Assessing Factors Affecting Adoption of Agricultural ... · Assessing Factors Affecting Adoption of...

Assessing Factors Affecting Adoption of Agricultural Technologies: The Case of Integrated Pest Management (IPM) in Kumi District, Eastern

Uganda

Jackline Bonabana-Wabbi

Thesis submitted to the faculty of the

Virginia Polytechnic Institute and State University

in partial fulfillment of the requirements for the degree of

Master of Science

in

Agricultural and Applied Economics

Daniel B. Taylor, Chair

Michael Bertelsen

Anya McGuirk

November 18, 2002

Blacksburg, Virginia

Keywords:

Integrated Pest Management, Adoption, Multivariate logit, Uganda

Copyright © 2002, Jackline Bonabana-Wabbi

Assessing Factors Affecting Adoption of Agricultural Technologies:

The Case of Integrated Pest Management (IPM) in Kumi District, Eastern Uganda

Jackline Bonabana-Wabbi

(Abstract)

Improper pesticide use on crops causes adverse effects on humans, livestock, crops

and the environment. Integrated pest management practices emphasize minimal use

of pesticides in controlling pests, and their adoption by farmers can reduce the use of

pesticides and their adverse impacts. The introduction of IPM CRSP activities in

Uganda to institutionalize IPM methods focused on priority crops in the country. This

study analyzed adoption of eight IPM technologies on cowpea, sorghum and

groundnuts. Low levels of adoption (<25%) were found with five of these technologies

while three technologies had high adoption levels (>75%). Results indicate that

farmers’ participation in on-farm trial demonstrations, accessing agricultural

knowledge through researchers, and prior participation in pest training were

associated with increased adoption of most IPM practices. Size of farmer’s land

holdings did not affect IPM adoption suggesting that IPM technologies are mostly scale

neutral, implying that IPM dissemination may take place regardless of farmer’s scale of

operation. Farmers’ perception of harmful effects of chemicals did not influence

farmers’ decisions in regard to IPM technology adoption despite their high knowledge

of this issue, suggesting that these farmers did not consider environmental and health

impacts important factors when choosing farming practices. Farmers’ managerial

capabilities were not important in explaining cowpea IPM technology adoption.

iii

Dedication

To my dad

and

my late mom

And to Bobby

iv

Acknowledgements

I would like to thank the United States Agency for International Development (USAID)

for funding this research through the Integrated Pest Management Collaborative

Research Support Program (IPM-CRSP), Grant Number LAG-G-00-93-00053-00. Many

thanks also go to the Office of International Research and Development at Virginia

Tech: its Director, Dr. S. K. De Datta, the management entity including Dr. Brhane

Gebrekidan, Dr Keith Moore, and Dr. Greg Luther for supporting this, and other

studies in Uganda. In addition, I am truly grateful to the IPM CRSP Uganda site chair

Dr. Mark Erbaugh at Ohio State University and the Uganda site coordinator Dr.

Samuel Kyamanywa for the opportunity to study at Virginia Tech.

I cannot say exactly how grateful I am to Prof Dan Taylor. His guidance in this study

was beyond measure. Dr. Taylor, I used to read acknowledgements by students you

have advised and always wondered how they could heap you so many endearments.

Now I know better. Your guidance is invaluable. Thank you also for providing me

facilities and various supplies that facilitated my comfortable study and stay at

Virginia Tech. In addition with Barbara, Alex and Claudia, I always had a family away

from home.

I would also like to extend my sincere thanks to Dr. Anya for reading through the last

draft and giving insightful comments. Many thanks go to Dr. Bertelsen for serving on

my committee and for providing valuable suggestions. Dr. Kasenge’s review of the first

three chapters was helpful in highlighting issues that would otherwise have gone un-

noticed.

Sincere thanks go to the farmers who volunteered to be interviewed. Without

sacrificing their valuable time to answer the survey questions, this study would not

have been possible. I am grateful to the field staff who assisted in collecting the data.

The District Agricultural Officer of Kumi Mr. Valdo Odeke was instrumental in making

the data collection process effective. Many thanks go to all my friends both in USA and

in Uganda. To my brothers Jacob and Josephat and sisters Joan and Dona, thank you

for encouraging me. Stella, your assistance with data entry is appreciated.

v

Lastly I would like to express my deepest gratitude to my husband, Bobby, for his love,

care and patience. Bobby, your emotional support lifted me up every day, encouraged

me and gave me a reason to always look towards my goals. For these, I cannot thank

you enough.

vi

Table of Contents

Abstract…………………………………………….………………………………………………. ….ii

Dedication………………………………………….…………………………………………………..iii

Acknowledgements………………………………..………………………………………………….iv

Table of Contents………………………………….………………………………………………….vi

List of Tables…………………………………………………………………………………………..ix

List of Figures …………………………………………………………………………………………xi CHAPTER 1 INTRODUCTION…..………………………………………………………..1 1.1 THE GENERAL PROBLEM....................................................................................................... 1 1.1.1 IPM Interventions on Cowpea, Groundnuts and Sorghum:....................................... 5 1.1.2 Rationale for IPM Interventions....................................................................................... 6 1.2 PROBLEM STATEMENT ........................................................................................................... 7 1.3 OBJECTIVES .......................................................................................................................... 10 1.3.1 General Objective ............................................................................................................. 10 1.3.2 Specific Objective ............................................................................................................. 10 1.4 HYPOTHESIS ........................................................................................................................... 10 1.5 SIGNIFICANCE OF THE STUDY ................................................................................................. 11 1.6 SUMMARY OF RESEARCH METHODS.................................................................................. 11 1.7 ORGANIZATION OF THESIS................................................................................................... 12 CHAPTER 2 LITERATURE REVIEW ........................................................................................... 13 2.1 OVERVIEW OF UGANDA ....................................................................................................... 13 2.1.1 Physical Characteristics ................................................................................................. 13 2.1.2 The Ugandan Economy ................................................................................................... 13 2.2 THE COLLABORATIVE RESEARCH SUPPORT PROGRAM (CRSP)...................................... 19 2.2.1 The IPM CRSP .................................................................................................................. 19 2.2.2 Why IPM? .......................................................................................................................... 20 2.2.3 IPM in Uganda .................................................................................................................. 22 2.3 TECHNOLOGY ADOPTION ..................................................................................................... 23 2.3.1 Measuring Adoption ........................................................................................................ 25 2.3.2 Determinants of Adoption .............................................................................................. 26 2.3.3 The Combined effect........................................................................................................ 33 2.4 SUMMARY.............................................................................................................................. 34 CHAPTER 3 METHODS ............................................................................................................... 35 3.1 THE STUDY AREA, SAMPLE, AND DATA COLLECTION TECHNIQUES ................................ 35 3.1.1 The Study Area................................................................................................................. 35 3.1.2 The Sample and Sampling Procedure .......................................................................... 37 3.1.3 Data Sources, Collection and Transformation ........................................................... 40

vii

3.2 DATA ANALYSIS TECHNIQUES AND THEIR LIMITATIONS ................................................... 43 3.2.1 Descriptive Analysis ........................................................................................................ 43 3.2.2 Crosstabs Chi-Square Tests .......................................................................................... 43 3.2.3 Discriminant Analysis:................................................................................................. 43 3.2.4 Analysis of Variance (ANOVA) .................................................................................... 43 3.2.5 Ordinary Least Squares (OLS) .................................................................................... 44 3.2.6 Correlation Analysis ..................................................................................................... 44 3.2.7 Tobit, Logit and Probit Models ...................................................................................... 44 3.3 DESCRIPTION OF CONCEPTUAL MODEL I .......................................................................... 46 3.4 EMPIRICAL MODEL I ............................................................................................................ 50 3.4.1 Explanation of Variables and Apriori Expectations................................................... 51 3.4.2 IPM Packages on Sorghum, Cowpea, and Groundnuts ............................................ 53 3.4.3 Sorghum models:............................................................................................................. 54 3.4.4 Cowpea models ................................................................................................................ 55 3.4.5 Groundnut models .......................................................................................................... 56 3.5 A TWO TIERED ANALYTICAL PROCESS ............................................................................... 56 3.5.1 Description of Conceptual Model II .............................................................................. 57 3.5.2 Empirical Model II ........................................................................................................... 58 3.6 COLLINEARITY DIAGNOSIS................................................................................................... 59 3.7 MODEL SELECTION............................................................................................................... 60 3.8 ANALYTICAL SOFTWARE ....................................................................................................... 62 3.9 SUMMARY.............................................................................................................................. 62 CHAPTER 4 RESULTS ................................................................................................................. 63 4.1 GENERAL DESCRIPTIVE ANALYSIS...................................................................................... 63 4.2 ADOPTION OF IPM PRACTICES - UNIVARIATE ANALYSIS.................................................. 66 4.2.1 Sorghum............................................................................................................................ 67 4.2.2 Cowpea .............................................................................................................................. 72 4.2.3 Groundnut ........................................................................................................................ 76 4.3 ADOPTION OF IPM PRACTICES - MULTIVARIATE ANALYSIS ............................................. 80 4.3.1 Multivariate analysis results: Sorghum....................................................................... 80 4.3.2 Multivariate analysis results: Cowpea ......................................................................... 83 4.3.3 Multivariate analysis results: Groundnut ................................................................... 86 4.4 ADOPTION OF IPM - MODEL FITTING................................................................................. 88 4.4.1 SORGHUM............................................................................................................................ 89 4.4.2 Cowpea .............................................................................................................................. 90 4.4.3 Groundnuts ...................................................................................................................... 91 4.5 TECHNOLOGY ADOPTION INDICES....................................................................................... 92 4.6 SUMMARY.............................................................................................................................. 95 CHAPTER 5 DISCUSSION AND CONCLUSIONS ......................................................................... 96 5.1 INTRODUCTION...................................................................................................................... 96 5.2 SUMMARY OF THESIS........................................................................................................... 96 5.3 SUMMARY OF METHODS: .................................................................................................... 98

viii

5.4 SUMMARY OF LEVEL OF IPM ADOPTION: .......................................................................... 98 5.5 SUMMARY OF FACTORS AFFECTING ADOPTION: ............................................................. 100 a) Economic factors: ................................................................................................................ 101 b) Social factors: ....................................................................................................................... 102 c) Management related factors: ............................................................................................. 102 d) Institutional factors: ........................................................................................................... 103 5.6 POLICY IMPLICATIONS AND CONCLUSIONS....................................................................... 103 5.7 FUTURE RESEARCH DIRECTION ....................................................................................... 107 REFERENCES ............................................................................................................................. 108 APPENDIX……………………………………………………………………………………………116 APPENDIX A: LIST OF ACRONYMS………………………………………………………………….116 APPENDIX B: MAP OF UGANDA SHOWING VEGETATION (APPENDIX B1)……………….117 MAP OF UGANDA SHOWING STUDY AREA (APPENDIX B2)…….………...118

MAP OF KUMI DISTRICT SHOWING LOCATION OF FOCAL POINTS AND SURROUNDING AREAS (APPENDIX B3)….…………………………………..…119

APPENDIX C: INTRODUCTORY LETTER……………………………………………………………120 APPENDIX D: SURVEY FORM…………………………………………………….…………………..121 APPENDIX E: SOURCES OF OFF-FARM INCOME (E1)……………………………………….…133 SORGHUM VARIETIES GROWN IN KUMI DISTRICT (E2)………………...…133 WEED SPECIES IN SORGHUM IN THE STUDY AREA (E3)……………...….134 COLLINEARITY DIAGNOSTIC RESULTS (E4)…………………………………..134 VITA……………………………………………………………………………………….…………..135

ix

List of Tables

TABLE 1.1: STATUS OF POPULATION AND FOOD AVAILABILITY IN DEVELOPING COUNTRIES .....................2

TABLE 2.1: CULTIVATED AREA OF MAJOR FOOD CROPS IN UGANDA ...................................................17

TABLE 3.1: SUMMARY OF SOCIAL, DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS OF KUMI DISTRICT

FARMING SYSTEM ....................................................................................................................37

TABLE 3.2: DESCRIPTION OF VARIABLES USED IN THE ANALYSES.........................................................42

TABLE 4.1: SUMMARY STATISTICS OF CONTINUOUS VARIABLES...........................................................65

TABLE 4.2: SUMMARY STATISTICS OF NON-CONTINUOUS VARIABLES ...................................................65

TABLE 4.3: CROPS GROWN: SUMMARY STATISTICS.............................................................................66

TABLE 4.4: PEST INCIDENCE ON SORGHUM ........................................................................................68

TABLE 4.5: CHARACTERISTICS OF FERTILIZER ADOPTERS AND NON-ADOPTERS IN SORGHUM

PRODUCTION – CONTINUOUS VARIABLES...................................................................................69

TABLE 4.6: CHARACTERISTICS OF FERTILIZER ADOPTERS AND NON-ADOPTERS IN SORGHUM

PRODUCTION – NON-CONTINUOUS VARIABLES ...........................................................................69

TABLE 4.7: STRIGA ADOPTERS VERSUS NON- ADOPTERS – CONTINUOUS VARIABLES ..............................70

TABLE 4.8: STRIGA ADOPTERS VERSUS NON- ADOPTERS – COMPARISON WITH NON-CONTINUOUS

VARIABLES...............................................................................................................................70

TABLE 4.9: CHARACTERISTICS OF CROP ROTATORS AND NON-CROP ROTATORS – CONTINUOUS VARIABLES

................................................................................................................................................71

TABLE 4.10: CHARACTERISTICS OF CROP ROTATORS AND NON-CROP ROTATORS – NON-CONTINUOUS

VARIABLES...............................................................................................................................71

TABLE 4.11: DISTRIBUTION OF COWPEA VARIETIES IN STUDY AREA.....................................................73

TABLE 4.12: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF TIMELY PLANTING FOR COWPEA

PRODUCTION – CONTINUOUS VARIABLES....................................................................................73

TABLE 4.13: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF TIMELY PLANTING FOR COWPEA

PRODUCTION –NON-CONTINUOUS VARIABLES.............................................................................74

TABLE 4.14: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF INTERCROPPING WITH CEREALS

FOR COWPEA PRODUCTION – CONTINUOUS VARIABLES................................................................74

TABLE 4.15: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF INTERCROPPING WITH CEREALS

FOR COWPEA PRODUCTION – CATEGORICAL VARIABLES ..............................................................75

TABLE 4.16: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF IMPROVED VARIETY FOR COWPEA

PRODUCTION – CONTINUOUS VARIABLES....................................................................................75

x

TABLE 4.17: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF IMPROVED VARIETY FOR COWPEA

PRODUCTION– NON-CONTINUOUS VARIABLES.............................................................................76

TABLE 4.18: GROUNDNUT VARIETIES, AND THEIR PERFORMANCE IN THE STUDY AREA..........................77

TABLE 4.19: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF CLOSE SPACING IN GROUNDNUT

PRODUCTION - NON-CONTINUOUS VARIABLES ............................................................................77

TABLE 4.20: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF CLOSE SPACING IN GROUNDNUT

PRODUCTION - NON-CONTINUOUS VARIABLES.............................................................................78

TABLE 4.21: CHARACTERISTICS OF ADOPTERS AND NON-ADOPTERS OF IMPROVED VARIETY IN

GROUNDNUT PRODUCTION – CONTINUOUS VARIABLES...............................................................78

TABLE 4.22: COMPARISON OF PEST OCCURRENCE AND CONTROL EFFORTS AMONG SORGHUM, COWPEA

AND GROUNDNUT CROPS...........................................................................................................79

TABLE 4.23A. MAXIMUM LIKELIHOOD ESTIMATES FOR FERTILIZER (FTIS) ADOPTION MODEL..............81

TABLE 4.23B. MAXIMUM LIKELIHOOD ESTIMATES FOR ECAT (CELOSIA) ADOPTION MODEL................82

TABLE 4.23C. MAXIMUM LIKELIHOOD ESTIMATES FOR ROTN (CROP ROTATION) ADOPTION MODEL....83

TABLE 4.24A. MAXIMUM LIKELIHOOD ESTIMATES FOR TPCP (TIMELY PLANTING) ADOPTION MODEL..84

TABLE 4.24B. MAXIMUM LIKELIHOOD ESTIMATES FOR ICCP (INTERCROPPING) ADOPTION MODEL ......85

TABLE 4.24C. MAXIMUM LIKELIHOOD ESTIMATES FOR ICPV (INTERCROPPING) ADOPTION MODEL ......86

TABLE 4.25A. MAXIMUM LIKELIHOOD ESTIMATES FOR CLSP (CLOSE SPACING) ADOPTION MODEL ......87

TABLE 4.25B. MAXIMUM LIKELIHOOD ESTIMATES FOR IGNV (IGOLA) ADOPTION MODEL....................88

TABLE 4.26: MAXIMUM LIKELIHOOD ESTIMATES FOR THE FITTED SORGHUM IPM ADOPTION MODELS ..89

TABLE 4.27: SUMMARY GOODNESS-OF-FIT TESTS FOR SORGHUM MODELS ............................................89

TABLE 4.28. MAXIMUM LIKELIHOOD ESTIMATES FOR THE FITTED COWPEA IPM ADOPTION MODELS....90

TABLE 4.29: SUMMARY GOODNESS-OF-FIT TESTS FOR COWPEA MODELS ..............................................90

TABLE 4.30: MAXIMUM LIKELIHOOD ESTIMATES FOR THE GROUNDNUT IPM ADOPTION MODELS .........91

TABLE 4.31: SUMMARY GOODNESS-OF-FIT TESTS FOR GROUNDNUT MODELS........................................91

TABLE 4.32: DISTRIBUTION OF TECHNOLOGIES ...................................................................................92

TABLE 4.33: CUMULATIVE LOGIT MODEL ESTIMATES FOR ADOPTION OF ‘ONETECH’ AND ‘TWOTECH’

SORGHUM TECHNOLOGIES ........................................................................................................92

TABLE 4.34: CUMULATIVE LOGIT MODEL ESTIMATES FOR ADOPTION OF ‘ONETECH’ ‘TWOTECH’ AND

‘THREETECH’ COWPEA TECHNOLOGIES ..................................................................................93

TABLE 4.35: CUMULATIVE LOGIT MODEL ESTIMATES FOR ADOPTION OF ‘ONETECH’ AND TWOTECH

GROUNDNUT TECHNOLOGIES ....................................................................................................94

xi

List of Figures FIGURE 2.1: THE ADOPTION CURVE ............................................................................25

FIGURE 3.1: RESPONDENT SELECTION..........................................................................39

FIGURE 3.2: LOGISTIC REGRESSION CURVE FOR [0,1] RESPONSE MODELS .........................49

FIGURE 3.3: COMPONENTS OF THE IPM PACKAGES ON COWPEA, SORGHUM, AND

GROUNDNUTS. ...................................................................................................54

FIGURE 3.4: MODEL BUILDING PROCEDURE ..................................................................61

FIGURE 4.1: FARM INPUT ACQUISITION: DISTRIBUTION OF PURCHASE DECISIONS ................67

FIGURE 4.2: DISTRIBUTION OF SORGHUM VARIETIES AS A PERCENT OF TOTAL SORGHUM

ACREAGE FOR THE SAMPLE .................................................................................67

FIGURE 4.3: REASONS FOR COWPEA DEFOLIATION .........................................................72

FIGURE 4.4: PEST OCCURRENCE AND CONTROL .............................................................79

FIGURE 5.1: LEVELS OF IPM ADOPTION……………………………………………..…………...98

FIGURE 5.2: FACTORS AFFECTING IPM TECHNOLOGY ADOPTION…………….…………...….100

Chapter 1 Introduction

1.1 The General Problem

In the past several years following the advent of the green revolution, concerted efforts to

raise food production resulted in substantial increments in global food output. The

distribution of the increase was heavily skewed towards the more developed nations while

other regions of the globe realized less than impressive increments. Food output in Africa

lags behind the rest of the world’s production levels. In the last decade, the continent’s

share of world food production was a meager 3.9%. By comparison, Asia, North America

and Europe produced 47.7%, 14.8% and 12.2% respectively (Oerke, et al., 1994). By 1990,

Africa’s population was 615 million and was projected to increase to 813 million by the

end of 2002 (FAOSTAT, 2002), a 32% population increase in just over a decade. Moreover,

even within Africa, there are variations in these trends with some countries exhibiting

higher population growth with low agricultural development. Sub-Saharan Africa’s

agricultural performance has been variably called the world’s foremost global challenge

(United Nations, 1997) and as “still very far behind” the rest of Africa (Odulaja and Kiros,

1996 p.86). Moreover, the region’s population is increasing, and is expected to account for

30% of the underdeveloped world by the year 2010 (Table1.1).

Low food production and high population growth rates inevitably lead to problems of per

capita consumption. Not surprising, the world’s most hungry people also live in the Sub-

Saharan region of the continent (von Braun, Teklu and Webb, 1999; Wilson, 2001).

According to FAO (The Food and Agricultural Organization), Sub-Saharan Africa is

expected to have 264 million chronically1 undernourished people by the year 2010 (FAO,

1996). Several demographers have studied the situation and hypothesized numerous ways

of avoiding the ‘Malthusian trap’ that is likely to envelop the continent. No wonder world

organizations such as FAO, the World Bank, and IFPRI (International Food Policy and

Research Institute) have defined their core objective towards increasing food output and

improving the quality of life for the rural poor on the continent.

IFPRI suggests that the supply of food will need to rise by around 70% by the year 2020 if

the 6.5 billion people who are expected to be living in developing countries, including

Uganda, are going to be food secure (Leisinger, 1996). With only 18 years to this deadline,

1 Chronically undernourished people are defined by FAO as those whose estimated annual food energy intake falls below that required to maintain body weight and support light activity.

2

food production has remained stagnant, or declined in most of Sub-Saharan Africa (von

Braun, Teklu and Webb, 1999; McCalla, 1999). IFPRI2 already realizes that food problems

in Sub-Saharan Africa will persist well beyond 2025 (McCalla, 1999).

Table 1.1 Status of population and food availability in developing countries Number of chronically undernourished people

(Millions)

Share of region’s Population (Percent)

Share of total undernourished

Population (Percent)

Region 1990-92 2010 1990-92 2010 1990-92 2010 East Asia 268 123 16 6 32 18 South Asia 255 200 22 12 30 29 Sub-Saharan Africa 215 264 43 30 26 39 Latin America and the Caribbean

64 40 15 7 8 6

Middle East and N. Africa

37 53 12 10 4 8

Total 839 680 21 12 100 100 Source: FAO (1996)

The goal of increasing food production is both externally and internally challenged by

various factors. External factors such as natural calamities like droughts and floods are

well beyond the control of the local subsistence farmer. Other broad external factors

include poor farming technologies and bad government policies. Internal factors include

pests, soil infertility, land availability and population increase with a subsequent rise in

food demand. Although these broad external and internal factors may not be directly

controllable, they can be influenced by human behavior. While the increase in population

will exacerbate, rather than improve the food availability situation (Wilson, 2001) -

especially if the population is malnourished - this study confines itself to perhaps what is

considered the most limiting factor to food production increase, that is insect pests and

diseases.

Africa’s overall crop loss due to pests stands at an astonishing 96.2% of its production

(Oerke et al., 1994). In Uganda, although literature does not provide quantitative losses, it

is estimated that crop losses due to pests are larger than those causes by drought, soil

infertility, or poor planting material (Kyamanywa, 1996). As a result, addressing the effects

of pests on Uganda’s agricultural production captures a lot of attention from both local

and international bodies. Not surprisingly, a number of agricultural research efforts are

2 Appendix A contains a list of acronyms used in this thesis.

3

currently underway aimed at reversing the trend of pest damage to Uganda’s agricultural

produce.3 As noted above, Uganda’s current 4.8% annual increase in crop production is

perhaps most attributable to the numerous agricultural research activities (Odulaja and

Kiros, 1996; United Nations, 1997) in the country, many of which encompass pest control

programs.

The aforementioned agricultural research is mainly supported through governmental, non-

governmental, private and other funding sources. Moreover, this research requires

sustained investment of resources. The success or failure of this research inevitably plays

a pivotal role in the continued investment in such programs in the country. Needless to

say, prioritization of funding for these research programs is necessary due to the limited

availability of scarce resources and competing uses of these resources for other

investments. As Alston, Norton and Pardley (1995) correctly state, research administrators

are increasingly facing sharper pressure to justify budgets and prioritize programs.

Probably the most important determinant of the effectiveness of such programs is the level

of adoption of innovations that these programs generate, and on their profitability

(Griliches, 1957; Caswell et. al, 2001). In addition, the faster the research can be

completed, the higher the turnover of benefits. Moreover, the more evident research results

are, the easier it is to justify the implementation of, and continued investment in research

programs. A common problem for many individuals and organizations is how to speed up

the rate of diffusion of a research program’s innovations (Rogers, 1995). Yet, speeding up

the rate of adoption of technologies requires knowledge of the underlying factors that

influence adoption decisions. It is therefore not unexpected that economists and others

conduct studies to determine these factors.

Rogers (1995) demonstrates that adoption of technologies depends on their characteristics:

compatibility with the existing values and norms, complexity, observability, trialability,

and relative advantage. This definition pertains to technologies in a variety of disciplines,

and may be as relevant in other fields as it is in agricultural related technologies. In dairy

production in 5 states in the US for instance, El-Osta and Morehart (1999) identify age of

operator, size of operation and specialization as important factors in increasing likelihood 3 This is in spite of IFPRI’s recent statement that international and national support for agricultural research is eroding due to perceptions of agriculture as a major source of environmental pollution (IFPRI, 2001) implying that those who fund research may shift their emphasis from agricultural research to natural resource management (Hassan, 1998; Wilson, 2001).

4

of technology adoption, while research by Caswell, et al., (2001) ascertains that high levels

of farm operator education are likely to induce adoption of management technologies.

Others say lack of adequate inputs and active information4 (Feder and Slade, 1984) may

be obstacles to adoption. These studies pertain to technologies in the developing countries

but could apply to less developed countries.

In developing countries, studies related to Integrated Pest Management (IPM) have not

been as prevalent as in developed countries. This realization has led to a number of recent

studies on IPM being done in the Philippines (Tjornhom, 1995), Jamaica (Ogrodowczyk,

1999; Patterson, 1996), Ecuador (Yamagiwa, 1998) and other developing nations. The

Ecuadorian study identifies over-valuation of the local currency for pesticide importers,

lowering the cost of pesticides, subsidized credit to farmers and exemption from sales

taxes as policies that encourage pesticide use and are thereby limiting adoption of pest

control alternatives such as IPM. The Ecuadorian study is in agreement with the

Philippines study which found that among others, the lower the cost of pesticides, the

more likely it is for farmers to use pesticides instead of IPM technologies.

No such study has however been done for adoption of IPM technologies in Uganda. IPM is

a set of technologies that aims at reducing pest damage to crops while emphasizing non-

chemical pest control methods. In Uganda, the rejuvenation of IPM activities through the

Integrated Pest Management Collaborative Research Support Program (IPM CRSP)5 efforts

was welcomed with much enthusiasm with expected benefits including higher yields,

reduced pests and reduced expenditures on pesticides. At the time, it was anticipated that

pioneering participating farmers would act as role models and other farmers would adopt

the practices thereafter. The IPM practices were thus expected to diffuse beyond the

original area of operation.

However, several years after its introduction, the activities of the IPM CRSP program have

not been evaluated in terms of adoption. According to a recent study evaluating farmer

knowledge and awareness of IPM in Uganda, IPM CRSP program involvement and

exposure by potential adopters was thought to explain the trend of adoption of targeted

4 Active information is that obtained purposively. Unlike passive information, active information involves costs to the information seeker in terms of time, cash or both. 5 More about the IPM CRSP is explained in later sections.

5

practices (Erbaugh et al., 2001). However, no attempt to quantify these assertions was

made.

Since the introduction of IPM CRSP research in Uganda, researchers have developed

several pest control strategies for important crops including cereals, legumes, vegetables

and other horticultural crops. In Uganda’s agricultural production arena, cereal and

legume crops are of major importance. In Eastern Uganda, three such crops include

sorghum (Sorghum bicolor), groundnut (Arachis hypogaea) and cowpea (Vigna unguiculata).

Their priority status is due to both acreage planted and their nutrition content.

Groundnuts and Cowpea are the second and third most important legume food crops in

Uganda after beans (IPM CRSP Annual Report, 2001), Sorghum is among the most

important cereal crops ranked third to maize and millet in providing for the carbohydrate

needs of the Ugandan diet (FAOSTAT, 2002). Current statistics (FAOSTAT, 2002) estimate

64,000ha, 208,000ha and 282,000ha as national acreage for cowpea, groundnuts and

sorghum while production levels are estimated at 64,000 Mt, 146,000 Mt and 423,000 Mt

respectively. These crops are often said to be food-security crops because of their drought

resistance. However, these crops are not without their share of problems. Pest attacks on

these crops calls for control strategies. A brief review of pest control activities and the

rationale for IPM intervention on the three crops is given below.

1.1.1 IPM Interventions on Cowpea, Groundnuts and Sorghum: The IPM CRSP field monitoring of 1996 revealed high pest levels on cowpea in eastern

Uganda. Major insect pests on cowpea identified included blister beetles (Epicauta spp.),

aphids (A. craccivora Koch), pod-borers (Maruca testularis) and thrips (Megalurothrips

sjostedti ) and leafhoppers (IPM CRSP Annual Report, 1996). Aphids (A. craccivora) cause

damage by sucking plant sap and damage pods by forming honeydew deposits. The pod-

borer (M. testularis) is reportedly worst during the rainy season. In Kumi district, pest

damage contributes to 24-48% of the total variation in cowpea grain yield with thrips (M.

sjostedti ) accounting for the greatest damage (Karungi et al., 1999). In addition, cowpea is

one of the crops that are consistently sprayed by farmers probably because pesticide

application has a significant effect on limiting the severity of diseases (Adipala et al, 1999).

By 1995, 92% of cowpea farmers in Kumi were using insecticides as their main pest

control strategy (IPM CRSP Annual Report, 1996). And, as an IPM practice, farmers are

increasingly planting with the first sign of rains to enable the cowpea crop to escape

damaging populations of certain pests by harvesting before peak pest populations.

6

Striga is considered a major pest of Sorghum (Parker, 1980) and was found to be the most

serious weed affecting sorghum yields in Uganda. In Kumi district, the parasitic weed has

a widespread distribution. Ninety seven percent of sampled farmers involved in the 1996

IPM CRSP participatory assessment were able to identify it (Erbaugh et al, 2001). The

gravity of the Striga problem is thought to stem from the fact that the seed evolved in such

a way that it only germinates naturally when in the vicinity of a sorghum root (Parker,

1980). Furthermore, the seed is very small and can persist for many years in the soil.

Two diseases are of major consequence to groundnut production in Kumi namely

groundnut rosette (GRV) and cercospora leafspot (Cercospora arachidicola) which

frequently lead to total crop failures. Major groundnut pests include aphids (A. craccivora),

thrips (M. sjostedti ) and leaf miners (Aroarema modeicella) (IPM CRSP Annual Report,

2001).

1.1.2 Rationale for IPM Interventions

Disease and insect infestation on many crops occurs simultaneously. Sorghum,

groundnuts and cowpeas are no exceptions. Therefore controlling insects and diseases

simultaneously necessarily calls for an integrated approach, which IPM packages address.

For cowpea, a number of studies revealed that cowpea production could be improved and

increased through well-defined IPM systems (Isubikalu, Erbaugh and Semana, 1997;

Jackai et al., 1985). Among the most promising technologies developed by IITA are

varieties resistant to Striga,6 aphids (A. craccivora Koch), and bruchids (Callosobruchus

maculates), improved storage techniques using solar drying, and the use of botanical

pesticides in the field and in storage (CGIAR, 2002). Current IPM CRSP practices

disseminated to farmers in Uganda for control of Cowpea insect pests have included close

spacing, and strategic insecticide application. In addition, well-timed defoliation, and

intercropping with Sorghum are encouraged.

In regard to Sorghum, although farmers are generally less likely to use pesticides on

cereals, heavy weed infestations cause considerable crop loss and therefore provide an

incentive for weed control. It is suggested in literature (Parker, 1980 and IPM CRSP

6 In West Africa, Striga is a problem on cowpea (CGIAR, 2002)

7

Annual Report, 1999) that crop rotation is one of the most potent methods for reducing

striga. Planting a more rapidly growing cultivar that suppresses and shades weeds,

growing resistant varieties and good irrigation, in addition to high nitrogen levels are

thought to reduce the weed. However, because of the nature of the weed, several studies

have demonstrated that no purely cultural control system is fully effective – thereby calling

for an integrated approach. IPM CRSP measures on striga in Uganda include intercropping

Sorghum and silver leaf desmodium, a legume that suppresses striga weeds; planting

resistant genotypes (such as Sekedo); crop rotations (Cotton/Sorghum/Cowpea); and

recommended fertilizer application (40-8-8 kg of NPK/ha). Other measures include seed

coating with herbicide, two weedings and other cultural practices involving modified

planting dates and crop management practices (IPM CRSP Annual Reports, 1998-2000).

On Groundnuts, practices developed by researchers include: early planting, manipulation

of plant density, planting a resistant variety, and minimum spray schedule of 2-3

Dimethoate or 1-2 sprays of Dimethoate and Dithane M45. The crop is also often

intercropped with maize as a control strategy (IPM CRSP Annual Reports, 1998-2000).

IPM CRSP researchers hypothesize that these alternative methods (IPM activities) can be

disseminated to more farmers through establishing field schools and through interacting

with other partners such as non-governmental organizations (IPM CRSP Annual Report,

1999). The effectiveness of this dissemination approach however, greatly depends on how

farmers perceive IPM CRSP activities. Moreover, these alternative methods require farmers

to abandon age-old methods involving the use of conventional non-farm inputs including

reducing their dependence on ‘reliable’ pesticides for control.

1.2 Problem Statement

In many countries, including Uganda, non-farm chemical inputs play a large role in

agricultural production, especially because of the need to increase production.

Unfortunately the use of some of these inputs is associated with degradation of the

environment, and health of living organisms, including humans. Mitigating the effects of

these “necessary evils” therefore became a focus for many research programs. Alternative

methods of production that reduce negative effects of chemicals and yet maintain at least

the same level of production are continuously sought. Alternatives such as cultural

8

methods, organic, and biological control methods are increasingly emphasized to improve

land productivity and control of pests.

One such alternative is Integrated Pest Management. As mentioned earlier, the IPM

approach emphasizes the use of non-chemical inputs and judicious use of chemical inputs

in production to reduce pest incidence on crops, thereby increasing farmers’ yields and

returns. This approach is recommended globally for increasing agricultural production

without upsetting the balance of nature while controlling pests. Although some literature

indicates uncertainty of IPM profitability (Abara and Singh, 1993), or profitability of some,

but not all parts of the total package (Smith, Wetzstein and Douce, 1987), several studies

demonstrate that benefits accrue from IPM. These include its effect on reducing pesticide

residue on crops, lessening the negative impacts of pesticides on the environment and

humans, lowering production costs, and increased pest management effectiveness.

A linear programming model developed in 1982 on a national level indicated that

widespread adoption of farming practices without the use of pesticides (and fertilizers)

would increase net farm incomes in the US (Olson, Langley and Heady, 1982). In an

evaluation of pest management characteristics Smith, Wetzstein and Douce (1987) showed

that different characteristics of pest management affected net benefits in Georgia, USA.

They specifically found that proper spraying7 and using beneficial insects significantly

increased net returns. In Virginia, Mullen, Norton and Reaves (1997) quantified annual

environmental returns of approximately $844,000 from implementation of the Virginia

peanut IPM program, while on Jamaican vegetable crops, IPM led to increase in profits on

all the three crops studied by Ogrodowczyk (1999). Furthermore, because of the high

potential IPM has in the Near East8, IPM implementation was stated as a necessary

requirement to improve crop protection in vegetable cultivation (Alebeek and Lenteren,

1992). In addition, in a study on the environmental and economic consequences of IPM in

Viticulture (Fernandez-Cornejo, 1996), IPM adoption was found to positively affect both

yields and profits in grape production.

7 They define ‘proper spraying’ as (number of sprays after the threshold - number of improper sprays)/total number of sprays. 8 Countries in the region referred to as ‘The Near East’ include Turkey, Jordan, Egypt, Tunisia and Morocco

9

Similar studies in Uganda show potential benefits from IPM adoption. Bashaasha et al.,

(2000) establish benefits ranging between Shs 101,378 and Shs 255,9089 from adopting

IPM systems in the control of striga in sorghum fields in Kumi district. In another study

assessing IPM systems in Groundnuts, Bonabana et al., (2001) established a Marginal

Rate of Return of 870% in adopting a disease resistant variety as an IPM strategy for

control of major groundnut insect pests in the same district. These net benefits translate

into profits for farmers. Therefore IPM has been demonstrated to be potentially profitable

and in such cases society can benefit from its adoption.

Economic theory suggests that practices proved to be profitable are likely to be adopted by

producers. Yet according to Giliomee (1994), IPM, a profitable venture, has not been widely

adopted. For instance, only 4% of all US farms are said to practice ‘true IPM’ (Ehler and

Bottrell, 2000). This pattern is also found in small-scale farming communities in Jamaica

(Patterson, 1996). In Uganda, only a few farmers use complete IPM packages (Kyamanywa,

1996). Moreover, extent and level of IPM use in Uganda is still largely unknown. As such,

several questions arise: What is the current level of adoption of IPM? How can adoption be

accelerated? What factors influence IPM adoption?

There is a general lack of understanding of the factors affecting the adoption of IPM

technologies in farming systems in Uganda. Moreover, as noted above, although economic

analyses show potential benefits, no attempt has been made to ascertain reasons for the

observed levels of adoption. Only with a thorough understanding of these factors can

further insight be developed concerning strategies to promote IPM.

Most of those who attempt to explain the adoption of IPM in Uganda base their assertions

on subjective beliefs about the conventional practices of smallholder farmers, and not on

analytical evidence. Therefore, an empirical description regarding factors affecting

adoption is necessary. Several underlying factors may be the cause of the observed level of

adoption. For example a complex set of interactions or conditions involving the technology

(IPM), the institution (administration), the potential/targeted adopter (the farmer) or the

general setting in which the technology is introduced may affect adoption. As Diebel,

Taylor and Batie (1993) state, these factors may either be barriers or enhancers of

adoption. It is therefore imperative to study these conditions – farmer’s social

9 1US$=1760UGShs (May 2001)

10

characteristics, economic setting, institutional factors and managerial aspects to identify

the conditions that are affecting IPM adoption.

1.3 Objectives

1.3.1 General Objective

This study’s objective is to establish whether social, economic, management and

institutional factors that affect adoption of IPM technologies on three major crops in Kumi,

Eastern Uganda.

1.3.2 Specific Objective

The specific objectives of the study include the following:

(i) To establish factors that affect adoption of IPM practices specific on cowpea,

groundnuts and sorghum.

(ii) To estimate the relative contribution of each factor in affecting adoption, thereby

establishing the factors that have the greatest impact on technology adoption.

(iii) To establish the level of adoption of eight IPM technologies in Kumi

Achieving the above objectives will be a major step towards designing a system that can

encourage adoption in the study area and up-scaling the adoption pattern to other

geographical areas with similar agro-ecological characteristics.

1.4 Hypothesis

The following hypotheses will be tested:

(i) Cost of a technology negatively influences its adoption while per capita farm income

positively influences technology adoption.

(ii) Farm size and education level of farmers positively influence technology adoption.

(iii) Adoption is negatively influenced by length of farming experience, farmer’s age and

household labor.

(iv) There is no significant difference in IPM adoption between men and women.

11

1.5 Significance of the study

By pointing out the factors that influence IPM technology adoption, this study will provide

guidance to the IPM administrators and researchers for enhancing the program’s

effectiveness. The added knowledge on which factors have the greatest influence on IPM

adoption will help administrators make more informed decisions on how to promote IPM

adoption.

Another benefit from the research will be provision of an explanation of the current state of

technologies used by farmers. Moreover, since IPM involves a variety of practices that are

specific to individual crops, measuring its adoption on various crops may provide a strong

case for increasing investment in various IPM research.

Also because of the importance of cowpea, sorghum and groundnut in the Eastern region,

it is envisioned that technological spillovers are likely outside of this study area. IPM

adoption on these three crops outside the study area could be projected.

In addition, this study will provide a basis for gauging how policy changes may affect

farmers. Policy issues that constrain or enhance the provision of inputs that are required

to carry out IPM practices have a direct effect on how IPM farmers react to them. The

results will provide useful information to enhance the success of the IPM CRSP project,

and indeed any other related program that attempts to introduce practices for adoption in

settings that are similar to those in this study area. Results of this study will thus have

implications well beyond the confines of the study area.

Finally, in Uganda the IPM CRSP is an externally funded project whose continued support

is dependent on the effectiveness of the program. Therefore, for continued funding, the

IPM CRSP must demonstrate benefits. Yet these benefits do not accrue if farmers do not

adopt the practices. A crucial step therefore seems to be to identify the forces that enhance

IPM adoption. This thesis aims to fulfill this important task.

1.6 Summary of Research Methods

Survey data were collected from a random sample of farmers in the study area. Using

statistical methods yield differences between adopters and non-adopters are obtained. A

12

multivariate logit analysis identifies factors and their relative importance in explaining

adoption of eight IPM technologies.

1.7 Organization of Thesis

The thesis is organized into five main chapters. Chapter 1 has presented an introduction

to the problem - the main thrust of the study, a delineation of underlying assumptions and

objectives of the study. Chapter 2 addresses the general theory and description of the

agricultural system in Uganda. Chapter 3 provides the methods of data collection and data

sources; a description of the study area, the sampling and analysis techniques, and

develops a conceptual framework used to analyze the empirical data, while Chapter 4

comprises the empirical results of the study and discussion. The final chapter gives a

summary, policy implications and conclusions of the study.

13

Chapter 2 Literature Review

This chapter describes Uganda’s economy with major emphasis on the agriculture sector

and IPM activities in the country. It also examines relevant literature on technology

adoption: trends, the process of adoption, measurement of adoption, and factors affecting

adoption.

2.1 Overview of Uganda

2.1.1 Physical Characteristics Uganda is a landlocked country located between –1.450N 29.940E and 4.250N 35.010E in

the Great Lakes region in Sub-Saharan Africa. The country has substantial natural

resources, including fertile soils, regular rainfall, and sizeable mineral deposits of copper

and cobalt. It has five major lakes, two major rivers and several seasonal and non-

seasonal rivers, swamps and wetlands providing a variety of fish species all year round.

The natural vegetation is mainly savanna grassland, woodland, bush land and tropical

high forest. See Appendix B1 for a map of geographical features of Uganda. Uganda has a

tropical climate, with two rainy seasons from December to February, and from June to

August, providing two crop-growing seasons. There are slight variations in rainfall from

one region to another. Over the last several years, the mean annual rainfall was 750mm in

the North East and 1,500mm in the high rainfall areas of the shores of Lake Victoria,

around the highlands in the East and southwestern region. Average temperatures of 210C

(70F) have prevailed in the last decade. The tropical climate with fertile soils, regular

rainfall and favorable temperatures enable production of a diversity of crops and livestock

(EIU, 2001).

2.1.2 The Ugandan Economy Uganda has had wide fluctuations in its economic performance. In the late 1960s political

instability caused by a dictatorial government and state-run intervention in almost all

sectors of the economy destroyed the country’s physical infrastructure and many

economic and social amenities. By the early 1980’s, Uganda had become one of the

poorest countries in the world. At that time, education and health systems broke down,

14

human indices10 were poor, and the civil service had been destroyed by low wages and

poor morale. Consequently, real GDP per capita was at its lowest (EIU, 2001).

Around the mid 1980s a new government came into power. This government, with the help

of donors – mainly the IMF and World Bank embarked on an economic recovery program

aimed at reducing poverty by rehabilitating infrastructure (economic, social and

institutional). Further, the recovery program encompassed civil service reform, revised

investment and incentive structures, and made a rapid move to a market-determined

exchange rate thereby giving the country a robust economic performance (EIU, 2001).

From the mid 1990s, an economic downturn set in again. Prior to this period, specifically

around 1994, Uganda had attained a GDP growth of 11.5%. By the end of 1996 real GDP

growth had dropped to 7%. Although this was one of the highest in the Sub-Saharan

region at the time (UN, 1997), it still represents a significant drop. This downturn has

continued to the present. In 1999 GDP was estimated at US $6.3 billion but by 2000, it

had dropped to US $5.9 billion and to US $5.7 billion by the end of 2001 (World Bank,

2002). Export of goods has been declining from US $639.2 million in 1996 to US $390.8

million in 2000, and remaining constant in 2001 (World Bank, 2002). The significant

decline in the 2000 exports was attributed to a drought in the third quarter of 2000 that

resulted in a low level of coffee production (from 236,200 tons in 1999 to 186,000 tons in

2000). Considering that coffee accounts for over 40% of the country’s export receipts, the

poor harvest severely depressed export receipts. In addition, coffee prices are falling. This

decline in coffee production is attributed to excess production (internationally) that

currently outstrips demand. The EIU (2001) estimated Uganda’s earnings from coffee fell

from US $162 million in 2000 to US $149 million in 2001.

In 2002 coffee export earnings continue to fall due to depressed prices. However,

predictions of a more robust economic performance indicate that real GDP growth may

increase to 6.6% in 2002 due to increased output in other agricultural activities: cotton,

tobacco, fish, maize and flowers. In addition non-agricultural sector development, which is

largely funded by donor inflows, is increasingly supporting growth (EIU, 2001).

10 Human indices include infant mortality rate, maternal mortality rates, birth weight, average life expectancy, literacy rates and percentages below/above poverty line.

15

Uganda’s terms of trade continue to deteriorate. In addition, the total outstanding debt is

increasing. While the total debt was estimated at US $3.48 billion in 1999 (World Bank,

2000), it had reached US $3.7 billion by 2000 (EIU, 2001). It is probably because of these

deteriorating terms of trade that Uganda was the first country to benefit from the Heavily

Indebted Poor Countries (HIPC) Initiative of the World Bank given that it was faced with a

serious debt problem (World Bank, 2000). Under the initiative, the country would be able

to reduce its external debt by 20% of the net present value (Uganda, 1998a) and to redirect

resources to priority poverty reduction efforts. HIPC relief in 2002 is expected to reduce

debt-service payments to US $106 million (EIU, 2001). Moreover, reports (Uganda, 1998a)

show that Uganda’s poverty reduction strategy is very effective, which has prompted the

World Bank to shift emphasis from project-based assistance to direct budgetary support

for the government’s poverty eradication plan.

Uganda’s current (mid-2002) population of 23.9 million (US Bureau of the Census, 2002)

is increasing, expected to reach 28 million by 2010. In 1991, the total population was 16.7

million and by 1999, it had reached 21.5 million. Between the period 1969 and 1980,

Uganda’s population grew at an average rate of 2.7% per annum and was expected to grow

even faster. However, the period 1980-1991 saw a decline in the growth rate to 2.5% per

annum. The current annual population growth rate of 2.9% partially outstrips the labor

force, which is lagging behind at a 2.7% annual increase. If this trend continues, it might

suggest an increase on stress caused by dependant populations on the working

population. However, Uganda’s urban population (12.5% in 1995) as a percentage of the

total population was about the lowest in Sub-Saharan Africa and other low-income

countries. This could present a higher opportunity for success and effectiveness of rural

based programs (CIA, 2000).

It is not possible to talk about Uganda’s economy without mentioning the country’s

agricultural sector. Agriculture accounts for the biggest proportion of the country’s GDP.

Its performance closely predicts the economy’s overall behavior. Agriculture has been

referred to as the ‘backbone of the Ugandan economy’. It is not surprising that the

devastating effects of the 1997 El Nino weather phenomenon on Uganda’s agricultural

output affected the whole economy. Agriculture provides food for domestic consumption,

raw materials for local industries, is the major source of export earnings, and employment.

In 1990, approximately 80% of the country’s total work force was employed in agriculture.

Uganda’s principal exports are coffee, fish and fish products, tea, cotton and tobacco. In

16

1999, the country generated US $549 million in exports (FoB) from agricultural trade.

Owing to the great contribution of this sector, more will be mentioned in the next section.

The second most important economic sector in Uganda is the service industry (including

but not limited to tourism, construction, hotels, and transportation). The service industry

provides critical support to the other sectors of the economy. This sector employs about

13% of the total work force. In a span of 10 years, its contribution to the economy has

been increasing from 28.7% in 1987, to 30.3% in 1992, and to 34.2% in 1997 (Uganda,

1998a). In 2001, the service industry accounted for over 70% of the value added to

manufactured goods in the country (Uganda Export Promotion Board, 2002).

During the same time frame, the share of the manufacturing sector (the third most

important sector) in the total national GDP steadily rose. In 1987, it contributed 4.9%, and

6.1% and 9.0% to GDP in 1992 and 1997 respectively. The manufacturing sector employs

about 6% of the total work force in Uganda. This sector is, however, heavily dependent on

imports of materials. Therefore increases in prices of imports hinder growth.

2.1.2.1 The Uganda Agricultural Sector

As mentioned in section 2.1.2.0 above, the agricultural sector makes the largest

contribution to Uganda’s economy. In 1989 it contributed 56.8% of the national GDP, a

percentage that has, however, been declining since. In 1999 agriculture’s share of GDP

was 41.9%, which represents a drop of over 14% in just a decade. This is despite its

increasing annual growth rate from 4.0% in 1990 to 4.6% in 1996. This is confirmation

that the value of agricultural production has declined both in absolute terms and in

relation to other sectors (Uganda, 1998a).

Of the country’s total area of approximately 236,046 sq. km, land area accounts for over

199,710sq. km of which 25% is arable, 9% is under permanent crops, another 9% under

permanent pasture and the rest under other uses (roads, buildings and other

infrastructure). Available information (World Bank, 1993) suggests that only 30% of the

total cultivable area is under use. In 1990, an estimated 4.6 million ha were cultivated,

and of this, 36% was under Coffee, Banana, Tea, and Sugar, while cereal crops like maize,

millet, and sorghum took up about 23%. Table 2.1 shows selected 1981 to 2001 figures of

land area under various food crops in the country.

17

The amount of cultivated area currently fluctuates around 3.6 million ha from its high of

5.5 million ha in 1978. The majority of farm production (80%) is carried out on an average

of one - two hectare farms. Cultivated area per farm shows a substantial increase in farm

size in the central and southwestern region.

Table 2.1: Cultivated Area of Major Food Crops in Uganda ('000 ha) Crop 1981 1986 1989 1990 1992 1994 1996 1998 2000 2001

Millet 293 341 374 373 396 412 400 401 384 389

Maize 260 321 566 401 438 563 584 616 629 652

Sorghum 170 207 245 240 250 260 271 280 280 282

Rice 12 18 73 39 50 55 58 64 72 76

Wheat 4 5 4 2 5 5 5 5 7 8

Sweet potato 33 407 300 413 442 473 516 544 555 572

Irish potato 24 19 35 32 37 44 53 60 68 73

Cassava 309 361 495 412 362 320 335 356 401 390

Beans 289 396 431 495 536 574 615 645 699 731

Field peas 18 17 27 24 26 28 29 31 29 36

Cowpeas 40 49 46 49 49 53 56 60 64 64

Groundnut 110 176 217 186 184 189 195 200 199 208

Pigeon peas 54 66 67 62 62 67 71 74 78 78

Soybeans 5 11 20 37 59 68 76 80 106 127

Sesame seed (Simsim) 70 70 133 124 143 158 172 179 194 203

Source: The World Bank (1993), FAOSTAT (2002)

More than 70% of the farms are primarily crop-production oriented. In the western areas,

over 90% of crop production farms are in monocrop stands while in the other regions

mixed cropping systems predominate (World Bank, 1993). Labor is primarily from family

sources. During peak seasons like land preparation, weeding and harvesting, hired labor

is used especially in the central region. In the northeastern region, labor-sharing

arrangements are common while in the north, communal labor is widely used.

Food crop production increased over the last decade in almost all crop categories with the

highest increase noticeable in cereal crops (FAOSTAT, 2002). In the same period, the

livestock sector experienced moderate increases and accounted for 17% of agricultural

GDP. More than 90 percent of agricultural output is consumed domestically.

18

Generally, cash crop production has experienced wide fluctuations. Coffee yields have

been low and declining over the past fifteen years. Recently there has been some increase

in cash crop production (mainly cotton). In January – August 2000, tea output increased

by 22% compared with the same period in 1999 (CIA, 2000). In 2001, tea production

experienced a 12% reduction. Coffee production increased from 196,800 tons in 1998 to

198,000 tons in 1999, but dropped to 186,000 tons (EIU, 2001). Cotton’s increase was

from 45,100 tons in 1998 to 46,000 tons in 1999.

Part of the increase in crop production was prompted by the high urban demand for food

(World Bank, 1993), favorable government policies (McCalla, 1999; CIA, 2000; EIU, 2001)

and the expansion in cultivated area for food crops (Odulaja and Kiros, 1996; also

illustrated in Table 2.1). Sound government policies, including the continued investment

in the rehabilitation of infrastructure, improved incentives for production and exports, and

reduced inflation also led to boosting of production. Incentives for production and export

included subsidizing producers of export crops (CIA, 2000).

2.1.2.2 The Ugandan Agricultural Research and Extension Network

Numerous agricultural research activities on major crops in the country are one of the

biggest contributing factors to Uganda’s increase in agricultural production. Uganda has

had a long tradition of crop research. Agricultural research began in 1908, the major focus

then, being the improvement of production of major export crops (such as cotton and

coffee) to increase Uganda’s share of these crops in the international market (Uganda,

1988). Progressively, research focus shifted, and grain crops such as beans and maize

were introduced to the research arena. The establishment of the National Agricultural

Research Organization (NARO) in 1992 was aimed at increasing the amount of research on

all major crops in the country (Kyamanywa, 1996).

Prior to the establishment of NARO, agricultural research activities were scattered and

uncoordinated in three ministries: the Ministries of Agriculture and Forestry, Animal

Industry and Fisheries, and Regional Cooperation (Uganda, 1988). However, with the

launching of the Rehabilitation and Development Plan 1987-1999, which aimed at a rapid

recovery of the agricultural sector and the improvement and stabilization of its

contribution to the GDP, the need for “organized research to contribute more effectively

and efficiently to development became even more urgent” (Uganda, 1988, p.1). As such,

NARO was formed to act as a catalyst in the development process of Uganda. Its aim was

19

to improve the efficiency of the utilization of resources allocated to research and enhance

research activities in the agricultural sector. In carrying out these duties, NARO’s board

was expected to pay special attention to obtaining, through government and other

appropriate sources, financial and other resources required for the implementation of the

National Agricultural Research (NAR) strategy and plans. These resources are obtainable

from donors, the international community, research agencies and the public.

Donor sources of funding are important to many developing countries, in part because of

the component of foreign exchange that they represent (Uganda, 1988). In the agricultural

sector, under the umbrella of NARO, a number of donor sources were identified. The

extent of investment by each varies depending on the scope of the activity and the progress

of the sector involved. The World Bank and other donors reacted positively to Uganda’s

economic reform effort and mounted an expanded level of donor support (World Bank,

2000).

Clearly, Uganda’s agricultural research stands to benefit from this expanded support.

Uganda currently is a member of a number of local, regional and international agricultural

organizations. These provide support to agricultural research in many forms. Some are

purely donors while many still are in collaborative agreements and/or partnerships

between the major funding body and the host country’s private and public institutions.

2.2 The Collaborative Research Support Program (CRSP)

2.2.1 The IPM CRSP Collaborative Research Support Programs were created by the United States Agency for

International Development (USAID) and the Board for International Food and Agriculture

Development (BIFAD) as a long-term mechanism to focus capabilities of US Land Grant

Colleges to carry out the international food and agricultural research mandate of the US

Government (IPM CRSP, 2001). In September 1993, the IPM CRSP was initiated under the

International Development and Food Assistance Act of 1975 (IPM CRSP, 2001), funded by

USAID and participating universities. In the USA, participating institutions include Ohio

State, Purdue, University of Georgia, Penn State, Montana State, USDA Vegetable lab and

Virginia Tech. Virginia Tech serves as the Management Entity for the IPM CRSP for the

program.

20

The IPM CRSP engages in research, education/training and information exchange through

collaborative partnerships among US and developing country institutions. These

developing countries include Bangladesh and Philippines in Asia, Albania in Eastern

Europe, Jamaica in the Caribbean, Ecuador and Guatemala in Latin America, and Mali

and Uganda in Africa. As some of its major objectives the IPM CRSP seeks to evaluate

appropriate participatory IPM approaches, describe technical factors affecting IPM and

identify and describe the social, economic and institutional factors affecting pest

management. In doing these, the IPM CRSP supports research activities in the host

countries. In Uganda, Makerere University and National Agricultural Research

Organization (NARO) are the participating institutions.

2.2.2 Why IPM? The concept of Integrated Pest management (IPM) was first conceived after World War II

when it was determined that a control system was required to check overuse or abuse of

pesticides used to control major pests of cotton in the USA. It required a compatible

control strategy, which was a mix of biological and reduced chemical control tactics. In

1972, IPM was formulated into national policy and under US president Jimmy Carter; an

interagency coordinating committee was formed in 1979 to ensure development and

implementation of IPM practices (Ehler and Bottrell, 2000).

The focus of IPM research is to reduce pesticide usage on crops while maintaining a high

level of pest control. In general, IPM calls for a much greater reliance on non-chemical

approaches to pest management (IFPRI, 1998) while maintaining agricultural production

and preserving profitability (Mullen, Norton and Reaves 1997). In doing this, IPM

encourages strategies that include greater dependence on biological approaches, cultural

approaches and judicious use of some pesticides. A broader definition of IPM is that given

by Wightman (1998):

“IPM consists of management activities carried out by farmers that maintain

the intensity of potential pests at levels below which they become pests, without

endangering the productivity and profitability of the farming system as a whole,

the health of the farm family and its livestock and the quality of the adjacent

and down stream environments.” (Wightman, IFPRI homepage, 1998).

21

Pests have been known to attack crops virtually at every stage of crop development: at pre-

germination, budding, flowering, harvest and in post harvest/storage thereby leaving the

crop with no "breathing space." This necessarily calls for pest control. Various methods of

pest control may be employed and can be categorized into two broad groups: Chemical and

non-chemical - each with its advantages and disadvantages.

The range of non-chemical options is diverse, including biological control, cultural control,

plant host resistance, sanitation and genetic transformations. Biological control, the use of

natural enemies of pests and entomopathogens is somewhat limited in its applicability and

its application for subsistence level farming although the potential for expanding its use is

great (Jackai, et al., 1985; Pimentel, 1986).

Chemical means have a number of benefits like ease in application (although not

necessarily safe application), effectiveness and fast action on target pests. However, their

disadvantages, especially in interfering with the ecosystem, are well documented.

Cultural methods include manipulation of planting dates and cropping patterns, such as

crop diversity and crop rotation. These methods achieve their pest control abilities from

having one or more crops in the rotational sequence that are resistant to a key pest. For

weed suppression, the success of rotation systems appears to be based on the use of crop

sequences that create varying patterns of resource competition, soil disturbance, and

mechanical damage to provide an unstable and frequently inhospitable environment that

prevents the proliferation of a particular weed species, (Liebman and Dyck, 1993).

Rotations offer an opportunity to increase production, either through direct yield increases

or through reductions in some of the inputs required for the present or next crop. Greater

benefits are usually obtained by rotating two distinctly unrelated crops. Crop diversity

makes the environment less favorable to certain pests while manipulation of planting time

avoids reduction in yields caused by pests. In addition, cultural controls are far less

ecologically disruptive than the standard chemical control practices.

However, cultural methods are often labor intensive (Pimentel, 1986). Considering that

most subsistence farms use family labor, one might infer that this should not be a

problem. However, with the fast paced life that is expected in the near future, and the

subsequent value of time, these two resources: time and labor will become constraints to

cultural control means. Furthermore in subsistence production systems, family labor is

22

often in short supply at times such as sowing, weeding and harvesting. In addition, these

methods may have added risks. For instance, in a bid to control the known pests, altering

planting time may create a more favorable environment to more destructive pests. Also

planting time manipulations may be constrained by climatic changes. Moreover the

effectiveness of these cultural methods is highly unpredictable (Pimentel, 1986).

In general, each method (biological, cultural or chemical) may contribute to pest

suppression. However, according to Jackai et al., (1985), no one method provides

satisfactory results. Hence, an integrated approach that avoids the use of a single control

tactic is necessary. In effect when several methods are employed, the amount of each

component (biological, cultural, chemical), including the use of pesticides in the package

may be reduced - this is the basic principle in Integrated Pest Management programs.

2.2.3 IPM in Uganda In many developing countries, IPM systems and practices have been pursued for over two

decades. In Uganda, early IPM practices were focused on coffee and cotton. This was

probably because of these crops’ importance as major cash crops and foreign exchange

earners for the country and hence the urgent need to protect them from devastating yield

loss due to pests. Post harvest systems were also developed under these early Uganda IPM

efforts by various agricultural research institutes in the country. Both cultural and

chemical methods were used to control pest populations on these crops. The system was

based on a careful analysis of pest populations and pest patterns and determining a

suitable strategy for their control. However, the period of political and civil strife saw the

collapse of this otherwise effective IPM system.

However, this was not the end of IPM efforts in Uganda. Kyamanywa (1996) mentions that

efforts to rejuvenate IPM were pursued in 1994 when under funding from the IPM CRSP,

the Uganda IPM Network was formed. Its initial focus was directed towards raising

knowledge and awareness of fundamental IPM concepts. Subsequently, efforts to develop

pest management alternatives for priority pests with an added emphasis on environmental

quality were incorporated and more aspects of agricultural production were considered.

The ‘new’ IPM CRSP crop focus expanded to include key food crops many of which were

grain crops - Beans, Maize, Cowpeas, Sorghum and Groundnut. Other additions to IPM

CRSP trials in Uganda included disease and pest control strategies on two high-value

23

horticultural crops: tomatoes and potatoes. Mold incidence on stored maize and

groundnuts and coffee wilt incidence are currently being investigated. Among the crops the

IPM CRSP has had active programs on a long-term basis include sorghum, groundnuts

and cowpeas. Thus this study focuses on these crops.

2.3 Technology Adoption

Various authors define the term “technology” in a variety of ways. Rogers (1995) uses the

words ‘technology’ and ‘innovation’ synonymously and defines technology as the design for

instrumental action that reduces the uncertainty in the cause-effect relationship involved

in achieving a desired outcome.

A more meaningful definition may be that a technology is a set of ‘new ideas’. New ideas

are associated with some degree of uncertainty and hence a lack of predictability on their

outcome. For a technology to impact on the economic system, blending into the normal

routine of the intended economic system without upsetting the system’s state of affairs is

required. This entails overcoming the uncertainty associated with the new technologies. It

therefore comes as no surprise that several studies set out to establish what these factors

are, and how they can be eliminated (if constraints) or promoted (if enhancers) to achieve

technology adoption.

Perhaps a clearer definition of the term ‘technology’ can be obtained from the work by

Enos and Park (1988), who, in their study of adoption of imported technology, define

technology as “the general knowledge or information that permits some tasks to be

accomplished, some service rendered, or some products manufactured” (p.9). Abara and

Singh (1993) explain that it is the actual application of that knowledge that would be

termed ‘technology’. Although in the Enos and Park (1988) study, the focus was non-

agricultural, this definition fits agricultural technologies too. From their definition, it is

clear that technology is aimed at easing work of the entity to which it applies. Most

technologies are therefore consequently termed ‘labor-saving’, ‘time-saving’, ‘capital-saving’

or ‘energy-saving’ and so forth. To economists this implies saving on resources that are

scarce.

Adoption is an outcome of a decision to accept a given innovation. Feder, Just and

Zilberman (1985) while quoting Roger’s earlier work of 1962 define adoption as “a mental

24

process an individual passes from first hearing about an innovation to final utilization”

p.256. Much scholarly interest on adoption falls in two categories: rate of adoption, and

intensity of adoption. It is usually necessary to distinguish between these two concepts as

they often have different policy implications. Rate of adoption, the relative speed with

which farmers adopt an innovation, has as one of its pillars, the element of ‘time’. On the

other hand, intensity of adoption refers to the level of use of a given technology in any time

period.

Clearly, a technology that is being adopted has an edge over conventional practices.

Usually, a technological innovation encompasses at least some degree of benefit for its

potential adopters (Rogers, 1995). In this study, a technology, as it relates to IPM, is a set

of practices (new or old) integrated into a package that aims to control specific pests on

select crops in a manner that is proven more effective than the conventional means.

Several stages precede adoption. Awareness of a need is generally perceived as a first step

in adoption process (Rogers, 1983). The other stages are: Interest, Evaluation, Acceptance,

Trial, and finally, Adoption (Lionberger, 1960). The Lionberger analysis also notes that

these stages occur as a continuous sequence of events, actions and influences that

intervene between initial knowledge about an idea, product or practice, and the actual

adoption of it. However, not all decisions involve a clear-cut sequence. In fact most recent

literature suggests that these stages may occur concurrently and some may/not occur in

adoption decision processes.

According to Cameron (1999) the dynamic process of adoption involves learning about a

technology over time. In fact many innovations require a lengthy period often of many

years from the time they become available to the time they are widely adopted (Lionberger,

1960; Rogers, 1995; Enos and Park, 1988). The average time between initial information

and final adoption varies considerably by person, place and practice. Alston, Norton and

Pardley (1995) demonstrate that the time after the initial investment in research through

the generation of pre-technology knowledge up to maximum adoption by producers

involves many long, variable and uncertain lags.

The literature on this subject (Griliches, 1957; Lionberger, 1960; Rogers, 1983; Alston,

Norton and Pardley, 1995), describes the process of adoption as taking on a logistic

nature. It increases with time (as the stock of knowledge increases), reaches a maximum

25

level, and later decreases as the technology depreciates or becomes obsolete. Fig 2.1 below

shows the shape that most adoption processes take.

Figure 2.1: The adoption curve Source: Alston, Norton and Pardley (1995)

The research stage may take up to 5 years and the development stage another five years.

However, this is a generalization. Some adoption processes are shorter or longer than the

model shown above. Underlying conditions may shorten or lengthen this period.

2.3.1 Measuring Adoption Although studies by Mullen, Norton and Reaves (1997), suggest that adoption of IPM is

usually a matter of degree, this is not to state that the measurement of adoption is simple.

In fact, in his study, Nowak, (1996) did agree that measuring the adoption of IPM can be

more complex than it sounds: “At first glance, it appears to be nothing more than a

question of whether a grower is or is not using a specific practice. Yet this simplistic view

quickly changes as one begins to assess how it is being used, where it is being used, and

the appropriateness of that use relative to actual pest conditions” (Nowak, 1996, p. 99).

Much more work is needed in refining methods and in compiling the data needed to

credibly measure and monitor IPM adoption.

The rate of adoption is usually measured by the length of time required for a certain

percentage of members of a system to adopt an innovation. Extent of adoption on the other

hand is measured from the number of technologies being adopted and the number of

10

30

Adoption Process Research and

Development lag

Adop

tion

of

new

Tec

hn

olog

y

Time (Years)

26

producers adopting them. The current study focuses on the extent of adoption and the

factors affecting it.

Depending on the technology being investigated, various parameters may be employed to

measure adoption. Measurements also depend on whether they are qualitative or

quantitative. For instance in the study investigating the adoption of improved seed and

fertilizer in Tanzania, Nkonya, Schroeder and Norman (1997) estimated the intensity of

adoption by examining the area planted to improved seed and the area receiving fertilizer.

For another study that investigated the adoption of use of single-ox technology, pesticide

and fertilizer use, the dependent variable was the number of farmers using pesticide and

fertilizer (Kebede, Gunjal and Coffin 1990).

In another study on factors affecting peanut producer adoption in Georgia, McNamara,

Wetzstein and Douce, (1991) used the producer’s decision to adopt or not to adopt and

subdivided respondents into two groups: Adopters and non-adopters. Similarly, farmers’

perceptions are examined in several studies including the one by Adesiina and Baidu-

Forson, (1995) and that by Tjornhom, (1995). In the former, farmers’ perception of

characteristics of sorghum and modern varieties are taken into account. In the latter,

farmer perceptions on harmful effects of pesticides on water quality, on health of

individuals and on natural enemies of insects are sought. Baidu-Forson, (1999) examined

farmers’ perceived utility from adopting half-crescent shaped earthen mounds - a land-

enhancing technology.

While direct qualitative attributes are harder to measure, several studies have used

estimates of probabilities (Shakya and Flinn, 1985; Harper et al, 1990; Green and

Ng'ong'ola, 1993; Kebede, Gunjal and Coffin 1990). In soliciting respondents’ subjective

perceptions, researchers capture the qualitative aspects that influence farmers’ decisions

probably because farmers’ technology choices are based on their subjective probabilities

(Feder, Just and Zilberman, 1985). Farmers’ perceptions are interpreted as perceived

profitability of a technology and translate into more resources being devoted to it – hence

adoption.

2.3.2 Determinants of Adoption A variety of studies are aimed at establishing factors underlying adoption of various

technologies. As such, there is an extensive body of literature on the economic theory of

27

technology adoption. Several factors have been found to affect adoption. These include

government policies, technological change, market forces, environmental concerns,

demographic factors, institutional factors and delivery mechanism. Although not tested, it

appears that only four of these broad items may be related to IPM technology adoption.

They include Market forces: availability of labor, technology resource requirements, farm

size, level of expected benefits, and level of effort required to implement the technology;

Social factors: Age of potential adopter, social status of farmers, education level and

gender-related aspects, household size, and farming experience; Management factors:

membership to organizations, the capacity to borrow, and concerns about environmental

degradation and human health of farmers; Institutional/technology delivery mechanisms:

information access, extension services, and prior participation in, and training in pest

control practices.

Some studies classify the above factors into broad categories: farmer characteristics, farm

structure, institutional characteristics and managerial structure (McNamara, Wetzstein

and Douce, 1991) while others classify them under social, economic and physical

categories (Kebede, Gunjal and Coffin 1990). Others group the factors into human capital,

production, policy and natural resource characteristics (Wu and Babcock, 1998) or simply

whether they are continuous or discrete (Shakya and Flinn, 1985). By stating that

agricultural practices are not adopted in a social and economic vacuum, Nowak (1987)

brought in yet another category of classification. He categorizes factors influencing

adoption as informational, economic and ecological.

There is no clear distinguishing feature between elements within each category. Actually,

some factors can be correctly placed in either category. For instance, experience as a factor

in adoption is categorized under ‘farmer characteristics’ (McNamara, Wetzstein and Douce,

1991; Tjornhom, 1995) or under ‘social factors’ (Kebede, Gunjal and Coffin 1990; Abadi

Ghadim and Pannell, 1999) or under ‘human capital characteristics’ (Caswell et al., 2001).

Perhaps it is not necessary to try and make clear-cut distinctions between different

categories of adoption factors. Besides, categorization usually is done to suit the current

technology being investigated, the location, and the researcher’s preference, or even to suit

client needs. However, as some might argue, categorization may be necessary in regard to

policy implementation.

28

Extensive work on agricultural adoption in developing countries was pioneered by Feder,

Just and Zilberman, (1985). Since then the amount of literature on this subject has

expanded tremendously. Because of this extensive literature, the following section provides

a review of selected factors as they relate to agricultural technology adoption.

2.3.2.1 Economic Factors Farm Size

Much empirical adoption literature focuses on farm size as the first and probably the most

important determinant. Farm size is frequently analyzed in many adoption studies

(Shakya and Flinn, 1985; Harper et al, 1990; Green and Ng'ong'ola, 1993; Adesiina and

Baidu-Forson, 1995; Nkonya, Schroeder and Norman 1997; Fernandez-Cornejo, 1998;

Baidu-Forson, 1999; Boahene, Snijders and Folmer, 1999; Doss and Morris, 2001; and

Daku, 2002). This is perhaps because farm size can affect and in turn be affected by the

other factors influencing adoption. In fact, some technologies are termed ‘scale-dependant’

because of the great importance of farm size in their adoption.

The effect of farm size has been variously found to be positive (McNamara, Wetzstein, and

Douce, 1991; Abara and Singh, 1993; Feder, Just and Zilberman, 1985; Fernandez-

Cornejo, 1996, Kasenge, 1998), negative (Yaron, Dinar and Voet, 1992; Harper et al, 1990)

or even neutral to adoption (Mugisa-Mutetikka et al., 2000). Farm size affects adoption

costs, risk perceptions, human capital, credit constraints, labor requirements, tenure

arrangements and more. With small farms, it has been argued that large fixed costs

become a constraint to technology adoption (Abara and Singh, 1993) especially if the

technology requires a substantial amount of initial set-up cost, so-called “lumpy

technology.” In relation to lumpy technology, Feder, Just and Zilberman, (1985) further

noted that only larger farms will adopt these innovations. With some technologies, the

speed of adoption is different for small- and large- scale farmers. In Kenya, for example, a

recent study (Gabre-Madhin and Haggblade, 2001) found that large commercial farmers

adopted new high-yielding maize varieties more rapidly than smallholders.

Furthermore, access to funds (say, through a bank loan) is expected to increase the

probability of adoption. Yet to be eligible for a loan, the size of operation of the borrower is

crucial. Farmers operating larger farms tend to have greater financial resources and

chances of receiving credit are higher than those of smaller farms.

29

A counter argument on the effect of farm size can be found in Yaron, Dinar and Voet,

(1992) who demonstrate that a small land area may provide an incentive to adopt a

technology especially in the case of an input-intensive innovation such as a labor-intensive

or land-saving technology. In that study, the availability of land for agricultural production

was low, consequently most agricultural farms were small. Hence, adoption of land-saving

technologies seemed to be the only alternative to increased agricultural production.

Further, in the study by Fernandez-Cornejo (1996), farm size did not positively influence

adoption.

The majority of the studies mentioned above consider total farm size and not crop acreage

on which the new technology is practiced. While total farm size has an effect on overall

adoption, considering the crop acreage with the new technology may be a superior

measure to predict the rate and extent of adoption of technology (Lowenberg-DeBoer,

2000). Therefore in regard to farm size, technology adoption may best be explained by

measuring the proportion of total land area suitable to the new technology.

Cost of Technology

The decision to adopt is often an investment decision. And as Caswell et al, (2001) note,

this decision presents a shift in farmers’ investment options. Therefore adoption can be

expected to be dependent on cost of a technology and on whether farmers possess the

required resources. Technologies that are capital-intensive are only affordable by wealthier

farmers (El Oster and Morehart, 1999) and hence the adoption of such technologies is

limited to larger farmers who have the wealth (Khanna, 2001). In addition, changes that

cost little are adopted more quickly than those requiring large expenditures, hence both

extent and rate of adoption may be dependent on the cost of a technology. Economic

theory suggests that a reduction in price of a good or service can result in more of it being

demanded.

Level of Expected benefits

Programs that produce significant gains can motivate people to participate more fully in

them. In fact, people do not participate unless they believe it is in their best interest to do

so. Farmers must see an advantage or expect to obtain greater utility in adopting a

technology. In addition, farmers must perceive that there is a problem that warrants an

alternative action to be taken. Without a significant difference in outcomes between two

options, and in the returns from alternative and conventional practices, it is less likely

30

that farmers, especially small-scale farmers will adopt the new practice (Abara and Singh,

1993). Farmers may receive little long-term benefits from IPM adoption, which negatively

influences adoption. A higher percentage of total household income coming from the farm

through increased yield tends to correlate positively with adoption of new technologies

(McNamara, Wetzstein, and Douce, 1991; Fernandez-Cornejo, 1996)

Off-farm hours

The availability of time is an important factor affecting technology adoption. It can

influence adoption in either a negative or positive manner. Practices that heavily draw on

farmer’s leisure time may inhibit adoption (Mugisa-Mutetikka et al., 2000). However,

practices that leave time for other sources of income accumulation may promote adoption.

In such cases, as well as in general, income from off-farm labor may provide financial

resources required to adopt the new technology.

2.3.2.2 Social Factors Age of Adopter

Age is another factor thought to affect adoption. Age is said to be a primary latent

characteristic in adoption decisions. However there is contention on the direction of the

effect of age on adoption. Age was found to positively influence adoption of sorghum in

Burkina Faso (Adesiina and Baidu-Forson, 1995), IPM on peanuts in Georgia (McNamara,

Wetzstein, and Douce, 1991), and chemical control of rice stink bug in Texas (Harper et

al., 1990). The effect is thought to stem from accumulated knowledge and experience of

farming systems obtained from years of observation and experimenting with various

technologies. In addition, since adoption pay-offs occur over a long period of time, while

costs occur in the earlier phases, age (time) of the farmer can have a profound effect on

technology adoption.

However age has also been found to be either negatively correlated with adoption, or not

significant in farmers’ adoption decisions. In studies on adoption of land conservation

practices in Niger (Baidu-Forson, 1999), rice in Guinea (Adesiina and Baidu-Forson,

1995), fertilizer in Malawi (Green and Ng'ong'ola, 1993), IPM sweep nets in Texas (Harper

et al., 1990), Hybrid Cocoa in Ghana (Boahene, Snijders and Folmer, 1999), age was either

not significant or was negatively related to adoption.

31

Older farmers, perhaps because of investing several years in a particular practice, may not

want to jeopardize it by trying out a completely new method. In addition, farmers’

perception that technology development and the subsequent benefits, require a lot of time

to realize, can reduce their interest in the new technology because of farmers’ advanced

age, and the possibility of not living long enough to enjoy it (Caswell et al., 2001; Khanna,

2001). Furthermore, elderly farmers often have different goals other than income

maximization, in which case, they will not be expected to adopt an income-enhancing

technology. As a matter of fact, it is expected that the old that do adopt a technology do so

at a slow pace because of their tendency to adapt less swiftly to a new phenomenon

(Tjornhom, 1995).

Education

Studies that have sought to establish the effect of education on adoption in most cases

relate it to years of formal schooling (Tjornhom, 1995, Feder and Slade, 1984). Generally

education is thought to create a favorable mental attitude for the acceptance of new

practices especially of information-intensive and management-intensive practices (Waller

et al. 1998; Caswell et al., 2001). IPM is frequently stated to be a complex technology

(Pimentel, 1986; Boahene, Snijders and Folmer, 1999). What is more, adoption literature

(Rogers 1983) indicates that technology complexity has a negative effect on adoption.

However, education is thought to reduce the amount of complexity perceived in a

technology thereby increasing a technology’s adoption. According to Ehler and Bottrell

(2000), one of the hindrances to widespread adoption of IPM as an alternative method to

chemical control is that it requires greater ecological understanding of the production

system. For IPM, the relevance of education comes to play in a number of ways. First,

effective IPM requires regular field monitoring of pests conditions to identify the critical

periods for application of a pesticide or other control measures (Adipala et al, 1999).

Farmers’ knowledge of insect life cycles is also crucial when precision is required about the

best stage of the life cycle to apply a particular control strategy. In addition, knowledge of

the possible dangers from improper use of particular practices may direct farmers to the

safest application procedure regarding a given control strategy especially where chemicals

are involved.

The ability to read and understand sophisticated information that may be contained in a

technological package is an important aspect of adoption. In the case of IPM, the ability to

comprehend pesticide application instructions and proper measurement required in

32

certain control strategies becomes useful. Furthermore, distribution of knowledge reduces

the risk of adopting a new technology. Increased education is thus expected to improve

IPM adoption.

In recent studies reviewed, including Daku (2002) and Doss and Morris (2001), education

positively affected IPM adoption. A study on IPM practices on potatoes identified level of

education as one of the major factors that positively affected the observed level of IPM

practices with Ohio potato growers (Waller et al, 1998). However, in adoption of IPM insect

sweep nets in Texas, higher education was negatively related to adoption (Harper et al.,

1990).

Gender Concerns

Gender issues in agricultural production and technology adoption have been investigated

for a long time. Most show mixed evidence regarding the different roles men and women

play in technology adoption. In the most recent studies, Doss and Morris (2001) in their

study on factors influencing improved maize technology adoption in Ghana, and Overfield

and Fleming (2001) studying coffee production in Papua New Guinea show insignificant

effects of gender on adoption. The latter study notes “effort in improving women’s working

skills does not appear warranted as their technical efficiency is estimated to be equivalent

to that of males” (p.155). Since adoption of a practice is guided by the utility expected from

it, the effort put into adopting it is reflective of this anticipated utility. It might then be

expected that the relative roles women and men play in both ‘effort’ and ‘adoption’ are

similar, hence suggesting that males and females adopt practices equally.

2.3.2.3 Institutional Factors

Information

Acquisition of information about a new technology demystifies it and makes it more

available to farmers. Information reduces the uncertainty about a technology’s

performance hence may change individual’s assessment from purely subjective to objective

over time (Caswell et al., 2001). Exposure to information about new technologies as such

significantly affects farmers’ choices about it. Feder and Slade (1984) indicate how,

provided a technology is profitable, increased information induces its adoption. However in

the case where experience within the general population about a specific technology is

limited, more information induces negative attitudes towards its adoption, probably

33

because more information exposes an even bigger information vacuum hence increasing

the risk associated with it. A good example is the adoption of recombinant bovine

Somatotropin Technology (rbST) in dairy production (McGuirk, Preston and Jones, 1992;

Klotz, Saha and Butler, 1995).

Information is acquired through informal sources like the media, extension personnel,

visits, meetings, and farm organizations and through formal education. It is important that

this information be reliable, consistent and accurate. Thus, the right mix of information

properties for a particular technology is needed for effectiveness in its impact on adoption.

Extension Contacts

Good extension programs and contacts with producers are a key aspect in technology

dissemination and adoption. A recent publication stated that “a new technology is only as

good as the mechanism of its dissemination” to farmers (IFPRI, 1995 p. 168). Most studies

analyzing this variable in the context of agricultural technology show its strong positive

influence on adoption. In fact Yaron, Dinar and Voet, (1992) show that its influence can

counter balance the negative effect of lack of years of formal education in the overall

decision to adopt some technologies.

2.3.3 The Combined effect Although most adoption literature concentrates on single technology adoption - for

example adoption of fertilizer (Green and Ng'ong'ola, 1993), improved varieties like beans

(Kato, 2000), hybrid cocoa (Boahene, Snijders and Folmer, 1999) and many more, other

studies investigate adoption of a combination of technologies such as improved varieties

and fertilizer (Nkonya, Schroeder and Norman 1997; Shakya and Flinn, 1985). As such,

some literature (Feder, Just and Zilberman, 1985; Rogers, 1995) suggests that adoption of

technologies may in effect be enhanced because of complementarities that exist between

the technologies.

Complementarities occur at two levels: at the factor level and at the technology level. At

the factor level, complementarities occur from the manner in which combinations of

factors act together to influence adoption (Lionberger, 1960). Additionally,

complementarities between factors occur where all inputs considered together have a

significant effect on adoption but when the influence of one is held constant, the

correlation between the other remaining inputs and technology adoption is greatly lowered

34

(Lionberger, 1960). As such where inputs that are critical for adoption are in short supply

– for instance water supply that is critical for irrigation technology adoption, the

unavailability may hinder adoption. Thus, crucial inputs must be readily available in order

to encourage adoption.

At the technology level, complementarities occur because one technology enhances the

positive impacts of another. For example in some cases, the high yield potential of seed

can be realized only if fertilizer is applied. In fact, in most studies addressing the use of

improved seeds and fertilizer, a complementary relationship is found between them. For

example, in Northern Tanzania, farmers tend to adopt improved maize seed in combination

with fertilizers (Nkonya, Schroeder and Norman (1997).

The site-specificity of agricultural practices leads to some authors asserting that adoption

studies in every region experiencing a technological change are warranted. This might be

because populations are heterogeneous and individual behavior is dynamic (Feder, Just

and Zilberman, 1985). Furthermore, there are numerous differences in factor endowments

and farmer characteristics among regions. Thus an adoption study on a technology in a

geographical setting does not imply that a similar study of the same technology is

unwarranted in another geographical setting. Moreover, even within a geographical

setting, different regions have varying adoption patterns for the same type of technology.

Yaron, Dinar and Voet (1992) assert that extrapolations of adoption results should be

avoided and that where possible region specific studies should be encouraged.

2.4 Summary

From the preceding discussions, and review of literature, it is clear that several factors

may help to explain the pattern of technology adoption. However, to attempt to include all

these in a model is generally not a viable option. Limited research funds may limit the

amount of data that can be collected. In addition, collinearity generally exists among a

number of these factors, precluding their inclusion in modeling efforts. Considering this

limitation, therefore, those factors hypothesized to exert the largest influence on

technology adoption, given the circumstances in the study area, are investigated in the

analyses. As discussed above, they include, market forces, social factors, management

characteristics, institutional factors and information delivery mechanisms. These factors

may act as either incentives or barriers to IPM adoption.

35

Chapter 3 Methods

This chapter contains a description of: the study area, the methods used in the collection

of data, and the sampling frame. A review of techniques used to analyze this kind of data,

including their limitations is presented, and the conceptual models used for data analysis

are developed.

3.1 The Study area, Sample, and Data collection techniques

3.1.1 The Study Area Kumi district is located around 1025’N, 35055’E in Eastern Uganda bordered by Mbale

district to the East, Soroti to the North and West, and Pallisa to the South.

Topographically the district lies at an altitude of between 1,036 and 1,127 meters above

sea level (masl). The vegetation is mainly short- grass savanna land. Appendix B2 shows

the location of Kumi in relation to other districts in the country. With an estimated land

area of 2,457 sq. km of which 350 hectares is forest reserve area, the district occupies

about 1.2 % of the country’s total land area (Uganda, 1998b). Administratively, Kumi

district is subdivided into three counties: Kumi, Bukedea and Ngora. In 1991, the district

had a population of 236,700 people (Rwaboogo, 1998) with a population density of 96

people per sq. km, which was among the least densely populated districts in eastern

Uganda. It has a gender ratio of 91 males: 100 females (Uganda, 1998b).

The district is mainly rural with only 5% of its population classified as urban. Agriculture

is the main economic activity. The district’s soils are broadly categorized as moderately

fertile sandy loams (World Bank, 1993) but with poor water holding capacity (FARMESA,

1998). Main crops grown include grains like millet, groundnuts, sorghum, rice, cowpea,

and soybeans, which are predominantly intercropped. Other crops include potatoes,

cassava, onions, sunflower and bananas. Crops are grown under a bimodal rainfall

pattern – the longer first rains from March to July and shorter second rains from

September to December. Annual rainfall ranges from 700-1,300mm/annum. Short

intermittent rainfall and poor water retention capacity of soils are said to be the major

constraints to increased food production in the area. Cotton is the major traditional export

crop grown in the area.

During the period 1980-1991, the districts of Eastern Uganda experienced a decline in

their population, which was attributable to out migration resulting from civil war. At that

36

time, Kumi district’s annual population growth rate was one of the lowest in the country at

minus 0.1%. Uganda’s eastern region has had one of the highest poverty indices since the

early 1990’s, although this figure is decreasing (Uganda, 1998b). In the years 1995/96 the

poorest 10% of the population in Kumi had 2.3% (approximately UShs 5,902 per capita)

share of the income while the richest 10% had over 31% of the share of income

(approximately UShs 560,000 per capita). Poverty in the eastern region can at least in part

be attributed to low levels of manufacturing, and service-related activities such as

transportation. Nationally, growth in these sectors has lead to substantial improvement in

living standards (Uganda, 1997).

Kumi district was selected for study for a number of reasons. The eastern region of

Uganda is the largest producer of groundnuts and second largest producer of sorghum in

the country. According to the 1995-96 National Household Survey (Uganda, 1997), the

eastern region produced 31% and 22% of the national totals of groundnuts and sorghum

respectively. In that survey, results for cowpea production in the region were not reported.

However, the 1996 IPM Participatory Appraisal in Eastern Uganda and the 1999 baseline

survey (Erbaugh et al., 2001) identified cowpea, sorghum and groundnuts as priority crops

in Kumi district. The crops are grown by about 80% of farmers. Considering the

importance of these crops in Kumi district, identifying the factors that induce IPM

adoption on these crops can be important in explaining factors affecting adoption for a

large part of the cowpea, sorghum and groundnut producing areas of the country. Finally,

Kumi is one of the IPM CRSP’s primary research sites in Uganda.

The study area covers two of the district’s three counties namely Kumi and Bukedea.

These counties are selected because the bulk of IPM CRSP work in the district has been

done there. A previous study examined the basic demographic, social and economic status

of Kumi and Iganga (Erbaugh et al., 2001). To put the current study in perspective, the

results of Kumi district study are summarized below.

37

Table 3.1 Summary of Social, Demographic and Economic Characteristics of Kumi District Farming System

Characteristics

Mean/% (n=100)

Age (Years) 43 Household size (Number of people) 9.8 Farm size (hectares) 5.45 Farm labor (Number of people) 6.8 Area in crops (hectares) 3.4 Educational level (Years) 6.6 Major sources of income

Agriculture (%) 76 Brewing (%) 9 Trading (%) 9 Salary (%) 8 Casual labor (%) 3

Total Household income range (‘000 Ushs) 152-300 Cultivation methods

Rent animal traction (%) 39 Own animal traction (%) 57 Rent tractors (%) 3 Own tractors (%) 1

Time spent on agricultural activities Full time (%) 35 Half time (%) 33 Less than half time (%) 32

Use of credit, labor and production inputs Formal credit (%) 30 Hire labor (%) 90 Exchange labor (%) 74 Fertilizer use (%) 4 Use of insecticides, herbicides and fungicides (%) 93

Source: Erbaugh et al., 2001

These results parallel the FARMESA study of 1998 in Kumi that revealed farm size range

of 2.5-10ha and household sizes of 6 persons on average and up to 20-25 people per

household. That study also identified off-farm income sources as mainly fishing, trading,

charcoal manufacture (burning) and casual labor. Average annual income for the area was

very difficult to estimate since not all earnings are in cash and farmers are reluctant to

provide information on actual household income. The FARMESA study estimated annual

income in Kumi in the range of $180-200, an equivalent of UShs180, 000-200,000.11

3.1.2 The Sample and Sampling Procedure To obtain respondents for this study, first, the four initial IPM on-farm trial farmers were

identified and traced back to sub county, parish and village level, which formed focal or

11 US$1=1,000 UShs in 1998

38

reference points – creating three focal sub-counties that is, Kumi sub-county, Malera sub-

county and Bukedeya sub-county (two of the active IPM farmers are located in Kumi sub-

county). Within each focal sub county, two parishes were selected: the focal parish in

which the active farmer(s) is (are) located, and another parish geographically neighboring

the focal parish. For example in Bukedeya sub-county, Bukedeya was the focal parish and

Okunguro was the adjacent parish. In addition, the initial IPM farmer group in Aturtur

sub-county (Kumi county) was identified and traced back to parish level (Refer to Figure

3.1 below).

Within each selected parish two villages were identified: the focal village and another

village within proximity to the focal village. In Bukedeya focal parish, for example,

Achabule was the focal village and Oswapai the adjacent village (Figure 3.1). With the help

of agricultural extension officials, farmers growing sorghum, cowpeas, and groundnuts on

their farms were identified in each village. Fifteen farmers were then randomly selected

from this group and interviewed.

39

KUMI DISTRICT

Okoba Omolokonyo

Okoba Parish

Odiding Ameje

Kabata**Parish

Kumi*S/County

Aturtur Apapayi

AturturParish

AturturS/County

Kumi County

Oswapai Achabule

Bukedeya** Parish

Kaloko Okunguro

OkunguroParish

BukedeyaS/County

Kalou Kachede

Kachede**Parish

Kakori Kabarwa

Kabarwa Parish

MaleraS/County

BukedeyaCounty

Green shade indicates focal points

Figure 3.1 Respondent Selection

County

S/County

Parish

Village

40

This selection procedure resulted in 210 respondents from fourteen villages. Using the

‘focal point’ criterion in respondent selection was reasonable. It ensured data collection

from farmers within a reasonable distance of the IPM CRSP area of operation (See

Appendix B3). It is assumed that farmers geographically close to IPM CRSP activities (as

opposed to those further away) are more likely to have had exposure to IPM and would be

more likely to use its practices.

3.1.3 Data Sources, Collection and Transformation Pre-testing of the questionnaire was conducted in March of 2002 by interviewing the active

IPM CRSP farmers. An informal group discussion was held with the district agricultural

officer and sub-county extension personnel from the selected sub-counties in order to

obtain general farming information on the study’s sub-counties prior to formal data

collection. Adjustments were made to the survey following the pre-testing. Several

questions were dropped and others added to ensure the correct format for data collection

and that the final survey questions were appropriate. To avoid respondent bias, and by

way of introduction of the research, a one-page statement of intent preceded the survey

questions (see Appendix C). Twelve sub-county extension personnel with Diploma-level

training in agriculture who were also conversant with the local language of respondents

undertook an intensive one-day training session on data collection techniques prior to the

survey. Each field assistant completed a total of seventeen questionnaires.

Data collection took place in April 2002. Both qualitative and quantitative primary data

were collected by way of open-ended and structured questions administered through

personal interviews with the selected respondents as outlined in the previous section. The

final coded questionnaire is contained in Appendix D. The questionnaire contained 8

sections. Section A obtained demographic information from respondents. Section B

contained background questions including information on general farming practices.

Section C was more specific containing questions about farmer’s knowledge of IPM

practices. Sections D-F had crop specific questions for sorghum, cowpea and groundnuts

respectively, and the IPM CRSP technologies for each. Sections G and H are each one-

question sections requiring information on producer’s use of pesticide and fertilizers and

sources of off-farm income.12 This last section was a modification from a more detailed

section prior to pre-testing.

12 During pre-testing, these were portrayed as ‘sensitive’ questions. Thus the final questionnaire left them as stand-alone questions.

41

Data from the coded questionnaire were transferred into excel spreadsheets. Using excel,

all binary variables represented by a ‘2’ from the questionnaire were recoded as ‘0’ in

conformity with the analytical software requirements (SPSS Inc. 1999) for ease in

interpretation of coefficients.13 Continuous variables were retained in the format of the

coded questionnaire. For non-continuous variables, responses were obtained to

statements about farmers’ perceptions to requirements of various IPM practices in

comparison with conventional practices. Possible responses were: “High”, “Equal”, “Low”,

or “Do not know” while others required: “No”, “Yes”, “Do not know.” In the analysis, some

of these categories were reduced to binary variables (i.e. 1=High, 0 otherwise, or 1=Yes, 0

otherwise) to aid in exposition without altering the basic conclusions (McGuirk, Preston

and Jones, 1992), and where responses were heavily skewed to one side the categories of

the independent variables were collapsed to eliminate ‘zero cell’ situations.14

The dependent variable, a binary response was farmers’ decision to use or not to use a

given practice. For testing multiple adoption decisions, the dependent variable was

obtained by summing the dependent variables for the different technologies and ranking

them to obtain an index of adoption. Thus a ‘0’ for example, represented farmers with no

adoption of IPM technology, ‘1’ for those with one IPM technology, and so on, for each crop

under study.

The potential variables used to explain adoption of the various practices include proxies

for four broad categories: Economic, social, management and institutional factors. To

allow for a comparison of results between different models, a common definition of

variables is adopted. Except for a series of core variables that are common to all models,

the complete set of potential variables in a model varies depending on the IPM technology

being investigated. Table 3.2 defines these variables. The rationale for including each

variable in a model is explained in section 3.4.

13 For binary variables, the value of ‘1’ represents the presence of a condition, while the value of ‘0’ represents the absence of a condition. The advantage of this format is that the average of such a variable is the probability of occurrence. 14 Zero cells in regression models yield point estimates of either zero or infinity which is undesirable and meaningless (Hosmer and Lemeshow, 2000).

42

Table 3.2 Description of variables used in the analyses Variable Name

Type Description [Value]

Economic Factors PEST Discrete Incidence of insects (INSECT)/weeds (WEED)/diseases (DZZ) on

crops [0=No, 1=Yes, 2=Don’t know] HIRE Binary If farmer hires labor [1=Yes, 0=No] FMSZ Continuous Total farm size (ha) YIELD Continuous Crop yield in last season (kg) [SGYD, CPYD, GNYD] FTANY Binary If farmer uses fertilizers on any other crops [0=No, 1=Yes] RACRE Continuous Proportion of total farm acreage under specific crop (ha) FMLBR Continuous Number of family members working on farm OFFLBR Continuous Family members working off the farm INCMSC Binary If farmer has off-farm income sources [0=No, 1=Yes] RFMLBR Continuous Proportion of family members working on farm RSCEREQ Discrete Resource requirements: Management Time (MGT), Labor (LBR),

Land (LND), Cost (COST), Knowledge/Skill (KNOW) for IPM practice (Fertilizer use FTIS, Crop rotation ROTN, Timely planting TPCP, Intercropping ICCP and Close spacing CLSP relative to conventional practices [1=High, 0=Otherwise]

Social Factors AGE Continuous Age of respondent MSF Dummy Farmer’s marital status [0=Not married 1=Married

2=Divorced/Widowed/Separated] HHSZ Continuous Number of household members (Persons) EDUC Continuous Number of years of formal schooling (Years) FMEXP Continuous Length of farming experience (Years) GENDER Binary Gender of farmer [0=Female, 1=Males] RFMEXP Continuous Proportion of farming years to age of respondent Management Factors BFCP Binary Whether farmer borrows to finance crop production

[0=No, 1=Yes] HARM Dummy Perception of hazardous effect of pesticides

[0=No harm, 1=Harm, 2=Don’t know] PURCH Binary Who makes input purchase decisions (Fertilizer, Pesticide, Seed,

Farm implement) [1=Exclusively Males, 0=Otherwise] ONFTR Binary If farmer participates in on-farm trial demonstrations

[0=No, 1=Yes] BFMORG Binary If farmer belongs to a farmer organization [0=No, 1=Yes] OWNIPM Binary If farmer ‘owns’ plots with any IPM recommended practice

[0=No, 1=Yes] VARIETY15 Binary If farmer grew improved variety [0=No, 1=Yes] Institutional Factors EXTS Dummy Frequency that farmer has had contacts with extension staff

[0=None, 1=Few, 2=Many] TRNNG Binary If farmer had had other training in pest control[0=No, 1=Yes] HDIPM Binary If farmer has heard of IPM [0=No, 1=Yes] INFOSC Continuous Number of farming information sources available to farmer INFOTYPE Information from MUK, MAAIF& Farm organizations (RSCH)/

NGOs, neighbors & friends (INFNNF) /radio & newspapers (MEDIA)

15 Only for improved cowpea and groundnut variety adoption models.

43

3.2 Data Analysis Techniques and their Limitations There are several methods of data analysis commonly used in survey work including:

Descriptive analysis including frequencies and means, Chi-square tests of association,

Analysis of Variance (ANOVA), Discriminant analysis, Correlation analysis, Ordinary Least

Squares (OLS), Tobit model analysis and Logit and Probit analyses. These methods are

briefly considered below.

3.2.1 Descriptive Analysis This method of analysis provides statistics that are used to describe the basic features of

the data in a study. They provide simple summaries of the characteristics of the sample

such as measures of dispersion and central tendency. The limitation with this analytical

procedure is that descriptive statistics do not show the relationship among the variables

and the influence that each variable may have on the response. Descriptive analysis does,

however often provide guidance for more advanced quantitative analyses.

3.2.2 Crosstabs Chi-Square Tests Cross tabulations are useful for summarizing categorical variables. The crosstabs chi

square test is used to measure whether there is some level of association among

categorical variables in two-way and multi-way contingency tables. Variables for which the

test statistic is significant at a set cut-off point are considered associated, while those for

which the test statistic is not significant are not associated. However, the test does not

indicate the direction, or even the magnitude of the association, thus it is not sufficient to

use this analytical approach alone.

3.2.3 Discriminant Analysis: This is a procedure for classifying observations into categories based on several

explanatory variables and a classification variable defining groups of observations. This

type of analysis is useful in finding linear combinations of quantitative variables that

provide maximal separation between classes or groups. However, when variables are not

linearly related, it provides inappropriate estimates. In addition, the procedure requires

that the predictor variables have a normal distribution.

3.2.4 Analysis of Variance (ANOVA) The Analysis of Variance is a procedure used to test the difference between two or more

means. ANOVA does this by examining the ratio of variability between two groups and

variability within each group. Its use in limited dependent variable models is limited

because it assumes that the dependent variable is continuous.

44

3.2.5 Ordinary Least Squares (OLS) The OLS approach is the most commonly used method in determining the quantitative

importance of various explanatory variables as they influence continuous dependent

variables. This approach’s limitation in limited dependent variable adoption studies is that

the assumption that the error term is normally distributed does not hold for such

regressions (since it is impossible to have a normal distribution with only a few values of

the dependent variable). As such, the estimated standard errors and t-ratios produced by

an OLS regression are biased. In addition OLS estimates can produce predictions other

than 0 or 1 for the dependent variable in dichotomous choice adoption models. But these

predictions cannot be interpreted as probabilities because probabilities outside the [0,1]

range are undefined. Use of ordinary linear regression in analyzing non-linear adoption

decisions thus gives biased and inconsistent results.

3.2.6 Correlation Analysis This analytical procedure can be used to examine pair-wise associations between

continuous variables. The degree of association is given by the Pearson r coefficient. The

sign of the Pearson r coefficient indicates the direction of the effect of the variables on each

other while the magnitude of the coefficient indicates the strength of effect. A Pearson r of

–1 indicates perfect negative linear association, an r of 0 indicates zero linear association,

while an r of +1 indicates a perfect positive linear association between the variables.

However when the variables are not linearly related, the results are biased. Furthermore,

this pair-wise approach between the dependent and independent variables may overlook

the interactions among variables affecting an adoption decision hence using this approach

alone may provide incomplete information. Nonetheless, correlations may identify variables

that are highly related to each other.

3.2.7 Tobit, Logit and Probit Models Logistic regression models are used when the dependent variable is categorical.

Categorical variables are defined as those for which the measurement scale consists of a

set of categories. For such responses, the use of continuous data analytical methods is

inappropriate. These models include Probit, Logit, and Tobit.

Coefficients of the Tobit model can be disaggregated to determine the effect of a change in

an individual explanatory variable on the probability of adopting. Its biggest advantage is

that the dependent variable can take on a number of discrete values. Several studies

including those on fertilizer adoption by Shakya and Flinn (1985); land enhancing

45

technologies by Baidu-Forson (1999); and sorghum variety adoption by Adesiina and

Baidu-Forson (1995) among others used Tobit models. The dependent variable in these

models is measured as the proportion of area under the improved technology.

Both probit and logit approaches are probabilistic dichotomous choice qualitative models.

These models are statistically similar (Amemiya, 1981), except that the probit model

assumes a normal cumulative distribution function (thus has fatter tails) while the logit

model assumes a logistic distribution of the dependent variable. Although parameter

estimates may differ in the two models because the two distributions have different scales,

Amemiya, (1981) and Agresti (1996), note that it would require enormous sample sizes to

have significant differences in the two models.16 Use of either model is thus discretionary.

Variants of the logit model include the ordinary logit (binary logit), the ordinal logistic,

nominal logistic and the multinomial logit. Binary logistic17 models are the most popular

type because binary data are a common type of categorical data - the response is either a

‘success’ or a ‘failure’. The ordinal logistic regression model is used when the dependent

variable is ordered while nominal logistic handles nominal categorical responses.

Multinomial logistic modeling is a special case of ordinary logistic approach, developed to

address the case where the dependent variable can take on more than two values that are

not ordered.

Probit models lack flexibility in that they do not easily incorporate more than one

prediction variable (Montgomery, Peck and Vining, 2001) unlike logit models. As such

these models are less widely used in limited dependent variable models. However, in

studies such as those by Klotz, Saha and Butler (1995), Fernandez-Cornejo (1996), and

Doss and Morris (2001), probit models were employed. Other studies of factors influencing

insect management technology adoption by Harper et al. (1990); fertilizer adoption by

Kebede, Gunjal and Coffin (1990), pesticide misuse by Tjornhom, (1995) and hybrid Cocoa

by Boahene, Snijders and Folmer, (1999), binary logit models were employed. In all these

studies of adoption, the dependent variable is constrained to lie between 0 and 1. The

16 This is especially true for univariate analysis. Amemiya (1981) argues that in multivariate analysis, the results of the two models may differ significantly 17 Some analytical packages may provide different procedures for ‘logit’ and ‘logistic models’. The main difference between the two is that logit models display coefficients while logistic models display both coefficients and odds ratios. The conclusions from both procedures are the same. In this thesis the term ‘logit model’ is used interchangeably with ‘logistic model’.

46

advantage with these models is that their assumptions are realistic for binary adoption

study data.

3.3 Description of Conceptual Model I Most econometric modeling used to estimate the effect of explanatory variables on the

observed economic phenomena employs linear models:

eXbay ii

n

i

++= ∑=1

(1)

Where y is a continuous random variable, X =X1…Xn are the variables that explain y, a is a

constant and b=b1…bn are the parameters that ultimately describe the effect a change in X

has on y. i denotes the i-th individual and n is the number of observations. But for

adoption decisions, such as adoption of IPM technologies, the random variable y is not

continuous. Instead it can be discrete or dichotomous.

When dichotomous,

p = P(Y=1 X ) (2)

is the probability that Y=1 X and

1-p=P(Y=0 X ). (3)

Y=1 X could, for example, mean adoption of IPM practices and Y=0 X mean non-

adoption given all Xs. Note that this adoption is an end-result of farmers’ decisions based

on economic theory. An economic unit (a farmer in this case) makes rational decisions to

maximize expected utility. The utility associated with each technology is a function of the

possible outcomes from adopting each technology, thus:

U0=f (b [X0]) (4)

U1=f (b [X1]) (5)

Where:

U1 ,U0, are the expected utility levels with and without the technology,

X1 ,X0, are socio-economic and other characteristics of farmers.

b=b1…bn are parameters that describe the effect of farmers characteristics on

utility.

47

When U1>U0, the assumption is that a farmer adopts a technology, or simply, that

Y=1 X in Equation 2. Now by substituting equation 1 into equation 2, it becomes:

p= P( eXban

iii ++ ∑

=1

) (6)

for Y=1 X . Equation 6 can be expressed as

p eXn

iii ++= ∑

=1

βα (7)

There is a similarity between equation 7 and equation 1. The outcome of a continuous

random variable, y, is replaced by p the probability18 of adoption. But equation 7 is linear,

hence would show probabilities of <0 and >1 at low levels and high levels of X respectively.

To ensure that p is positive and restricted to the [0,1] range, equation 7 is reformulated as:

eX

eX

eeXYp

+∑+

+∑+

+==

βα

βα

1)|1( (8)

Where:

p(.) = Probability that an IPM technology is adopted

α = Constant term

X = A set of core explanatory variables

β = A vector of unknown parameters

e = Disturbance term

Reformulation19 of equation 8 yields

eXeXYp

XYp +∑+==−

= βα

)1(1

)1( (9)

This is the odds ratio, or the probability of adoption of IPM packages divided by the

probability of non-adoption. Transforming Equation 9 into a logistic function gives

ln ∑=

++=

=−

= n

i

X eXYp

XYpii

1)1(1

)1(βα (10)

18 The quantitative expression of the chance that an event will occur, or the number of times an event occurs divided by the number of times the event could occur.

19 From simple algebra, if w

wv

+=

1, then )1( vwv −= and w

vv

=−1

48

Equation 10 is also known as the logit(p). By defining p

p−1

as the odds of adoption and

modeling p with the logistic function above, it is equivalent to estimating a linear

regression model where the continuous outcome y has been replaced by the logarithm of

the odds of adoption.20 Thus the logit model is linear in the explanatory variables.

To estimate a logistic model the method of maximum likelihood estimation (MLE) is more

appropriate than Ordinary Least Squares because MLE gives unbiased and efficient

estimates21 (Amemiya, 1981; Agresti and Finlay, 1997). Maximum likelihood finds the

function that will maximize the ability to predict the probability of the dependent variable

based on what is known about the independent variables. Thus, a maximum likelihood

estimate is the value of the parameter that is most consistent with the observed data in

that if the parameter equaled that estimate, the observed data would have a greater

chance of occurring than if the parameter equaled any other possible value.

One major limitation with logistic regression is that the parameter estimates are difficult to

interpret, as the coefficients do not have a direct interpretation. For instance, β in

equation 10 is not the change in probability per unit change in the independent variable.

One way to ease this interpretation issue is to calculate the marginal probabilities for each

parameter estimate. Equation 8 above allows the determination of a change in farmer’s

adoption behavior if the independent variables change by a given amount and is measured

by taking the first derivative of Equation 8:

∑+=

∂

=∂+

+∑

XY

e

eX

XYpeXb

eXb

2)1(

)1( (11)

Which is,

= )]1(1)[1( XYpXYp =−=β (12)

Thus, the marginal probability for the logistic distribution is the parameter estimate for

the logit multiplied by a standardization factor. The standardized factor is the probability

of adoption multiplied by the probability of non-adoption and is given by:

20 Or the probability that an event occurs divided by the probability that the event does not occur 21 Maximum likelihood is an iterative process that starts out with a ‘guestimate’ of what the logit coefficients should be and determines the direction and size of the change in logit coefficients which will increase the log likelihood. The process continues until there are very small improvements in the log likelihood.

49

[ ]2)exp(1()exp(β

βX

X+

(13)

This change in probability of adoption is not constant. It increases or decreases depending

on the value of X. Thus for a continuous variable X, at relatively high values, a large

change will give a relatively smaller change in the probability of adoption (McCullagh and

Nelder, 1989). For instance where farmers’ characteristics have a mean value and p=0.5, a

straight line drawn tangent to the logistic curve (Figure 3.2) has a slope 0.25β given fixed

levels of the other Xs.22 When the value of X is larger than the mean, say p=0.9 (or smaller

than the mean, say p=0.1), then the change in probability is given by 0.09 β which is a

smaller value. A convenient measure is to evaluate the change in probability at the sample

means of the explanatory variables (Pascale, 1998).

Probability (p)

1

0Variables (X)

Fig 3.2 Logistic regression curve for [0,1] response models (Source: Agresti and Finlay, 1997 p.577) However when the independent variables are not continuous, several authors (see Pascale,

1998) suggest a different method for obtaining marginal probabilities. The change in the

probability of a success (Y=1) that results from changing X from zero to one, holding all

other variables at some fixed values, denoted by X*, is given by the difference:

P(Y=1 | X=1, X*) - P(Y=1 | X=0, X*) (14) 22 When p=0.5, the odds p/(1-p) = 1 and the logit log[1]= 0. At that point the slope would be 0.25 β , and X= -

α / β . This is also the largest slope of any logistic curve. Simply put, a one unit increase in X relates approximately

to a 0.25 β increase in the probability of adoption holding other factors constant. The predicted probability of

adoption of IPM technologies i s below 0.5 for X values less than -α / β and above 0.5 for X values greater than -

α / β . This X value is sometimes called the median effective level EL50 because it represents a level at which each outcome has a 50% chance. Values of p range from (0 ≤ 1), odds are in the range of (0 ∞≤ ), and logits (- ∞≤∞ ). An odds ratio r =p/(1-p) of 1 means that both adoption and non-adoption have equal chances of occurring. This text is adapted from Agresti and Finlay (1997, p.578).

50

Where X* is the value of all other variables in the model. An approach commonly used is to

set values of X* to represent a "typical case." A "typical case" is defined by setting all

dummy variables to their modal values and all other variables to their mean values

(SHAZAM, 2002).

The sign of an estimated coefficient gives the direction of the effect of a change in the

explanatory variable on the probability of a success (adoption in this case). Whenβ >0, the

probability of adoption increases as the level of factors (X) increases. When β =0, the

binary response is independent of X. For different variables, the rate of change increases

as the absolute magnitude of β increases.

The success of any logistic regression is assessed by establishing several tests: the correct

and incorrect classifications of the dependent variable, the Wald statistic, the pseudo R-

square measures, and the likelihood tests. The difference between the log likelihood of the

constrained model (Log Lo) with only the intercept term in the model and the log likelihood

of the unconstrained model (Log Lmax) is the model chi-square which tests the significance

of the logistic model. Several pseudo R2 measures exist including the McFadden’s R2 also

known as the likelihood ratio index (LRI) and is obtained by (1-Log Lmax/Log Lo). Low

values of McFadden’s R2 are typical in logit models. The number of iterations in a logistic

model is an indication of presence of multicollinearity problems. For a normal dataset

convergence is reached in 4-5 iterations. The Wald statistic is a measure of independence

of response and predictor variables. It is obtained by squaring the ratio of the logit

coefficient to its standard error. Wald statistics tests that individual independent variables

have no effect on the probability of the response occurring. In other words, that the

probability of adoption of IPM technologies is independent of the variables (X’s) in the

model.

3.4 Empirical Model I Farmers’ decision to adopt or not to adopt a technology is assumed to be the outcome of a

complex set of factors related to the farmers’ objectives and constraints. In other words,

there are certain factors – including market forces, social, institutional, and management

factors that affect the likelihood that farmers adopt a technology. Thus if each farmer and

each technology can be classified based on a core set of variables, then it is possible that

the probability of a farmer adopting that technology could be estimated.

51

3.4.1 Explanation of Variables and Apriori Expectations Variables used in the models are discussed in the following section. The variable

abbreviation used in the model appears in parenthesis following the more descriptive

variable name.

Age (AGE) As farmers advance in age, risk aversion increases and adopting a new

technology seems less likely. This variable is expected to negatively affect the adoption of

most technologies.

Experience (FMEXP) This is measured by the number of years of farming. Experienced

farmers are assumed to have tried out a number of profitable technologies. Hence the

variable is expected to positively affect close spacing, improved variety, early planting, and

fertilizer use but negatively affect intercropping adoption because of this practice’s

relatively low skill requirement.

Relative Experience (RFMEXP) Age and experience are not the same. The two variables

hence have distinct influences on farmer’s adjustment to change. In this analysis another

variable is introduced. It is measured by the number of years of farming as a percentage of

farmer’s age. Older farmers with say 10 years of farming experience and younger farmers

with the same length of farming experience have different relative farming experience.

Higher relative experience will be positively associated with adoption of improved varieties

and less for labor-intensive practices.

Years of Formal Education (EDUC) Respondents’ exposure to education will increase the

farmers’ ability to obtain, process and utilize information relevant to the adoption of IPM

technologies. More education is expected to reduce a producer’s information acquisition

costs. Again, with the exception of intercropping and close spacing, this variable is

expected to have a positive influence on IPM technology adoption. These two practices do

not heavily draw on educational qualities of farmers and thus their adoption is not likely

to be positively correlated with education.

Family Labor (FMLBR) A large number of family members (relative to household size)

working on the farm reduces the farms’ external labor requirements and is hence assumed

to positively affect adoption of labor-intensive IPM technologies such as close spacing.

52

Off Farm Labor (OFFLBR) A large number of family members working off the farm for

extra income can increase financial resources available on the farm. It is anticipated that

the effect of this variable on adoption of all practices will be positive.

Respondent’s Gender (GENDER) Female and male farmers are likely to play different

roles in technology adoption, depending on the nature of the technology. The effect of this

variable is indeterminate.

Number of Information Sources (INFOSC) Sources of information are relevant in

adoption as sources expose the potential adopter to the new technology. The direction of

effect may be mixed. For this study, access to outside information is expected to be

positively correlated with IPM adoption.

Type of Information Source (INFOTYPE) Information from various sources may have a

different impact on farmer’s perception of farming practices. For ‘exotic’ practices such as

using improved seed or intercropping sorghum with exotic celosia, information from

research sources such as MUK, MAAIF and farmer organizations (RSCH) may be more

influential than information from neighbors and friends (INFNNF) or from media sources

(MEDIA). These latter sources however may be important in adoption of practices like crop

rotation, timely planting and close spacing.

Perception of Damage by Pesticides (HARM) A negative perception on the effect of

pesticides is likely to positively influence adoption of all technologies that do not have a

chemical component.

Total Farm Size (FMSZ) Farmers’ total land holding may serve as a good proxy for wealth

and status and income levels. This variable is likely to have a positive effect on adoption of

most practices.

Area under Crops (RACRE) The proportion of land allocated to a specific crop signifies the

importance of that crop to the farmer relative to the others on the farm. For important

crops, adoption is expected to be high therefore this variable is expected to be positively

correlated with adoption.

53

Crop Yield (YIELD) Higher anticipated yields from a crop may increase the probability of

adoption of even more yield-enhancing technologies. Therefore variables GNYD (groundnut

yield) and CPYD (cowpea yield) are expected to be positively correlated with adoption of

improved varieties of groundnut and cowpea respectively.

Pest Incidence (WEED, INSECT, DZZ) The incidence of pests on farmer’s fields presents

an economic problem that highlights the need for adoption of control strategies. Thus this

variable is anticipated to have a positive relationship with adoption. However, ex-post, that

is, after farmers adopt the practice (and the problem has been solved) this variable may be

negatively related increased adoption.

Frequency of Extension Contacts (EXTS) Extension is a source of information about

better farming practices. Frequent extension contacts are expected to positively impact

adoption of all IPM technologies on cowpea, sorghum and groundnuts.

3.4.2 IPM Packages on Sorghum, Cowpea, and Groundnuts Because crops are different in nature and different pests attack them, the IPM CRSP

developed different control strategies for the different crops. Three technologies on

sorghum, three on cowpea and two on groundnuts are analyzed in this study (Figure 3.3).

The individual practice may not be a new phenomenon, however, the particular

combination into a set of practices for pest control is a “new idea” developed and

disseminated by the IPM CRSP. Consequently the proposed models differ slightly based on

the specific characteristics of the technology for each crop. Variables hypothesized to have

theoretical importance to each model are used in the exploratory stage of model building.

In this section models for each crop and each technology are developed and described.

Sub-section 3.4.3 considers the models for sorghum, 3.4.4 considers those for cowpea and

3.4.5 outlines groundnut models.

54

Fertilizer applicationCelosia

Crop rotationSeed coating with herbicides*

Sorghum

Seed dressing*Defoliation*

Minimum spray schedule*Close spacing*

IntercroppingEarly planting**

Improved variety**

Cowpea

Close spacingImproved varietyTimely planting*

Minimum spray schedule*

Groundnuts

Integrated Pest ManagementPackage

* IPM technologies not investigated23 ** Non-IPM CRSP pest control technologies investigated Figure 3.3 Components of the IPM packages on Cowpea, Sorghum, and Groundnuts.

3.4.3 Sorghum models: Three technologies developed for striga control that are examined in this study are:

Fertilizer application in sorghum fields, intercropping sorghum with celosia argentia, and

crop rotations involving legumes such as cowpea and groundnuts. The following models

are proposed for the preliminary analyses.

Model specification24

I a) Fertilizer Use in Sorghum (FTIS)

The sorghum model can be summarized as follows:

FTIS= 0β + 1β COST+ 2β INCMSC+ 3β FTANY+ 4β FTISLBR+ 5β MGT+ 6β WEED+

7β VARSKDO (15)

Fertilizer use in sorghum is expected to be inversely related to its cost, its high skill,

management time and labor requirements, but positively related to the ability to use it on

any other crops (FTANY), the importance of the crop to be protected (VASKDO), weed

infestation (WEED) and the ability to pay for it (INCMSC).

I b) Intercropping sorghum with celosia argentia (ECAT)

Striga control using an exotic legume celosia is anticipated to be related to sources of

information, contacts with extension personnel, researchers and active participation of

farmers in pest control activities.

23 These practices were mostly found to have either 100% adoption or 0% adoption. In this case, the dependent variable becomes a constant, hence does not provide enough variability in models if used. In addition, the fitted probability is either zero or one and this leads to failure to converge (McCullagh and Nelder, 1989). Finally, some of these practices are “new” and it is too early to evaluate their adoption. 24 The variables are defined in Table 3.2 and explained in detail in section 3.5.

55

ECAT= 0β + 1β INFOSC+ 2β ONFTR+ 3β OWNIPM+ 4β TRNNG+ 5β EXTS (16)

I c) Crop rotations involving legumes (ROTN)

Crop rotation adoption is expected to be influenced mainly by social and economic factors

because of the more “indigenous” aspect of this practice. Institutional and management

aspects of farmers are not expected to be influential.

ROTN= 0β + 1β VARIETY+ 2β OFFARM+ 3β MGT+ 4β COST+ 5β LBR+ 6β LND+

7β FMEXP (17)

3.4.4 Cowpea models A combination of practices involving close spacing of seed, accompanied by 3 sprays at

budding, flowering and podding, in addition to intercropping cowpea with cereal crops

including sorghum were found to significantly reduce pest infestation in cowpea fields. As

explained earlier, defoliation, seed dressing, close spacing and insecticide application on

cowpeas are not analyzed in this study. Although early (timely) planting and growing

improved cowpea variety are not practices disseminated by IPM CRSP on cowpea, farmers’

potential to adopt these practices, may be an indicator of their responsiveness to other

technological changes, of which IPM is part. Therefore, adoption of these technologies was

investigated. The following are the hypothesized models:

Model specification

II a) Intercropping cowpea with cereal crops (ICCP)

Intercropping is expected to be influenced by the type of information sources, the presence

of weed problems and availability of hired labor on farms.

ICCP= 0β + 1β MEDIA+ 2β WEED+ 3β OWNIPM+ 4β HIRE (18)

II b) Timely planting (TPCP)

Planting early at the onset of the planting season is anticipated to be positively related

availability of labor and adequate land at planting time, frequency of extension contacts

and farmers perception of the harmful effect of chemicals in pest control as compared to

cultural methods.

TPCP= 0β + 1β HARM+ 2β EXTS+ 3β TPCPLBR+ 4β TPCPLND (19)

II c) Improved cowpea variety (ICPV)

It is expected that growing an improved cowpea variety is positively influenced by access to

information from researchers, the incidence of diseases and availability of off-farm income

sources to purchase improved seed.

56

ICPV= 0β + 1β RSCH+ 2β DZZ+ 3β INCMSC (20)

3.4.5 Groundnut models Groundnut research developed the following technologies for groundnuts: Close spacing,

early planting, minimum spray schedules and planting groundnut rosette resistant

varieties. These are rather inexpensive technologies; therefore chances that relatively less

wealthy farmers may adopt them could be high.

Model specification

III a) Close spacing (CLSP)

CLSP= 0β + 1β BFMORG+ 2β INCMSC+ 3β IGNV+ 4β TRNNG (21)

III b) Improved Variety (IGNV)

IGNV= 0β + 1β HHSZ+ 2β EXTS+ 3β FMLBR+ 4β RFMLBR+ 4β RSCH (22)

3.5 A Two Tiered Analytical Process The nature of IPM is that an approach often consists of a package of component

technologies, and this nature calls for a two-tiered analytical process of IPM adoption.

Each set of technologies is considered effective in controlling a pest, or a set of pests, and

could thus be individually adopted. Therefore, rather than require that all components of

an IPM package must be employed in order to consider a farmer an adopter, the

discussion above considers the adoption of each practice individually. This approach

allows the use of simple logit models to estimate the relationships between various factors

and the decision to adopt or not to adopt.

However, this approach alone is not sufficient for examining the extent and intensity of

adoption. Feder, Just and Zilberman (1985) argue that adopters do not have a binary

choice; that there are varying stages of adoption, hence variations even in the class of

adopters. In fact, adopting farmers may choose to adopt a subset of the technological

package, or all of the components of a package. In such a case, the use of dichotomous

models may misrepresent decisions made by these farmers.

A second approach hence considers that technologies can be complementary. From 3.4.3

above, for example, a sorghum farmer can be said to be an adopter of intercropping,

celosia, fertilizer individually, or a combination of one or more other practices. These

options are possible because farmer’s decisions to use these practices need not be

57

simultaneous or sequential (unlike in Kebede et al., 1990). In addition, although

technologies are parts of an overall IPM package, they are not necessarily technically

interdependent. In this study thus, unlike several adoption studies, a two-tiered process of

analysis is employed first, to identify adopters and non-adopters of a single technology

(explained in section 3.4. above), and then within the class of adopters, to consider the

level of adoption.

3.5.1 Description of Conceptual Model II When the outcome of y is able to assume a set of discrete ordinal categories such as 1, 2,

3, etc (also known as multicategory or polytomous responses), binary logistic regression

alone is inappropriate to use to estimate model parameters25 (Agresti and Finlay, 1997).

For such responses, cumulative (Ordinal) logit analysis that incorporates orderings in

responses potentially has a greater power to explain behavior than the ordinary

multicategory logits (Agresti, 1996) and is thus considered adequate to ensure that the

statistical assumptions of the model are met.

Suppose the dependent variable can take on three values: 1 (one technology adopted), 2

(two technologies), 3 (three technologies adopted), and let

p1=P(Y=1) and p2=P(Y=2), p3=P(Y=3) (23)

The ordinal logistic regression models the relationship between the cumulative logits of Y,

that is,

log

− 1

1

1 pp

= log

+ 32

1

ppp

and (24)

log

+−

+)(1 21

21

pppp

= log

+

3

21

ppp

(25)

The model assumes a linear relationship for each logit (like in ordinary logit) but with

parallel regression lines, so that for each cumulative logit the parameters of the models are

the same except for the intercept a.

log

− 1

1

1 pp

= a 1 + bX∑ (26)

25 Described as dependent variables with more than two categories of response.

58

log

+

3

21

ppp

= a 2 + bX∑ (27)

From the above equations,

− 1

1

1 pp

=exp( 1a + bX∑ ) and (28)

+

3

21

ppp

= exp( 2a + bX∑ ) (29)

=

− 1

1

1 pp

*12

aa

(30)

Equations 28-30 imply that the odds ratio for Y=1 versus Y=2 or 3 and for Y=1 or Y=2

versus 3 are the same. If parameter b>0 then p1, the predicted probability of (Y=1) as well

as cumulative probability of (Y=1 or Y=2), p1+p2, are higher for higher values of X. If b<0,

p1 and p1+p2 are lower for higher values of X.

The estimation of a cumulative logit model employs maximum likelihood estimation after

transforming the dependent variable (the decision and level of adoption) into a logit

variable (the natural log of the odds of adoption occurring or not) like in an ordinary

logistic regression. In this way, the model estimates show changes in the log odds of

adoption, and not the changes in adoption itself as OLS regression does.

3.5.2 Empirical Model II This latter approach gives technology adoption indices based on the level of adoption of

various technologies. The hypothesized relationship that exists between factors affecting

adoption of sorghum (SGTECHS), cowpea (CPTECHS) and groundnut (GNTECHS)

technologies can be represented thus:

SGTECHS= f (INFOSC, TOTCROPS, YDSKDO, GENDER, BFMORG, BFCP,

ONFTR, HDIPM, TRNNG, RSCH, WEED, OWNIPM) (31)

CPTECHS= f (AGE, HHSZ, FMLBR, EBYD, EBACRE, INCMSC, BFMORG, SEEDPURCH,

IMPLPURCH, ONFTR, OWNIPM, EXTS, RSCH, SSNMGT, SSNCOST, SSNKNOW,

INTCRPKN, ICCPLND, EBMGT, INSECT, DZZ, WEED) (32)

59

GNTECHS= f (HHSZ, FMLBR, TOTGNYD, IGOLAYD, INFOSC, INCMSC, BFMORG,

EXTS, HDIPM, INFNNF, RSCH, SPACEKN) (33)

The dependent in the CPTECHS model is a multi-category variable with ‘1’ representing

one technology adopted, ‘2’ and ‘3’ representing respective levels of technologies adopted.

For sorghum and groundnut technologies the multi-category dependent variables

SGTECHS and GNTECHS take on two ordered values (1,2).

3.6 Collinearity Diagnosis Linear dependencies among regression variables make it difficult to separate out the

unique role of each independent variable on the response variable. Each independent

variable may be nearly redundant in the sense that it can be predicted well using the

others. This is the problem of multicollinearity. This problem may dramatically impact the

usefulness of a regression model and lead to inappropriate conclusions being drawn from

incorrect parameter estimates and confidence intervals, particularly, small changes in the

data values may lead to large changes in the estimates of the coefficients. The problem is

more likely to arise the more independent variables that there are in the model. It is thus

important to test for multicollinearity and remedy it prior to regression modeling.

Although simple correlations between continuous variables and associations between non-

continuous variables provide some guidance about potential multicollinearity, the

preferred method to test for multicollinearity is the use of the Variance Inflation Factor

(VIF). VIF is the reciprocal of tolerance and is given by 1/(1- R2). When the VIF is high, the

R2 value is high and the interpretation of the coefficients becomes unreliable. Although

logistic regression lacks a perfect analogue for R2 as in OLS, Garson (1998), Allison, (1999)

suggest the use of collinearity diagnostic statistics produced by linear regression analysis.

Most diagnostics use a VIF of >10 (tolerance <0.1), as rule of thumb. However, for survey

type research, Garson (1998) recommends a more stringent VIF of ≤ 4 (tolerance ≥ 0.25) to

test for multicollinearity. Hosmer and Lemeshow (1989) suggest examining values of

estimated standard errors, and estimated slope coefficients. Very large values of these

coefficients are an indication of multicollinearity.

60

3.7 Model selection The number of regression models to be evaluated when building the best possible model is

often large when the set of candidate regressors is large. Although a number of variables in

the model may serve equally well as predictors, it is redundant to use them all as they

make the model more difficult to interpret. In addition, large models have high potential

for multicollinearity (see above, and Agresti, 1996). To avoid these problems, formal

variable selection is necessary to ensure retention of a model that has a limited number of

explanatory variables that are useful in explaining the relationship. These selection

procedures ensure that the final model contains the best26 variables.

In linear regression models, two major methods include the ‘all possible regressions except

the intercept’ and the automated selection procedures such as Forward selection,

Backward selection or step-wise selection. Selection is based on values of R2, R2adj.27, and

Mallow’s Cp statistic. A high R2-value shows that the portion of response explained by the

variables in the model is high. The Cp statistic describes how well each model fits

compared to the full model with all the predictors (P denotes the number of parameters in

the model) and the lower the Cp value the better the model. Backward elimination involves

a step wise approach where the procedure tries to remove the most non-significant

regressor. This process starts with a complex model progressively eliminating variables

that have the largest P-value thus retaining variables that make significant partial

contributions to predicting the outcome. Forward selection methods work in the opposite

direction. Step-wise procedures combine both methods.

However, these automated selection procedures have been criticized because they are not

guaranteed to find the correct (most practical) model. And as many authors (Hosmer and

Lemeshow, 2000; Agresti and Finlay, 1997) suggest, the use of computer algorithm alone

to select variables for the final model is inappropriate. Inclusion of certain variables of

special interest even when they are not statistically significant may be more important

than reliance on computer-generated models. With non-linear regression models, use of R2

content procedures may not be possible. As such, model selection procedures in logistic

regression mainly involve stepwise procedures through likelihood ratio tests.

26 ‘Best’ here is used to refer to a model that is not too small as to be underfit and not too large as to be overfit. Underfit models miss important information and may be biased while overfit models may suffer the undesirable effects of multicollinearity. 27 R2adj. corrects R2 to closely reflect the goodness of fit of the model.

61

Clearly, there are many methods of building a regression model. Some authors recommend

beginning with a large model and reducing it to a simpler model with fewer explanatory

variables, while others recommend starting with a more parsimonious model, increasing

its size to incorporate other variables of importance. Hosmer and Lemeshow (1989)

recommend this latter approach through the procedure outlined below:

(1) Univariate analysis; selection of variables of significant relationship individually

with the response variable using chi-square tests of significance, two-sample mean

tests, crosstabs symmetric and directional measures.

(2) Multivariate analysis using all variables retained from the univariate analysis step

and including other variables of theoretical significance,

(3) Elimination of unimportant variables from the multivariate analysis step (using

Wald Tests and Likelihood ratios).

This is the approach followed in the current study. In particular, for this study Figure 3.4

illustrates the model building process.

START

Test for Multicollinearity

Any Highly Correlated Variables?Yes Use Judgement

Drop Variables

Use each Variable Separately to Explain the Response

Run Estimation using all Significant Variables

Run Final Model

Drop

DropNo

YesYes

Are Any Significant?

Does Explanatory Power Improve? Are any Theoretically Important?NoNo

Yes

Figure 3.4 Model building procedure

62

The procedure ensures retention of variables that explain the underlying complexity with

the simplest model.

The criterion used for entry or removal of a variable from a model is the α level. A cut-off

of 0.05 is the preferred significance level for most research. However use of a less stringent

α level may be necessary to allow more variables to become candidates for inclusion in the

multivariate model and to ensure the selection procedure does not stop too early resulting

in underfit models (Montgomery, Peck and Vining, 2001). Regressors with p-values greater

than α are considered for dropping out of the model.

3.8 Analytical software Two statistical packages are used to take advantages of different features in both

programs. Statistical Package for Social Scientists (SPSS) is especially useful in obtaining

descriptive statistics, comparison of means and running the multivariate logit model. The

Statistical Analysis System (SAS) package was used to test for collinearity and

multicollinearity diagnostics and running cumulative logistic models.

3.9 Summary This chapter provided a background description of the study area, an explanation of how,

and what data was obtained, and analytical methods used to obtain results for the thesis.

Based on this framework, the chapter highlighted the major strengths of this thesis:

(i) The broad number of technologies analyzed (Figure 3.3)

(ii) The extensive number of explanatory variables considered (Table 3.2), to make

the empirical model more reliable,

(iii) The use of primary data in the analysis, and,

(iv) The study’s sufficiently large sample, necessary for reliability in detecting partial

effects of variables on the response using maximum likelihood techniques.

63

Chapter 4 Results

This chapter is presented in five sections. Section 4.1 summarizes the general descriptive

information from the survey; 4.2 gives results of running univariate empirical models and

4.3 gives multivariate estimations, 4.4 gives results of model fitting procedures. Section

4.5 gives empirical results for conceptual model II of chapter 3, and 4.6 is a concluding

section. Each section is organized into three sub-parts, corresponding to each of the crops

in the study: sorghum, cowpea and groundnuts.

4.1 General Descriptive Analysis

Survey responses were obtained from 212 respondents (104 females and 108 males) from

16 villages in 8 parishes28 with a combined farming experience of 4,409 years and 2,090

acres under various crops. Sorghum, cowpea and groundnut acreage accounted for 13%,

12% and 18% of farmers’ total cropland respectively.29 Total acreage owned by producers

was approximately 1,412 acres indicating that producers in the study area rented a large

proportion (at least 32%) of the land they cropped. Tables 4.1-4.3 summarize some of the

descriptive statistics for the sample.

Average household size (HHSZ) was 8.17 people. The mean farm labor (FMLBR) of 4.29

household members shows that over half of the household members worked on the farm.

Twenty eight percent of the producers borrowed to finance crop production (BFCP), while

those who did not borrow cited credit unavailability as a major obstacle to credit

acquisition. Information pertaining to agriculture was obtained from a number of sources

including the Ministry of Agriculture staff (mainly sub-county agriculture/extension

assistants), from friends, neighbors, and the media (radio). Other information sources

included farmers’ organizations, bulletins, newspapers, Makerere University researchers,

and NGOs. Thirty two percent of producers belonged to a local farmer’s organization

(BFMORG). Over 87% of farmers hired laborers (HIRE) to work on their fields. However,

payment for this labor took on many forms including payment in kind (live animals, part

of farm produce, local brew), cash or exchange of labor.

28 The original sample size was 210 (see Chap. 3). However, two farmers from two different villages in neighboring Parish asked to be interviewed. Thus in total 212 farmers were interviewed. 29 1acre = 0.4047ha

64

Significant differences exist between levels of education (EDUC) for males and females,

with men having more years of formal education (Table 4.1). The means of the other

continuous variables are not significantly different by gender at α =0.05. At a less

stringentα of 20%, significant differences exit between sorghum acreage (SGACRE) owned

by females and males. As shown in Table 4.2, membership in farmers’ organizations

(BFMORG), borrowing to finance production (BFCP), and pesticide use on other crops

(PTANY) were significantly different for female and male farmers with more females

belonging to organizations (39% vs. 24%) and borrowing (32% vs. 24%) than males, while

on the other hand, males were more inclined to use pesticides (50% vs. 42%) on crops they

grew than females. None of the other categorical variables presented in Table 4.2 showed a

significant difference by gender.

Half of the respondents had heard of the term Integrated Pest Management30 (through

MUK, MAAIF, farmer organizations, and others), and were aware of its benefits and

requirements although only 66 (61.7%) of those who heard of IPM had used IPM

recommendations on the crops they grew. Although 84% of farmers (including those who

had not previously heard of IPM) agreed that pesticides were harmful to crops, animals,

birds, humans and other living creatures, they generally perceived IPM practices to require

more knowledge, more labor, more cost and more management time than current

practices. Indeed a good number, (93%) applied pesticides on crops they grew. Note that

14% of the respondents had participated in some form of on-farm trials (ONFTR). Testing

for association between farmers’ perception of harm by chemicals (HARM) and their gender

found no significant differences.

30 Farmers had varying descriptions of their practices, some of which were in fact, IPM practices.

65

Table 4.1: Summary statistics of continuous variables

Total (N=212) Females (n=104) Males (n=108) Variables Mean31 Std. Dev. Mean Std. Dev. Mean Std. Dev. Sig.32 AGE (Years) 39.9 15.2 39.31 14.89 40.45 15.55 0.584 EDUC (Years) 5.62 3.83 4.94 3.8 6.28 3.77 0.011 FMEXP (Years) 20.8 13.6 21.57 13.56 20.06 13.61 0.419 INFOSC (Number) 3.33 1.67 3.34 1.81 3.31 1.53 0.925 HHSZ (No. of people) 8.17 4.82 8.1 4.85 8.23 4.82 0.839 FMLBR (No. of people) 4.29 2.5 4.1 2.38 4.47 2.61 0.264 OFFARM (No. of people) 1.31 2.15 1.26 2.12 1.36 2.18 0.732 FMSZ (Acres) 6.66 6.06 6.56 6.29 6.76 5.85 0.803 SGACRE (Acres) 1.12 0.62 1.05 0.56 1.19 0.66 0.112 CPACRE (Acres) 1.29 0.79 1.22 0.75 1.36 0.82 0.207 GNACRE (Acres) 1.75 1.22 1.75 1.22 1.76 1.22 0.935 TOTSGYD (Bags) 2.02 1.61 1.91 1.49 1.71 2.02 0.357 TOTCPYD (Bags) 1.73 1.73 1.62 1.92 1.52 1.51 0.676 TOTGNYD (Bags) 6.67 5.31 6.61 5.11 6.72 5.51 0.878

Table 4.2: Summary statistics of non-continuous variables33 Total Females Males

Variable (%) (%) (%) Sig.34

INCMSC (% yes) 65.0 65.0 64.0 0.467 PTANY (% yes) 46.0 42.0 50.0 0.162 FTANY (% yes) 6.00 6.70 4.60 0.358 BFMORG (% yes) 32.0 39.0 24.0 0.012 HIRE (% yes) 85.0 84.0 86.0 0.379 BFCP (% yes) 28.0 32.0 24.0 0.138 ONFTR (% yes) 14.0 13.0 15.0 0.466 OWNIPM (% yes) 31.0 30.0 32.0 0.398 HARM (% yes) 84.0 80.0 88.0 0.202 HDIPM (% yes) 50.0 45.0 56.0 0.206 TRNNG (% yes) 26.0 23.0 29.0 0.352

During pre-testing of the questionnaires all farmers said that they could not provide an

accurate estimate of farm and off-farm income due to the irregular nature of sales and off-

farm work. These questions were thus eliminated from the survey. However, some income

information was obtained. A significant number of respondents (65%) had sources of

income (INCMSC) outside the farm including petty trading, selling local brew, and

remittances from family members who work off the farm. Cross tabulations by gender

31 Values indicate the mean of the quantitative attribute. 32 Testing difference in means between males and females 33 Values of categorical variables indicate the proportion of farmers taking on particular qualitative attributes. 34 Testing for significant differences between males and females

66

revealed no significant association between gender and off-farm income source (Table 4.2).

Appendix Table E1 shows the frequency table of income sources in the study area.

Farmers grew a wide range of crops. All sampled farmers grew groundnuts on plots

ranging from 0.1 to 6 acres. Table 4.3 below shows the summary statistics of crops grown.

Highest total acreage was in groundnuts with a mean of 1.85 acres per farmer. Grain

crops including sorghum, cowpea, groundnuts, millet, maize, beans, greengrammes (a

popular leguminous crop called ‘Choroko’), rice, sunflower, soybean, simsim (sesame)

represented over 67% of the farmed area.

Table 4.3: Crops Grown: Summary Statistics Total area

in crop Min area

Max area Mean area

Crops

Percent35 sampled farmers

growing crop (%)

Acres Acres Acres Acres

Std. Dev. of mean

Sorghum 99.06 262.00 0.10 4 1.236 0.62 Cowpea 98.58 278.88 0.13 6 1.315 0.78 Groundnuts 100.00 392.68 0.10 6 1.852 1.22 Cassava 96.23 383.50 0.50 8 1.809 1.27 Millet 78.30 245.53 0.10 7 1.158 1.13 Potatoes 68.87 164.33 0.10 8 0.775 0.94 Maize 60.38 143.30 0.20 8 0.676 0.87 Beans 21.70 35.30 0.25 3 0.167 0.42 Greengrammes 15.09 37.15 0.25 5 0.175 0.55 Cotton 13.68 42.50 0.50 8 0.200 0.76 Rice 13.21 26.90 0.25 2 0.127 0.35 Sunflower 10.85 27.00 0.50 3 0.127 0.43 Tomato 8.96 8.60 0.02 1.5 0.041 0.17 Soybean 5.19 9.15 0.01 2 0.043 0.22 Eucalyptus 3.77 7.20 0.20 3 0.034 0.25 Simsim 2.83 5.00 0.50 1 0.024 0.14 Banana 2.36 1.86 0.01 1 0.009 0.08 Others 5.66 19.55 0.25 8 0.092 0.61

4.2 Adoption of IPM Practices - Univariate Analysis

Figure 4.1 below contains summary statistics some of which relate to components of some

of the IPM practices. Eighty eight percent of farmers did not purchase fertilizers. In

contrast 93% of farmers purchased pesticides, and 96% purchased seed. While more

men than women are sole purchasers of farm supplies, if the percentages in the “both”

35 The high percentages of people growing sorghum, cowpea and groundnuts are expected since the selection procedure targeted these farmers. Nonetheless, these crops’ popularity in the area can be seen from the mean acreage devoted to these crops relative to other crops in the farmer’s fields.

67

category are added to the “female” category over all it might be reasonable to say that men

and women have about the same levels of purchasing activities.

Pesticide Purchase

Males46%

Females9%

Both 38%

Do not purchase

7%

Seed Purchase

Don’t Buy4%

Females15%

Males39%

Both42%

Fertilizer Purchase

Both3%

Females3%

Males6%

Do not purchase

88%

Figure 4.1 Farm input acquisition: distribution of purchase decisions

4.2.1 Sorghum

Farmers generally rotated sorghum some with groundnuts (34%) and with cassava (20%).

The crop was grown mainly in the second season. Popular sorghum varieties included

Seredo, Sekedo - the two improved local varieties, Eidima, other local varieties (that

farmers could name) and other unnamed varieties. Figure 4.2 below shows the

distribution of these varieties. A complete list of local varieties grown in Kumi is presented

in Appendix Table E2

Distr ibut ion of Sorghum Var iet ies

Local var iety named

49% (135)Eidima

8% (22.1)

Sekedo19% (45.6)

Seredo13% (33)

Variety not known/named11% (26.3)

Figure 4.2: Distribution of sorghum varieties as a percent of total sorghum acreage (values in parentheses are total acreages by variety) for the sample

68

Assuming that land attributes and other factors were the same across varieties, Sekedo

had the highest productivity (with 1.76 bags/acre) compared to Seredo (1.57 bags/acre) or

Eidima (1.33 bags/acre). Farmers perceived that growing improved varieties involved high

costs (60%), and had high knowledge requirements (43%) but also felt that labor (87%),

land (83%) and management time (92%) requirements were less than or equal to those for

local varieties.

In the crop season preceding this survey, 66% of sorghum farmers reported that their crop

was harmed by insects, mainly sorghum shoot fly (Atherigona soccata) (41%) and stem

borers (Chilo partellus, Busseola fusca and Sessamia calamistis) (22% ). Notable in Table

4.4 below is that although the occurrence of problems was slightly higher with insects,

more farmers focused on controlling weeds than insects. This finding might suggest that

insects are not destructive during sorghum development, that is, why even though

incidence is higher compared to weeds and diseases, controlling them is not as high a

priority to farmers as controlling weeds is. The major weed in sorghum fields was striga

found in 40% of farmers’ fields . Other common weeds included star grass (Cynodon

dactylon), spear grass (Imperata cylindrica), couch grass (Digitaria scalarum) and goat weed

(Epimedium grandiflorum). Appendix Table E3 gives the complete list of weed species

reported in farmers’ fields.

Control strategies on striga varied tremendously. Celosia argentia, an exotic legume that is

reported by farmers to “chase” striga was a fairly new strategy and was employed by about

3% of respondents. However, physical weeding, rotation with cassava (migyera variety),

sweet potatoes and other unidentified local weeds were common striga control practices.

Also effective against striga is fertilizer use and crop rotations involving legume crops like

cowpea and groundnuts. These were practiced by 3% and 92% of sorghum farmers

respectively. Adverse impacts of disease on sorghum were reported by only 40% of the

respondents.

Table 4.4: Pest Incidence on Sorghum (n=210) Pest problems

Occurrence n (%)

Control Attempted n (%)

Insects 139 (66%) 51 (36.7%) Weeds 131 (62%) 103 (78.6%) Diseases 84 (40%) 15 (17.9%)

69

4.2.1.1 Fertilizer use (FTIS)

Farmers in the study area generally do not apply fertilizer on sorghum. Although not

significantly different (Table 4.5), producers who use fertilizer on sorghum tended to be

younger (AGE), with more years of formal education (EDUC) than non-adopters. In

addition, these farmers had less farming experience (FMEXP) and smaller total farm sizes

(FMSZ). As might have been anticipated, as indicated in Table 4.6 if farmers use fertilizer

on other crops (FTANY), they are likely to use it on sorghum. Tables 4.5 and 4.6 show

comparative statistics of other social and economic features for the two types of farmers.

Table 4.5: Characteristics of Fertilizer Adopters and Non-Adopters in Sorghum Production – Continuous Variables

Total (N=210) Non adopters (n=204) Adopters (n=6) Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.

Sig.*

AGE (Years) 39.89 15.20 40.03 15.19 35.17 16.22 0.305 EDUC (Years) 5.62 3.83 5.59 3.79 6.50 5.58 0.568 HHSZ (No. of people) 8.13 4.83 8.14 4.87 7.86 3.67 0.878 FMSZ (Acres) 6.66 6.06 6.75 6.11 3.53 1.89 0.144 FMEXP (Years) 20.80 13.57 20.87 13.54 18.33 16.00 0.620 TOTCROPS (Number) 6.99 1.66 6.98 6.14 7.29 2.06 0.628 SGACRE (Acres) 1.13 0.61 1.13 0.62 1.14 0.38 0.958 INFOSC (Number) 3.33 1.66 3.00 1.29 3.34 1.68 0.596 FMLBR (Number) 4.27 2.52 4.25 2.53 4.86 2.27 0.532 OFFARM (Number) 1.32 2.15 1.34 2.18 0.86 1.07 0.561 TOTSGYD (Bags threshed) 2.04 1.60 2.03 1.62 2.18 0.98 0.808 * Testing for significant differences between adopters and non-adopters

Table 4.6: Characteristics of Fertilizer Adopters and Non-adopters in Sorghum Production – Non-continuous Variables

Total Non-adopters Adopters Variable (N=210) (n=204) (n=6)

Sig.**

FTANY (% yes) 5.70 4.40 50.0 0.004 BFCP (% yes) 28.0 29.0 00.0 0.134 BFMORG (% yes) 32.0 33.0 00.0 0.101 TRNNG (% yes) 26.0 27.0 00.0 0.164 ONFTR (% yes) 14.0 15.0 00.0 0.392 HARM (% yes) 84.0 84.0 83.0 0.225 FTPURCH (% yes) 7.14 6.37 33.0 0.061 FTISLBR (% yes) 41.0 40.0 83.0 0.043 RSCH (% yes) 69.0 70.0 33.0 0.079 WEED (% yes) 69.0 69.0 83.0 0.549 OWNIPM (% yes) 31.0 31.0 33.0 0.613 GENDER (% Males) 50.0 50.0 50.0 0.351 ** Testing for significant differences between adopters and non-adopters.

70

4.2.1.2 Intercropping with root crops, celosia and other striga ‘chasers’ (ECAT)

Eleven percent of interviewed farmers practiced striga control using one or more of these

methods. Table 4.7 shows no significant differences between adopters and non-adopters in

terms of the continuous variables. However, in Table 4.8 it can be seen that adopters were

more likely to belong to farmer organizations (BFMORG), to have participated in some form

of on farm trials (ONFTR), and to have had some form of (even non-IPM) pest management

training (TRNNG) than non-adopters. Forty six percent of striga control adopters obtained

agricultural information from Makerere University researchers, compared to 19% for non-

adopters while 92% of adopters obtained their information for Ministry of Agriculture

extension staff. Not surprisingly, since more adopters belonged to farmer organizations

than non-adopters, significantly more adopters obtained agricultural information from

these organizations than non-adopters.

Table 4.7: Striga adopters versus non- adopters – continuous variables Adopters (n=24) Non-Adopters (n=186)

Variable Mean Std. Dev. Mean Std. Dev. Sig.*

AGE (Years) 40.13 18.48 39.95 14.83 0.937 EDUC (Years) 5.92 3.51 5.59 3.89 0.691 FMEXP (Years) 21.75 16.68 20.68 13.17 0.716 FMSZ (Acres) 7.83 7.98 6.51 5.78 0.316 HHSZ (No. of people) 7.63 6.27 8.23 4.62 0.562 FMLBR (No. of people) 3.54 1.86 4.37 2.57 0.128

* Testing for significant differences between adopters and non-adopters.

Table 4.8: Striga adopters versus non- adopters – comparison with non-continuous variables

Adopters Non-Adopters Variable Mean (n=24) Mean (n=186)

Sig.*

BFMORG (% yes) 58.0 28.0 0.004 GENDER (% Males) 75.0 47.0 0.011 HDIPM (%yes) 71.0 48.0 0.099 ONFTR (%yes) 38.0 11.0 0.002 OWNIPM (%yes) 63.0 27.0 0.001 TRNNG (%yes) 67.0 21.0 0.000 INFOTYPE

Neighbors (%yes)*2 88.0 74.0 0.106 Radio (%yes)*1 79.0 72.0 0.319 MAAIF (%yes)*3 92.0 62.0 0.002 Friends (%yes) *2 67.0 56.0 0.233 Newspapers (%yes) *1 25.0 23.0 0.319 MUK (%yes) *3 46.0 19.0 0.006 Farmers organizations (%yes) *3 37.5 17.0 0.023 NGOs (%yes) *2 33.0 18.0 0.076

SGDZZ (%yes) 50.0 74.0 0.153 HARM (%yes) 96.0 82.0 0.313

* Testing for significant differences between adopters and non-adopters. *1, *2 and, *3 aggregated into MEDIA, INFNNF and RSCH respectively in multivariate analysis.

71

4.2.1.3 Sorghum and Crop rotation (ROTN)

This practice involved rotation of sorghum with legume crops including cowpea and

groundnuts. No significant difference was found between adopters and non-adopters when

examining continuous variables (Table 4.9). As indicated in Table 4.10, there was no

significant difference between where adopters and non-adopters obtained agricultural

information. At α of 0.2, there are significant differences in off-farm labor (OFFARM),

management time constraint (MGT), availability of off-farm income (INCMSC) and weed

incidence (WEED) between farmers who used crop rotation and those who did not (Table

4.10). Non-adopters considered crop rotation as management time demanding, and had

higher weed incidence than adopters.

Table 4.9: Characteristics of crop rotators and non-crop rotators – Continuous variables

Adopters (n=194) Non-Adopters (n=16) Variables Mean Std. Dev. Mean Std. Dev.

Sig.*

AGE (Years) 40.32 15.37 35.75 13.37 0.179 EDUC (Years) 5.70 3.88 4.69 3.36 0.397 FMEXP (Years) 21.04 13.65 18.00 13.61 0.401 OFFARM (No. of people) 1.40 2.20 0.37 1.02 0.066 FMSZ (Acres) 6.74 6.03 6.14 6.76 0.566 INFOSC (Number) 3.30 1.67 3.56 1.65 0.542


Table 4.10: Characteristics of crop rotators and non-crop rotators – non-continuous variables

Adopters Non-Adopters Variables Mean (n=194) Mean (n=16) Sig.** INFOTYPE

Radio (%yes) 72.0 88.0 0.139 Newspaper (%yes) 23.0 19.0 0.481 Friends (%yes) 58.0 50.0 0.349 Neighbors (%yes) 74.0 88.0 0.192 Farmers organization (%yes) 20.0 13.0 0.361 MAAIF (%yes) 64.0 81.0 0.128 MUK (%yes) 22.0 25.0 0.500

BFMORG (%yes) 32.0 25.0 0.395 RSCREQ

MGT (%yes) 22.1 38.0 0.189 COST (%yes) 28.0 25.0 0.253 KNOW (%yes) 24.0 13.0 0.604 LBR (%yes) 18.0 31.0 0.464 LND (%yes) 09.0 13.0 0.803

ONFTR (%yes) 14.0 19.0 0.407 INCMSC (%yes) 63.0 81.0 0.120 IMPLPURCH (%yes) 45.0 75.0 0.023 WEED (%yes) 67.0 94.0 0.058


72

4.2.2 Cowpea

Cowpea is generally grown in the second season, sowing two weeks after the start of rains.

The seed is mainly broadcast rather than precision planted. Although cowpea has

outstanding potential for intercropping, the crop in the study area is seldom intercropped.

However, when intercropped, maize or sorghum is sparsely spaced within the cowpea plot.

This finding is similar to that of the Consultative Group on International Agricultural

Research (CGIAR) findings in West Africa where cowpea is intercropped with millet, maize,

yam and sorghum (CGIAR, 2002).

Figure 4.3 Reasons for Cowpea Defoliation

During growth, the plant is defoliated several times for reasons ranging from the food

value of the leaves, better health of the plant, and as a control strategy for pests. Figure

4.3 shows the breakdown of these reasons.

Ebelat, a local but highly improved variety is popular among the farmers, grown on fields

ranging from 0.13 to 6 acres (Table 4.3). Other varieties included Large White, Brown tan,

Ecirikukwai and a black seeded type from Kenya, popular for its high yield (Table 4.11).

Selling1%

Medicinal Purposes

2%

Pest Control18%

Good Yield6%

Good Plant Health31%

Food Value42%

73

Table 4.11. Distribution of Cowpea varieties in study area Varieties grown Total Acreage

(Acres) Total Yield

(Bags threshed) Average Yield

(Bags/acre) Ecirikukwai 23.00 26.33 1.14 Ebelat 210.13 246.36 1.17 Large white 14.50 14.27 0.97 Brown tan 0.75 1.00 0.98 Black seeded Kenyan 26.50 32.95 1.24 Un-named varieties 4.00 4.17 1.04

Pest occurrence on the cowpea crop was high. Major insect pests included aphids (A.

craccivora), bollworms (Heliothis armigera), pod-borers (M. testularis) and stinkbugs

(Nezara viridula). Diseases on cowpea were diverse including cowpea mosaic virus (CMV),

leaf rust (Uromyces vignae), and anthracnose (Colletotrichum lindemuthianum). However,

disease control was not as intense as insect control probably because the vectors of the

disease were the insect pests. Disease control strategies were directly employed on only

15% of disease infected cowpea plots compared to 79% of insect-infested plots.

4.2.2.1 Timely planting (TPCP)

Farmers who practiced early planting sowed cowpea seed at the first sign of rains in either

rainy season. Adopters had smaller households (HHSZ), and were less likely to borrow

(BFCP) than non-adopters. However, as indicated in Table 4.12, these differences were not

significant. At a less stringent cut-off of 20%, FMEXP, INCMSC, FMLBR, ONFTR and

OWNIPM were significant (Table 4.12 and 4.13).

Table 4.12: Characteristics of adopters and non-adopters of timely planting for cowpea production – continuous variables

Adopters of Timely planting

Non-Adopters of Timely planting

Variables Mean (n=43) Std. Dev. Mean (n=169) Std. Dev. Sig.* AGE (Years) 41.96 16.10 39.52 14.99 0.320 EDUC (Years) 5.67 3.63 5.63 3.89 0.921 HHSZ (Number of people) 7.49 3.47 8.35 5.15 0.304 FMEXP (Years) 23.47 15.07 20.20 13.20 0.149 FMSZ (Acres) 7.34 7.88 6.53 5.53 0.412 INFOSC (Number) 3.42 1.68 3.30 1.67 0.683 FMLBR (Number of people) 3.81 2.29 4.41 2.59 0.145 OFFARM (Number of people) 0.95 1.70 1.39 2.25 0.221 * Testing for significant differences between adopters and non-adopters.

74

Table 4.13: Characteristics of adopters and non-adopters of timely planting for cowpea production –non-continuous variables

Adopters of Timely planting

(n=43)

Non-Adopters of Timely planting

(n=169) Variables % %

Sig.** BFMORG (%yes) 35.0 31.0 0.365 BFCP (%yes) 23.0 28.0 0.292 HIRE (%yes) 89.0 87.0 0.306 INCMSC (%yes) 74.0 62.0 0.091 ONFTR (%yes) 21.0 13.0 0.120 OWNIPM (%yes) 42.0 29.0 0.066 ** Testing for significant differences between adopters and non-adopters.

4.2.2.2 Intercropping with cereals (ICCP)

At α of 20%, farmers who intercropped cowpea (adopters) were significantly younger

(AGE), and had shorter farming experience (FMEXP) (Table 4.14). Significantly more

adopters than non-adopters hired labor (HIRE), applied fertilizer on other crops (FTANY)

and had used some IPM recommendation on their farm (Table 4.15). There was no

significant difference between adopters and non-adopters and their use of the other

sources of information except the media.

Table 4.14: Characteristics of adopters and non-adopters of intercropping with cereals for cowpea production – continuous variables

Intercropping Adopters (n=48)

Intercropping Non-adopters (n=161)

Variables Mean Std. Dev Mean Std. Dev Sig.*

AGE (Years) 36.38 14.73 41.11 15.23 0.068 EDUC (Years) 6.10 3.73 5.50 3.86 0.323 HHSZ (Number of people) 7.50 3.76 8.37 5.13 0.279 FMEXP (Years) 18.02 13.42 21.73 13.62 0.107 FMSZ (acres) 6.18 4.83 6.85 6.40 0.533 INFOSC (Number) 3.44 1.62 3.29 1.68 0.598 FMLBR (Number of people) 4.00 2.21 4.37 2.61 0.378 OFFARM (Number of people) 1.27 1.65 1.31 2.29 0.911


75

Table 4.15: Characteristics of adopters and non-adopters of intercropping with cereals for cowpea production – categorical variables

Intercropping Adopters (n=48)

Intercropping Non-adopters (n=161)

Variables % % Sig.**

INFOTYPE Radio (% yes) 83.0 70.0 0.049 Newspaper (%yes) 29.0 21.0 0.116 Friends (%yes) 58.0 57.0 0.479 Neighbors (%yes) 73.0 76.0 0.410 Farmer organization (%yes) 19.0 21.0 0.485 MAAIF (%yes) 69.0 64.0 0.334 MUK (%yes) 27.0 21.0 0.219 NGO (%) 15.0 21.0 0.216

BFMORG (%yes) 31.0 32.0 0.551 BFCP (%yes) 25.0 28.0 0.382 HIRE (%yes) 92.0 86.0 0.092 ONFTR (%yes) 19.0 13.0 0.208 OWNIPM (%yes) 44.0 28.0 0.026 TRNNG (%yes) 21.0 28.0 0.235 HARM (%yes) 87.0 83.0 0.304 FTANY (%yes) 10.0 04.0 0.112

** Testing for significant differences between adopters and non-adopters.

4.2.2.3 Improved Variety (ICPV)

The improved local high yielding cowpea variety Ebelat was grown by over 90% of

producers in the study area. Though not significantly different (Tables 4.16 and 4.17),

generally these farmers were younger (AGE) and had less education (EDUC). There were

significant differences between adopters and non-adopters and their membership in

farming organizations (BFMORG), and access to off-farm income sources (INCMSC). More

adopters than non-adopters had off-farm income sources, while non-adopters belonged

more to organizations than adopters.

Table 4.16 Characteristics of adopters and non-adopters of Improved variety for cowpea production – continuous variables

Adopters Non Adopters Variables Mean (n=191) Std. Dev. Mean (n=18) Std. Dev. Sig.*

AGE (Years) 39.77 14.89 41.00 18.19 0.729 EDUC (Years) 5.60 3.79 5.86 4.32 0.768 HHSZ (Number) 8.31 4.91 6.81 3.83 0.176 FMEXP (Years) 20.72 13.46 21.52 14.94 0.797 FMSZ (Acres) 6.67 6.03 6.58 6.45 0.946 FMLBR (Number) 4.31 2.53 4.06 2.36 0.684 CPACRE (Acres) 1.30 0.77 1.21 0.93 0.635 * Testing for significant differences between adopters and non-adopters.

76

Table 4.17 Characteristics of adopters and non-adopters of Improved variety for cowpea production– non-continuous variables

Adopters (n=191)

Non Adopters (n=18)

Variable % % Sig.**

INCMSC (% yes) 66.0 48.0 0.057 BFCP (% yes) 26.0 43.0 0.187 BFMORG (% yes) 30.0 48.0 0.071 INFOTYPE

Farmer organization (% yes) 19.0 28.0 0.282 MAAIF (% yes) 64.0 78.0 0.179 MUK (% yes) 23.0 11.0 0.196

** Testing for significant differences between adopters and non-adopters.

4.2.3 Groundnut

All 212 respondents grew groundnuts, sowing two weeks after the onset of the first rains.

Three planting methods were used: measuring plant spacing within and between rows

(36%), chop and drop method without specific plant to plant measurement (45%), and

broadcasting the seed into the ground (24%). A few producers had groundnuts under two

different planting systems. Measurement of plant spacing was perceived as time

consuming, labor intensive, and costly. In addition, the practice was perceived to require

more land and knowledge for correct plant spacing. Fifty five percent of farmers

intercropped groundnuts, mostly with Maize. Intercropping was perceived to be a land-

saving practice. Eight varieties of groundnuts were grown in the study area. The popular

variety Igola-1, a hybrid from India is resistant to most fungal and bacterial diseases and

was grown by over 76% respondents (Table 4.18). Land resources were not constraints to

growing Igola-1, although the cost of purchasing seed was perceived to be high. Etesot, a

local improved variety was not popular because of persistent crop failure due to pests.

Table 4.18 contains summary statistics on area and yield estimates of the varieties grown.

77

Table 4.18: Groundnut varieties, and their performance in the study area Variety

Percent farmers with

variety36

Total Area in variety (Acres)

Total Yield

(Bags threshed)

Productivity (Bags/acre)

Erudurudu Red 12.7% 30.00 140.50 4.64 Igola-1 76.9% 185.68 712.55 3.84 Otira 17.9% 43.25 147.00 3.39 Matudda 1.9% 4.00 13.50 3.38 Etesot 38.2% 82.70 282.75 3.41 Serenut 1 2.4% 6.25 20.00 3.20 Ebaya 7.5% 16.30 46.50 2.85 Erudurudu White 10.8% 22.25 49.91 2.24 Serenut 2 0.4% 2.25 1.00 0.44

4.2.3.1 Close spacing (CLSP)

Close spacing involved sowing seed in measured plots of 10cm by 30cm or 15cm by 30cm.

Table 4.19 shows that significant differences exist between adopters and non-adopters of

close spacing in terms of yield of groundnuts (TOTGNYD), with adopters of close spacing

obtaining higher yield than non-adopters. Adopters tended to grow the improved Igola-1

(and on a wider scale), had off-farm income sources (INCMSC), and belonged more to

farmer organizations (BFMORG) than non-adopters (Table 4.20).

Table 4.19. Characteristics of adopters and non-adopters of close spacing in Groundnut production - Non-continuous variables

Low plant density Close Spacing Variables Mean (n=187) Std. Dev. Mean (n=25) Std. Dev.

Sig.*

EDUC (Years) 5.71 3.79 5.00 4.19 0.323 FMSZ (Acres) 6.48 5.82 8.01 7.63 0.237 GNACRE (Acres) 1.76 1.21 1.72 1.28 0.881 FMLBR (Number) 4.22 2.56 4.80 2.22 0.273 OFFARM (Number) 1.26 1.84 1.72 3.76 0.312 IGOLAYD (Bags threshed) 3.46 3.91 2.76 2.36 0.393 IGOLACRE (Acres) 0.87 0.89 0.99 0.73 0.496 TOTGNYD 6.21 4.86 7.32 5.85 0.132 * Testing for significant differences between adopters and non-adopters.

36 Does not necessarily add up to 100% as individual farmers often planted more than one variety

78

Table 4.20. Characteristics of adopters and non-adopters of close spacing in Groundnut production - non-continuous variables Variable

Low Plant Density (n=187)

Close Spacing (n=25)

Sig.**

INFOSC Ministry of agric (% yes) 63.6 76.0 0.160 Makerere Univ. researchers (% yes) 19.8 40.0 0.026

HIRE (% yes) 84.5 84.0 0.543 IGNV (% yes) 74.9 92.0 0.040 GENDER (% male) 48.0 56.0 0.245 INCMSC ((% yes) 57.0 75.0 0.008 BFMORG (% yes) 24.0 42.0 0.005 INSECT (% yes) 63.1 64.0 0.913 DZZ (% yes) 58.3 48.0 0.809 WEED (% yes) 57.8 72.0 0.085 ** Testing for significant differences between adopters and non-adopters.

4.2.3.2 Resistant variety (IGNV)

Igola-1 was grown by about 77% of the respondents on plots averaging 1.15 acres

representing over 64% of all groundnut acreage in the sample area. Igola-1 average yield

was 4.4bags/farmer. Of significant difference (Table 4.21) was that adopters had bigger

households (HHSZ) than non-adopters and had a greater number of household members

providing farm labor (FMLBR). These improved variety adopters also did less broadcasting

than the non-adopters. Most Igola-1 farmers used close spacing. No differences existed

between adopters and non-adopters in terms of non-continuous variables.

Table 4.21: Characteristics of Adopters and Non-adopters of Improved Variety in Groundnut Production – Continuous variables

With Igola-1 Without Igola-1 Variables Mean (n=163) Std. Dev. Mean (n=49) Std. Dev.

Sig.*

EDUC (Years) 5.712 3.839 5.327 3.832 0.539 FMSZ (Acres) 6.887 6.246 5.913 5.371 0.325 HHSZ (No. of people) 8.59 4.91 6.76 4.27 0.019 AGE (Years) 40.34 15.55 38.41 14.05 0.437 FMLBR (No. of people) 4.48 2.52 3.63 2.32 0.036 GNACRE (Acres) 1.78 1.27 1.67 1.03 0.579 TOTGNYD (Bags shelled) 6.87 5.20 5.99 5.67 0.309 FMEXP (Years) 21.37 14.04 18.88 11.80 0.260 * Testing for significant differences between adopters and non-adopters.

Pest incidence on groundnuts was fairly high, although little effort was put into controlling

them (Figure 4.4). The common practice for controlling weeds and diseases involved hand

removal of weeds and rouging out of diseased plants while 88% of control practices on

insect pests involved spraying the crop with pesticides (Table 4.22). The most reported

79

constraint to chemical use was the high knowledge requirement involved in the practice.

Figure 4.4 shows pest incidence and the relatively low level of control efforts. Note that

although insect incidence was highest, control practices were more focused on weeds. This

may not suggest that weed incidence was more severe, rather that weed control practices

were more accessible to farmers than other pest controls.

59.40%

65.10%

63.20%

24.60%

57.10%

7.40%

Weeds Insects Diseases

Pest and Disease Occurrence and Control (Percent of Farmers report ing)

Inc idence (%) Contro l ef for ts (%)

Figure 4.4: Pest occurrence and control

Table 4.22 below shows a comparative summary of pest occurrences and control efforts on

the three crops considered in this study in the study area.

Table 4.22 Comparison of pest occurrence and control efforts among sorghum, cowpea and groundnut crops Sorghum Cowpea Groundnuts Crop

Incidence (%)

Control (%)

Incidence (%)

Control (%)

Incidence (%)

Control (% )

Insects 66.2 36.7a 92.2 79.3c 63.2 24.6c Diseases 40.0 17.9a 93.3 15.4c 57.1 7.4a Weeds 62.4 80.2b 77.0 40.4b 59.4 65.1b a Predominantly rouging b Predominantly hand removal c Predominantly chemical spray

80

4.3 Adoption of IPM Practices - Multivariate Analysis

This section presents results of multivariate estimations in tabular form. Tables 4.23-4.25

give results of the estimations for sorghum, cowpea and groundnut adoption models

respectively. Model goodness-of-fit tests and model implications are presented in tables for

each model. The marginal probability is evaluated at the mean of the continuous variable

and mode of the non-continuous variables.

Correlation analysis (between quantitative variables) and measures of associations

(between qualitative variables) were run prior to logit regression modeling. Examination of

correlation coefficients and measures of association indicate that models were subject to

multicollinearity. For instance FMEXP (length of farming experience) was highly correlated

with relative length of farming experience RFMEXP (r=0.88) and AGE (r=0.87), and also

HHSZ (household size) was correlated with size of family labor FMLBR (r=0.73). Therefore,

RFMEXP, AGE and HHSZ were eliminated from models. Appendix Table E4 shows these

associations and variables retained for estimating logit models.

4.3.1 Multivariate analysis results: Sorghum

4.3.1.1 Fertilizer use (FTIS)

From Table 4.23a below, Wald tests show that three factors are significant in explaining

adoption of fertilizer in sorghum. Overall, the estimated model has a strong explanatory

power, as the included variables correctly predict 97.6% of the observations. The model

chi-square of 19.888 corresponds to a p-value of 0.001 [df=5]. This shows that the model

is significant, that variables in the model other than the intercept term are useful in

explaining fertilizer adoption.

The test of the null hypothesis that FTANY (fertilizer use on other crops) coefficient is zero

against the alternative hypothesis has a Wald test statistic of 7.474 corresponding to

p=0.006 therefore the null hypothesis is rejected in favor of the alternative. The sign of the

coefficient for FTANY is as expected. Based on this data set, on average using fertilizer on

other crops in the producer’s fields is likely to increase the probability of its use on

sorghum by 0.025. High labor requirement involved in fertilizer application does not

discourage its use as seen from the FTISLBR (fertilizer labor constraint) variable’s positive

coefficient.

81

Important to note is the sign of the coefficient on RSCH (information from researchers)

that suggests that information from researchers may be negatively associated with

fertilizer use. Farmers obtain information that the use of a high cost input on relatively low

value sorghum results in minimal net returns. Such information is more likely obtained

from researchers than from other sources, hence explaining the sign of the RSCH

coefficient. Neither gendered differentiated input acquisition (FTPURCH) nor size of land

holdings (FMSZ) had an effect on fertilizer adoption.

Table 4.23a Maximum Likelihood Estimates for Fertilizer (FTIS) Adoption Model37 Logit Standard Significance

Variables Coefficient Errors Wald Odds Ratio Level Marginal

Probability

Constant -3.916 1.556 6.333 0.020 0.012 FMSZ -0.204 0.203 1.009 0.815 0.315 -0.0002FTISLBR 2.081 1.262 2.720 8.015 0.099 0.0077FTANY 3.164 1.157 7.474 23.656 0.006 0.0246FTPURCH 0.639 1.260 0.258 1.895 0.612 0.0010RSCH -1.524 1.033 2.176 0.218 0.140 -0.0040Goodness-of-fit tests38 Initial –2log likelihood = 54.491 -2log likelihood = 34.603 Model chi-sq. = 19.888 (0.001)[df=5] Classification = 97.6% McFadden’s R2 = 0.365 Iterations = 7

4.3.1.2 Intercropping with Celosia

Results of the celosia adoption model in Table 4.23b indicate that five variables: gender

(GENDER), family labor (FMLBR), membership in farm organizations (BFMORG), disease

incidence (DZZ) and prior training in pest control (TRNNG) have a significant influence on

celosia adoption (at the 20% level). The model overall correctly predicts 91.0% of the

variations in the response and is highly significant (p=0.000).

The positive value on the gender coefficient indicates that males are more likely to adopt

this practice. Farmers’ prior training in pest control increased their likelihood of adopting

celosia strategies. The variable farm labor size (FMLBR) was significant at 1%. A one-

person increase in family labor multiplies the odds of adopting celosia by 0.792 (which is a

37 Because the level of adoption of this t echnology was extremely low only a few individually significant variables from univariate estimation are used in the multivariate estimation (Amemiya, 1981, Hosmer and Lemeshow, 2000) 38 See section 3.3 for a description of goodness-of-fit tests

82

20.8% decrease in likelihood of adoption). There is a significant difference between

producers whose fields had disease occurrence DZZ(1) and those without DZZ(0). High

disease incidence in sorghum crops is negatively related to celosia adoption implying that

producers who do not adopt the practice would more likely have higher disease incidence

than those who adopt. The positive value on BFMORG indicates that membership in local

farm organizations increases the likelihood of adoption of celosia technologies.

Table 4.23b. Maximum Likelihood Estimates for ECAT (Celosia) Adoption Model Variables

B

S.E.

Wald

Sig.

Exp(B)

Marginal Probability

Constant -5.116 1.774 8.318 0.004 0.006 FMLBR -0.233 0.137 2.879 0.090 0.792 -0.0204INFOSCES 0.248 0.216 1.317 0.251 1.281 0.0217SKDOYD 0.124 0.269 0.212 0.646 1.131 0.0108GENDER 1.937 0.660 8.623 0.003 6.939 0.0816BFMORG 0.868 0.648 1.796 0.180 2.382 0.1604ONFTR 0.770 0.643 1.432 0.231 2.159 0.1408TRNNG 1.182 0.566 4.360 0.037 3.262 0.2273INFNNF 0.405 1.303 0.097 0.756 1.500 0.0301RSCH 1.361 1.186 1.316 0.251 3.898 0.0701WEED (1) -0.692 0.594 1.359 0.244 0.501 -0.1255WEED (2) 0.190 1.467 0.017 0.897 1.209 0.0333DZZ (1)39 -1.082 0.626 2.988 0.084 0.339 -0.2053DZZ (2) -0.538 0.890 0.365 0.546 0.584 -0.0962 Goodness of fit tests Initial –2log likelihood =149.261 -2log likelihood = 100.721 Model chi-sq.= 48.539 (0.000)[df=13], Classification = 91.0%, McFadden’s R2 = 0.325 Iterations=6

4.3.1.3 Crop Rotation

Influence of the explanatory variables on the probability of adopting crop rotation is shown

in Table 4.23c. This model containing eight variables correctly predicts 92.9% of the

variation in adoption probability. From Wald statistic tests, all economic variables except

pesticide use on other crops (PTANY) are significant at least at the 20% level. Insignificant

variables include borrowing potential of farmers (BFCP) and sources of information

(MEDIA and RSCH). 39 The reference category with dummy variables is the absence of the value category, that is, when the value of the category is zero, that category is used as the reference, in this case, DZZ(0) represents farmers without disease incidence and the other categories are compared with it. DZZ(1) variable compares farmers who had the disease with those who did not.

83

INCMSC has a negative coefficient implying that having off-farm income sources reduces

the likelihood of adopting crop rotation by 65.8%. This conforms to this practice’s

relatively low resource requirements – practicing crop rotation does not usually require

large cash outlays. As expected, high management requirements involved in crop rotation

(ROTNMGT) reduce this practice’s adoption. In addition, producers who adopted crop

rotation had lower weed incidences (WEED) than non-adopters. Switching from male farm

implement purchases to female-purchases (IMPLPURCH) has a positive effect on adoption

of the practice. When males purchase farm implements, the probability of adoption

decreases by 0.0185, ceteris paribus.

Table 4.23c. Maximum Likelihood Estimates for ROTN (Crop Rotation) Adoption Model Variables B S.E Wald Exp(B) Significance Marginal

Probability Constant 5.919 1.381 18.368 372.005 0.000 INCMSC -1.072 0.709 2.288 0.342 0.130 -0.0123 PTANY -0.589 0.580 1.030 0.555 0.310 -0.0052 IMPLPURC -1.367 0.650 4.420 0.255 0.036 -0.0185 BFCP -0.610 0.587 1.080 0.544 0.299 -0.0054 MEDIA -0.638 0.839 0.577 0.529 0.447 -0.0031 RSCH -0.450 0.712 0.400 0.638 0.527 -0.0023 ROTNMGT -1.090 0.615 3.143 0.336 0.076 -0.0126 WEED (1) -0.513 0.720 0.509 0.599 0.476 -0.0026 WEED (2) -1.955 1.151 2.883 0.142 0.090 -0.0377 Goodness-of-fit tests Initial –2log likelihood =113.133 -2log likelihood = 89.746 Model chi-sq.= 23.387 (0.009) [df=9] Classification = 92.9% McFadden’s R2 =0.207 Iterations=6

4.3.2 Multivariate analysis results: Cowpea

4.3.2.1 Early Planting

Four economic factors: pest (INSECT and WEED) occurrences that suggest the need for

pest control, availability of off-farm income sources (INCMSC) and seasonal labor

constraints (TPCPLBR) affect the adoption of timely planting. The other variables in the

model are not significant at the 20% level (Table 4.24a). The model correctly predicts

82.8% of the observations.

84

The positive coefficient on the WEED variable implies that a high weed incidence in

farmers’ plots may induce producers to practice timely planting as a control strategy,

increasing the probability of adoption by 0.043. There is a significant difference between

producers whose crop was not affected (WEED(1)) and those who do not know if insect

pests affected their crop (WEED(2)). Sowing is a labor-intensive agronomic practice. Thus

producers plant early to avoid the peak labor requirement season hence the positive

coefficient on TPCPLBR (seasonal labor constraints). Results also show that having off

farm sources of income (INCMSC) positively influences adoption of timely planting. As

measured by the marginal probability, insect incidence (INSECT) had the largest impact on

adoption of timely planting.

Table 4.24a. Maximum Likelihood Estimates for TPCP (Timely Planting) Adoption Model Variables

B

S.E.

Wald

Sig.

Exp(B)


Constant -1.198 1.031 1.349 0.245 0.302 FMEXP 0.017 0.014 1.373 0.241 1.017 0.0008FMLBR -0.121 0.095 1.629 0.202 0.886 -0.0062TOTCROPS -0.105 0.133 0.627 0.428 0.900 -0.0054GENDER -0.327 0.398 0.672 0.412 0.721 -0.0195INCMSC 0.614 0.440 1.952 0.162 1.848 0.0417FTPURCH -1.081 1.114 0.941 0.332 0.339 -0.0353IMPLPURCH -0.006 0.409 0.000 0.989 0.994 -0.0003ONFTR 0.211 0.549 0.147 0.701 1.235 0.0119TRNNG 0.577 0.474 1.479 0.224 1.781 0.0385TPCPLBR 0.983 0.441 4.959 0.026 2.672 0.0788TPCPLND 0.699 0.680 1.057 0.304 2.012 0.0493INSECT (1) -0.434 0.567 0.584 0.445 0.648 -0.0271INSECT (2) 2.250 1.374 2.684 0.101 9.489 0.2986WEED (1) 0.632 0.423 2.228 0.136 1.881 0.0432WEED (2) -0.922 0.755 1.492 0.222 0.398 -0.0320 Goodness-of-fit tests Initial –2log likelihood = 212.453 -2log likelihood = 179.048 Model chi-sq. =33.405 (0.004) [df=15] Classification = 82.8%, McFadden’s R2 =0.157 Iterations=4

4.3.2.2 Intercropping

In this model (Table 4.24b), from the Wald tests, four variables are significant at least at

the 20% level. The model correctly predicts 78.9% of the observations. The level of farmers’

experience (FMEXP) is significant at 20%. FMEXP, a social factor, in many technology

studies is found to positively affect adoption. In the statistical sense, however, the

85

hypothesis is not supported in this adoption model. The marginal effect of FMEXP is –

0.0029 implying that a one year increase in FMEXP decreases the probability of adoption

of intercropping by 0.29%. Two economic factors, fertilizer use on other crops (FTANY) and

weed incidences (WEED) are significant in predicting adoption of this practice. Fertilizer

use on other crops increases the odds of adoption of intercropping. The effect of this

variable is unexpected. However, since, intercropping involves growing cowpea with other

crops, fertilizer application may be for other components in the intercrop and not

necessarily for cowpea. A high incidence of weeds (WEEDS) on cowpea may influence

farmers to adopt intercropping practices as a weed control strategy.

Table 4.24b. Maximum Likelihood Estimates for ICCP (Intercropping) Adoption Model Variables

B

S.E.

Wald

Sig.

Exp(B)


Constant -3.072 1.023 9.011 0.003 0.046 FMEXP -0.021 0.014 2.135 0.144 0.980 -0.0029TOTCROPS 0.111 0.105 1.115 0.291 1.118 0.0153FTANY 0.908 0.687 1.750 0.186 2.480 0.1633HIRE 0.643 0.606 1.126 0.289 1.903 0.0705MEDIA 0.464 0.456 1.035 0.309 1.590 0.0542ICCPLBR 0.070 0.390 0.032 0.858 1.072 0.0098ICCPLND 0.770 0.716 1.157 0.282 2.160 0.1336WEED (1) 0.749 0.381 3.874 0.049 2.116 0.1293WEED (2) 0.796 0.518 2.361 0.124 2.217 0.1391 Goodness-of-fit tests Initial –2log likelihood = 225.248 -2log likelihood = 209.132 Model chi-sq. =16.117 (0.064) [df=9] Classification = 78.9%, McFadden’s R2 =0.072 Iterations=4

4.3.2.3 Improved Variety

From the Wald tests, three variables are important in explaining the Ebelat variety

adoption. The model correctly predicts 91.9% of the variations in the response. Having off-

farm income sources (INCMSC) increases the likelihood of adoption of improved cowpea

varieties (Table 4.24c). Information from informal sources including friends, NGO’s and

neighbors (INFNNF) increase the probability of adoption by 0.17 while information from

researchers (RSCH) does not have the same effect. Information from Makerere University,

Ministry of Agriculture staff and farmers’ organizations (RSCH) had a negative impact on

adoption of Ebelat variety. Holding other factors constant, farmer’s membership in local

86

organizations (BFMORG), and their borrowing concerns (BFCP) were not related to their

decision to grow the improved Ebelat variety.

Table 4.24c. Maximum Likelihood Estimates for ICPV (Intercropping) Adoption Model Variables B

S.E. Wald Sig.

Exp(B)


Constant 2.271 1.293 3.084 0.079 9.693 FMLBR 0.120 0.114 1.111 0.292 1.127 0.0175INCMSC 1.061 0.526 4.073 0.044 2.888 0.1081BFMORG -0.310 0.554 0.312 0.576 0.734 -0.0499IMPLPURCH 0.704 0.554 1.611 0.204 2.021 0.0811BFCP -0.305 0.567 0.290 0.590 0.737 -0.0490INFNNF 0.907 0.633 2.055 0.152 2.478 0.1710RSCH -2.161 1.076 4.030 0.045 0.115 -0.1534 Goodness-of-fit tests Initial –2log likelihood =122.674 -2log likelihood = 104.956 Model chi-sq. = 17.718 (0.013)[df=7] Classification = 91.9% McFadden’s R2 = 0.144 Iterations=6

4.3.3 Multivariate analysis results: Groundnut

4.3.3.1 Close spacing

From the Wald statistic test eight factors are significant in explaining the adoption of close

spacing. The model correctly predicts 68.9% of the variations in the response.

Membership in local farm organizations (BFMORG), having off-farm employment

(INCMSC), informal information sources (INFNNF), and growing the improved Igola-1

(IGNV) increase the likelihood of adoption of close spacing. Males are more likely to adopt

the practice while fertilizer use on other crops (FTANY) is negatively related to close

spacing. Producers’ ability to hire labor (HIRE) for farm operations involving close spacing

was positively significant at 20%. Land constraints (CLSPLND) had a positive effect on

practicing close spacing.

87

Table 4.25a. Maximum Likelihood Estimates for CLSP (Close spacing) Adoption Model Variables B S.E.

Wald Sig. Exp(B)


Constant -1.786 0.773 5.336 0.021 0.168 IGOLAYD -0.037 0.053 0.496 0.481 0.963 -0.0222GENDER 0.411 0.316 1.698 0.193 1.509 0.0745INCMSC 0.890 0.335 7.049 0.008 2.436 0.1984FTANY -1.236 0.824 2.250 0.134 0.291 -0.1692BFMORG 0.792 0.377 4.420 0.036 2.208 0.1902HIRE 0.636 0.485 1.720 0.190 1.888 0.0913IGNV 1.081 0.458 5.564 0.018 2.948 0.1717BFCP 0.052 0.359 0.021 0.885 1.053 0.0189TRNNG 0.480 0.398 1.453 0.228 1.616 0.0865INFNNF -0.732 0.475 2.377 0.123 0.481 -0.1558RSCH -0.287 0.374 0.590 0.443 0.750 -0.0728CLSPLBR -0.340 0.345 0.970 0.325 0.712 -0.1450CLSPLND 0.676 0.465 2.114 0.146 1.966 0.1081 Goodness-of-fit tests Initial –2log likelihood = 287.751 -2log likelihood = 253.436 Model chi-sq. = 34.316 (0.001)[df=13] Classification = 69.8% McFadden’s R2 =0.119 Iterations=3

4.3.3.2 Igola

In Table 5.25b below, results showed that all explanatory variables, except farmers’

practice of close spacing (CLSP) and number of family members providing family labor

(FMLBR) were insignificant at the 20% level. The positive estimated coefficients of CLSP

and (FMLBR) imply that, ceteris paribus, adoption of Igola-1 is expected to be higher if

producers plant under high density and if the number of family members providing farm

labor increases. This result is nor surprising as for better performance of Igola-1, growing

it at high plant densities (close spacing) is necessary. This shows that the two IPM

technologies (CLSP and IGNV) are complementary.

88

Table 4.25b. Maximum Likelihood Estimates for IGNV (Igola) Adoption Model Variables B

S.E. Wald Sig.

Exp(B)


Constant 0.604 0.744 0.658 0.417 1.829 FMLBR 0.140 0.080 3.064 0.080 1.150 0.0309INCMSC 0.365 0.354 1.064 0.302 1.441 0.0749ONFTR 0.626 0.592 1.119 0.290 1.870 0.1210INFNNF -0.547 0.590 0.860 0.354 0.578 -0.1077RSCH 0.060 0.375 0.026 0.873 1.062 0.0133CLSP 0.768 0.373 4.246 0.039 2.156 0.1433DZZ (1) -0.126 0.423 0.089 0.765 0.881 -0.0283DZZ (2) -0.507 0.509 0.992 0.319 0.602 -0.1196 Goodness-of-fit tests Initial –2log likelihood = 229.232 -2log likelihood = 214.074 Model chi-sq. = 15.158 (0.056)[df=8] Classification = 77.4% McFadden’s R2 =0.066 Iterations=4

4.4 Adoption of IPM Practices - Model Fitting

While there may be many independent variables that could potentially be included in the

model, including all possible variables does not always lead to the best predictions. The

results shown below are an attempt to obtain the best fitting model while minimizing the

number of parameters through model selection procedures suggested by Hosmer and

Lemeshow (1989). Tables 4.26, 4.28 and 4.30 are aggregated summary tables of results of

this procedure for sorghum, cowpea and groundnut models respectively. Tables 4.27, 4.29

and 4.31 show summary statistics of respective goodness-of-fit tests.

89

4.4.1 Sorghum

Table 4.26. Maximum Likelihood Estimates for the fitted sorghum IPM Adoption Models (FTIS, ECAT and ROTN) Practice

Variable

Estimate (B)

Std. Error

Wald

Test Sig.

Exp(B)

Marginal Probabilities

FTIS Constant -4.845 1.196 16.423 0.000 0.008 FTANY 3.255 1.013 10.322 0.001 25.922 0.0333 FTISLBR 2.294 1.218 3.546 0.060 9.913 0.0022 RSCH -1.737 1.006 2.983 0.084 0.176 -0.0064 ECAT Constant -4.538 1.176 14.885 0.000 0.011 FMLBR -0.253 0.137 3.428 0.064 0.776 -0.0224 GENDER 1.399 0.547 6.535 0.011 4.051 0.0721 BFMORG 0.888 0.540 2.702 0.100 0.411 0.1666 ONFTR 0.967 0.561 2.971 0.085 2.630 0.1831 TRNNG 1.493 0.517 8.342 0.004 4.451 0.3028 RSCH 2.008 1.068 3.534 0.060 7.446 0.0840 ROTN Constant 4.581 1.000 20.994 0.000 97.617 INCMSC -1.169 0.696 2.823 0.093 0.311 -0.0104 IMPLPURCH -1.345 0.636 4.473 0.034 0.260 -0.0132 ROTNMGT -1.083 0.595 3.316 0.069 0.339 -0.0091 WEED (2) -2.211 1.096 4.071 0.044 0.110 -0.0370

Table 4.27 Summary Goodness-of-fit tests for sorghum models Sorghum Technologies Statistic FTIS ECAT ROTN Initial –2log likelihood 54.491 149.261 113.133 -2 log likelihood 36.960 109.334 93.398 Model Chi-sq (p-value)

17.531 (0.001)

39.927 (0.000)

19.736 (0.003)

Percent Prediction 98.1% 91.0% 92.9% McFadden’s R2 0.32 0.267 0.174

Comparing Table 4.23a and Table 4.26 shows that by refitting the fertilizer adoption model

the correct percent prediction increases by 0.5% , and the Wald tests show that all retained

variables are significant atα =0.1. The overall model is highly significant (p=0.001). The

fitted ECAT and ROTN models do not show an improvement in correctly predicted

responses. However, since they contain fewer variables they are better since they are less

costly in terms of data collection40 and in simplicity compared to the full models.

40For the current study, the cost of data collection including all variables is considered a sunk cost. However, for future research on these technologies, results of this procedure are important.

90

The retained variables do not show any change in direction of effect, that is, they retain

the same sign. However, the magnitude of their effect changes. In the fitted fertilizer

adoption model, the effects of variables FTANY and RSCH are enhanced while the effect of

farm labor (FMLBR) decreases. For the crop rotation model, all variables retained exerted a

reduced effect. In the fitted celosia adoption model, the effect of gender was diminished

while the other variables exerted an increased effect in the fitted model compared to the

full model.

4.4.2 Cowpea

Table 4.28. Maximum Likelihood Estimates for the Fitted cowpea IPM Adoption Models (TPCP, ICCP and ICPV) Practice

Variable

Estimate

Std. Error

Chi-Square Statistic

Test Sig. Exp(B)


TPCP Constant -1.453 0.644 5.082 0.024 0.234 FMEXP 0.024 0.014 3.178 0.075 1.025 0.0033 FMLBR -0.153 0.085 3.216 0.073 0.858 -0.0212 TPCPLBR 0.885 0.386 5.250 0.022 2.423 0.1597 INSECT (2) 2.634 1.338 3.874 0.049 13.924 0.5692 WEED (1) 0.752 0.402 3.504 0.061 2.121 0.1312 ICCP Constant -2.150 0.826 6.774 0.009 0.116 FMEXP -0.025 0.014 3.296 0.069 0.976 -0.0032 TOTCROPS 0.136 0.102 1.769 0.184 1.145 0.0175 WEED (1) 0.841 0.371 5.142 0.023 2.318 0.1416 WEED (2) 0.869 0.508 2.929 0.087 2.384 0.1472 ICPV Constant 2.882 1.220 5.584 0.018 17.846 INFNNF 1.016 0.602 2.846 0.092 2.762 0.0606 RSCH -2.185 1.045 4.369 0.037 0.112 -0.3935

Table 4.29 Summary Goodness-of-fit tests for cowpea models Cowpea Technologies Statistic TPCP ICCP ICPV Initial –2log likelihood 212.453 225.248 122.674 -2 log likelihood 186.869 214.466 111.189 Model Chi-sq 25.584

(0.001) 10.782

(0.029) 11.189

(0.011) Percent Prediction 81.3% 77.5% 91.4% McFadden’s R2 0.12 0.048 0.091

At 10% all five variables of the TPCP model are significant. The fitted model correctly

predicts 81.3% of the variation in adoption and is significant at 0.1%. The full model has a

higher correct prediction percentage. On grounds of goodness-of-fit tests, the fitted model

might be said to be a poorer model than the full model. However, for practical

91

considerations this five-variable model performs better than the 16-variable model (Table

4.24a). The eight-variable full ICCP model has a correct classification of 78.9% which is

only 1.4% higher than that for the smaller fitted model. The fitted model is still significant

and is hence better in terms of data collection cost than the full model. Fitting the ICPV

model with just three variables results in a lower model chi-sq, but higher percent

classification. This is a better model than the full model and significance is higher

(p=0.011).

In this analysis no change in directional effects of the retained variables is noticed. The

effect of all variables in the fitted TPCP and ICCP models is enhanced while in the ICPV

model the effect of informal information (INFNNF) is reduced while that of researcher

information (RSCH) is enhanced.

4.4.3 Groundnuts

Table 4.30. Maximum Likelihood Estimates for the groundnut IPM Adoption Models (CLSP, IGNV) Practice

Variable

Estimate (B)

Std. Error

Chi-Square Statistic

Test Sig

Exp(B)

Marginal Effects

CLSP Constant -2.019 0.460 19.260 0.000 0.133 INCMSC 0.795 0.320 6.187 0.013 2.215 0.1922 FTANY -1.349 0.808 2.790 0.095 0.260 -0.2191 BFMORG 0.900 0.322 7.817 0.005 2.459 0.2183 IGNV 0.844 0.377 5.015 0.025 2.325 0.1568 GENDER 0.487 0.304 2.561 0.110 1.627 0.0985 IGNV Constant 0.196 0.359 0.298 0.585 1.216 FMLBR 0.154 0.077 3.972 0.046 1.166 0.0322 ONFTR 0.690 0.574 1.448 0.229 1.994 0.1225 CLSP 0.850 0.364 5.452 0.020 2.340 0.1444

Table 4.31 Summary Goodness-of-fit tests for groundnut models Groundnut Technologies Statistic CLSP IGNV Initial –2log likelihood 287.751 229.232 -2 log likelihood -261.714 216.734 Model Chi-sq 26.038

(0.000) 12.498

(0.006) Percent Prediction 67.9 76.9% McFadden’s R2 0.09 0.05

The fitted CLSP model contains 3 variables significant at 5%. This model is highly

significant [Model Chi-sq is 26.038 (p=0.000)]. Percent correct prediction is lower in the

92

fitted model by 1.9%. Two variables in the fitted IGNV model are significant at the 5% level

and percent correct prediction is lower than that for the full model by 0.5%.

The effect of the retained variables in the fitted IGNV model is increased compared to the

full model. With the CLSP model, the retained variables have a mixed effect. The impact of

variables FTANY (fertilizer use), BFMORG (membership in farmers organizations) and

GENDER are increased, while that of INCMSC (sources of off-farm income) and IGNV

(growing Igola-1) is reduced.

4.5 Technology adoption indices.

Table 4.32 below gives simple summary statistics of technology adoption indices.

Table 4.32 Distribution of technologies Description Sorghum (n=210) Cowpea (n=209) Groundnuts (n=212) Non-Adoption either technology

No sorghum technology adopted (n=15)

No cowpea technology adopted (n=14)

No groundnut technology adopted (n=36)

One Tech Any one of (celosia, fertilizer use, crop rotation) technologies adopted (n=166)

Any one of (improved variety, intercropping, Timely planting) technologies adopted (n=116)

Any one of (close spacing, improved variety) technologies adopted (n=101)

Two Tech Any combination of two technologies adopted (n=29)

Any combination of two technologies adopted (n=70)

Two technologies adopted (n=75)

Three Tech Three technologies adopted (n=0)

Three technologies adopted (n=9)

Three technologies adopted (n=0)

Variables for the cumulative logit models were obtained using procedures explained in

Fig 3.4.

Table 4.33 Cumulative Logit Model Estimates for Adoption of ‘ONETECH’ and ‘TWOTECH’ Sorghum Technologies Variables

Estimate

Std. Error

Chi Sq

Sig.

Odds Ratio


INTERCEPT = 2 -2.892 1.050 7.584 0.006 0.055 INTERCEPT = 1 1.895 1.033 3.362 0.067 6.653 GENDER 0.579 0.362 2.560 0.110 1.784 0.1275BFMORG 0.595 0.427 1.943 0.163 1.813 0.1471BFCP -0.798 0.406 3.876 0.049 0.450 -0.1684ONFTR 0.427 0.520 0.673 0.412 1.533 0.1053TRNNG 0.809 0.429 3.560 0.059 2.246 0.1995WEED (1) -1.277 0.956 1.785 0.182 0.279 -0.2419

93

From the analysis above in Table 4.33, only the variable ONFTR (participation of farmers

in on-farm trials) does not have a significant effect on adoption of TWOTECH and

ONETECH, but variables BFMORG (membership in farmers’ organizations), GENDER,

TRNNG (prior training in pest control) and WEED (weed incidence) do at the 20% level.

The 2 fitted regression lines are as follows:

Logit(p1) = 1.895 + 0.579GENDER + 0.595BFMORG - 0.798BFCP +

+ 0.81TRNNG-1.28WEED

Logit(p1+p2) = -2.892 + 0.579GENDER + 0.595BFMORG - 0.798BFCP +

+ 0.81TRNNG-1.28WEED

Where p1 is the probability of adoption of any one sorghum technology and p2 is the

probability of adoption of two technologies. Positive coefficients on the variable BFMORG

indicate that farmers’ membership in farm organizations is associated with increased

adoption of at least one sorghum technology. Estimated odds of 1.81 for this variable

indicate that the likelihood of adoption increase almost two-fold when farmers belong to

organizations than when they do not. The adoption of TWOTECH and ONETECH decline

when the availability of crop financing (BFCP) increases, that is, when producers switch

from not borrowing, to borrowing for crop production, adoption decreases. The positive

coefficient on TRNNG variable indicates that the more training farmers obtain, the more

likely they are to adopt one or two IPM sorghum technologies.

The likelihood ratio test statistic is equal to 24.9887, which corresponds to a p-value of

0.0003 hence the model is significant.

Table 4.34 Cumulative Logit Model Estimates for Adoption of ‘ONETECH’ ‘TWOTECH’ and ‘THREETECH’ Cowpea Technologies Variables

Estimate

Std. Error

Chi Sq.

Sig.

Odds Ratio


INTERCEPT = 3 -7.191 1.202 35.764 0.000 0.0007 INTERCEPT= 2 -3.770 1.146 10.821 0.001 0.0231 NTERCEPT = 1 -0.754 1.090 0.478 0.489 0.4705 FMLBR -0.110 0.059 3.522 0.061 0.8958 -0.00000 EBACRE 0.632 0.227 7.741 0.005 1.8814 0.00006 ONFTR 0.378 0.399 0.897 0.343 1.4594 0.00005 TPCPLBR 0.618 0.322 3.679 0.055 1.8552 0.00047 TPCPLND 2.009 0.589 11.624 0.001 7.4559 0.00006 INSECT 2.028 0.997 4.138 0.042 7.5989 0.00006

94

From Table 4.34, at the 20% level, Chi square tests show that participation in on-farm

trials (ONFTR) is not significant in explaining the adoption of various levels of cowpea

technologies. However, availability of family labor (FMLBR), the acreage in improved variety

(EBACRE), insect incidence (INSECT) and labor and land constraints at the time of

planting (TPCPLBR and TPCPLND) are significant in explaining the three levels of cowpea

technology adoption. The negative coefficient on FMLBR (the availability of farm labor on

farms) indicates that the variable is associated with reduced adoption of any cowpea

technology.

Fitted logit models are thus:

Logit(p1) = -0.754- 0.110FMLBR + 0.632EBACRE + 0.378ONFTR

+0.618TPCPLBR+ 2009TPCPLND + 2028INSECT

Logit(p1+p2) = -3.770- 0.110FMLBR + 0.632EBACRE + 0.378ONFTR


Logit(p1+p2+p3) = -7.191 - 0.110FMLBR + 0.632EBACRE + 0.378ONFTR


Table 4.35 Cumulative Logit Model Estimates for Adoption of ‘ONETECH’ and TWOTECH Groundnut Technologies Variables

Estimate

Std. Error

Chi Sq.

Sig.

Odds Ratio


INTERCEPT = 2 -1.897 0.464 16.708 0.000 0.150INTERCEPT = 1 0.671 0.441 2.318 0.128 1.956IGOLAYD 0.243 0.054 20.052 0.000 1.275 0.0548INCMSC 0.826 0.290 8.104 0.004 2.284 0.2010BFMORG 0.775 0.338 5.241 0.022 2.171 0.1883ONFTR -0.144 0.428 0.113 0.737 0.866 -0.0317RSCH 0.153 0.310 0.244 0.621 1.165 0.0336CLSPLBR -0.390 0.300 1.694 0.193 0.677 -0.0924

Based on the Chi square test, at the 0.05 level RSCH and ONFTR do not have a significant

effect on the probability of adoption of ONETECH and TWOTECH. In this model, higher

yield of Igola-1 (IGOLAYD) is positively related to adoption of groundnut pest control

technologies. Availability of off-farm income (INCMSC) and farmers’ membership in farm

organizations (BFMORG) positively influences their adoption of technologies.

From Table 4.35 the odds ratio indicate that the likelihood of adoption of two technologies

(TWOTECH) versus adoption of one (ONETECH) or none (ZERO TECH) increase by 27.5%

95

for a unit increase in yield of the improved variety (IGOLAYD). This is also true for

likelihood of adopting two or one technology versus adoption of none.

The fitted logit models are thus:

Logit(p1) = 0.671+0.243IGOLAYD+0.826INCMSC+0.775BFMORG-

0.390CLSPLBR

Logit(p1+p2) = -1.897+0.243IGOLAYD+0.826INCMSC+0.775BFMORG-

0.390CLSPLBR

The model with the independent variables included is significant. The likelihood ratio test

statistic of 46.66 corresponds to a p-value less than 0.0001 indicating that variables in the

model were important in explaining adoption.

4.6 Summary

Based on both univariate and multivariate estimations, measures of the overall fit of

estimated equations are relatively high. Results show that the variables included in each

model explain the variability of the dependent variables, as shown by the values of the

McFadden’s R2. In addition, the correctly predicted percent is high, ranging from 69.8% to

97.6%. Overall, models were significant at the 0.05 level (except ICCP and IGNV,

significant at the 0.1 level). For both univariate and multivariate models however,

coefficients of many variables are not different from zero (at the 0.05 level), as shown by

the Wald tests.

The model fitting procedures attempted to find the most important variables explaining

adoption. From section 4.4, some models performed relatively poorly in terms of goodness-

of-fit in relation to the full model. For these models (with the exception of Ebelat adoption

model (ICPV), crop rotation adoption model (ROTN) and Igola-1 adoption models (IGNV)),

the McFadden’s R2 and the correctly predicted percentage was lower than the full models.

The overall significance of the fitted models was improved ranging from p=0.000 to

p=0.029. These fitted models (Tables 4.26, 4.28 and 4.30) have substantially fewer

variables, yield better estimates of the effect of the significant variables and therefore are

empirically better than the full more complex models in Tables 4.23-4.25. The results

indicate the importance of parsimony – that emphasis needs to be placed only on a few

important variables, as this is less costly in terms of data collection.

96

Chapter 5 Discussion and Conclusions

5.1 Introduction This chapter summarizes the study, gives policy implications and recommendations for

future research. Section 5.2 gives the summary of objectives of the research, and the

thrust of the thesis. Section 5.3 summarizes analytical methods used and a comparison of

these methods, while section 5.4 and 5.5 give a summary of level of adoption and

discussion of factors affecting IPM technology adoption. Section 5.6 presents policy

implications and study highlights. Section 5.7 provides implications for further research.

5.2 Summary of Thesis Improper use of pesticides in controlling pests on crops can cause adverse effects on

humans and livestock through ingestion, inhalation and contact; degradation of soils,

water, and the general environment wherein it acts as a non-point pollution source.

Integrated pest management practices emphasize minimal use of pesticides in controlling

pests on farmers’ fields. Thus the adoption of IPM can reduce the use of pesticides and

their accompanying problems. In addition IPM has been commended for its role in

increasing farm production, net farm incomes and environmental benefits. In general, IPM

methods have been demonstrated to be profitable. The introduction of IPM CRSP activities

in Uganda to institutionalize IPM methods as pest control practices focused mainly on

sorghum, cowpea and groundnuts. These crops have priority status in terms of

agricultural area devoted to their production and their production levels.

Among the technologies encouraged to control striga on sorghum were intercropping the

crop with celosia argentia, fertilizer application and crop rotation with legumes; while in

controlling several cowpea pests including aphids (A. craccivora), blister beetles (Epicauta

spp.), pod-borers (M. testularis), thrips (M. sjostedti ) and leafhoppers, intercropping the

crop with cereals, close spacing, defoliation and timely planting were identified as potential

methods. In groundnut production, planting at high plant densities and growing improved

Igola-1 were methods found to control groundnut rosette virus (GRV) and cercospora leaf

spot (Cercospora arachidicola), the two being the most important diseases affecting

groundnuts in Uganda.

97

The emphasis of this study was to determine the level of IPM technology adoption and the

economic, social, management and institutional factors that influence farmers’ decision to

adopt these practices. In Uganda, where pests are the leading cause of crop loss, this

study is important in a number of ways:

(i) The adoption of the integrated pest management technologies acts as a measure of

success of IPM CRSP research efforts. Thus, establishing the level of adoption of

these technologies indicates to IPM researchers farmers’ preferences of alternative

pest control practices. This might suggest to program administrators ways to direct

IPM research efforts to those preferred practices to enhance their increased

adoption.

(ii) Determining the unsuccessful IPM strategies could contribute to increased research

on reducing problems associated with those IPM practices and enhance their

adoption.

(iii) IPM practices, like other technologies, are not introduced in a vacuum: the

intended adopters have their own pest control systems . Their pest control practices

are expected to change upon introduction of IPM technologies. In order to effectively

introduce technologies requires that researchers understand the social, economic

and management factors of targeted adopters and the institutional characteristics

that may either inhibit or enhance IPM adoption. Such an understanding might

highlight the importance of integrating effective management, social and economic

aspects of the introduced IPM program into farmers’ current farming systems.

Results from this study may be important in explaining adoption of similar technologies in

other areas with similar economic, social and other characteristics.

Chapter 2 presented background information on the agricultural and research system in

Uganda. It also reviewed literature pertaining to studies establishing factors affecting

agricultural technology. The review of literature revealed an extensive number of studies

related to adoption of technologies mainly in developed countries. Limited studies on

adoption of agricultural technologies in Uganda were found, and in particular the body of

literature on factors regarding IPM adoption was generally lacking. The review revealed

98

many factors related to technology adoption elsewhere, pointing to the importance of

incorporating them in IPM adoption studies in the Ugandan context.

5.3 Summary of Methods: Using a pre-tested questionnaire, primary data was collected from 212 farmers in sixteen

villages in eight parishes in Kumi district, Eastern Uganda. Simple statistics on the data

and testing for multicollinearity problems was done prior to advanced econometric

modeling. SPSS and SAS statistical packages were used to run univariate and multivariate

analyses. Univariate analysis established whether there were significant differences

between adopters and non-adopters in terms of various characteristics. Multivariate logit

analysis was deemed appropriate to use to model farmers’ decisions to adopt or not to

adopt. Ordered logit estimation was done to find the factors that are responsible in

explaining varying levels of adoption. Model fitting procedures attempted to find the

smallest model that best explained adoption. Using marginal probabilities, the most

influential factors affecting each technology adoption pattern were determined.

Important to note is that the three procedures resulted in approximately the same

outcomes. For a given model, variables from either procedure retained the same sign on

their coefficients (although different in magnitude) indicating a similar effect on the

response regardless of the method used. Thus, conclusions drawn from each method apply

to all methods employing the same model. However the above analytical methods

progressively yielded smaller models and for practical considerations, smaller models,

which are generally less costly in terms of data collection, are preferable. Thus, results of

model fitting procedures were considered most appropriate and were discussed.

5.4 Summary of Level of IPM Adoption: Figure 5.1 shows levels of adoption of the eight IPM practices studied.41.

SORGHUM COWPEA GROUNDNUTPractice % Practice % Practice %FTIS 3 TPCP 21 CLSP 12ECAT 11 ICCP 23 IGNV 77ROTN 92 ICPV 91

Fig 5.1 Levels of IPM Adoption

41 In Figure 5.1 FTIS, ECAT, ROTN, TPCP, ICCP, ICPV, CLSP and IGNV are the IPM practices: fertilizer use in sorghum, intercropping with celosia, crop rotation, timely planting, intercropping cowpea with cereals, growing improved cowpea varieties, close spacing and growing an improved groundnut variety respectively.

99

The level of adoption was measured as the percent of farmers taking on a particular

practice. Farmers in the study area generally tended not to apply fertilizers on sorghum.

The high expense involved in fertilizer input purchase may be a limiting factor in its wide

adoption. In terms of marketability, sorghum is considered a low-value crop, and use of an

expensive input would not be consistent with economic rationality of producers’ decision

making.

Intercropping with sorghum with celosia and other weed chasers also had a low level of

adoption. Eleven percent of farmers intercrop the sorghum crop with weed chasers.

However, considering the short time span since the release of this technology, the 11%

level of adoption might be an indicator of the attractiveness of this technology.

Consequently in time its adoption rate might be quite high.

Both crop rotations involving legumes and intercropping with cereals are considered

indigenous practices. These practices are important in providing carbohydrates and

proteins in alternate seasons or in the same growing season, in improving soil conditions

in addition to reducing pest populations. Twenty three percent and 92% of farmers in the

study area practiced intercropping and crop rotation respectively, suggesting a low

preference for intercropping as compared to crop rotation.

Timely planting involves planting early at the on-set of rains. The importance of this

practice was explained above. Twenty one percent of cowpea farmers planted early. This

level of adoption is the lowest of the three cowpea pest technologies, the highest being 91%

with improved cowpea variety adoption.

Reasons for having high plant density groundnut plots were presented. Close spacing

involved measuring plant-to-plant distance both within rows and between rows. Twelve

percent of farmers practiced close spacing in groundnut production.

The highest levels of adoption were registered with crop rotation and improved varieties of

cowpea and groundnut. Ninety two percent, 91% and 77% of farmers practiced crop

rotation, grew Ebelat and Igola-1 respectively.

100

5.5 Summary of Factors Affecting Adoption: To summarize the factors affecting IPM adoption requires revisiting the hypotheses of the

study in chapter 1. Results of hypothesis tests vary depending on the technology.

Hypothesis 1 stated that cost of a technology was negatively associated with IPM adoption.

In this study, the direct cost of a technology was not important in explaining its adoption.

However, test of this hypothesis through proxies such as labor, land and skill availability

showed mixed effects of these variables depending on technology. Labor constraints at

planting time positively influenced timely planting while availability of fertilizer as input

positively influenced its adoption in sorghum but negatively influenced the adoption of

close spacing in groundnuts. The availability of off-farm income to farmers had mixed

effect on adoption. It was positively related to adoption of close spacing but negatively with

adoption of crop rotation. Hypothesis two stated that the size of farm holdings positively

influences adoption does not hold true. Farm size was not important in IPM adoption. The

level of education did not show significance with adoption. Hypothesis three is proved

correct in the intercropping model but the reverse holds true with the timely planting

model. The size of household labor force negatively influences celosia adoption but

positively affects growing improved cowpea and groundnut varieties. Hypothesis four holds

true for all technologies except for celosia adoption model.

Practice Economic Social Management Institutional

Sorghum FTIS FTANY+ RSCH-ECAT FMLBR- GENDER+ ONFTR+ RSCH+

DZZ- TRNNG+ROTN WEED- IMPLPURCH-

INCMSC-

Cowpea ICCP WEED+ FMEXP-ICPV INFNNF+

RSCH-TPCP WEED+ FMEXP+

INSECT+TPCPLBR+

Groundnut IGNV FMLBR+CLSP INCMSC+ BFMORG+

FTANY- IGNV+ONFTR+

Fig 5.2 Factors Affecting IPM Technology Adoption

101

Figure 5.2 above42 gives a summary of significant factors affecting IPM adoption (using α =0.1)

a) Economic factors: Two economic factors affect fertilizer adoption in sorghum. Fertilizer use on other crops

(FTANY) in the farmer’s cropping system promotes its use in sorghum. This is in fact the

most influential factor in fertilizer adoption as gauged from the high value of its marginal

probability. The positive coefficient on the variable representing labor constraints in

fertilizer use (FTISLBR) is unexpected as it indicates that high labor requirements involved

in fertilizer use do not negatively influence its adoption.

Economic factors that are important in explaining adoption of celosia and other Striga

chasers include availability of farm labor and disease incidence, both factors affecting

adoption negatively. High availability of unpaid family labor (FMLBR) negatively affects

adoption of celosia technologies. The negative sign on the disease variable (DZZ) shows the

positive relationship between celosia adoption and disease control, that is, farmers who

adopt celosia have a low level of disease as they are more likely to be actively engaged in

disease control on their farms.

In the sorghum crop rotation model, 80% of the significant variables are economic factors.

The most important variable explaining the adoption of crop rotation was weed incidence.

The negative sign of the coefficients for the weed variable (WEED) imply that farmers who

adopt crop rotation are less prone to experience weed problems. This variable is a proxy for

the level of expected benefits from adoption of a technology. Availability of off-farm income

(INCMSC) acts as a hindrance to adoption of crop rotation. That is, farmers with more

income appear to prefer to use their finances in other practices other than crop rotation.

High management time requirements involved in crop rotation (ROTNMGT) also act as a

barrier to this practice’s adoption.

A number of economic factors are important in IPM adoption in cowpea production. Crop

losses due to high pest incidences (WEED and INSECT) provide an incentive for pest

control through practicing timely planting. In addition labor constraints at planting time 42 The sign on each variable indicates the direction of each factor on adoption of IPM practices. FTANY, FMLBR, DZZ, WEED, INSECT, INCMSC, TPCPLBR, GENDER, FMEXP, ONFTR, IMPLPURCH, RSCH and TRNNG respectively are the variables: Fertilizer use on other crops, size of family labor resource, disease, weed and insect incidence, availability of off-farm income, labor constraint at time of planting, sex of farmer is male, length of farming experience, on-farm trial participation, male-driven input purchase decisions, information from researchers and farmers prior attainment of pest control training.

102

(TPCPLBR) induce farmers to plant early to avoid peak labor demands. This is important to

ensure the cowpea crop reaches maturity before the pest populations peak.

Intercropping cowpea with cereals is positively influenced by weed incidence (WEED) in the

cowpea plots implying that perhaps, as a weed control strategy, farmers who experience

high weed incidences are induced to intercrop. Farmers growing many crops (TOTCROPS)

perceive the need to intercrop cowpea as a way of reducing land availability constraints;

hence it is probable that they practice intercropping both as a land-saving technology and

as a pest control strategy.

None of the economic factors examined in the study were related to improved cowpea

adoption. In groundnuts, close spacing was positively influenced by availability of off-farm

income (INCMSC), but negatively by use of fertilizer on other crops (FTANY). High farm

labor availability (FMLBR) positively influences adoption of the improved groundnut

variety.

b) Social factors: Social factors were generally not related to sorghum technology adoption except celosia.

The positive coefficient on the gender variable (GENDER) indicates that males were more

likely to adopt celosia than females. In groundnut production the gender variable was

positively associated with practicing close spacing.

Farm experience (FMEXP) positively influenced timely planting of cowpea. Farmers with

accumulated farming experience probably acquire knowledge of seasonal changes that

signal the approaching sowing season and thus prepare resources necessary for sowing. In

addition, these farmers may have acquired encouraging returns from the practice and thus

continue with it anticipating continued benefits. Both these aspects could influence

farmers’ inclination to plant at the on-set of rains. On the other hand, accumulated

farming experience acted as a barrier to intercropping cowpea with cereal crops. It is

probable that past experience with poor performance of cowpea intercrops may discourage

increased practice of intercropping.

c) Management related factors: In the fertilizer model, management factors played no significant role, while with celosia,

farmers’ participation in on-farm trials (ONFTR) increased the likelihood of the practice’s

103

adoption in sorghum. In the crop rotation model, when males purchase implements

(IMPLPURCH) the probability of practicing crop rotation in sorghum reduces, as seen from

this variable’s negative coefficient.

None of the management factors analyzed in this study were related to cowpea technology

adoption. In groundnut production, however, results show that adoption of close spacing

was induced by farmers’ membership in organizations, participation in on-farm

demonstrations, and the variety farmers grew. Farmers who grow the improved groundnut

variety or belong to farmers’ organizations are more inclined to practice close spacing.

Ideas obtained from farmers organizations may be related to planting at high plant density

because of the benefits gained from either improved yields or from less pest pressure on

the close spaced crop.

d) Institutional factors: In sorghum models, three institutional factors affect the adoption of celosia and fertilizer

adoption. Information from researchers does not positively influence farmers to use

fertilizer, while it has a pronounced positive effect on celosia adoption. In addition,

attaining pest control training increases the probability of celosia adoption

Adoption of improved Ebelat cowpea variety does not seem to be positively influenced by

information from researchers. This finding is not unexpected. Growing an improved

cowpea variety as a pest control strategy was not an IPM recommendation in the study

area (see Section 3.4.4). This technology was included in this analysis to examine how

responsive farmers were of other potential technological changes. Nonetheless, farmers’

access to informal sources of information like friends, neighbors and others had a positive

effect on the likelihood of this technology’s adoption. Groundnut technologies were

generally not affected by institutional factors.

5.6 Policy Implications and Conclusions Results from this analysis reinforce similar findings by other researchers. That labor is

important in adoption models is evident in Bartel and Lichtenberg (1987) and in Green

and Ng’ong’ola (1993) among others. Bartel and Lichtenberg (1987) found that it is not the

availability of labor, but rather how skilled the labor is that would be important in

technology adoption. In their study of factors affecting fertilizer adoption in Malawi, Green

104

and Ng’ong’ola (1993) found that the availability of regular labor positively influenced a

practice’s adoption.

Farm labor availability in this study positively influenced growing of improved groundnut

variety Igola-1. This variable was positively correlated with household size suggesting that

a big household yielded a large family labor force. In general, big households have larger

food demands than smaller ones. The improved disease-resistant varieties were also high

yielding. Therefore the high involvement of family members in growing high yielding

varieties is consistent with households’ food consumption requirements.

The most influential variables in celosia adoption are institutional/informational factors,

including farmers’ access to information from researchers and training in pest control

activities. These services have been part of an ongoing IPM CRSP study involving farmer

field schools. The big influence they have suggests that continuing and/or intensifying

their activities would further enhance technology adoption.

Another important factor with a positive influence on celosia technology adoption was

farmers’ participation in on-farm trial demonstrations. It should be noted that celosia

technology is largely a ‘new’ technology, and farmers are likely to attach a higher risk

premium on such a technology than on the more ‘indigenous’ practices. Its adoption is

thus expected to be enhanced more through farmers having hands-on experience than

would be the case with the more indigenous technologies. This suggests that the

introduction of such ‘exotic’ practices should be preceded by encouraging higher farmer

participation in on-farm trial demonstrations as a means of increasing farmers’ practical

experience with the introduced technologies.

The most important variable that was related to fertilizer use in sorghum was the

availability of fertilizer for use on other crops (FTANY) with the largest associated change in

probability. This variable is a measure of farmers’ willingness to pay for a high cost input

in production. The positive sign of the variable implies that availability of fertilizer input

for the producer’s other crops would benefit the sorghum crop in terms of pest control if

adopted in sorghum production.

The positive effect that the variable off-farm income (INCMSC) had on adoption of close

spacing highlights how essential availability of non-farm earnings may be in financing the

105

purchase of inputs necessary for practicing close spacing. Results also showed that

females were more inclined to borrow to finance crop production than males. In the event

that the borrowed capital is directed to purchasing these inputs, then providing accessible

credit to women farmers would enhance the adoption of this practice.

Males were more likely than females to adopt celosia technology. Celosia technology is an

exotic control method and accessibility to such technologies is mostly a preserve for males.

To change this, programs that target both gender groups would be necessary to ensure

equitable adoption of practices between males and females.

None of the management factors analyzed in the study were related to cowpea technology

adoption. This suggests that high managerial capacity of farmers may not be an important

aspect in efforts to disseminate cowpea technologies. Management factors in several

studies (McNamara, Wetzstein, and Douce, 1991; Waller et al., 1998) were found to hinder

technology adoption. In the latter study the more intensive management effort required for

integrated pest management hindered potato farmers from adopting these technologies.

The finding here that management factors do not play an important role in cowpea

technology adoption implies that introduction of cowpea IPM technology in Uganda can

take place regardless of cowpea farmers’ managerial capability.43

Many factors that were theoretically hypothesized to be influential in explaining adoption

patterns of technologies showed no relationship with the dependent variable. Farmers’

perception of the harmful effect of chemicals did not influence farmers’ decisions in regard

to IPM technology adoption. This is in spite of farmer’s high knowledge about this issue. A

plausible explanation would be that these farmers do not consider environmental and

health impacts important considerations when choosing farming practices. A similar result

was also found in the analysis of adoption of non-chemical methods for controlling olive

pests in Albania (Daku, 2002). Educational programs geared to increasing awareness

about the effects of chemicals and the effectiveness of alternative methods of pest control

could transform this attitude and hence influence farmers to adopt IPM practices.

43 Recall: Factors under this broad category of management included ability for farmers to borrow for crop production, membership in farmers’ organizations, input purchase decision making, and participation in on-farm trial demonstrations.

106

The level of education was not an important factor in explaining adoption. The variable

EXTS (contacts of farmers with extension personnel) showed multicollinearity with other

variables and was eliminated in models. This is not to say that this factor does not

influence technology adoption. The variable that it was correlated with: accessibility to

information from Makerere University and Ministry of Agriculture researchers (RSCH) was

important in explaining adoption of fertilizer, improved cowpea variety and celosia. In

studies on adoption of sweep net and treatment thresholds in Texas (Harper et al, 1990),

producer’s contacts with extension were significant but negative in their effect on

technology adoption. In this study, the effect of information from researchers (RSCH) on

adoption of fertilizer and improved cowpea variety was negative.

The effect of size of farm holdings (FMSZ) was unimportant in adoption decisions. A study

analyzing factors affecting adoption of new bean varieties in Uganda found a similar result

(Mugisa-Mutetikka, 2000). In the current study, in the fertilizer adoption model where this

variable was not eliminated at the preliminary analysis stage, its effect was negative

(although insignificant). That this variable was not significant in explaining adoption might

suggest that IPM technologies are mostly scale neutral. This finding is particularly

important for IPM dissemination in the study area implying that IPM practices could be

introduced to farming systems regardless of the farmer’s scale of operation.

Females were less educated than males. And perhaps to make up for this, they strive to

acquire information and skills by belonging in farmers’ organizations. However,

membership in farmers’ organization was not a significant factor in adoption of many

practices except close spacing of groundnuts (CLSP) and celosia (ECAT). In fact, for the

case of celosia adoption, this variable exerted a negative influence on the probability of

adoption. The most plausible explanation is that information obtained in the organizations

may not have contained IPM-content. Providing IPM-content information at farm

organization meetings might enhance dissemination of these technologies and in

particular this would target women farmers whose membership in farm organizations was

significantly higher than males, subsequently promoting their adoption of IPM practices.

Overall, it appears that these policy changes are mostly applicable to institutional and

management factors. Economic and social factors could be effected through institutional

changes. Also important to note is that it appears that the more ‘exotic’ an introduced

practice is, the more its adoption will be dependent on informational aspects of the

107

implementing program. This argues for the intensification of training and educational

programs for potential adopters of that practice.

5.7 Future Research Direction Adoption of IPM technologies is dependent on a number of factors which are dynamic both

in terms of geographic setting and in time. Thus adoption can be said to be site specific.

The site specificity of adoption has an implication on the extensive applicability of the

policy implications stated in this study in that they may have a somewhat limited bearing

over a large area. According to Yaron, Dinar and Voet (1992) the site specificity of adoption

argues for region specific adoption studies.

In addition, as social, economic and other factors change, it is imperative that this study

be revisited in line with the changing socio-economic and other demographic changes.

Subsequent to establishment of the level adoption and factors affecting adoption, rate of

adoption studies may be appropriate to examine the effect time has on adoption of IPM

technologies.

Subsequent to establishing factors influencing adoption, a study on adaptation of farmers

to IPM practices may be a necessary step. Such a study would examine how farmers

adjust their economic, social and other conditions to accommodate the introduced IPM

practices.

To take this study further, a study to examine the effect of distance of farmers from the

focal points on adoption is necessary. It is anticipated that close proximity of farmers to

IPM activities (with focal farmers) may increase the likelihood of adoption of IPM

technologies. In addition, exposure of farmers to IPM activities over a wider geographical

space might facilitate more widespread adoption. However at this point these assertions

cannot be explicitly made. Therefore, examining the level and intensity of adoption of

farmers in various locations relative to the focal points may be important in highlighting

the importance of distance in adoption studies. An initial objective of this study was to

conduct such an analysis. However, limited availability of (GPS) Global Positioning System

units precluded the collection of this data.

108

References Abadi Ghadim, K. A., and D. J. Pannell. “A Conceptual Framework of Adoption of an

Agricultural Technology.” Agricultural Economics 21 (1999): 145-154. Abara, I. O. C. and S. Singh. “Ethics and Biases in Technology Adoption: The Small Farm

Argument.” Technological Forecasting and Social Change. 43 (1993): 289-300. Adesiina, A.A. and J. Baidu-Forson. “Farmers’ Perceptions and Adoption of New

Agricultural Technology: Evidence From Analysis in Burkina Faso and Guinea, West Africa.” Journal of Agricultural Economics. 13 (1995):1-9.

Adipala, E., C.A. Omongo., A. Sabiti., J.E. Obuo., R. Edema., B. Bua and M.W.

Ogenga-Latigo. “Pests and diseases of cowpea in Uganda: Experiences from a diagnostic survey.” African Crop Science Journal. 7(4) (December 1999):465.

Agresti, A. “An Introduction to Categorical Data Analysis.” New York: John Wiley

and Sons, Inc., 1996. Agresti, A. and B. Finlay. Statistical Methods for the Social Sciences. 3rd Edition. New

Jersey, Upper Saddle River: Prentice-Hall, 1997. Alebeek, van F.A.N and J.C van Lenteren. Integrated Pest Management for Protected

Vegetable Cultivation in the Near East. FAO Plant Production and Protection Paper 114. The Food and Agricultural Organization of the United Nations, Rome, 1992.

Allison, P. D. Logistic Regression Using the SAS System: Theory and Application. SAS

Institute, Inc., Cary, NC, 1999. Alston, J.M., G.W. Norton and P.G. Pardey. Science under Scarcity: Principles and Practice

of Agricultural Research Evaluation and Priority Setting. Ithaca: Cornell University Press, 1995.

Amemiya, T. “Qualitative Response Models: A Survey.” Journal of Economic Literature.

The American Economic Association. 19 (1981):1483-1536. Baidu-Forson, J. “ Factors Influencing Adoption of Land-enhancing Technology in the

Sahel: Lessons From a Case Study in Niger.” Journal of Agricultural Economics. 20(1999):231-239.

Bashaasha, B., V. Kasenge., D.B. Taylor., J. Olupot and M. Erbaugh. “An Economic

Assessment of the Integrated Striga Management Strategy for Small-scale Sorghum Farmers in Uganda.” Annual Report, IPM CRSP, 2000.

Boahene, K., T.A.B. Snijders, H.Folmer. “An Integrated Socio-Economic Analysis of Innovation Adoption: The case of Hybrid Cocoa in Ghana.” Journal of Policy Modeling. 21(2) (November 1999):167-184.

Bonabana, J., D. B. Taylor., V. Kasenge., B. Bashaasha and M.J. Erbaugh. “Assessing the

Economic Impacts of IPM CRSP Strategies on Groundnut Diseases in Kumi-Uganda.” Annual Report, IPM CRSP, 2001.

109

Cameron, L.A. “The Importance of Learning in the adoption of High-Yielding Variety Seeds.” American Journal of Agricultural economics. 81(February 1999):83-94.

Caswell, M., K. Fuglie., C. Ingram., S. Jans and C. Kascak. Adoption of Agricultural

production practices: Lessons learned from the US. Department of Agriculture area studies project. Washington DC. US Department of Agriculture. Resource Economics Division, Economic Research service, Agriculture Economic Report No. 792. January 2001.

CGIAR, Consultative Group on International Agricultural Research. (Online) URL:

http:///www.cgiar.org September 2002. CIA, The Central Intelligence Agency - The World Factbook, 2000. Country Focus: Uganda

Daku, L. “Assessing farm-level and aggregate economic impacts of olive integrated pest

management programs in Albania.” PhD. Dissertation. Virginia Polytechnic Institute and State University, 2002

Diebel, P.L., D.B Taylor and S.S. Batie. “Barriers to low-input agriculture adoption: A Case

study of Richmond county, Virginia.” American Journal of Alternative Agriculture. 8(3) (1993):120-127.

Doss, C.R and M.L Morris. “How Does Gender Affect the Adoption of Agricultural

Innovation? The Case of Improved Maize Technologies in Ghana.” Journal of Agricultural Economics. 25(2001):27-39.

Ehler, L.E and D.G Bottrell. “The illusion of Integrated Pest Management. Issues in

Science and Technology.” Bell and Howell Information and Learning Company, 2000, pp. 61-64.

EIU, The Economist Intelligence Unit. EIU Country Report: Uganda. (Online)

http://www.eiu.com. April 2001. El-Osta, H.S and M.J. Morehart. “Technology Adoption Decisions in Dairy Production and

the Role of Herd Expansion.” Agricultural and Resource Economics Review. 28 (1999): 84-95.

Enos, J.L and W.H Park. The Adoption of Imported Technology: The Case of Korea.

New York: Croom Ltd, 1988. Erbaugh, J.M., S. Kyamanywa, P. Kibwika, D. Taylor, V. Odeke and E. Mwanja.

“Baseline II: A Follow-up survey of farmer pest management practices and IPM knowledge diffusion.” Working paper 01-2. IPM CRSP. October, 2001.

FAO, Food and Agricultural Organization of the United Nations. Nineteenth FAO Regional

conference for Africa. World Food Summit: Food Security situation and issues in the Africa region. Ouagadougou, Burkina Faso, 16-20 April 1996. (Print)

FAOSTAT, The Food and Agricultural Organization of the United Nations Production

Databases [Online], URL: http://www.fao.org. January 20, 2002, October 18, 2002, May 2002.

110

Feder, G. and R. Slade. “The acquisition of information and the adoption of new Technology.” American Journal of Agricultural Economics. American Agricultural Economics Association. 66 (August 1984):312-320.

Feder, G., E. R. Just and D. Zilberman. “Adoption of Agricultural Innovations in

Developing Countries: A Survey.” Economic Development and Cultural Change. 33 (1985):255-298.

Fernandez-Cornejo, J. “Environmental and Economic Consequences of Technology

Adoption: IPM in Viticulture.” Agricultural Economics, 18 (1998):145-155. Fernandez-Cornejo, J. “The Microeconomic Impact of IPM Adoption: Theory and

Application.”Agricultural and Resource Economic Review. 25 (October 1996):149-160.

Gabre-Madhin, E.Z. and S. Haggblade. Success in African Agriculture: Results of an

Expert Survey. International Food Policy Research Institute. Washington DC. June 2001.

Garson, D. (1998). North Carolina State University. Logistic Regression Class notes.

(Online), URL: www2.chass.ncsu.edu/garson/ (November 12, 2001). Giliomee, J. H. “Integrated Pest Management in the African Context.” African Crop

Science Proceedings. 1(1994): 346-349. Green, D.A.G., and D.H. Ng’ong’ola. “Factors Affecting Fertilizer Adoption in Less

Developed Countries: An Application of Multivariate Logistic Analysis in Malawi.” Journal of Agricultural Economics. 44 (1) (January 1993):99-109.

Griliches, Z. “Hybrid Corn: An Exploration in the Economics of Technological Change.”

Econometrica.. 25(1957):501-522. Harper, J. K., M. E. Rister, J. W. Mjelde, B. M. Drees, and M. O. Way. “Factors influencing

the adoption of insect management technology.” American Journal of Agricultural Economics. 72(4) (November 1990):997-1005.

Hassan, B. Maize Technology Development And Transfer: A GIS Application For Research

Planning in Kenya. CAB International, Walford, UK. CIMMYT and Kenya Agricultural Research Institute, 1998.

Hosmer, D. W. and S. Lemeshow. Applied Logistic Regression. A Wiley-Interscience

publication. Wiley Series in Probability and Mathematical Statistics. New York: John Wiley & Sons, 1989.

___ 2nd Edition. John Wiley & Sons. 2000. IFPRI, International Food Policy Research Institute. Pest Management and Food Production:

Looking into the future, Vol. 20, no. 52. September 1998. ___ Applying science to sub-Saharan Africa’s food needs. Reported by Ellen Wilson. 2020

Vision. News and Views. February 1995 (Online) http://www.ifpri.org/pubs/books/ufa/ufa_ch26.pdf accessed May, 2001.

111

IPM CSRP, Integrated Pest Management Collaborative Research Support Program homepage. http://www.ag.vt.edu/ipmcrsp March 9th 2001.

IPM CRSP, Integrated Pest Management Collaborative Research Support Program.

Office of International Research and Development (OIRD) Outreach Division. Virginia Tech. Annual Reports. 1995-2001.

Isubikalu, P., J.M. Erbaugh and A.R Semana. “Understanding Farmer Knowledge

Regarding Cowpea Production and Pest Management: A Case Study of Kumi, Pallisa and Katakwi Districts in Eastern Uganda.” MS thesis, Makerere University, 1997.

Jackai, L.E.N, S.R. Singh, A.K. Raheja, F. Wiedijk. “Recent Trends in the Control of

Cowpea Pests in Africa.” Cowpea Research, Production and Utilization. S.R. Singh and K.O. Rachie. eds. pp. 233-243. Institute of International Tropical Agriculture (IITA) World Cowpea Research Conference. John Wiley & Sons, 1985.

Karungi, J., E. Adipala., P. Nampala., M.W. Ogenga-Latigo and S. Kyamanywa. “Population

dynamics of selected cowpea insect pests as influenced by different management strategies in eastern Uganda.” African Crop Science Journal. 7(No.4) (1999):487-495

Kasenge, V. 1998. Socio-economic factors influencing the level of Soil Management

Practices on Fragile Land. In Proceedings of the 16th Conference of Soil Science Society of East Africa (Eds.: Shayo-Ngowi, A.J., G. Ley and F.B.R Rwehumbiza), 13th-19th, December 1998, Tanga, Tanzania pp.102-112, 1998.

Kato, E. “An analysis of factors affecting adoption of K131 bean variety by women groups

in Luuka County, Iganga district.” MS thesis, Makerere University, 2000. Kebede, Y., K. Gunjal and G. Coffin. “Adoption of New Technologies in Ethiopian

Agriculture: The case of Tegulet-Bulga District, Shoa Province.” Journal of Agricultural Economics. 4(1990):27-43.

Khanna, M. “Sequential Adoption of Site-Specific Technologies and its Implications for

Nitrogen Productivity: A Double Selectivity Model.” American Journal of Agricultural Economics. 83(1) (February 2001):35-51

Klotz, C., A. Saha, and L. J. Butler. “The Role of Information in Technology Adoption: The

Case of rbST in the California Dairy Industry.” Review of Agricultural Economics, 17(1995): 287-298.

Kyamanywa, S. “Current Status of IPM in Uganda.” Proceedings of the IPM Networking in

Sub-Saharan Africa Workshop, 14-16 October 1996. IPM Networking in Sub-Saharan Africa. Gebrekidan B, Amirault J.P, Abate T. Eds. pp 28-36. Virginia Tech, 1996.

Leisinger, K.M. Sociopolitical Effects of new Biotechnologies in Developing countries.

IFPRI.2020 Vision Brief 35. July 1996. Liebman, M and E. Dyck. “ Crop rotation and intercropping strategies for weed

Management.” Ecological Applications. 3 No.1 (1993): 92-122.

112

Lionberger, H.F. Adoption of New Ideas and Practices. Iowa State: University Press, 1960.

Lowenberg-DeBoer, J. “Comments on Site-Specific Crop Management: Adoption Patterns and Incentives.” Review of Agricultural Economics. 22(1) (2000):245-247.

McCalla, A.F. “Prospects for Food Security in the 21st Century: with special Emphasis

on Africa.” Journal of Agricultural Economics. 20(1) (March 1999):95-103. McCullagh, P. and .A. Nelder. Generalized Linear Models. 2nd edition. Monographs on

Statistics and Applied Probability. New York: Chapman and Hall, 1989.

McGuirk, A. M., W.P. Preston and G.M. Jones. “Introducing Foods Produced using Biotechnology: The Case of Bovine Somatotropin.” Southern Journal of Agricultural Economics. (July 1992):209-223.

McNamara, K. T., M.E. Wetzstein, and G.K. Douce. “Factors Affecting Peanut Producer

Adoption of Integrated Pest Management.” Review of Agricultural Economics. 13 (1991):129-139.

Montgomery, D.C., E.A. Peck, and G.G. Vining. Introduction to Linear Regression Analysis.

3rd Edition. Wiley Series in Probability and Statistics. A Wiley-Interscience Publication. New York, New York: John Wiley & Sons, Inc. 2001

Mugisa-Mutetikka, M., A.F. Opio., M.A. Ugen., P. Tukamuhabwa., B.S. Kayiwa, C.

Niringiye and E. Kikoba. “Logistic Regression Analysis of Adoption of New Bean Varieties in Uganda.” Unpublished, 2000.

Mullen, J.D., G.W. Norton and D.W. Reaves. “An Economic Analysis of Environmental

Benefits of Integrated Pest Management.” Journal of Agricultural and Applied Economics. Southern Agricultural Economics Association. 29 (2) (December 1997):243-253

Nkonya, E., T. Schroeder, and D. Norman. “Factors Affecting Adoption of Improved Maize

Seed and Fertilizer in Northern Tanzania.” Journal of Agricultural Economics. 48 No. 1 (January 1997):1-12

Nowak, P. “Practical Considerations in Assessing Barriers to IPM Adoption.” Proceedings

of the Third National IPM Symposium/Workshop. (Eds.) S. Lynch, C. Green, and C. Kramer-LeBlanc. Washington, DC, USDA-ERS Misc. Pub No. 1542, 1996.

Nowak, P. “The Adoption of Agricultural Conservation Technologies: Economic and

Diffusion Explanations.” Rural Sociology. 52(2) (1987): 208-220. Odulaja, A. and F.G Kiros. “Modeling Agricultural Production on Small-scale Farmers in

Sub-Saharan Africa: A Case Study in Western Kenya.” Journal of Agricultural economics. 14(1996): 85-91.

Oerke, E. C., H.W Dehne., F. Schoenbeck and A. Weber. Crop Production and Crop Protection: Estimated Losses in Major Food and Cash Crops. Amsterdam, The Netherlands: Elsevier, 1994.

113

Ogrodowczyk, J.D. “Policies Affecting Production Practices and Adoption of Integrated Pest Management for Jamaican Farmers in Ebony Park, Clarendon.” MS thesis, Virginia Polytechnic Institute and State University, 1999.

Olson, K.D., J. Langley and E.O. Heady. “Widespread Adoption of Organic farming

Practices: Estimated Impacts on US agriculture.” Journal of Soil and Water Conservation. 37(January-February 1982):41-45.

Overfield, D. and E. Fleming. “A Note on the Influence of Gender Relations on the

Technical Efficiency of Smallholder Coffee Production in Papua New Guinea.” Manuscript. Journal of Agricultural Economics. (2001):153-156.

Parker, C. “Integrated Control of weeds in Sorghum.” Elements of Integrated Control of

Sorghum pests, FAO Plant Production and Protection Paper. 1980, Food and Agricultural Organization of the United Nations.

Pascale, G. Calculating marginal probabilities in PROC PROBIT. (Online)

http://www.ats.ucla.edu/stat/sas (1998). Accessed October, 2002. Patterson, K-A. “The political ecology of nontraditional agricultural exports and an IPM

project in Jamaica.” MS. Thesis. Virginia Polytechnic Institute and State University, 2000.

Pimentel, D. (ed.) Some Aspects of Integrated Pest Management. Department of

Entomology, Cornell University, Ithaca, NY, 1986. Rogers, E.M. Diffusion of Innovations. 3rd Edition. New York: The Free Press, 1983. __4th Edition. New York: The Free Press, 1995. Rwaboogo, M. O. Uganda Districts: Information handbook. 1997/98 Edition. Fountain

Publishers, 1998. Shakya, P. B. and J. C. Flinn, “Adoption of Modern Varieties and Fertilizer Use on Rice in

the Eastern Tarai of Nepal.” Journal of Agricultural Economics. 36(3) (September 1985):409-419.

SHAZAM. (Online) URL: http://econometrics.com Accessed October, 2002 Smith, G.S., M.E. Wetzstein and G.K Douce. “Evaluation of Various Pest management

Characteristics.” Southern Journal of Agricultural Economics. Southern Agricultural Economics Association. (December 1987):93-101.

SPSS for Windows. Release 10.0.1 SPSS Inc. October 1999. Strengthening National Agricultural Research Systems in Eastern and Central Africa:

A framework for Action. The World Bank Technical paper no. 290. The World Bank, Washington, DC, 1995.

Tjornhom, J.D. “Assessment of Policies and Socio-Economic Factors Affecting Pesticide

Use in the Philippines.” MS. thesis, Virginia Polytechnic Institute and State University, 1995.

114

Uganda Exports Promotion Board. Uganda Services Exporters Association. (Online)

URL: http://www.ugandaexportsonline.com. October, 2002. Uganda, Republic of, Background to the Budget 1998/99 (Budgeting for Poverty

Eradication). Ministry of Finance, Planning and Economic Development. Kampala. 1998a.

Uganda, Republic of, Establishment of National Agricultural Research Organization

(NARO). International Service for National Agricultural Research (ISNAR), 1988 Uganda, Republic of, Report on Uganda National Household Survey 1995-96 (Third

Monitoring Survey) Vol. 1. Main results of the Crop-Survey Module. Statistics Department. Ministry of Finance, Planning and Economic Development. Entebbe, 1997

Uganda, Republic of, Statistical Abstract. Ministry of Finance, Planning and Economic

Development. Entebbe, 1998b United Nations, Food and Agriculture Organization. The State of Food and Agriculture.

Rome, 1997. United States Bureau of the Census. International Data Base. (Online) URL:

http://www.census.gov accessed September 25, 2001, September 29, 2002. von Braun, J., T. Teklu and P. Webb. Famine in Africa: causes, responses and Prevention.

Book Review. John Hopkins University Press for the International Food Policy Research institute (IFPRI). Baltimore. pp 238, 1999.

Waller, B.E., C.W. Hoy., J.L Henderson., B. Stinner and C. Welty. “Matching Innovations

with Potential Users: A Case Study of Potato IPM practices.” Agriculture, Ecosystems and Environment. 70(1998):203-215.

Wightman, J. A, International Food Policy Research Institute: 2020 Vision News and

Views. November 1998. Wilson, E. “Famine and Poverty in the 21st Century.” The Unfinished Agenda: Perspectives

on Overcoming Hunger, Poverty, and Environmental Degradation. Per Pinstrup-Andersen and Rajul Pandya-Lorch. eds. 2001. Pp. 1-6.

World Bank Group Countries: Uganda, 2000. World Bank Participation Sourcebook. (Online)

www.worldbank.org/wbi/sourcebook. Accessed May, 2002 World Bank, Uganda Agriculture: A World Bank Country Study. The World Bank

Washington, D.C. 1993. Wu, J., and B.A. Babcock. “The Choice of Tillage, Rotation and Soil Testing Practices:

Economic and Environmental Implications.” American Journal of Agricultural Economics. 80 (August 1998):494-511.

115

Yamagiwa, T.J. “Analysis of Policies Affecting Pesticide Use in Ecuador.” MS thesis, Virginia Polytechnic Institute and State University, 1998.

Yaron, D., A. Dinar and H. Voet. “Innovations on Family farms: The Nazareth Region in

Israel.” American Journal of Agricultural Economics. American Agricultural Economics Association. (1992):361-370.

116

Appendix A: List of Acronyms

ANOVA Analysis Of Variance

BIFAD Board for International Food and Agriculture Development

CGIAR Consultative Group on International Agricultural Research

CIA Central Intelligence Agency

EIU Economic Intelligence Unit

FARMESA Farm-Level Applied Research Methods in Eastern and Southern Africa

FAO Food and Agricultural Organization of the United Nations

FoB Free on Board

GDP Gross Domestic Product

HIPC Heavily Indebted Poor Countries

IFPRI International Food Policy Research Institute

IITA International Institute of Tropical Agriculture

IMF International Monetary Fund

IPM Integrated Pest Management

IPM CRSP Integrated Pest Management Collaborative Research Support Program

MAAIF Ministry of Agriculture Animal Industry and Fisheries (Uganda)

MUK Makerere University (Uganda)

NARO National Agricultural Research Organization (Uganda)

NARS National Agricultural Research System

NGO Non-Governmental Organization

OLS Ordinary Least Squares

SAS Statistical Analytical System

SPSS Statistical Package for Social Scientists

UN United Nations

US United States

USAID United States Agency for International Development

USDA United States Department of Agriculture

VIF Variance Inflation Factor

117

KU MI

Uganda districts

Vegatation typesBuilt-up AreasBushlandButyrospermum SavannaDry Acacia SavannaDry Combretum SavannaDry ThicketsFarmlandForest/Savanna MosaicsGrass SavannaGrass SteppeHerbacious SwampHigh Alt ForestHigh Alt Moor & HeathMed Alt Moist Ever ForMed Alt Moist Semi-dec ForMoist Acacia SavannaMoist Combretum SavannaMoist ThicketsOpen WaterPalm SavannaSeasonal WetlandsSwamp ForestTree & Shrub SteppeWoodland

Latitude and Longitudes

200 0 200 Miles

NUganda: Vegetation

Appendix B1: Map of Uganda showing vegetation

Source: ArcView GIS, 2002

118

ppendix A2: Map of Kumi District Showing Location of Focal Points and surrounding area: Respondent Locations

ÊÚ

Uganda DistrictsKUMI District

ÊÚ Kampala

90 0 90 180 Miles

N

EW

S

UGANDA

#

KUMIMoroto

KapchorwaMbale

#

Tororo

#

Iganga

Soroti

#

Pallisa

Kamuli

Jinja

Appendix B2: Map of Uganda showing Study Area

118

119

##

#

#

EilorAkabwai

Olemukan

Ebukalin

BUKEDEYA

NGORA

KUMI

Kumi District CountiesBukedeaKumiNgora

Distance from Focal Points50005000 - 1000010000 - 15000

# Focal Farmers

20 0 20 Miles

NMap of Kumi

Appendix B3: Map of Kumi District Showing Location of Focal Points and surrounding areas

120

Appendix C: Introductory Letter

Dear Farmer,

As you may already know, your district is one of the areas in the country where

the USAID-supported Integrated Pest Management (IPM) project has been

operating for the past several years. With this project’s effort the use of IPM

has been advanced as an alternative to conventional pest and disease control.

Scientists on the project working initially with pioneering farmers have

developed practices that are available for use by farmers. However it is not

clear how the developed practices are currently being used and what the

perception of farmers regarding these technologies in the project’s area of

influence are.

This questionnaire is designed to obtain this information. You have been

selected as a source of this information. The attached questionnaire will ask

about several aspects of your farming. Confidentiality will prevail if you so wish

since your name will not appear on the questionnaire. Results of this study will

be used by IPM researchers and program administrators to evaluate the

program in the areas of its operation.

Feel free to provide any additional information that you think may be useful in

this analysis. Your responses will be highly appreciated.

Sincerely,

Jackline Bonabana

Department of Agricultural and Applied Economics,

Virginia Tech

121

Appendix D: Factors Affecting Adoption of IPM Technologies in Kumi-Uganda

- Survey Form This questionnaire is designed to solicit for your responses on factors affecting adoption on IPM. Your responses will be used for academic purposes only and are highly appreciated. Date of Interview _________________________ Interviewer___________________ District _________________________ County:______________________ Sub-county: __________________________ Parish_______________________ Village: _________________________ Respondent ID_________________ A: Demographic Information 1. Age ________ Years (Code:A1) 2. Marital status (Code:A2) _1___Single _2___Married _3___Divorced _4___Widowed _5___ Separated _6___ Other specify __________________________________ 3. Gender ____Female (1)_____Male (2) (Code:A3) 4. Total number of years of schooling ____________________ (Code:A4) 5. Number of people living in your household? ____________________ (Code:A5) 6. For how long have you been a farmer? ____________________Years (Code:A6) B: General/Background 7. What is the total size of your farm? ____________________Hectares (Code:B7) 8. What crops do you grow? (Tick all that apply) ______________ What is the acreage on each

Crop Yes (1) No (2) Code (B81) Acreage Code (B82) 1 Cowpea (B811) (B821) 2 Sorghum (B812) (B822) 3 S. Potato (B813) (B823) 4 G. Nuts (B814) (B824) 5 Millet (B815) (B825) 6 Beans (B816) (B826) 7 Maize (B817) (B827) 8 Bananas (B818) (B828) 9 Tomato (B819) (B829) 10 Cotton (B8110) (B8210) 11 Simsim (B8111) (B8211) 12 Rice (B8112) (B8212) 13 Soyabean (B8113) (B8213) 14 Cassava (B8114) (B8214) 15 Eucalyptus (B8115) (B8215) 16 Sunflower (B8116) (B8216) Others (Specify) 17 (B8117) (B8217)

122

9. How do you access agricultural information? Sources Yes (1) No (2) Code 1 Radio B91 2 Newspapers B92 3 Bulletins B93 4 Friends B94 5 Neighbours B95 6 Farmers’ organisations B96 7 MAAIF Extension staff B97 8 MUK researchers B98 9 NGO B99 10 Others B910

10. How many extension/NGO contacts do you have in a period of a year ___None (1) ___Few (2) ___Many (3) ___Don’t know (4) (Code:B10) 11. Do you belong to a farmer organization? ___Yes (1) ___No (2) (Go to 13) (Code:B11) 12. How many times do you attend meetings in a period of one year? ___________ (Code:B12) 13. Do you ever borrow to finance crop production? ____Yes (1) ____No (2) (Code:B13) 13.1. If no, why?

Reason Yes (1) No (2) Code 1 Not available B1311 2 Interest Rate is high B1312 3 Don’t know B1313 4 Other reason B1314

13.2 What rate is normally charged _______(Code:B132) 14. How many household members work on the farm _______(Code:B14) 15. How many household members work off the farm? _______(Code:B15) 16. Do you ever hire laborers to work on your farm? ____Yes (1) ____No (2) (Code:B16) 17. Who decides what inputs to buy?

Male(1) Female(2) Both(3) Do not buy (4)

Code

1 Fertilizers B171 2 Seed B172 3 Pesticides B173 4 Farm Implements B174

C. Knowledge of IPM 18. Have you ever heard of the term IPM? (Enumerator prompts by defining IPM) ___Yes (1) ___No (2) (Go to 27) ___Don’t Know (3) Go to 27 (Code:C18) 19. When and where did you first hear of the term IPM?

When (Code:C191) Yes (1) No (2) 1 1994-1996 2 1997-1999 3 2000-2002

Where (Code:C192) Yes (1) No (2) 1 Makerere University 2 MAAIF 3 Farmer Organisation meetings

20. Have you ever been invited to attend IPM meetings?

___Yes (1) ___No (2) (Go to 24) (Code:C20) 21. How many times? ___ (Code:C21)

123

22. Did you attend? ___Yes(1) ___No(2) (Go to 24) (Code:C22) 23. How many times? ___ (Code:C23) 24. Have you ever participated in IPM on-farm trial demonstrations?

___Yes(1) ___No(2) (Code:C24) 25. Have you ever tried out any of the methods as specified by IPM?

___Yes(1) ___No(2) (Code:C25) 26. What is your view on the requirements of practicing IPM? (Tick all that apply).

Compared to the conventional means, practicing general IPM involves 1. Mgt time

(C261) 2. Cost (C262)

3. Knowledge (C263)

4. Labor (C264)

5. Land (C265)

6. Other (Specify) (C266)

More (1) Equal (2) Less (3) Don’t Know (4)

27. Have you ever attended other training on pest control? ___Yes (1) ___No (2) (Go to 29) ___Don’t know (3) (Go to 29) (Code: C27)

28. What was the training about? Explain (Code: C28) 1._________________________________________________________________________ 2. _________________________________________________________________________ 3. _________________________________________________________________________ 29. Do you think chemicals can sometimes be harmful? (Code: C29) ___Yes (1) ___No (2) (Go to next section) ___Don’t know (3) (Go to next section) 30. Do you think chemicals can harm crops? (Code: C30) ___Yes (1) ___No (2) ___Don’t know (3) 31. Do you think chemicals can cause sickness to humans? (Code: C31) ___Yes (1) ___No (2) ___Don’t know (3) 32. Do you think chemicals can cause sickness to farm animals? (Code: C32) ___Yes (1) ___No (2) ___Don’t know (3) 33. Do you think chemicals can contaminate drinking water? (Code: C33) ___Yes (1) ___No (2) ___Don’t know (3) 34. Do you think chemicals can cause sickness to other living organisms, birds, fish and

water creatures, including natural enemies of insects? ___Yes (1) ___No (2) ___Don’t know (3) (Code: C34) D. Crop Specific Sorghum I. Technologies a. Crop Rotation technology 1. The last time you grew sorghum, are there crops that you grew on that piece of land before

you planted sorghum? ___Yes (1) ___No (2) ___Don’t know (3) (Code:D1) 2. What rotation did you use? Yes(1) No(2) Code:D2

1 Groundnuts/Sorghum D21 2 Groundnut/Sorghum/Cassava D22 3 Sorghum/Cotton D23 4 Cotton/Sorghum/Cowpea D24 Other

124

2.1. What is your view on the requirements of rotating sorghum/cowpeas? (Tick all that apply). If rotating Sorghum with Cowpeas is not your conventional means, compared to the conventional means, it involves 1. Mgt time

(L211)

2. Cost (L212)

3. Knowledge (L213)

4. Labor (L214)

5. Land (L215)

6. Other (Specify) (L216)

More (1) Equal (2) Less (3)

2.2. If given a random choice to rotate sorghum/cowpea or not to rotate them, which one would you choose? ___Rotate (1) ___Not rotate (2) (CodeD221) b. Variety technology 3. What varieties did you grow last season? And on what amount of land? Variety Yes

(1) No (2)

Variety Code: D31

Amount of land on which it was grown (Acres)

Acreage Code: D32

1 Seredo D311 D321 2 Sekedo D312 D322 3 Eidima D313 D323 4 Local D314 D324 5 Don’t know D315 D325 Other 6 D316 D326

3.1. What is your view on the requirements of growing improved Sorghum varieties? (Tick all that apply). Compared to the conventional means, it involves

1. Mgt time (M311)

2. Cost (M312)

3. Knowledge (M313)

4. Labor (M314)

5. Land (M315)

6. Other (Specify) (M316)

More (1) Equal (2) Less (3)

c. Celosia Argentia technology 4. Have you heard about the plant that controls (chases) the striga weed? ___Yes (1) ___No (2) (Go to 6) (Code: D41)

What is it? ___Celosia(1) ___Local weed (2) (Code: D42) 5. Did you grow that plant in your/sorghum field the last season you grew sorghum?

___Yes (1) ___No (2) (Code: D5) 6. In what season did you grow sorghum the last time you grew sorghum?

___First (1) ___Second (2) ___Both (3) ___Don’t know (4) (Code: D6) d. Fertilizer technology 7. Do you use fertilizer in sorghum fields? ___Yes (1) ___No (2) (Code: D7) 7.1 What is your view on the requirements of fertilizer use? (Tick all that apply).

If fertilizer use on sorghum is not your conventional means, compared to the conventional means, fertilizer use involves

125

1. Mgt time

(N711) 2. Cost (N712)

3. Knowledge (N713)

4. Labor (N714)

5. Land (N715)

6. Other (Specify) (N716)

More (1) Equal (2) Less (3) Don’t know

II. General 8. How much was the total sorghum yield last season Variety Yield the last time you

grew crop (bags) Yield Code:

1 Seredo D81 2 Sekedo D82 3 Eidima D83 4 Local2 D84 5 Don’t know D85 6 Other (specify) D86

9. In the last crop season that you grew sorghum, was your sorghum crop harmed by insects/diseases/weeds? Yes (1) No (2) Don’t know (3) Code: D9 1 Insects D91 2 Diseases D92 3 Weeds D93

10. Name the most important weed/insect/disease on the Sorghum crop the last season you grew the crop____

Name Code 1 Insects D101 2 Diseases D102 3 Weeds D103

11. How did you get rid of the problem?

Hand Removal/rouging

Code (D111)

Inter-cropped

Code (D112)

Changed Variety in next season

Code (D113)

1 Insects D1111 D1121 D1131 2 Diseases D1113 D1122 D1132 3 Weeds D1123 D1133

Insecticide Code D1141 Fungicide Code D1151 Did not control Yes

(1) No (2)

No. of sprays (D1142)

Yes(1) No(2) No. of sprays Code (D1152)

Tick Code (D116)

1 Insects D1161 2 Diseases D1162 3 Weeds

D1163

126

E. Crop Specific: Cowpea I. Technologies a. Time of planting 12. In what season did you grow cowpea the last time you grew the crop?

___First (1) ___Second (2) ___ Both (3) ___Don’t know (4) (Code: E12) 13. In the last crop season, when did you plant cowpea in the field relative to the start of

rains? Time of planting First Code (E131) Second Code (E132) 1 First sign of rains E1311 E1321 2 One week after first sign of rains E1312 E1322 3 Two weeks after first sign of rains E1313 E1323 4 Towards the end of the rainy

season E1314 E1324

13.1. What is your view on the requirements of timely planting? (Tick all that apply). Compared to the conventional means, timely planting involves 1. Mgt time

(O1311) 2. Cost (O1312)

3. Knowledge (O1313)

4. Labor (O1314)

5. Land (O1315)

6. Other (Specify) (O1316)


b. Plant Spacing 14. How did you plant cowpea in the ground? Yes No (Code: E14) 1 Broadcast E141 2 Chop and drop E142 3 In lines E143 4 Other……………………… E144

15. Would you please show me how you space your cowpea seeds when planting? (Enumerator takes measurements) Measurements Yes (1) No (2)

Code E15

1 (15x30) cm E151 2 (15x45) cm E152 3 (15x60) cm E153 4 (30x10) cm E154 5 (30x45) cm E155

15.1. What is your view on the requirements of close spacing plants at (30X10)? (Tick all that

apply). Compared to the conventional means, it involves 1. Mgt time

(P1511) 2. Cost (P1512)

3. Knowledge (P1513)

4. Labor (P1514)

5. Land (P1515)

6. Other (Specify) (P1516)


127

c. Intercropping 16. Did you intercrop cowpea the last time you grew the crop? ___Yes(1) ___No (2)

(Go to 18) (Code E16) 17. What crops did you intercrop cowpea with? (Code: E17) Crops Yes (1) No (2) Code 1 Beans E171 2 Maize E172 3 Groundnuts E173 4 Sorghum E174 5 S. Potato E175 6 Soyabean E176 7 Cotton E177 8 Millet E178

17.1. What is your view on the requirements of practicing IPM? (Tick all that apply). If intercropping is not your conventional means, compared to the conventional means, intercropping involves 1. Mgt time

(Q1711) 2. Cost (Q1712)

3. Knowledge (Q1713)

4. Labor (Q1714)

5. Land (Q1715)

6. Other (Specify)(Q1716)


d. Defoliation 18. Do you ever remove cowpea leaves from the plant in the field? ___Yes (1) ___No (2)

(Go to 21) (Code: E18) 19. How many times do you do this in a growing season? ___ (Code E19) 20. What reason did you have for doing this? (Enumerators not to prompt) (Code: E20) Reason Yes (1) No (2) (Code: E20) 1 Food value E201 2 Pest control E202 3 Disease control E203 4 Medicinal value E204 5 Weed control E205 6 Flowering/Plant Health E206 7 Other E207

e. Variety 21. What varieties of Cowpea do you usually grow? (Code: D21) Variety Yes (1) No (2) Code 1 Ecirikukwai E211 2 Ebelat E212 3 Large White E213 4 Brown Tan E214 5 Other……………………………. E215

128

21.1 What is your view on the requirements of growing an improved cowpea variety? (Tick all that apply). If growing improved cowpea varieties is not the conventional means, compared to the conventional means, growing an improved variety involves

1. Mgt time (S2111)

2. Cost (S21112)

3. Knowledge (S2113)

4. Labor (S2114)

5. Land (S2115)

6. Other (Specify) (S2116)

More (1) Equal (2) Less (3) Don’t Know

II. General 22. How much was the total cowpea yield the last time you grew the crop? (Code:E22) Variety Yield (bags) Yield code 1 Ecirikukwai E221 2 Ebelat E222 3 Large White E223 4 Brown Tan E224 5 Other………………………… E225

23. In the last crop season was your crop harmed by insects/diseases/weeds? Yes (1) No (2) (Go to 26) Don’t know

(3) (Go to 26) Code: E23

1 Insects E231 2 Diseases E232 3 Weeds E233

24 Which was the most important weed/insect/disease on the cowpea crop? Name Code: E24 1 Insects E241 2 Diseases E242 3 Weeds E243

25. How did you get rid of the problem? (Code E25)

Hand Removal

Code (E251)

Inter-cropped

Code (E252)


Code (E253)

1 Insects E2511 E2521 E2531 2 Diseases E2522 E2532 3 Weeds E2513 E2523 E2533

Insecticide Code (E2541) Fungicide (Code E2551) Did not control Yes

(1) No (2)

No. of sprays Code (E2542)

Yes (1)

No (2)

No. of sprays Code (E2552)

Tick Code (E256)

1 Insects E2561 2 Diseases E2562 3 Weeds

E2563

F. Crop Specific: Groundnuts I. Technologies a. Time of planting 26. In what season did you grow groundnuts the last time you grew the crop? ___First (1) ___Second (2) ___Both (3) ___Don’t know (4) (Code: F26) 27. In the last crop season, when did you plant cowpea in the field relative to the start of

129

rains? Time of planting First Code (F271) Second Code (F272) 1 First sign of rains F2711 F2721 2 One week after first sign of rains F2712 F2722 3 Two weeks after first sign of rains F2713 F2723 4 Towards the end of the rainy

season F2714 F2724

b. Plant Spacing 28. How did you plant groundnuts in the ground? Yes No Code (F28) 1 Broadcast F281 2 Chop and drop F282 3 In lines F283 4 Other……………………… F284

29. Would you please show me how you space your groundnut seeds when planting? (Enumerator takes measurements) Measurements Yes (1) No (2) Code F29 1 (15x30) cm F291 2 (15x45) cm F292 3 (15x60) cm F293 4 (30x10) cm F294 5 (30x45) cm F295

29.1 What is your view on the requirements of measuring plant spacing? (Tick all that apply). Compared to the conventional means, it involves

1. Mgt time (U2911)

2. Cost (U2912)

3. Knowledge (U2913)

4. Labor (U2914)

5. Land (U2915)

6. Other Specify (U2916)


c. Intercropping 30. Did you intercrop groundnuts the last time you grew the crop? ___Yes(1) ___No (2)

(Go to 32 ) (Code F30) 31. What crops did you intercrop groundnuts with? (Code: F31) Crops Yes (1) No (2) Code 1 Beans F311 2 Maize F312 3 Cowpeas F313 4 Sorghum F314 5 S. Potato F315 6 Soyabean F316 7 Cotton F317 8 Millet F318

31.1 What is your view on the requirements of, intercropping? (Tick all that apply).

If intercropping is not the conventional means, compared to the conventional means, it involves

130

1. Mgt time (V3111)

2. Cost (V3112)

3. Knowledge (V3113)

4. Labor (V3114)

5. Land (V3115

6. Other Specify (V3116)


d. Variety technology 32. What varieties did you grow last season and on what amount of land? (Code: F32) Variety Yes

(1) No (2)

Variety Code: F321

Amount of land on which it was grown (Acres)

Acreage Code: F322

1 Igola1 F3211 F3221 2 Rudu White F3212 F3222 3 Rudu (Red) F3213 F3223 4 Etesot F3214 F3224 5 Serenut1 F3215 F3225 6 Serenut2 F3216 F3226 7 Tanto F3217 F3227 8 Matudda F3218 F3228 9 Roxo F3219 F3229 10 Don’t know F32110 F32210

32.1 What is your view on the requirements of growing improved groundnut varieties? (Tick all that apply). If growing improved groundnut varieties is not the conventional means, compared to the conventional means, variety change involves

6. Other (Specify) 1. Mgt time (W3211)

2. Cost (W3212)

3. Knowledge (W3213)

4. Labor (W3214)

5. Land (W3215) W3216)

More (1) Equal (2) Less (3) Dont know

e. Spray Schedule 33. Do you use any chemical spray on groundnut fields (by variety)? ___Yes

(Tick all that apply) ___No (Go to 38) Variety Yes (1) No (2) Variety

Code: F331 Number of sprays in the last season you grew crop

Code: F332

1 Igola1 F3311 F3321 2 Rudu White F3312 F3322 3 Rudu (Red) F3313 F3323 4 Etesot F3314 F3324 5 Serenut1 F3315 F3325 6 Serenut2 F3316 F3326 7 Tanto F3317 F3327 8 Matudda F3318 F3328 9 Roxo F3319 F3329 10 Don’t know F33110 F33210

34. How do you decide which chemical to use? _____________________________________________________________________ (Code: F34)

131

35. How do you decide how much to use? _____________________________________________________________________(Code: F35)

36. How do you decide when to spray? _____________________________________________________________________(Code: F36)

37. At what stage of crop development did you use them in the last season? __________________________________________________________ (Code: F37) 37.1 What is your view on the requirements of spraying using the IPM recommended

practice? (Tick all that apply). Compared to the conventional means, it involves 1. Mgt time

(X3711) 2. Cost (X3712)

3. Knowledge (X3713)

4. Labor (X3714)

5. Land (X3715)

6.Other (Specify) (X3716)


II. General 38. How much was the total (In shell) groundnut yield the last time you grew the crop? Variety Yield the last time you

grew crop (Bags) (In-shell) Yield Code: F38

1 Igola1 F381 2 Rudu White F382 3 Rudu (Red) F383 4 Etesot F384 5 Serenut1 F385 6 Serenut2 F386 7 Tanto F387 8 Matudda F388 9 Roxo F389

39. In the last crop season was your crop affected by insects/diseases/weeds? Yes

(1) No (2) (Go to 42)

Don’t know (3) (Go to 41)

Code: F39

1 Insects F391 2 Diseases F392 3 Weeds F393

40. Which was the most important weed/insect/disease on the groundnut crop? Name Code: F40 1 Insects F401 2 Diseases F402 3 Weeds F403

41. How did you get rid of the problem? (Code F41)

Hand Removal

Code (F411)

Inter-cropped

Code (F412)


Code (F413)

1 Insects F4111 F4121 F4131 2 Diseases F4112 F4122 F4132 3 Weeds F4113 F4123 F4133

132

Insecticide Code F4141 Fungicide Code F4151 Did not control Yes (1)

No (2)

No. of sprays Code (F4142)

Yes (1)

No (2)

No. of sprays F4152

Tick Code (F416)

1 Insects F4161 2 Diseases F4162 3 Weeds

F4163

G: General for all crops 42. Did you use any chemicals on any other crops in the last growing season? Yes (1) No (2) Code (G42) 1 Pesticides G421 2 Fertilizers G422

H: Household Income (We do not want to know the amount, but just the sources) 43. Do you have other sources of household income outside the farm? ___Yes (1)___No (2)

If Yes, please name them_______________________________________________ (Code H43)

133

Appendix E: Tables Appendix Table E1: Distribution and Sources of off-farm income in study area.

Farmers with income source Off-farm income source n %

Professional/Skill 5.18 Salary/Pension 7 3.30 Allowance from LC 2 0.94 Contracts for feeder road construction 2 0.94

Manual jobs 30.66

Casual labor 31 14.62 Selling out labor 11 5.19 Building 1 0.47 Charcoal burning 4 1.89 Brick making 8 3.77 Bicycle riding (Bodaboda) 10 4.72

Remittances 8.02

From children, sisters 12 5.66 From friends 5 2.36

Business 45.28

Petty trading 15 7.08 Local brew 68 32.08 Food vending 7 3.30 Selling sisal 2 0.94 Arts and Crafts 1 0.47 Baking 1 0.47 Selling firewood 1 0.47 Grinding mill 1 0.47

Appendix Table E2: Sorghum varieties grown in Kumi District Variety Farmers with variety n % Abili 4 1.90 Eera 17 8.10 Ekoli 10 4.76 Emumwailo arengan 3 1.43 Epurpur 1 0.48 Erepet 24 11.43 Serena 20 9.52 Etanzaniat 1 0.48 Ilungole 2 0.95 Ilodir 18 8.57 Eterai 1 0.48 Red head 3 1.43 Ekonokamu 1 0.48 Ikanyawa 1 0.48 Black variety 2 0.95 White head 4 1.90

134

Appendix Table E3: Weed species in sorghum in the study area Frequency of Occurrence Weed name (local)

n % Striga 83 39.52 Spear grass 16 7.62 Couch grass 13 6.19 Ekolet 4 1.90 Goat weed 2 0.95 Esioto 1 0.48 Simama 3 1.43 Emuriat 1 0.48 Ekosile 1 0.48 Ipunuka 1 0.48 Ekoropot 1 0.48 Emoppim 2 0.95 Esirike 2 0.95 Esisio 1 0.48 Appendix Table E4: Collinearity Diagnostics Results

Retained Dropped*

FMEXP RFMEXP (r=0.82), AGE (r=0.87)FMLBR HHSZ (r=0.72)IMPLPURCH SEEDPURCH (r=0.69)RSCH EXT (r=0.56)ONFTR OWNIPM (r=0.49), HDIPM (r=0.40)YDSKDO ACRESKDO (r=0.86)

SGDZZ SGINSECT (r=0.49) ROTNMGT ROTNLBR (r=0.55), ROTNCOST (r=0.62)EBACRE EBYD (r=0.51)TPCPLBR TPCPMGT (r=0.58), TPCPCOST (r=0.64)EBMGT EBLBR (r=0.52)CPWEED CPDZZ (r=0.60)

ICCPMGT ICCPCOST (r=0.66), ICCPKNOW (r=0.45),ICCPLBR (r=0.59)

IGOLAYD IGOLACRE (r=0.67), TOTGNYD (r=0.6)GNDZZ GNINSECT (r=0.57), GNWEED (r=0.53)CLSPLBR CLSPCOST (r=0.62), CLSPKNOW (r=0.46),

CLSPMGT (r=0.46)

* The dropped variables are correlated with the retained variables(as measured by the value of thecorrelation coefficient)

Groundnut

Correlations between Variables

General

Sorghum

Cowpea

135

Vita

Jackline Bonabana-Wabbi was born in Kampala, Uganda to Mr. and Mrs. J.

Baitwababo in 1972 and went through the education system in Uganda. She was

awarded a two-year high-school scholarship that saw her through to University

education. She completed her Bachelor of Science in Agriculture from Makerere

University in 1997. After graduating, she worked in the Office of the Vice President

under contract with the Danish International Development Agency (DANIDA). Her

duties involved providing technical agricultural support to the Office of the Vice

President through research on agricultural issues and linking the Office to the

Ministry of Agriculture, Animal Industry and Fisheries. At the end of the

contractual arrangement, she was hired as an Assistant Lecturer by Makerere

University. She received her Master of Science in Agricultural and Applied

Economics at Virginia Polytechnic Institute and State University in December

2002. After graduation, she resumed her duties at Makerere University.

Assessing Factors Affecting Adoption of Agricultural ... · Assessing Factors Affecting Adoption of...

Documents

Transcript of Assessing Factors Affecting Adoption of Agricultural ... · Assessing Factors Affecting Adoption of...