong signs though the extension variable is 
statistically significant. The mean predicted probability o f adoption was 0.88 among adopters 
whereas it is only 0.13 among non-adopters.
The factors that influence both the rate and intensity o f adoption are: age, available family 
labour, fallow period, farmers perception about the varieties, farm size and the distance to the 
nearest purchase point for improved seed. Education and farm type are found not to be 
significant while the extension contact variable neither showed the expected sign nor is it 
significantly associated with the dependent variable. The estimated rate o f adoption is 0.40 
(40% of sampled farmers) while the intensity of use, measured as the proportion of sorghum 
area under improved varieties, is 56.33% following a decision to adopt.
It is found that farmers’ choice of variety to plant depends on yield, maturity period, market 
value, taste and suitability for local dishes, among others.
Finally, the economic surplus model framework is employed to estimate a rate o f  return o f 8% 
for sorghum research. A sensitivity analysis is conducted using three different scenarios, an 
increase in the adoption rate to 1% of sorghum area, a 10% decline in yield and a 10% 
increase in yield. The results reveal that the first scenario has the highest rate of return of 
109%.
84
University of Ghana          http://ugspace.ug.edu.gh
5.2 Conclusions
An important aspect of agricultural development is the adoption o f innovations by farmers. In 
order to overcome some o f the constraints to sorghum production, the Savanna Agricultural 
Research Institute (SARI) has for over 30years embarked on a sorghum improvement 
' programme, through which a number of improved varieties have been released to farmers. 
However, the fact that a technology has been released or recommended does not guarantee its 
widespread adoption. It is therefore important to identify the factors that influence farmers’ 
adoption decision with respect to the technology.
%
The influence of each of the explanatory variables was found to be more on the probability of 
adoption than the intensity of use. This means that the chances o f getting a farmer to plant 
improved sorghum are better than getting a user to plant more. The intensity o f use estimated 
at 0.56 for adopters implies that the question o f acceptability o f the improved varieties is yet 
to be addressed. This explains why the estimated area under improved sorghum varieties is 
less than 1 percent whereas the percentage o f farmers planting improved sorghum varieties is 
as high as 40 percent of the sample. Farmers are simply not convinced about the superiority 
of the improved varieties.
There is the need to redirect more attention to farmers’ preferences and socio-economic 
conditions. This is against the background that such a re-orientation would enhance 
acceptability and hence adoption. This will ultimately improve the impact o f Sorghum 
research and hence justify further investments.
85
University of Ghana          http://ugspace.ug.edu.gh
5.3 Policy Implications
Policy implications of the findings of this study are outlined in this section. The estimated 
Probit and Tobit models have identified factors that are needed to boost adoption. Improving 
farmers’ perception about the improved varieties is essential in influencing adoption and the 
■ intensity of use. There is the need to effectively educate farmers to understand the capabilities 
o f the varieties and the production procedures for optimum results. This can best be achieved 
by forging strong research-extension-farmer linkages to ensure that while farmers are being 
educated, there is an effective feedback mechanism to convey their opinions and preferences 
to researchers. Incorporating farmers’ concerns in Sorghum research will help to develop 
varieties that meet widespread acceptance. There is room for further varietal improvement 
with qualities of traditional varieties that farmers' desire most. In this regard efforts should be 
directed at improving the yield potential of such traditional varieties as Chare/Nyoso, Dawelle 
and Latuor/Gon as these already enjoy widespread acceptance among sorghum farmers.
It is recommended that extension efforts should focus more on the younger farmers while 
research is made more sensitive to farmer resource levels. This is against the background that 
younger farmers are predisposed to adopt the improved varieties, and access to family labour 
and non-farm income also enhance adoption. Also, in the face of limited access to credit, 
empowering farmers to engage in non-farm income generating activities could be an effective 
strategy of increasing farm level investment in general, and adoption of farm technologies in 
particular.
86
University of Ghana          http://ugspace.ug.edu.gh
Farmers’ market access to improved seed is essential in influencing the intensity o f adoption. 
Therefore, it is necessary to intensify efforts at improving distribution networks, which could 
better be achieved through improved infrastructure and active involvement o f farmers in 
acquisition of inputs. To ensure that improved seeds (and other inputs) are readily available 
' to farmers there is the need for government to integrate the activities of extension agents, 
NGOs and the Seed Growers Association in the area. Also, as threats of land scarcity grow, 
there is the need for more strategic research to meet the technological needs for land use 
intensification. The Sorghum varieties are proving to meet that kind of need among farmers in 
the Upper West region.
Finally, the prospects for Sorghum research are good and there is the need for further 
investment. Though the rate of return of 8% may seem low, it is good. This is because of a 
possible underestimation of benefits since the estimated cost included more than just the 
development of varieties. The results of sensitivity analysis show that effective technology 
transfer is essential for reaping the benefits of new technologies. Hence more efforts in this 
direction will make prospects for Sorghum research even better.
87
University of Ghana          http://ugspace.ug.edu.gh
5.4 Suggestions for Future Research
The database for conducting reliable impact assessment of agricultural research/projects is 
very weak. No systematic adoption rates are available on the various recommended 
technologies, neither are there consistent and reliable costs data on most of these 
technologies. Yet the results of impact assessments obey the “Garbage-In-Garbage-Out” 
principle. Hence, there is the need for research institutions in collaboration with extension 
organisations to pursue a systematic policy to build up and maintain a comprehensive database 
for the purpose.
Such a database should include time-series information on:
Quantities o f inputs used, prices per unit of input used, quantities of output produced, prices 
per unit of output, adoption rates for the technology or quantities o f improved seed planted by 
or distributed to farmers, research costs, and the costs of technology transfer. Also, the 
recommended input use rates, and the demand and supply elasticities of the product concerned 
should be available.
University of Ghana          http://ugspace.ug.edu.gh
REFERENCES
Abatania, L. N. (1999). Sustainability of Maize-Based Cropping Systems in Northern
Ghana. An MPhil thesis submitted to the Department of Agricultural Economy and 
Farm Management, University of Ghana, Legon.
Adesina, A. and Zinnah (1992). “Ex-Post Assessment of the Adoption and Impacts of 
Mangrove Swamp Rice Varieties in West Africa.” A Paper presented at the 
MSU/AID Symposium on the Impact of Technology on Agricultural 
Transformation in Africa, October 14-16,1992.
Ahmed, M., Salih, A., and J. Sanders (1992). “The Impact o f Hageen-Dura I in Sudan.”
Paper presented at the Symposium on the Impact of Technology on Agricultural 
Transformation in Africa, October 14-16, 1992, Washington, D C.
Akibonde, L. (1969). “Relationship between the Socio-economic Characteristics of
Farmers i f in British Columbia and their Contacts with District Agriculturist.” 
Unpublished MS Thesis, University of British Columbia.
Akino, M. and Y. Hayami (1975). “Efficiency and Equity in Public Research: Rice Breeding 
in Japan’s Economic Development” . American Journal o f  Agricultural 
Economics, 57:1-10.
Akinola A. A., and T. Young (1985). “An Application of the Tobit Model in the Analysis of 
Agricultural Innovation Adoption Process: A study of the Use of Cocoa Spraying 
Chemicals among Nigerian Cocoa Farmers” Oxford Agrarian Studies, Band 14,
26-51
Akinola, A. A. (1987). “An Application of Probit Analysis to the Adoption of Tractor Hiring 
Service Scheme in Nigeria. Oxford Agrarian Studies, 26
Alao J. A. (1973). “Community Structure and Farmers Farm Practice Adoption” Bulletin o f  
Rural Economics and Sociology. Vol. 8 No. 2:281-312.
Aldrich, J. H. and Nelson, F. D. (1984). Linear Probability. Logit and Probit Models. 
Sage,BeverlyHills,CA, 95p.
Alston J. M., Norton G. W., and P. G. Pardey (1995). Science Under Scarcity: Principles and 
Practices for Agricultural Research Evaluation and Priority Setting. Cornell 
University Press, Ithica,NY.
Amemiya, T. (1984). “Tobit Models: A Survey”. Journal o f  Econometrics, 24:3-61.
Atokple, I. D. K. (1992), “An overview of Sorghum and Millet improvement in Ghana” 
Nyankpala Agricultural Experiment Station, Tamale.
89
University of Ghana          http://ugspace.ug.edu.gh
Atokple, I. D. K. 1993. “Current Progress towards the Utilization of Sorghum in Brewing 
Lager beer in Ghana”. In: Processing and Industrial Utilization of Sorghum and 
Related Cereals in Africa. Proceedings of the Regional Symposium on the Processing 
and Industrial Utilization of Sorghum and Related Cereals in Africa held at 
Ouagadougou, OAU-SAFGRAD, Ouagadougou, Burkina Faso.
Berkson J. " The Application of the Logistic Function to Bio- Assay" Journal o f  American 
Statistical Association. 39: 357-365
Boughton D. and B. H. de Frahan (1994). Agricultural Research Impact Assessment: The 
Case of Maize Technology Adoption in Southern Mali. Department of Agricultural 
Economics and Department o f Economics, Michigan State University, East 
Lasing, Michigan48824-1039 USA, MSU International Working Paper No.41
Brennan J.F. (1949). "Evaluation of Parameters in the Gompetz and Makeham Equations" 
Journal o f  American Statistical Association 44:116-121
Byerlee, D. (1986). “From Adaptive Research to Farmer Recommendations and Extension 
Advice”. Agricultural Administration and Extension 231-244,
CIMMYT Economics Program. 1993. The Adoption of Agricultural Technology: A Guide 
for Survey Design. Mexico, D.F,: CIMMYT.
Coulibaly, O. N. (1996). “Adoption of Improved Technologies in the West African Semi- 
Arid Tropics: Success Stories and Challenges” In, Achieving Greater Impact from 
Research Investments in Africa, (edited) by Breth, Steven A. Mexico City, Sasakawa 
Africa Association 1996,
Dagenais, M. G. (1969) "A Threshold Regression Model" Econometrica 37: 193-203
Davis, J. A. (1981). “Comparison o f Procedures for Estimating Returns to Research Using 
Production Functions”, Australian Journal o f  Agricultural Economics, 25, 60-72.
Denning, G. L. (1991). “ Intensifying Rice-based croping Systems in the Rainfed Lowlands 
of Iloilo, Phillipines: Results and Implications”. In Planned Change in Farming 
Systems: Progression in On-Farm Research (edited) by Robert Tripp. Chichester, UK: 
John Wiley.
Dickson K. B. And G. Benneh (1970). A New Geoerahpv of Ghana. Longman Group UK 
Ltd, New Revised Ed.
90
University of Ghana          http://ugspace.ug.edu.gh
Edwin, J. A. (1996). The adoption of improved Rice varieties in inland valley swamps in
Kambia and Port Loko Districts of Sierra Leone. A report presented at the 4lh plenary 
session o f thje Rice economics task force of West Africa (RETWA). WARDA, 
Bouake, CoAte d’Ivoire. 5lh -8 ,h March, 1996.
Eicher, C. K. and J. M. Staatz (ed) (1984). Agricultural Development in the Third World. 
John Hopkins University Press, Baltimore.
.Engmann, S. C. (1993). An Evaluation of Production. Marketing and Utilization of Soybean 
in Ghana: A Case Study of the Greater Accra Region. ABSc. Dissertation submitted 
to the Department of Agricultural Economy and Farm Management, University of 
Ghana, Legon.
Evenson, R. E., Landau, D., and D. Ballou (1987). “Agricultural Productivity Measurement 
for the US States”, In, Evaluating Agricultural Research and Productivity.
Proceedings of a Symposium, Atlanta, Georgia, January 29-30, Miscellaneous 
Publication 52-1987 Minnesota. Agricultural Experiment Station.
FAO (1986). African Agriculture: Input Supply and Incentive Policies. Annex III: Rome.
Feder, G. ; R. E. Just and D. Zilbermann (1985). “Adoption of Agricultural Innovations in 
Developing Countries: A Survey”, Economic Development and Cultural Change, 33: 
255-298.
Finney D.J. (1952). Probit Analysis. Cambridge, England Cambridge University Press.
Francis, D. G. (1976). “Population Characteristics and Response to Rural Development 
Activities” . Bulletin o f  Rural Economics and Sociology. 10(1): 21-28.
Frolich W., and S. S. J, Buah (1991a). Country Report o Sorghum Improvement in Ghana.
A paper presented at the Workshop of West and Central Africa Sorghum Research 
Network. March 8-15, 1991.
Garcia, E. V., Garza, FI. N., Mendoza, S. M., andG. F., Gambero (1997). “Institutional 
Supports and Agricultural Technology: the Case of Maize, Sorghum Soybean and 
Safflower in the Ejido Celaya, Tmaulipas” . Agrociencia, 31(4): 465-472.
Ghana Statistical Service (1984) Population Census. Demographic and Economic 
Characteristics. Upper West Region.
Gittinger, J. P. (1982). Economic Analysis of Agricultural Projects. The Johns Hopkins 
University Press, Baltimore, Maryland, 21218 U.S.A. 2nd ed.
91
University of Ghana          http://ugspace.ug.edu.gh
Gladwin, C. (1989). Ethnographic Decision Tree Modelling. Sage Qualitative Research 
Methods Series No. 19. Beverly Hills, Califonia: Sage
Goldberger, A. S. (1964). Econometric Theory. John Wiley and Sons Inc., New York.
Griliches, Zvi (1957) "Hybrid corn: An Exploration in the Economics of Technological 
Change" Econometricci 25: 501-522
.Griliches, Zvi. (1958). “Research Costs and Social Returns: Hybrid Corn and Related 
Innovations”. Journal o f  Political Economy 66:419-31.
Hailu Z. (1990), The Adoption of Modern Farm Practices in African Agriculture: Empirical 
Evidence about the Impacts of Household Characteristics and Input Supply Systems 
in the Northern Region of Ghana. Nyankpala Agricultural Research Report No.7.
Harlan, J. R., and de Wet J. M. J. (1972). “A Simplified Classification of Cultivated 
Sorghum”. Crop Science 12: 172-176.
Haugerud, A. (1988). “Anthropology and Interdisciplinary Agricultural Research in 
Rwanda”. In: Anthropology of Development and Change in East Africa. West 
View Press, 1988, p.137-160.
Hill L. and P. Kau (1973). “Application of Multivariate Probit to a Threshold Model of
Grain Dryer Purchasing Decisions”. American Journal o f  Agricultural Economics 
55: 19-27.
Hounkpe, E.R.C. (1999). Adoption of Agricultural innovations: A case study of maize 
narrow crib in Southern Benin. An M Phil thesis, Department of Agricultural 
Extension, University of Ghana, Legon.
Howard, J., G., Chitalu, and S., Kalonge (1992). “ The Impact of Investments in Maize
Research and Dissemination in Zambia: Preliminary Results” . Paper presented at the 
Symposium on the Impact of Technology on Agricultural Transformation in Africa, 
October 14-16, 1992, Washington, D.C.
Hunter, J. M. (1967). "Population Pressure in a part of the West Africa Savanna, A Study of 
Nangodi, N.E. Ghana" Annals o f  Association o f  American Geographers 57: (1)
Jayne, T. S., Khatri Y,, Thirtle C. and Reardon T. (1994). “A Profit Function Approach to 
Productivity Change in Zimbabwean Agriculture: Implications for Research and 
Policy”, American Journal o f  Agricultural Economics, 76:613-618.
Judge, G. G.; W. E. Griffiths; R.C. Hill; T. C. Lee and H. Lutkepol (1985). Introduction to 
the Theory and Practice of Econometrics. John Wiley & Sons, Inc. New York, 2nd ed.
92
University of Ghana          http://ugspace.ug.edu.gh
Kebede, Y. Gunjal, K. and G. Coffin (1990). Adoption of New Technologies in Ethiopian 
Agriculture: The Case of Tegulet-Bulga District, Shoa Province. Agricultural 
Economics, 4(1990):27-43. Elsevier Science Publishers B. V.
Kelly T. G., Ryan J. G., and B. K , Patel. (1995). “Applied Participatory Priority Setting in 
International Agricultural Research: Making Trade-offs Transparent and Explicit” 
Agricultural Systems, 49(2): 177-216.
. Kinsey, J. (1984). “Probit and Tobit Analysis in Consumer Research” Proceedings of the 
30lh Annual Conference of the American Council on Consumer Interests, Atlanta, 
Goergia.
Kislev, Y., and Slichori-Bachrach, N. (1973). “The Process of an Innovation Cycle”.
American Journal o f  Agricultural Economics 55: 28-37.
Kristjanson, J., Swallow, B., Krusha, P., de Leeuw and S.,Nagda (1999). Using Economic 
Surplus Model to Measure Potential Returns to International Livestock Research:The 
Case of Trypanosomiasis Vaccine Research. IRLI Impact Assessment Series 4,
IRLI (International Livestock Research Institute), Nairobi, Kenya.
Laker-Ojok, Rita (1992). “The Rate of Return to Agricultural Research in Uganda: The 
Case of Oilseeds and Maize.” Michigan State University, East Lasing.
Lin, J. Y. (1991). “Education and Innovation Adoption in Agriculture: Evidence from
Hybrid Rice in China” American Journal o f  Agricultural Economics, 73:713-723.
Lopez-Pereira, M.A., D. Gonzalez-Rey, and J.H. Sanders (1991). “The Impacts ofNew 
Sorghum Cultivars and Associated Technologies in Honduras” Proceedings, 
International Sorghum & Millet CRSP Conference, July 1991,Corpus Christi,TX. 
Lincoln, NE: University of Nebraska, INTSORMIL Publication No. 92-1.
Lu, Y., Cline P.and Quance, L. (1979). Prospects for Productivity Growth in US Agriculture. 
Agriculture Economic Report, No. 435 ESCS, USDA Washington D C.
Maddala, G. S. (1983). Limited-Dependent and Qualitative Variables in Econometrics. 
Econometric Society Monographs No. 3. Cambridge University Press.
Mansfield, E. (1961). “Technical Change and the rate of Innovation” Econometrica 
29:441-463.
93
University of Ghana          http://ugspace.ug.edu.gh
Manyong, V. M.; K. E. Dashiell; B. Oyewole and G. Blahut (1998). “Spread of New Soybean 
Varieties in a Traditional Soybean Growing Area of Nigeria” In, Towards 
Sustainable Farming Systems in sub-Saharan Africa. Edited by Bezunah, T., S., 
Ouedraogo, J. M., Menyonga, Jean-Didier, Zongo and M., Ouedraogo. AAFSRET 
Secretariat and OAU/STRC-SAFGRAD, Ouagadougou, Burkina Faso. 1998.
Masters, W. A., with B. Coulibaly., D. Sanogo., M. Sidibe., A. Williams., J. H. Sanders and 
J. Lowenberg-DeBoer, (1996). The Economic Impact of Agricultural Research: A 
Practical Guide. Department of Agricultural Economics, Purdue University, West 
Lafeyette, IN.47907.
Mazzucato V. And S. Ly (1994). An Economic Analysis of Research and Technology
Transfer of Millet. Sorghum, and Cowpeas in Niger. Michigan State University, East 
Lasing, Michigan 48824-1039 USA MSU International Working Paper No. 40, 1994
McDonald, J. F. and A. Moffit (1980). “The Uses of Tobit Analysis”. Review o f  Economics 
and Statistics, LXII, 318-321.
Mellor, J. W. (1966). The Economics of Agricultural Development. Itheca, NY: Cornell 
University Press.
MOFA (Ministry of Food and Agriculture) (1997). Agriculture in Ghana: Facts and Figures. 
Policy Planning, Monitoring and Evaluation Division, Presbyterian Press, Accra.
Mueller, R. A. E., and H. G. P. Jansen (1988). “Farmer and Farm Concepts in Measuring 
Adoption Lag” . Agricultural Economics: 121-124.
Nelson, Richard (1968). “A Diffusion Model of International Productivity Differences in 
Manufacturing Industry” American Economic Review. 58:1219-1248
Njoku, J.E., (1993). Determinants of Adoption of Improved Oil-Palm Production
Technologies in Imo State. Nigeria. Winrock International Institute for Agricultural 
Development, African Rural Social Science Series, Research Report No. 10.
Nkonya, E., P. Xavey, H. Akonaay, W. Mwangi, P. Anandajayasekeram, H. Verkujil, D. 
Martella, and A. Moshi (1998). Adoption of Maize Production Technologies in 
Northern Tanzania. Mexico, International Maize and Wheat Improvement Center 
(CIMMYT), The United Republic of Tanzania, and Southern African Center for 
Cooperation in Agricultural Research (SACCAR).
Norton, G. and J. Davis (1981). “Evaluating Returns to Agricultural Research: A Review” 
American Journal o f  Agricultural Economics 63.4:685-99
94
University of Ghana          http://ugspace.ug.edu.gh
Nweke, F. I. (1988). “Collecting and Interpreting Relevant Data Required to Link
Research Results from varying Environments: An Economist’s Perspective” In; 
International Institute for Tropical Agriculture. Cassava-Based Cropping Systems 
research 1. Ibadan, Nigeria.
Nyankpala Agricultural Experiment Station, Annual report. Various issues. Nyankpala 
Agricultural Experiment Station, Tamale.
Nyaribo-Roberts, F. and Ospina, E. (1992). “Methodologies and Impacts of Small
Ruminant Technologies on Small Farm Systems in Kenya.” A Paper presented at the 
MSU/AID Symposium on the Impact of Technology on Agricultural Transformation 
in Africa, October 14-16,1992.
Oehmke J.F. and E. W. Crawford (1993). The Impact of Agricultural Technology in
Sub-Saharan Africa: A Synthesis of Symposium Findings. Michigan State University, 
East Lasing, Michigan48824-1039 USA MSU International Working Paper No. 14.93
Panin, A. (1988). Hoe and Bullock Farming Systems in Northern Ghana. A Comparative 
Socio-economic Analysis. Nyankpala Agricultural Research Report No. 1.
Petzel, T. (1976). “Economics and the Dynamics of Supply.” PhD Dissertation, University 
of Chicago, 1976.
Pindyck R.S. and Rubinfeld D.L. (1991). Econometric Models and Economic Forecasts 
MaGraw-Hill, Inc. New York. 3rd ed.
Poison, R. S. and D. S. C. Spencer (1992). The Technology Adoption Process in Subsistence 
Agriculture: The Case of Cassava in South-Western Nigeria. IITA Research No.5.
Roe, T. (1983). “Risks and Information: Farm Level Impediments to Transforming
Traditional Agriculture”. In, J. J. Molnar and A. A. Clonts (eds), Transferring Food 
Production Technologies to Developing Nations. Westview, Boulder, CO, pp.85- 
105.
Rogers, E.M. (1983). Diffusion of Innovations. Macmillan, Landon; New York.
Ruthenberg, H. (ed. by H. Janke). (1985V Innovations Policy for Small Farmers in the 
Tropics: The Economics of Technical Innovations for Agricultural Development. 
Oxford.
Ruttan, V. W. and Y. Hayami. (1984). “Induced Innovations Model of Agricultural
Development.” (Edited) by C. K Eicher and J. M. Staatz. Agricultural Development 
in the Third World. John Hopkins University Press, Baltimore, (1984).
95
University of Ghana          http://ugspace.ug.edu.gh
Ryan, B. and N. Gross. (1943). “The Diffusion of Hybrid Seed Corn in Two Iowa 
Communities.” Rural Socioloy 8:15-24 RS(E).
Sanders, J. H., and J. D., Vitale (1998). “Technology Development for Traditional Cereals 
in the Sahelian Countries”. In, Towards Sustainable Farming Systems in sub-Saharan 
Africa. Edited by Bezunah, T., S., Ouedraogo, J. M., Menyonga, Jean-Didier, Zongo 
and M., Ouedraogo. AAFSRET Secretariat and OAU/STRC-SAFGRAD, 
Ouagadougou, Burkina Faso. 1998.
Schipprack, W. and H. Mercer-Quarshie (1984), “Sorghum improvement in Ghana’'. In: 
Sorghum improvement in West Africa, Proceedings of the Regional Workshop held 
at Ouagadougou, November 27-30, 1984. ICRISAT: Ouagadougou.
Schuh G. H, (1989). “Extension Support Activities and Technology Utilization for Poor 
Small Farmers”. In: Technology Systems for Small Farmers Issues and Options 
edited by Abbas M. Kesseba (1989), Westview Press Inc. In coop with IFAD, 5500 
Central Avenue, Boulder, Colorado80301.
Schuh, E. and H. Tollini (1979). Costs and Benefits of Agricultural Research: The state of 
the Arts. World Bank Staff Working Paper No. 360 Washington, DC.
Schultz, T. W. (1964). Transforming Traditional Agriculture. New Haven, Connecticut, 
USA. Yale University Press.
Schwartz, L; Sterns, J. and Oehmke, J. (19993). “Economic Returns to Cowpea Research, 
Extension, and Impact Distribution in Senegal.” Agricultural Economics 8:161-171.
Shakya, P. B. and J. C. Flinn (1985). “Adoption of Modern Varieties and Fertilizer Use on 
Rice in the Eastern Tarai of Nepal” . Journal o f  Agricultural Economics 36(3): 
509-519.
Songsore J. (1985). Intra-regional Community Flows in Sub-Saharan Africa: the Case of Wa 
Town and North Western Ghana. Occasional Paper, No. 19, Centre for Development 
Studies, University College, Swansea.
Songsore, J. and A. Denkabe (1995). Challenging Rural Poverty in Northern Ghana: The 
Case of the Upper-West Region, The University of Trondheim, Centre for 
Environment and Development, UNIT-SMU.
Sperling, L., and M. E., Loevinsohn. (1993). “The Dynamics of Adoption: Distribution and 
Mortality of Bean Varieties among Small Farmers in Rwanda” . Agricultural Systems, 
41(4)441-453, (UK).
96
University of Ghana          http://ugspace.ug.edu.gh
Sterns J. A. And R. H. Bernsten (1994), Assessing the Impact of Sorghum and Cowpea 
Research and Extension in Northern Cameroon. Department of Agricultural 
Economics and Department of Economics, Michigan State University, East Lasing, 
Michigan48824-1039 USA MSU International Working Paper No.43, 1994.
Swartz, L. A., J. A., Sterns and J.F., Oehmke (1993). “ The Economic Payoff to 
Agricultural Research and Extension: The Case of Cowpea in Senegal.” 
Agricultural Economics, 8:161-171.
Tarde, G. (1969). The Laws of Innovation. pl27 translated by E. C. Parsons, New York,
Holt, reprinted 1969. University of Chicago Press.
Teal, F. and Dinwiddy, C. (1996). Principles of Cost-Benefit Analysis for Developing 
Countries. Cambridge University Press.
Terbobri P., L. Abatania., I. D. K. Atokple and R. A. Kanton (1998). Evaluation of the
Adoption of improved Sorghum Varieties in Ghana. A Research Proposal submitted 
to the West and Central Africa Sorghum Research Network.
Thirtle C. and P. Bottomley (1989). “The Rate of Return to Public Sector Agricultural 
R&D in the UK, 1965-80”, Applied Economics, 21,1063-1086.
Thirtle, C. (1999). “Productivity and the Returns to Levy-Funded R&D for Sugar Production 
in the Eastern Counties of England”, Journal o f  Agricultural Economics,
50(3): 450-467.
Tomek, W.G. (1968) “Regression Analysis with a Limited Dependent variable” Journal o f  
farm  Economics 50:445-447
White, K. J. (1978). Shazam: A General Computer Program for Econometric Methods. 
Econometrica 46:239-240.
World Bank (1999). What is the Impact of Agricultural Research in Africa? 
http ://www. worldbank. org/ afr/affsr/impact 1. htm
Yapi, A., Kergna, A. O., Bebrah, S. k., Sidibe, A., Sanogo, O., Bantilan, M. C. S., and
P.K., Joshi (1998). Impact of Sorghum and Millet Research in Mali. In, Assessing 
Joint Research Impacts. Proceedings of an International Workshop on Joint Impact 
Assessment of NARS-ICRISAT Technologies for the Semi-Arid Tropics,
Patancheru, India, 2-4 Dec. 1996
Yaron D., Dinar A. and H. Voet (1992). “Innovations on Family Farms: The Nazareth Region 
oflsrael”. American Journal o f  Agricultural Economics 74:361-70.
97
University of Ghana          http://ugspace.ug.edu.gh
Zellner, A., and Tong-Hung Lee, (1965) “Joint Estimation of Relationships Involving 
Discrete Random Variables”, Econometrica 33:382-394
Zinnah, M. M., J. L. Compton and A. A. Adesina (1993). “Research-Extension-Farmer 
Linkages within the Context of Generation, Transfer and Adoption of Improved 
Mangrove Swamp Rice Technology in West Africa”. Quarterly Journal o f  
International Agriculture (Germany). 32(2): 201-214.
98
University of Ghana          http://ugspace.ug.edu.gh
Appendix la: Rate of Return Calculation
YEAR Ct K, Pt Qt('000)kg B, Bnel
1966 232309.4 -232309
1967 328878 -328878
1968 465594.8 -465595
1969 564979.5 -564979
1970 799840.3 -799840
.1971 512999.2 3.56E-05 7360.406 172.7 45306.38 -467693
1972 914926.2 3.36E-05 11191.05 152.4 57251.42 -857675
1973 1295261 3.43E-05 14228.62 166.6 81269.67 -1213991
1974 1355768 3.81E-05 11820.33 176.8 79651.44 -1276117
1975 1503499 4.68E-05 13888.89 135.1 87728.81 -1415771
1976 1715510 3.4E-05 24253.73 188.5 155532.3 -1559978
1977 1367390 3.52E-05 25210.08 140 124079.5 -1243311
1978 446221.1 3.52E-05 5811.138 137.4 28066.03 -418155
1979 444885.5 3.35E-05 5456.654 159.7 29158.4 -415727
1980 424310 3.44E-05 29838.71 156.2 160482.2 -263828
1981 383578.8 3.34E-05 26291.08 170.7 149956 -233623
1982 445824 6.96E-05 25547.45 125.9 223890.9 -221933
1983 28912.33 7.48E-05 25996.53 105.8 205748.3 176836
1984 17219.81 4.01E-05 18202.5 176 128509.1 111289.3
1985 15938.86 3.57E-05 17000 185 112391.6 96452.76
1986 9281.142 0.001709 13175.72 128.1 2884919 2875638
1987 3594.751 0.001216 11809.43 205.9 2956769 2953175
1988 2445.341 0.001740 17812.38 177.6 5505319 5502874
1989 1885.011 0.001662 15847.86 215 5663919 5662034
1990 1442.801 0.003440 11235.78 136 5256865 5255422
1991 1165.92 0.001195 14200.25 241.4 4095601 4094435
1992 1524.905 0.001379 17885.8 258.8 6381629 6380104
1993 795.033 0.000792 14372.39 328.3 3736024 3735229
1994 714.8427 0.000803 11983.91 323.9 3117237 3116522
1995 468.7497 0.000750 3654.116 360.1 986892.1 986423.4
1996 503.2884 0.000634 1264.741 353.4 283495.1 282991.8
1997 535.7087 0.001015 1205.946 332.6 407035.9 406500.2
1998 538.4116 0.000844 687.2138 355.4 206212 205673.5
Rate of return (2d.p.) 0.08
Source: SARI Accounts, PPMED  o f  M oFA  and A uthor’s computation.
99
University of Ghana          http://ugspace.ug.edu.gh
Appendix lb: Sensitivity Analysis Assuming Adoption Rate Increases to 1% of 
Sorghum Area
YEAR ct Kt Pt Qt('000)kg Bt Bnet
1966 232309.4 -232309
1967 328878 -328878
1968 465594.8 -465595
1969 564979.5 -564979
,1970 799840.3 -799840
1971 512999.2 0.008221 7360.406 172.7 10450673 9937673
1972 914926.2 0.007743 11191.05 152.4 13205995 12291069
1973 1295261 0.007908 14228.62 166.6 18746203 17450942
1974 1355768 0.008792 11820.33 176.8 18372931 17017163
1975 1503499 0.010785 13888.89 135.1 20236111 18732612
1976 1715510 0.007847 24253.73 188.5 35876119 34160609
1977 1367390 0.008109 25210.08 140 28621008 27253618
1978 446221.1 0.008108 5811,138 137.4 6473898 6027677
1979 444885.5 0.007718 5456,654 159.7 6725871 6280986
1980 424310 0.007942 29838,71 156.2 37017903 36593593
1981 383578.8 0.007707 26291.08 170.7 34589859 34206280
1982 445824 0.016056 25547.45 125.9 51644161 51198337
1983 28912.33 0.017255 25996.53 105.8 47459272 47430360
1984 17219.81 0.009253 18202.5 176 29642776 29625556
1985 15938.86 0.008243 17000 185 25925000 25909061
1986 9281.142 0.00859 13175.72 128.1 14498566 14489285
1987 3594.751 0.007808 11809.43 205.9 18984835 18981241
1988 2445.341 0.008471 17812.38 177.6 26798728 26796282
1989 1885.011 0.008555 15847.86 215 29148173 29146288
1990 1442.801 0.011362 11235.78 136 17361532 17360089
1991 1165.92 0.004686 14200.25 241.4 16061901 16060735
1992 1524.905 0.00603 17885.8 258.8 27911678 27910153
1993 795.033 0.002731 14372.39 328.3 12886287 12885492
1994 714.8427 0.002471 11983.91 323.9 9589525 9588810
1995 468.7497 0.00254 3654.116 360.1 3342602 3342134
1996 503.2884 0.002006 1264.741 353.4 896764.3 896261.1
1997 535.7087 0.003135 1205.946 332.6 1257320 1256784
1998 538.4116 0.002626 687.2138 355.4 641445.4 640907
Rate of return (2d.p.) 1.09
Source: SARI Accounts, PPMED  o f  M oFA  and A uthor’s computation
100
University of Ghana          http://ugspace.ug.edu.gh
Appendix lc: Sensitivity Analysis Assuming Yield Increases by 10% over the Current
1.35t/ha.
YEAR c, K, P. Qt('000)kg B, Bnel
1966 232309.4 -232309
1967 328878 -328878
1968 465594.8 -465595
1969 564979.5 -564979
.1970 799840.3 -799840
1971 512999.2 4.57E+08 7360.406 172.7 55347.75 -457651
1972 914926.2 6.47E+08 11191.05 152.4 70370.42 -844556
1973 1295261 9.16E+08 14228.62 166.6 99673.31 -1195588
1974 1355768 1.11E+09 11820.33 176.8 96676.56 -1259092
1975 1503499 1.58E+09 13888.89 135.1 104636.3 -1398863
1976 1715510 1.01E+09 24253.73 188.5 190905.6 -1524604
1977 1367390 1.8E+09 25210.08 140 151788.4 -1215602
1978 446221.1 2.56E+09 5811.138 137.4 34334.13 -411887
1979 444885.5 2.68E+09 5456.654 159.7 35852.11 -409033
1980 424310 2.97E+09 29838.71 156.2 196736.3 -227574
1981 383578.8 3.39E+09 26291.08 170.7 284971.4 -98607.4
1982 445824 2.7E+09 25547.45 125.9 741049.4 295225.4
1983 28912.33 8.83E+08 25996.53 105.8 849033.6 820121.3
1984 17219.81 8.8E+08 18202.5 176 727627.6 710407.8
1985 15938.86 8.4E+08 17000 185 1025865 1009926
1986 9281.142 7.6E+08 13175.72 128.1 3509250 3499969
1987 3594.751 8.84E+08 11809.43 205.9 3631147 3627552
1988 2445.341 57333154 17812.38 177.6 6705731 6703286
1989 1885.011 34164097 15847.86 215 6892398 6890513
1990 1442.801 31638633 11235.78 136 6245232 6243789
1991 1165.92 18432347 14200.25 241.4 5379247 5378081
1992 1524.905 7142770 17885.8 258.8 8078114 8076589
1993 795.033 4861338 14372.39 328.3 5477614 5476818
1994 714.8427 3749287 11983.91 323.9 4690736 4690021
1995 468.7497 2871174 3654.116 360.1 1474081 1473612
1996 503.2884 2321346 1264.741 353.4 453142.3 452639
1997 535.7087 3037612 1205.946 332.6 577588.1 577052.4
1998 538.4116 1584501 687.2138 355.4 206212 205673.5
Rate of return (2d.p.) 0.10
Source: SARI Accounts, PPMED of MoFA and Author’s computation
101
University of Ghana          http://ugspace.ug.edu.gh
Appendix Id: Sensitivity Analysis Assuming Yield Decreases by 10% from the Current
1.35t/ha.
YEAR c, Kt Pt Qt('000)kg B, Bnet
1966 232309.4 -232309
1967 328878 -328878
1968 465594.8 -465595
1969 564979.5 -564979
■ 1970 799840,3 -799840
1971 512999.2 4.57E+08 7360.406 172.7 35265.01 -477734
1972 914926.2 6.47E+08 11191.05 152.4 44132.43 -870794
1973 1295261 9 .16E+08 14228.62 166.6 62866.02 -1232395
1974 1355768 1.11E+09 11820.33 176.8 62626.32 -1293142
1975 1503499 1.58E+09 13888.89 135.1 70821.29 -1432678
1976 1715510 1.01E+09 24253.73 188.5 120159 -1595351
1977 1367390 1.8E+09 25210.08 140 96370.65 -1271020
1978 446221.1 2.56E+09 5811.138 137.4 21797.94 -424423
1979 444885.5 2.68E+09 5456.654 159.7 22464.7 -422421
1980 424310 2.97E+09 29838.71 156.2 124228.2 -300082
1981 383578.8 3.39E+09 26291.08 170.7 178492.5 -205086
1982 445824 2.7E+09 25547.45 125.9 534115.1 88291.08
1983 28912.33 8.83E+08 25996.53 105.8 617404.5 588492.2
1984 17219.81 8.8E+08 18202.5 176 476979.2 459759.4
1985 15938.86 8.4E+08 17000 185 654075 638136 1
1986 9281.142 7.6E+08 13175.72 128.1 2260587 2251306
1987 3594.751 8.84E+08 11809.43 205.9 2282392 2278797
1988 2445.341 57333154 17812.38 177.6 4304907 4302462
1989 1885.011 34164097 15847.86 215 4435441 4433556
1990 1442.801 31638633 11235.78 136 4268498 4267056
1991 1165.92 18432347 14200.25 241.4 2811956 2810790
1992 1524.905 7142770 17885.8 258.8 4685143 4683618
1993 795.033 4861338 14372.39 328.3 1994435 1993640
1994 714.8427 3749287 11983.91 323.9 1543737 1543022
1995 468.7497 2871174 3654.116 360.1 499703.5 499234.7
1996 503.2884 2321346 1264.741 353.4 113847.8 113344.5
1997 535.7087 3037612 1205.946 332.6 236483.7 235948
1998 538.4116 1584501 687.2138 355.4 107073.8 106535.4
Rate of return (2d. p.) 0.06
Source: SARI Accounts, PPMED of MoFA and Author’s computation
102
University of Ghana          http://ugspace.ug.edu.gh
APPENDIX 2
THE DETERMINANTS OF ADOPTION, AND IMPACT OF IMPROVED 
SORGHUM VARIETIES IN THE UPPER WEST REGION, GHANA.
Questionnaire for Farmer Survey
Background: A Study of Socio-economic, Technological and Institutional Factors Affecting 
the Adoption of Improved Sorghum Varieties by Farmers in the Upper West 
Region, Ghana.
District code ........................ Interviewer's code..................................
Village cod e ........................ D a te ........................
Name of household head ...................................................................................................
Section A: Socioeconomic factors
1. Age of respondent in years.
a) 21-30 b) 31-40 c) 41-50 d) 51-60 e) above 60
2. Sex of respondent. a) male b) female
3. What is your level of education?
a) none b) non-formal c) basic d) secondary e) tertiary
4. Marital status of respondent a) single b) married c) separated d) divorced
e) widow f)widower
5. What is the total number of members of your household?.........................
6. What is the total number of household members who really work on the household 
farm?
No. of children..............  No. of women ...........  No of men.
7. What is your major occupation? a) farming b) business c) other (specify)
8 Do you have any source of non-farm income? a) yes b) no
9 If yes, name them a ) ...............  b)...................  c ) ..................  d ) ...........
10. Is any member of your household in paid employment? a) yes b) no
103
University of Ghana          http://ugspace.ug.edu.gh
11. Do any of your wives brew pito a) yes b) no
12.Do you grow sorghum? a) yes b) no
13. If yes, how long have you been growing sorghum of your own as a farmer? .......... years
14. What is the estimated size of your farm?
Plot Number Crops grown
Size of 
plot (ha)
Size under
improved
sorghum
Average yield of 
sorghum(kg/ha)
Crop codel. Sorghum 2. Millet 3. Cowpea 4. Maize 5. other
15. Do you use inorganic fertilizer? a) yes b) no
16. Do you use organic fertilizer?
17. If  yes, indicate the source.
18. How much does it cost you to undertake each of the following activities on an acre 
of sorghum?
Activity Cost (0/ha) or (0/acre) Source of labour
Ploughing
Harrowing
Sowing/Planting
Weeding
Harvesting
Threshing
Winnowing
Other (specify)
Total cost
104
University of Ghana          http://ugspace.ug.edu.gh
19. How much does it cost you to grow an acre/hectare of sorghum in terms o f  ?
Type/Input Seed Fertilizer Land Pesticide Other
Traditional
Framida
Naga white
Kadaga
Kapaala
Dorado
Other
20. Would you say have enough land for farming? a) yes b) no
21. Is it easy to obtain/acquire farmland in this area? a ) yes b) no
22. Is it easy to get hired labour when needed? a) yes b) no
Section B: Technological factors
23. Which varieties of sorghum do you grow/plant?
Variety Year l sl planted Varieties no longer grown
Traditional
Naga white
Framida
Kadaga
Kapaala
Dorado
Other
24. Why do you no longer plant ?
25.Give reasons for the varieties you grow.
Reasons Traditional Framada Naga white Kadaga Kapaala Other
Good yield
Does not 
lodge
Good for 
pito
Family food
Other
105
University of Ghana          http://ugspace.ug.edu.gh
26. Name the best local variety of sorghum you grow................
27. Name the best improved variety of sorghum you grow.......
28. Area under sorghum crop has been   over the years.
Tick one Possible reason
Decreasing
.Same
Increasing
29. The number of farmers growing improved sorghum has been  over the years.
Tick one Possible reason
Decreasing
Same
Increasing
30. In your opinion, has sorghum yields been  ?
Tick one Possible reason
Decreasing
Same
Increasing
31. In your opinion, what are the most important factors that could prevent you from 
growing improved sorghum?
32. Where do you obtain information on non-traditional sorghum varieties?
a) extension officer b) NGOs c) other farmers d) friends 
e) farmers' group
106
University of Ghana          http://ugspace.ug.edu.gh
33. What extra inputs are required for the growing of improved sorghum?
Input/ha Extra input required
Framida Kadaga Naga White Kapaala Dorado
Labour
Improved seed 
.Fertilizer 
Pesticide 
Other (specify)
34. For the period starting from 1988, please indicate the varieties grown.
Year Variety
Framida Naga white Kadaga Kapaala Dorado
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
Size of current 
acreage (ha)
35. How do you grow your sorghum? a) monocrop b) intercrop
36. What method of planting did you use for your sorghum in 1999?
a) row planting/lines b) random c) broadcasting.
37. Which will you plant in a normal year? a) Traditional b) improved
38. Which varieties give you better yield when rains are
a) average.............. b) below average.... c) above average.............
107
University of Ghana          http://ugspace.ug.edu.gh
39. How do improved sorghum varieties compare with the traditional in terms of
Characteristic
Improved variety is 
Less/same or better/ more
Framida Naga
white
Kadaga Kapaala Dorado
Disease tolerance
Pest tolerance
Drought tolerance
Varietal yield
Suitability for local dishes
Maturity period
Taste quality
Marketability
Fertilizer requirement
Storage
Weed suppression
Section C: Institutional factors
40. Where do you obtain your improved sorghum seed?
a) extension officer b) NGOs c) other farmers d) market
e) farmers' group f) select seed from own farm g) other (specify)
41. How often does the extension officer visit you as a farmer? a) once a week
b) once a month c) once a year d) never
42. Has the extension officer ever discussed your problems regarding sorghum production 
with you? a) yes b) no
43. Did he discuss with you the type of variety that would be suitable to you? 
a) yes b)no
44. Do you use inorganic fertilizer on your sorghum crop? a) yes b) no
45. If no, w hy?.................
46. Where do you obtain your fertilizer? a) extension officer b) market 
c) FASCOM d) NGOs e) Farmers' group/association
108
University of Ghana          http://ugspace.ug.edu.gh
47. How much is the cost of a bag of fertilizer? (j£/50kg bag)
48. Are agricultural inputs (improved seed, fertilizer, etc) readily available when you need 
them? a) yes b) no
49. What is the distance of the nearest sale point for agricultural inputs in your area? km
50. Is there any credit facility available to you for growing improved sorghum? a)yes b) no
51. If yes, name the source, a) bank b) credit union c)NGOs d) other specify........
52. What are the major uses to which you put your sorghum crop?
Uses % allocation
Sale
Consumption
Pito
Other
53. Is there a ready market for improved sorghum? a) Yes b) no
54. Do you get a good price for your improved sorghum? a) yes b) no
* Secondary data/information was collected from the researchers at SARI, Seed Growers 
Association, Agricultural Extension Officers and PPMED of MOFA for impact aspect of the 
study.
109
University of Ghana          http://ugspace.ug.edu.gh