agricultural biodiversity, farm level technical …€¦ · this dissertation is dedicated to: to...
TRANSCRIPT
i
AGRICULTURAL BIODIVERSITY, FARM LEVEL TECHNICAL EFFICIENCY
AND CONSERVATION BENEFITS: AN EMPIRICAL INVESTIGATION
THIS DISSERTATION IS SUBMITTED TO THE FACULTY OF BUSINESS,
QUEENSLAND UNIVERSITY OF TECHNOLOGY FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
MAY 2012
K.M.R. Karunarathna
B. A (Economics) Hons., University of Peradeniya, Sri Lanka
M.Sc. (Environmental Economics), University of Peradeniya, Sri Lanka
School of Economics and Finance
QUT Business School
Queensland University of Technology
Gardens Point Campus
Brisbane, Australia
ii
Statement of Original Authorship
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the
best of my knowledge and belief, the thesis contains no material previously
published or written by another person except where due reference is made.
…………………………………….
K. M. R. Karunarathna
21st May, 2012
iii
This dissertation is dedicated to:
To my loving husband, Wasantha son, Kavindu and daughter, Disuni
To my mother, father and all who helped me to make it true
iv
ACKNOWLEDGEMENTS
I greatly acknowledge the assistance I received from numerous individuals and
institutions for completing this research. Special thanks should go to my advisers,
Professor Clevo Wilson and Professor Tim Robinson, for their constant support and
guidance throughout my graduate program. Their kindness, patience, and continual
coaching are greatly appreciated. They encouraged me to carry out this interesting
dissertation research and for their invaluable advice, guidance, endless
encouragement and untiring efforts to make it a success. They provided a stimulating
environment with productive discussion throughout the dissertation research that
helped make me a better researcher. I am grateful to them for their support and
wisdom, and the kindhearted assistance extended to me throughout the study period.
I am also thankful for the invaluable help and encouragement I received from my
dissertation committee members Dr. Mark McGovern, Dr. Henri Burgers, Prof. Tim
Robinson and Prof. Clevo Wilson. I also would like to thank the panel members of
my PhD confirmation seminar, especially Dr. Louisa Coglan, for her constructive
comments.
People who are living in Anuradhapura, Kurunegala and Ampara districts deserve
my thanks for their cooperation in the data gathering effort. I greatly appreciate the
help given by many individuals including enumerators and government officers
during the data collection process. I thank the University of Peradeniya for granting
me study leave, staff members in the Department of Economics and Statistics who
encouraged me to pursue my postgraduate studies at the Queensland University of
Technology in Australia.
v
I must recognize the constant help given by my colleagues at the School of
Economics and Finance, for their assistance and cooperation throughout the course
of study. I am also thankful for the invaluable help and encouragement I received
during my QUT life from Dr. Tony Sahama in the faculty of IT. I also should thank
to Dr. Jeanette who helped me correct English in this dissertation. I thank
participants of local and international conferences for providing useful feedback and
facilitating discussion on this work that I have presented to them. I have benefited a
lot from working with them.
I gratefully acknowledge the role of Queensland University of Technology for
providing financial support for my graduate studies. It is only with the help of QUT’s
IPRS scholarship, I was able to undertake this study in Australia. I therefore
acknowledge and thank QUT for awarding me this scholarship. Further, I gratefully
acknowledge the role of National Centre for Advanced Studies in Humanities and
Social Sciences (NCAS) for providing financial support for my PhD research. I am
also thankful to Professor Tim Robinson, former head of the school, School of
Economics and Finance, and all other administrative staff of the faculty of business
for their invaluable service received during my study period at QUT.
Last but not least I wish to express my deep gratitude to my husband, Wasantha for
his understanding, patience and encouragement throughout my graduate studies. I am
indebted to my loving son, Kavindu and daughter, Disuni. As I had to spend
considerable time on this study, they missed their mum during the time in the first
few years in their life. Finally, I am deeply grateful to my beloved mother for her
invaluable contribution throughout my life. I also owe a debt of gratitude to my late
vi
father. I also acknowledge my brother, sisters and their families, for their
unconditional love inspiration and encouragement throughout my life.
vii
TABLE OF CONTENTS
STATEMENT OF ORIGINAL AUTHORSHIP …………………………… ii
DEDICATION………………………………………………………………. iii
ACKNOWLEDGEMENTS ……………………………………………….... iv
TABLE OF CONTENTS…………………………………………….……… vii
LIST OF TABLES……………………………………………………….….. xi
LIST OF FIGURES…………………………………………………………. xii
LIST OF ABBRIVIATION…………………………………………………. xiii
ABSTRACT……………………………………………………………….…
xv
CHAPTER 1: INTRODUCTION……………………………………….………. 1
1.1 Overview …………………………………………………….…..……. 1
1.2 Motivation …………………………………………………….………. 13
1.3 Expected contributions of the study…………………………….……... 16
1.4 Structure of the thesis………………………………………….…….…
18
CHAPTER 2: STATUS AND TRENDS OF BIODIVERSITY IN SRI LANKA 20
2.1 Biodiversity wilderness area in the world………………………….….. 20
2.2 Biodiversity in Sri Lanka ……………………………….………….…. 22
2.3 Present status and future challenges of biodiversity…………….…….. 26
2.4 Agricultural biodiversity in the country…………………………….…. 33
CHAPTER 3: DATA SOURCES AND DESCRIPTION……………….……….. 39
3.1 Introduction ……………………………………………….….………... 39
3.2 Selecting appropriate sample size…………………………….………... 40
3.3 Selecting respondents for the survey…………………………………… 44
3.4 Field survey and its content……………………………………………. 47
3.5 Design choice experiment survey……………………………………… 49
viii
CHAPTER 4: FARMERS’ VALUATION OF AGRICULTURAL
BIODIVERSITY
57
4.1 Introduction………………………………………………………..…..… 57
4.2 Literature review on valuation of agricultural biodiversity……….…...… 60
4.3 Random utility models………………………………………………...… 64
4.4 Choice experiment method …………………………………………....… 70
4.5 Choice experiment design and model selection……………………..….... 76
4.6 Empirical approach to choice experiments study……………….…....….. 82
4.7 Socio-economic profile of sample respondents……………………....….. 91
4.8 Data cording and estimation procedure……………………………….….. 94
4.9 Result of the conditional logit model (CLM)………………………….…. 96
4.10 Result of the CLM including attributes and socioeconomic variables…. 103
4.11 Result of the random parameter logit model………………………….… 108
4.12 Estimating welfare changes with changing attributes and their level…... 110
4.13 Summary and key findings……………………………………………… 116
CHAPTER 5: FACTORS INFLUENCING FARMERS’ DEMAND FOR
AGRICULTURAL BIODIVERSITY
119
5.1 Introduction ………………………………………………………..….. 119
5.2 Literature review on demand for agricultural biodiversity…………...... 121
5.3 Derivation of demand for agricultural biodiversity………..……….….. 128
5.4 Empirical model specification and relevant variables……….…….….. 135
5.5 Theoretical approaches for the relevant models…………..…….….….. 143
5.5.1 Poisson regression model……………………………………..... 144
5.5.2 Negative binomial (NB2) regression model………………….… 148
5.5.3. Empirical tests for different count data models……………...… 152
5.6 Socio-economic characteristics of the households………………….… 155
5.7 Determinants of crops variety demand…….…………………….…… 158
5.8 Determinants of livestock variety demand.……………………….….. 166
5.9 Summary and key findings …………………………………………… 169
ix
CHAPTER 6: FARMERS’ PREFERENCES FOR DIFFERENT FARMING
SYSTEMS
173
6.1 Introduction………………………………………………………...…. 173
6.2 Literature review on farmer’s preference for different farming systems 175
6.3 Methods of explaining farmer’s preferences………………………….. 180
6.4 Factors influencing the selection of landrace cultivation………..……. 187
6.5 Factors influencing the selection of organic farming …………………. 190
6.6 Farmers’ demand for mix farming system…………………………….. 194
6.7 Summary and key findings……………………………………….…... 198
CHAPTER 7: AGRICULTURAL BIODIVERSITY AND FARM LEVEL
EFFICIENCY
201
7.1 Introduction…………………………………………………………… 201
7.2 Literature on agricultural biodiversity and farm level efficiency……... 204
7.3 Method of estimating farm level technical efficiency……………..….. 209
7.4 Empirical model of estimation………………………………………… 215
7.5 Estimates for parameters of stochastic frontier production function….. 220
7.6 Estimating marginal productivity and input elasticity………….…...… 226
7.7 Variations of technical efficiency………………………….………...... 228
7.8 Results of the inefficiency model……………………………………... 233
7.9 Summary and key findings………………………………………….… 238
CHAPTER 8: CONCLUSIONS AND POLICY IMPLICATIONS…………… 241
8.1 A summary of findings and discussion………………………….……. 241
8.2 Policy implications……………………………………………….…… 247
8.3 Limitations of the study and further research………………….………
251
BIBLIOGRAPHY……………………..………………………………………...
255
x
APPENDIX A (1): Defining agricultural biodiversity..............................................
288
APPENDIX A (2): TEV of agricultural biodiversity on small-scale farms.............. 289
APPENDIX A (3): Defining TEV of agricultural biodiversity on farms……….… 290
APPENDIX B: Number of described species in the World………………….….… 291
APPENDIX C: Biodiversity wilderness areas in the world…………………….…. 292
APPENDIX D (1): Topography in Sri Lanka………………………………….….. 293
APPENDIX D (2): Major climatic zones in Sri Lanka…………………………..... 294
APPENDIX E: Protected areas under department of wildlife in Sri Lanka……..… 295
APPENDIX F: List of protected areas of Sri Lanka……………………………..... 296
APPENDIX G: Map showing survey areas in Sri Lanka………………………...... 297
APPENDIX H: Questionnaire used in the survey……………..………….…….….. 298
APPENDIX I(1): A sample choice set is given to the respondent…………….….. 322
APPENDIX I(2): Description of 36 choice sets of the choice experiment………... 323
APPENDIX J: Descriptive statistics of the sample respondents.……………….… 324
APPENDIX K: Zero inflated Poisson / negative binomial regression model….….. 327
APPENDIX L: MLE of parameters and point estimates of TE………………....… 330
APPENDIX M: Derivatives of elasticities using translog production function…… 335
APPENDIX N(1): List of crops varieties on small-scale farms…………………… 336
APPENDIX N(2): List of livestock breeds on small-scale farms……………….….
337
xi
LIST OF TABLES
Tables Page
Table 2.1: The list of recorded species in different taxonomic groups………... 27
Table 2.2: Estimated number of selected species …………………………….. 29
Table 2.3: Natural ecosystem richness…………………………………........... 32
Table 3.1: Estimating minimum sample size for each district……………….... 43
Table 3.2: Details of the survey areas………………………………………..... 45
Table 4.1: Classifications of small-scale farm attributes in the CE survey…… 85
Table 4.2: Attributes and their levels……………………………………….…. 87
Table 4.3: Example of a choice set………………………………………....…. 89
Table 4.4: Individual attributes for the estimation of CL and RPL models….... 90
Table 4.5: Regression results of the CL model ………………………….…..... 99
Table 4.6: Test of independence of irrelevance alternatives…………………... 102
Table 4.7: CL model including attributes and socioeconomic variables….…... 107
Table 4.8: Regression results of the RPL model……………………..…….…. 109
Table 4.9: Implicit price estimates for attributes………………………...….… 111
Table 4.10: Estimates of WTA for various scenarios: Ampara……………..… 113
Table 4.11: Estimates of WTA for various scenarios: Anuradhapura……..….. 114
Table 4.12: Estimates of WTA for various scenarios: Kurunegala……..…..… 114
Table 4.13: Simulation total welfare gains to the districts…………….…....… 115
Table 5.1: Definition of the agricultural biodiversity…………………..…...… 135
Table 5.2: Definition of potential explanatory variables ……………………... 136
Table 5.3: Explanatory variables used in the demand model…………………. 142
Table 5.4: Summary of the econometric models to be used for the analysis….. 143
Table 5.5: Poisson regression results for crops variety model…………….….. 161
Table 5.6: Poisson regression results for animal variety model………….…… 167
Table 6.1: Definition dependent variables in different models…………….….. 183
Table 6.2: Definition of potential explanatory variables ……………………... 184
Table 6.3: Explanatory variables and their expected signs……………………. 186
Table 6.4: Probit regression results for landrace production model………..…. 189
Table 6.5: Probit regression results for organic production model……….…… 191
Table 6.6: Probit regression results for agro-diversity model…………………. 195
Table 7.1: ML estimates for parameters of the production function……….…. 225
xii
Table 7.2: Estimated elasticities and marginal productivity of each input……. 227
Table 7.3: Frequency and percentage distribution of the technical efficiencies. 229
Table 7.4: Average TE, value of actual and potential output with land size….. 231
Table 7.5: Average efficiency with farm type……………………………...…. 232
Table 7.6: ML estimates for parameters of the inefficiency model……….…... 234
LIST OF FIGURES
Figures Page
Figure 1.1: Summary of the three main sections of the thesis…………….. 10
Figure 7.1: Stochastic frontier production function……………………….. 211
xiii
LIST OF ABBRIVIATION
ASC Alternative Specific Constant
BCAP Biodiversity Conservation Action Plan
CBD Conservation on Biological Diversity
CS Compensating Surplus
CEM Choice Experiment Method
CVM Contingent Valuation Method
CL Conditional Logit
DSDs Divisional Secretary Divisions
DFC Department of Forest Conservation
EEZ Exclusive Economic Zone
EEPU Environmental Economic Policy Unit
EU European Union
FAO Food and Agriculture Organization
GDP Gross Domestic Production
GLR Generalised Likelihood Ratio
GM Genetically Modified
HYV High Yield Varieties
IBEC Biodiversity and Environmental Conservation
IIA Independence of Irrelevant Alternatives
IID Independently and Identically Distributed
IFPRI International Food Policy Research Institute
IUCN International Union for Conservation of Nature
LKR Sri Lankan Rupees
MLE Maximum Likelihood Estimator
MNL Multinomial Logit
NB Negative Binomial
NBM Negative Binomial Model
NCS National Conservation Strategy
NEAP National Environmental Action Plan
NGOs Non Government Organizations
PGRC Plant Genetic Resource Centre
xiv
PM Poisson Model
RPL Random Parameter Logit
RUM Random Utility Models
TE Technical Efficiency
TEV Total Economic Values
TWTP Total Willingness to Pay
TWTA Total Willingness to Accept
UK United Kingdom
USD US Dollars
VC Variance-covariance
WTA Willingness to Accept
WTP Willingness to Pay
ZIP Zero-inflated Poisson
ZINB Zero-inflated Negative Binomial
xv
ABSTRACT
The issues involved in agricultural biodiversity are important and interesting areas for the
application of economic theory. However, very little theoretical and empirical work has been
undertaken to understand the benefits of conserving agricultural biodiversity. Accordingly,
the main objectives of this PhD thesis are to: (1) Investigate farmers’ valuation of
agricultural biodiversity; (2) Identify factors influencing farmers’ demand for agricultural
biodiversity; (3) Examine farmers’ demand for biodiversity rich farming systems; (4)
Investigate the relationship between agricultural biodiversity and farm level technical
efficiency. This PhD thesis investigates these issues by using primary data in small-scale
farms, along with secondary data from Sri Lanka. The overall findings of the thesis can be
summarized as follows.
Firstly, owing to educational and poverty issues of those being interviewed, some policy
makers in developed countries question whether non-market valuation techniques such as
Choice Experiment (CE) can be applied to developing countries such as Sri Lanka. The CE
study in this thesis indicates that carefully designed and pre-tested nonmarket valuation
techniques can be applied in developing countries with a high level of reliability. The CE
findings support the priori assumption that small-scale farms and their multiple attributes
contribute positively and significantly to the utility of farm families in Sri Lanka. Farmers
have strong positive attitudes towards increasing agricultural biodiversity in rural areas. This
suggests that these attitudes can be the basis on which appropriate policies can be introduced
to improve agricultural biodiversity.
Secondly, the thesis identifies the factors which influence farmers’ demand for agricultural
biodiversity and farmers’ demands on biodiversity rich farming systems. As such they
provide important tools for the implementation of policies designed to avoid the loss
agricultural biodiversity which is shown to be a major impediment to agricultural growth and
sustainable development in a number of developing countries. The results illustrate that
certain key household, market and other characteristics (such as agricultural subsidies,
percentage of investment of owned money and farm size) are the major determinants of
demand for agricultural biodiversity on small-scale farms. The significant household
characteristics that determine crop and livestock diversity include household member
participation on the farm, off-farm income, shared labour, market price fluctuations and
household wealth. Furthermore, it is shown that all the included market characteristics as
well as agricultural subsidies are also important determinants of agricultural biodiversity.
Thirdly, it is found that when the efficiency of agricultural production is measured in
practice, the role of agricultural biodiversity has rarely been investigated in the literature.
The results in the final section of the thesis show that crop diversity, livestock diversity and
mix farming system are positively related to farm level technical efficiency. In addition to
these variables education level, number of separate plots, agricultural extension service,
credit access, membership of farm organization and land ownerships are significant and
direct policy relevant variables in the inefficiency model. The results of the study therefore
have important policy implications for conserving agricultural biodiversity in Sri Lanka.
1
CHAPTER ONE
INTRODUCTION
1.1 Overview
Biological diversity provides all of mankind’s food requirements, numerous medicines
and industrial products. Agricultural biodiversity1 (see Appendix A.1 for more details) is
a sub-set of general biodiversity which is essential for global food production, livelihood
security and sustainable agricultural development (Brookfield, 2001; Pascual and
Perrings, 2007). Agricultural biodiversity includes all forms of life directly relevant to
agricultural production. In addition to providing direct benefits to farmers, agricultural
biodiversity improves ecological processes by regulating climate, maintaining soil
quality, providing protection from erosion, storing nutrients and breaking down
pollution (Thrupp, 1988; FAO, 1999). Some societies also value biodiversity for cultural
reasons as it maintains the aesthetic value of landscapes (Nagarajan et al., 2007).
Despite all these benefits previous experience has shown that population growth,
inequity, inadequate economic policies and institutional systems have mainly
contributed towards the increasing loss of agricultural biodiversity in the world (Ayyad,
2003; Ganesh and Bauer, 2006). Low levels of education and lack of integrated research
on natural ecosystems and their innumerable components may exaggerate the process,
1 FAO, (1999a) defined agricultural biodiversity as the variety and variability of animals, plants and
micro-organisms that are used directly or indirectly for food and agriculture, including crops, livestock,
forestry and fisheries. It comprises of the diversity of genetic resources (varieties, breeds) used for food,
fodder, fibre, fuel and pharmaceuticals. It also includes the diversity of non-harvested varieties that
support production (soil micro-organisms, predators, pollinators), and those in the wider environment that
support agro-ecosystems (agricultural, pastoral, forest and aquatic) as well as the diversity of the agro-
ecosystems.
2
especially in developing countries. While the loss of habitats may occur through clearing
land for agriculture, specialisation of agricultural practices reduces farm level crops,
genetic or livestock diversity (Swanson, 1999).
Neoclassical economic theory predicts that specialisation in one kind of variety or
technology is the profit maximising solution for a farmer and that it is costly to maintain
a diverse portfolio of species, varieties and management systems due to several reasons.
These reasons include time and management intensity of diversity maintenance and high
opportunity costs associated with not specialising in particular varieties with the highest
current economic return (Brush et al., 1992; Smale et al., 2001; Gauchan and Smale,
2003). But in reality, it has been observed that contrary to economic theory, farmers,
especially in developing countries often prefer to maintain a diverse portfolio of
varieties and to continue employing traditional agricultural technologies, even when
modern technologies and high yielding varieties (HYVs) are available to them. Several
explanations have been found for this persistence in management of agricultural
biodiversity on farms. These include farmers’ attitudes towards risk (in yield, income,
price and consumption) and their need to compensate for market imperfections in
satisfying household demands for diversity in consumption.
Many farmers manage high levels of agricultural biodiversity on farms to keep options
open for possible future benefits of diversity, such as being sources of new varieties.
Many farm families use agricultural biodiversity as a way of spreading out labour needs
to ensure that limited labour supplies are used more efficiently. There are also cultural
benefits (e.g. cuisine, ritual, prestige, payment, gift, social ties) attached to agricultural
3
biodiversity. Equally, agricultural biodiversity is found to have positive impacts on
overall productivity and soil quality. In recognition of agricultural biodiversity
importance, international agreements such as the Convention of Biological Diversity
(CBD) and the international institutes such as International Food Policy Research
Institute (IFPRI) and Institute of Biodiversity and Environmental Conservation (IBEC)
encourage the design of policies that convey economic incentives for farmers to
conserve agricultural biodiversity (CBD, 2002). The number of economic studies that
have attempted to explain the reasons for on farm conservation and the means by which
this method of conservation can be strengthened, are however small compared to the
magnitude of the problem of loss of agricultural biodiversity in farmers’ fields
throughout the world.
Modern agricultural methods and technologies have brought spectacular increases in
food production (Tilman et al., 2002), but not without high environmental costs. Efforts
to boost food production, for example, through direct expansion of cropland and
pastures, have negatively affected the capacity of ecosystems to support food production
and to provide other essential services. Food production will undoubtedly be affected by
external factors such as climate change. But the production and distribution of food itself
is also a major cause of climate change. As food production becomes increasingly
industrialised, with fewer niches available for varieties other than those targeted for
production, a rapid decline in the diversity of varieties used has been observed. These
major changes in production have lead to simplified and less resilient agro-ecosystems,
reducing not only the number of niches but also the range of products and their
distribution over time and space (FAO, 1999b). There is ample research which indicates
4
that modern agricultural methods and technologies can generate large environmental and
social costs. A substantial contribution to sustaining agricultural biodiversity can
therefore be made through continued support of producer organisations working with
small-scale farm producer groups to conserve, develop and use sustainably food and
agricultural genetic resources including plant, animal and aquatic.
As mentioned above, agricultural biodiversity is eroding and resources available for
conservation are limited, implying economic valuation (especially estimation of total
economic value) can play an important role in ensuring an appropriate focus for
conservation efforts (UNEP, 1995; Drucker et al., 2001). As Swanson et al. (1997) state,
in order to design policies and programmes that both encourage maintenance of
agricultural biodiversity on farm and ensure that economic and agricultural development
occur, it is necessary to establish the value of what it is that needs to be conserved.
The direct and indirect benefits of conserving farm level biodiversity can be numerous,
especially in semi-subsistence economies. The measurement of economic values of
services provided by agricultural biodiversity can be done on the basis of total economic
values (TEV). TEV consists of use and non-use values. Diagrammatically, the TEV
framework can be expressed as shown in Appendices A.1 and A.2. Benefits obtained by
individuals using agricultural biodiversity are defined as use values. Use values of
agricultural biodiversity include, direct, indirect, portfolio values and option values2
(Brown, 1990; Primack, 1993; Swanson, 1996; Evenson et al., 1998). On the other hand,
bequest values, altruistic values, existence values and cultural values of agricultural
2Option values can be placed under both use and non-use values. It includes future direct and indirect use
values.
5
biodiversity are considered under non-use values (Krutilla, 1967; Brown, 1990; Primack,
1993; Evenson et al., 1998).
In this study, five indicators (components) are used to capture the use and non-use
values of agricultural biodiversity. They are: crop diversity (number of crops varieties
that are grown on the farm), livestock diversity (number of livestock varieties on the
farm), mixed farming systems (integration of crop varieties and livestock breeds),
landrace cultivation (whether a farm contains crop varieties that have been passed down
from the previous generation and/or has not been purchased from a commercial seed
supplier) and organic production (when industrially produced and marketed chemical
inputs are not used in farm production). Among these five indicators, the first two
represent agricultural biodiversity while last three represent the different farming
systems which help maintain biodiversity under rich farming practices. More details
about using these variables to capture farmers’ valuation of agricultural biodiversity are
found in studies conducted byBenin et al. (2003), Benin et al. (2004), Bellon (2004),
Birol et al. (2006), Nagarajan et al. (2007), Birol et al. (2008) and Hadgu et al. (2009). It
is evident that economic values of conserving these components can only be calculated
based on a comprehensive identification of the environmental and social values of the
ecosystem services that they provide.
Commercial direct use value of agricultural biodiversity can be a relatively small
component of their total use value in agriculture (Drucker et al., 2005). Many values are
not captured well in market prices and hence investments in conservation may not occur
optimally (Swanson, 1996). This is one of the reasons why farmers’ activities gradually
6
reduce agricultural biodiversity. Some of the other possible reasons why farmers may
tend to destroy agricultural biodiversity can be explained as follows. Firstly, most
benefits of conserving agricultural biodiversity are long-term (and inter-generational)
and not traded in the market. For example, by cultivating different crops and livestock,
soil fertility can be improved. However, farmers may not take into account these long-
term benefits. Secondly, poor farmers with lower levels of education may not be aware
about the total benefits of conserving agricultural biodiversity. They may consider only
the short-term direct use benefits and may select the specialisation of cash crops as a
mean of increasing income in the short term. However, single crops are more vulnerable
to the rapid spread of disease, this greatly heightens the vulnerability of resource-poor
farmers. Thirdly, sales promotion activities and credit facilities have promoted the
cultivation of modern crop varieties using pesticides and chemical fertilisers. Such a
system can increase short-term yields while destroying the resilience of agro-ecosystems
in the long-term. Fourthly, high discount rates will decrease the future value of
agricultural biodiversity and provide some incentives to increase present consumption
which in turn can increase the degradation of biodiversity. These reasons show that as
long as farmers underestimate the total benefits of conserving agricultural biodiversity,
there will be simplified and less resilient agro-ecosystems, thus reducing the number of
services provided by them in the long-run.
Although much theoretical as well as empirical work has investigated various aspects of
agricultural biodiversity there is still a considerable lack of understanding of what social
benefits could be achieved from conserving agricultural biodiversity in developing
countries. Economics to some extent provides us with the analytical tools to assist in
7
guiding towards socially desirable outcomes. However, little theoretical and empirical
work has been undertaken in this area of research. This means that there exists a gap in
the theoretical and empirical literature, addressing practical issues utilising correct
economic instruments in this area. This thesis examines three main issues that arise in
the area of agricultural biodiversity in the context of Sri Lanka. The focus of the thesis
allows for the study of direct and tangible issues facing policy makers. After reviewing a
large number of studies, existing models and empirical work, the shortcomings that exist
in their application are identified. They are:
(1) Farmers’ valuation of agricultural biodiversity is not properly explained. As a result
social welfare losses due to loss of agricultural biodiversity have not been adequately
estimated. It is evident that management of agricultural biodiversity requires
measurement, and measures of diversity to some extent. It is thus necessary to measure
and disentangle some of the separate benefits of agricultural biodiversity in order to
formulate appropriate policies. However, many of the goods and services provided by
different components of agricultural biodiversity are crucial, but not always quantifiable
in monetary terms. Many of these goods and services are not traded in the market place
and do not have an obvious price or commercial value. The danger is that if these
unpriced values are not included in the decision-making process, the final decision may
favour outcomes which do have a commercial value and decision makers may not have
full awareness of the consequences for biodiversity conservation. Therefore, it is of
paramount importance to understand the true value of agricultural biodiversity and to
estimate the welfare change of the society with the change of agricultural biodiversity.
The first section of this thesis attempts to capture farmers’ valuation of agricultural
8
biodiversity. This objective will help to determine the economic value of conserving
agricultural biodiversity to society.
(2) Factors affecting the conservation of agricultural biodiversity are not adequately
identified in the literature. The literature shows that, despite the emphasis placed by
policy decision-makers on increasing the conservation of biodiversity in small scale-
farms3, it is increasingly becoming degraded in many agricultural areas (see, for
example, Matson et al., 1997; Perrings, 2001; Brookfield et al., 2002; Mattison and
Norris, 2005). Therefore, it is important to understand which factors are contributing to
decreasing agricultural biodiversity in small-scale farms. In the second section of this
thesis farmers’ demand for agricultural biodiversity and environmentally rich farming
systems such as organic farming and landrace cultivation are estimated. This objective
will help understand and identify factors influencing the degradation of agricultural
biodiversity in small-scale farms.
(3) No previous analysis has investigated the links between agricultural biodiversity and
farm level technical efficiency. Some studies reveal that crop diversity is positively
related to agricultural productivity of small-scale farms (see, for example, Di Falco and
Perrings, 2003). They also find that inter-species’ crop genetic diversity is positively
related to mean income and negatively related to the variance of income. While
increasing productivity on farms, diverse farming systems help farmers manage some
3 A small-scale farm is defined as any farm which is less than one hectare. We only concentrate on small-
scale farms in this study. This is due to three reasons. First, small-scale farms are the most common type
of farms in rural areas in Sri Lanka. Second, maintaining diverse farming systems with the objective of
acquiring family food consumption is a common characteristic of small-scale farms rather than large-scale
farms. Third, some indicators of agricultural biodiversity that we considered in this study such as animal
diversity, landrace cultivation and organic production can commonly be seen in small-scale farms in the
country.
9
resources, such as labour, optimally. It also helps to increase farm revenues by
minimising market risks which is a common problem in developing countries. For
example, in a particular season prices of some crops or livestock can decrease while
others can increase. Therefore, maintaining more diverse farming systems help farmers
manage unnecessary risks in the markets. In the third section of this thesis we investigate
the relationship between agricultural biodiversity and farm level efficiency.This type of
study allows us to analyse the effects of agricultural biodiversity on farm level technical
efficiency.
The overall objective of this thesis is to address some of the issues related to the above
mentioned three sections in the context of Sri Lanka’s agriculture. Accordingly, the
thesis has three separate sections. The structure of the three main sections and
subsequent studies are summarised in Figure 1.1. The first section of the thesis analyses
farmers’ valuation of agricultural biodiversity. The choice experiment (CE) method
which is one of the most widely used and a preferred technique is used for this purpose.
The results are then used to estimate the likely welfare gains under various hypothetical
scenarios. The results of the study will enable policy decision-makers to better
understand the relevant issues and thereby take appropriate action to mitigate some of
the adverse issues in this field.
The second section of the thesis examines the demand for agricultural biodiversity in
small-scale farms in Sri Lanka. This section consists of two studies. The first study
analyses farmers’ demand for crops and livestock varieties respectively while the second
10
study examines farmers’ demand for landrace cultivation, mixed farming and organic
farming systems. This section attempts to identify the different market and non-market
Figure 1.1: Summary of the three main sections of the thesis
factors which are important for increasing agricultural biodiversity on small-scale farms.
An agricultural farm household model is used for this purpose. The motivations of the
second section of this thesis are threefold. Firstly, this study investigates whether
farmers within a semi-subsistence economy allocate farm resources (e.g. land or
household time endowment) to the production of food crops and thus have higher levels
Agricultural Biodiversity
Demand Estimation
Technical Efficiency Conservation Benefits
Agricultural
Household Model
Agricultural
biodiversity
Primary data:
Three districts
Choice Experiment
Approach
Welfare change
estimation
Primary data:
Three districts
Stochastic Production
Frontier Approach
Efficiency gains
with AB
Primary data:
Three districts
Different
farming systems
11
of agricultural biodiversity, or to cash crops4, and have a subsequent loss of agricultural
biodiversity. Secondly, the empirical research that has investigated farmers’ preferences
of environmentally friendly farming systems is limited in the literature (Van Dusen,
2000; Smale et al., 2001). Therefore, it is important to identify different factors which
support an increase in landrace cultivation, mixed farming systems and organic farming
systems. Thirdly, a common finding of previous studies in this area shows that market
development is one of the causes of agricultural biodiversity loss on farms in most
developing countries (Smale et al., 2001). This study attempts to investigate this finding
using semi-subsistence farm level data in Sri Lanka.
The third section of the thesis investigates the relationship between agricultural
biodiversity and farm level technical efficiency. The stochastic production frontier
approach is used to estimate farm level technical efficiency. There is increasing evidence
(Adams et al., 2004; Agrawal and Redford, 2006) to show that agricultural biodiversity
conservation can in turn facilitate increasing productivity and farm level efficiency in
small-scale farming. However, the existing scientific knowledge regarding agricultural
biodiversity and its link with farm level technical efficiency has not been fully
examined. The existing literature does not assess the value of ecosystem services to the
poor and the implications of these links for development policy. As a result, the need for
proper estimation of costs and benefits of conserving agricultural biodiversity, as well as
the demand for introducing appropriate policy regimes for managing them is increasing
(Romstad et al., 2000).
4Cash crops are those which are produced for the purpose of generating cash or money. The products are
therefore intended to be marketed for profit. A specialized farming system is the mostly preferred farming
system for the cultivation of cash crops.
12
Among these sections, the first and the second sections deal with the important aspects
of conserving agricultural biodiversity while the third section investigates the
relationship between agricultural biodiversity and farm level technical efficiency. The
implications of these findings will help illustrate the importance of conserving natural
resources in agriculture. This study will help implement policies to reduce degradation
of biodiversity that can be hypothesized to be increasingly posing a major impediment to
agricultural growth and sustainable development in many developing countries.
Therefore, the findings of the study will provide useful policy implications.
Sri Lanka is an ideal representative country for this type of study. This is because the
country, being largely agricultural, historically has had phases of agricultural policy
development on the basis that development of agriculture would lead to the overall
development of the nation and would thus help to eradicate poverty. It has been later
realized that the increasing efforts to raise agricultural growth has cost the country in
terms of land as well as biodiversity degradation (Anon, 1999). Sri Lankan agriculture
today has a dual structure consisting of large-scale, mechanised farms alongside semi-
subsistence, small-scale farms managed with family labour and traditional practices.
These Sri Lankan small-scale farms5have a range of local varieties of trees, crops and
livestock breeds, as well as soil micro-organisms.
Agricultural scientists describe small-scale farms as micro-agro ecosystems that are rich
in several components of agricultural biodiversity. Many expect that as a result of
5Small-scale farms are semi-subsistence in nature and are the most common type of farms in rural area in
Sri Lanka. These farms are privately owned, labour intensive and has a traditional production system that
maintains a high level of agricultural biodiversity in Sri Lanka.
13
continued economic transition, the dual structure of Sri Lankan agriculture and the share
of home-produced food will eventually disappear. So the private provision of public
goods generated by small-scale farms management cannot be sustained in the long
run.In addition to that, the disappearance of the rural-based multi-crops farming system
has affected rural communities in Sri Lanka in many ways (Anon, 1999). Therefore, it is
necessary to implements agri-environmental schemes to advance the use of specified
farming methods in rural areas, but so far the role of small-scale farms within these
schemes has not been elucidated. This study identifies the least-cost options for
including farming communities in Sri Lankan’s agri-environmental schemes, by
characterising those who value agricultural biodiversity in their small-scale farms most.
The motivations for undertaking this type of study are explained in the following
section.
1.2 Motivation
The overall aim of this study is to estimate the conservation benefits of agricultural
biodiversity in small-scale farms with special reference to Sri Lanka. The results of the
study can be used to develop/implement economically profitable and environmentally
feasible agro-ecosystems in any country. It is clear that understanding these issues is
crucial when formulating policies to upgrade livelihood of rural households and
enhancing agricultural biodiversity. A study of this nature also helps to develop a
sustainable agricultural system that minimises the social cost of using natural resources.
Lack of sufficient incentives for managing farm level agricultural biodiversity could be
one of the constraints in conserving biodiversity in most developing countries. The first
14
aim of this thesis is to review the current state of knowledge associated with agricultural
biodiversity and to identify gaps in our knowledge base in this area. Secondly,
appropriate economic methodologies are applied to analyse the three main research
questions that have been highlighted in the introduction. The overall objective of the
thesis is to establish a case for increasing the sustainable use of agricultural biodiversity
in improving people’s well-being and food and nutrition security.
Agricultural biodiversity provides a wide range of direct and indirect benefits to the
farming community (see, Appendices A.2 and A.3 for more details). However, many
human activities contribute to unprecedented rates of biodiversity loss, which threaten
the stability and continuity of ecosystems as well as their provision of goods and
services. In this context, several studies have been conducted to identify the possible
monetary values based on farmers’ preference of agricultural biodiversity. However,
most studies do not use a uniform, clear measurement framework that enables the
exploration of the use of both market and non-market benefits. Moreover, existing
studies only analyse welfare changes without considering crop heterogeneity and
regional heterogeneity simultaneously. The first section of this research attempts to
identify farmers’ valuation of agricultural biodiversity using the choice experiment
technique. This methodology helped estimate the welfare changes to society due to
changes in agricultural biodiversity. Heterogeneous farms from three districts in Sri
Lanka were selected. The implications of these findings will help illustrate the benefits
of conserving diverse farming systems in small-scale agriculture in developing
countries.
15
Although the contribution of small-scale farms to household survival in developing
countries is very important, only a few studies are available in this field. Moreover,
existing studies have taken into account the market value of only crop or livestock
diversity. In this study, the demand for crop diversity, livestock diversity, mixed farming
systems, landrace cultivation and organic production are estimated using farm household
survey data. This will help identify factors influencing the degradation of agricultural
biodiversity. The results will provide information for policy makers to implement a
farming system that provides maximum benefits to themselves and society.
The loss of biodiversity may impair ecosystem functions while decreasing farm level
productivity. A number of experimental studies have been performed or are emerging in
this area (see, for example, Johnson et al., 1996). However, most of these studies are
restricted to experimental work in the field of science rather than economic analysis. In
this context, the third section of this research attempts to investigate the relationship
between agricultural biodiversity and efficiency. To the best of my knowledge, no
economics study has attempted to examine this relationship before. The results will be a
novel contribution to the existing literature. In this study, it is expected to calculate farm
level technical efficiency and investigate its links with important variables that are
directly linked with agricultural biodiversity. Stochastic production frontier method is
used for this purpose. The results will show a way of increasing farm level technical
efficiency which is a major challenge in developing countries, including Sri Lanka.
Rural people use and manage agricultural biodiversity in order to improve their
livelihoods. However, there is an increasing interest in the opportunities that
16
conservation in a broader production landscape could afford as a means to overcome
poverty. Much has been written on the loss of managed biodiversity under threats from
commercial and intensified agricultural production. However only a limited amount of
work has been conducted on how farm households manage their resources so as to
sustain and enhance them. The overall findings of this study will help conserve the
agricultural biodiversity in small-scale farms which can in turn help design poverty
alleviation policies, especially in developing countries. In the next section the expected
contribution of this thesis will be explained.
1.3 Expected contributions of the study
The strategic roles of agricultural biodiversity in food and nutritional security and
income generation have been insufficiently documented and understood. Systematic
identification and investigation of such roles are needed to build on scattered research so
far. This in turn requires the development, testing and diffusion of tools, methodologies
and strategies that strengthen the mutually reinforcing contribution of biodiversity to
livelihoods and livelihoods to biodiversity conservation. Most of the world’s agricultural
biodiversity is found in small-scale agricultural areas in developing countries (Smale et
al., 2001). Hence, an essential element of the research is to strengthen the benefits from
agricultural biodiversity realised by communities in these areas. The key hypothesis is
that biodiversity, given certain interventions and support, can be used to improve
nutrition and livelihood options, and in so doing creates incentives for the conservation
of its diversity in order to achieve a sustainable farming system.
17
This research contributes to developing a biodiversity rich agricultural system across
different ecological and socio-economic contexts. It also evaluates the effects of
different farming systems (with different biodiversity levels) on farmers’ wellbeing. In
the first section a stated preference method Choice modelling) is used to investigate
farmers’ preferences for biodiversity rich agricultural systems. The second section of the
research attempts to identify the influencing factors for conserving agricultural
biodiversity. The third section contributes to the existing literature showing the
agricultural biodiversity and its link with productivity and farm level efficiency which is
a missing part of the economics literature. This will directly help make suitable policies
to implement most appropriate agricultural systems for small-scale farms. In general,
this study will show the importance of the conservation and sustainable use of
agricultural biodiversity on farms.
The overall policy goal of the study is to increase awareness and generate support for
investment in conservation and development of agricultural biodiversity. The research
aims at sharing innovative ideas, research methods and findings in areas of agricultural
biodiversity conservation to the existing literature. It will provide an opportunity to
make necessary policies that provide incentives to protect biodiversity at farm level that
generate regional as well as global benefits in the future. This research will also identify
the weaknesses and the gaps that exist in this field. It will help develop mechanisms,
approaches and pathways for strengthening engagement on agricultural biodiversity for
food and nutrition security and environment in the future. This will include establishing
a platform for actions for supporting and strengthening research, development and
policies in agricultural biodiversity.
18
The overall findings of this PhD research will help implement policies to reduce
degradation of biodiversity that is increasingly posing a major impediment to
agricultural growth and sustainable development. The research findings could also be
used to develop/implement economically profitable and environmentally feasible agri-
ecosystems in any country. Understanding the benefits of conserving biodiversity and its
variations are of paramount importance when designing policies in this field. In general,
the research contributes to the sustainable use of agricultural biodiversity to improve
farmers’ well-being. For this purpose, we first attempt to explain some previous
economic models and identify the shortcomings of previous studies. Then we apply
appropriate economic models to analyse relevant issues mentioned above, which is an
extension of the conventional work in this field. In this context the results of the study
help policy makers understand the real issues and come up with appropriate solutions.
The way of carrying out this task is explained in the next section.
1.4 Structure of the thesis
This PhD research addresses issues related to agricultural biodiversity in small-scale
farms which are extremely important in the context of conserving agricultural
biodiversity as well as improving the livelihood of farmers in the agricultural sector in
Sri Lanka. The task of analysing these issues is accomplished in the following manner.
The thesis is presented in eight chapters. This first chapter defines the research problem.
Chapter two provides background information on the present status of biodiversity in Sri
Lanka, which includes: biodiversity wilderness area in the world; biodiversity in Sri
Lanka; present status and trends of biodiversity and future challenges in agricultural
19
biodiversity. In Chapter three the conduct of the survey, the data collection method and
data sources are explained.
Chapter four investigates farmers’ preferences for different attributes of agricultural
biodiversity. It also analyses the welfare changes to society due to changes in
agricultural biodiversity. The fifth chapter estimates demand for agricultural
biodiversity. It attempts to identify the determinants of crop diversity and livestock
diversity. Chapter six focuses on the farmers’ preference for agricultural biodiversity
rich farming systems. This chapter investigates the important factors for selecting mixed
farming systems, landrace cultivation and organic production systems. Chapter seven
focuses on investigating the relationship between different variables that represent
agricultural biodiversity and farm level technical efficiency.
The final Chapter provides a brief summary of the thesis with a discussion of the results
within a policy framework. Particular attention is paid to highlighting the key findings
and policy constraints. This Chapter attempts to clearly define where the information
gathered from this thesis fits within the larger social, political and economic discussions
on agricultural biodiversity loss, economic growth and policy failure. It presents some
concluding remarks, while highlighting obvious gaps in the literature. The importance of
the analysis undertaken in this study, along with the limitations and remaining future
research areas, are also highlighted in the final Chapter.
20
CHAPTER TWO
STATUS AND TRENDS OF BIODIVERSITY IN SRI LANKA
2.1 Biodiversity wilderness area: a global prospective
Biodiversity for food and agriculture includes the components of biological diversity
that are essential for feeding human populations and improving the quality of life
(Adams et al., 2004). It includes the variety and variability of ecosystems, animals,
plants and micro-organisms at the genetic, species and ecosystem levels, which are
necessary to sustain human life as well as the key functions of ecosystems. Biodiversity
is usually explored at three levels; genetic diversity, species diversity and ecosystem
diversity (Brock and Xepapadeas, 2003). Genetic diversity is the variety of genes within
a species. Each species is made up of individuals that have their own particular genetic
composition. This means a species may have different populations, each having different
genetic compositions. To conserve genetic diversity, different populations of a species
must be conserved. Species diversity is the variety of species within a habitat or a
region. Species are grouped together into families according to shared characteristics.
The number of globally identified species under each category is given in Appendix B.
Ecosystem diversity is the variety of ecosystems in a given place. An ecosystem is a
community of organisms and their physical environment interacting together
(Brookfield, 2001; Brock and Xepapadeas, 2003). An ecosystem can cover a large area,
such as a whole forest, or a small area, such as an agricultural farm. It is a community of
organisms and their physical environment interacting together.
21
Biodiversity is crucial to the maintenance of many ecosystem services such as regulation
of chemical composition of the atmosphere, food production, supply of raw materials,
water provision, nutrients’ recycling, biological control of populations of flora and
fauna, use of genetic resources and leisure activities (Brookfield and Stocking, 1999;
Brookfield, 2001). Biodiversity continues to decrease at unprecedented rates as human
development and expansion result in the fragmentation and loss of habitat for flora and
fauna (Di Falco and Chavas, 2009). The loss of biodiversity is expected to continue at an
unchanged increasing pace in the coming decades as well (Drucker et al., 2005). Key
underlying drivers for the loss of biodiversity such as global population and economic
activity are expected to keep on growing. Between 2000 and 2050, the global population
is projected to grow by 50 per cent and the global economy to quadruple (Slingenberg et
al., 2009). The need for food, fodder, energy and wood will unavoidably lead to a
decrease in and unsustainable use of natural resources.
Biodiversity is the basis of agriculture (see, Appendix A.1). As mentioned in the
introduction, biodiversity is the origin of all species of crops and domesticated livestock
and the variety within them. It is also the foundation of ecosystem services essential to
sustain agriculture and human well-being (Diwakar and Johnsen, 2009). Biodiversity
and agriculture are strongly interrelated because while biodiversity is critical for
agriculture, agriculture can also contribute to conservation and sustainable use of
biodiversity (Brookfield, 2001). Indeed, sustainable agriculture both promotes and is
enhanced by biodiversity. Maintenance of this biodiversity is essential for the
sustainable production of food and other agricultural products and the benefits these
provide to humanity, including food security, nutrition and livelihoods. As highlighted
22
by Slingenberg et al. (2009) during the last decades, worldwide biodiversity has been
lost at an unprecedented rate in all the ecosystems, including agro-ecosystems.
According to the FAO (1999), it is estimated that about three-quarters of the genetic
diversity found in agricultural crops and livestock has been lost over the last century,
and this genetic erosion will further continues in the future. Therefore, understanding the
important causes of agricultural biodiversity loss is important for conserving
biodiversity in the world. A map showing the biodiversity wilderness area in the world is
given in Appendix C. As can be seen, Sri Lanka is identified as a biodiversity wilderness
area. In this context, the next section provides a brief overview about the biodiversity in
Sri Lanka.
2.2 Biodiversity in Sri Lanka
Sri Lanka is an island with a total land area of 6,570,134 hectares, a coastline of 1,600
km and an Exclusive Economic Zone (EEZ) that extends up to 320 km beyond the
coastline (Department of Census and Statistics in Sri Lanka, 2010). Total cultivated land
and forest cover comprise 39 per cent and 24 per cent, respectively. The country is one
of the smallest, but biologically diverse countries in Asia (Sanjeeva, 2003).
Consequently it is recognized as a biodiversity hotspot of global and national
importance. It’s varied climate and topography conditions have given rise to rich species
diversity, believed to be the highest in Asia in terms of unit land area (Kotagama, 2002).
Many of the species are endemic, a reflection of the island's separation from the Indian
subcontinent. This is especially relevant for mammals, amphibians, reptiles and
flowering plants. These species’ are distributed in a wide range of ecosystems which can
23
be broadly categorized into forest, grassland, aquatic, coastal, marine and cultivated
(Ministry of Environment and Natural Resource in Sri Lanka, 2007). The diversity of
ecosystems in the country has therefore resulted in a host of habitats, which contain high
genetic diversity.
In the broader context, biodiversity in Sri Lanka includes species diversity, genetic
diversity and ecosystem diversity (Ministry of Environment and Natural Resources in
Sri Lanka, 2007). An interesting feature of this species diversity is its high degree of
endemism, which is observed in several taxonomic groups. A large proportion of these
endemic species is found in the wet zone in the south western region of the island.
Genetic diversity is another component of biodiversity that is important but not well
investigated (Bellon, 2004). Almost all of the available information is confined only to
economically important agricultural crops. The Plant Genetic Resource Centre (PGRC)
at Gannuoruwa, Peradeniya, Sri Lanka has collected and preserved propagative material
of a large number of species from various agro-climatic zones of the country. For
example, the PGRC has germoplasm materials of 3,194 traditional varieties and
cultivars, and 17 wild relatives of rice (Ministry of Environment and Natural Resources
in Sri Lanka, 2007).
There is a wide range of ecosystem diversity under different climatic conditions in the
island. The topography in Sri Lanka and the major climatic zones are shown on the maps
in Appendix D.1 and D.2. The major natural ecosystems are forests, grasslands, inland
wetlands, and coastal and marine ecosystems (Kotagama, 2002). There are also
agricultural ecosystems. Forests vary from wet evergreen forests (both lowland and
24
mountain), dry mixed evergreen forests to dry thorn forests. Grasslands are found in
mountains and low country wetlands include a complex network of rivers and freshwater
bodies. Marine ecosystems include sea-grass beds, coral reefs, estuaries and lagoons and
mangrove swamps. Contemporary issues in relation to the diversity of valuable
ecosystems are: deforestation, soil erosion, threatened wildlife populations (as a result of
both poaching and urbanisation), coastal degradation from mining activities and
increased pollution. Most of these issues can be controlled by using appropriate policies.
The Environmental Economic Policy Unit (EEPU) in Sri Lanka is responsible for the
formulation and deployment of policy conserving and protecting Sri Lanka’s native
natural capital. Although the EEPU is attempting to address these issues, the short term
development goals that encourage economic growth over unsustainable resource use
have generated a number of issues. There are numerous policies, laws, action plans and
institutions involved in the conservation of Sri Lanka’s biodiversity. Although most of
the laws relate directly or indirectly to biodiversity conservation, implementation has
been sluggish (Sanjeeva, 2003). Therefore, adopting suitable policies focusing on rural
communities, encompassing both economic development and ecological conservation
efforts would aid Sri Lanka in retaining long-term value in its natural capital.
There are many legislative enactments that deal with the protection of biological
resources in the country. In 1980, The National Environmental Act Constituted the
Central Environmental authority and established a National Conservation Strategy
(NCS) to protect biodiversity in the country. In 1988, the NCS was adopted to deal with
environmental degradation (Ministry of Environment and Natural Resource in Sri
Lanka, 2007). In 1991, the National Environmental Action Plan (NEAP) was adopted
25
for a four year period. Based on the outcomes of its implementation, it was revised in
1994, for the period 1995-98. Over the years these environmental policy frameworks
have influenced and helped shape several sectoral and national development plans. The
National Conservation Strategy, the National Environmental Action Plan, the Forestry
Sector Master Plan, the National Coastal Zone Management Plan, and Coastal 2000, are
some of the policy documents that have addressed biodiversity conservation in the
country (Ministry of Environment and Natural Resource in Sri Lanka, 2007).
The Sri Lanka Biodiversity Conservation Action Plan (BCAP) was adopted in 1998. The
BCAP has identified four broad areas of ecosystem diversity, namely forests, wetlands,
coastal and marine systems, and agricultural systems. Under each ecosystem, the main
issues have been identified and the recommended actions and the implementing
institutions defined. At the regional level, biodiversity action plans have been developed
(Ministry of Environment and Natural Resource in Sri Lanka, 2007). The International
Union for Conservation of Nature (IUCN) is currently working on developing a legal
framework to safeguard traditional knowledge relating to the use of medicinal plants.
However, shortages of trained manpower and financial assistance, and weak legislation
have affected the successful implementation of policies in this field. As a result the
country’s biodiversity is continuing to decrease. Therefore, studies in this area would
provide enormous benefits for conserving biodiversity in the future. The next section
provides details about the present status and future challenges of biodiversity in Sri
Lanka.
26
2.3 Present status and future challenges of biodiversity
Sri Lanka has the highest biodiversity per unit area of land among Asian countries in
terms of flowering plants and all vertebrate groups except birds (Kotagama, 2002).
According to the Ministry of Environment and Natural Resources in Sri Lanka (2007)
the vegetation of Sri Lanka supports over 3,350 species of flowering plants and 314
species of ferns and fern allies. There is also considerable invertebrate faunal diversity.
The vertebrate fauna include 51 species of teleost fishes, 39 species of amphibians, over
125 species of reptilia, over 435 species of birds, 96 species of mammals including 38
species of marine mammals (IUCN, 2007). Among the vertebrates, there are 65 species
of freshwater fishes indigenous to Sri Lanka, of which about half is endemic. Many of
these species are riverine or marsh dwelling and occur mainly in the wet zone streams.
In addition, there are 22 species of introduced fish which are consumed for food. There
are about 350 species of marine fish which include ornamental fishes and food species
such as seer, tuna and skipjack (Ministry of Environment and Natural Resources in Sri
Lanka, 2007). Table 2.1 summarises overview of the status of some species in Sri
Lanka.
27
Table 2.1: The list of recorded species in different taxonomic groups
Taxonomic group No. of species Percentage of world flora
Sri Lanka World
Angiosperms 3,771 250,000 1.50
Gymnosperms 314 650 48.30
Pteridophytes 348 10,000 3.48
Bryophytes 566 17,500 3.23
Liverworts 222 - -
Lichens 661 13,500 4.80
Fungi 1,920 46,000 4.17
Algae *2,260 70,000 3.22
Virus/Bacteria (NA) 8,050 -
Source: Ministry of Environment and Natural Resources in Sri Lanka (2007)
Note: *Fresh water
In terms of species, genes and ecosystems, Sri Lanka has a very high biodiversity and is
one of the 18 hot spots in the world (IUCN, 2007). The wet zone rainforests have nearly
all of the country’s woody endemic plants and about 75 per cent of the endemic animals
(Ministry of Environment and Natural Resources in Sri Lanka, 2007). The genetic
diversity of agricultural crops is quite remarkable, with 3,000 accessions of rice being
recorded. The biodiversity of coastal and marine ecosystems provide over 65 per cent of
the animal protein requirement of the country. The Ministry of Environment and Natural
Resources in Sri Lanka (2007) provides detailed information about the diversity of
different species. Accordingly, in terms of plant species diversity, vegetation supports
over 3,368 species of flowering plants (of which 26 per cent are endemic) and 314
species of ferns and fern allies (of which 57 are endemic). Species diversity is also high
among mosses (575), liverworts (190), algae (2,260) and fungi (1,920).
28
The provisional list of ‘threatened’ faunal species of Sri Lanka includes over 550
species, of which over 50 per cent are endemic (Ministry of Environment and Natural
Resources in Sri Lanka, 2007). The crop genetic diversity in the country is also high,
especially for Oryza sativa. In addition to the diversity seen in coarse grains, legumes,
vegetables, roots and tubers and spice crops, there are over 170 species of ornamental
plants. In addition to that, domesticated animals provide a large number of benefits to
rural households. Among domesticated animals of economic value are some indigenous
species of buffalo, cattle, fowl and fish. Table 2.2 provides the status of estimated
number of selected species in Sri Lanka.
The major threat to biodiversity in Sri Lanka is the ever-increasing demand for land for
human habitation and related developmental activities. Poor land use planning,
indiscriminate exploitation of biological resources, weak enforcement of legislation and
the absence of an integrated conservation management approach are other threats to
biodiversity. Sri Lanka has established 501 protected areas, accounting for 26.5 per cent
of the total land area of the country. Sri Lanka has also two Ramsar sites and two
Biosphere Reserves. The biological resources of coastal and marine ecosystems provide
nearly 70 per cent of the protein requirements of the country and generate employment
for about 500,000 people (Ministry of Environment and Natural Resources in Sri Lanka,
2007). Biodiversity also contributes directly to the national economy in the form of
revenue from National Parks and other wildlife reserves, while it’s potential to promote
eco-tourism could be a significant income generator in the future.
29
Table 2.2: Estimated number of selected species
Taxonomic group No. of species (endemic) Percentage in Sri Lanka
World Sri Lanka
Vertebrate Fauna
Pisces 21,723 82 0.37
Amphibia 5,150 106+ 2.05
Reptilia 5,817 171 2.93
Aves 9,026 482 5.34
Mammalia 4,629 91 1.96
Invertebrate Fauna
Butterflies - 243 -
Dragonflies - 120 -
Freshwater Crabs - 51 -
Freshwater Shrimps - 23 -
Theraphosid spiders - 7+ -
Land molluscs - 246 -
Bees - 148 -
Aphids - 84 -
Ants - 181 -
Ticks - 27 -
Spiders - 501+ -
Marine Fauna
Echinoderms - 213 -
Marine Molluscs - 228 -
Sharks - 61 -
Rays - 31 -
Marine Reptiles - 18 -
Marine Mammals - 28 -
Source: Ministry of Environment and Natural Resources in Sri Lanka (2007)
30
Forests in Sri Lanka cover 1,933,000 hectares. The dense forest cover in Sri Lanka has
decreased by 23 per cent, mostly in the dry zone during the period 1956 to 2003. The
rate of deforestation from 1960 to 1990 has been estimated at 42,000 hectares per year
(Ministry of Environment and Natural Resources in Sri Lanka, 2007). Between 1990 and
2000, Sri Lanka lost an average of 26,800 hectares of forests per year. This amounts to a
rate of 1.14 per cent average annual deforestation. Between 2000 and 2005 the rate
accelerated to 1.43 per cent per annum. Threats to natural forest ecosystems in the wet
zone are mainly due to the expansion of tea, rubber, oil palm and other cash crops
(Department of Census and Statistics in Sri Lanka, 2007).
In the dry zone the cultivation of cash crops, large-scale development schemes like the
Accelerated Mahaweli Development Project and shifting cultivation have impacted on
natural forests. Mangrove ecosystems on the other hand, are threatened by the
reclamation of land, urbanisation and prawn culture. Dry zone ecosystems are also
disturbed by cyclones, which fortunately are not frequent. The construction of large
reservoirs continues to reduce the extent of natural ecosystems, particularly in the
lowland wet and intermediate zones. Some of the most important wet zone forests in
terms of biodiversity are the Peak Wilderness Sanctuary (22,379 hectares), the
Kanneliya-Dediyagala-Nakiyadeniya Reserve (10,139 hectares), the Sinharaja Forest
(11,280 hectares), the Knuckles Range of Forests (21,650 hectares) and the Horton
Plains National Park (3,159 hectares). These forests are also important hydrologically as
they protect the headwaters of all of Sri Lanka's main rivers (Ministry of Forestry and
Environment in Sri Lanka, 1999).
31
Most Sri Lankan habitats are officially protected by the Department of Forest
Conservation (DFC) and the Department of Wildlife Conservation (DWLC). Protected
areas under the DWLC are shown by the map given in Appendix E. These areas include
national parks, strict nature reserves, jungle corridors, and sanctuaries. Approximately
30 percent of the nation’s land area falls under some level of natural resource
management. Protected areas of which these are 501 in Sri Lanka are directly
administrated by DFC and DWLC. Among the world heritage sites, Sinharaja Forest
Reserve is an example of a national heritage forest. There are 32 forests categorized as
conservation forests including Knuckles Mountain Range. Total of all categories of
areas protected is 1,767,000 hectares. Protected areas in Sri Lanka account for 26.5
percent of total areas (Ministry of Environment and Natural Resources in Sri Lanka,
2007). This is a higher percentage of protected areas than in all of Asia and much of the
World. The natural ecosystem richness in the country is shown by the Table 2.3.
The list of protected areas in Sri Lanka is given in Appendix F. All sites contain endemic
species that are found nowhere else, and are therefore considered irreplaceable, with
several sites having more than 100 globally threatened species. All of these sites
technically have some form of protection, but there is an urgent need to strengthen the
management and monitoring of these areas. Additionally, landscape-scale conservation,
particularly reforestation and conservation of biological corridors will be required for
biodiversity to persist in the severely fragmented regions, even in the short term. One of
the most important reserves is the Sinharaja Forest Reserve, which encompasses 50 per
cent of the remaining lowland rainforest vegetation in Sri Lanka. Portions of the reserve
have been protected since 1875, and it was declared a World Heritage Site in 1989. Sixty
32
five per cent of Sri Lanka's 220 endemic tree and woody climber species and 270 species
of vertebrates have been recorded there (Ministry of Environment and Natural Resources
in Sri Lanka, 2007). Although public awareness of Sinharaja's biodiversity is growing,
the reserve still faces threats. People from neighbouring villages encroach on the reserve
via logging roads to collect non-timber forest products.
Table 2.3: Natural ecosystem richness
Types Categories Extent (hectares)
Forests Tropical lowland rainforests 141,506
Tropical lower-montane forests 68,616
Tropical upper-montane forests 3,108
Lawland dry-monsoon forests 243,886
Lawland semi-evergreen forests 1,090,981
Arid zone scrublands 464,076
Riverine forests 22,435
Grasslands Wet /Dry pathana grasslands 65,000
Savannahs -
Freshwater wetlands River and streams 5,913,800
Thalawas, Damanas, Villus 10,000
Marshes -
Swamp forest -
Brackish water wetlands Salt marshes 23,819
Mangroves 12,500
Lagoons and Estuaries 158,017
Coastal ecosystems Coral reefs -
Sea grass beds 33,573
Sea shores/beaches 11,788
Mud flats 9,754
Sand dunes 7,606
Source: Ministry of Environment and Natural Resources in Sri Lanka (2007)
33
Biodiversity is essential for ecosystem services and hence for human well-being. It goes
beyond the provisioning for material welfare and livelihoods to include security,
resiliency, social relations, health and choices. Therefore, during the last few decades,
the importance of community participation in biodiversity conservation has gained much
recognition in the country. However, degradation of biodiversity is still occurring at an
alarming rate in the country. The threats to biodiversity have several underlying causes.
They are population growth, trade pressures, political instability, perverse incentives,
economic performance, poverty, lack of law enforcement, poor protection standards,
lack of awareness and lack of information about the value of biodiversity. Biodiversity is
integral to key development sectors such as agriculture and livestock, forestry, and
fishing or tourism. More than 8.5 million people depend on biodiversity and on basic
ecosystems goods and services for their livelihoods (Sanjeeva, 2003). Since the poor
farmers are particularly dependent on the goods and services supplied by biodiversity,
development strategies that ignore their protection undermine poverty alleviation and are
therefore counterproductive. For this reason, it is crucial for development and poverty
alleviation strategies and programs in the country to prioritise biodiversity, especially
agricultural biodiversity which is an important component of general biodiversity. In the
next section we will discuss the present trend and issues related to agricultural
biodiversity in Sri Lanka.
2.4 Agricultural biodiversity in the country
The conservation of biological diversity is of special significance to Sri Lanka in the
context of its predominantly agriculture-based economy and the high dependence on
34
many plant species for food, medicines and domestic products (Jeremy Carew-Reid,
2002). Over one third of the plant species in the country are used in indigenous medical
practice, and many of these species are growing scarce due to habitat destruction and
over-collection. Sri Lanka has been an agrarian-based society. At present the agricultural
sector’s gross domestic production (GDP) contributes 20 per cent to the country's GDP,
second only to the manufacturing sector (Central Bank in Sri Lanka, 2009). Currently,
an estimated 8.9 million families are engaged in farming, and nearly 70 per cent of the
country's labour force is dependent upon the agricultural sector for its income and
sustenance (Department of Census and Statistics in Sri Lanka, 2010). The Sri Lankan
agricultural sector is dominated by small-holders, and over 55 per cent of farming
families in the country cultivate small holdings of less than 0.44 hectares. The
agricultural landscape of the country consists mainly of rice paddies, covering 780,000
hectares of cultivated land, and the plantation sector amounting to about 772,000
hectares (Department of Census and Statistics in Sri Lanka, 2010). The plantation crops
are tea, rubber, coconut and sugarcane, and on a smaller scale, coffee, cocoa, cinnamon,
pepper, clove and other spices.
Agricultural crop biodiversity in the island includes Oriza sativa with its 2,800
accessions and seven wild relatives; seven coarse grain species and their traditional
cultivars, maize and sorghum; 14 grain legumes species; eight cucurbitaceous; two
solonaceous and four other vegetable (bean, okra, amaranth, chilli) species; 17 root and
tuber crop species (Wijesinghe et al., 1993). The economically useful spices are eight
species of cinnamon, elettaria cardamomum, three pepper species with seven wild
relatives, cloves, nutmeg, betel nut, vanilla, chilli, and ginger. Others that are of
35
importance include citronella, three species of oil crops and two fiber crops. The
horticultural species are banana with nine cultivars and two wild relatives, citrus, and
over 15 other fruit species (Ministry of Environment and Natural Resources in Sri
Lanka, 2007). The rich and diverse ecosystems of the country harbour many wild
relatives of cultivated species, and the gene pools represented by these wild plants are a
resource of considerable potential value that could be used for the genetic improvement
of cultivated plants. Plant products such as fruits, fibre, spices, kitul sap, bamboo and
rattan are used as raw material for many small scale industries which provide financial
security to rural populations.
Paddy cultivation receives the highest attention in the agricultural sector (Central Bank
in Sri Lanka, 2009). Rice constitutes the staple food of the population and is the
backbone of Sri Lanka's agriculture and its ancient culture. There are varieties of rice
which are resistant to pests and adverse climatic and soil conditions, exhibit variations in
grain size and quality, and show differences in rate of maturing. There is also significant
crop genetic diversity among spices of commercial importance. Other crops in this
sector include over 100 species used as items of food. Many of these, such as onion,
potato and vegetables, remain a small farmer activity, and most fruit species are grown
in home gardens. Grain legumes and root and tuber crops also show a rich genetic
variability, as do fruit crops such as banana, mango and citrus. Similarly, there are many
varieties of vegetables such as cucurbits, tomato and eggplant. Out of 170 plant species
of ornamental value, 74 are endemic, and many species of orchids and foliage plants of
commercial importance occur naturally in forests (Department of Census and Statistics
in Sri Lanka, 2010). Grain legumes, such as cow pea, green gram, black gram, winged
36
bean, and soya bean constitute an important source of protein for most Sri Lankans,
particularly in rural areas, and are increasingly used for crop diversification. Winged
bean, in particular, shows much genetic variability as is evident in the seed colour, pod
size and flower colour. A few crops, such as chilli and cashew, are grown on a semi-
commercial scale (Department of Census and Statistics in Sri Lanka, 2010). A good
many field crops also continue to be harvested from shifting cultivation plots in the dry
zone. This method of agriculture has caused widespread forest destruction in the dry
zone where it has adversely affected overall biodiversity in the country.
Sri Lanka has a large number of vegetables, including both temperate and tropical
species, cultivated throughout the country. Among these, cucurbits, tomato and eggplant
exhibit high genetic diversity. There are also a fair number of root and tuber crops, of
which cassava, dioscorea and innala show considerable genetic variation. Sweet potato,
although introduced to this country, is naturalized and has high genetic variability. There
is also considerable genetic variation among a wide range of fruit crops, such as citrus,
mango, avocado and jak that are grown mainly in home gardens. Other fruit crops such
as durian, pomegranate, rambutan, guava and papaw have also been in cultivation for a
long time and exhibit a wide range of genetic diversity. Fruit crops such as wood apple
and velvet tamarind are a source of income for the dry zone farmers, and are harvested
from forests for sale. Of concern is the fact that harvesting of the latter species from
forests is destructive as it involves chopping down of large fruit bearing branches to
facilitate collection.
37
Among domesticated animals of economic value are wild species of buffalo, cattle and
fowl (Department of Census and Statistics in Sri Lanka, 2010). The local cattle show
high resistance to disease and tolerance of internal parasites. Likewise, the local breeds
of poultry are resistant to tropical diseases. In the livestock industry, the animals
commonly reared comprise neat cattle (1,644,000), buffalo (760,900), goats (535,200),
sheep (11,400), pigs (84,800) and poultry (9,136,600). The indigenous cattle have a very
low genetic potential for milk production, but are resistant to diseases and have the
ability to feed on coarse grasses. Several foreign breeds of cattle have been introduced to
the country over the last four decades in an effort to boost milk production. The local
backyard breed of scavenging poultry that are resistant to tropical diseases and were
commonly found in many village households prior to the 1960s are fast disappearing due
to the strong preference for imported germplasm (Ministry of Environment and Natural
Resources in Sri Lanka, 2007).
There are several reasons for the loss of agricultural biodiversity in rural areas in Sri
Lanka. After the green revolution, the adoption of modern varieties of seeds reached
from 12 to 67 per cent (Ministry of Environment and Natural Resources in Sri Lanka,
2007). Access to, and use of, a wide range of agricultural biodiversity is threatened by
this simplification of production systems. Secondly, as food production becomes
increasingly industrialised, we are witnessing a rapid decline in the diversity of varieties
used. The FAO (2007) estimates show that more than 90 per cent of crop varieties have
disappeared from farmers’ fields in the past 100 years. Agricultural plant varieties are
continuing to disappear at two per cent a year. Livestock breeds are being lost at five per
cent annually (FAO, 2007). The current extinction rate of species ranges from
38
approximately 1,000 to 10,000 times higher than natural extinction rate (Benton, 2001).
This is true for Sri Lanka as well. Thirdly, government incentives to specialise crops
(e.g. fertiliser subsidies) have also badly affected the agricultural biodiversity. Single
crops are more vulnerable to the rapid spread of disease. As government incentives
encourage farmers to increasingly produce crops for the market to obtain income, their
immediate dependence on agricultural biodiversity tends to diminish and they grow
fewer crops and a lesser number of varieties. Hence, commercial food production often
goes hand-in-hand with the reduction of cultivated crop or animal diversity.
Existing agricultural biodiversity has to be conserved in order to ensure access to it now
and in the future. This necessarily involves human intervention. Small-scale farms make
a substantial contribution to agricultural production, and it is estimated that there are
now a total of around 1.33 million small-scale farms in Sri Lanka, accounting for about
367,800 hectares of cultivated land (Department of Census and Statistics in Sri Lanka,
2010). Small-scale farms constitute a traditional system of perennial cropping for a wide
range of valuable crops and are considered important sites for in-situ conservation of
different components of agricultural biodiversity. This thesis will focus attention only on
small-scale farms in the country as a means to conserving agricultural biodiversity in the
future.
39
CHAPTER THREE
DATA SOURCES AND DESCRIPTION
3.1 Introduction
As mentioned in Chapter one, this PhD thesis consists of three main sections concerned
with different aspects of agricultural biodiversity. We use primary data along with
secondary data for the analysis. Secondary data were obtained from the Ministry of
Agriculture in Sri Lanka, Department of Census and Statistics, Central Bank of Sri
Lanka (various years), and various published books and articles. The Ministry of
Environment and Natural Resources in Sri Lanka provided data related to biodiversity
degradation in Sri Lanka. In addition to, data provided by the International Union for
Conservation of Nature (IUCN) and Food and Agricultural Organisation (FAO) were
used to explain the main issues in this area in the country.
The farm household data from three agricultural districts (Anuradhapura, Ampara and
Kurunegala) are used for the main analysis. A map showing in these three districts is
shown in Appendix G. There are at least three reasons for selecting farms in these
districts as representative farms for this study. Firstly, most of the farms in those districts
maintain a higher diversity which enables us to capture the market and non-market
benefits. Secondly, the diversity between the districts is significant. It will help us to
capture the benefits under heterogeneous systems. Thirdly, the loss of agricultural
biodiversity is increasing rapidly in these districts with modern agricultural practices.
40
Therefore, it is expected that farms in these districts may be representative farms which
will assist in an understanding of the issues in this field.
The determination of sample size is an important task for many researchers.
Inappropriate, inadequate, or excessive sample sizes continue to influence the quality
and accuracy of research (Bartlett et al., 2001). Generally, the actual sample size of a
survey is a compromise between the desired level of precision, the survey budget and
operational constraints such as budget and time. According to Wunsch (1986) two of the
most consistent flaws in data collection include (1) disregard for sampling error when
determining sample size and (2) disregard for response and non-response bias. This
clearly indicates that in developing a quantitative survey design, determining sample
size and dealing with non-response bias is essential. The following section will discuss
the selection of appropriate sample size in each district for this study.
3.2 Selecting appropriate sample size
The choice of survey population obviously depends on the objective of the survey
(Lukas, 2007). Given the survey population, a sampling strategy has to be determined.
Possible strategies include a simple random sample, a stratified random sample or a
choice-based sample (Dattalo, 2008). A simple random sample is generally a reasonable
choice. One reason for choosing a more specific sampling method may be the existence
of a relatively small but important sub-group which is of particular interest to the study.
Another reason may be to increase the precision of the estimates for a particular sub-
41
group (Bartlett et al., 2001). In practice the selection of sample strategy and sample size
is also largely dependent on the budget available for the survey.
Louviere et al. (2000) provide a formula to calculate the minimum sample size. Equation
3.1 provides the size of the sample, n, as determined by the desired level of accuracy of
the estimated probabilities, . Let p be a true proportion of the relevant population, a is
the percentage of deviation between and p that can be accepted and β is the confidence
level of the estimations such that: for a given n. Given this, the
minimum sample size is defined as:
2
1
2 21
1
pa
pn
(3.1)
where 2/11 is the inverse cumulative distribution function of a standard normal
distribution [N~(0,1)] taken at (1-α/2). Note that n refers to the size of the sample and
not the number of observations. Since each individual makes R succession of choices in
a choice experiment, the number of observations will be much larger (a sample of 500
individuals answering eight choice sets each will result in 4,000 observations). One of
the advantages of choice experiments is that the amount of information extracted from a
given sample size is much larger than, for example, using referendum based methods
and, hence, the efficiency of the estimates is improved. The formula above is only valid
for a simple random sample and with independency between the choices. A more
detailed explanation about this issue is found in studies carried out by Ben-Akiva and
Lerman (1985) and Louviere et al. (2000).
)ˆPr( appp
p
p
42
Pilot survey information was used to decide the minimum sample size in each district. In
agricultural research, a 90 per cent confidence interval is normally used (Bartlett et al.,
2001). It gives the level of risk the researcher is willing to take that the true margin of
error may exceed the acceptable margin of error. α is assumed to be 10 per cent. Based
on the pilot survey conducted in August 2010, estimation for the true choice proportion
of the relevant population is obtained. The level of allowable deviation as a percentage
between and p is assumed as 10 per cent (an equals 0.1). The parameters required to
estimate sample sizes and their calculations are reported in Table 3.1.
An estimate of the inverse cumulative normal distribution function is obtained using a
Microsoft Excel worksheet. It is clear that the cumulative distribution function (CDF) of
a normal distribution is the probability that a standard normal variable will take a value
less than or equal to z [P(Z ≤ z)] where z is some established numerical value of Z. Table
3.1 shows estimated true choice proportion of the population for each district. These
values are estimated using information provided by the pilot survey in these districts.
For example, for Anuradhapura district it is assumed that the researcher tolerates the
sampled proportion of decision makers, being within ± 10 per cent of the true
population proportions, P, and that the estimated population proportions of selecting
Farm A, Farm B and Neither Farm A or B are 0.41, 0.37 and 0.22 respectively. For the
pilot survey the number of choice scenarios used was eight per household. The Z statistic
was calculated using NORMINV (1-α/2, 0,1) formula in an Excel worksheet. This
formula can be used in Excel to calculate the inverse normal distribution function for
different normal distribution functions with varying means, standard deviations and at
varying levels of α. We entered a mean of zero and a standard deviation of one into the
p
p
43
NORMINV formula. This suggests that we are using a standard normal distribution such
that Z ~ N(0,1).
Table 3.1: Estimating minimum sample size for each district
Anuradhapura P a^2 1-P R Z^2 Total Observation n
Farm A 0.41 0.01 0.59 8.00 2.71 389.33 48.67
Farm B 0.37 0.01 0.63 8.00 2.71 460.67 57.58
None 0.22 0.01 0.78 8.00 2.71 959.24 119.90
Total 226.16
Ampara P a^2 1-P R Z^2 Total Observation n
Farm A 0.28 0.01 0.72 8.00 2.71 695.71 86.96
Farm B 0.46 0.01 0.54 8.00 2.71 317.61 39.70
None 0.26 0.01 0.74 8.00 2.71 770.04 96.25
Total 222.92
Kurunegala P a^2 1-P R Z^2 Total Observation n
Farm A 0.45 0.01 0.55 8.00 2.71 330.68 41.33
Farm B 0.35 0.01 0.65 8.00 2.71 502.46 62.81
None 0.20 0.01 0.80 8.00 2.71 1,082.22 135.28
Total 239.42
Note: Although optimal sample size for Anuradhapura, Ampara and Kurunegala are 226, 222 and 239, we
collected data covering 251, 247 and 248 households in these districts respectively. This allows us to
adjust the sample size after removing erroneous and irrational data points.
44
The total number of observation column provides the final number of observations that
is used to calculate the minimum sample size. The value of this column should be
divided by the number of choice scenarios, in this case, eight. This will provide the
minimum sample size for each study area. Using this information we calculated a
sample size of 226, 223 and 239 for study areas in Anuradhapura, Ampara and
Kurunegala respectively. However, in the survey we used only six choice sets as we
found that answering eight choice sets was difficult for respondents. We obtained data
from 251, 248 and 247 farmers in Anuradhapura, Ampara and Kurunegala respectively.
The method of selecting respondents for the survey in each district is explained in the
next section.
3.3 Selecting respondents for the survey
Several steps are involved in selecting farm households for the survey. Firstly, we
identified diverse farms located at divisional secretariat (DS) level in each district. Then
we selected one divisional secretariat regime in each district randomly. Secondly, four
villages in each divisional secretariat regime were selected randomly. During the third
stage, selections of households were done based on the name list provided by village
officers of the representative villages. We assigned random numbers to represent each
farm household address and used this number to select the households for the interview
(e.g. each third number). The survey was carried out during a two month period (Sept-
Oct. 2010).
45
The procedure explained in Section 3.2 identifies the minimum required sample size.
However, in practice the response rates are typically well below 100 per cent. Bartlett
(2001) recommends over sampling as a solution. For example, if it is anticipated that a
response rate of r per cent would be achieved based on prior research experience, the
required sample size to be selected to the survey can be calculated as rnSn / where
Sn = sample size adjusted for response rate. Details of survey areas, population and
sample size are provided in Table 3.2.
Table 3.2: Details of the survey areas
Research Area
(District and DS)
Villages Population
Size
Sample
Surveyi
Sample
Sizeii
Observationsiii
Anuradhapura
(Kahatagasdigiliya)
Puliyankadawela
Kudapattiya
Kaneddawewa
Kubukgollawa
1,352 288 247 4,446
Ampara
(Uhana)
Veeragoda
Udayagiriya
Himidurawa
Varankada
1,338 273 248 4,464
Kurunegela
(Paduvasnuwara)
Hathapola
Veediyagala
Kadavalagedara
Hidagahawawa
1,442 279 251 4,518
Note: i. This is the number of people selected for the survey allowing for the non-respondent households.
ii. This is the valid number of data points collected from the survey. A few survey questionnaires in
each district were dropped due to incomplete or erroneous reporting. These numbers were 8, 5 and 7 for
Anuradhapura, Ampara and Kurunegala districts respectively.
iii. This is the total number of possible observations for the choice experiment study.
46
The respondent rate is estimated based on pilot survey information. They were 94, 92
and 96 for Anuradhapura, Ampara and Kurunegala districts. However, actual response
rates for Anuradhapura, Ampara and Kurunegala districts were 88, 92 and 87 per cent
respectively. Population size in four selected villages in Anuradhapura district is 1,352.
Of them, 288 households were selected using the household list provided by the village
officer. We interviewed 255 households. However, only 247 survey forms could be used
to analyse the data as a few survey forms had to be dropped due to incomplete or
erroneous recording. Population size for the selected four villages in Ampara district is
1,338. Only 273 households were selected for the interview in these villages. After
dropping a few incomplete questionnaires 248 households could be used in the analysis.
Four selected villages in Kurunegala district have 1,442 households in total. Of them,
279 were selected for the survey. However, we used 251 households for the analysis.
The total number of possible observations in Anuradhapura, Ampara and Kurunegala
districts are 4,446, 4,464 and 4,518 respectively6.
It is commonly accepted that no survey can achieve success without a well-designed
questionnaire. It is a common challenge for many researchers with inappropriate,
inadequate, or excessive questions that can influence the quality and accuracy of
research (Bartlett et al., 2001). On the other hand, collecting inadequate information
provides data constraint in the analysis. Therefore, careful attention is needed in each
step of designing the questionnaire. The next section will discuss the process for
designing the field survey and its content.
6This number is estimated using the number of respondents, number of options and the number of choice
scenarios. For example, in Anuradhapura number of respondents was 247. The number of options was
three while the number of choice scenarios was six. Hence the total number of observations is 247*3*6 =
4,446.
47
3.4 Field survey and its content
We use primary data along with secondary data for the analysis. Survey data were
collected covering approximately 746 farmers in three agricultural districts in Sri Lanka.
In August 2010, a pilot survey was conducted to obtain the necessary information for the
main survey in certain randomly chosen areas of the Anuradhapura, Ampara and
Kurunegala districts. The main survey was started at the beginning of September 2010
and completed at the end of October 2010. Surveys in all districts were carried out by
administering a questionnaire through a face-to-face interview with the head or any
other working member of the households.
A questionnaire designed to capture the various aspects of agricultural biodiversity was
validated in a pilot survey and in a number of focus group discussions. The final
questionnaire was then adjusted. The gathering of data was carried out carefully by a
trained group of researchers under the close supervision of their search team. The
interviews took place in the interviewee’s home. The participants were informed about
the purpose of the study and gave verbal consent. A field supervisor reviewed the quality
of the data gathered and entered it into a database for analysis. It was confirmed that the
survey questions were clearly understood by respondents and obtained appropriate
information regarding agricultural biodiversity, its different components and each
farmer’s attitudes towards conserving it.
The questionnaire used for the survey had six main sections. Section A covered general
information about the small scale farm and the methods farmers use to cultivate it. This
48
section had 15 main questions following a few other sub questions. Section B sought
details of the different components of agricultural biodiversity and the level of efficiency
on the farm. At the beginning of this section the enumerator gave a broad introduction
on diverse farming systems, practiced in different areas in Sri Lanka and then narrowed
attention to the farming system in the survey areas. Then details about different types of
benefits that farmers can obtain by having a diverse farming system are gathered. In
addition to that farming practices, different crops varieties and livestock breeds, cost and
production data were also collected. Further, data on inputs as well as outputs were
obtained in detail for estimating farm level efficiency. Section C dealt with evaluating
poverty, income and expenditure. Household income and expenditure, food availability,
their health situation and details of agricultural as well as non-agricultural debt were
obtained in this section. Section D collected information about farmers’ preference for
agricultural biodiversity on farms. This is the CE part of the questionnaire. More details
about this section are provided in Section 3.5. Section E measured the farmer’s attitudes
towards different components of agricultural biodiversity on farms while Section F
covered various socio-economic and demographic features such as age, gender, level of
education, marital status, occupation and the size of the dwellings and total family
income. The questionnaire used for this survey is shown in Appendix H.
Prior to conducting the survey, the enumerators attended training conducted by the
researcher. They were briefed on the CE procedure, the idea of economic valuation, the
background of the study. Role-play exercises were used to expose the enumerators to the
ways of obtaining cooperation from the respondents. They were also made aware of
possible biases (like strategic and starting-point bias) during interviews and ways to
49
minimise these. The enumerators were taken for a brief tour to familiarise them with the
areas of the study sites and also to meet the village heads to seek their help in getting
respondents to cooperate in the survey.
The CE part is the most important section of the questionnaire and it needs expert
knowledge and careful attention. In a CE, individuals are presented with a choice set or
series of choice sets that are framed with various attributes and attribute levels and are
asked to choose one bundle at a varied set of price and attribute levels. Consumers’
willingness to accept (WTA) compensation payment for each attribute is then computed
from estimates of econometric models. An intrinsic problem that all researchers face in
designing a survey questionnaire is how much information or complexity to incorporate.
Specifically, these issues may include which attributes should be used, how many levels
of each attribute need to be considered, how many alternatives need to be presented in
each choice set, and how many choice sets should be included in each questionnaire.
More detail about the way of addressing these issues is explained in Chapter Four. The
process for designing CE questions for this survey is briefly explained in the next
section.
3.5 Design choice experiment (CE) survey
The overall objective of the CE part of the study is to estimate the possible private
benefits that could be achieved from conserving agricultural biodiversity. Under this
method a sample of people is asked to choose their most preferred alternatives from a
sequence of grouped options that relate to different agricultural biodiversity
50
management strategies. Each option is described in terms of its agricultural biodiversity
outcomes and a monetary cost to be borne personally by the respondent. By analysing
the choices made by respondents it is possible to infer the tradeoffs that people are
willing to make between money and greater benefits of agricultural biodiversity. This in
turn allows the estimation of changes of private benefits with changing levels of
agricultural biodiversity.
Experimental levels of the six agricultural biodiversity attributes described below were
identified through prior knowledge and literature in this field. A monetary attribute in
terms of required additional labour days is included in order to estimate welfare
changes7. The monetary attribute in this CE is a proxy, measuring the labour costs that
farmers have to allocate for receiving the benefits of agricultural biodiversity. This
attribute represents WTA compensation which is measured as a cost rather than a
benefit. Farm attributes and their levels used in this study are explained below.
Farm attributes and their levels include:
1. Crop species diversity. This is measured by the total number of crop species that are
grown in the small-scale farm in a given season. For example, a farm with tomatoes,
beans and carrots has in total 3 different crops. We present this with four levels of crop
diversity: 3, 7, 10, and 15 varieties.
7This indirect measure is preferred over a direct monetary attribute because most (if not all) of the outputs
and functions of farms that result in agricultural biodiversity are not traded in the markets, but consumed
by the farm families themselves. Hence, they are not likely to be familiar with a direct monetary measure.
The proxy monetary attribute can easily be converted into actual monetary units by using secondary data
on labour costs.
51
2. Mixed crop and livestock diversity. This attribute investigates whether a farmer
prefers an integrated crop and livestock production system over a system that is
specialised in crops or livestock.
3. Organic production. This attribute investigates whether a farmer prefers organic
methods of production over a system using chemical fertiliser and pesticides. For
example, when a farmer sells small-scale farm crops that are produced entirely with
organic methods, these products are certified as organic. We asked farmers to think
about their imaginary farm and decide whether or not they prefer a farm in which they
produce crops with entirely organic methods.
4. Landrace cultivation. This attribute investigates whether a farmer prefers to have a
farm in which a landrace is grown as opposed to none. A landrace cultivation is defined
as a crop variety that was passed down from their ancestors and is very resistant to any
disease. In general these varieties are called traditional varieties in rural agricultural
areas in Sri Lanka. Varieties that were introduced after the agricultural modernisation
programs, took place during the 1960s are called modern varieties.
5. Estimated benefits in terms of decreasing household’s food expenditure. This is
defined as a percentage of decreasing household food expenditure under different policy
options. Farmers receive these benefits as the diverse farming system increases their
self-sufficiency level. It indicates the additional benefits that farmers are going to
receive when they are accepting a new policy. We present this attribute with three levels
of percentages: 5, 10, and 15.
52
6. Estimated costs in terms of additional labour days. This is defined as a percentage of
additional labour requirements under different policy options. It indicates the additional
costs that farmers have to bear when they are accepting a new policy. The percentages
that are presented to them are 10, 20 and 30.
The first four attributes reflect the various attributes of agricultural biodiversity found in
the small-scale farms in Sri Lanka. The fifth factor represents benefits that farmers can
receive in terms of receiving foods from their farms under different policy options. The
last factor is the monetary attribute in terms of additional labour costs that farmers have
under different policy options. As compared to willingness to pay (WTP), willingness to
accept is measured as a benefit rather than a cost (Freeman, 2003). In order to estimate
this benefit, a monetary attribute in terms of additional labour costs that farmers are
willing to offer is included. The size of the hypothetical small-scale farm is fixed as one
acre8 in area in each case (this is the average small-scale farm size in Sri Lanka).
There are several different design types in the literature to obtain a choice set. One is a
full factorial design which consists of all possible choice situations (Bennett and
Blamey, 2001). With this design all possible effects (main and interaction effects) can be
estimated. However, for a practical study the number of choice situations in a full
factorial design is too large. Therefore, most people rely on fractional factorial designs.
However, within this class there exist many different types of designs. One could
randomly select choice situations from the full factorial, but clearly this is not the best
8This small-scale farm size was chosen from the agricultural census survey conducted in 2002 (Census of
Agriculture, 2002).
53
way of doing it. Another way to select choice situations in a structured manner, such that
the best data from the stated CE will be produced in estimating the model (Hensher et
al., 2005). A fractional factorial design consists of a subset of choice situations from the
full factorial. The most well-known fractional factorial design type is the orthogonal
design, which aims to minimise the correlation between the attribute levels in the choice
situations (Kuhfeld, 2005). However, these orthogonal designs have limitations and
cannot avoid choice situations in which a certain alternative is clearly more preferred
over the others (hence not providing much information). More recently, several
researchers have suggested another type of fractional factorial designs, so-called
efficient designs (Hensher et al., 2005; Scarpa and Rose, 2008).
Instead of merely looking at the correlation between the attribute levels, efficient designs
aim to find designs that are statistically as efficient as possible in terms of predicted
standard errors of the parameter estimates. Essentially, these designs attempt to
maximise the information from each choice situation. In case any information about the
parameters is available, then efficient designs will always outperform orthogonal designs
(Kessels et al., 2006). This is due to the fact that efficient designs use the knowledge of
the prior parameters to optimise the design in which the most information is gained from
each choice situation (e.g. dominant alternatives can be avoided as the utilities can be
computed). While efficient designs outperform the orthogonal designs, prior parameter
estimates need to be available (Hensher et al., 2005). Therefore, efficient designs rely on
the accuracy of the prior parameter estimates. As we do not have the prior parameter
values for our estimation in this study we used orthogonal design to generate the number
of choice situation in this study.
54
Three reasons can be given to justify using orthogonal design in this study. Firstly, it
allows for an independent estimation of the influence of each design attribute on choice.
Secondly, with the absence of prior parameter, there is no way to apply efficient design
in this study. Thirdly, the common use of orthogonal designs in stated choice studies is
largely a result of historical impetus. In the past, the experimental design literature has
been primarily concerned with linear models (such as linear regression models), where
the orthogonality of data is considered important (Scarpa and Rose, 2008). In linear
regression models, this is because (a) orthogonality ensures that the model will not
suffer from multicollinearity, and (b) orthogonality is thought to minimise the variances
of the parameter estimates, which are taken from the variance-covariance (VC) matrix of
the model (Hensher et al., 2005). The VC matrix of a linear regression model is given in
Equation 3.2.
1'2 XXVC (3.2)
where 2 is the model variance, and X is the matrix of attribute levels in the design or in
the data to be used in estimation. Fixing the model variance, the elements of the VC
matrix for linear regression models are minimised when the X matrix is orthogonal. A
design that results in a model where the elements contained within the VC matrix are
minimised is preferable, for two reasons (Hensher et al., 2005). Firstly, such a design
will produce the smallest possible standard errors, and hence maximise the t-ratios
produced from that model. Secondly, an orthogonal design will produce zero-off
diagonals in the models VC matrix, thus ensuring that the parameter estimates are
55
unconfounded with one another (or no multicollinearity problem). As such, orthogonal
designs, at least in relation to linear models, meet the two criteria for a good design
(Scarpa and Rose, 2008). They allow for an independent determination of each attributes
contribution on the dependent variable, and they maximise the power of the design to
detect statistically significant relationships (e.g. maximise the t-ratios at any given
sample size).
In this study orthogonal design is used to generate the number of choice situations. A
large number of unique farm profiles can be constructed from the six attributes and their
levels. An orthogonalisation procedure was used to recover only the main effects,
consisting of 36 pair-wise comparisons of different farm profiles. These were randomly
blocked to six different versions, with six choice sets. In face-to-face interviews, each
farmer was presented with six choice sets. The questionnaire used for this survey is
shown in Appendix H. More details about the implementation of the choice experiment
study are given in Chapter Four.
Hypothetical farms in pairs on a series of cards were generated and then farmers were
asked- to indicate out of the pair, which type of farm they preferred for each card. Each
set contained two farm profiles and an option to select neither. The farmers who took
part in the choice experiment were by and large those responsible for making decisions
in the farms. Enumerators explained the context in which choices were to be made; (a)
farm size is an acre; and (b) that attributes of farms had been selected as a result of prior
research and were combined artificially. More details about the CE survey are given in
Chapter Four.
56
The next four chapters provide methodology, literature and results of the analysis related
to the three main sections in this thesis. Each study is carried out as a separate study and
presented as a separate chapter in the thesis. Chapters Five, Six and Seven used a total of
746 observations for their main analysis while Chapter Four used 12,006 observations
for its pool data analysis. In most of the cases, the analysis is carried out using district
wise data and pool data separately. This type of analysis will help to understand the
heterogeneity across different districts.
57
CHAPTER FOUR
FARMERS’ VALUATION OF AGRICULTURAL BIODIVERSITY
4.1 Introduction
The valuation of nonmarket goods is one of the principal issues addressed by
environmental economics research (Bishop and Romano, 1998; Champ et al., 2004).
When competitive markets exist, market prices are the appropriate measure of social
valuation. However, in practice, all markets do not function exactly in the manner
assumed by economic theory. In such cases market prices are not the best available
approximate measure of social values of goods and services (Portney, 1994; Freeman,
2003). For example, all benefits of diverse farming practice provided by small-scale
farms are not marketed in rural areas. However, it is extremely important to analyse the
role of subjective well-being received from these farms for informing policy decisions.
The value of agricultural biodiversity can be measured in a variety of ways. However,
the range of agricultural biodiversity valuation techniques can be considered under two
headings that reflect the continuum from pure market to pure non-market techniques
(Freeman, 2003). The first method uses revealed preference techniques because people’s
preferences for agricultural biodiversity protection are revealed through their actions in
related markets. The second method uses stated preference techniques. These are
valuation techniques that require people to state the strength of their preferences and
58
hence reveal the values they enjoy through structured questionnaires (Bishop and
Romano, 1998). This method does not involve any reliance on market data.
For market based valuation techniques, the benefit generated by agricultural biodiversity
must be bought and sold in markets. The techniques are most suitable for applications
where direct use benefits are involved. As both consumer and producer receive the
benefits, consumer surplus and producer surplus can be used to measure the total
benefits received from use value of agricultural biodiversity9. Therefore, it is clear that if
there are sufficient observations of trade, it is possible to use standard economic
techniques to estimate values for both buyers and sellers (Freeman, 2003). For example,
if a species is under threat of extinction, the cost of a captive breeding program may be
used to estimate the benefit being provided by its continued survival. Another approach
involves the estimation of how much it would cost to replace the lost agricultural
biodiversity benefit with a substitute. This replacement cost technique is widely used in
various analyses because of its reliability as well as the simplicity of capturing the
relevant cost.
Limitations in the range of agricultural biodiversity value types that can be estimated
using either the market based or revealed preference techniques, led to the development
of stated preference techniques (Champ et al., 2004). In this type of technique a sample
of people are asked about their preferences for a biodiversity sensitive asset under a
9Observations of market supply (the marginal costs of suppliers) and prices received through transactions
recorded in markets allow the estimation of profits enjoyed by producers (known technically as the
producers’ surplus). Observations of market demand (the marginal values of consumers) and price paid
allow the estimation of the net benefit received by consumers when they purchase the biodiversity derived
goods or service involved. This is known as the consumers’ surplus.
59
hypothetical set of circumstances. A number of different methods have been developed
to inquire about peoples’ preferences. The first stated preference technique to be
developed was the contingent valuation method (CVM)10
. Originally, this method
required that a sample of people be asked the amount they would be willing to pay to
secure an improvement in a particular aspect of agricultural biodiversity. More recently,
this technique has been refined to accommodate a dichotomous choice version that
involves people being asked if they would or would not support a proposal to improve
agricultural biodiversity given some personal monetary cost.
A widely used stated preference technique is the CE method11
. Under this method a
sample of people is asked to choose their most preferred alternatives from a sequence of
grouped options that, in the case of this study relate to different biodiversity
management strategies. Each option is described in terms of its agricultural biodiversity
outcomes and a personal monetary cost to be borne personally by the respondent. By
analysing the choices made by respondents it is possible to infer the tradeoff that people
are willing to make between money and greater biodiversity benefits. This in turn allows
the estimation of values for agricultural biodiversity changes. In this study, particular
effort is given to using the CE method for valuation of different attributes of agricultural
biodiversity.
The next section critically looks at the existing research that is directly linked to
valuation of agricultural biodiversity in different countries. It provides the context for
10
The idea of CVM was first suggested by Ciriacy-Wantrup (1947), and the first study ever done was in
1961 by Davis (1963). 11
For a detailed explanation of choice experiment design techniques, please see Louviere et al., 2000;
Bennett and Blamey, 2001; Bateman et al., 2003; Drucker et al., 2005; Hensher, et al., 2005.
60
the present research by looking at what work has already been done in this field. It also
identifies the shortcomings of existing work and highlights the importance of carrying
out the present work.
4.2 Literature review on valuation of agricultural biodiversity
There have been some studies that have employed CE method or CVM to value crop
diversities, livestock diversities and other types of farming practices in different
countries. Hanley et al. (1998) employed the CE method to aid the design of agri-
environmental programs that yield the highest benefitsin Scotland. They also valued the
components of a Scottish agri-environmental scheme, which offers payments to farmers
in return for adoption of conservation practices. Scarpa et al. (2003) estimated the value
of animal genetic resources to farm families, who produce and consume them, by
comparing the value of attributes of creole pigs to those of more productive, but less
well adapted exotic breeds in Yucatan, Mexico. Kontoleon (2003) investigated
consumers’ perceptions of genetically modified (GM) food and found that consumers
across the European Union (EU) were willing to pay more to obtain information on the
GM content in their food supplies.
Using the CE method, Lusk et al. (2003) investigated consumers’ preferences for beef
produced with hormones in the United States. Ndjeunga and Nelson (2005) estimated
farmer valuation of crop varieties, whereas Birol (2004) estimated farmer valuation of
several components of agricultural biodiversity in Hungarian home gardens. In this
study she applied the CE method to estimate farmers’ valuation of agricultural
61
biodiversity using primary data collected in three environmentally sensitive areas of
Hungary. Her findings show the variation in values farmers assign to home gardens
across regions and households. The CE method was used to investigate farmers’
valuation of agricultural biodiversity of maize varieties, using 414 farm households from
three states of Mexico by Birol et al. (2006). The results revealed that there is a
considerable heterogeneity in farmers’ preferences for Milpa diversity and GM maize
across and within the three states.
Ouma et al. (2007) used mixed logit and latent class models to examine preferences for
cattle traits with a focus on heterogeneity among cattle keepers, using CE data of 506
cattle-keeping households in Kenya and Ethiopia. The findings indicated the existence
of preference heterogeneity based on cattle production. Ruto et al. (2008) investigated
buyers’ preference for indigenous breads and Roessler et al. (2008) assessed farmers
preferences and trade-offs for pig breeding for a list of adaptive and productive traits
using the CE method. Further, Zander and Drucker (2008) provided empirical evidence
for the high economic value of the Borana breeds using CE surveys.
A CE method was employed to elicit the preferences and a random parameter logit
(RPL) model was used to estimate the relative importance of the preferred attributes of
indigenous cows in Central Ethiopia by Kassie et al. (2009). They identified the relative
weights assigned to the preferred traits of the indigenous cow population in the most
dominant crop-livestock mixed production system. The results show that fertility,
disease resistance and calf vigour traits are at least as important as milk provided by
cows. The location the cows are brought from is an important attribute for buyers. The
62
findings suggest that the smallholder community in this part of Ethiopia depends on
semi-subsistence agriculture and so livestock development interventions should focus on
a multitude of reproductive and adaptive traits that stabilise the herd structure rather than
focusing on traits that are only important for commercial purposes.
Poudel and Johnsen (2009) soughtto advance the application of CVM to document the
economic value of crop genetic resources based on farmers’ willingness to pay for
conservation. According to them landholding size, household size, education level,
socio-economic status, gender of respondent, number of crop landraces grown, and
knowledge of biodiversity influence the willingness to pay for in situ conservation,
whereas only landholding size and household size influence the willingness to pay for ex
situ conservation. The CE approach was employed to investigate Ethiopian farmers’
crop variety preferences and estimate the mean willingness to pay for each crop variety
attributes by Asrat et al. (2009). They also identified household-specific and institutional
factors that governed the preferences. However, the costs and benefits estimated from
these studies cannot be generalized for all countries. The range in benefits is extremely
sensitive to assumptions concerning socioeconomic characteristics and the discount rate.
Recently, a choice experiment method was used by Kikulwe et al. (2011) to estimate
farmers’ valuation of agricultural biodiversity in the milpa system, and examined their
interest in cultivating genetically modified (GM) maize.
Although these studies identified the importance of small-scale farms for conserving
agricultural biodiversity, literature on economic valuation of both crop and livestock
resources in small-scale farms are very limited in developing countries. This is because
63
assigning monetary values to crop and livestock resources are complicated in
subsistence farming systems (Gauchan, 2004) and, therefore, a challenging area of
study. Furthermore, the above review has demonstrated that most studies have tended to
simply value a particular biological resource such as species, habitat or ecosystem
service in agriculture. As a result these studies have only provided limited information
on the value of the different attributes of agricultural biological diversity. Accordingly, it
is obvious that more conceptual and theoretical work is needed to develop a better
understanding of feasible, cost-effective approaches to valuing multiple attributes of
agricultural biodiversity in developing countries.
Among the environmental valuation methods, the CE method is considered to be the
most appropriate method for valuing the multiple benefits of small-scale farms
attributes. This is because the CE method allows for estimation not only of the value of
the environmental good as a whole, but also of the implicit values of its attributes
(Hanley et al., 1998; Bateman et al., 2003). This approach has a theoretical grounding in
Lancaster’s attribute theory of consumer choice (Lancaster, 1966) and an econometric
basis in models of random utility (Luce, 1959; McFadden, 1974). Therefore, in the next
section, the theoretical explanation for the random utility model (RUM) is provided.
64
4.3 Random Utility Models (RUM)
The CE model is of the class of multinomial choice models used to analyse the discrete
response data produced by the survey instrument12
. The CE methods rely on the random
utility model framework to provide a utility theoretical interpretation of the discrete
responses observed from the respondents. Garber-Yonts (2001) provided the basic steps
of the RUM and a derivation of WTP compensation that is explained below. Given a set
of alternatives An, presented to an individual n, the probability that any one alternative i
is chosen is given by:
(4.1)
where Uin is the utility that individual n achieves by choosing alternative i. According to
the random utility theory, the utility which is not directly observable can be partitioned
into a deterministic component and a random component (Hanemann, 1984; Ben-Akiva
and Lerman 1985; Garber-Yonts, 2001). The accompanying assumption is that the
individual knows their utility function with certainty, however with other measurement
errors, utility can be stochastic:
(4.2)
12
The principal alternative method of WTP elicitation is using open ended questions to which the
respondent provides a direct statement of the amount they would pay to gain an economic benefit, or
alternatively, accept in compensation or forego. Although this elicitation method is much simpler to
analyse from a statistical perspective, it has been shown to be problematic in eliciting accurate responses
(Arrow et al., 1992). The advantage of closed-ended, discrete response elicitation questions with respect to
realism and incentive compatibility are purchased at the cost of greater statistical complexity.
),Pr()/( njjninn AVUUAiP
ininin VU
65
where Vin is the mean and the random disturbance of the stochastic random utility
function. The specification of Vin includes a vector of attribute of alternative i, Xin,
which includes a price or bid variable, and a vector of characteristics of the respondent,
Hn, including income (Garber-Yonts, 2001). Thus model can be written as Equation 4.3:
(4.3)
where the deterministic component is here specified as linear in parameters, though the
function f(.) can be nonlinear. However, when choosing the functional form, there is a
trade-off between the benefits of assuming a less restrictive formulation and the
complications that arise from doing so. This is especially relevant for the way income
enters the utility function (Garber-Yonts, 2001). A simpler functional form (e.g. linear in
income) makes estimation of the parameters and calculation of welfare effects easier, but
the estimates are based on restrictive assumptions (Ben-Akiva and Lerman, 1985). Most
often researchers have been inclined to use a simpler linear in the parameters utility
function. Another important thing is that the error term enters the utility function as an
additive term. This assumption, although restrictive, greatly simplifies the computation
of the results and the estimation of welfare measures. With the indirect utility specified
as above, the individual seeks to maximise utility such that:
(4.4)
inninin HXfU ),('
jnnjninninnn HXfHXfPAiP ),(),(()/( ''
jiAjiHXfHXfPAiP ninjnnjnninnn ,,));(),(),(()/( ''
66
It becomes clear that unless Hn enters the function f(.) nonadditively, it appears
identically on both sides of the inequality and cancels out of the function. Thus, Hn must
enter nonadditively if the effects of respondent characteristics on choice are to be
measured (Garber-Yonts, 2001). If εin and εjn are assumed to be extreme value
independently and identically distributed (IID) with scale parameter µ, then ε*
=εjn - εinis
logistically distributed (Ben-Akiva and Lerman, 1985). This distributional assumption
approximates the normal distribution which leads to the multinomial logit (MNL) model
for the choice probabilities (McFadden, 1974; Ben-Akiva and Lerman, 1985). This is the
simplest version of the analysis of multinomial outcomes when comparing with
conditional logit (CL) model and RPL model. MNL model can be given as Equation 4.5:
(4.5)
Since µappears as a multiplicative constant on every parameter of the model, it is not
identifiable. A common assumption employed by users of MNL models is that the scale
parameter, µ, is equal to one, which has a homoscedastic disturbances (Garber-Yonts,
2001). Empirical observations about this assumption found that it was not significantly
different that one (Xu, 1997; Adamowicz et al., 1998). Therefore, we adhere to this
assumption in this study. The log likelihood function for the MNL model can be written
as Equation 4.6:
(4.6)
),(),(''
//)/( njn
n
jnV
nin
n
jnVjnV HXf
Aj
HXf
Ajnn eeeeeAiP
)],(ln),([)/(ln ''
njnAjninn Ai innn Ai in HXfHXfsAiPsLnnn
67
where sin=1if alternative i is chosen by individual n, otherwise sin= 0. Garber-Yonts
(2001) provides the details explanation about the derivatives of all Equations related to
MNL. The necessary first order conditions to maximise the likelihood function are
obtained by setting the first derivative of Equation 4.6 with respect to the parameter
vector equal to zero:
(4.7)
Estimation of the parameters of this model can be done by using maximisation of the
multinomial likelihood. This usually requires numerical procedures, and Fisher scoring
or Newton-Raphson often work rather well. McFadden (1974) argues that, under certain
conditions, ln L in Equation 4.6 is globally concave so that a solution to Equation 4.7
exists and is unique. Thus the maximum likelihood estimator of β is consistent,
asymptotically normal, and asymptotically efficient.
Estimation of Hicksian welfare effects from the MNL choice probabilities follows the
method outlined by Hanemann (1984) and Hanemann and Kanninen (1999). Given a
quantity change in the level of a public good from to , the compensating surplus
which exactly offsets the utility gain of the change is the level of B which provides the
equality:
(4.8)
68
where v is indirect utility, p is the vector of market prices, a X is vector of attributes
other than the bid level B, y is income, H is a vector of the socio-demographic
characteristics, and is a random error term. The objective is to obtain the solution for
the expected value of which is the maximum WTP for the
change from to Assuming the additive separability of the cost attribute of the
individual’s indirect utility function, we can express the deterministic part of utility as
shown in Equation 4.9:
(4.9)
where B is the specified bid level alternative i, and is associate parameter. The
following measures Total WTP/Total WTA (TWTP/TWTA) for a change in the
attributes of a good from state i to state j aggregated over all observations (Hanemann,
1984; Adamowicz et al.,1994; Xu, 1997; Garber-Yonts, 2001 ):
(4.10)
If the mean value of TWTP/TWTA for the change in all attributes from state i to state j
is for interest, Equation 4.10 simplifies to:
(4.11)
69
where f(X,H) is evaluated at the sample mean value of H, recalling that H drops out of
the Equation if it enters f(.) additively. The TWTP/TWTA for the “part-worth” of the
change of an individual attribute k from state i to state j, holding other attributes
constant, further simplifies to Equation 4.12:
(4.12)
Finally, as adopted by Hanemann et al. (1991); Xu (1997) and Garber-Yonts, (2001) the
Hicksian compensated demand curve, depicting marginal WTP/WTA for attribute k at
level i, is given as Equation 4.13:
(4.13)
In choice modelling applications to agricultural biodiversity, different components of
agricultural biodiversity as well as monetary factors should be included as attributes of
the options in a choice set. Thus, choice modelling allows one to obtain compensating
surplus estimates so that one can account for the welfare change generated by a bundle
of changes in relevant attributes. It is also possible to determine the relative importance
of these attributes to people in making their choices.
Haneman and Kanninen (1999) make an important distinction between the conventional
regression techniques used in analysis of open ended WTP data and the limited
dependent variable models used in conjunction with discrete choice elicitation methods.
With the former, the investigator obtains an estimate of the mean WTP conditional on
the regressors. The later estimates the entire conditional cumulative distribution function
70
(cdf) of the dependent variable. The preferred measure of central tendency by which to
summarise the estimated cdf is therefore at the discretion of the investigator, and its
selection can significantly alter the results of the analysis (Garber-Yonts, 2001).
It is clear that the choice experiment technique is an application of the characteristics
theory of value combined with random utility theory (see, for example, Thurstone, 1927;
Lancaster, 1966; Manski, 1977). In this method, respondents are asked to choose
between different bundles of (environmental) goods, which are described in terms of
their attributes, or characteristics, and the levels that these take. The CE approach is
essentially a structured method of data generation. It relies on carefully designed choice
tasks that help reveal the factors influencing choice. Designing a CE technique also
requires careful definition of the attribute levels and ranges. Furthermore, the choice
experiment approach involves the use of statistical design theory to construct choice
scenarios which can yield parameter estimates that are not confounded by other factors.
In the next section, we discuss the main steps to be followed when applying CE method
for environment valuation.
4.4 Choice experiment method
Since the CE method paves the way to estimate farmers’ preferences for agricultural
biodiversity in small-scale farms, this method is used to analyse the data gathered from
personal interviews with farmers. It is the most appropriate for valuing attributes of
small-scale farms, considering their multiple benefits and functions. This method, which
is based on farmers choosing between hypothetical (biodiversity enhanced agricultural
71
system) farms, enables estimation of the value of new small-scale farm attributes, which
are outside farmers’ current set of experiences (Adamowicz et al., 1994). As mentioned
in the previous section, the CE method has its theoretical grounding in Lancaster’s
model of consumer choice (Lancaster, 1966). Lancaster proposed that consumers derive
satisfaction not from goods themselves, but from the attributes they provide. To
illustrate the basic model behind choice experiments, assume that farm families have a
utility function of the form:
(4.14)
where for any farm family a given level of utility will be associated with any
alternative small-scale farm Utility derived from any of the small-scale farm
alternatives depend on the attributes of the small-scale farm and the social and
economic characteristics of the farm family , since different families may receive
different levels of utility from these attributes. According to the random utility model,
the utility of a choice comprises of a systematic (deterministic) component, and an
error (random) component, , which is independent of the deterministic part and
follows a predetermined distribution (Hanemann et al., 1991):
(4.15)
The systematic component can be explained as a function of the characteristics of the
small-scale farm and of the social and economic characteristics of the farm family.
Accordingly, Equation 4.15 can be expressed as
.
,i
.j
ijX
iZ
ijT
ije
),( iijij ZXUU
ijijij eTU
iiijij eZXTU ),(
72
Given an error part in the utility function, predictions cannot be made with certainty and
the analysis becomes one of probabilistic choice (Bateman et al., 2003). Consequently,
choices made between alternative small-scale farms will be a function of the probability
that the utility associated with a particular small scale-farm option is higher than that
for other alternative small scale-farm. Hence, the probability that farm family will
choose small-scale farm over all other options is given by:
where .
We assume that the relationship between utility and attributes follows a linear path in the
parameters and variables. We further assume that the error terms are identically and
independently distributed with a Weibull distribution13
(Greene, 1997). These
assumptions ensure that the probability of any particular alternative j being chosen can
be expressed in terms of logistic distribution. This specification is known as the CL
model (McFadden, 1974; Greene, 1997; Maddala, 1999) which has the following
general form:
(4.16)
The components of Xij are typically called the attribute of the choices. However, Zi
contains characteristics of the individual and is, therefore, the same for all choices.
Equation 4.16 is the probabilistic response function and it shows that, given all other
13
Weibull distribution is a continuous probability distribution. For further details about the basic properties
of this distribution, please see Greene (1997).
)( j
i
j n
ininijijij eTeTprobP nj
J
j iij
iij
ij
ZX
ZXP
1
''
''
)exp(
)exp(
73
options the probability that farmers i selecting the option j type small-scale farm. The
CL model generates results for a conditional indirect utility function of the form:
(4.17)
where is the alternative specific constant (ASC), that captures the effects in utility
from any attributes not included in choice specific attributes (Rolfe et al., 2000). The
number of small-scale farm attributes considered is m and the number of social and
economic characteristics of the farm family employed to explain the choice of the small-
scale farm is . The vectors of coefficients are attached to the vector of attributes
and to a vector of socio-economic factors that influence utility, respectively.
The CE method is consistent with utility maximisation and demand theory (Bateman et
al., 2003). When parameter estimates are obtained, welfare measures can be estimated
from the CL model using the following formula:
(4.18)
where is the compensating surplus welfare measure, is the marginal utility of
income (generally represented by the coefficient of the monetary attribute in the CE) and
and represent indirect utility functions of alternative i (with subscript 0 indicating
the base situation and 1 indicate the changed situation) before and after the change under
consideration. For the linear utility index the marginal value of change within a single
attribute can be represented as a ratio of coefficients, reducing Equation 4.18 to 4.19:
kkmmij ZZZXXXT ............... 22112211
k )(X
)(Z
CS
0iT 1iT
)exp(ln)exp(ln 01
i
i
i
i TT
CS
74
(4.19)
Equation 4.19, the implicit prices (W) for the various small-scale farm attribute can be
calculated. These demonstrate the marginal rate of substitution between cost and the
attribute in question. This is the same as the marginal welfare measure (WTP or WTA)
for a change in any of the attributes.
An alternative model specification to the CL model is random parameter logit (RPL)
model which is increasingly becoming popular in CE studies. The advantage of RPL
model is that it accounts for consumers’ taste heterogeneities and also relaxes the
Independence of Irrelevant Alternatives (IIA) assumption of the CL model. It also
provides a flexible and computationally practical econometric method for any discrete
choice model derived from random utility maximisation (McFadden and Train, 2000).
More importantly preferences are in fact heterogeneous and accounting for this
heterogeneity enables estimation of unbiased estimates of individual preferences and
enhances the accuracy and reliability of estimates of parameters of the model and total
welfare (Greene, 1997). Furthermore, accounting for heterogeneity enables prescription
of policies that take equity concerns into account. This is because an understanding of
who will be affected by a policy change in addition to understanding the aggregate
economic value associated with such changes is necessary (Boxall and Adamowicz,
2002). Formally, the random utility function in the RPL model is given by:
(4.20) )]),([ iijij ZXUU
iablemonetary
attributeWvar_
75
As with the CL model, indirect utility is assumed to be a function of the choice attributes
(Xj), with parameters β, which, due to preference heterogeneity, may vary across
respondents by a random component µ, and by the social, economic and attitudinal
characteristics (Zi), namely income, education, household size and farmers’ attitudes to
agricultural biodiversity. By accounting for unobserved heterogeneity, Equation 4.16
now becomes:
(4.21)
Since this model is not restricted by the IIA assumption, the stochastic part of utility
may be correlated among alternatives and across the sequence of choices via the
common influence of µi. Treating preference parameters as random variables requires
estimation by simulated maximum likelihood (Kikulwe et al., 2011). In general, the
maximum likelihood algorithm searches for a solution by simulating n draws from
distributions with given means and standard deviations. Probabilities are calculated by
integrating the joint simulated distribution. Recent applications of the RPL model have
shown that this model is superior to the CL model in terms of overall fit and welfare
estimates (Breffle and Morey, 2000; Layton and Brown, 2000; Carlsson et al., 2003;
Kontoleon, 2003; Lusk et al., 2003; Morey and Rossmann, 2003).
Even if unobserved heterogeneity can be accounted for in the RPL model, the model
fails to explain the sources of heterogeneity (Boxall and Adamowicz, 2002). This can be
done by including interactions of respondent-specific social, economic and attitudinal
J
j iiij
iiij
ij
ZX
ZXP
1
''
''
])(exp[
])(exp[
76
characteristics with choice specific attributes and/or with ASC in the utility function.
This enables the RPL model to pick up preference variation in terms of both
unconditional taste heterogeneity (random heterogeneity) and individual characteristics
(conditional heterogeneity), and hence improve model fit (e.g. Revelt and Train, 1998;
Morey and Rossmann, 2003; Kontoleon, 2003). In the context of empirical application
of choice experiment model, choice experiment design as well as model selection steps
are extremely important. Therefore, the next section discusses basic steps of choice
experiment design and selecting the appropriate model for econometric estimation.
4.5 Choice experiment design and model selection
A choice experiment is a highly structured method of data generation, relying on
carefully designed tasks (experiment) to reveal the factors that influence choices (Hanley
et al., 1998). Experimental design theory is used to construct profiles of the
environmental good in terms of its attributes and levels of these attributes. Profiles are
assembled in choice sets, which are in turn presented to the respondents, who are asked
to state their preferences14
.
In the CE method, respondents are presented with panels of choices with two or more
alternatives each, where each alternative is a bundle of attributes which are specified at
different levels in each alternative (Louviere et al., 2000). The inclusion of a price or
cost attributes permits estimating the effect of cost on the respondents’ choice. For
example a farmer may choose from a number of different farm scenarios in her choice
14
For a detailed explanation of choice experiment design techniques, please see Louviere et al. (2000);
Bennet and Blamey (2001);Bateman et al. (2002) and Hensher, et al. (2005),
77
set, each of which exhibits variation in an array of attributes such as crops diversity,
livestock diversity, mixed farming system, landrace cultivation and organic production.
A farmer chooses the type of farm in a given season depending on the balance of
preferences for different attributes and the degree to which they are represented at a
given farm. In a survey context, the researcher should identify the essential attributes
and levels of the environmental goods in question and designs the choice question to
reveal the structure of the respondents’ preferences (Bateman et al., 2002).
Adamowicz et al. (1999) provided several stages of designing a CE study. They are as
follows:
1. Identification of relevant attributes
2. Selection of measurement unit for each attribute
3. Specification of the number and magnitude of the attribute levels
4. Experimental design
5. Model estimation
6. Use of parameters to simulate choice
The first three steps are involved in developing a concise and sufficiently complete
representation of the valuation scenario which will provide the survey respondent with
an appropriate information set on which to base statements of preference. This phase
uses information obtained from secondary sources, experts in the field, focus groups and
personal interviews in order to refine the informational content of the survey instrument.
The selection of attributes in relation to the choices of interest is very important in
framing a CEexercise. According to Blamey et al. (2000) attribute selection needs to
78
take place from both the perspectives of the end-user (the population of interest) and the
decision-makers/resource managers to ensure that the attributes are not only easily
identifiable, but produce policy-relevant information.
Another goal of the attribute selection process is to minimise the number of attributes as
the use of a large number of attributes is likely to lead to lower data reliability due to the
excessive cognitive burden it would place on respondents (Mogas et al., 2002).
Identification of appropriate attribute ranges is another basic framing task in choice
experiment, as a failure to accept trade-offs indicates that the range of attribute levels
offered is not salient (Johnson et al., 2000). In determining how many attributes to
include in a study design, there is often a trade-off between describing tradeoffs
accurately (requiring more attributes) and minimising choice and experimental design
complexity (requiring fewer attributes). Louviere et al. (1993) claim to have successfully
administered surveys with up to 32 choice tasks, though this requires scaling down the
number of alternatives and attribute levels. Boxall et al. (2002) suggests that respondents
can endure large numbers of choice sets but sets with more than six alternatives tend to
exceed cognitive limits. Louviere et al. (1993) suggest that the average choice
experiment survey employs seven attributes, four choice sets and four alternatives per
set, though they note that there is a great deal of variability and this average does not
constitute a best practice.
79
Experimental design15
is the next important aspect of choice modelling and it is
concerned with how to create the choice sets in an efficient way or how to combine
attribute levels into profiles of alternatives and profiles into choice sets. In practice, a
design is developed in two steps: (i) obtaining the optimal combinations of attributes and
attribute levels to be included in the experiment and (ii) combining those profiles into
choice sets. A starting point is a full factorial design, which is a design that contains all
possible combinations of the attribute levels that characterise the different alternatives.
A full factorial design is, in general, very large and not tractable in a choice experiment
(Louviere et al., 2000). Therefore, we need to choose a subset of all possible
combinations, while following some criteria for optimality and then construct the choice
sets. The standard approach used in most research has been to use orthogonal designs,
where the variations of the attributes of the alternatives are uncorrelated in all choice
sets. More recently researchers in marketing have developed design techniques based on
the Doptimal criteria for non-linear models in a choice experiment context. However,
there can be some problems with these more advanced design strategies due to their
complexity, and it is not clear whether the advantages of being more statistically
efficient outweigh the problems (Scarpa and Rose, 2008)16
.
The next step of choice experiment involves econometric model selection and
estimation. The most common model estimated in economics literature has been the
MNL model, and the most common estimation criterion is maximum likelihood. The
15
This step is much more complex in choice experiments in that the experimental design is critical to
producing a dataset that will yield estimable parameters for the attributes in an econometric model of
preferences. 16
For example, utility balance in more advanced design makes the choice harder for the respondents, since
they have to choose from alternatives that are very close in terms of utility.
80
MNL model is easy to estimate, and interpretation is straightforward. However, there are
also examples of other choice model specifications such as the CL model and RPL
model. Selection between the MNL and CL depends on whether the researcher is
interested in including socioeconomics variables in addition to the choice attribute into
the model. If researcher uses only choice attributes, the MNL model can give higher
accuracy of the model fits. However, if the researcher uses choice attributes as well as
socioeconomic variables in the model, the CL model provides more accurate results
(Rolfe et al., 2000). In empirical settings, inclusion of social and economic
characteristics is also beneficial in avoiding IIA violations, since social and economic
characteristics relevant to preferences of the respondents can increase the systematic
component of utility while decreasing the random error (Rolfe et al., 2000; Bateman et
al., 2003).
The MNL model relies on the assumption of the independence of irrelevant
alternatives17
. The IIA arises from the assumption about the IID of the error term. IID of
error term means that it has an extreme value error distribution. The IIA means that the
probability of choosing an alternative is dependent only on the options from which a
choice is made, and not on any other options that may exist. If the IIA/IID is violated,
the estimates derived from the model could be biased and not generate accurate values
for inclusion in cost benefit analysis (Ben-Akiva and Lerman, 1985). The IIA property
allows the addition or removal of an alternative from the choice set without affecting the
structure or parameters of the model. This assumption has three main advantages.
17
The independence of irrelevant alternatives means that, all else being equal, a person’s choice between
two alternative outcomes is unaffected by what other choices are available.
81
Firstly, the model can be estimated and applied in cases where different members of the
population face different sets of alternatives. For example, in the case of the farm choice
model, households living in one area may not have one component of agricultural
biodiversity. Secondly, this property simplifies the estimation of the parameters in the
MNL and CL models. Third, this property is advantageous when applying a model to the
prediction of choice probabilities for a new alternative. On the other hand, the IIA
property may not properly reflect the behavioral relationships among groups of
alternatives (Hensher et al., 2005). That is, other alternatives may not be irrelevant to the
ratio of probabilities between a pair of alternatives. In some cases, this will result in
erroneous predictions of choice probabilities.
There are various reasons why IIA/IID violation could occur. One possibility is the
existence of random taste variations (that is heterogeneity). To account for this, a model
which includes socioeconomic variables in addition to the attributes in the choice sets
can be estimated (Bennett and Blamey, 2001). The socio-economic information could be
included in two different ways. The first is by interactions with the attributes in the
choice sets. The second method includes the socio-economic information through
interactions with the alternative specific constants. These interactions show the effect of
various socio-economic characteristics on the probability that a respondent will choose
particular options.
Alternative model specifications to the MNL models are the CL and RPL. The CL
model allows us to estimate the effect of choice-specific variables on the probability of
choosing a particular alternative. The CL model also assumes the IIA property, which
82
states that the relative probabilities of two options being chosen are unaffected by
introduction or removal of other alternatives. In other words, the probability of a
particular alternative being chosen is independent of other alternatives. If the IIA
property is violated then the CL model results will be biased and hence a discrete choice
model that does not require the IIA property, such as the RPL model, should be used. To
test whether the CL model is appropriate, the Hausman and McFadden (1984) test for
the IIA property can be employed. In this case, whether or not IIA property holds can be
tested by dropping an alternative from the choice set and comparing parameter vectors
for significant differences. A RPL model is a generalisation of a standard multinomial
logit model. The advantages of a RPL model are that (i) the alternatives are not
independent (the model does not exhibit the independence of irrelevant alternatives
property) and (ii) there is an explicit account for unobserved heterogeneity.
In this study we followed all these steps in order to increase the accuracy as well as
reliability of the results of this study. We carefully designed the CE survey and used
appropriate econometric techniques for the analysis. The way of approaching each step
of the choice experiment study is explained in the next section.
4.6 Empirical approach to choice experiments study
As discussed in the previous section, a starting point of CE study involves studying the
attributes and attribute levels used in previous studies and their importance in the choice
decisions (Green and Srinivasan, 1990). The selection of attributes should be guided by
the attributes that are expected to affect respondents' choices on agricultural biodiversity,
83
as well as those attributes that are policy relevant in this field. Information obtained from
previous studies was used as the base for selecting the attributes and relevant attribute
levels to include in the first round of focus group discussion in this study18
. The focus
group discussion can provide information about credible minimum and maximum
attribute levels. It was found that crop diversity, mixed farming systems, organic
production and landrace cultivation were the most important attributes of agricultural
biodiversity used in previous studies. In addition to that it is necessary to include a
monetary attribute for calculating welfare measures (Rolfe et al., 2000).
In this study first we attempted to define the biodiversity rich farms in terms of their
attributes and the levels of these attributes in study areas. The most important attributes
and their levels were identified in consultation with experts from the Ministry of
Environment in Sri Lanka, drawing on the results of informal interviews and workshops
with traditional small-scale farmers in the study sites, focus group discussions and a
thorough review of previous research in this area in the country. The chosen small-scale
farm attributes used in this study are reported in Table 4.1. The attributes shown in Table
4.1 were found to be of the most interest to both potential respondents and agricultural
officers in traditional agricultural districts in Sri Lanka.
18
The task in a focus group is to determine the number of attributes and attribute levels, and the actual
values of the attributes. Attributes are identified from prior experience, secondary research and/or primary,
exploratory research. It is also important to identify any possible interaction effect between the attributes.
84
This monetary attribute is specified in terms of required additional labour days is
included in order to estimate welfare changes19
. The monetary attribute in this CE is a
proxy, measuring the labour costs that farmers have to allocate for receiving the benefits
of agricultural biodiversity. This attribute represents WTA compensation which is
measured as a benefit rather than a cost. It is clear that the first five attributes reflect the
various attributes of agricultural biodiversity found in the farms in Sri Lanka. The sixth
factor represents benefits that farmers can receive in terms of reducing family food
expenditure under different policy options. The last factor is the monetary attribute in
terms of additional labour costs that farmers have to use under different policy options.
19
This indirect measure is preferred over a direct monetary attribute because most (if not all) of the
outputs and functions of farms that result in agricultural biodiversity are not traded in the markets, but
consumed by the farm families themselves. Hence, they are not likely to be familiar with a direct
monetary measure. The proxy monetary attribute can easily be converted into actual monetary units by
using secondary data on labour costs.
85
Table 4.1: Classifications of small-scale farm attributes in the CE survey
Farm attributes Definitions
Crop species diversity The total number of crops that are grown in the farm
Livestock diversity The total number of animal species on the farm
Mixed farming system Mixed crop and livestock production, representing diversity in
agricultural management system
Landrace cultivation Whether or not the farm contains a crop variety that has been
passed down from the previous generation and/or has not been
purchased from a commercial seed supplier.
Organic production Whether or not industrially produced and marketed chemical
inputs are applied in farm production
Expenditure Own farms’ contribution to reduce family food expenditure
Estimated labour cost Estimated cost in terms of additional labour requirement
Notes: These attributes are common to most agricultural districts in Sri Lanka. However, the importance
of different attributes can be different in different areas. More details of all attributes are given under
section 3.5 in Chapter three.
After identifying the attributes for a particular experiment, the analyst must assign
values or levels to each attribute. These levels should be chosen to represent the relevant
range of variation in the present or future interest of respondents. In general, focus group
discussions will indicate the level of the attributes as well as the best way to present
them. Though commonly presented in words and numbers, attribute levels may be
presented using pictures. To the extent that visual representations of attribute levels are
utilised, it is likely that respondents will perceive levels more homogeneously, likely
leading to more precise parameter estimates in the modelling stage (Alpizar et al., 2001).
86
We presented choice set using pictures of the different attributes and their levels. A
sample choice set is given in Appendix I.1. In this study crop species diversity is
explained as different levels, while mixed farming system, landrace cultivation and
organic production variable are given as binary variables. The animal diversity variable
was dropped from the choice set as including this variable could provide choice sets
which cannot be interpreted. This is because it is directly linked with mixed farming
systems. The levels of relevant variables were identified through the pilot survey that
conducted in August 2010. Attribute levels used in this study are given in Table 4.2.
As credibility plays a crucial role in choice modelling, the researcher must ensure that
the attributes selected and their levels can be combined in a credible manner (Layton and
Brown, 1998; Alpizar et al., 2001). Therefore, experimental design, where different
types of hypothetical farms are created plays an important role in choice modelling. A
large number of different types of farms (combinations of attributes) could be
constructed from this number of attributes and levels. The number of farms that can be
generated from six attributes, 1 with 4 levels, 2 with 3 levels and remaining 3 with 2
levels is 288. This means that it would be possible to generate 41*3
2*2
3=288 alternatives
from these, simply by considering all the possible combinations or complete factorial
design. Clearly it would not be practical to ask respondents to consider simultaneously
288 possible alternatives. It is not necessary to do so. The answer lies in the use of
statistical experimental designs. Therefore, the fractional factorial design is used to
create an optimal number of choice options for the survey. In our case the minimum
number of choice options which could be used in the survey was 12. However 36 choice
options were used and randomly blocked them into six different versions (each has six
87
options). The 36 choice options are given in Appendix I.2. Using the Dptimal procedure
in Engine an experimental design was undertaken to recover only the main effects,
consisting of 36 pair wise comparisons of farms profiles.
Table 4.2: Attributes and their levels
Attributes Levels
Crop species diversity 3, 7,10 and 15
Mixed farming system Mixed crop and livestock production vs.
specialized crop or livestock production (If Yes
they maintain mixed crop and livestock
otherwise No)
Organic production Organic production vs non-organic production
(If Yes organic production, otherwise No)
Landrace cultivation Whether farm contains landraces or not
(If Yes farm contains landraces otherwise No)
Decrease in food expenditure
(in percentage)
5 %, 10% and 15%
Estimated cost in terms of
additional labour days
10%, 20% and 30%
Note: i. Upper and lower bound of the crop species diversity, food expenditure change and additional
labour requirements are estimated using pilot survey information.
ii. These attributes are common to most agricultural districts in Sri Lanka. However, it is important
to note that attributes can be different in different areas.
The questionnaire is usually a paper and pencil task that is presented through an
interviewer. While its main content will be six choice scenarios through which the
respondent will be guided, it may also include sections requesting socio-demographic,
economics, attitudinal and past behaviour data. The questionnaire used for this study
was developed using the results from nine focus groups’ discussions and a pre-test. A
88
pre-test of 30 respondents was undertaken in August 2010 in three districts. On the basis
of the pre-test, only minor modifications to the questionnaire were required. In the
questionnaire, respondents were told that the development of the choice experiment
questionnaire was based on focus group studies. Nine focus groups discussion were
conducted for both potential respondents (6) and agricultural officers (3) to ensure that
inputs for choice sets were correctly specified. The purposes of the focus group studies
were to determine attributes relevant to respondents and agricultural managers and test a
draft questionnaire. More details about selecting sample size and the content of the
questionnaire are provided in Section 3.2 in Chapter three.
Before the interview it was confirmed whether the respondents were generally those
responsible for farm production decision making. An introductory section explained to
the respondents the context in which choices were to be made, described each attribute
and explained that the key attributes of farms had been selected as a result of prior
research and were combined artificially in the choice sets. Respondents were told that
their names and individual choices were confidential and that completion of the exercise
would provide information to agricultural policy makers in summary form. In face-to-
face interviews, each respondent was presented with six choice sets showing various
options for the different farms, the one of which was an example given in Table
4.3.Respondents were told that three sets of possible options had been prepared and were
then asked for their preferred choice from each set of options. Before answering the
choice sets, respondents were requested to keep in mind their available income, food
consumption expenditure, available labour, size of the land and other things on which
they may consider when making a decision. They were also reminded that different
89
types of farms may have cost and benefits for them in the future. There was not any item
non-response, in other words all the choice sets were answered due to the advantage of
the in person interviewing. Therefore, a total of 4,488 (748*6) choices could be elicited
from a total of 746 farm families.
Table 4.3 shows an example of a choice set used for a choice experiment. In the survey,
the enumerator asked: “Assuming that the following farms were the ONLY choices you
have, which one would you prefer to cultivate?” Each choice set consists of two
different profiles and one common profile. We presented different options to the
respondents six times and asked them to select only one option each time. Neither is a
“status quo” alternative and it is common to all choice sets. The sample population in
each area was randomly divided into six, each sub-sample receiving one of the six
versions of the choice experiment.
Table 4.3: An example of a choice set
Farm Characteristics Farm
(A)
Farm
(B)
Total number of crop varieties grown on a farm 10 7 Neither
Small-scale
farm (A) nor
Small-scale
farm (B):
Crops is combined with livestock/poultry production Yes No
Farm crops produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No No
Expenditure reduction (in percentage) 15% 10%
Estimated cost in terms of additional labour
requirement (in percentage)
20% 10%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
90
In addition to the main attributes, it is required to obtain some socio-economic and
household characteristics which can be used as interaction term for the estimation of the
CL and RPL models. Some of the individual specific attributes that can be used in the
estimation of the CL and RPL models are reported in Table 4.4. These types of
characteristics have been commonly used by previous studies in choice experiment.
More details about including socio-economic and household characteristics are given by
Birol (2005) and Rolf et al. (2000).
Table 4.4: Individual attributes for the estimation of CL and RPL models
Individual attributes Definitions
Age Age of the small-scale farm decision maker
Family Size Total number of family members in the farm family
Farm ownership Type of ownership of the land
Business vehicle Household owns a business vehicle or not
Experience Experience of small-scale farm decision maker in years
Off farm employment Number of family members employed off-farm
Education level Education of the small-scale farm decision maker
Attitudes Farmers’ attitudes towards to agricultural biodiversity
Number of plots Numbers of plots own to farmer
Note: These attributes were selected to be used in the CL model as interaction terms with the main
attributes. After a preliminary run of the model, the most important five variables were selected in the
final model.
Under the CE method a sample of people is asked to choose their most preferred
alternatives from a sequence of grouped options that relate to different agricultural
91
biodiversity management strategies. Each option is described in terms of its agricultural
biodiversity outcomes and a personal monetary cost to be borne personally by the
respondent. By analysing the choices made by respondents it is possible to infer the
tradeoffs that people are willing to make between money and greater benefits of
agricultural biodiversity. This in turn allows the estimation of changes of private benefits
with changing agricultural biodiversity.
Socio economic aspects such as community, gender, age, marital status, literacy level,
income, expenditure, savings and indebtedness provide a base for studying the impact of
any program. Therefore, before estimating the models it is important to know the basic
socio-economic profile of the respondents in each district. The most important socio-
economic variables are explained in the next section with reference to the respondents
and their families.
4.7 Socio-economic profile of sample respondents
In Appendix J.1, J.2 and J.3 some descriptive statistics of the respondents are presented.
The mean value of age was slightly higher in Anuradhapura samples and men displayed
a higher response-rate in all three districts. The average number of persons in the
household was 5, 4 and 5 in Ampara, Anuradhapura and Kurunegala samples
respectively. Although agriculture was the dominant source of household income,
monthly income from non-farm activities was approximately Rs. 1,750, Rs. 2,300 and
Rs. 2,200 per household, which accounted for almost 7, 8 and 10 per cent of the total
household income in Ampara, Anuradhapura and Kurunegala samples respectively. The
92
mean labour usage per season was 96 man-days for the three samples. This is expected,
given the tedious labour intensity for all agricultural work in semi-subsistence economy.
There was low usage of external input (Rs. 2,110 in capital) as a result of the small size
of farms in the study areas.
Rice was cultivated by the most number of households (532), followed by various types
of vegetables and cash crops. The maximum number of crop varieties cultivated by any
household was nine. Percentage of households that have cultivated between one and nine
crop varieties is as follows. Approximately 14 per cent cultivated one variety only, while
18, 22, 31 and 15 per cent cultivated two, three, four and more than five varieties,
respectively in all districts. Only 54, 46 and 72 households used a modern variety of
seeds in Ampara, Anuradhapura and Kurunegala samples respectively. Approximately
68, 72 and 66 per cent of households in respective samples used mixed farming systems
where they have crops as well as livestock.
The average number of years of education is 8, 9 and 8 in Ampara, Anuradhapura and
Kurunegala samples respectively. In the survey it was found that a few farmers have not
attended any schools. Any interviewee whose education level is less than three years
was not included into the choice experiment study. This is to make sure that everyone
could understand the trade-offs between different alternatives. Approximately 1 per cent
of the respondents had a diploma or degree and 26 per cent of the interviewees have
passed the ordinary level examination. Such relatively high education levels may be
attributed to the reliable results of the choice experiment part of this study.
93
About 22 per cent of the respondents were within the range 18-30 years old. The most
frequent age class was 30-55 years (65 per cent). Around 23 per cent and 6.2 per cent of
the cases fell within the age-ranges 42-55 and more than 55 years, respectively. The
average age of respondents was approximately 41 years for the three samples. The main
income source of families was agricultural income. Average monthly agricultural
income was Rs. 22,844, Rs. 26,109 and Rs. 29,074 for Ampara, Anuradhapura and
Kurunegala samples respectively. The majority of household expenditure was spent on
food, followed by health and personal care, and transport. Approximately 51, 62 and 39
per cent of the farm outputs were used for family consumption in Ampara,
Anuradhapura and Kurunegala samples respectively.
We compared the above mentioned sample averages with district averages for small-
scale farmers which were provided by the Department of Census and Statistics in Sri
Lanka. In most of the cases, sample averages are similar to the population averages for
these districts and hence the results reported in this chapter of the thesis could be
generalized for the entire population of these districts. Given this general information
about the respondents, next, the results of this analysis should be investigated. Before
explaining the results, it is important to know the way of coding the data and the
estimating procedure in this analysis. Therefore, data coding is explained in the next
section.
94
4.8 Data coding and estimation procedure
Data coding is one of the important parts of the choice experiment model. In this study
the data were coded according to the levels of the attributes. Attributes with 2 levels
entered the utility function as binary variables that were effects coded (Louviere et al.
2000). Crops diversity variable is used as a continuous variable. Consequently the crop
species diversity attribute took levels 3, 7, 10 and 15. For the mixed farming system,
landrace cultivation and organic farming method were coded as effect coding method.
For example, if a farm family selected the mixed farming system, it was entered as 1 and
if they selected specialised crop or livestock production, it was entered as –1. For the
organic production attribute, organically produced farms were entered as 1 and those
farms that were not produced organically were entered as –1. For the landrace attribute,
those farms that contained a landrace were entered as 1 and those without were effects
coded as –1. Farm contribution to family expenditure reduction and labour requirement
are transformed into monetary values when estimating the models.
The percentage values of the levels given to farmers of possible family food expenditure
reduction due to own farm consumption is 5, 10 and 15. On average farm families spend
approximately Rs. 12,00020
for their monthly food consumption. Accordingly, the value
of net expenditure reduction due to own farm food consumption can be represented as
Rs. 600, Rs. 1,200 and Rs. 1,800 for the three levels respectively. The percentage values
of additional labour requirement were given as 10, 20 and 30. On average, farm families
need approximately 16 labour days per month for their farm cultivation. Daily average
20
Exchange rate at the time of survey was 1 US$ = LKR 115(approximately).
95
wage rate per person per day was Rs.45021
. Accordingly, value of the cost of accepting
alternative farms can be represented as Rs. 720, Rs. 1,440 and Rs. 2,160 for the three
levels respectively. In this way the levels used for expenditure reduction and labour
requirement variables were entered in a cardinal-linear form.
The attributes for the ‘neither farm’ option were coded with zero values for all attributes.
The alternative specific constants were equalled to 1 when either farm A or B was
chosen and to 0 when ‘neither farm’ alternative was chosen. In other words, in this
model the ASC is specified to account for the proportion of choice of participation in
small-scale farm production. Choice data were converted from wide to long format with
a program coded in LIMDEP 9.0 NLOGIT 4.0. This data conversion step was necessary
to estimate models with multiple responses from each respondent, a format similar to
panel data.
First, we estimated the CLM. The IIA property of this model is tested using a procedure
suggested by Hausman and McFadden (1984). This test involves constructing a
likelihood ratio test around different versions of the model where choice alternatives are
excluded. If IIA holds then the model estimated on all choices (the full choice set)
should be the same as that estimated for a sub-set of alternatives (Bateman et al., 2003).
It is found that the IIA conditions are not violated any of the case. Therefore, the IIA
tests performed indicate that the model fully conform to the underlying IIA conditions.
Then social and economic characteristics were we included as interaction terms and test
21
This varies between Rs. 500 and Rs. 400 depending on various factors (gender, period and area). For
example, men’s wage rate is slightly higher than female. Wage rate in the harvesting period is greater than
other period.
96
whether there was an improvement of the results. It was found that there was no
improvement by including any social-economic characteristics as the interaction term.
As the next step of the analysis, the RPL model was used in order to take into account
preference heterogeneity. We estimated basic the RPL model which includes only
attributes as well as the extended RPL model that includes some socio-economic
variables. When comparing with the RPL results with the CL results it was found that
basic CL results were better in term of overall fit of the model and number of significant
variables. Therefore, the results of the basic CL model could be used to simulate welfare
change of the society when changing different attributes and their level of agricultural
biodiversity. The results of the CL model are discussed in the next section.
4.9 Results of the CL model
It is well known that the choice experiment is designed with the assumption that the
observable utility function would follow a strictly additive form. This study explored a
variety of different specifications of the utility functions to identify the best specification
of the data. These tests include both formal statistical tests and informal judgments about
the signs, magnitudes, or relative magnitudes of parameters based on our knowledge
about the underlying behavioral relationships that influence different farms choice.
Different researchers have different styles and approaches to the model development
process. One of the most common approaches is to start with a minimal specification
which includes those variables that are considered essential to any reasonable model
(Hensher et al., 2005). Working from this minimal specification, incremental changes
97
are proposed and tested in an effort to improve the model in terms of its empirical fit to
the data while avoiding excessive complexity of the model. Another common approach
is to start with a richer specification which represents the model developer’s judgment
about the set of variables that is likely to be included in the final model specification.
The first of these methods were adopted in this study for the specification of a model
choice as it was the most appropriate approach for those new to discrete choice
modelling. As a formal statistical process, different model specifications were compared
according to higher log-likelihood value criterion in this study. Most appropriate
specification was found to be the model with the linear version of the six attributes of
the study. Accordingly, the CL model was specified so that the probability of selecting a
particular alternative is a function of attributes of the alternatives and of the alternative
specific constant. Indirect utility received by the farm attributes take the form:
(4.22)
where β0 refers to the alternative specific constant and β1-6 refers to the vector of
coefficients associated with the vector of attributes describing farms characteristics. The
results of the estimated basic CL model for the separate district and pool data set are
presented in Table 4.5. All attributes in the model were statistically significant at
conventional levels, and their signs were as expected. The overall fit of the model as
measured by McFadden’s R2 was also good by conventional standards used to describe
)()()()( 4_3_2_10 landracefarmorganicfarmmixdiversitycropij XXXXT
)6exp5 ()( labourenditure XX
98
probabilistic discrete choice models (Ben-Akiva and Lerman, 1985). The results indicate
that the indirect utility function takes the following form:
(4.23)
(4.24)
(4.25)
(4.26)
As shown in Table 4.5 we estimated models for three samples separately. In addition to
that pool data model was estimated. This type of estimation allowed us to compare
relative values of attributes in different regime. It also helped in understanding the
heterogeneity of the results among different districts.
All of the farm attributes are statistically significant at 10 per cent level implying that
any single attribute increases the probability that a farm is selected, other things
remaining equal. Since the underlying sample is statistically significant, these
)(090.0)(096.0)(119.0)(028.0832.1 ___)( landracefarmorganicfarmmixdiversitycropAmparaij XXXXT
)exp (00045.0)(00023.0 labourenditure XX
)(112.0)(077.0)(095.0)(019.0028.5 ___)( landracefarmorganicfarmmixdiversitycropraAnuradhapuij XXXXT
)(00024.0)(00018.0 exp labourenditure XX
)(243.0)(064.0)(092.0)(018.0984.2 ___)( landracefarmorganicfarmmixdiversitycropKurunagalaij XXXXT
)exp (00069.0)(00034.0 labourenditure XX
)(144.0)(079.0)(077.0)(021.0711.2 ___)_( landracefarmorganicfarmmixdiversitycropdataPoolij XXXXT
)exp (00045.0)(00025.0 labourenditure XX
99
Table 4.5: Regression results of the CL model for separate districts and pool data
Variables Ampara Anuradhapura Kurunegala Pool data
ASC 1.832(0.199)* 5.028(0.445)* 2.984(0.219)* 2.711(0.123)*
Crop diversity 0.028(0.009)* 0.019(0.008)** 0.018(0.009)** 0.021(0.005)*
Mixed system 0.119(0.041)* 0.096(0.041)** 0.092(0.042)** 0.077(0.021)*
Organic farms 0.096(0.041)** 0.077(0.040)*** 0.064(0.041)*** 0.079(0.023)*
Landrace cultivation 0.090(0.041)** 0.112(0.042)* 0.243(0.043)* 0.145(0.024)*
Expenditure reduction 2.3E-04(8.4E-05)* 1.8E-04(8.4E-05)* 3.4E-04(8.9E-05)* 2.5E-04(4.9E-05)*
Labour requirement -4.5E-04(6.9E-05)* -2.4E-04(6.8E-05)* -6.9E-04(7.2E-05)* -4.5E-04(4.0E-05)*
LR chi2(7) 648.47 1338.59 989.37 2786.68
Prob > chi2 0.000 0.000 0.000 0.000
Pseudo R2 0.127 0.264 0.199 0.183
N 4,032 4,032 3,942 12,006
Note: i. Standard errors are shown in brackets.
ii. *denotes significant at 1% level while ** and *** indicates significant variables at 5% and 10% level respectively.
100
parameters represent preference estimates of farm families for farms attributes in 3
environmentally different areas of Sri Lanka. In the Ampara district organic farm and
landrace cultivation variables are significant under five per cent level while all other
attributes are significant at one per cent level. In Anuradhapura district crop diversity and
mixed farm variables are significant under five per cent level while all other attributes
except the organic farms attribute are significant at one per cent level. This is similar to the
results of Kurunegala sample. However, the organic farm attribute of Anuradhapura and
Kurunegala districts are significant at 10 per cent level. Interestingly, all variables in the
pool data model are significant at one per cent level. When the additional labour
requirement attribute is used as the normalising variable, it can be seen that the almost all
attributes are significantly contributing towards the welfare in rural agricultural society in
Sri Lanka. The positive sign on the ASC coefficient implies that respondents are highly
responsive to changes in existing farms attributes level and they make decisions that are
closer both to rational choice theory and the behaviour observed in reality (Hensher et al.,
2005).
Investigation of the results in each regime reveals that the findings of the study are
strikingly in line with those as predicted by economic theory. It is obvious that regions
where food markets as well as road infrastructure are fully developed, farmers’ demand for
crop species diversity and mixed farming is highly significant and organic farm and
landraces are relatively less significant. In contrast to that, in the relatively isolated region
where community level markets are lacking and distance to the nearest towns are large,
organic farming method and landrace cultivation are significantly and positively demanded
by the farmers. However, in contrast to our findings, Birol (2004) found that farmers’
101
demand for crop species diversity in home gardens was positive and significant in rural
isolated areas, more so than in areas where market as well as transport facilities were easily
accessible.
The overall fit of all models can be measured by Pseudo R2and it is reasonable when
considering probabilistic discrete choice models (Hensher et al., 2005). We used Swait-
Louviere log likelihood ratio test in order to test whether there is a significant regional
heterogeneity of the farm families’ utility for different attributes. The rejection of the null-
hypothesis would imply that farmers in different districts have different preferences for
farms and their attributes. It is found that Swait-Louviere log likelihood ratio test rejects the
null hypothesis that the regression parameters are equal at five per cent significance level.
This implies that farm families in each of the three regions have distinct preferences for
different farms and their attributes.
As the next step of the analysis, the IIA property of all models is tested using a procedure
suggested by Hausman and McFadden (1984) and contained within NLOGIT 4.0. This test
involves constructing a likelihood ratio test around different versions of the model where
choice alternatives are excluded22
. If IIA holds then the model estimated on all choices (the
full choice set) should be the same as that estimated for a sub-set of alternatives (Bateman
et al. 2003). It was found that the IIA property is not violated implying that the conditional
logit estimates do not hold any bias that could have resulted from inclusion of the ‘neither’
option. The test results are reported in Table 4.6 for all versions including pooled model
22
It is evident that maximum likelihood of conditional logit is consistent and efficient if the model is correctly
specified. A consistent but inefficient estimator is obtained by estimating the model on a restricted set of
outcomes. If IIA holds and the dropped choices are irrelevant, the estimates of the parameters should be the
same.
102
without the constant. The results of Hausman-McFadden test reported in Table 4.6 strongly
provide the evidence of holding IIA assumption for each sample in our data set. However,
as mentioned previously, CL model assumes homogeneous preferences across farm
families in each district.
Table 4.6: Test of independence of irrelevance alternatives
Ampara χ2 D.O.F Probability
Scenario A 66.35 6 0.000
Scenario B 84.52 6 0.000
Scenario C 65.49 6 0.000
Anuradhapura
Scenario A 38.47 6 0.000
Scenario B 113.64 6 0.000
Scenario C 278.01 6 0.000
Kurunegala
Scenario A 198.15 6 0.000
Scenario B 143.00 6 0.000
Scenario C 34.99 6 0.000
Pool data
Scenario A 128.21 6 0.000
Scenario B 254.79 6 0.000
Scenario C 479.29 6 0.000
Note: The Hausman-McFadden test is based on the comparison of two estimators of the same parameters.
One estimator is consistent and efficient if the null hypothesis is true (IIA holds), while the second estimator
is consistent but inefficient.
103
In general preferences across families are in fact heterogeneous. Accounting for this
heterogeneity enables estimation of unbiased estimates of individual preferences and
enhances the accuracy and reliability of parameter estimates and hence total welfare (Rolfe
et al., 2000; Bateman et al., 2003). Furthermore, accounting for heterogeneity enables
prescription of policies that take equity concerns into account (Birol, 2004).
There are two standard ways of accounting for preference heterogeneity. First, it can be
done by separating the respondents into various groups (segments) and estimating the basic
model for each group separately. Estimating the CL model for each district separately is
one way of doing this. Second, it is possible to accounting for preference heterogeneity by
using household and decision-maker level characteristics directly as interaction terms.
Interaction of individual-specific social and economic characteristics with choice specific
attributes or with ASC of the indirect utility function is a common solution to dealing with
the heterogeneity. However, the main problem with this method is multicollinearity, which
occurs when too many interactions are included in the estimation. In this context, the model
needs to be tested down, using the higher log-likelihood criteria (Bateman et al., 2003;
Birol, 2004). Therefore, as the next step of the analysis, CL model is estimated using five
socioeconomic variables as interaction terms.
4.10 Results of the CL model including attributes and socioeconomic variables
In order to account for heterogeneity of preferences across farm families, interactions of
household-specific socioeconomic characteristics with choice-specific attributes were
104
included in the utility function. The use of socioeconomic variables as independent
variables is justified under the hypothesis that socioeconomic characteristics are separate
factors influencing behavioural intentions and behaviour (Lynne et al., 1988; Rolfe et al.,
2000; Bateman et al., 2003). As discussed in section 4.2, in random utility models the
effects of social and economic characteristics on choice cannot be examined in isolation but
as interaction terms with choice attributes. It is not possible to include interactions between
many household specific characteristics and the six farm attributes when estimating the CL
models due to possible multicollinearity problems (Hensher et al., 2005). Therefore, only
five important household specific characteristics are selected. They are; age of the
respondent (age), whether farmer owned a farm or not (landownership), education level of
the respondent (education), household size (hhs) and number of family members who have
off farm employment (offfarm). Accordingly, indirect utility received by the farm attributes
and interaction with socioeconomic characteristics can be respecified as follows:
(4.27)
It is clear that in model 4.27 five socioeconomic variables are included in addition to the
attributes from the choice sets. The total number of coefficients in the full model is 36. We
tested various interactions of the six farm attributes with the household-level characteristics
mentioned above. An initial run of the model with all interaction terms reveal that a large
))(...)()(...)(
)(...)()(...)(
)(...)()(...)(
))())())(
3026exp25exp21
2016_15_11
_10_6_5_1
6exp54_3_2_10
offfarmlabouragelabourofffarmenditureageenditure
offfarmlandraceagelandraceofffarmfarmorganicagefarmorganic
offfarmfarmmixedagefarmmixedofffarmdiversitycropagediversitycrop
labourenditurelandracefarmorganicfarmmixdiversitycropij
ZXZXZXZX
ZXZXZXZX
ZXZXZXZX
XXXXXXT
105
number of variables are insignificant for all three models. Then we estimated the
correlation matrix and it was revealed that there was a higher level of correlation and
multicollinearity among these household-level variables. Estimation of variance inflation
factor further provided the evidence about higher correlation among household level
variables. To address this limitation, independent variables were eliminated based on
variance inflation factors, which were calculated by running ordinary least square
regressions between each independent variable23
. Then the results of the correlation matrix
were also used for further eliminating some of the interaction terms. The estimated results
of the final models are reported in Table 4.7.
This specification of the model was not significantly different from the previous
specification. In particular, the model did not reveal a higher level of parametric fit
compared with the first model. Most of the interaction terms of all three models are not
significant. Further, including the interaction terms has reduced the significance of some of
the attributes of the models. Therefore, it can be concluded that the improvement in model
fit was not significant. The Hausman-McFadden test also revealed that the CL model
without interactions is a better fit for the data than the CL model with interaction.
Among the significant interactions, households with higher ages in Anuradhapura and
Kurunegala had a higher preference for crop variety diversity. Higher age households in
Kurunegala district had higher preferences for mixed farming systems. Demand for a
23
Those independent variables for which VIFj > 5 indicate clear evidence that the estimation of the
characteristic is being affected by multicollinearity (Maddala, 2000).
106
landrace cultivation in the small-scale farm also increased with land ownership. This
implies that farmers who have their own land are likely to select traditional varieties for
their cultivation. Farmers who have the ownership of the land have higher probability of
using organic farming methods as well. More educated farmers were more likely to select
organic farming methods in Ampara sample. This implies that land ownership as well as
education has a significant impact on agricultural biodiversity in all regimes. As expected,
off farm employment has significant negative impact on biodiversity improvement in these
areas. Preferences of farm families for small-scale farms without land race cultivation may
reflect the effect of government subsidies for purchasing the seed of modern varieties on
agricultural biodiversity maintained in farms.
The interaction between the demand for crop varieties and the number of members in the
family is positive and highly significant in all models. We included interaction between
organic production and the number of members in off farm employment in the family to see
whether this variable provided good results in this analysis. However, this variable was
highly correlated with other variables. As a result this variable was removed from the final
version of the model. The demand for crop species diversity decreased with the number of
household members employed off farm. It was found that households with higher number
of members in the family were more likely to choose more mixed farming systems that
would provide more foods for household consumption. The overall model is significant at
the one per cent level. Compared to basic CL model, the explanatory power of the model
has not changed significantly.
107
Table 4.7: CL model including attributes and socioeconomic variables
Variables Ampara Anuradhapura Kurunegala
ASC 1.65(7.07)* 5.60(10.98)* 2.42(9.66)*
Crop diversity 0.27(7.22)* 0.35(2.96)* 0.06(1.63)***
Mixed system 0.33(1.92)*** 0.83(1.96)*** 0.34(1.74)***
Organic farms 0.59(4.21)* 0.51(1.42) 0.23(1.14)
Land race cultivation 0.22(1.53) 1.19(1.42) 0.31(1.50)
Expenditure 2.1E-04(0.90) 4.3E-02(2.91)* 1.1E-02(2.19)**
Labour -6.6E-04(-3.11)* -5.5E-04(-2.54)** -1.9E-02(-3.30)*
Crops_age 3.2E-04(0.73) 1.6E-02(2.49)** 1.9E-02(2.20)**
Mixed_Age 7.6E-04(0.22) 2.1E-03(0.60) 9.6E-02(2.14)**
Crops_ownership 0.29(22.62)* 5.1E-03(1.47) 8.5E-02(1.87)***
Organic_ownership 0.50(5.32)* 1.2E-04(0.04) 8.3E-02(1.80)***
Landrace_ ownership 0.31(3.31)* 1.2E-05(1.39)*** 7.1E-05(5.84)*
Crops_education 0.012(4.39)* 5.4E-07(0.57) 2.7E-06(2.06)*
Mixed_education 0.03(2.56)** 0.29(2.61)* 0.03(1.81)***
Organic_education 0.04(2.58)** 0.98(1.20) 0.14(1.33)
Landrace_education 0.01(0.55) 0.51(1.11) 0.14(1.29)
Expenditure_education 2.9E-06(0.13) 0.98(1.19) 0.36(3.31)
Labour_education 4.6E-06(2.26)** 4.5E-03(3.14)* 2.7E-04(0.96)
Crops_hhs 5.2E-02(2.22)* 4.9E-03(2.28)** 7.8E-06(2.26)**
Mixed_hhs 1.6E-02(0.08) 0.11(2.08)* 0.05(3.25)*
Landrace_hhs -2.6E-05(-1.09) -0.02(-0.04) -0.36(4.47)*
Crops_offfarm -0.03(-2.49)** -0.13(-2.40)** -0.23(-2.84)*
Mixed_offfarm 0.05(0.79) 5.5E-04(3.16)* 1.2E-03(5.89)*
Labour_Offfarm 8.3E-05(0.96) 3.4E-05(1.74)*** 7.2E-05(3.09)*
LR chi2(25) 1676.23 3225.88 2141.64
Prob > chi2 0.000 0.000 0.000
Pseudo R2 0.229 0.233 0.240
N 4032 4032 3942
Note: i. *denotes significant at 1% level while ** and *** indicates significant variables at 5% and 10 %
level.
ii. t values are in parenthesis.
108
An alternative method to account for preference heterogeneity is the use of the RPL model.
We next estimate the results using the RPL model to investigate whether there is an
observable improvement of the results. The RPL model is one of the fully flexible versions
of the discrete choice models because its unobserved utility is not limited to the normal
distribution. It decomposes the random parts of utility into two parts. One has the
independent, identical type 1 extreme value distribution, and the other representing
individual tastes can be any distribution. It is also characterised by accommodating
heterogeneity as a continuous function of the parameters. Therefore, as the next step of the
analysis, we ran the RPL model and the results of it are explained in the next section.
4.11 Results of the RPL model
Running the RPL model requires an assumption to be made about the distribution of
preferences for each attribute. The main candidate distributions are normal and log normal.
The former allows preferences to range between positive and negative for a given attribute,
the latter restricts the range to being of one sign only. Further, treating preference
parameters as random variables requires estimation by simulated maximum likelihood. This
means that the maximum likelihood algorithm searches for a solution by simulating m
draws from distributions with given means and standard deviations. Probabilities can be
calculated by integrating the joint simulated distribution. In this study the RPL model was
estimated using NLOGIT 4.0. All the parameters were specified to be independently
normally distributed and distribution simulations were based on 500 draws. The results of
the RPL estimations for the separate districts are reported in Table 4.8.
109
Table 4.8: Regression results of the RPL model for separate districts and pool data
Variables Ampara Anuradhapura Kurunegala Pool data
ASC 1.638 (0.191)* 4.743(0.477)* 2.468(0.213)* 2.304(1.121)*
Crop diversity 0.024(0.009)* 0.020(0.009)** 0.015(0.008)*** 0.018(0.005)*
Mixed system 0.076(0.041)*** 0.557(0.041) 0.135(0.041)* 0.059(0.021)**
Organic farms 0.157(0.044)* 0.092(0.044)** 0.154(0.045)* 0.136(0.025)*
Landrace cultivation 0.048(0.042) 0.090(0.043)** 0.206(0.044)* 0.107(0.024)*
Expenditure 2.1E-04(9.7E-05)** 8.9E-05(1.0E-04) 0.5.4E-04(1.1E-04)* 2.6E-04(5.5E-05)*
Labour -6.1E-0.4(7.8E-05)* -2.7E-04(7.8E-05)* -8.4E-04(8.4E-05)* -5.6E-04(4.6E-05)*
Log likelihood -1226.65 -923.40 -1040.72 -3294.469
Simulation 500 500 500 500
ρ2 0.187 0.164 0.191 0.248
N 1344 1344 1314 4003
Note: i. Standard errors are shown in brackets.
ii. *denotes significant at 1% level while ** and *** indicates significant variables at 5% and 10 % level respectively.
110
The results of the RPL model are quite similar in sign and magnitude to the CL
model where preferences are assumed to be homogenous. The crop diversity
coefficient for the standard CL model is 0.021 whereas it is 0.018 for the RPL for
pool data model. Pool data coefficients of the mixed farming systems are 0.077 and
0.059 for the CL model and RPL model respectively. The CL model contains all
positive and significant choice attributes except landrace cultivation in Ampara
district and mixed farming in Anuradhapura district with similar magnitudes to the
RPL results. The major difference between the two models is with regard to the
mixed farming system and landrace coefficients. The landrace cultivation variable
was not significant in Ampara sample while mixed farming system variable was not
significant for Anuradhapura sample for RPL model while these variables are highly
significant in the CL model. The CL model, unlike the RPL model, displays the
significant results for all variables. The log likelihoods are almost the same for the all
three models the CL model and with RPL model. Therefore, the Swait Louviere Log
Likelihood ratio test results of the test cannot reject the null hypothesis that the RPL
model and CL model estimates are equal. Hence no improvement in the model fit can
be achieved with the use of a RPL model. Accordingly, it can be concluded that the
CL model is sufficient for analysis of the data set presented in this study.
4.12 Estimating welfare changes with changing attributes and their level
Comparing the results of different models reveals that the basic CL model provides
the most significant results of the data. Therefore, the results of the CL model
reported in Table 4.5 can be used to calculate the value assigned by the farm families
to farm attributes. Point estimates of the WTA a change in one of the attributes in the
111
choice sets can be found by estimating implicit prices. Implicit prices are the
marginal rates of substitution between the attribute of interest and the monetary
attribute. This is equal to the ratio of the coefficient of one of the non-monetary
attributes and the monetary attributes. Equation 4.19 is used to estimate the implicit
prices for each attribute. Estimates of implicit prices for each of the non-monetary
attributes in the choice sets are reported in Table 4.9.
Table 4.9: Implicit price estimates for attributes
Variables Ampara Anuradhapura Kurunegala
Crop diversity 60.31 81.30 26.69
Mixed system 260.48 392.74 133.70
Organic farms 209.76 318.17 92.98
Landrace cultivation 197.32 460.06 352.49
Note: all implicit prices are estimated using the result of the basic CLM.
These estimates indicate that, for example, farmers’ valuation of the additional
benefits that farmers could obtain per month in increasing crop diversity by one is
Rs. 60, 81 and 27 in Ampara, Anuradhapura and Kurunegala farmers. It is clear that
farmers in Anuradhapura have placed relatively high values on organic farms and
landrace cultivation. This is expected as most farmers in these districts use their farm
products for their own consumption. However, these estimates are based on a ceteris
paribus assumption where we assume that all other parameters are held constant
except the attribute for which the implicit price is being calculated. Implicit prices,
however, do not provide estimates of compensating surplus. Estimating the overall
WTA for a change from the current situation requires more substantial calculations.
112
This is because the attributes in the choice sets do not capture all of the reasons why
respondents might choose to increase agricultural biodiversity. To estimate overall
WTA it is necessary to include the alternative specific constant. As discussed in
Section 4.2, the alternative specific constant captures systematic but unobserved
information about why respondents chose a particular option (unrelated to the choice
set attributes). Therefore, following Equation was used to estimate the consumer
surplus in different areas:
(4.28)
To illustrate this process, estimates are provided for six alternative scenarios. The
current situation and six scenarios are provided. These six household profiles were
generated to describe the variation in WTA within the sample, based on the farm
characteristics that were found to affect the households’ preferences for different
compensation plan attributes. The current situation is identified as a farm with three
crop varieties and no mixed farming as well as not organic farms or landrace
cultivation. We also assume contribution of farms to reduce household expenditure is
five per cent. We changed these characteristics for the rest of the profile gradually
and estimated change of the CS under each profile. Estimated change of CS in each
district is given in Table 4.10, 4.11 and 4.12.
Estimates of WTA the six scenarios in Ampara district are presented in Table 4.10.
These are marginal estimates, showing willingness to accept a change from the
current situation. Compared to the average household profile, household profiles
ASCCS enditurelandracefarmorganicfarmmixdiversitycrop
labour
exp___
1
113
three, five and six were WTA significantly higher amounts. The CS values indicate
that the value attached to scenario one was Rs. 4,802, 5,382 and 4,865 in Ampara,
Anuradhapura and Kurunegala sample respectively. That is, the average benefits that
each household can obtain by increasing crops diversity from three varieties to seven
varieties with having a mixed farming system. This shows that farmer welfare could
be easily increased by shifting farming practice to more diverse systems in rural
areas in Sri Lanka.
WTA value estimates for the six household profiles in the three regions disclose a
few main interesting findings. First, all attributes have positive use value in all
samples areas. This result shows that farm families in study area have strong
preference to increase agricultural biodiversity. It is clear that all diversity
components are valued highly by all types of households in study area. Second,
farmers’ valuation of different attributes is different in different areas.
Table 4.10: Estimates of WTA for various scenarios: Ampara
Crops
diversity Mixed
farm LR OP Consumption
(%) CS (Rs.)
As a
percentage
of average
income
Status quo 3 0 0 0 5 -
Scenario 1 7 1 0 0 10 4,802 3.50
Scenario 2 10 1 1 0 10 5,192 3.79
Scenario 3 15 1 1 1 15 5,993 4.37
Scenario 4 7 0 1 0 5 4,449 3.25
Scenario 5 10 1 1 1 15 5,481 4.00
Scenario 6 15 1 0 1 10 6,932 5.06
Note: Crops diversity represent the number of crops in the farm. Mixed farm, landrace cultivation and
organic farm variables are dummy variables while the expenditure variable provides a percentage of
the farms contribution to reduce family expenditure.
114
Table 4.11: Estimates of WTA for various scenarios: Anuradhapura
Crops
diversity
Mixed
farm
LR OP Consumption
(%)
CS
(Rs.)
As a
percentage
of income
Status quo 3 0 0 0 5 -
Scenario 1 7 1 0 0 10 5,382 3.44
Scenario 2 10 1 1 0 10 5,944 3.79
Scenario 3 15 1 1 1 15 7,255 4.63
Scenario 4 7 0 1 0 5 4,863 3.10
Scenario 5 10 1 1 1 15 6,530 4.17
Scenario 6 15 1 0 1 10 9,112 5.82
Note: Crops diversity represent the number of crops in the farm. Mixed farm, landrace cultivation and
organic farm variables are dummy variables while the expenditure variable provides a percentage of
the farms contribution to reduce family expenditure.
Table 4.12: Estimates of WTA for various scenarios: Kurunegala
Crops
diversity
Mixed
farm
LR OP Consumption
(%)
CS
(Rs.)
As a
percentage
of income
Status quo 3 0 0 0 5 -
Scenario 1 7 1 0 0 10 4,865 2.79
Scenario 2 10 1 1 0 10 5,038 2.89
Scenario 3 15 1 1 1 15 5,824 3.34
Scenario 4 7 0 1 0 5 4,524 2.59
Scenario 5 10 1 1 1 15 5,598 3.21
Scenario 6 15 1 0 1 10 5,774 3.31
Note: Crops diversity represent the number of crops in the farm. Mixed farm, landrace cultivation and
organic farm variables are dummy variables while the expenditure variable provides a percentage of
the farms contribution to reduce family expenditure.
Results show that most of the attributes are highly valued by Anuradhapura farmers.
For example, crops diversity is relatively valued highly by Anuradhapura farmers
than farmers in other two districts. Third, the demand for the farms with organically
produced products as well as landrace cultivation is relatively higher than that of
115
other attributes. This is revealed by relative high value of these two when comparing
with other attributes.
These per household estimates can be extrapolated to estimate the total benefit that
could be achieved for the total district. This type of analysis can provide the possible
social welfare estimates which can be used to inform appropriate policies in the
future. According to the Census and Statistics of Sri Lanka, the number of farmers
who cultivate less than 0.25 acre in Ampara, Anuradhapura and Kurunegala districts
are 67,778, 26,351 and 90,104 respectively. The total number of farmers who
cultivate less than 1 acre for the same districts is 80,778, 76,823 and 152,042
respectively. Using this secondary information, we estimated possible social welfare
gains under different profiles for different districts. Table 4.13 reports the results.
Table 4.13: Simulation total welfare gains to the districts (Rs. million / per season)
Ampara Anuradhapura Kurunegala
Total
WTA
As a
percentage
of average
income
Total
WTA
As a
percentage
of average
income
Total
WTA
As a
percentage
of average
income
Scenario 1 387.86 3.50
413.47 3.44
739.67 2.79
Scenario 2 419.42 3.79
456.65 3.79
765.98 2.89
Scenario 3 484.07 4.37
557.36 4.63
885.52 3.34
Scenario 4 359.41 3.25
373.60 3.10
687.82 2.59
Scenario 5 442.77 4.00
501.69 4.17
851.09 3.21
Scenario 6 559.94 5.06
699.98 5.82
877.96 3.31
Note: Total welfare gain is estimated using the total number of small-scale farm in these three
districts.
Results in Table 4.13 clearly show that improving agricultural biodiversity in rural
areas in Sri Lanka enables significantly increased social welfare. That is, the average
116
benefits that can be obtained by increasing crop diversity from three varieties to
seven varieties through having a mixed farming system are Rs. 387 Rs. 413 and
Rs.739 million per season in Ampara, Anuradhapura and Kurunegala sample
respectively. The results of this type of analysis can also be used to estimate values
associated with a range of scenarios resulting from different ecosystem management
practices. Government policy makers can use these value estimates, and estimates of
the value of any change in Sri Lanka to determine which scenarios are likely to have
the greatest net benefit for the community. From the empirical analysis, scenario six
produced the highest willingness to accept. This type of aggregate WTA can be
compared to aggregate costs in a cost-benefit analysis framework to assess net
welfare change in the society when introducing new policies to increase agricultural
biodiversity.
4.13 Summary and key findings
The research reported in this chapter of the thesis represents one of the first attempts
to use choice modelling to investigate farmers’ preference for different attributes of
agricultural biodiversity that can be seen in small-scale farm in Sri Lanka. We
applied the choice modeling approach to identify the possible benefits of conserving
agricultural biodiversity in the country. The first of the two CL models presented
here was found to be robust, being statistically significant, having relatively high
explanatory power and having identically and independently distributed error terms.
Therefore, the result of that model is used to analyse the welfare changes in the
society. The study provides important information for policy-makers considering the
117
consequences of changes in the condition or quality of an ecosystem in small-scale
farms in rural agricultural areas.
Four conclusions can be drawn from this study. Firstly, owing to educational and
poverty issues, some policy makers in developed countries are suspicious of whether
non-market valuation techniques like CVM and CE method can be applied in
developing countries like Sri Lanka. This CE study has demonstrated that carefully
designed and pre-tested nonmarket valuation techniques can be applied in developing
countries without any doubt. Secondly, farmers have strong positive attitudes
towards increasing agricultural biodiversity in rural areas. This is evident from the
results obtained from CL model. Thirdly, the study illustrates that there is a
possibility to improve agricultural biodiversity using appropriate policies in the
country. Finally, the application of CE study appears promising by its potential to
model complex and simultaneous trade-offs in the field of ecological management.
The choice experiment technique can be used to model a variety of simultaneous
trade-offs which involve a mixture of environmental and socio-economic factors.
The results provide a tool for decision makers to use in prioritising ecosystem
management options in the agricultural sector.
In general, the findings of the choice experiment support the priori assumption that
small-scale farms and their multiple attributes contribute positively and significantly
to the utility of farm families in Sri Lanka. To the extent that the findings are
representative of other rural areas in the country as they confirm that small-scale
farms continue to be a vital for that nation since the benefits to farms are overall
positive and high. The value estimates reported in this chapter represent lower
118
bounds since only the private, use values of small-scale farms were estimated. The
results reveal that differences between regions, in terms of market integration,
infrastructure quality and agro-ecological condition, affect small-scale farmers’
private valuation. Our results indicate that in isolated regions farmers highly value
organic farming methods and landrace cultivation practices. The CE study discloses
the farm family and regional characteristics that are important to consider in
designing programs or policies to conserve or enhance the agricultural biodiversity
and other attributes of Sri Lankan farms.
It is clear that various attributes of agricultural biodiversity provide direct and
indirect benefits and advantages which meet human needs in different ways. Putting
a value on these benefits is extremely difficult, but decision makers often call for
them to be expressed in monetary terms. To this end, in this study we present the
results of a CE study designed to shed light on poor farm households’ preferences for
various farm attributes and these households’ trade-offs among these attributes. The
findings presented here are, therefore, expected to inform the design of efficient,
effective, equitable, and targeted compensation and livelihood diversification
policies in the country. The results of this study will suggest how economic policies
may be designed and appropriately implemented in the future in Sri Lanka.
119
CHAPTER FIVE
FACTORS INFLUENCING FARMERS’ DEMAND FOR AGRICULTURAL
BIODIVERSITY
5.1 Introduction
Agricultural biodiversity is of fundamental significance to human societies,
providing socio-cultural, economic and environmental benefits (Mozumder and
Berren, 2007). It is essential to food security and poverty alleviation in rural
economies. The conservation and sustainable use of all aspects of agricultural
biodiversity may presents opportunities for enhancing soil fertility, naturally
controlling pests, reducing the use of pesticides while increasing yields and incomes
(Brock and Xepapadeas, 2003). Diversified agricultural production also offers
opportunities to expand new markets and further increase the level of food security
for rural households (Birol, 2004; Ceroni et al., 2005). The underlying causes for the
loss of agricultural biodiversity are extremely complex. They are closely related to
the needs of increasing food demands, growing market pressure, agricultural
development policies, demographic, economic and social factors (Mozumder and
Berren, 2007). Many agricultural practices such as reliance on monoculture,
exotic/cross breeds, high yielding varieties, mechanization, and misuse of
agricultural chemicals have caused negative impacts on agricultural biodiversity at
all levels in the long term. Such loss of biodiversity may be accompanied by the loss
of cultural diversity of traditional communities (see Appendix A.1), and their
impoverishment (Franks, 1999).
120
Conserving and sustainable use of agricultural biodiversity may provide local,
national and global benefits (Bardsley, 2003). The global interest in maintaining
agricultural biodiversity is linked to the fact that most species important to
agriculture may be of benefit not only to the region of their origin, but other regions
of the globe as well. Additionally the conservation and sustainable use of associated
agricultural biodiversity can contribute to maintaining the health and quality of the
global environment, by, for example, providing habitats for wildlife, protecting
watersheds, and reducing the use of harmful chemicals (Gauchan et al., 2005).
Consequently, using agricultural biodiversity sustainably may provide
environmental, economic and socio-cultural benefits on national, regional and global
scales (Hengsdijk et al., 2007). Therefore, understanding the underlying causes of
degradation of agricultural biodiversity would help to integrate global environmental
imperatives into existing sustainable development efforts in the appropriate regions
and countries.
This chapter aims at identifying the determinant factors of conserving crop variety
diversity (richness in crop varieties) and livestock variety diversity (richness in
animal breeds) which are important parts of agricultural biodiversity. A farm
household model is used to predict farmer demand for crop variety and livestock
variety using small-scale farms data in Sri Lanka. Farm households who are most
likely to sustain observed levels of agricultural biodiversity are described
statistically. Findings can assist those who formulate agri-environmental policy in Sri
Lanka to design efficient programs that incorporate family farm management. The
next section provides the context for the present research by looking at what work
121
has already been done in the field of agricultural biodiversity. It critically looks at the
existing research that is significant to the work carried out in this study.
5.2 Literature review on demand for agricultural biodiversity
Several studies have used econometric models to identify the determinants of
diversity in livestock and crops in developing or transitional economies. Detailed
case studies, conducted in Peru (potato), Turkey (wheat), and Mexico (maize), have
sought to identify some of the important factors that positively and negatively affect
the conservation of agricultural biodiversity (Brush et al., 1992; Meng, 1997; Van
Dusen, 2000; Smale et al., 2002). However, most of these studies (Brush et al., 1992;
Franks, 1999) on in situ conservation of agricultural biodiversity on farms
concentrate on diversity within a single crop or animal bread. When analysing the
multiple benefits of the farms under semi-subsistent rural areas, concentration on
variety diversity is more important than considering a single crop.
According to Fafchamps (1992) crop diversity may be particularly important for
farmers with limited opportunities to trade and participate in markets. He identified
agro-ecological heterogeneity and imperfect markets with high transaction costs in
rural areas as contributing factors to the demand for agricultural biodiversity. Brock
and Xepapadeas (2003) develop a conceptual framework for valuing biodiversity
from an economic perspective. They consider biodiversity important because of a
number of characteristics or services that it provides or enhances. This study shows
that a more diverse system could attain a higher value even though the genetic
distance of the species in the more diverse system could be almost zero. Mauricio
(2004) argues that crop diversity maintained by farming household’s results from the
122
interplay between a demand and a supply for this diversity. According to them
interventions to support on farm conservation can be conceptualised by the way they
influence these two factors. Demand interventions should increase the value of crop
diversity for farmers or decrease the farm-level opportunity costs of maintaining it,
while supply interventions should decrease the costs of accessing diversity.
Bunning and Hill (1996) present a gender perspective on farmers' rights and illustrate
with several case studies that attempt to identify the different roles and
responsibilities of women on conserving crop diversity. This study explains women’s
role in the conservation, development and utilisation of less common crops and
varieties, and in the management of high-diversity home gardens.
The theory of impure public goods was used by Heisey et al. (1997) to demonstrate
why farmers may not grow wheat cultivars with the socially desirable level of rust
resistance. They argue that farmers may grow cultivars that are high yielding though
susceptible to rust. Furthermore, many farmers may grow cultivars with a similar
genetic basis of resistance. This study shows three ways of reducing expected rust
losses. They are (a) more diversified genetic background in released wheat cultivars;
(b) greater spatial diversity in planted cultivars; or (c) use of temporally changing list
of cultivars known to be rust resistant. Yield trade-offs associated with these policies
illustrate potential costs of increasing genetic diversity.
Meng (1997) investigated the diversity of traditional varieties of wheat on Turkish
farms. He analyzed the impacts of a combination of factors, including missing
markets, farmers’ attitudes towards risk and environmental constraints on wheat
123
diversity outcomes. According to this study, regional effects, off-farm income and
distance from markets significantly explain diversity of traditional varieties of wheat
on Turkish farms. Franks (1999) discussed the value of plant genetic resources for
food and agriculture in the United Kingdom (UK). According to him the UK’s agri-
environmental conservation schemes do not prioritize the conservation of genetic
diversity of agricultural crops.
According to Van Dusen (2000) agro-ecological and market characteristics
significantly affect the levels of diversity maintained by households. He developed a
theoretical model in which a household's decision to plant a milpa variety is linked to
household specific, agro-ecological, and market variables. The empirical
methodology in this study uses a Poisson regression. The results from the regressions
of household level diversity showed that a range of household, village,
environmental, and market conditions affect the diversity outcomes. Market
integration, measured by distance to a regional market, use of hired labour, and
international migration, were found to negatively affect diversity outcomes. Agro-
ecological conditions, measured by the number of plots, plots with different slopes,
and the high altitude region, were all found to positively increase agricultural
biodiversity in the study area.
According to Maikhuri et al. (2001) environmental, biological, socio-cultural and
economic variations in the Himalayas have led to the evolution of diverse and unique
traditional agro ecosystems, crop species, and livestock, which help the traditional
mountain farming societies to sustain themselves. It was found that the loss of
agricultural biodiversity and the changing socio-cultural and economic dimensions
124
and their impacts on the sustainability of Himalayan agro ecosystems are emerging
as major causes of concern at local/regional/national levels. This study also discusses
the appropriate options to meet these challenges.
Di Falco and Perrings (2003) investigated the impact of providing financial
assistance to farmers in maintaining crop biodiversity in an uncertain setting. The
findings reveal that risk aversion is an important driving force for crop biodiversity
conservation24
. Li-zhi Gao (2003) investigated genetic erosion of rice and its possible
impacts on the Chinese economy. The result of this study finds that genetic erosion
can significantly affect the future yield of any crop in China. Meanwhile Scarpa et al.
(2003) show that for Creole pigs in Mexico, the respondent’s age, years of schooling,
size of the household and the number of economically active members of the
household were important factors in explaining breed trait preferences. Accordingly
younger, less educated and lower income households placed relatively higher values
on the attributes of indigenous piglets compared to exotics and their crosses.
A farm household model was used to identify the factors affecting inter and intra
crop species diversity of cereal crops in the northern Ethiopian highlands by Benin et
al. (2003). They compared factors explaining the inter-specific diversity and infra-
specific diversity. This study found that a combination of factors related to the agro-
ecology of a community, its access to markets, and the characteristics of its
households and farms significantly affect both inter-specific and intra-specific
diversity of cereal crops. Their findings showed that agro-ecological, market,
24
Risk averse farmers can hedge against uncertainty they face by allocating land to different crop
species.
125
household and community level characteristics affect increasing agricultural
biodiversity at the farm level.
An empirical approach was employed to understanding the determinants of farmers'
access to and use of, crop genetic resources by Van Dusen (2005). He also
investigated the impacts of farmer behaviour on crop populations. In the same year
Van Dusen et al. (2005) carried out an empirical case study about farmer
management of rice genetic resources in two communities of Nepal. The decision-
making process of farm households is modelled and estimated in order to provide
information for the design of community-based conservation programs. Gauchan et
al. (2005) investigated the socioeconomic, market and agro ecological determinants
of farmers’ maintenance of rice diversity at the household level. They assessed
spatial rice diversity at the farm level using household survey data. Findings of this
study are useful for designing policies for farm conservation programs. Winters et al.
(2005) studied potato diversity managed on farms in Peru. Their findings showed
that the diversity of potato varieties managed on farms increases with the size of the
land owned, number of different plots cultivated, distance to the nearest market and
wealth indicators at a diminishing rate.
The two-stage tobit procedure was used to identify the determinants of on-farm
variety diversity in a rain fed ecosystem in Nepal by Ganesh and Bauer (2006). The
results identified motivating factors for variety diversification such as heterogeneous
production environments, risk consideration and farmers’ participation in the
markets. Wilson and Tisdell (2006) investigated how specialisation of production of
commodities in the agriculture sector leads to the concentration of genetic materials.
126
Isakson (2007) investigated how the participation of Guatemalan peasants in the
market economy is related to on-farm conservation of crop genetic diversity in three
crops: maize, legumes, and squash. He found that participation in markets is not
inherently detrimental to the provisioning of crop genetic resources. However,
without proper protections in place market participation may unleash processes that
contribute to genetic erosion over time. Nagarajan et al. (2007) investigated the
determinants of biological diversity of millet crops in the semi-arid regions in India.
This analysis is based on data collected through sample surveys of farmers and
traders in selected sites of Karnataka and Andhra Pradesh, combined with cultivar
taxonomies developed with geneticists and applied to seed samples. Findings in this
study demonstrate that millet crop diversity levels at both scales of analysis are
significantly influenced by seed system parameters, factors which related studies
have omitted. In particular, the presence of active local (formal and informal) seed
markets enhances millet richness among and within farming communities.
Accordingly, crop improvement strategies oriented toward local seed markets could
provide important benefits and incentives to farm households living in these areas.
An agricultural household model was developed with missing market for a
subsistence crop that arises from non-market values of the crops by Arslan (2007).
This study theoretically derived household-specific shadow prices for maize and
empirically estimated these shadow prices for rural farmers in Mexico. The results
suggest that the value of traditional maize varieties for subsistence farmers is
significantly higher than market prices for maize. Pascual and Perrings (2007)
distinguished between the proximate and fundamental causes of biodiversity loss in
terms of decentralized behaviour of farming households. Special attention is paid to
127
the interplay between micro-economic decisions and the macro-economic factors
(institutional and market conditions) that determine the effects of government
policies.
According to the above literature review it is clear that a large number of studies
have been conducted in the area of agricultural biodiversity. They have addressed
various issues in this field. However, it is obvious that more conceptual and
theoretical work is needed to understand the factors influencing farmers’ demand for
agricultural biodiversity in developing countries. For example, analysis including
direct policy relevant variables to demand for crop variety diversity and animal
variety diversity is not properly explained in the literature. Moreover, although a
wider cross-section of case studies has been conducted in commercially-oriented
farming systems, an analysis of subsistence oriented farming systems is required in
order to generalise and validate the empirical findings (Ceroni et al., 2005).
In the next section, the theoretical model that is used to analyze the demand for
agricultural diversity is explained. The behavioural model employed to explain the
farm households’ production and consumption decisions is based on the semi-
subsistence model of the farm household in rural economy (Singh et al., 1986; de
Janvry et al., 1991; Taylor and Adelman, 2003; Birol et al., 2005). Firstly, we explain
the way of deriving demand for agricultural biodiversity using basic farm household
model. Second, the empirical approach of different model estimation is discussed.
The background to the general model is provided in the next section.
128
5.3 Derivation of demand for agricultural biodiversity
In order to estimate demand for agricultural biodiversity we use a basic model
developed by Singh et al. (1986); Taylor and Adelman (2003) and Van Dusen and
Taylor (2005). A similar model was used by Birol et al. (2005) to analyse four
components of agricultural biodiversity found on family farms in Hungary. The
utility a household derives from various consumption combinations and levels
depends on the preferences of its members. Preferences are in turn shaped by the
characteristics of the household, such as the age or education of its members, and
wealth (Birol, 2004). Choices among goods are constrained by the full income of the
household, total time (T) allocated to farm production (F) and leisure (l), and a fixed
production technology represented by G(.). Suppose a farm family maximises his/her
utility over consumption of market purchased goods, Cm, leisure, Cl, and owned farm
outputs, Cf. The utility is maximised subject to budget, time, and production
technology constraints respectively. Household utility is influenced by a vector of
household characteristics h . The utility function is assumed to be quasi-concave
with positive partial derivatives (Birol, 2004; Van Dusen and Taylor, 2005). The
prices of all market purchased goods, inputs and wages are exogenous, and
production is assumed to be riskless. The model can be written as follows:
);,,( hflm CCCUU (5.1)
Constraints:
mmxe CpXpwFIwTI (income constraint) (5.2)
0);,,( fXFQG (technology constraint) (5.3)
129
TCLF ld (time constraint) (5.4)
Equation 5.1 gives the utility function of a representative household, while Equation
5.2 gives the full income constraint. Full income is composed of value of stock of
total time owned by the household T, exogenous income Ie, the values of household
management input used in the small-scale farm production F, other variable inputs
required for production of small-scale farm outputs, X and market commodities
consumed by the farm family, Cm. The household faces a production constraint for
the production technology on the small-scale farm (Equation 5.3). It gives the
relationship between farm inputs F, X and all outputs Q, and has the properties of
quasi-convexity, increasing in outputs and decreasing in inputs (Taylor and
Adelman, 2003). The vector, f represents the fixed agro-ecological features of the
small-scale farm, such as soil quality and land shape. The household also faces a
time constraint. Labour use in small-scale farm cultivation F is one use of labour
which competes with other uses, including off farm employment Ld and leisure Cl.
The household is driven toward the goal of increasing diverse farming within the
family farm because of uncertainty, unreliable or missing markets, as well as the
desire to consume fresh food. This phenomenon brings about an additional constraint
that induces the household to equate small-scale farm output demand and supply,
resulting in an endogenous, shadow price for small-scale farm outputs (Singh et al.,
1986; Birol, 2004). This can be written as follows:
)(ZCQ ff (5.5)
130
Qf and Cf denote the quantity supplied and consumed of small-scale farm produce,
and Z is a vector of exogenous characteristics related to availability and access to
markets. This equality condition implicitly defines the shadow prices for small-scale
farm outputs under missing market, which guides production decisions (Birol, 2004).
The production and consumption decisions of the household cannot be separated
when labor markets, markets for other inputs, or product markets are imperfect.
Then, prices are endogenous to the farm household and affected by the costs of
transacting in the markets (Taylor and Adelman, 2003). The specific characteristics
of farm households (represented by vector h ) and physical access to markets
(represented by vector Z) influence the magnitude of transaction costs and hence, the
effective price governing the household’s choices (Van Dusen and Taylor, 2005).
The household maximises utility subject to constraints explained in Equations 5.2,
5.3, 5.4 and 5.5. This maximisation results in the following Lagrangian Equation 5.6:
)();,( , XpwFCpwCIwTCCCUL xmmlehflm
);,,()]([ ffff XFQGZCQ
(5.6)
However, when all relevant markets function perfectly, farm production decisions
are made separately from consumption decisions (Birol, 2004). In this context, full
income in a single decision-making period is composed of the net farm earnings
(profits) from crop or livestock production (Qf), of which some may be consumed on
farm and the surplus sold, and income that is exogenous to the season’s crop/animal
breads and variety choices, such as stocks carried over, remittances, pensions, and
other transfers from the previous season (Ie). The household maximises the net farm
131
earnings subject to constraints and then allocates these with other income among
consumption goods (Smale et al., 2001).
Farm production decisions, such as crop/animal breeds and variety choices, are
driven by net returns, which are determined only by wage, input and output prices
(w, px and po) and farm physical characteristics (represented by vector βf)25
. This will
only change the full income budget constraint adding farm profit as an income and
market prices have some role to play in decision making (Singh et al., 1986; Meng et
al., 1998 and Smale et al., 2001). Accordingly,
0)( ZCQ ff and additional
income constraint can be added to the Equation 5.6. It can be given as
0)]([ pZCQ ff where p0 is the output prices of the commodities that are produced
by the small-scale farms and has a market.
Assuming interior solutions exist, the optimal set of choice variables are given by the
solutions to the first order conditions. The first order necessary conditions with
respect to decision variables are:
0// mmm pCUCL (5.7)
0// wCUCL ll (5.8)
0)(/ mmxel CpXpICFTwL (5.9)
0/ fGwFL (5.10)
25
When comparing farmers among communities located in a broader geographical area, one can see
that their decisions are also affected by factors that vary at a regional level but that they themselves
cannot influence. These include several fixed factors hypothesized to affect variation in the diversity
maintained among regions, such as agro-ecological conditions or infrastructural development, or the
ratio of labor to land.
132
0/ xx GpXL (5.11)
0);,,(/ fXFQGL (5.12)
0// ff CUCL (5.13)
0/ ff GQL (5.14)
Equations 5.7 and 5.8 imply the optimal demand for market purchased goods and
leisure respectively. These equations show that the marginal utility the household
receives from each commodity equals to Lagrange multiplier, , times its market
price, mp and w respectively. Equation 5.9 is the full income constraint, which
ensures that the net full income received is spent. Equation 5.10 and 5.11 represent
the optimal amount of each input required in the small-scale farm, determined by the
equality between the Lagrange multiplier, , times the price of the input and its
marginal product.
Equation 5.12 ensures being on the transformation function. The optimal demand for
small-scale farm output is given by Equation 5.13. This condition implies that the
marginal utility obtained from consuming small-scale farm products is equal to its
shadow price, . The supply of the small-scale farm output is given by Equation
5.14. This implies that the marginal cost of producing small-scale farm products
equals to its shadow price. Substituting for the shadow price in 5.13 and 5.14, it
can be shown that the marginal utility of small-scale farm outputs is equal to the
marginal cost of small-scale farm outputs and to the shadow price (Birol, 2004).
Similar derivation could be found in the study carried out by Birol (2004) in order to
estimate the demand for attributes in home garden in Hungary:
133
f
f
GC
U
(5.15)
The endogenous shadow price is household-specific, depending on the household
characteristics that affect access to markets and consumption demand, such as
wealth, education, age, household composition. Agro-ecological features of the
small-scale farm such as soil quality or irrigation enter the equation through their
effect on supply. Fixed factors related to market transactions costs and observed
market prices also influence the shadow prices of small-scale farm outputs (Feder
and Umali, 1993). The shadow price, , can therefore be expressed as a function of
all exogenous prices and household, agro-ecological and market characteristics:
),,;,,(* ZwPP fhxm (5.16)
The general solution to the household maximisation problem yields a set of optimal
choices for production, inputs demand and consumption demand as given in
following Equations:
):,,(*
fxff wpQQ (5.17)
):,,(*
fx wpFF (5.18)
):,,(*
fx wpXX (5.19)
):,,(*
hmii wpCC I =m, l, f (5.20)
Equation 5.17 is the optimal supply of small-scale farm outputs while Equation 5.18
provides the expression for optimal demand of household labour in small-scale farm
134
production. Equation 5.19 gives the optimal demand for all other inputs to small-
scale farm production and Equation 5.20 is the optimal demand for market purchased
goods (m), household produced goods (f) and leisure (l).
Substituting these solutions for the shadow price (Equation 5.16) into small-scale
farm output production and consumption solutions (Equations 5.17 and 5.20), the
optimal production of small-scale farm outputs is seen to be a function of all
exogenous variables:
),,;,,(* ZwPPQQ fhxmff (5.21)
We assume that the household does not value diversity itself rather than the direct
benefits of it. Therefore, diversity is not explicitly in the utility function. The
diversity within a given household is the result of the choice of which crops to
produce, subject to constraints. This ‘diversity outcome’ in the constrained case takes
the form of a derived demand for number of varieties resulting from the farmer’s
utility maximisation subject to income, production, and market constraints.
Following Van Dusen and Taylor (2005) the level of agricultural biodiversity
maintained on the small-scale farms, which is a direct outcome of the production and
consumption choices of the farm household, is a function of all prices, and
characteristics of the households, markets, and of the small-scale farm plots. This
relationship can be given as shown in Equation 5.22:
),,;,,( * ZwPPQBDBD fhxmf (5.22)
It becomes clear that conceptual approach used in this study to analyse the demand
for agricultural biodiversity is based on the theory of the farm household model
135
developed by Singh et al. (1986); Taylor and Adelman (2003) and Van Dusen and
Taylor (2005). Some of the interesting applied economic analyses of agricultural
biodiversity based either on the farm household model or a model of variety choice
are Brush et al. (1992); Meng (1997); Smale et al. (2001) and Birol (2004). Studies
in this area commonly use count data analysis or Logit/Probit model for empirical
estimation.
In this study crop or livestock diversity was taken as count number. This is a discrete
variable ranging between zero and nine in our sample. It is preferred in this study as
a measure of agricultural biodiversity because it is simple to construct and yet
elaborate enough to describe the richness of species. The empirical model
specification, relevant variables and theoretical background behind each model are
explained in subsequent sections below.
5.4 Empirical model specification and relevant variables
In this study agricultural biodiversity is investigated in terms of crop diversity and
livestock diversity. The definitions of these variables are given in Table 5.1.
Table 5.1: Definition of the agricultural biodiversity
Components Definitions
CD The total number of crops that are grown in the farm
AD The total number of animal species in the farm
Note: In this analysis we investigate the influencing factors for crop variety and livestock variety
selections. Multi-crops and multi-livestock practices are the most important farming practices that can
be seen on small-scale farms in Sri Lanka.
136
In order to understand the important determinants of variety demands, different types
of policy relevant variables are selected. Importance of these variables were
understood by the information gathered from the pilot survey as well as information
provided by the agricultural specialist in this area. All collected variables are divided
into three main categories namely household characteristics, market characteristics
and other characteristics. Table 5.2 provides the definition of all variables used in the
regression analysis.
Table 5.2: Definition of potential explanatory variables
Variables Definition
Household characteristics
EXP Experience of farm decision maker (number of years)
OWN Household owns a business vehicle or not: dummy- 1 if Yes, Otherwise 0
HMP Household member’s participation in agricultural activities (%)
GEN Decision maker, male or female: dummy- 1 if Male, Otherwise 0
INC Off farm income of the family (Rs. 000)
SHL Shared labour (number in the last season)
WLH Household wealth: dummy- 1 if wealthier, Otherwise 0
Market characteristics
NMA Number of market access days per week (number)
DIMK Distance to the nearest market (KM)
DSN Direct sales or not (intermediary) : dummy- 1 if Yes, Otherwise 0
PRIF Price fluctuation of the output(index)i
Other characteristics
AS Receiving agricultural subsidize: dummy- 1 if Yes, Otherwise 0
IOM Percentage of investment of owned moneyii
Note: i. Price fluctuation indexes were constructed using average unit price changes over the last two
seasons for crops and livestock outputs.
ii. This variable is created by taking the percentage value of own money invested to total farm
investment in the last season. Total farm investment includes own money plus borrowing for the last
season.
137
All these independent variables are based directly on the questionnaire responses.
During the survey we collected some variables related to farm specific characteristic
such as irrigation water availability, soil fertility and land shape. However, these
variables were dropped from the analysis due to three reasons. Firstly, these variables
are not important for determining animal diversity. Secondly, most of these variables
are relatively less policy relevant and beyond the farmers’ control. Thirdly, in order
to avoid the over identification problem, some of the variables had to be dropped
from the analysis.
It is clear that some variables are defined as numbers (such as number of years in
experience in farming) while other variables are defined as dummy variables.
Experience in farming is one of the important variables used in the analysis.
Experience of household head in agricultural activities is expected to have a
quadratic relationship in selecting a diverse farming system (Van Dusen, 2000), as
younger households may be more willing to try out different crops and varieties,
while older households with more experience in farming may be more set in their
production activities and are less likely to try new crops and varieties. Therefore, it is
hypothesised that demand for agricultural biodiversity will decrease with experience
in farming. Owning a business vehicle can have a positive correlation with
agricultural biodiversity. This is because a business vehicle can help farmers to take
different products into different markets. Given the limited market places as well as
market access days in rural areas in developing countries, business vehicles can be
used to sell farm products directly in the market. This will avoid an intermediary
transaction.
138
A household member’s participation in agricultural activities is one of the important
variables used in this analysis. This variable shows the number of mandates received
from members of the family (except household head) for agricultural work during the
last season. Participation rate captures the family labour availability for farming
activities. In general, the number of members in the family is expected to have a
positive effect on diversity through its effects on preferences and overall labour
capacity. Considering the household preferences, it is clear that when the family size
increases, expenditure on food consumption26
also increases.
Diverse or more productive farming systems can help minimise household
expenditure on food consumption. However, diverse farming systems mean that the
labour requirement is also higher. Therefore, large families with higher participation
rates may not face any labour constraints for maintaining diverse farming systems.
Gender variable can give different results since it depends on their preference.
Women household heads are thought to influence diversity in positive as well s in
negative ways. It is expected that a women’s knowledge in seed selection and
management would contribute towards increased richness. On the other hand, their
low economic position such as lack of skills in ploughing may influence their
decisions to grow high number of varieties.
Off farm income is expected to have a negative correlation with agricultural
biodiversity. The reason is that farmers who have other types of income sources are
less likely to maintain diverse farming practices due to managerial impossibility and
labour constraints. Shared labour shows the strength of social capital in rural area.
26
A positive correlation can be expected between agricultural biodiversity and income spent on food
consumption as well.
139
This variable shows the exchange labour quantities in a given cultivating season. As
this helps reducing labour constraints, it is expected to have a positive correlation
with diverse farming systems.
We created dummy variables to differentiate whether households are wealthier or
not. Three things were considered for making this decision. Firstly, we classified
houses as luxury/ upper middle class, ordinary and small house/cottage. Secondly,
the facilities available to their house are investigated. Under this category, telephone,
electricity, pipe water, vehicle road to the house, water sealed toilet and attached
bathrooms were considered. Thirdly, durable assets are considered. They include
vehicles, threshing machines, water pumps and motorcycles. If a household belongs
to a luxury/ upper middle class or ordinary house and has at least four of the afore
mentioned facilities with at least two of the asset varieties, that household is
identified as a wealthier household. It is hypothesised that wealth is negatively
correlated with agricultural biodiversity. This is because wealth helps reduce the risk
of having family household needs for poor farmers in rural area.
A few interesting market characteristics as explained in the Table 5.2 were used to
see whether these variables are important determinants of agricultural biodiversity.
Market infrastructure operates in several ways that may not be dissociable in a given
location at one point in time. For example, the more removed a household is from a
major market centre, the higher the costs of buying and selling on the market and the
more likely that the household relies primarily on its own production for subsistence.
This implies that the more physically isolated a community or household, the less
specialised its production activities. On the other hand, as market infrastructure
reaches a village, new trade possibilities may emerge, adding crops and production
140
activities to the portfolio of economic activities undertaken by its members. The
theory of the household farm predicts that the higher the transactions costs faced by
individual households within communities, the more we would expect them to rely
on the diversity of their crop and variety choice to provide the goods they consume.
Consistent with this hypothesis, Van Dusen (2000) found that the more distant the
market, the greater the number of maize, beans, and squash varieties grown by
farmers. In Andean potato agriculture, Brush et al. (1992) found proximity to
markets to be positively associated with the adoption of modern varieties, but this
adoption did not necessarily decrease the numbers of potato types grown.
We hypothesised that the number of market access days is expected to have a
positive correlation with agricultural biodiversity as it helps minimise the risk of
selling the surplus. Distance to the nearest market is one of the important variables
used in this study. It is hypothesised that when the distance to the nearest market is
higher, farmers are less likely to maintain a diverse system27
. This is because
whenever farmers face market constraint, they are less likely to have diverse output
for market. A direct sale variable is included to see whether it has some impact on
selecting a diverse farming system. It is expected that farmers who sell their output to
market directly are more likely to maintain a diverse farming system. A variable to
capture price fluctuation on agricultural biodiversity is used in this analysis. This
variable is created for average output price changes for crops and livestock over the
last two cultivation seasons. It is expected that the coefficient of this variable has a
positive correlation with agricultural biodiversity.
27
This may not be a reasonable hypothesis for rural subsistence area. This is because their main
purpose of production is the consumption. However, farmers in semi-subsistence area have two main
objectives of their farming. One is consumption while other is revenue purpose by selling the surplus
to the market.
141
Among the other characteristics, receiving agricultural subsidy and own money
investment in the farm are important policy relevant variables. Receiving agricultural
subsidy helps reduce financial constraint of rural farmers. It is expected that this
variable has a positive impact on selecting more a diverse farming system. Farmers
can finance their expenditure for the agriculture in different ways. Some farmers use
their own saving while others borrow money from formal or informal sources.
Borrowing agricultural inputs from informal sources is also common practice in rural
areas in Sri Lanka. For example, some farmers borrow seeds or pesticides or fertiliser
from village shops with the promise of paying after selling their product28
. We
included a variable to understand this behaviour and agricultural biodiversity.
We hypothesised that the percentage of own money contribution to total farm
expenditure has a positive correlation with agricultural biodiversity. This is because
farmers often borrow money in order to maintain a specialisation system with a
marketing purpose. It is clear that the relevance of these variables for the different
models can be different. For example, although agricultural subsidy is important for
determining crop varieties, it is not an important determinant for animal varieties.
This is because agricultural subsidy policies in the country only focus on the crop
sector. Therefore, the subsidy variable is not included for the animal variety model.
Theoretically, possible signs in different variables are given in Table 5.3.
28
In this case interest paid is very high. It is around 20 per cent per month in most rural areas.
142
Table 5.3: Explanatory variables used in the demand model
Variable
s
Definitions
CD
AD
Household characteristics
EXP Experience of farm decision maker - -
OWN Household owns a business vehicle or not + +
HMP Household member’s participation + +
GEN Decision maker, male or female +/- +/-
INC Off farm income of the family - -
SHL Shared labour + +
WLH Household wealth - -
Market characteristics
NMA Number of market access days per week +/- +/-
DIMK Distance to the nearest market +/- +/-
DSN Direct sales or not (intermediary) +/- +/-
PRIF Price fluctuation of the output + +
Other characteristics
AS Receiving agricultural subsidize - NA
IOM Percentage of investment of owned money + + Note: Definitions of all variables are given in Table 5.2. Expected signs in each variable are provided
in this Table. As shown in the Table, some variables can take positive or negative depending on the
situation.
A summary of the models to be used for the empirical estimation is provided in
Table 5.4. The Poisson model (PM) or Negative binomial model (NBM) for count
data may be the suitable model for estimating the determinants of the farm family’s
decision about how many crop and livestock species to cultivate on the farm (see, for
example, Greene, 1997). Negative binomial regression is used to estimate count
models when the Poisson estimation is inappropriate due to overdispersion. In a
Poisson distribution the mean and variance are assumed to be equal (Winkelmann,
2008). When the variance is greater than the mean the distribution is said to display
over dispersion. When over dispersion is an issue in the data, the negative binomial
model should be used (Hilbe, 2011).
143
Table 5.4: Summary of the econometric models to be used for the analysis
Different components of
agricultural biodiversity
Econometric
model
Definition
Crop species diversity
and
Livestock diversity
Poisson model Suitable model for estimation of
count data, based on Poisson
distribution, but restricted by the
assumption that the sample mean
equals sample distribution
Negative
binomial
model
Suitable model for estimation of
count data, based on Poisson
distribution, however, unlike the
Poisson model, it is not based on the
assumption that the sample mean
equals sample distribution
Note: Theoretical explanations about these models are given in Section 5.5. Before estimating the
final models, different tests were performed to find most appropriate model for the each estimation.
In the next section log-linear models for count data under the assumption of a
Poisson error structure are explained. These models have many economic
applications, not only to the analysis of counts of events, but also in the context of
models for contingency tables and the analysis of various incidents. We introduce the
Poisson regression model and discuss the rationale for modelling the logarithm of the
mean as a linear function of observed covariates. Then the negative binomial model
is discussed. As an extension, zero-inflated Poisson and negative binomial models
are explained in the Appendix K.
5.5 Theoretical approaches for the relevant models
A count variable is a variable that takes on nonnegative integer values. Both
variables that are of interest in this study come as counts. For example, crop
144
diversity and animal diversity. These variables have two important characteristics in
common: there is a natural upper bound, and the outcome will be zero for at least
some members of the population. In order to analyse this type of variable, the
Poisson or negative binomial model can be used. The theoretical approaches for all
these models are explained below.
5.5.1 Poisson regression model
Poisson distribution is a discrete probability distribution that expresses the
probability of a number of events occurring in a fixed period of time if these events
occur with a known average rate and independently of the time since the last event
(Greene, 1997). In other words, it is used to model the number of events occurring
within a given time interval.The theoretical basis for using this type of count data
models is very important for interpretation of estimation results. Poisson model
expresses the natural logarithm of the event or outcome of interest as a linear
function of a set of predictors. The dependent variable is a count of the occurrences
of interest variables. Typically, one can estimate a rate ratio associated with a given
predictor or exposure. In other words, the typical Poisson regression model expresses
the log outcome rate as a linear function of a set of predictors (Winkelmann, 2008).
For the ith
observation, i = 1 to n, let i denote the mean value of yi given xi.
Supposeix
i e 10 (this insures that i is positive) and yi= i + i , where i is a
random error term. Then
. Thus, there is a “log-linear” relationship
between y and x. Since each yi has a Poisson distribution with mean i , the
probability of yi given xi is:
ii x10)ln(
145
!y
e)y(P
i
y
ii
ii
=!
)( 10
)( 10
i
y
i
x
y
xe ii
(5.23)
where yi is a non-negative integer valued random variable. Estimates of the
coefficients 0 and 1 are obtained by forming the likelihood function and choosing
values of 0 and 1that maximise the likelihood (that maximise the log-likelihood).
That is, 0 and 1 are maximum likelihood estimates. In Poisson regressions, as in
logistic regression, the model deviance is used to measure the goodness of fit of the
Poisson regression model, and the change in deviance is used to test whether 1 is
significantly different from zero (Greene, 1997; Winkelmann, 2008). The functional
form of the parameterisation for the conditional mean can be given as following
Equation 5.24:
)'exp()/( ixxyE (5.24)
The Poisson model assumes that the conditional mean, i , is equal to the conditional
variance. Overdispersion is when the conditional variance exceeds the conditional
mean and is considered to be heteroskedastic (Wooldridge, 2002). The standard
approach of estimating the model is using a form of maximum likelihood estimation,
either using a Newton-Ralphson algorithm or the iterative reweighted least squares,
which is used by the generalized linear model approach (Wooldridge, 2002; Hilbe,
2005).
146
The maximum likelihood estimator (MLE) of the parameter is obtained by
maximising the log likelihood function29
. The Poisson log-likelihood function may
then be derived as follows:
)!ln()ln([);(1
iii
n
i
i yyyl
(5.25)
As
, it can be substituted into above equation.
)!ln()'exp()'exp([);(1
iii
n
i
i yxxyyl
(5.26)
Equation 4.26 can be expresses in terms of the log-gamma function as
)1(ln)'exp()'exp([);(1
iii
n
i
i yxxyyl
(5.27)
The first derivative of the Poisson log-likelihood function, in terms of its coefficient
value can be derived as follows:
)]'exp([1
ii
n
i
i xxyl
(5.28)
Solving for parameter estimates entails setting Equation 5.28 to zero and solving it.
Resulting solution determine the parameter estimates for β. In the estimated model,
the conditional mean function is assumed to be correctly specified and the MLE is
consistent, efficient, and asymptotically normally distributed. Since the mean is equal
to the variance, any factor that affects one will also affect the other. Thus, the usual
assumption of homoscedasticity would not be appropriate for Poisson data.
29
The likelihood function is a transformation of the probability function for which the parameters are
estimated to make the given data most likely.
)exp( ' ii x
)1(ln)!ln( yyi
)]exp([ '
1
iii
n
i
i xxxyl
147
The Poisson regression model is also considered as a non-linear regression to be
estimated using maximum likelihood methods. In the empirical setting, this model is
typically used either to summarise predicted counts based on a set of explanatory
predictors, or for the interpretation of exponentiated estimated slopes, indicating the
expected changes or difference in the incidence rate ratio of the outcome based on
changes in one or more explanatory predictors (Wooldridge, 2002). In this context,
empirical model specification of the Poisson model can be given as follows:
i
i
UIOMASPRIFDSN
DIMKNMAWLHSHLINCGENHMPOWNEXPY
13121110
9876543210
(5.29)
where iY is a count dependent variable that represents the diversity indices, namely
crops or livestock, and all other independent variables are as explained in Table 5.3.
Significant variables in this model will provide important insights into the parameters
that must be taken into account in order to design policies in this field. The
predictions based on this econometric model enable us to profile households that are
most likely to sustain current levels of crops diversity and animal diversity because
they reveal the greatest preference for them.
The regression explaining the richness of all crop varieties grown and animal
varieties maintained in their farms can be estimated using a Poisson regression with
the assumption of mean equals variance. However, if the statistical tests for sample
data reveal overdispersion, a negative binomial model, an extension of the Poisson
regression model that allows the distribution of the variance to differ from the
distribution of the sample mean has to be used (Greene, 1997). Therefore, the
theoretical explanation of negative binomial model is explained in the next section.
148
5.5.2 Negative binomial (NB2) regression model
The assumed equality of the conditional mean and variance functions is typically
taken to be the major shortcoming of the Poisson regression model30
. Many
alternatives have been suggested by different authors (Cameron and Trivedi, 1986).
The most common is the negative binomial model, which arises from a natural
formulation of cross-section heterogeneity. It is clear that the negative binomial
model is employed as a functional form that relaxes the equidispersion restriction of
the Poisson model. Therefore, negative binomial regression is used to estimate count
models when the Poisson estimation is inappropriate due to overdispersion (Hilbe,
2005). It is possible to generalise the Poisson model by introducing an individual,
unobserved effect into the conditional mean as follows31
:
(5.30)
where the disturbance ωi reflects either specification error as in the classical
regression model or the kind of cross-sectional heterogeneity that normally
characterises micro-economic data. Then, the distribution of yi conditioned on xi and
ui remains Poisson with conditional mean and variance i :
!
)(),;(
i
y
ii
u
y
ueuyf
iii
(5.31)
30
In a Poisson distribution the mean and variance are equal. When the variance is greater than the
mean the distribution is said to display over dispersion. Although econometricians have modified the
Poisson regression model to deal with over dispersion, a popular alternative has been the use of the
negative binomial regression model. 31
This is known as the NB2 model because it has a quadratic variance function. The error term reflects
unobserved heterogeneity and is distributed gamma.
iii x 'log
iii ulogloglog
149
This can be assumed as a Poisson model with gamma heterogeneity where the
gamma noise has a mean of one (Greene, 2000). The conditional mean of y under
gamma heterogeneity is thereby expressed as µu rather than as only µ. As a result,
the unconditional distribution )/( ii xyf can be derived from the following
expression:
ii
i
y
ii
u
uugy
ueuxyf
iii
)(!
)(),;(
0
(5.32)
The unconditional distribution of y is specified by the definition of g(u). For this
model a gamma distribution is given u = exp(ε) where (Winkelmann,
2008). Assuming a mean of 1 to the gamma distribution, it is possible to have the
following Equation 5.33:
i
u
i
i
y
ii
u
dueuy
ueuxyf i
iii
1
0)(!
)(),;(
(5.33)
The gamma nature of u is evident in the derivation from above Equation 5.33 to
following Equation 5.34:
i
y
i
u
i
y
i duuey
uxyf iii
i
1)(
0
)(
)()1(),;(
(5.34)
We can continue the derivation further by moving to the left of the integral, with the
remaining terms under the integral equating one. More details about the derivation of
these Equations can be found in Hilbe (2011):
0ln x
150
i
i
y
i
i
i
y
i y
y )(
)(
)()1( (5.35)
Further solution of this integration can be continued as follows:
i
ii
y
i
y
ii
i
i
y
i yy
uxyf
11
)()()1(
),;(
iy
iii
i
y
yuxyf
1
)()1(
)(),;(
iy
iii
i
y
yuxyf
/1
11
/1
1
)()1(
)(),;(
(5.36)
As we derives of the NB2 model, inverting the gamma scale parameter (θ) yields the
negative binomial heterogeneity or over dispersion parameter (α). Accordingly, the
resulting negative binomial probability mass function can be written as follows:
iy
iii
i
y
yuxyf
1
11
1
1
)/1()1(
)/1(),;(
/1
(5.37)
In this form the heterogeneity parameter is inversely related to the amount of Poisson
over dispersion in the data (Hilbe, 2005). When we have in deriving the
parameterisation of the negative binomial, y and α are assumed to be integers.
However, this assumption does not have to obtain when it is used as the
distributional basis of a regression model. As a count data model, the negative
binomial response y should consists of non-negative integer values while α should
take positive rational values (Winkelmann, 2008).
151
The negative binomial model can be estimated by using maximum likelihood
method. The likelihood function for the negative binomial probability function can
be given as follows:
1ln)1(ln
1ln)1ln(
1
1lnexp),;(
1
iii
i
i
i
n
i
yyyyL
(5.38)
The log-likelihood is obtained by taking the natural log of both sides of the Equation
5.37. As with the Poisson models, the function becomes additive rather than
multiplicative. Therefore, log-likelihood function can be written as follows:
1ln)1(ln
1ln)1ln(
1
1ln),;(
1
iii
i
ii
n
i
yyyyl
(5.39)
The negative binomial log-likelihood, parameterised in terms of β (model
coefficients) can be expressed as follows:
1ln)1(ln
1ln)]'exp(1ln[
1
)'exp(1
)'exp(ln),;(
1
i
ii
i
ii
n
i
j
y
yxx
xyyl
(5.40)
Maximum likelihood principles define estimating Equations as the derivatives of the
log-likelihood function. It is clear that ML estimates of the model parameters are
determined by setting the first derivative of the log-likelihood with respect to model
parameters (β) to zero and solving the resulting Equation. As Poisson model is a
variety of the negative binomial model, a test of the distribution is often carried out
152
by testing the hypothesis θ = 0 using the Wald or likelihood ratio test. In the present
study, empirical model specification of the negative binomial model can be given as
follows:
i
i
UIOMASPRIFDSN
DIMKNMAWLHSHLINCGENHMPOWNEXPY
13121110
9876543210
*
(5.41)
where*
iY is a count dependent variable that represent the diversity indices such as
crop diversity or livestock diversity and all other independent variables are as
explained in Table 5.3. Significant variables in this model will provide important
insights into the parameters that must be taken into account in order to design
policies in this field.
As noted in the previous section, the Poisson model imposes the transparently
restrictive assumption that the conditional variance equals the conditional mean. The
typical alternative is the negative binomial model. The model can be motivated as an
attractive functional form simply in its own right that allows over dispersion.
However, in the empirical context, model selection should be done using some
statistical test (Winkelmann, 2008). Therefore, the next section discusses the way of
selecting an appropriate model for the data used in this study.
5.5.3 Empirical tests for different count data models
Count outcomes are commonly encountered in many economic applications, and are
often characterised by a large proportion of zeros. Although Poisson or negative
binomial regression models have often been used to analyse count outcomes, the
153
resulting estimates are likely to be inefficient, inconsistent or biased with the
presence of excess zeros (Hilbe, 2005). Several models belonging to the family of
generalised linear models are available for performing regressions with excess zeros
and dispersion32
(Winkelmann, 2008). For example, zero-inflated Poisson (ZIP) and
zero-inflated negative binomial (ZINB) are specifically developed for count
outcomes with excess zeros and dispersion. Theoretical aspects of using these types
of zero-inflated models are discussed in the Appendix K.
In the empirical model the phenomenon of having zeros can be a concern in this
study. This is because farmers who do not cultivate any crops (only livestock) and
farmers who do not have livestock (only crops) provide zero outcomes for crops and
livestock varieties diversity models respectively. When the farmers have only
livestock, the crop variety index becomes zero while when they have crops only, the
livestock diversity index becomes zero. As discussed in Appendix K, the issue of
excess zeros can be dealt with through the application of ZIP / ZINB regression
models. Besides ZIP or ZINB models, two-part or hurdle models are commonly
applied in count data with excess zeros (Hilbe, 2005).
However, from the preliminary investigation, it was found that farmers who have
only livestock varieties are very few in our samples. It is 7, 10 and 9 per cent of total
samples in Ampara, Anuradhapura and Kurunegala respectively33
. As a result the
excess zero issue was not a problem when estimating crops variety diversity. When
estimating the determinants of animal variety diversity, mixed farming system (both
32
Method of addressing excessive zero counts were first introduced by Lambert (1992). Zero-inflated
models are two-part models, consisting of both binary and count model sections. 33
When estimating crop variety diversity we have included two categories, farmers who cultivate
crops only and farmers who maintain a mixed farming system (crops and livestock).
154
livestock and crops) and farms that have only animals are included. Farms that have
only crops were recorded as zero diversity farms here. The percentage of zero values
in samples of Ampara, Anuradhapura and Kurunegala were 19, 16 and 22. It clearly
shows that there is not an excess zero issue here too. Therefore, for both analyses,
either Poisson or negative binomial model could be used.
In addition to this preliminary observation, one can use the asymptotically normal
Wald type t statistic defined as the ratio of the estimate of α to its standard error. If
the t statistic falls outside (–1.96, 1.96) interval, we reject the null hypothesis that α
equals zero (reject the Poisson model at five per cent significance level). Another
way to test the null hypothesis of α equals zero is to use the likelihood ratio statistic,
which is approximately chi-square distributed with one degree of freedom when the
null hypothesis is true (Hilbe, 2005). Both the likelihood ratio test and the Wald type
t test are asymptotically equivalent (Winkelmann, 2008). In empirical context, both
provide the similar conclusions about selecting the appropriate model.
Grootendorst (1995) introduced steps to choose the best model among the ZINB,
ZIP, NB, and Poisson models. If the Vuong test shows that the ZINB model is
rejected in favour of the NB model, the splitting mechanism with excess zero is
rejected. In this case, we will estimate the NB model and test if the heterogeneity
parameter α is significant by using the t-test; a significant α suggests that
unobservable heterogeneity accounts for dispersion.
On the other hand, if the Vuong test shows that the NB is rejected in favor of the
ZINB model, we will test if the parameter α in the ZINB model is significant. If the
155
estimate of α is also significant, both the splitting mechanism and individual
heterogeneity account for dispersion (Hilbe, 2005). Preliminary test for over
dispersion shows that it is not a problem in district sample data or pool data. This
type of result can be expected due to two reasons. Firstly, the range of crops variety
variation for all data is between zero and nine while animal variety it is zero and five.
It shows low level of variation of our count variables. Secondly, a majority (72 per
cent) of farmers have cultivated three to six crops and two to three animal varieties
(63 per cent). This type of result helps conclude that there is no overdispersion issue
in our data. Therefore, we selected the Poisson model as the best model for
interpreting the results. The STATA version 11.0 as well as Nlogit 4.0 version of the
program was used for the empirical analyses.
5.6 Socio-economic characteristics of the households
We estimated the diversity regression equations for selecting crops varieties and
animal varieties. Most farmers in a given district cultivate or maintain approximately
similar crops or livestock. Rice, different types of vegetables and cash crops are
found to be the common type of crops that farmers cultivate34
. Animal breeds include
manly cattle, chickens, goat, pigs and buffalos. Most households cultivated between
two and six types of crop varieties. In the case of animals, most households
maintained two to three animal varieties in the study areas. On average,
approximately 92 per cent of the sample respondents’ main occupation is farming.
Approximately eight per cent of respondents are employed in the government or
private sector. Their main income source is the salary from the job while an
34
We only included seasonal crops in this analysis. This implies that any variety that takes more than 6
months to harvest is excluded from the survey. Appendix N.1 and N.2 provide the list of crop varieties
and animal breeds which were found in small-scale farms in the study area).
156
agricultural practice is a secondary activity to them. However, some households (32
per cent) have some other source of secondary income in addition to their farm
income. Some farmers (approximately 23 per cent) work as waged labourers on some
days in the month and this provides some additional income for poor farmers to meet
day to day expenses.
Socio-demographic characteristics of the household such as the age, the education of
its members, and household size could be some of the significant factors that
determine the diversity of crops and livestock they grow. However, in the present
study we only use directly policy relevant variables. The average experience of
farming is 24, 19 and 26 years in Ampara, Anuradhapura and Kurunegala samples
respectively. Approximately 78 per cent of respondents are male while 22 per cent
are female. Household members’ participation in agricultural activities is very high
in rural communities in Sri Lanka. Average participating rates were 87, 92 and 96
per cent of the total number of households (greater than 14 years old) in Ampara,
Anuradhapura and Kurunegala districts respectively.
Off farm income is not significant for most households as their main income source
is determined by the farm output. Earnings as a waged labourer, small-scale business
and government family allowance (Samurdi allowance) are among the most common
off farm income sources in rural areas. One of the interesting aspects of rural
households is explained by shared labour. This variable represents the magnitude of
social capital. Average number of shared labour per season is 12, 21 and 18 Ampara,
Anuradhapura and Kurunegala respectively. On average it is approximately 17 per
cent of their labour usage in a given season.
157
According to the criteria that we used to isolate wealthier families from others, a
significant percentage of families belong to other category. For example only 31, 18
and 22 per cent of the respondents were identified as wealthy families in Ampara,
Anuradhapura and Kurunegala district respectively. A significant difference could be
observed in the number of market access days in different districts. It varies 1 to 7
days per week in different districts. There are different types of markets where
farmers could sell their products. One type of market which is commonly called a
‘weekly fair’ could be functioning properly in some villages. In this case farmers
could directly sell their products. However, intermediary traders also come to the
village and purchase various items. Some farmers sell their product to intermediary
traders. In general, informal discussions with farmers reveal that marketing is the
biggest problem for all areas. This is because in some seasons there is no demand for
their product while in other seasons they do not get an expected price.
It was revealed that one of the main objectives of their agricultural activities is to
meet the family food requirement. The marketable surplus of small-scale farms in
rural areas is relatively small. After the harvesting most households maintain a stock
of foods until the next harvesting season approaches. It was found that the
consumption rate of some of the crop and livestock products are as high as 98 per
cent of their output.
Average family consumption rate of rice was approximately 73 per cent while the
consumption rate of some vegetable varieties was approximately 95 per cent. Some
farmers cultivate cash crops for marketing purposes in small-scale farms. The
average marketable surplus of cash crops such as Chilis and Onions were
158
approximately 79 and 86 per cent respectively in the study areas. The distance to the
nearest market is relatively higher in the Anuradhapura district sample. However,
average price fluctuations are similar in all three districts. Moreover, a significant
different could not be observed for receiving subsidies for cultivating crops in
different districts. This is expected as input or output subsidy policies were handled
by the government in Sri Lanka. For example, any farmer who has his own land is
eligible for receiving fertiliser subsidies in any given season.
Given this general information about the respondents, it is interesting to investigate
the results of this analysis. Estimated results are reported with their interpretation in
the next sections. As we were covering three separate districts, data were analysed in
two ways for each model that represents agricultural biodiversity. Firstly, separate
regressions were run for district wise data separately. Secondly, the pool data model
was run after combining three data sets together. A dummy variable is included in
the pool data model to capture the effects of regional fixed factors for Anuradhapura
and Kurunegala, as compared to Ampara. The next section discusses the
determinants of crops variety diversity in separate district data and pool data models.
5.7 Determinants of crops variety demand
As explained in the previous section, we use simple richness measures or counts of
the number of crop varieties the household plants as our basic measures of species
diversity at the household level. In order to model crop species diversity, we use a
Poisson regression because of the discrete, count nature of the dependent variable.
This econometric approach can be linked to the theoretical model through a random-
159
utility framework involving a series of discrete decisions of whether to plant
individual crops (Wooldridge, 2002). In order to check for over or under-dispersion,
the estimated Poisson model was tested against the negative binomial regression
models, resulting in failure to reject the Poisson model. Therefore, we used the
Poission regression for interpreting final results. The more detailed explanation about
the way of selecting the appropriate model using different criteria was given in
Section 5.5.3.
Marginal effects provide a way to measure the effect of each covariate on the
dependent variable. The marginal effect of one covariate is the expected
instantaneous rate of change in the dependent variable as a function of the change in
that covariate, while keeping all other covariates constant. We reported only
marginal effects for all regression models.These coefficients indicate how a one unit
change in an independent variable alters the count dependent variable. For the crops
variety diversity, four Poisson regression models were estimated: three for separate
districts data and one for the pool data for all districts. The estimated results of the
four regression models are reported in Table 5.5.
The results show that experience in farming is highly significant in all models and
has shown a positive coefficient value. This result is not consistent with our initial
hypothesis. We expected that younger households may be more willing to try out
different crops and varieties, while older households with more experience in
farming may be more set in their production activities and less likely to try new crops
and varieties. However, this type of hypothesis can be expected in a more
commercialized farming system. In a semi-subsistence farming system, we found
160
that the farmers who have more experience in farming are likely to maintain a more
diverse farming system. This is because more experienced farmers may have a better
understanding about the benefits of having a diverse farming system than less
experienced farmers. Further, this implies that the human capital and access to
information are favourable for growing a wider range of crop varieties in rural areas
in Sri Lanka. It is also obvious that farmers’ experience is highly correlated with
their age. Therefore, this variable can serve as a proxy for farmers’ age. Owning a
business vehicle is not significant in Anuradhapura sample35
. However, it is a
significant variable for the other three models. The possible implication is that
farmers who have a business vehicle are more likely to maintain a diverse farming
system. This is because having a business vehicle may help reduce market
transaction costs for selling any surplus of their farm.
Household members’ participation variable is highly significant in all models. This
variable shows labour support provided by family members for their farming. It is
clear that more active household labour in agriculture generally contributes
positively to crop diversity. A diverse farming system requires more labour time and
results are consistent with the theory. As hypothesised, households headed by men
grow more diverse varieties. This might be associated with the skill or requirement
for frequent manual work for cultivating more varieties. The influence of this
variable is uniform and significant across all models.
35
Ownership of business vehicle in Anuradhapura sample is relatively smaller than other two samples.
It is 8 per cent in Anuradhapura sample while 21 and 26 per cent in Ampara and Kurunegala samples.
161
Table 5.5: Poisson regression results for crops variety model
Variables Ampara Anuradhapura Kurunegala Pool data
EXP 0.022(0.004)* 0.016(0.003)* 0.018(0.004)* 0.011(0.002)*
OWN 0.325(0.118)* 0.073(0.129) 0.267(0.118)** 0.208(0.096)**
HMP 0.009(0.002)* 0.006(0.001)* 0.007(0.002)* 0.008(0.001)*
GEN 0.181(0.131)**** 0.456(0.121)* 0.235(0.106)** 0.263(0.078)*
INC -0.016(0.006)** -0.004(0.005) -0.002(0.001)** -0.004(0.002)***
SHL 0.033(0.007)* 0.029(0.008)* 0.019(0.007)* 0.037(0.005)*
WLH -0.445(0.115)* -0.222(0.111)** -0.024(0.068) -0.126(0.056)**
NMA 0.152(0.036)* 0.086(0.020)* 0.066(0.023)* 0.154(0.019)*
DIMK -0.124(0.032)* -0.094(0.022)* -0.105(0.025)* -0.097(0.015)*
DSN 0.350(0.112)* 0.647(0.129)* 0.495(0.110)* 0.387(0.068)*
PRIF 0.008(0.002)* 0.002(0.001)*** 0.007(0.001)* 0.004(0.001)*
AS -0.216(0.153)**** -0.444(0.162)* -0.3840.108)* -0.405(0.094)*
IOM 0.021(0.004)* 0.003(0.001)*** 0.009(0.002)* 0.010(0.001)*
Anuradhapura - - - 0.813(0.141)*
Kurunagala - - - 0.286(0.114)**
N 248 247 251 746
Pseudo R2 0.207 0.181 0.256 0.208
Wald chi2(13) 634.37 1206.75 1696.93 2469.98
Note: i. Definitions of the variables used in the regression analysis are shown in the Table 5.3. In the
pool data analysis, Ampara is used as the base district when creating dummy variables.
ii. Standard errors are shown in brackets. *, **, *** and **** denotes the significant variables at
1%, 5%, 10% and 20% level of significance respectively.
iii. Marginal effects on the count dependent variable are reported in this Table. These coefficients
indicate how a one unit change in an independent variable alters the count dependent variable.
Off-farm income of the household has been included, and is measured as the sum of
(the value of) remittances, pension and salary from other employment. This type of
exogenous income can be used to hire labour and purchase other inputs (e.g.,
improved seed) for their cultivation. Off-farm income can release the cash income
constraint faced by some farmers, enabling them to shift their focus from growing
varieties for sale to growing the varieties they may prefer to consume. Moreover,
162
higher off farm income implies that more members of the family are involved in
economic activities other than agriculture. This means less labour availability to
maintain a diverse farming system. In this context, off-farm income may enable them
to specialise in the most profitable crops and varieties. However, literature related to
off-farm income and crop diversity shows ambiguous results. In Mexico, Bellon and
Taylor (1993) found that off-farm employment was associated with higher levels of
maize diversity. Meng (1997) found the existence of off-farm labour opportunities to
have no statistically significant effect on the likelihood of growing wheat landraces
in Turkey.
The result of this study shows that off-farm income has significant negative effect on
crop variety diversity. One of the possible reasons is that when the off-farm income
is higher farmers attempt to purchase most of the food they need for consumption
from the market. Therefore, the incentive for having diverse system, mainly focusing
on family consumption, is less. Another reason can be the labour constraint. A
significant portion of off-farm income comes as off-farm employment. If farmers are
employed in other places, the incentive to maintain a diverse farming system is less
as it needs a relatively higher amount of labour.
Shared labour is another interesting variable used in this analysis. This variable
shows the number of mandates a particular household exchange with other
households during the last crop season. Shared labour is one of the important social
capitals in rural areas in Sri Lanka. This variable shows a significant positive
correlation with crop variety diversity. Shared labour helps reduce the direct cost of
hiring people for agricultural activities.
163
The coefficient on household wealth is negative and significant. The greater the
wealth of the household, the less likely the household is to plant a diverse set of
crops. This finding is consistent with a risk motivation for ‘investing’ in diversity.
Decreasing risk aversion and greater ability to self-insure gives wealthy households
less incentive to invest in a portfolio of crop varieties. The wealth effect is not
necessarily limited to risk. Wealth may be a proxy for networks, information, and
access to outside market opportunities in the presence of various kinds of market
imperfections. In the state of Puebla, Mexico, Van Dusen (2000) found that the
greater the wealth of the household, as measured by house construction and
ownership of durable goods, the less likely the household is to plant a diverse set of
maize, beans, and squash varieties.
The relationship between markets and the conservation of agricultural biodiversity is
complex. As the analysis in this study has shown, higher rates of market participation
are not necessarily associated with higher measures of crop diversity. Sometimes,
higher market participation can contribute to the erosion of crop diversity over time.
Market infrastructure operates in several ways that may not be dissociable in a given
location at one point in time. For example, the more removed a household or
community is from a major market centre, the higher the costs of buying and selling
on the market and the more likely that it relies primarily on its own production for
subsistence. This implies that the more physically isolated a community or
household, is the less specialised its production activities36
. On the other hand, as
36
The theory of the household farm predicts that the higher the transaction costs faced by individual
households within communities as a function of their specific social and economic characteristics, the
more we would expect them to rely on the diversity of their crop and variety choice to provide the
goods they consume. Consistent with this hypothesis, Van Dusen (2000) found that the more distant
the market, the greater the number of maize, beans, and squash varieties grown by farmers.
164
market infrastructure reaches a village, new trade possibilities may emerge, adding
crops and production activities to the portfolio of economic activities undertaken by
its members. We have included four market related variables in this study. They are
the number of market access days per week, distance to the nearest market, direct
sales or not (intermediary) and price fluctuation of the output in the previous season.
An increase in the level of market access can increase level of total diversity in a
farmer’s field. This is because, farmers could maximise their return from diverse
output if they can easily access the market. As expected, the coefficient of this
variable is significant in all four models and has a positive sign. The distance to the
nearest market is another interesting variable used in the analysis. The distance of the
household farm to the nearest market, which is a major component of the cost of
engaging in market transactions related to seed, labour, other inputs, and farm
produce, is hypothesised to affect negatively crop diversity. This means that
households further from markets are less likely to produce a diverse farming system
in a semi-subsistence area. Households further from markets are less responsive to
diversity selection due to the higher transaction cost of market access, which limits
interaction with the market and results in more autarkic behaviour. Households
closer to the market will select more crops as expected, providing evidence of market
participation when transaction costs are low. This is what the result of this study has
shown.
Price fluctuation of output is another interesting market characteristics used in this
analysis. This variable is a proxy for risk of future return of farm output.
Interestingly, market price fluctuation is, as expected, positively related to variety
165
demand. The higher the market price fluctuation, the higher the likelihood that a
household is to cultivate more crops on their farms. This is because, this could help
farmers to minimise the risk of their return. Receiving agricultural subsidies is
another interesting variable used in the analysis. This variable is significant in all
models and has taken negative coefficient value. This implies that agricultural
subsidies are likely to reduce crop diversity in rural areas. This is because most of the
agricultural subsidy scheme in the country focuses on specialisation crops. As a
result, if farmers receive subsidies they have to maintain a single variety system or
specialised system.
The last variable that we included in this model is the percentage of own money
invested for agricultural activities over the last season. As hypothesised, when the
percentage of own money expenditure is higher, their variety selection is also higher.
This coefficient is significant in all models in the analysis. In addition to these
findings, the pool data results show that heterogeneity among districts is significant.
This is expected as we have selected three districts to represent different aspects of
agricultural biodiversity in the country.
In general, the findings suggest that some farm households, market and other
characteristics have a greater impact on variation in crops diversity levels across
small-scale farms in Sri Lanka. Farmers’ choices and cultivation of different crops
diversity and their possible implications for conservation policy are indicated by the
significance of marginal probabilities of the explanatory variables in this analysis. In
the next section, we will investigate important variables for determining animal
variety diversity.
166
5.8 Determinants of livestock variety demand
The development of high-performing livestock and poultry breeds has greatly
contributed to increase food production. Within the agricultural context, animal
biodiversity is the genetic variability (or diversity) between breeds and within breeds
of the same species. However, in this study we only focus genetic variability between
breeds as the variability of the breeds in the same species is not significant in Sri
Lanka. Therefore, as the next step of this analysis we investigate the determinants of
livestock variety demand in separate district data and pool data. We included all
variables which were included in the crop variety model except agricultural subsidy
into this model. The estimated result of the Poisson regression model is given in
Table 5.6.
The results in Table 5.6 show that experience in agricultural activities is highly
significant in Anuradhapura and the pool data model. This variable is significant
under 5 per cent level of significance for samples in Ampara and Kurunegala. All
models show a positive coefficient value implying that farmers who have more
experience in farming are likely to maintain a diverse livestock farming system. This
is expected as livestock farmers need special knowledge to maintain them. Owning a
business vehicle is not a significant determinant of livestock varieties as the
coefficients are not significant in Ampara and Anuradhapura samples while it is
weakly significant in the Kurunegala sample. This is because livestock farms are
mainly maintained for the family consumption purpose in rural areas in Sri Lanka.
Household members’ participation variable is highly significant in all models. This
variable shows labour support provided by family members for their farming. It is
167
clear that more active household labour participation generally contributes positively
to variety diversity. The gender variable is not significant in Ampara and Kurunegala
sample. However, it is significant at 20 per cent and 5 per cent level of significance
for Anuradhapura and the pool data model. The negative coefficient implies that
households headed by women grow more diverse animal varieties. Livestock
diversity is a small-scale business in most areas in the country. Women can easily
manage it from home as it does not need as much attention as crops.
Table 5.6: Poisson regression results for animal variety model
Variables Ampara Anuradhapura Kurunegala Pool data
EXP 0.003(0.001)** 0.018(0.003)* 0.008(0.003)** 0.010(0.001)*
OWN 0.023(0.035) 0.079(0.155) 0.195(0.124)**** 0.085(0.054)****
HMP 0.004(0.001)* 0.003(0.002)**** 0.004(0.001)* 0.004(0.001)*
GEN -0.014(0.035) -0.182(0.118)**** -0.056(0.103) -0.095(0.047)**
INC -0.006(0.001)* -0.019(0.003)* -0.016(0.002)* -0.014(0.001)*
SHL 0.008(0.003)** 0.019(0.010)** 0.026(0.006)* 0.021(0.004)*
WLH -0.096(0.054)*** -0.366(0.133)* -0.591(0.129)* -0.410(0.055)*
NMA -0.117(0.022)* -0.104(0.021)* -0.027(0.025) -0.063(0.011)*
DIMK 0.017(0.001)** 0.032(0.024)**** 0.042(0.016)** 0.029(0.012)*
DSN 0.103(0.037)* 0.162(0.120)**** 0.131(0.123) 0.082(0.051)***
PRIF 0.001(0.000)** 0.003(0.001)* 0.002(0.001)*** 0.002(0.000)*
IOM 0.002(0.000)* 0.009(0.001)* 0.001(0.001)* 0.004(0.001)*
Anuradhapura - - - 0.614(0.091)*
Kurunagala - - - 0.402(0.080)*
N 241 243 242 726
Pseudo R2 0.443 0.297 0.378 0.362
Wald chi2(12) 290.09 254.81 321.11 757.38
Note: i. Definitions of the variables used in the regression analysis are shown in the Table 5.3. In the
pool data analysis, Ampara is used as the base district when creating dummy variables.
ii. Standard errors are shown in brackets. *, **, *** and **** denotes the significant variables at 1%,
5%, 10% and 20% level of significance respectively.
iii. Marginal effects on the count dependent variable are reported in the table. These coefficients
indicate how a one unit change in an independent variable alters the count dependent variable.
168
The results show that off-farm income has a significant negative effect on animal
variety demand. One of the possible reasons is that when the off-farm income is
higher farmers attempt to purchase most of the food they need for consumption from
the market. Therefore, the incentive for having a diverse system, mainly focusing on
family consumption is less. Another reason can be the labour constraint. A
significant portion of off-farm income comes as off-farm employment. If farmers are
employed in other places, an incentive to maintain a diverse farming system is less as
it needs a relatively higher amount of labour. Shared labour is one of the important
social capitals in rural areas in Sri Lanka. This variable shows a significant positive
correlation with animal variety diversity. The coefficient for household wealth is
negative and significant. The greater the wealth of the household, the less likely the
household is to have a diverse set of animals.
The coefficient for the number of market access day’s variable is significant at one
per cent in Anuradhapura and has shown a positive sign. It is less significant in
Ampara and Anuradhapura while no significant result is found in Kurunegala model.
The distance to the nearest market is another variable used in the analysis.
Households further from markets are less responsive to diversity selection due to the
higher transaction cost. Households closer to the market will select more marketed
items, providing evidence of market participation. However, the results show that
households who are living far away from the market are more likely to maintain a
diverse farming system. This shows the subsistence nature of the livestock farming
system in these areas. When the households are away from the market, they are more
likely to maintain a diverse livestock system for their own consumption. In general,
this variable is less significant in this analysis.
169
The variable representing the direct sales or not is weakly significant in the
Anuradhapura sample and not significant in the Kurunegala sample. Price fluctuation
of output is another variable used in this analysis. This variable is a proxy for risk of
future return of output. Interestingly market price fluctuation is, as expected,
positively related to variety demand. The higher the market price fluctuation, the
higher likelihood a household is to maintain diverse livestock system. This is because
this could help farmers to minimise the risk of their return. The last variable that we
included in this model is the percentage of own money invested for farm activities
over the last season. As hypothesised, when the percentage of own money
expenditure is higher, variety selection is also higher. This coefficient is significant
in all models in the analysis. In addition to these findings, pool data results show that
heterogeneity of animal varieties among districts is significant.
The results show that some households, market and other characteristics have a
greater impact on variation in livestock diversity levels across small-scale farms in
Sri Lanka. In the next section the main conclusions drawn from this study is
explained.
5.9 Summary and key findings
A study on the current status of agricultural biodiversity and its determinants is
useful for policy decision makers in order to conserve agricultural biodiversity in
rural areas in the country and hence improve farmer livelihoods. In this context, it is
important to know if farmers promote diversity and what are the determinants of it.
This study investigated this issue using farmers demand for crop and livestock
170
varieties. It is found that maintaining on-farm diversity has received increasing
attention as a strategy for mitigating production risk and protecting food security in
rural areas in Sri Lanka. For poorer farmers with small land holdings, crop and
animal variety diversification increases options for coping with variable
environmental and market conditions. Also due to the existence of imperfect
markets, farmers may grow different varieties to meet their consumption
requirement. On the one hand this practice increases their food security. On the other
hand, it provides more fresh food with high nutrition content. Farmers may also sell
some of the surplus to the market so as to buy their family needs (clothes and other
goods/commodities). This may motivate farmers to grow the varieties that can be
sold in the market for cash.
We find that the key variables promoting diversity are household characteristics,
market characteristics, and some of the other characteristics such as percentage of
own savings invested for agriculture. One of the main conclusions drawn from this
study is that the centrality of markets in shaping diversity does not suggest a trade-
off between development and diversity. This is because as integration with outside
markets increases, the level of diversity on farms can also be increased for crops.
Further, we found that households with more experience, more labour availability
and more foods required for consumption can grow more diverse crops or livestock
because they have the resources to do so. Greater total crops or livestock assets are
associated with greater experience.
In rural farms in Sri Lanka, wealth in livestock can ensure against any crop
production risks that might arise when fewer crops are grown. Households living
171
further away from markets could demand fewer crops or higher livestock breeds.
Access to roads and markets were insignificant factors. Location of farm contributes
to higher levels of crop diversity. However, off-farm income, wealth and agricultural
subsidies were shown to be negatively related with agricultural biodiversity in small-
scale farms in Sri Lanka. Furthermore, output price fluctuations is one of the
important variables that provided significant results in all the models.
Despite the rich agricultural biodiversity in Sri Lanka, the impacts of socio-economic
change upon diverse farming systems in the country has received little attention. This
research has helped to fill the gap by investigating how different forms of market
provisioning and other variables shape the on-farm conservation of agricultural farm
biodiversity in Sri Lanka. It is clear that policies that affect a household’s labour
supply and its composition are therefore likely to have a major impact on most
components of agricultural biodiversity in the country. Educational campaigns, and
recognising the possible importance of women in variety choice and seed
management are also relevant. The information provided by analysis of all models is
directly policy relevant and appropriate policies can be designed to control the
identified factors. The predictions from the models estimated above enable us to
identify the types of families that are most likely to sustain the agricultural
biodiversity. Profiles can be used to design targeted, least cost, incentive mechanisms
to support conservation as part of national environmental programs.
In each statistical analysis conducted, whether descriptive or econometric, the
regional heterogeneity has emerged. Hence, any agri-environmental policy or
programs that aim to support the management of current levels of agricultural
172
biodiversity in rural areas in Sri Lanka will need to recognise the heterogeneity of
these traditional farms and their context. Furthermore, any policy or program that
affects the wealth, education or labour participation of family members, or the
formation of food markets within settlements, will influence their choices. As we
argued in Chapters six and seven, farmers maintain diversity for many reasons other
than those explained in this chapter. There are a number of other reasons and aspects
that we should consider when designing policies in this field. More details and
different aspects of these issues are discussed in Chapters six and seven. In the next
chapter we discuss the farmers’ preference for different farming system such as
organic farming, landrace varieties and mixed farming practices while the efficiency
aspects of small-scale farms are discussed in Chapter seven.
173
CHAPTER SIX
FARMERS’ PREFERENCES FOR DIFFERENT FARMING SYSTEMS
6.1 Introduction
Organic farming and landrace cultivation are increasingly disappearing in most rural
areas in developing countries. Continued landrace loss and disappearing organic
farming methods in developing countries can be attributed to several factors. Firstly,
the diffusion of modern cultivars which, being more productive, under high inputs at
least, rapidly substituted landraces when agriculture became a market-oriented
activity. Secondly, social-economic and cultural transformation of the society has
increased demand for more commercialised farming practice. Thirdly, some of the
other factors include the constant reduction in rural populations, the constant
simplification of productive processes due to high manpower costs and problems
with passing information from one generation to the next are serious threats for the
on-farm maintenance of landraces or existing organic farming methods in rural areas
(Negri, 2003). These factors have significantly changed the traditional mixed
farming system as well.
It is well known that landraces possess a wide range of genes useful for quality
breeding, specialty uses, and their variability of characteristics. The best means of
their conservation is if the materials are still available within the farming system.
However, except for rare cases, there are only several remaining traditional landraces
presently in agriculture. The economic environment of the farm household
significantly determines the extent of genetic diversity in agriculture, selecting
174
organic or mixed farming system. Economic development predominantly had a
negative impact on agricultural biodiversity due to escalating inorganic farming as
well as using modern varieties in specialised farming systems. Since the long term
costs of losing biodiversity rich farming practices is significant, it is important to
understand the influencing factors for selecting landrace, organic and mixed farming
systems in small-scale farms in developing countries.
In some rural areas in Sri Lanka, landraces are still cultivated, mainly with traditional
methods. Compared to commercial varieties, these landraces may be less productive
and more variable, but better adapted to the specific climatic conditions. Moreover,
their product has market desirable quality traits (easy cooking, tasteful). Organic
farmers can profit from the physiological and qualitative characteristics of such
genetic material adapted to local conditions with possible tolerance to diseases and
weed competition. Consumer preferences of high quality product with good
physicochemical characteristics are also an important factor when selecting cultivars
adapted to organic farming (Ghaouti et al., 2008). In this context, the objective of
this chapter of the thesis was to investigate the determinant factors of selecting
organic farming method, landrace cultivation and mixed farming system in small-
scale farms in Sri Lanka. The results will contribute to the better exploitation of local
plant material and give us important information about conservation of landrace
cultivation, organic farming and mixed farming systems which are directly related to
improving agricultural biodiversity in small scale-farms in Sri Lanka.
Farm households who are most likely to maintain farms with landrace cultivation,
organic farming systems and mixed farming systems are described statistically in this
175
study. Findings can assist those who formulate agri-environmental policy in Sri
Lanka to design efficient program that incorporate small-scale family farm
management. The next section provides the context for the present research by
looking at what work has already been done in this field. It critically looks at the
existing research that is significant to the work carried out in this study.
6.2 Literature review on farmers’ preference for different farming systems
There are a number of studies that have analysed farmers’ preferences for landrace
cultivation, organic farming and mixed farming systems in different countries. Brush
et al. (1992) investigated the effects of the adoption of modern varieties of potato on
the diversity of potato varieties on Andean farms. They found that adoption of
modern varieties to be one of the principal causes of agricultural biodiversity loss.
Their findings reveal that farmers adapted only partially to modern varieties of potato
and they continue to employ traditional technologies and to maintain crop diversity
on farms. According to Brush et al. (1992) the loss of biological resources in
agricultural systems due to the introduction of high-yielding varieties is a potential
cost of agricultural development. Their econometric analysis using data from Peru
indicates that the adoption of high-yielding potato varieties results in a reduction but
not a complete loss of biological diversity on individual farms and a possible loss in
aggregate diversity. They concluded that on-site conservation of seed resources may
be a viable complement to the off-site methods now in place.
A study conducted by Brush (1995) presented three cases of on-going maintenance
of landraces by farmers who have also adopted high-input technology, including high
176
yielding crop cultivars. These cases are potatoes in the Andes of Peru, maize in
southern Mexico, and wheat in western Turkey. These cases suggest that on-farm
conservation of landraces can be decoupled from traditional farming practices.
Factors that promote in situ conservation are the fragmentation of land holdings,
marginal agricultural conditions associated with hill lands and heterogeneous soils,
economic isolation, and cultural values and preference for diversity. Landraces are
likely to persist in patches and islands of farming systems in regions of crop
domestication and diversity, and these patches provide potential sites for
conservation programs.
Conventional and ecologically sound agriculture were compared for the specific case
of corn production by Pimentel (1997). As opposed to conventional agriculture, the
ecological agricultural system used manure as a substitute of inorganic fertiliser to
provide soil nutrients. This modified system also adopted tillage to substitute
herbicides and used crop rotation for insect control and no pesticides. In addition to
environmental benefits (e.g. reduced soil erosion and reduced fossil energy
consumption), the modified ecologically-sound system produced higher corn yield
(15.7 per cent more) at a reduced cost (36 per cent less). Heisey et al. (1997)
demonstrated that higher levels of latent genetic diversity in modern wheat varieties
would have generated costs in terms of yield losses in some years in the Punjab of
Pakistan. In other years, the mixed of varieties and their spatial distribution across
the region generated both lower overall yields and less diversity than was feasible.
Tsegaye (1997) looked at crop diversity in Ethiopia and the role that women play in
the development and conservation of crop genetic resources.
177
Howard-Borjas (1999) examined the role of women in plant genetic resource
management and concludes that integration of gender perspectives in plant genetic
resource management programs is necessary if such initiatives are not to fail. Smale
et al. (2001) studied farmers demand for traditional varieties of maize in a region of
Mexico where cultivation of modern varieties of the crop is negligible. They found
that farmers continue to cultivate traditional varieties of maize because they receive
private benefits. Mulatu and Belete (2001) studied the effectiveness of farmers'
participatory varietal evaluation on sorghum crops in the Kile-Bisidimo plains of
eastern Ethiopia for three consecutive years, 1995-1997. The study aimed at
providing farmers with alternatives to their landrace to enable them to overcome crop
losses and to identify farmers' varietal selection criteria for inclusion in future
breeding work. The study also confirmed that increasing farmers' access to their
preferred varieties would result in a faster rate of diffusion through farmer-to-farmer
seed exchange. Benin et al. (2004) pointed out that in less favoured areas such as the
highlands of Ethiopia, farmers manage risk through land allocation to crops and
varieties since they cannot depend on market mechanisms to cope. Farmers also
grow traditional varieties that are genetically diverse and have potential social value.
According to them supporting the maintenance of crop and variety diversity in such
locations can address both the current needs of farmers and future needs of society,
though it entails numerous policy challenges. The result of this study shows that
growing modern varieties of maize or wheat does not detract from the richness or
evenness of these cereals on household farms.
A survey was conducted covering 408 households to understand the role of
socioeconomic, cultural and environmental factors in determining the rice varietal
178
diversity in two contrasting eco sites in Nepal by Rana et al. (2005). The results of
this study suggested that land, livestock number and use of chemical fertiliser have
significant positive influence on landraces diversity on-farm. Other factors like total
land area and membership in farmers’ groups have significant, but negative influence
on landrace diversity. According to them resource-rich households maintain
significantly higher varietal diversity on-farm than that of the resource-poor
households. Reviewing the conservation biology literature, Hole et al. (2005)
conclude that organic farming increases biodiversity at every level of the food chain.
Degraded soil also could be restored through improved agricultural practices. Such
evidence supports the promotion of alternative agricultural practices to achieve
sustainable food supplies.
The degree of urbanisation and the availability of infrastructure contributed more
strongly to genetic erosion as compared to climatic conditions (Keller et al., (2006)).
Farmers’ training encouraged exotic vegetable cultivation and reduced traditional
vegetable diversity. At the same time, indigenous knowledge on how and where to
collect, cultivate and prepare traditional vegetables was disappearing. Mozumder and
Berrens (2007) investigated the empirical relationship between the intensity of
inorganic fertiliser use and biodiversity risk. Using cross-country biodiversity risk
indices, their statistical estimates indicate that the amount of inorganic fertiliser use
per hectare of arable land is significantly related to increasing biodiversity risk.
Robust findings across various specifications hold after controlling for heterogeneity
across countries, including the scale of agricultural production.
179
Sharma et al. (2007) investigated the relationship between landraces and rice
diversity using 183 landraces of rice adapted to the lowlands and the hills in Nepal.
Abdelali-Martini et al. (2008) assessed gender roles as a determinant factor of
managing agricultural biodiversity. According to them increased empowerment
actions of women through alternative sources of income options are needed to
enhance their role in conservation and sustainable use of agricultural biodiversity.
Arslan and Taylor (2008) investigated how shadow prices guide farmers' resource
allocations. They estimated the shadow prices of maize using data from a nationally
representative survey of rural households in Mexico. According to them shadow
prices were significantly higher than the market price for traditional, but not
improved maize varieties.
The CVM was used to document the economic value of crop genetic resources based
on the farmers' willingness to pay for conservation by Diwakar and Johnsen (2009).
A total of 107 households in Kaski, Nepal were surveyed in November 2003. Their
mean willingness to pay was USD 4.18 for in situ and USD 2.20 for ex situ
conservation per annum. Landholding size, household size, education level, socio-
economic status, sex of respondent, number of crop landraces grown, and knowledge
on biodiversity influenced the willingness to pay for in situ conservation, whereas
only landholding size and household size influenced the willingness to pay for ex situ
conservation. The respondents were willing to contribute more for in situ than ex situ
conservation because of the additional effect of direct use and direct involvement of
the farmers in in-situ conservation.
180
According to the above literature review it is clear that a large number of studies
have been conducted to show the benefits of landrace cultivation, organic farming
method and mixed farming system. They have addressed various issues in this field.
However, it is obvious that more empirical work is needed to understand the
determinants of the farmers’ demand for landrace cultivation, organic farming
method and mixed farming system in developing countries. As these farming
practices enhance the agricultural biodiversity, any conservation program that is
targeted to increase the farm level biodiversity should take into account the
influencing factors for maintaining these farming practices. Available studies in this
area also have focused in commercially-oriented farming systems. Therefore, an
analysis of semi-subsistence oriented farming systems is required in order to
generalise and validate the empirical findings. In the next section, the method of
explaining farmers’ preferences is discussed.
6.3 Methods of explaining farmers’ preferences
When economic behaviour is expressed as a continuous variable, a linear regression
model is often adequate to describe the impact of economic factors on this behaviour.
However, there are a variety of economic behaviours where the continuous
approximation is not possible. In such cases binary dependent variable method can
be used to estimates the parameters (Wooldridge, 2002). Probit and logit models are
among the most widely used members of the family of generalised linear models in
the case of binary dependent variables. In probit models, the link function relating
the linear predictor µ= xβ to the expected value µ is the inverse normal cumulative
distribution function, Φ-1
(λ) = µ. In the logit model the link function is the logit
181
transform, ln (λ/1- λ) = µ. Given the similarities between the two types of models,
either model will give identical substantiative conclusions in most application37
. As
sample size in this study is relatively large, we use Probit regression model to
analyse the dummy dependent variables that represent agricultural biodiversity rich
farming systems namely landrace, organic farming method and mixed farming
system.
Bernoulli random variable is the basis of binary choice model (Wooldridge, 2002). If
N observations are available, then the likelihood function of binary dependent
variable can be written as following Equation 6.1:
ii y
i
N
i
y
i PPL
1
1
)1(
(6.1)
The Probit model arises when Pi is specified to be given by normal cumulative
distribution function evaluated at ix' . Let )'( ixF denote the cumulative
distribution function. Then, the likelihood function of Probit models can be given as
following Equation 6.2:
ii y
i
N
i
y
i xFxFL
1
1
)'(1)'(
(6.2)
Then, the log-likelihood function is given by Equation 6.3:
))'(1ln()1()'(lnln1
iiii
N
i
xFyxFyL (6.3)
The first order conditions arising from Equation 6.3 are nonlinear function.
Therefore, we have to obtain the ML estimates using numerical optimisation
37
If one multiplies a Probit estimate by a factor, one gets an approximate value of the corresponding
Logit estimates. Empirical support for the recommendations regarding both the similarities and
differences between the probit and logit models can be traced back to results obtained by Chambers
and Cox (1967). They found that it was only possible to discriminate between the two models when
sample sizes were large and certain extreme patterns were observed in the data.
182
methods. The maximum of likelihood is solved by differentiating the function with
respect to each of the β and setting the partial derivatives equal to zero. Following
Greene (2000) and Gujarati (2003), the empirical model can be generally expressed
as follows:
(6.4)
According to the Equation 6.4, the decision of the ith
farmers to select landrace
cultivation method or organic farming method depends on household, market and
other characteristics. In this model the dependent variables represent whether farmer
selects landrace cultivation (LR), organic production (OP) and mixed farming system
(MIX). Empirical model specification is given in Equation 6.5:
SFIOMASPRIFDSNDIMKNMA
FATWLHSHLINCGENHMPOWNEXPZi
1514131211109
876543210
*
(6.5)
where *
iZ is a dummy dependent variable that represent that represent farmer’s
preference on different farm type. We used eight independent variables in landrace
cultivation and organic farming models and 13 independent variables for mixed
farming model. Significant variables in these models will provide important insights
into the parameters that must be taken into account in order to design policies in this
field. The definitions of the dependent variables are given in Table 6.1.
In order to understand the important determinants of these farming practices,
different types of policy relevant variables are selected. Importance of these variables
were understood by the information gathered from the pilot survey as well as
information provided by the specialist in this area.
'* XZi
183
Table 6.1: Definition of dependent variables in different models
Variables Definitions
LR Whether or not the farm contains a crop variety that has been passed
down from the previous generation and/or has not been purchased from
a commercial seed supplier. Farm contains a landrace vs. farm does not
contain a landrace
OP Whether or not industrially produced and marketed chemical inputs are
applied in farm production
MIX Mixed farms that include crop and livestock production, representing
diversity in agricultural management system
Note: As mentioned previously, these farming systems are common in small-scale farms in Sri Lanka.
All collected variables are divided into three main categories namely household
characteristics, market characteristics and other characteristics. Table 6.2 provides
the definition of all independent variables used in the regression analysis.
All these independent variables are based directly on the questionnaire responses. It
is clear that some variables are taken numbers while other variables are defined as
dummy variables. Experience in farming is one of the important variables used in the
analysis. Experience of household head in agricultural activities is expected to have a
positive relationship with landrace, organic and mixed farming system. This is
because younger households may be more willing to try out modern varieties and
modern farming practice, while older households with more experience in farming
may be more set in their production activities and less likely to try modern farming
practices. Therefore, we hypothesised that demand for the organic farming method,
landrace and mixed farming system would increase with experience in farming.
184
Table 6.2: Definition of potential explanatory variables
Variables Definition
Household characteristics
EXP Experience of farm decision maker (number of years)
OWN Household owns a business vehicle or not: dummy- 1 if Yes, Otherwise 0
HMP Household member’s participation in agricultural activities (%)
GEN Decision maker, male or female: dummy- 1 if Male, Otherwise 0
INC Off farm income of the family (Rs. 000)
SHL Shared labour (number in the last season)
WLH Household wealth: dummy- 1 if wealthier, Otherwise 0
FAT Farmers’ attitudes towards to ABi: dummy- 1 if Positive, Otherwise 0
Market characteristics
NMA Number of market access days per week (number)
DIMK Distance to the nearest market (KM)
DSN Direct sales or not (intermediary) : dummy- 1 if Yes, Otherwise 0
PRIF Price fluctuation of the input(index)ii
Other characteristics
AS Receiving agricultural subsidize: dummy- 1 if Yes, Otherwise 0
IOM Percentage of investment of owned money
SF Size of the farm (hectare)
Note: i. In the questionnaire we asked, ‘what is your general attitude towards agricultural biodiversity’
and possible answer were; very positive, positive, normal, negative and strongly negative. First three
answers were corded as positive while other two were corded as negative when creating dummy
variable.
ii. Price fluctuation indexes were constructed using average unit price changes over the last two
seasons for crops and livestock outputs and inputs.
Gender can give different results as it depends on their preference. Women
household heads are thought to influence selecting landrace and organic farming
method in positive and negative ways. It is expected that women’s conservative
attitudes would contribute towards selecting landrace and organic farming method.
On the other hand, their lack of ability to undertake more labour intensive work may
influence their decisions to grow modern varieties. Farmers’ attitudes towards
185
agricultural biodiversity is an important policy variable used in the analysis. Before
including this variable in the models, a correlation matrix was obtained to test
whether this variable is correlated with other independent variables. It was found that
the correlation coefficients are less than 0.47. Therefore, this variable is included in
the empirical models in order to investigate whether there is an impact of farmers’
attitudes on the conservation of agricultural biodiversity.
Four interesting market characteristics as explained in the Table 6.2 were used to see
whether these variables are important determinants of selecting mixed farming
systems. However, two variables that represent market characteristics are used to see
whether these variables are important for selecting landrace and organic farming
method. It is hypothesised that farmers who are more isolated from markets are more
likely to select organic farming methods and landrace cultivation. In this context, as
the distance to the nearest market is higher, farmers are more likely to maintain
landrace and organic farming methods. Input price fluctuation is an important
variable used in this analysis. This variable was created using average input price
changes (by taking the different between maximum and minimum unit prices) over
the previous two cultivation seasons. It is expected that the coefficient of this
variable has a positive correlation with selecting landrace and organic farming
method. Explanatory variables and their expected signs for different models are
given in Table 6.3.
186
Table 6.3: Explanatory variables and their expected signs
Variable
s
Definitions
LR
OP
MIX
Household characteristics
EXP Experience of farm decision maker + + +
OWN Household owns a business vehicle or not NA NA +
HMP Household member’s participation NA NA +
GEN Decision maker, male or female - + +
INC Off farm income of the family NA NA -
SHL Shared labour NA NA +
WLH Household wealth NA NA -
FAT Farmers’ attitudes towards to AB + + NA
Market characteristics
NMA Number of market access days per week NA NA +
DIMK Distance to the nearest market - - +
DSN Direct sales or not (intermediary) NA NA +
PRIF Price fluctuation of the input/output + + +
Other characteristics
AS Receiving agricultural subsidize - - -
IOM Percentage of investment of owned money + + + SF Size of the farm (hectare) - - NA
Note: i. Only relevant variable are included in each model in order to minimise over identification
problem
ii. Price fluctuation of input is used as explanatory variable in models of landrace cultivation and
organic production decision as well. This variable is created by taking average unit price changes over
the last two seasons for crops and livestock outputs and inputs. However, price fluctuation of output is
used in mixed farming model.
Among the other characteristics, receiving agricultural subsidies, farm’s own
investment in their farm and farm size are important policy relevant variables in the
model. Agricultural input subsidies were crucial instruments in the green revoluation
strategy introduced in the 1960s to increase output and productivity. Agricultural
input subsidies that are known to have an adverse effect on the environment include
pesticides, fertilizers and irrigation. These subsidies provide incentive for farmers to
select specialised crops which are dependent on chemical inputs and irrigation.
Moreover, heavy subsidies on inputs potentially distort the relative costs of factors of
production leading to inefficient allocation of inputs. This applies particularly where
187
inputs are substitutes, rather than cases where they are complementary. Therefore,
receiving agricultural subsidies is used as an important policy variable in this study.
It is expected that receiving agricultural subsidy variable will have a positive impact
on selecting landrace and organic farming method. Farmers who borrowed money
for their cultivation are less likely to select organic and landrace varieties. This is
because farmers often borrow money in order to maintain a specialisation system
with marketing purpose. Size of the farm is expected to have negative impact on
selecting landrace and organic farming method. This is because when the farm size is
larger, farmers are more likely to maintain a specialisation farming system with
modern varieties.
It is clear that the relevance of these variables for selecting landrace cultivation,
organic farming and mixed farming system can be different. As shown in Table 6.3
we used eight independent variables to estimate landrace cultivation and organic
production regression model. However, 13 independent variables are used in the
mixed farm model. Significant variables in these models will provide important
insights into the parameters that must be taken into account in order to design
policies in this field. In the next section, we will investigate the empirical results of
the analysis.
6.4 Factors influencing the selection of landrace cultivation
The loss of diversity in planting materials threatens the livelihoods of millions of
small holders who have local seeds as their major source of planting materials. This
188
is because the loss in diversity weakens the possibility to combine complementary
planting materials which are adaptable to moisture, temperature, and soil type
variability (Chavas and Holt, 1996). It would also reduce the available pool of
genetic materials for breeding to enhance productivity and ensure environmental
stability (Salvatore et al., 2010). Therefore, it is important to understand the main
variables that affect farmers’ decisions for selecting landrace cultivation in rural
areas in Sri Lanka. This section of this analysis uses Probit models to determine
which factors are more likely to contribute to farmers’ decisions on selection
landrace cultivation in their farms. We have included eight important variables which
were explained in Table 6.3 for this purpose. The results of the model estimations are
shown in Tables 6.4.
The results in Table 6.4 show that experience in agricultural activities is significant
for Ampara and pool data models for selecting landrace cultivation. However, this
variable is not significant for Anuradhapura and Kurunagala samples. The gender
variable is significant in all models. The negative coefficient implies that households
headed by women are more likely to use landrace cultivation in their farm. This
shows the conservative nature of women. Household attitude toward the conservation
of agricultural biodiversity is one of the interesting variables used in this analysis.
The estimation results clearly show that household positive attitude toward
agricultural biodiversity is more likely to continue with landrace cultivation. The
coefficient of this variable is highly significant in all models and provides expected
sign. Distance to the nearest market variable is significant in Anuradhapura,
Kurunegala and pool data models under 10 per cent and 1 per cent respectively. The
189
implication is that when the distance to the nearest market is higher, probability of
cultivating landraces is also higher. Meng (1997) also found that cultivation of wheat
landraces was positively associated with their relative isolation from markets in
Turkey. In Andean potato agriculture, Brush et al. (1992) found proximity to markets
to be positively associated with the adoption of modern varieties. In southeast
Guajanuato, Mexico, Smale et al. (2001) found that the better the market
infrastructure in a region the greater the area households allocated to any single
maize landrace but the greater the evenness in the distribution of landraces across the
region. It is clear that the result of this study is consistent with these previous
findings.
Table 6.4: Probit regression results for landrace production model
Variables Ampara Anuradhapura Kurunegala Pool data
EXP 0.017(0.002)* 0.010(0.008) 0.006(0.005) 0.023(0.003)*
GEN -0.232(0.078)* -0.207(0.131)**** -0.332(0.124)* -0.297(0.069)*
FAT 0.174(0.061)* 0.738(0.176)* 0.372(0.109)* 0.338(0.062)*
DIMK 0.002(0.016) 0.012(0.063)*** 0.109(0.033)* 0.065(0.018)*
PRIF 0.002(0.001)*** 0.006(0.003)** 0.017(0.003)* 0.006(0.001)*
AS -0.328(0.065)* -0.738(0.188)* -0.403(0.105)* -0.527(0.050)*
IOM 0.002(0.001)*** 0.004(0.003)**** 0.011(0.003)* 0.005(0.001)*
SF -0.132(0.031)* -0.329(0.215)*** -0.470(0.143)* -0.245(0.041)*
Anuradhapura - - - -0.325(0.082)*
Kurunagala - - - -0.024(0.004)*
N 236 229 232 697
Pseudo R2 0.411 0.929 0.881 0.721
LR chi2(8) 76.69 35.63 66.50 216.39
Note: i. In the pool data analysis, Ampara is used as the base district when creating dummy variables.
ii. Standard errors are shown in brackets. *, **, *** and **** denotes the significant variables at
1%, 5%, 10% and 20% level of significance respectively.
iii. Marginal effects of probit models are reported in the table.
190
Price fluctuation of input is another interesting variable used in this analysis. The
results show that when the market price fluctuation of inputs is higher, the
probability of selecting landrace cultivation is higher. This is expected as input price
fluctuation can increase risk in farming by adding an additional cost component to
farmers. Receiving agricultural subsidies is another interesting variable used in the
analysis. This variable is significant in all models and has taken a negative
coefficient value. This implies that agricultural subsidies are likely to reduce the
probability of having a landrace cultivating system in rural areas.
We also included the percentage of the farm’s own money invested on farm activities
over the last season as an independent variable. This coefficient is significant in all
models in the analysis and has taken the expected sign. It implies that when the
percentage of own money expenditure is higher farmers are more likely to use
landrace systems. The size of the farm is an important variable used in this model.
The coefficient of this variable shows that relatively small farms are more likely to
use landrace cultivation. In addition to these findings, pool data results show that
heterogeneity among districts is significant. In general, the findings suggest these
variables have a greater impact on landrace cultivation across households in Sri
Lanka.
6.5 Factors influencing the selection of organic farming
In this section we investigated the important variables for determining the decision of
having an organic farming system. The econometric results for this model are weaker
statistically because of the smaller percentages of farmers engaged in organic
191
production relative to other models explained in the previous chapters, though they
are consistent with hypotheses based on economic theory. The results of the Probit
models for organic farming method are given by Table 6.5.
Table 6.5: Probit regression results for organic production model
Variables Ampara Anuradhapura Kurunegala Pool data
EXP 0.009(0.002)* 0.033(0.006)* 0.012(0.005)* 0.014(0.002)*
GEN -0.142(0.066)** -0.257(0.129)** -0.396(0.119)* -0.259(0.053)*
FAT 0.115(0.050)** 0.232(0.127)*** 0.278(0.163)*** 0.182(0.054)*
DIMK 0.015(0.012) 0.119(0.026)* 0.111(0.019)* 0.067(0.011)*
PRIF 0.006(0.001)* 0.003(0.001)* 0.009(0.003)** 0.006(0.001)*
AS -0.329(0.102)* -0.317(0.126)** -0.286(0.168)*** -0.314(0.060)*
IOM 0.005(0.001)* 0.005(0.002)* 0.009(0.002)* 0.006(0.001)*
SF -0.052(0.031)*** -0.165(0.083)** -0.038(0.080) -0.077(0.035)**
Anuradhapura - - - 0.135(0.082)***
Kurunagala - - - 0.076(0.027)*
N 233 229 232 694
Pseudo R2 0.396 0.756 0.748 0.624
LR chi2(8) 103.57 127.04 77.32 222.42
Note: i. In the pool data analysis, Ampara is used as the base district when creating dummy variables
ii. Standard errors are shown in brackets. *, ** and *** denote the significant variables at 1%,
5% and 10% level of significance respectively.
iii. Marginal effects of probit models are reported in the table.
It is clear that experience in farming is significant for all models while the gender
variable is highly significant for the Kurunegala and pool data models. This implies
that more experienced farmers are more likely to maintain organic farming systems.
Household attitude toward the conservation of agricultural biodiversity is one of the
interesting variables used in this analysis. The estimation results clearly show that
households with positive attitudes towards agricultural biodiversity are more likely to
continue with organic farming. The coefficient of this variable is highly significant in
192
all models and provides the expected sign. Distance to nearest market variable is
significant for Anradhapura, Kurunegala and pool data models. However, price
fluctuation of input is significant in all models. Results show that when the market
price fluctuation is higher, the probability of selecting organic systems is higher.
Receiving agricultural subsidies is another variable used in the analysis. This
variable is significant in all models and has taken a negative coefficient value. This
implies that agricultural subsidies are likely to reduce the probability of having
organic farms in rural areas. The percentage of own money invested on farm
activities variable was used in this model. This coefficient is significant in all models
in the analysis and has taken the expected sign. The last variable that we used in this
analysis is the size of the farm. The coefficient of this variable shows that small
farms are more likely to use organic farming system. Since organic techniques
require labour to substitute for chemicals in pest and disease control, larger farms
reduce the likelihood that they are used. In addition to these findings, pool data
results show that heterogeneity among districts is significant.
Organic farming has proved to be more cost-effective and eco-friendly than
conventional farming. The nutritional value of food is largely a function of its
vitamin and mineral content. In this regard, organically grown food is dramatically
superior in mineral content to that grown by modern conventional methods. A major
benefit to consumers of organic food is that it is free of contamination with health
harming chemicals such as pesticides. It is also known that organically grown food
tastes better than conventionally grown food. Furthermore, organically grown foods
193
can be stored longer and do not show the latter’s susceptibility to rapid mold and
rotting.
The survey results in this study show that organic farming reduces the production
cost by about 30 – 40 per cent since it does not involve the use of chemical fertilisers
and pesticides, which thus makes organic farming very cost-effective. There is a
discerning market of consumers who recognise the greater food value of organic
produce and are willing to pay premium prices for it. However, the existence of a
price premium for organic products are not significant in Sri Lanka. Although there
are some significant benefits of organic farming, it has a cost premium as well.
Organic farming requires greater interaction between a farmer and his/her crop for
observation, timely intervention and weed control for instance. It is inherently more
labor intensive than chemical/mechanical agriculture so that, naturally a single
farmer can produce more crops using industrial methods than he/she could by solely
employing organic methods. Organic farmers do not have a convenient chemical fix
on the shelf for every problem they encounter. A detailed analysis of these costs and
benefits are beyond the scope of this study.
In general, the findings of the analysis in Section 6.4 and 6.5 suggest that these
variables have a great impact on selecting landrace system and organic farming
system across small-scale farms in Sri Lanka. Farmers’ choices for landrace
cultivation as well as organic farming systems and their possible implications on
conservation policy are indicated by the significance of marginal probabilities of the
explanatory variables in the models. It is clear that these findings can assist those
who formulate agri-environmental policies in Sri Lanka to design efficient program
194
that incorporate small-scale farm management. In the next section farmer’s demand
for mixed farming system is explained.
6.6 Farmers’ demand for mixed farming system
Risk exposure and risk management are inherent components of agricultural
activities. Farmers face various forms of risks, ranging from vagarious climatic
conditions, pests and pathogens, and price volatility. In the presence of efficient
insurance markets, farmers may insure themselves effectively to manage these risks.
However, in the absence of perfect insurance markets, as is often the case in
developing countries, exposure to such risks is likely to affect the ex-ante production
choices (Fafchamps, 1992; Chavas and Holt, 1996; Kurosaki and Fafchamps, 2002).
In developing countries, farmers' choice for farm diversification may reflect an
insurance mechanism designed to reduce production risk. A growing body of
research suggests that mixed farming system contributes to increase agricultural crop
yield, and to reduce production risk (Smale et al., 1998; Di Falco and Chavas, 2009;
Salvatore et al., 2010). In this section we investigate the determinants of mixed farms
in separate district data and pool data. The dichotomous choice of whether or not to
raise crops together with livestock in the farm is estimated with a univariate probit
model.
Table 6.6 presents the results of the mixed farms regression model. The decision to
maintain a mixed farming system is assumed to be a function of household
characteristics, market characteristics and some of the other characteristic. Results
show that most of the included variables are significant for determining mixed
195
farming systems. It is also evident that for all regions taken together, household
characteristics as a set are highly significant determinants of the decision to raise
both crops and livestock when comparing with other characteristics.
Table 6.6: Probit regression results for mixed farm model
Variables Ampara Anuradhapura Kurunegala Pool data
EXP 0.007(0.003)*** 0.018(0.009)*** 0.022(0.008)* 0.009(0.003)**
OWN 0.106(0.080)**** 0.121(0.175) 0.165(0.156) 0.187(0.067)*
HMP 0.006(0.002)** 0.017(0.005)* 0.006(0.003)*** 0.008(0.001)*
GEN 0.122(0.090)**** 0.280(0.179)**** 0.571(0.151)* 0.153(0.081)***
INC -0.009(0.003)** -0.014(0.004)* -0.003(0.006) -0.008(0.002)*
SHL 0.014(0.006)** 0.148(0.035)* 0.060(0.013)* 0.043(0.007)*
WLH -0.005(0.065) -0.486(0.162)* -0.186(0.145) -0.161(0.064)**
NMA 0.130(0.030)* 0.111(0.049)** 0.065(0.035)** 0.058(0.024)**
DIMK -0.061(0.025)** -0.232(0.069)* -0.072(0.027)* -0.009(0.004)*
DSN 0.276(0.177)**** 0.004(0.245) 0.099(0.136) 0.152(0.076)**
PRIF 0.004(0.001)*** 0.007(0.002)* 0.014(0.003)* 0.005(0.001)*
AS -0.274(0.112)** -0.418(0.194)** -0.646(0.128)* -0.368(0.064)*
IOM 0.004(0.001)** 0.015(0.003)* 0.004(0.002)**** 0.007(0.001)*
Anuradhapura - - - 0.137(0.091)****
Kurunegala - - - 0.160(0.072)**
N 248 247 251 746
Pseudo R2 0.879 0.894 0.833 0.806
LR chi2(13) 92.49 48.25 118.95 129.54
Note: i. In the pool data analysis, Ampara is used as the base district when creating dummy variables.
ii. Standard errors are shown in brackets. *, **, *** and **** denotes the significant variables
at 1%, 5 %, 10% and 20% level of significance respectively.
iii. Marginal effects of probit models are reported in the table.
The results in Table 6.6 show that experience in agricultural activities is highly
significant in all models and shows a positive coefficient value implying that farmers
who have more experience in farming are likely to maintain mixed farm. The reason
196
may be, with the experience, they can understand the possible benefits of having a
mixed farming system. Owning a business vehicle is weakly significant in Ampara
sample while it is highly significant for pool regression model. It is clear that
business vehicles help farmers reduce the transaction costs for marketing output.
Household members’ participation variable is highly significant in all models. It is
clear that more active household labour participation generally contributes positively
to maintain mixed farming systems. The gender variable is significant in all models.
The positive coefficient implies that, households headed by men maintain more
diverse or mixed farming systems.
The results show that off-farm income has a significant negative effect on mixed
farms. This is expected when considering family food requirement as well as labour
requirements. It is clear that a significant portion of off-farm income comes as off-
farm employment. If they are employed in other places, the incentive to maintain a
diverse farming system is less as it needs a relatively higher amount of labour. As
mentioned previously, shared labour is one of the important social capitals in rural
areas. This variable shows a significant positive correlation with mixed farming
systems. The coefficient for household wealth is negative and significant. The greater
the wealth of the household, the less likely the household is to have a mixed farming
system. The coefficient for the number of market access day’s variable is significant
in all models and has shown a positive sign. The distance to the nearest market is
another interesting variable used in the analysis. The results show that households
who are close to the market are more likely to maintain mixed farming systems. This
is because their transaction costs are likely to be less. When the households are away
197
from the market, they are less likely to maintain diverse farming systems as their
market transaction cost is expected to be high.
The variable representing direct sales or not is significant only Ampara and pool data
models. This implies that farmers who can directly sell their output are more likely to
maintain a diverse farming system. Price fluctuation of output is another interesting
variable used in this analysis. This variable is a proxy for risk of future return of farm
output. Results show that when the market price fluctuation is higher, the probability
of selecting mixed farming systems is higher. This is expected as it shows the way of
managing risk of the farm. This could help farmers to minimise the risk of their
return. Receiving agricultural subsidies, another variable used in the analysis, is
significant in all models and has taken a negative coefficient value. This implies that
agricultural subsidies are likely to reduce the probability of having a mixed farming
system in rural areas. The last variable that we included in this model is the
percentage of own money invested for farm activities over the last season. As
hypothesised, when the percentage of own money expenditure is higher, the
probability of selection of a mixed farming system is also higher. This coefficient is
significant in all models in the analysis. In addition to these findings, pool data
results show that heterogeneity among districts is significant.
In general, the findings of this analysis suggest that household, market and other
characteristics have a great impact on determining mixed farms levels across small-
scale farms in Sri Lanka. Farmers’ choices on selection of mixed farming systems
and their possible implications for conservation policy are indicated by the
198
significance of marginal probabilities of the explanatory variables in this analysis. In
the next section the main conclusions drawn from this chapter are explained.
6.7 Summary and key findings
Although the benefits of environmentally rich farming systems in Sri Lanka are
clear, the impacts of socio-economic change upon agricultural biodiversity in the
country have received little attention. A study on the current status of agricultural
biodiversity is useful for policy decision makers in order to make policies for
conservation in rural areas in the country. It is clear that the different farming
practices that farmers use is directly related with agricultural biodiversity. Therefore,
it is important to know the determinant factors for selecting landrace cultivation,
organic farming and mixed farming systems. This chapter of the thesis investigated
this issue using small-scale farms in Sri Lanka. We found that the key variables
promoting landrace cultivation, organic farming and mixed farming systems are
household characteristics, market characteristics, and some of the other
characteristics such as percentage of farmers’ own money spent for agriculture.
The results show that gender is an important variable to determine the landrace
cultivation. It shows that female dominant farms are more likely to select landrace
varieties. Farmers’ positive attitudes towards agricultural biodiversity have a
significant impact on selecting landrace varieties. In addition to that farms size, input
price fluctuations, agricultural subsidies and percentage of own money investment
are found to be among important factors when taking decisions related to
maintaining landrace cultivation. An interestingly agricultural subsidy is one of the
199
important variables that provided significant results in all models. It implies that the
existing subsidy program in Sri Lanka has negatively affected choices about
cultivating landrace varieties. Investigation of profiles of farm families that are most
likely to cultivate landraces and use organic farming reveals that they have less
income compared to those farm families who are not likely to cultivate landraces.
They are more agriculturally-based, with less off-farm employment and are more
isolated from the markets.
Among the important variables in the organic farm model, farmers’ attitudes towards
agricultural biodiversity, input price fluctuations, agricultural subsidies and farm size
are found to be the most significant variables. It is clear that most of the variables
used in the mixed farm model are significant and have taken expected signs. We
found that households with more experience, more labour availability and less off
farm income are more likely to have a mixed farming system. The results also show
that the market characteristics as well as agricultural subsidies are important
determinants for selecting mixed farming systems. Off-farm income, wealth and
agricultural subsidies have been shown to be negatively related to mixed farming
systems in small-scale farms in Sri Lanka.
The information provided by analysis of all models is directly policy relevant and
appropriate policies can be designed to control them. The predictions from the
models estimated above enable us to identify the types of families that are most
likely to increase the agricultural biodiversity in Sri Lanka. Accordingly, household
profiles can be used to design targeted, least cost incentive mechanisms to support
conservation as part of the national environmental program. This study contributes to
the literature by providing insights into farmers’ landrace cultivation, organic
200
farming and mixed farming preferences, using small-scale farm household data in a
typical developing country setting. It also identifies the household contextual factors
that govern these decisions.
201
CHAPTER SEVEN
AGRICULTURAL BIODIVERSITY AND FARM LEVEL EFFICIENCY
7.1 Introduction
Technological innovation and the more efficient use of production technologies are
the main strategies of achieving productivity growth in agriculture (Hoang and Coelli,
2009). However, in developing countries most new agricultural technologies have
only been partially successful in improving productivity. This is often due to a lack
of ability or desire to adjust input levels by the producers because of their familiarity
with traditional agricultural systems or because of institutional constraints (Binam et
al., 2004). These considerations suggest that the best option to assist developing
countries to raise productivity is increasing efficiency. If farmers are not effectively
using existing technology, then efforts designed to improve efficiency may be more
cost-effective than introducing new technologies (Belbase and Grabowski, 1985).
The presence of shortfalls in efficiency means that output can be increased without
requiring additional conventional inputs and without the need for new technology. If
this is the case, empirical measures of efficiency are needed to determine the
magnitude of the gains that could be obtained by improving performance in
agricultural production with a given technology. In this chapter of the thesis farmers’
ability to select a production system and its relationship with farm level technical
efficiency is investigated.
202
There are several important reasons for measuring the farm level technical efficiency
of agricultural production. Firstly, if farmers are not making efficient use of existing
technologies, then efforts designed to improve efficiency would be more cost
effective than introducing a new technology as a means of increasing output
(Shapiro, 1983). Secondly, measuring efficiency leads to sustainable resource
savings, which has important implications for both policy formulations and farm
management (Bravo-Ureta and Evenson, 1994). Thirdly, it is only through measuring
efficiency and separating its effects from the effects of the production environment
that one can explore hypotheses concerning the sources of efficiency differential.
Fourthly, identification of sources of inefficiency is important to the institution of
public and private policies designed to improve performance of agriculture (Bozoglu
et al., 2007).
Biodiversity conservation of agricultural land is an objective that has received a
considerable attention from policy makers in recent years (Widawsky and Rozelle,
1998; Winters et al., 2005). This is because agricultural production can play an
important role on maintaining environmental friendly farming system in the long run.
Moreover, experience shows that production can be intensified (more production per
unit of area) while reducing inputs and lowering the environmental degradation in
agriculture through improving biodiversity in the agricultural sector (Winters et al.,
2005). However, enhancement of biodiversity appears not to be explicitly recognised
as a proper target or a positive output when production efficiency is measured in
practice. We hypothesise that this ignorance may cause biases in traditional
efficiency calculations and such incomplete measures may therefore discriminate
against environmentally benign technologies.
203
This Chapter of the thesis aims at delivering empirical evidence on the links between
technical efficiency and agricultural farm biodiversity by analysing farm level data
collected from 746 small-scale farmers in Sri Lanka. To the best of our knowledge,
this is the first attempt that investigates farm level technical efficiency and
biodiversity in Sri Lanka or any other country. It is believed that agricultural
biodiversity increases farm level technical efficiency due to three reasons. First,
farmers believe that they can utilise family labour optimally when they maintain a
diverse agricultural system (Brookfield et al., 2002). For example, different crops
may require labour in different time periods and family labour can easily be
distributed among different crops and/or livestock in order to obtain maximum
benefits. Second, a diverse farming system minimises external risk that farmers often
face. For example, if a farmer has both crops and livestock this will minimise the risk
from drought or water shortage. That is, while crops can be devastated, the farmer
still can derive an income from livestock. Third, a biologically rich farming system
can improve soil fertility and minimise input costs in the long run.
The next section will summarise the existing empirical studies of agricultural
biodiversity and farm level technical efficiency. This type of analysis helps to
identify what work has already been undertaken in this field. It also helps in
understanding the shortcomings of existing work and highlights the importance of
conducting the present research. As shown in the literature review, no studies in this
area analyses the relationship between agricultural biodiversity and farm level
technical efficiency using small-scale farms data in developing countries. Therefore,
the results of the study will be a novel contribution to the literature in this area.
204
7.2 Literature on agricultural biodiversity and farm level efficiency
Agricultural biodiversity is found to have some positive impacts on overall
productivity and soil quality (Heisey et al., 1997; Widawsky and Rozelle, 1998;
Meng et al., 2003). It also can affect farm level efficiency through the management
of scare resources in a diverse farming system. Belbase and Grabowski (1985)
estimated a deterministic Cobb-Douglas production frontier model to investigate
efficiency in Nepalese agriculture. According to this study, average technical
efficiency level of mainstream agricultural crops is found to be 80 per cent. Based on
the efficiency measures obtained from all crops, correlation analysis showed that
nutritional levels, income, and education were significantly related to technical
efficiency, while no relationship was found for farming experience. Parikh and Shah
(1995) presented a review of the various approaches to efficiency measurement and
conducted empirical analyses of cross-sectional data from 397 sample farmers in the
North-West Frontier Province of Pakistan. Their results show that small farms were
relatively more efficient than large farms in the study area.
The technical efficiency and productivity of maize producers in Ethiopia were
analysed by Seyoum et al. (1998). This study compared the performance of farmers
within and outside the program of technology demonstration. Using Cobb-Douglas
stochastic production functions, their empirical results showed that farmers who
participate in the program are more technically efficient with a mean technical
efficiency equal to 94 per cent compared with those outside the project whose mean
efficiency was equal to 79 per cent. Smale et al. (1998) found that the production
environment determines the sign of the relationship between diversity and
205
productivity for wheat varieties in the Punjab of Pakistan. For instance, among
rainfed districts, genealogical distance and a greater number of different varieties
grown of smaller areas were associated with both higher mean yields and more yield
stability. New evidence on technical efficiency and its sources were presented by
examining the cost behaviour of 387 farms in five irrigated districts of Punjab by
Burki and Shah (1998). They concluded that farm efficiency is positively related to
formal schooling of farm operators and negatively related to farm size. The age of
farm operators is shown to have no effect on efficiency. Dairy farms are also the
subject of a paper by Hadri and Whittaker (1999) where the efficiency of a small
panel of dairy farms in the south-west of England was considered in the context of
their use of potentially polluting agrochemicals. This study showed a positive
relationship between technical efficiency and farm size. However, there is a
negligible negative relationship between farm size and use of contaminants in farms.
A stochastic production frontier methodology was used to investigate the technical
efficiency of organic and conventional olive-growing farms by Vangelis et al.
(2001). Findings indicated that the organic olive-growing farms examined exhibit a
higher degree of technical efficiency (relative to their production frontier) than do
conventional olive-growing farms. Reasons may include lower profit margins and
restrictions on inputs permitted, thus forcing organic farmers to be more cautious
with input use. However, both input and output-oriented technical efficiency scores
were still relatively low for both types of olive-farming. Wilson et al. (2001)
examined the technical efficiency of a cross-section of cereal farmers in Eastern
counties. According to them, the technical efficiency index across production units
ranged from 62 to 98 per cent. The objectives of maximising annual profits and
206
maintaining the environment are positively correlated with, and have the largest
influence on, technical efficiency. Moreover, those farmers who seek information,
have more years of managerial experience, and have a large farm, are also associated
with higher levels of technical efficiency.
The efficiency of smallholder rice farmers were investigated by Sherlund et al.
(2002) in Côte d’Ivoire while controlling for environmental factors that affect the
production process. Apart from identifying factors that influence technical
efficiencies, the study found that the inclusion of environmental variables in the
production function significantly changed the results: the estimated mean technical
efficiencies increased from 36 per cent to 76 per cent. Karagiannis et al. (2002) also
analysed the efficiency of dairy farms in England and Wales. Binam et al. (2004)
examined factors influencing technical efficiency of groundnut and maize farmers in
Cameroon. Using a Cobb-Douglas production function they find mean technical
efficiencies to be in the region of 73 per cent and 77 per cent. They also concluded
that access to credit, social capital, and distance from the road and extension services
are important factors explaining the variations in technical efficiencies. Testing the
relationship of wheat variety diversity to productivity and economic efficiency in
China, Meng et al. (2003) found that although evenness in morphological groups
contributed to higher per hectare costs of wheat produced, potentially important cost
savings were apparent for some inputs, such as pesticides. A greater concentration of
cooperative market associations in regions of southern Italy contributed to greater
diversity of durum wheat varieties, with positive effects on productivity (Di Falco,
2003).
207
Hadley (2006) estimated stochastic frontier production functions for eight different
farm types (cereal, dairy, sheep, beef, poultry, pigs, cropping and mixed) for the
period 1982 to 2002. Differences in the relative efficiency of farms were explored by
the simultaneous estimation of a model of technical inefficiency effects. The results
showed that factors such as farm or herd size, farm debt ratios, farmer age, levels of
specialisation and ownership status are significant variables in the efficiency
function. Idiong (2007) provided estimates of technical efficiency and its
determinants using data obtained from 112 small scale rice farmers. The results
indicated that rice farmers were not fully technically efficient. The mean efficiency
obtained was 77 per cent indicating there was a 23 per cent allowance for improving
efficiency. The results also showed that farmers’ educational level, membership of
cooperative/farmer association and access to credit significantly and positively
influenced the farmers’ efficiency.
A study conducted by Bozoglu and Ceyhan (2007) explored determinants of
technical inefficiency in the Samsun province of Turkey. Farm managers from 75
randomly selected farms were interviewed for farm level data in the 2002-2003
production periods. Research results revealed that the average output of vegetable
farms in Samsun could increase by 18 per cent under prevailing technology. The
technical efficiency of the sample vegetable farms ranged from 0.56 to 0.95 (0.82
average). The variables of schooling, experience, credit use and participation by
women negatively affected technical inefficiency. However, age, family size, off-
farm income and farm size showed a positive relationship with inefficiency.
208
There are a few studies that indirectly concentrate on agricultural biodiversity and
farm level efficiency. Czech (2003) investigated the role of technology in agriculture
in conserving biodiversity. Latruffe et al. (2004) analysed technical efficiency and its
determinants for a panel of individual farms in Poland specialised in crop and
livestock production in 2000. Technical efficiency is estimated using stochastic
frontier analysis and the determinants of inefficiency are also evaluated. Latruffe et
al. (2005) analysed the technical and scale efficiency of Polish farms using data
envelopment method. Efficiency differences are measured according to farm
specialisation, in crop or livestock, at two points in time during transition, 1996 and
2000. Their findings indicate that livestock farms are on average, more technically
and scale efficient than crop farms. Scale efficiency is high for both specialisations.
Haji (2006) estimated technical, allocative and economic efficiencies and identifies
their determinants in smallholders’ vegetable-dominated mixed farming system of
eastern Ethiopia. An econometric analysis using a Tobit model indicates that asset,
off-farm income, farm size, extension visits and family size were the significant
determinants of technical efficiency. On the other hand assets, crop diversification,
consumption expenditures and farm size had a significant impact on allocative and
economic efficiencies.
According to the above review of previous studies, it becomes clear that a large
volume of literature deals with farm level technical efficiency in various contexts.
However, none of the studies consider the causal relationship between agricultural
biodiversity and farm level technical efficiency in a semi-subsistence economy. As a
result, there is a need to focus on small farms, the primary farming system in Asia,
Africa and Latin America. This study attempts to fill these gaps in the literature. The
209
primary focus of the chapter is to investigate the link between agricultural farm
biodiversity and farm level technical efficiency. It is also expected to identify some
of the other factors that affect inefficiency in small-scale farms in rural areas in Sri
Lanka. The results of this study will provide the necessary information for
policymakers to evaluate the social benefits of conservation of agricultural
biodiversity in rural areas in developing countries. The relevant theoretical and
empirical approaches are explained in the following section.
7.3 Method of estimating farm level technical efficiency
The term efficiency of a farm can be defined as its ability to provide the largest
possible quantity of output from a given set of inputs. The modern theory of
efficiency dates back to the pioneering work of Farell (1957) who proposed that the
efficiency of a farm consists of technical and allocative components, and the
combination of these two components provides a measure of total economic
efficiency. Technical efficiency measures how well the individual farm transforms
inputs into a set of outputs based on a given set of technology and economic factors
(Aigner et al., 1977; Kumbhakar and Lovell, 2000). It is measured either as input
conserving oriented or output-expanding orientation (Jondrow et al.,1982; Coelli,
1995). Accordingly, this section begins with a description of the basic stochastic
production frontier model, where output is specified as a function of a non-negative
random error which represents technical inefficiency, and a symmetric random error
which accounts for statistical noise. It also shows how the estimated parameters of
the model can be used to predict the technical inefficiencies of farms.
210
We use the stochastic frontier production function approach to estimate farm level
technical efficiency in small-scale farms in Sri Lanka38
. The advantage of using
stochastic frontier models are: (1) It introduces a disturbance term representing
statistical noise, measurement error and exogenous shocks beyond the control of
production units which would other-wise be attributed to technical inefficiency, (2) It
provides the basis for conducting statistical tests of hypothesis regarding the
production structure and the degree of inefficiency. The estimation of frontier
function and efficiency can be completed either in one stage or in two stages. This
method has been used extensively in the past two decades to analyse technical
efficiency. In this study, the model of Battese and Coelli (1995) is used in
accordance with the original models of Aigner et al. (1977) and Meeusen and van
den Broeck (1977). The general form of the stochastic frontier production can be
defined by:
iiii UVxfY exp, i = 1,2.............N (7.1)
Yi refers to the output obtained by farm i, xi is the vector of different inputs used and
β is a vector of parameters to be estimated. The model is such that the possible
production, Yi, is bounded above by the stochastic quantity,
.
Therefore,the term stochastic frontier is used. The error components Vi are assumed
to be independently and identically distributed as ),0( 2
vN . This is associated with
random factors such as random errors, errors in the observation and measuring of
data, which are not under the control of the farm (Coelli et al., 2005). The error
components, Ui are non-negative truncations of the ),0( 2
uN distribution that can be
38
Coelli (1996) observed that 30 out of 40 studies on application of frontier models to agriculture have
used stochastic frontier production functions.
)exp(),( ii Vxf
211
half normal, truncated normal, exponential distribution or gamma distribution. The
truncated normal frontier model is due to Stevenson (1980) while the gamma model
is due to Green (1990). The log-likelihood functions for these different models can
be found in Kumbhakar and Lovell (2000). The model explained by Equation 7.1 can
be expressed as follows:
iiii UVXY lnln 10 (7.2)
iiii UVXY lnexp 10 (7.3)
iiii UVXY expexplnexp 10 (7.4)
First component of the right hand side of Equation 7.4 gives the deterministic
component while second and third components give noise and inefficiency parts. The
basic structure of the stochastic frontier model is explained in Figure 7.1 in which the
productive activities of two farms, represented by i and j, are considered.
Source: Coelli et al. (2005).
Observed output, iY
Observed output, jY
xi xj Inputs X
×
Figure 7.1: Stochastic frontier production function
×
×
×
Deterministic production
function, Y = f(X;β)
Frontier output,
*
iY if 0iV
Output
Y
Frontier output,
,*jY if 0jV
),( jxfY
),( ixfY
212
In this case, the deterministic component of the frontier model has been drawn to
reflect the existence of diminishing returns to scale. Values of the input are measured
along the horizontal axis and outputs are measured on the vertical axis. Farm i uses
inputs with values given by the vector of xi and producers output Yi, but the frontier
output Yi*, exceeds the value on the deterministic production function, ),( ixf ,
because its productivity is associated with favourable conditions for which the
random error, Vi is positive. However, farm j uses inputs with values given by the
vector xj and producers output, Yj, which has corresponding frontier output, Yj*,
which is less than the value on the deterministic production function,
),( ixf ,
because its productive activity is associated with unfavourable conditions for which
the random error Vj is negative. In both cases the observed production values are less
than the corresponding frontier values, however, the (unobservable) frontier
production values lie above or below the deterministic production function
depending on the existence of favourable or unfavourable conditions beyond the
farms’ control (Coelli et al., 2005).
Accordingly, random variables Ui are assumed in capturing technical inefficiency.
Given the assumptions of the stochastic frontier model, inference about the
parameters of the model can be based on the maximum likelihood estimators (Aigner
et al., 1977). The parameter γ can be calculated using information of the variance of
two error terms (2
u and2
v ). More details about the method of obtaining parameters
are given by Coelli et al. (2005). Given the assumptions of the stochastic frontier
model, inference about the parameters of the model can be based on the maximum
likelihood estimators (Aigner et al., 1977). Battese and Corra (1977) considered the
213
parameter,22vu ,
which is bounded between zero and one.
It is clear that22
vu , the coefficient of
is bounded between
zero and one. If the γ equals zero, the difference between farmers yield and efficient
yield is entirely due to statistical noise. On the other hand γ = 1 indicate the
difference is entirely due to less than efficient use of technology (Coelli et al., 2005).
If 22
vu , then the more δ is greater than one, the more production is dominated
by technical inefficiency. The closer it is to zero, the more the discrepancy between
the observed and frontier output is dominated by random factors beyond the control
of the farmer (Coelli, 1995). The technical efficiency of individual farms can be
estimated by using the conditional distribution of Ui given the fitted values of error
term and the respective parameters.
The technical efficiency of an individual farm is defined in terms of the ratio of the
observed output to the corresponding frontier output, conditional on the levels of
inputs used by that farm (Coelli and Battese, 1996). It is the factor by which the level
of production for the farm is less than its frontier output. The technical efficiency of
farm in the context of the stochastic frontier production function is the same
expression as for the deterministic frontier model (Coelli et al., 2005). Although the
technical efficiency of a farm associated with the deterministic and stochastic frontier
models are the same, they have different values for the two models (Battese, 1992).
As shown in figure 7.1, technical efficiency of farm j is greater under the stochastic
frontier model than for the deterministic frontier. However, for a given set of data,
the estimated technical efficiencies obtained by fitting a deterministic frontier will be
less than those obtained by fitting a stochastic frontier, because the deterministic
)/( 222vuu
)/( 222vuu
214
frontier will be estimated such that no output values will exceed it (Battese, 1992).
Given the deterministic frontier model, the frontier output for the ith
farm is
, and the technical efficiency for the i
th farm, denoted by TEi
is that:
*
i
ii
Y
YTE
(7.5)
(7.6)
(7.7)
Technical efficiencies for individual farms are predicted by obtaining the ratio of the
observed production values to the corresponding estimated frontier values.
)( ,
iii xfYTE where
is either the maximum likelihood estimator or the
corrected Ordinary Least Squares estimator for β. It measures the output of the ith
farm relative to the output that could be produced by a fully-efficient firm using the
same input vector.
Once the inefficiency component of the production function is separated, its
determinant should be identified. For this purpose it is assumed that the average level
of technical inefficiency is a function of factors believed to affect technical
inefficiency as shown below:
(7.8)
)exp();(*ii VxfY
)exp(),(
)exp(),(
ii
iiii
Vxf
UVxfTE
)exp( ii UTE
iii gZU
215
where ig is a random variable distributed with mean value of zero and variance 2g .
That is, 20 ,N~ gig . The random variable Ui is defined by the truncation of the
normal distribution. In this study, we propose the use of the more flexible truncated
normal distribution that allows for a wider range of distributional shape (Coelli et al.,
2005). The assumption of truncated normal distribution for the Ui’s is an approach
that was suggested by Stevenson (1980) by generalising the assumption of half-
normal distribution. In the half normal distribution Ui are assumed to be the positive
half of a normally distributed variable with mean zero . Kumbhakar
and Lovell (2000) state that individual efficiency scores, as well as the composition
of the top and bottom efficiency scoredeciles, are not affected by the distributional
assumptions of the inefficiency component, Ui, and suggest the use of relatively
simple distributions such as a half normal or an exponential distribution. Complete
details of the MLE derivatives are shown in Appendix L.
7.4 Empirical model of estimation
This section explains the empirical method of estimating agricultural biodiversity
and farm level technical efficiency. As explained in the previous section, since the
basic stochastic frontier model was first proposed by Aigner et al. (1977) and
Mueeusen and Van den Broeck (1977), various other models have been suggested
and applied in the analysis of cross sectional and panel data. However, the empirical
model of technical efficiency in this study was based on the stochastic production
function proposed by Battese and Coelli (1995). In the first phase of the empirical
analysis, technical efficiency effects for a cross section of farmers is modeled in
terms of input variables in the production process. Rural agricultural households
),0(~ 2ui NU
216
generally cultivate different crops on their farms. Therefore, this practice renders the
single crop production model to be infeasible. In the case of multiple outputs, the
dependent variable in the production model is measured in terms of the total value of
agricultural outputs or production. Inputs can be categorised into four groups: land,
labour, capital and other inputs. It is assumed that the capital use in agriculture is
homogenous across the households. The translog production function is used since it
captures the interaction effects of the variables39
.
Estimation of the translog production function was performed using Frontier version
4.1 (Coelli, 1996).Accordingly, the stochastic frontier model to be estimated is
defined by:
(7.9)
where ln represent the natural logarithm. The subscript i, indicates the ith
farmer in
the sample (i = 1,2……..,n).
iYln represents the natural logarithm of the value of farm output (VFOUT)
1ln X represents the natural logarithm of the total area of land (in acres) under
cultivation (LAND).
39
The translog production function developed by Christiansen et al. (1973) is the most prevalent
functional form used in stochastic frontier analysis literature for a number of reasons. First, it provides
some degree of generality as it is a second order approximation to an arbitrary functional form. Other
familiar functional forms such as the Cobb Douglas and CES are special cases of the translog function
so these common forms are encompassed by the translog production function. Second, the translog
function allows for varying returns to scale and for technological progress to be both neutral and
factor augmenting. Additionally, partial elasticities of substitution are allowed to vary and elasticity of
scale can vary with output and input proportions.
iU
iVXXXY ki
kj k
ijjkij
j
ji
4 44
1
0 lnlnln
217
2ln X represents the natural logarithm of labour in man dates40
(LAB)
3ln X represents the natural logarithm of capital expenditure (CAP)
4ln X represents the natural logarithm of other cost: raw materials (COS)
j ’s are unknown parameters to be estimated
iV ’s are assumed to be independent and identically distributed normal random errors
having zero mean and unknown variance; ;2
v iU ’s are non-negative random
variables, called technical inefficiency effects, which are assumed to be
independently distributed such that iU is defined by the truncation (at zero) of the
normal distribution with mean, i and variance 2
u . The model for the technical
inefficiency effects specifies that the technical inefficiency effects of the stochastic
frontier are a function of the age, education, household size, number of separate
plots, agricultural extension services, credit access, membership of a farm
organisation, land ownership and different variables that represent agricultural farm
biodiversity. Some of these variables are assumed to be directly related to farmers’
management skills, while the others could impact on their technical efficiency
through availability of labour for timely management of farming activities or
incentives for increasing efficiency in farms.
Older farmers are expected to increase technical inefficiency (Battese and Coelli,
1992; Burki and Terrell, 1998) partly because older farmers tend to be less adaptable
to new technical developments. It is hypothesised that increased formal education,
40
Labour is measured by the number of adult family members working (greater than 14 years old).
This includes family labour as well as hired labour. However, there is no measure of individual
intensity of work such as number of hours per week. Since farmers cannot exactly remember the
number of hours worked each day, it was not possible to obtain this information.
218
ceteris paribus, is expected to reduce technical inefficiency. Expected sign of the
household size is negative. This is because when the household size increases,
available labour for agricultural activities is higher and farmers will not face any
labour constraint in their farming. The number of separate plots may increase
inefficiency if farmers cannot manage them well. More advice from extension
workers, ceteris paribus, is expected to reduce technical inefficiency effects, which
can be categorised as institutional characteristics. Agricultural credit access and
being a member of farmer organisation could increase technical efficiency while land
ownership will have negative sign as it affects the farmer managerial power of the
farm. We included three variables to capture effect of agricultural farm biodiversity
on farm level technical efficiency. They are crop diversity, livestock diversity and
mixed farming system. It is hypothesised that all these variables result in contribution
to decrease farm level technical inefficiency in small-scale farms.
Accordingly the empirical inefficiency model can be set out as shown in Equation
7.10:
iiiiiiiii ZZZZZZZZU 88776655443322110
iiii gZZZ 1111101099
(7.10)
iZ1 is the age of the responded in years (AGE)
iZ2 is the formal education of the responded in years (EDU)
iZ3 is the household size (HS)
iZ4 is number of separate plots (FS)
iZ5 is agricultural extension services contacts (AEC):Dummy variables if Yes 1, otherwise 0
219
iZ6 is credit access: Dummy variables if Yes 1, otherwise 0
iZ7 is member of a farm organization: Dummy variables if Yes 1, otherwise 0
iZ8 is the land ownership (LO): Dummy variable if owned 1, otherwise 0
iZ9 is crop species diversity (CSD): Dummy variable if multi-crops farm 1, otherwise 0
iZ10 is livestock diversity (LD): Dummy variable if multi-livestock farm 1, otherwise 0
iZ11 is mixed farm (AD): Dummy variable if mixed farm 1, otherwise 0
The econometric estimation strategy requires some of the assumption about
functional forms and distribution of error components. Given functional and
distributional assumptions, maximum-likelihood estimates (MLE) for all parameters
of the stochastic frontier production and inefficiency model defined by Equations 7.9
and 7.10 is simultaneously estimated using the program, FRONTIER 4.1 (Coelli,
1996). The technical efficiency of a farmer is between 0 and 1 and is inversely
related to the level of the technical inefficiency effects (Battese and Coelli, 1995).
Technical efficiency can also be predicted using the FRONTIER program, which
calculates the maximum-likelihood estimator of the predictor for Equation 7.6 that is
based on its conditional expectation, given the observed value of (Vi-U
i) (Battese and
Coelli, 1988). More details about obtaining maximum-likelihood estimator is given
by Coelli et al. (2005).
The next section statistically evaluates the predictions of the model on agriculture
biodiversity and farm level technically efficiency. The main interest lies in
quantifying the effect of technical inefficiency in rural agricultural areas in Sri
Lanka. A series of statistical tests were performed to decide the functional form and
presence of inefficiency effects. Then the first stage of the estimation was done by
220
using the translog production function followed by the finding of factors associated
with technical inefficiency. In the second stage, prediction of technical efficiency
was used to analyse the distribution of technical efficiency among different farmers.
A comparison of the results among different districts is made.
7.5 Estimates for parameters of stochastic frontier production function
As explained in the previous section, the econometric method using the stochastic
frontier production function was used to estimate the technical efficiency of the
farmers and the factors that influence inefficiency. The stochastic frontier production
function model has the advantage of allowing simultaneous estimation of individual
TE as well as its determinants.Following Battese and Coelli (1995), maximum
likelihood estimation is used to simultaneously estimate the parameters of stochastic
production frontier and the factors contributing to inefficiency. The software
program FRONTIER 4.1 is used for estimation. The total value of output of the farm
was modelled in terms of four input variables, namely, size of the land (plot), labour,
capital expenditure and expenditure on row materials. Last variable mainly includes
the expenditure on seeds, pesticides and fertiliser.
Various tests of null hypotheses for the parameters in the frontier production
functions and in the inefficiency models are performed at the beginning of the
empirical estimation. First, Frontier 4.1 allows various choices in relation to the
model’s functional form and inefficiency distribution. In this study, hypothesis tests
based on the Generalised Likelihood Ratio (GLR) test were conducted to select the
functional form. The null hypothesis here is that Cobb-Douglas is an adequate
221
representation of the data. The likelihood ratio test statistic λ = -2ln[L(H0)] -
ln[L(H1)] where ln[L(H0)] and ln[L(H1)] represent the values of the log-likelihood
function under the null (H0) and alternative hypothesis (H1). The likelihood-ratio
statistic, λ = -2log[Likelihood (H0)]–log[Likelihood (H1)] has approximately χ2ρ
distribution with ρ equal to the number of parameters assumed to be zero in the null
hypothesis (Battese and Coelli, 1992; Coelli, 1995).
The LR test shows that the Cobb-Douglas is rejected; indicating that the more
general form of the translog model fits this data better for all models. The LR test
shows that some combination of the squared and cross product terms in the translog
model improve the fit of the model. Second, the distributional assumptions were
tested based on the previously explained likelihood ratio test statistic. The truncated-
normal assumption is strongly accepted. Third, the truncated-normal translog
specification was tested for the existence of a frontier. The test result rejects the H0:
0 (i.e. 2
u = 0 and therefore no inefficiency exists), at the 1 per cent level for
Ampara, Anuradhapura and Kurunegala survey data using the appropriate tables
derived by Kodde and Palm (1986).
As explained in Chapter three, we identified 248, 247 and 251 observations as
completed survey questionnaires in Ampara, Anuradhapura and Kurunegala district
respectively. However, when estimating the efficiency model, we had to drop some
observations as a few respondents had not answered all the questions related to the
creation of the required variables in the efficiency model. For example, a few
households had not answered the question related to organic farm methods and
landrace cultivation or some had mentioned that they use both methods, that is,
222
landrace varieties as well as modern varieties. Furthermore, some farmers used
chemical as well as organic fertilisers. In such cases we removed these observations
from the models. After removing inconsistent observations, 238, 242 and 243
household level observations in Ampara, Anuradhapura and Kurunagala districts
could be used to estimate the efficiency model.
The socio-economic characteristics of the respondents are presented in Table in J.1,
J.2 and J.3 in the Appendix J. The study revealed that a majority of household heads
(94 per cent) were males on average. The age of the farmers ranged between 16 and
64 years. A majority of the respondents (65 per cent) were between the age of 30 and
55 years. The mean age was 41 years. This implies that the majority of the farmers
were at an economically active age and could therefore make a positive contribution
to farm production. Most respondents (98 per cent) were married. This contributed
widely to the use of family labour by the households as the wives and children
constituted the labour force. The literacy level among the farmers in the study area
was high. Chemical fertilisers were applied to 52 per cent of the plots while hybrid
varieties were the type of seed used on 49 per cent of the plots. In the study areas, 58
per cent of respondents had secondary education.
A majority of the respondents (68 per cent) had more than 10 years of farming
experience, which indicated the managerial ability of the farmers could be assumed
to be reasonably good. The study also revealed that a large proportion of the
respondents (67 per cent) were members of a farmers’ organisation. As well, most
farmers had used the services of agricultural extension officers. Around 42 per cent
had obtained credit for their farms. The household size of most respondents (88 per
223
cent) ranged between 2 and 5 members. Given that a large household size also means
more mouths to feed, large households generally produce a smaller market surplus
(Minot et al., 2006). However, in traditional agriculture, the larger the household size
the more labour force is available for farm activities.
Crop diversity varies between one and nine while livestock diversity varies between
one and five. The size of the farm can affect the diversification decision in both
ways. In some areas, the larger the farm size, the higher the tendency of
diversification of crop production thus leading to production for home consumption
and for sale (Birol, 2004). However, heterogeneity of the farm should be important in
this case. For example, suppose only part of the large farm has received irrigation
water, then farmers will attempt to diversify farming according to the water
requirement. Some farmers diversify their farms according to the soil quality or
shape of the land. On the other hand if the physical characteristic of the farm is
homogeneous, there is a higher probability to select a specialization system. On
average 68 per cent of farmers had mixed farming systems including both crops and
livestock. A relatively higher percentage of farmers in Ampara and Kurunegala
maintain mixed farming systems.
The maximum-likelihood estimates for the parameters of the translog production
function defined by Equation 7.9 are presented in Table 7.1. From the results all
except a few interaction variables had the expected positive signs suggesting that
more output would be obtained from the use of additional quantities of these
variables, ceteris paribus. The coefficients of the land variable were positive and
statistically significant at one per cent level in all models. The coefficients of labour
224
inputs were positive and highly significant indicating its importance in agricultural
production. The capital variable had a positive sign, which conforms to a priori
expectations. This indicated that higher capital use would result in higher crop yield.
The coefficient of the raw material input was positive as expected and statistically
significant at one per cent level. The significance of the variables derives from the
fact that they are major land augmenting inputs in the sense that they improve the
productivity of land thus leading to increased yield. In addition to this most of the
square variables and interaction terms provide expected signs and are statistically
significant.
The production function estimates indicate the relative importance of factor inputs in
agricultural production. The coefficients of all factors have the expected signs and
magnitudes. Land appears to be the most important factor of production with the
coefficient values of 0.39, 0.39 and 0.25 in Ampara, Anuradhapura and Kurunegala
districts respectively. Labour appears as the second most important factor for
Anuradhapura while row material is the second most important factor for farms in
Kurunegala district. The role of raw material in Ampara district is relatively less
important as a significant number of farmers were using organic methods and
landrace cultivation in this district. In addition to these variables, results show that
most of the interaction terms are significant at an acceptable margin and have
expected signs.
225
Table 7.1: Maximum-likelihood estimates for parameters of the production function
Ampara Anuradhapura Kurunegala
Variable Coefficient Coefficient Coefficient
Constant 0.172 (8.28)* 0.683 (3.42)* 0.351 (1.20)
Land 0.390 (22.83)* 0.394 (23.99)* 0.252 (12.29)*
Labour 0.158 (2.67)* 0.374 (2.25)* 0.167 (8.37)*
Capital 0.167 (6.78)* 0.126 (9.58)* 0.197 (18.55)*
Raw Material 0.032 (2.18)* 0.117 (11.72)* 0.241 (2.39)*
Land*Land 0.033 (1.85)** 0.021 (6.09)* 0.180 (10.36)*
Labour*Labour 0.055 (1.38) 0.253 (8.38)* 0.294 (1.88)**
Capital* Capital 0.168 (4.12)* 0.414 (1.77)** 0.022 (1.91)**
Raw material* Raw mate. 0.036 (2.69)* 0.006 (7.17)* 0.179 (10.82)*
Land*Labour -0.032 (-0.74) -0.283 (5.83)* -0.007 (-0.24)
Land* Capital 0.163 (4.36)* 0.187 (6.75)* 0.095 (3.29)*
Land* Raw Material 0.059 (2.58)* 0.018 (1.75)** 0.223 (8.76)*
Labour*Capital 0.063 (1.12) 0.105 (3.05)* 0.039 (11.11)*
Labour*Raw Material -0.126 (-3.45)* -0.063 (-2.11)* -0.153 (4.56)*
Capital*Raw Material -0.004 (-0.10) -0.041 (-2.52)* -0.021 (-0.59)
Model Variance 0.658 (9.02)* 0.820 (11.21)* 0.743 (11.08)*
Variance Ratio 0.713 (11.01)* 0.629 (3.59)* 0.671 (2.05)*
Log Likelihood function -277.083 -260.235 -464.258
Number of observation 238 242 243
Note: i. t ratios are given in the parenthesis. * denotes significant variables at 1% level while **
indicates significant at 5% level of significant.
ii. All estimated first order coefficients in the translog model fall between zero and one,
satisfying the monotonicity condition that all marginal products are positive and diminishing at the
mean of inputs.
The parameter γ = σu2/σ
2 lies between zero and one with a value equal to zero
implying that technical inefficiency is not present and the ordinary least square
estimation would be an adequate representation and a value close or equal to one
implying that the frontier model is appropriate. The values of γ are 0.71, 0.62 and
0.67 for Ampara, Anuradapura and Kurunegala districts and they are statistically
226
significant at the one per cent level which implies that more than half of the residual
variation is due to the inefficiency effect. This also implies that systematic influences
that are unexplained by the production function were the dominant sources of
random errors. In other words, the shortfall of observed output from the frontier
output is primarily due to factors which are within the control of the small-scale
farmers in the sample (Amos et al. 2004).
7.6 Estimating marginal productivity and input elasticity
As the second step we estimated output elasticities of each input. This is given by the
first derivative of the translog production function with respect to each variable. The
values of explanatory variables in the translog stochastic frontier model were mean-
corrected by subtracting the means of the variables so that their averages were zero.
Therefore, the first order parameters provide direct output elasticities for the
individual inputs at the mean values. Estimates of elasticities and marginal
productivity are given in Table 7.2. These coefficients can be interpreted as
elasticities of real output with respect to inputs (land, labour, capital and raw
material). The land size and labour provide relatively higher output elasticities. This
is because land and labour are the most important production inputs for semi-
subsistence agricultural areas.
227
Table 7.2: Estimated elasticities and marginal productivity of each input
Note: All equation for estimating output elasticities and marginal products for translog stochastic
frontier model is given in Appendix M. All elasticities are estimated using mean values of respective
variables.
Table 7.2 displays the mean estimates of input elasticity for each area as calculated
using Equations M.9, M.10, M.11 and M.12 in Appendix M. It becomes clear that
the average value across the sample for output elasticity of land is 0.477 while that
for labour is 0.397. Average output elasticities of capital and raw materials are 0.176
and 0.187 respectively. All elasticities are positive indicating that, as these inputs are
increased, output increases. Returns to scale are determined by summing all values of
elasticities. If the sum is less than one decreasing returns are indicated; if greater than
one increasing returns to scale are indicated. By adding coefficients of elasticities
together the returns to scale for Ampara, Anuradhapura and Kurunegala districts are
shown to be 0.75, 1.01 and 0.83 respectively. This implies there are decreasing
returns to scale for at least Ampara and Kurunegala districts farms. Constant returns
to scale hold for farms in the Anuradhapura district.
Table 7.2 also provides the estimated marginal productivity for each input. Marginal
productivity of land per acre is Rs. 11,349, 9,114 and 9,236 for Ampara,
Ampara Anuradhapura Kurunegala
Elasticities
Marginal
Productivity Elasticities
Marginal
Productivity Elasticities
Marginal
Productivity
Land 0.449 11,349 0.567 9,114 0.416 9,236
Labour 0.394 393 0.433 349 0.365 372
Capital 0.097 0.795 0.229 0.951 0.203 1.104
Row material 0.222 0.852 0.233 0.399 0.106 0.912
228
Anuradhapura and Kurunegala districts respectively. This provides the value of using
additional acre of land for the farming practice in different districts. Results clearly
show that marginal productivity of the land in Ampara district is relatively higher
than that of other district. One of the possible reasons could be that relatively larger
number of small-scale farmers in Ampara district use landrace cultivation and
organic farming method which could help them maintains higher soil fertility in the
long run. Interestingly, estimated marginal productivity of labour is relatively lower
than the existing wage rate in rural areas in Sri Lanka. Average wage rate varies
between Rs. 400 and Rs. 450 depending on peak or off-peak time. Also in some
areas there is a marginal different of the daily wage between women and men (it is
Rs. 450 for men while Rs. 400 for women). However, estimated marginal
productivity of labour is found of the range between Rs. 350 and Rs. 400. Marginal
productivity of capital is Rs. 0.97, Rs. 0.95 and Rs. 1.10 for Ampara, Anuradhapura
and Kurunegala districts respectively. Marginal productivity of raw material is
relatively lower in Anuradhapura district when compared with the other two districts.
7.7 Variations of technical efficiency
As the third step of the analysis, we examine the distribution of technical efficiency
of farmers in different regions. The results are presented in Table 7.3. The average
resource-use efficiency in the sample for Ampara, Anuradhapura and Kurunegala are
0.692, 0.511, and 0.685 respectively. This implies that about 30.8, 48.9 and 31.5 per
cent higher levels of production could be achieved without additional resource for
Ampara, Anuradhapura and Kurunegala districts respectively. From the distribution,
229
the most efficient farmers in terms of resource use in Ampara district sample have an
index of 92.13 per cent and the least efficient farmers in the same district have a
Table 7.3: Frequency and percentage distribution of the technical efficiencies
Ampara Anuradhapura Kurunegala
Efficiency-
range
Number
of farms
Percentage
(%)
Number of
farms
Percentage
(%)
Number of
farms
Percentage
(%)
0.00 - 0.40 1 0.42 42 17.36 3 1.23
0.41 - 0.45 3 1.26 27 11.16 6 2.47
0.46 - 0.50 22 9.24 34 14.05 8 3.29
0.51 - 0.55 10 4.20 36 14.88 12 4.94
0.56 - 0.60 14 5.88 43 17.77 36 14.81
0.61 - 0.65 21 8.82 31 12.81 41 16.87
0.66 - 0.70 32 13.45 14 5.79 46 18.93
0.71 - 0.75 46 19.33 5 2.07 34 13.99
0.76 - 0.80 59 24.79 3 1.24 37 15.23
0.81 - 0.85 18 7.56 6 2.48 14 5.76
0.86 - 0.90 9 3.78 1 0.41 5 2.06
0.91-1.00 2 0.84 0 0.00 1 0.41
Note: Number of farms used for this analysis are 238, 242 and 243 Ampara, Anuradhapura and
Kurunegala district respectively. Descriptive statistics shows that the average farm size in
Anuradhapura farms is relatively higher that of other two districts.
resource use efficiency of 22.13 per cent. However, the most efficient farmer in
Anuradhapura sample have index of 81.62 per cent and the least efficient ones have a
resource use efficiency of 16.25 per cent in the same district. The highest efficiency
level of the Kurunegala sample is recorded as 92.62 per cent while minimum is 24.35
per cent. A wide variation of the technical efficiency level among farmers in
different districts is evident by these figures.
230
The inability of any of the farmers to operate on the frontier could be attributed to
certain factors ranging from technical constraint, socioeconomic factors and
environmental factors. Specifically, scare inputs may be allocated to various users on
the basis of their marginal shadow values thereby preventing farmers from reaching
the efficiency frontier. The distribution of the inefficiency estimates shown in this
study agree with previous work carried out in other peasant farming settings in this
area (Burki and Shah, 1998; Coelli and Battesse, 1996). In the present study,
approximately 10 per cent of sample farmers in Ampara and Kurunegala had a mean
technical efficiency of less than 0.50 and approximately 70 per cent had a mean
technical efficiency in the range of 0.50 - 0.80 for the same districts. On average the
predicted TEs for the farmers in all districts ranged from 0.16 to 0.92. The mean TE
of 0.63 indicated that the average farmer produced about 63 per cent of maximum
attainable output for given input levels in the study area.
Next we estimated average efficiency levels for different land size. The purpose of
this analysis is to investigate whether there is a direct link between farm level
efficiency and farm size. The average estimates of technical efficiencies by farm-size
categories are presented in Table 7.4.
It is clear that producers in relatively larger farms are as efficient as the producers in
relatively smaller farms. This implies that there is no difference of mean technical
efficiency between different farm sizes. We also estimated actual output as well as
potential output under each land size category. It is clear that actual output and
potential output increase with land size holding mean technical efficiency as the
same level. As the average technical efficiency level in each land size is almost the
231
same in all districts, it can be concluded that land size does not change the farm level
technical efficiency among small-scale farms in Sri Lanka.
Table 7.4: Average TE, value of actual and potential output (Rs.) with land size
Ampara
Farm size (acres) Number Efficiency average Actual output Potential output
0.00-0.50 141 0.675 11,802.62 15,421.72
0.51-1.00 16 0.747 28,491.95 35,478.13
1.10-1.50 17 0.673 31,834.70 41,832.12
1.51-2.00 25 0.756 42,933.24 53,287.06
2.10-2.50 39 0.695 44,651.56 57,846.57
Anuradhapura
Farm size (acres) Number Efficiency average Actual output Potential output
0.00-0.50 55 0.505 7,610.79 11,253.28
0.51-1.00 42 0.528 19,796.06 28,557.33
1.10-1.50 24 0.499 22,719.72 33,789.43
1.51-2.00 44 0.486 29,204.10 43,289.87
2.10-2.50 77 0.523 36,081.82 52,660.34
Kurunegala
Farm size (acres) Number Efficiency average Actual output Potential output
0.00-0.50 43 0.701 13,591.16 17,491.87
0.51-1.00 69 0.682 24,040.00 31,465.66
1.10-1.50 67 0.671 28,857.81 38,121.13
1.51-2.00 26 0.695 40,279.51 52,076.89
2.10-2.50 38 0.689 48,861.82 63,805.41
Note: Potential output represents the value of actual output plus output loss due to inefficiency. The
value of inefficiency is estimated using coefficients of inefficiency in each category.
In the next step we calculated the average efficiency level with farm type. We
divided farms into single variety, multiple variety and mixed system. Single variety
includes farms that have only one crop variety or one livestock variety. Multiple
232
varieties include farms that have more than one crop variety or more than one
livestock variety. Mixed system includes farms that have both crops and livestock.
This type of analysis provides us evidence about the farm level technical efficiency
and diversity in the farm. The result of the average efficiency under different farming
systems is given in Table 7.5.
Table 7.5: Average efficiency with farm type
Ampara Anuradhapura Kurunegala
Category Farms Efficiency Farms Efficiency Farms Efficiency
Single variety only 16 0.583 54 0.312 18 0.497
More than one variety only 64 0.680 56 0.510 69 0.698
Mixed(crops and livestock) 158 0.812 132 0.710 156 0.859
Total 238 0.691 242 0.511 243 0.685
Note: Single variety and more than one variety include only single and multiple variety crops or
livestock. The mixed category include single variety or/and multiple variety crops and single or/and
multiple variety livestock.
The average efficiency level for single variety is 0.58, 0.31 and 0.49 for farms in
Ampara, Anuradhapura and Kurunegala districts respectively. However, these
numbers increased to 0.68, 0.51 and 0.69 in the same districts for farms which have
more than one variety. The highest average efficiency is recorded for farms which
have a mixed farming system. This means that the technical efficiency level of
farmers who have both crops and livestock is relatively higher than other categories
for all districts. For examples, average technical efficiency of mixed system farms in
Ampara, Anuradhapura and Kurunegala is 0.81, 0.81 and 0.86 respectively. It is
therefore clear the mixed farming system is more efficient than other farm systems in
each district. This evidence encourages us to investigate the agricultural biodiversity
and farm level technical efficiency in the formal efficiency analysis.
233
Information given in Table 7.5 shows those farms with single varieties is relatively
higher in Anuradhapura district. It was observed that most farms of this category in
Anuradhapura district had cultivated rice as the single crop. More than one variety
farms in Ampara, Anuradhapura and Kurunegala districts were 26, 23 and 28 per
cent. However, this category only includes farms that have more than one variety of
crops or livestock. Mixed farms are relatively higher in all three districts. This is
because the mixed farming system is the most common farming system in most rural
small-scale farms in Sri Lanka. Results of this study clearly show that technical
efficiency level of this type of farming system is relatively higher.
7.8 Results of the inefficiency model
As the final step of the analysis, the variables of the inefficiency model were
modeled to explain the determinants of inefficiency of production among farmers in
three districts. The TE difference between farmers could be due to farm-specific or
farmer-specific variables. The sign of the variables in the inefficiency model is very
important in explaining the observed level of TE of the farmers. A negative sign
would imply that the variable had the effect of reducing technical inefficiency, while
a positive coefficient would indicate increasing inefficiency. The results are
presented in Table 7.6 and indicate that all the included variables except age had the
expected sign.
234
Table 7.6: Maximum-likelihood estimates for parameters of the inefficiency model
Ampara Anuradhapura Kurunegala
Variable Coefficient Coefficient Coefficient
Constant 0.785 (1.96) 1.079 (5.40)* 0.585 (1.99)**
Age 0.016 (3.05)* 0.038 (5.74)* 1.217 (4.20)*
Education -0.015 (-5.29)* -0.011 (-5.40)* -0.035 (-3.74)*
HH size -0.026 (-4.77)* -0.034 (-5.14)* -0.021 (-6.36)*
Number of plots 0.018 (2.57)* 0.071 (6.93)* 0.014 (5.16)*
Extension services -0.059 (-1.76)** -0.018 (-1.74)** -0.083 (-2.62)*
Credit -0.069 (-3.14)* -0.022 (-1.64)*** -0.035(-5.77)*
MFO -0.045 (-1.61)** -0.031 (-2.06)* -0.052 (-6.99)*
Land ownership -0.131 (-3.77)* -0.030 (-2.41)* -0.078 (-1.92)**
Crop diversity -0.066 (-1.62)*** -0.038 (-1.72)** -0.021 (-7.48)*
Animal diversity -0.042 (-2.24)* -0.026 (-2.11)* -0.017 (-8.01)*
Mixed farmi -0.045 (-2.13)* -0.017 (-2.25)* -0.078 (-13.19)*
Note: i. mixed farm variable show the farm has a mixed system or not. A mixed system includes
single variety or/and multiple variety crops and single or/and multiple variety livestock.
ii. t ratios are given in the parenthesis. * denotes significant variables at 1% level and **
indicates significant at 5% level while *** denotes significant variables at 10% level of significant.
The estimated coefficients in the inefficiency model are of particular interest to this
study. This is because these estimated coefficients of the inefficiency function
provide explanations for the relative technical efficiency levels among individual
farms. Most of the coefficients of the explanatory variables in the inefficiency model
are found to have expected signs. The age coefficient is positive in all three models,
which indicates that the older farmers are more inefficient than the younger ones.
This variable is significant at one per cent level for Ampara and Anuradhapura
district while it is significant at five per cent level for Kurunegala sample. The
positive coefficient of age suggests that age led to technical inefficiency of the
farmers (Seyoum et al., 1998; Amos et al., 2004; Ogunyinka and Ajibefun, 2004). A
235
possible explanation could be that the general ability to supervise farming activities
decreases as farmers advanced in age.
The negative estimate for education variable implies that farmers with greater years
of schooling tend to be less inefficient. The relationship is relatively strong, because
the coefficient is very high relative to its estimated standard error in all three models.
The coefficient of education is significant at one per cent level. It can therefore be
assumed that farmers with greater years of formal schooling tend to be more
technically efficient. This agrees with the findings of Ajibefun and Aderinola (2003)
who reported that farmers in Southwestern Nigeria become more technically efficient
with more years of formal schooling. These data asserted that more years of formal
education and new technologies were imperative to better understand and adapt the
technologies, which subsequently make it possible to move close to the frontier.
The predicted coefficient of household size was negative and significant at one per
cent for sample of Ampara farmers while it was significant at five per cent for
Anuradhapura and Kurunegala farmers. The negative coefficient is in agreement with
the hypothesised expected sign and implies that as the number of persons (adult) in a
household increases, efficiency also increases. This is because more adult members
in a household mean that additional quality labour would be available for carrying
out farming activities in a timely fashion, thus making the production process more
efficient (Villano and Fleming, 2006; Shehu et al., 2007).
The number of separate plots may increase inefficiency if farmers cannot manage
them well. The result of this study shows that the greater the number of plots grown
by each household, the lower the farm level technical efficiency (see, Table 7.6).
236
This variable is significant at one per cent level in all three models. The probable
reason is that the separate plots can affect farmer managerial ability. When the
number of plots is higher, farmers need additional time to look after them which can
lead increasing farm level technical inefficiency. One of the main issues faced by
rural farmers is that they had to protect their crops or livestock from predators. The
location of different plots in different places means that farmer ability to overcome
this problem is less. Given this variable is highly significant in all three models this
appears to confirm our assumption.
The coefficient of extension contact is negative and significant, suggesting that such
contact increases farm level technical efficiency because farmers are able to use
modern techniques of farming involving land preparation, planting, application of
agro-chemicals (for example, fertiliser) and harvesting. This finding confirms the
results of Xu and Jeffrey (1998) that extension visits to farmers are important in
reducing farm inefficiency. The coefficients of availability of agricultural credit and
becoming a member of a farm organisation were also statistically significant for all
three models and had the expected signs. Credit access can remove farmers’ financial
constraints and thereby increase the farm level efficiency. It can also be assumed that
being a member of a farm organisation helps farmers improve managerial skills as it
provides training programs with necessary information during the crop season.
The results also show that land ownership has a negative impact on inefficiency. A
similar conclusion was drawn by Ajibefun and Aderinola (2003) and Amos et al.
(2004) in their analysis. This implies that farmers who cultivate their own land are
more efficient than those who cultivate land that is leased. This is because farmers
237
who own land have added motivation to cultivate more efficiently as they have an
incentive to maintain their land for long-term benefits. In general, agricultural land
market does not function well in Sri Lanka. A number of market distortions could be
observed of the land market in rural areas. The possible policy implication is that
steps should be taken to reduce imperfections that exist in the agricultural land
market.
The estimated coefficients of the variables that represent agricultural biodiversity are
of central interest to this study. This is because the estimated coefficients of the
inefficiency function provide an explanation of the way in which they contribute to
farm level technical efficiency in small-scale farms in a semi-subsistence economy.
The results show that crop diversity is significant at one per cent level for
Anuradhapura sample while it is significant at five per cent level for the other two
districts. The animal diversity variable is significant at one per cent level for
Anuradhapura and Kurunegala sample while it is significant at five per cent level for
Ampara sample with expected signs. This implies that diverse farms are more
efficient than the other farms. Accordingly, we find that the higher the crops or
livestock diversification, the higher the farm level technical efficiency. Possible
reasons include: farmers can utilise family labour optimally when they maintain a
diverse agricultural system; a diverse farming system minimises external risks that
farmers often face and a biologically rich farming system can improve soil fertility
and minimise input costs in the long run.
Variable that captures the mixed farming system is highly significant in all models.
This implies that the efficiency level of the farms which maintain mixed farming
238
systems is higher than that of other farms. This result is consistent with our
hypothesis that given the semi-subsistence nature of the rural faming system, farmers
can improve their efficiency level significantly by adopting mixed farming systems
for their farms. The results indicated in Table 7.6 show a significant decrease in farm
household inefficiency with the mixed farming system.
7.9 Summary and key findings
This study provides an economic analysis of farm household efficiency among rural
households in Sri Lanka, where crop and livestock farming generate a large part of
household income. Using stochastic frontier analysis, the results show the potential
of encouraging mixed farming systems as a driving force for output growth.
Econometric analysis of survey data shows that land size, labour, capital expenditure
and expenditure on raw materials are important inputs and are strongly associated
with the total output. The analysis reports evidence of farm level technical
inefficiency and its determinants. Results of this study show the potential for large
gains in real output if technical efficiency is increased. The results depict a wide gap
between farmers who are relatively poor in their efficiency performance (20 per cent)
and those who are highly efficient (more than 90 per cent). In particular this study
shows that the output value of farms in the study area can be increased with the
current levels of inputs and technology if less efficient farmers are encouraged to
follow the resource utilisation pattern as well as farm types that have already been
adopted by the most efficient farmers.
239
Among the significant variables in the inefficiency model, level of education,
number of separate plots, agricultural extension service, credit access, membership of
farm organisation and land ownership are direct policy relevant variables. This
means that all these variables can be controlled by using appropriate policies in the
country. More farmers in rural areas are not aware about the possible benefits that
they could gain by following their more efficient peers. It is also found that the
variables that were used to represent agricultural farm biodiversity (crop diversity,
animal diversity, mixed farming) are significant determinants of farm level technical
efficiency in rural small-scale farming in Sri Lanka. In general, the analyses of
determinants of inefficiency clearly indicated that households which have access to
agricultural extension services, credit facilities and those who maintain more diverse
or a mixed farming system with higher levels of diversification are more likely to be
more efficient than the other households.
What policy interventions would be appropriate to increase efficiency at rural
household level? The results suggest that policy makers could place more emphasis
on rural agricultural extension services to increase the probability that farmers will
adopt mixed farming system with more diversification. The analysis of farm level
technical efficiency indicates that maintaining more diverse farming systems is
crucial to reducing inefficiency and improves the welfare of rural households in Sri
Lanka. This fact has particular implications for policies required to sustain gains in
agricultural productivity and efficiency. Agricultural advisory services, rural credit
organisations and other stakeholders working for rural development should clearly
tailor their messages and services to meet the identified needs of rural farmers.
240
Designing formal and informal education programs that will improve farmers’
abilities to improve efficiency is extremely important. The emphasis should be on
providing education that will help farmers understand the socioeconomic and policy
conditions governing their farming activities. A further initiative would be taken to
strengthen the capacity of farmers through farmer centred training workshops geared
towards managerial efficiency as well as resource use efficiency. This could be done
in a collaborative manner involving the government, district assemblies and NGOs.
Government also need to intensify its agricultural extension services program by
training and deploying qualified extension officers. The officers, in turn, should
intensify farmer education on input use.
It is notable that the agricultural extension officer-farmer ratios, as well as extension
contact with farmers in the study area, are low. There is, therefore, a need to motivate
and train the existing extension officers to work more effectively and to train more
officers. It is also suggested that (i) an appropriate policy or regulation that
recognises and encourages the effective use of agricultural land be formulated by
state authorities at various levels; (ii) farmers should be encouraged to move to more
diverse farming practice and (iii) the role of educational programs in improving
efficiency should be highlighted. It is clear that the inefficiency effects in this
particular instance reinforce other empirical evidence from other developing
countries (Ali and Chaudry, 1990; Parikh and Shah, 1995; Shehu et al., 2007). In
general, the study has revealed that most of the farmers in Sri Lanka are not fully
technically efficient and, therefore, there is capacity to improve efficiency by
addressing some important policy variables that negatively and positively influence
farmers’ levels of technical efficiency.
241
CHAPTER EIGHT
CONCLUSIONS AND POLICY IMPLICATIONS
8.1 A summary of findings and discussion
Sustainable agricultural development is widely acknowledged as a critical
component in a strategy to combat both poverty and environmental degradation. Yet,
sustainable agricultural development remains an elusive goal, particularly in many of
the poorest regions of the world. Biodiversity degradation continues to be a key
factor in unsustainable agricultural systems, despite decades of research focus on
different issues related to agricultural biodiversity conservation (Brush et al., 1992;
Ceroni et al., 2005).
The prevailing economic explanation for the continuing trend toward agricultural
biodiversity degradation in many parts of the world is that economic incentives often
encourage degradation and discourage conservation. These incentive problems have
been attributed to poor farmers’ high discount rates, lack of markets, high transport
costs and other market imperfections, adverse government policies and insecure
property rights (Di Falco and Perrings, 2003). From this perspective, the challenge
facing researchers and policy analysts is to understand the factors causing
agricultural biodiversity degradation and design mechanisms that will provide
farmers in developing countries with the economic incentives needed to adopt more
sustainable land use and management practices with environmental rich farming
242
systems. This research analysed these issues using small-scale farm data in Sri
Lanka. The main findings of the study are summarised below.
First, the research reported in Chapter four of the thesis represents one of the first
attempts to use the CE approach to investigate farmers’ preference for different
attributes of agricultural biodiversity that is present in small-scale farms in
developing countries. We applied the CE approach to identify the potential benefits
of conserving agricultural biodiversity in Sri Lanka. Four conclusions can be drawn
from this chapter.
Firstly, owing to educational and poverty issues, some policy makers in developed
countries are suspicious of whether non-market valuation techniques like CVM and
CE can be applied in developing countries such as Sri Lanka. This CE study has
demonstrated that carefully designed and pre-tested nonmarket valuation techniques
can validly be applied in developing countries Secondly, farmers have strong
positive attitudes towards increasing agricultural biodiversity in rural areas. This is
evident from the results obtained from the CLM. Thirdly, the study illustrates that in
Sri Lanka it is possible to improve agricultural biodiversity using appropriate policies
in which draw on the finding of this study. Finally, the application of the CE
approach appears promising, given its capacity to model complex, simultaneous
tradeoffs involved in ecological management. The CE technique can be used to
model a variety of simultaneous tradeoffs which involve a mixture of environmental
and socio-economic factors. The results provide a tool for decision makers to use in
prioritising ecosystem management options in rural agricultural areas.
243
Secondly, a study on the current status of agricultural biodiversity and its
determinants is shown to be a useful input for policy decisions makers concerned
with conserving agricultural biodiversity in rural areas and hence improvement of
farmer livelihoods. In this context, it is important to know which farmers are
promoting diversity and what the determinants are. Chapter five of this thesis
investigated this issue using information derived from farmers’ demand for crop and
livestock varieties.
It is found that maintaining on-farm diversity has received increasing attention as a
strategy for mitigating production risk and protecting food security in rural areas of
Sri Lanka. For poorer farmer’s small land size, crop and animal variety
diversification increases the options for coping with variable environmental and
market conditions. As well, due to the existence of imperfect markets, farmers grow
different varieties to meet their consumption requirements. On the one hand, this
practice increases their food security. On the other hand, it provides more fresh food
with high nutrition content. Farmers may also sell some of the surplus to the market
so as to buy their family needs (clothes and other goods/commodities). This may
motivate farmers to grow the varieties that can be sold in the market for cash. We
therefore find that the key variables promoting diversity are household
characteristics, market characteristics, and some of the other characteristics such as
percentage of their own money spent for agriculture. One of the main conclusions
drawn from this study is that the centrality of markets in shaping diversity does not
suggest a trade-off between development and diversity. This is because as integration
with outside markets increases, the level of crop diversity on farms can also be
increased.
244
Thirdly, although the benefits of environmentally rich farming systems in Sri Lanka
are clear, the impacts of socio-economic change upon agricultural biodiversity in the
country have received little attention. Chapter six of the thesis investigated the
farmers’ preferences for different farming systems such as landrace cultivation,
organic and mixed farming practices. We find that the key variables promoting
landrace cultivation, organic farming and mixed farming systems are household
characteristics, market characteristics, and some of the other characteristics such as
percentage of their own money spent for agriculture.
The results show that gender, farmers’ positive attitudes towards agricultural
biodiversity, farms size, input price fluctuations, agricultural subsidies and
percentage of own money investment are found to be important factors when taking
decisions to maintain landrace cultivation. Investigation of profiles of farm families
that are most likely to cultivate landraces and organic farming reveals they have less
income compared to those farm families that are not likely to cultivate landraces.
They are more agriculturally-based, with less off-farm employment and more
isolated from the markets.
Among the important variables in organic farming models, farmers’ attitudes
towards agricultural biodiversity, input price fluctuations, agricultural subsidies and
farm size are found to be the most significant variables. Organic farming has proven
beneficial for many farmers, but the yield of organic farming has not been
substantial. Many farmers can be encouraged to undertake organic farming if the
benefits could be shown to them. There have also been instances where farmers have
opted for organic farming on account of reduced production costs compared to
245
conventional farming. Low productivity, increased time required to yield, and the
requirement of specialised skills have been some of the disadvantages of organic
farming. However, organic farming contributes towards providing quality food and
also protecting agricultural soils.
It is clear that most of the variables used in the mixed farming model are significant
and have taken expected signs. We found that households with more experience,
more labour availability and less off farm income are more likely to have mixed
farming systems. The results also show that the market characteristics as well as
agricultural subsidies are important determinants for selecting mixed farming
systems. Off-farm income, wealth and agricultural subsidies have been shown to be
negatively related with mixed farms in small-scale farms in Sri Lanka. Possible
policy implications related to agricultural subsidies is that given the government's
limited resources and competing demands, the best use of funds which are allocated
for agricultural development is to improve rural infrastructure/technology and to
build market linkages rather than using them for wasteful subsidies which have no
long-term development impacts.
Fourthly, Chapter seven of this thesis provides an economic analysis of farm
household efficiency among rural households in Sri Lanka, where crop and livestock
activities generate a large part of household income. Using stochastic frontier
analysis, the results show the potential of encouraging mixed farming systems as a
driving force of output growth. Econometric analysis of survey data shows that land
size, labour, capital expenditure and expenditure on raw materials are important
246
inputs and are strongly associated with the total output. Results of this study show
the potential for large gains in real output if technical efficiency is increased.
The results depict a wide gap between farmers who are relatively poor in their
efficiency performance and those who are highly efficient. In particular this study
shows that the output value of farms in the study area can be increased with the
current levels of inputs and technology if less efficient farmers are encouraged to
follow the resource utilisation patterns and farm types that have already been adopted
by the most efficient farmers. Among the significant variables in the inefficiency
model education level, number of separate plots, agricultural extension service, credit
access, membership of farm organisation and land ownerships are direct policy
relevant variables. This means that all these variables can be controlled by using
appropriate policies in the country. More farmers in rural areas are not aware about
the possible benefits to be gained by following their more efficient peers.
It is also found that crop diversity, animal diversity and mixed farming systems are
significant determinants of farm level technical efficiency in rural small-scale farms
in Sri Lanka. In general, the analysis of determinants of inefficiency clearly indicated
that households which have access to agricultural extension services, credit facilities
and those who maintain more diverse or a mixed farming system with higher levels
of diversification are more likely to be more efficient than those who are not.
247
8.2 Policy implications
There are number of important policy implications that arise from the findings of the
thesis. Some of the major implications are discussed as follows. First, the findings of
the choice experiment which support the assumption that small-scale farms and their
multiple attributes contribute positively and significantly to the utility of farm
families in Sri Lanka. To the extent that the findings are representative of other rural
areas in the country they confirm that small-scale farms continue to be a vital for that
nation since the benefits to farms are overall positive and high. The value estimates
reported in this analysis represent lower bounds since only the private use values of
small-scale farms were estimated. The results reveal that differences between
regions, in terms of market integration, infrastructure quality and agro-ecological
condition, affect small-scale farmers’ private valuation. The CE study discloses the
farm family and regional characteristics that are important to consider in designing
program or policies to conserve or enhance the agricultural biodiversity and other
attributes of Sri Lankan farms.
Second, it is clear that various attributes of agricultural biodiversity provide direct
and indirect benefits and advantages which meet human needs in different ways.
Putting a value on these benefits is difficult, but decision makers often call for them
to be expressed in monetary terms. To this end, in this study we present the results of
a CE study designed to shed light on subsistence farm households’ preferences for
various farm attributes and these households’ trade-offs among these attributes. The
findings presented here are therefore expected to inform the design of efficient,
effective, equitable, and targeted compensation and livelihood diversification
248
policies in the country. Such economic policies would be designed and appropriately
target the future conservation of agricultural biodiversity in Sri Lanka.
Third, analysis in Chapter five has attempted to fill the gap by investigating how
different forms of market provisioning and other variables shape the on-farm
conservation of agricultural farm biodiversity in Sri Lanka. It is clear that policies
that affect household labour supply and its composition are therefore likely to have a
major impact on most components of agricultural farm biodiversity in the country.
Educational campaigns on variety choice and seed management are also relevant.
The information provided by analysis of all models is directly policy relevant and
appropriate policies can be designed to control them. The predictions from the
models estimated above enable us to identify the types of families that are most
likely to sustain the agricultural biodiversity. Profiles can be used to design targeted,
least cost incentive mechanisms to support conservation as part of national
environmental and agricultural programs.
Fourth, in each statistical analysis conducted, whether descriptive or econometric,
regional heterogeneity is observed. Hence, any agri-environmental policy or program
that aims to support the management of current levels of agricultural biodiversity in
rural areas in Sri Lanka will need to recognise the heterogeneity of these traditional
farms and their context. Furthermore, any policy or program that affects the wealth,
education or labour participation of family members, or the formation of food
markets within settlements, will influence their choices. It is hoped that these
analyses will contribute to advancing the economics methods used to analyse the
prospects for on farm conservation, where evidence demonstrates that the expected
249
social benefit-cost ratio of on farm conservation is high. The relationship between the
diversity maintained by individual household farms and the diversity maintained
from the perspective of the community as a whole will also be essential for the
design of policy instruments.
Fifth, the information provided by the analysis of all models in Chapter five is shown
to be of high policy relevance. Specifically the predictions from these models enable
us to identify the types of farm families that are most likely to increase the
agricultural biodiversity in Sri Lanka. Accordingly, household profiles can be used to
design targeted, least cost incentive mechanisms to support conservation different
traditional farming system in the country. This study contributes to the literature by
providing insights into farmers’ landrace cultivation, organic farming and mixed
farming preferences, using small-scale farm household data in a typical developing
country setting. For example, ‘agricultural subsidies’ variable is significant in all the
models. It implies that the existing subsidy program in Sri Lanka has negatively
affected choices about cultivating landrace varieties and organic farming systems.
Therefore, steps should be taken to rethink the existing subsidy program in the
country. Furthermore, the results of the study also identify the household contextual
factors that govern these decisions.
Sixth, on farm conservation of crop diversity poses obvious policy challenges in the
design of appropriate incentive mechanisms and in terms of possible trade-offs
between conservation and productivity or other social objectives. It is clear that sales
promotion activities and credit facilities have promoted the cultivation of modern
crop varieties using pesticides and chemical fertilisers. Such a system can increase
250
short-term yields while destroying the resilience of agro-ecosystems in the long-term.
Policy decision-makers should take necessary action to minimise the impacts of such
programs while showing the benefits of agricultural biodiversity. Progress has also
been hampered both by ideological debates that are based on limited information,
and by the high cost involved in assembling the sort of large-scale scientific
databases that would be necessary to improve the quality of that information.
Furthermore, biological diversity has many components that are interrelated within a
continually evolving agro-ecosystem, and analysing causal relationships in any
component over a brief time horizon obviously leads to partial, static conclusions.
Seventh, designing formal and informal education programs that will improve
farmers’ efficiency should be a high priority. The emphasis should be on providing
education that will help farmers understand the socioeconomic and policy conditions
governing their farming activities. A further initiative would be to strengthen the
capacity of farmers through farmer centered training workshops geared towards
managerial efficiency as well as resource use efficiency. This could be done in a
collaborative manner involving the government, district assemblies and NGOs.
Government also needs to intensify its agricultural extension services program by
training and deploying qualified extension officers. The officers, in turn, should
intensify farmer education on input use.
Eight, it is notable that the agricultural extension officers-farmer ratios, as well as
agricultural extension contact with farmers in the study area, are low. There is
therefore a need to motivate and train the existing extension officers to work more
effectively and to train more officers. It is also suggested that (i) an appropriate
251
policy or regulation that recognises and encourages the effective use of agricultural
land be formulated by state authorities at various levels; (ii) farmers should be
encouraged to move to more diverse farming practice and (iii) the role of educational
program in improving efficiency should be highlighted. There is, therefore, a need to
design appropriate policies focusing on rural small-scale farms in Sri Lanka.
Nine, the results suggest that policy makers could fruitfully place more emphasis on
rural agricultural extension services to increase the probability that farmers will
adopt mixed farming systems with more diversification. The analysis of farm level
technical efficiency indicates that maintaining more diverse farming systems is
crucial to reducing inefficiency and improving the welfare of rural households in Sri
Lanka. This fact has particular implications for policies required to sustain gains in
agricultural productivity and efficiency. Agricultural advisory services, rural credit
organisations and other stakeholders working for rural development should clearly
tailor their messages and services to meet the identified needs of rural farmers.
8.3 Limitations of the study and further research
It is important to be conscious of the possible limitations of the study. It is also
important to consider some of extensions to this study. These are explained below.
First, all data used in this thesis are primary data collected through a field and CE
survey and should be considered fairly reliable. However, there is the possibility that
during interviews the interviewer asks the specific questions in a biased way. In
order to reduce this problem the survey was pre-tested on focus group discussion.
252
From the feed-back of the focus group discussion it was understood that the
questions were seen as unproblematic and in that sense the data collected is judged to
be reliable. However, it also noted that the answers from respondents in the survey
may be biased towards their own individual preferences. This means that the
respondent in the choice experiment may answer in a way that does not coincide with
his behaviour in reality.
Second, sample data used in this thesis are not representative samples of all Sri
Lankan farmers. We only selected the more diverse farming areas for this study.
Therefore, further research covering different climatic and social groups in this area
is needed to generalise the results of this research to Sri Lanka. This is another area
for future research. Moreover, obtaining accurate information from farmers was a
major challenge that was faced when collecting data. However, the data is as
accurate as possible since the trained research team was observing their behaviour
for at least a two month period. The validity of the data gained through interviewing
village level officers, agricultural officers as well as leaders of farmers’ organisations
was constantly validated during the data collection period.
Third, some of the important variables such as influences of agribusiness in
promoting chemical, seed and other products41
were not used in the analysis in
Chapters five and six. During the survey we collected some variables related to farm
specific characteristics such as irrigation water availability, soil fertility and land
shape. However, these variables were dropped from the analysis in order to avoid the
over identification problem. Comprehensive analysis covering all these variables
41
However, most of these variables do not play a significant role in small-scale farms in the country.
253
with a large sample could provide more accurate and relevant information in order to
design policies in this field. Furthermore, the results of the demand for agricultural
biodiversity show that all are positively valued in terms of extra labour required.
However, some farmers are likely to have inadequate knowledge of the long run
health effects and sustainability benefits from these possible changes, which will bias
their valuations downwards.
Methodological advances may be required to relate policies to diversity outcomes
measured at various geographical scales or levels of aggregation in the same farming
system. Specific issues for further social science research include the relationship of
seed management practices, seed markets, tenure and soil conservation practices to
diversity conservation, and the possible application of bio-economic models to the
analysis of species and genetic diversity interactions with farming systems also
require study. For policy purposes, it will be important to better understand the
particular institutional and social elements that cause communities to behave
differently in terms of conservation agricultural biodiversity in small-scale farms in
Sri Lanka in the future.
Fourth, it is clear that agricultural biodiversity is strongly determined by spatial
heterogeneity and temporal variability of the environment. Spatial heterogeneity at
the habitat, landscape and country levels play an important role in controlling
agricultural biodiversity dynamics. Dynamic biotic processes such as interspecific
competition and mutualistic interactions are important for the generation and
variation of agricultural biodiversity. Lack of knowledge about central processes
determining the spatial distribution of species in communities and ecosystems is a
254
serious problem for planning conservation measures. However, this study does not
focus on the implementation of management practices adapted to dynamic in situ
preservation of genetic resources. It does not aim at identifying new practices of
managing varietal diversity based on interaction at different levels of farmer,
commercial, and institutional seed systems.
Fifth, the simplest measure of diversity we use is a count of varieties. While counts
of varieties provide a relatively straightforward measure of richness, they suffer two
important limitations. One shortfall is that the count measures are not weighted
according to the area cultivated by a particular household. Thus, a household that
cultivates three seed lots on three hectares of land has the same diversity score as a
household that cultivates three seed lots on one hectare of land, even though the
former manages less diversity per unit of land. A second limitation of count
measures is that they do not capture the evenness of a distribution. This is another
area for future research
Sixth, in Chapter seven, some of the important factors that could play a major role in
the inefficiency function were not analysed. For example, the roles of the social
institutions and government agricultural policies can emerge as significant factors
behind technical efficiency of farmers. These factors were not targeted since the
primary purpose of this study was to investigate agricultural biodiversity and farm
level efficiency. Furthermore, some areas of further research under efficiency
measurements should be considered. These include: comparing stochastic and DEA
frontier analyses; investigating district or regional variations of technical efficiency
and investigating technical efficiency and productivity changes in using panel data.
255
BIBLIOGRAPHY
Abdelali-Martini, M., Amri, A., Ajlouni, M., Assi, R., Y. Sbieh and A. Khnifes.
2008. Gender dimension in the conservation and sustainable use of agro-biodiversity
in west Asia. The Journal of Socio-Economics, 37(1): 365-383.
Adamowicz, W., J. Louviere and M. Williams. 1994. Combining stated and revealed
preference methods for valuing environmental amenities. Journal of Environmental
Economics and Management,26(3): 271-292.
Adamowicz, W., Boxall, P. C., J. Louviere and M. Williams. 1998. Stated preference
approaches for measuring passive use values: choice experiments and contingent
valuation. American Journal of Agricultural Economics, 80(1): 65-75.
Adamowicz, W. L., Boxall, P. C., Louviere, J., J. Swait and M. Williams. 1999.
Stated-preference methods for valuing environmental amenities. In: Valuing
environmental preferences: Theory and practice of the contingent valuation method
in the US, EU, and developing countries, I. J. Bateman and K. G. Willis (Eds.),
Oxford University Press, pp: 460-479.
Adams, W. K., Aveling, R., Brockington, D., Dickson, B., Elliot, J., Hutton, J., Roe,
D., B. Vira and W. Wolme. 2004. Biodiversity conservation and the eradication of
poverty. Science, 306(5699): 1146-1149.
Agrawal, A. and K. Redford. 2006. Poverty, development and biodiversity
conservation: shooting in the dark? Working Paper No. 26, Wildlife Conservation
Society, New York, USA.
256
Aigner, D. J., C. A. K. Lovell and P. Schmidt. 1977. Formulation and estimation of
stochastic frontier production function models. Journal of Econometrics, 6(1):21-37.
Ajibefun, I. A. and A. Aderinola. 2003. Determinants of technical efficiency and
policy implications in traditional agricultural production: empirical study of Nigerian
food crop farmers. Work in progress report presented at the bi-Annual research
Workshop of AERC, Nairobi- Kenya.
Ali, M and M. A. Chaudry. 1990. Inter-regional farm efficiency in Pakistan’s Punjab:
a frontier production function study. Journal of Agricultural Economics, 41(1):62-
74.
Alpízar, F., F. Carlsson and P. Martinsson. 2001. Using choice experiments for non-
market valuation. Economic Issues, 8(1): 83-109.
Amos, T. T., D. O. Chikwendu and J. N. Nmadu. 2004. Productivity, technical
efficiency and cropping patterns in the Savanna Zone of Nigeria. International
Journal of Food Agriculture and Environment, 2(2): 173-176.
Anon, A. 1999.Biodiversity Conservation in Sri Lanka: A Framework for Action.
Ministry of Forestry and Environment, Colombo, Sri Lanka.
Arslan, A. 2007. Farmers' Subjective Valuation of Subjective Crops: The Case of
Traditional Maize in Mexico. PhD Thesis, Agricultural and Resource Economics,
Kiel Institute for the World Economy, UC Davis.
257
Arslan, A. and J. E. Taylor. 2008. Farmers’ subjective valuation of subsistence crops:
the case of traditional maize in Mexico. Working Paper, Kiel Institute for the World
Economy, Kiel, Germany.
Asrat, S., Yesuf, M., F. Carlsson and E. Wale. 2009. Farmers’ preferences for crop
variety Traits: lessons for on-farm conservation and technology adoption. EfD
Discussion Paper 09-15, Environment for the Future Initiative and Resources for the
Future, Washington DC.
Ayyad, M. A. 2003. Case studies in the conservation of biodiversity: degradation and
threats. Journal of Arid Environments,54(1): 165-182.
Bardsley, D. 2003. Risk alleviation via in situ agrobiodiversity conservation: drawing
from experiences in Switzerland, Turkey and Nepal. Agriculture, Ecosystems and
Environment, 99: 149-157.
Bartlett, J. E., J. W. Kotrlik and C. C. Higgins. 2001. Organizational research:
determining appropriate sample size in survey research. Information Technology,
Learning, and Performance Journal, 19(1): 43-50.
Bateman, I. J., Carson, R. T., Day, B., Hanemann, M., Hanley, N., Hett, T., Jones-
Lee, M., Loomes, G., Mourato, S., Ozdemiroglu, E., Pearce, D. W., R. Sugden and J.
Swanson. 2002. Economic Valuation with Stated Preference Techniques.
Cheltenham: Edward Elgar.
258
Bateman, I. J., Carson, R. T., Day, B. W., Hanemann, M., Hanley, N., Hett, T.,
Jones-Lee, M., Loomes, G., Mourato, S., Ozdemiroglu, E., Pearce, D. W., R. Sugden
and S. Swanson. 2003. Guidelines for the Use of Stated Preference Techniques for
the Valuation of Preferences for Non-Market Goods. Cheltenham: Edward Elgar.
Battese, G. E. and G. S. Corra. 1977. Estimation of a production frontier model: with
application to the Pastoral Zone of Eastern Australia. Australian Journal of
Agricultural Economics, 21(3): 169-I79.
Battese, G. E. 1992. Frontier production functions and technical efficiency: a survey
of empirical applications in agricultural economics. Agricultural Economics, 7(3-4):
185-208.
Battese, G. E. and T. J. Coelli. 1988. Prediction of firm-level technical efficiencies
with a generalised frontier production function and panel data. Journal of
Econometrics, 38(3): 387-399.
Battese, G. E. and T. J. Coelli. 1992. Frontier production functions, technical
efficiency and panel data with application to paddy famers in India. Journal of
Productivity Analysis, 3(1): 153-169.
Battese, G. E. and T. J. Coelli. 1995. A model for technical inefficiency effects in a
stochastic frontier production function for panel data. Empirical Economics, 20(2):
325-332.
Belbase, K. and R. Grabowski. 1985. Technical efficiency in Nepalese agriculture.
Journal of Development Areas, 19(4): 515-525.
259
Bellon, M. R. 2004. Conceptualizing interventions to support on-farm genetic
resource conservation. World Development, 32(1): 159-172.
Bellon, M. R. and J. E. Taylor. 1993. Folk soil taxonomy and the partial adoption of
new seed varieties. Economic Development and Cultural Change, 41(4): 763-786.
Ben-Akiva, M. and S. R. Lerman. 1985. Discrete Choice Analysis. Theory and
Application to Travel Demand. The MIT Press, Cambridge.
Benin, S., Gebremedhin, B., Smale, M., J. Pender and S. Ehui. 2003. Determinants of
cereal diversity in communities and on household farms of the northern Ethiopian
highlands. Environment and Production Technology Division, Discussion Paper No.
105, International Food Policy Research Institute, Washington DC.
Benin, S. B., Smale, M., Gebremedhin, B., J. Pender and S. Ehui. 2004. The
economic determinants of cereal crop diversity on farms in the Ethiopian highlands.
Agricultural Economics, 31(2): 197-208.
Bennett, J. and R. Blamey. 2001. The Choice Modelling Approach to Environmental
Valuation, Edward Elgar, Cheltenham, UK.
Benton, M. J. 2001. Biodiversity on land and in the sea. Geological Journal, 36(3-4):
211-230.
Binam, J. N., Tonye, J., Wandji, N., G. Nyambi and M. Akoa. 2004. Factors
affecting the technical efficiency among smallholder farmers in the slash and burn
agriculture zone of Cameroon. Food Policy, 29(5): 531-545.
260
Birol, E. 2004.Valuing Agricultural Biodiversity on Home Gardens in Hungary: An
Application of Stated and Revealed Preference Methods. PhD Thesis. University
College London, UK.
Birol, E., G. Bela and M. Smale. 2005. The role of home gardens in promoting
multi–functional agriculture in Hungary. Euro Choices, 4(3): 14-26.
Birol, E., M. Smale and A. Gyovai. 2006. Using a choice experiment to estimate
farmers’ valuation of agricultural biodiversity on Hungarian small farms.
Environmental and Resource Economics, 34(4): 439-469.
Birol, E., E. R. Villalba and M. Smale. 2008.Farmer preferences for milpa diversity
and genetically modified maize in Mexico: a latent class approach. Environment and
Development Economics, 14(4): 521-540.
Bishop, R. and D. Romano. 1998. Environmental Resource Valuation Applications
of the Contingent Valuation Method in Italy. Kluwer Academic Publishers, Boston.
Blamey, R., Rolfe, J., J. Bennett and M. Morrison. 2000. Valuing remnant vegetation
in Central Queensland using choice modelling. The Australian Journal of
Agricultural and Resource Economics, 44(3): 439-456.
Boxall, P. C. and W. L. Adamowicz. 2002. Understanding heterogeneous preferences
in random utility models: A latent class approach. Environmental and Resource
Economics,23: 421- 446.
261
Bozoglu, M., Ceyhan, V., Cinemre, H.V., K. Demiryurek and O. Kilic. 2007.
Important factors affecting trout production in the Black Sea Region, Turkey. Czech
Journal of Animal Science, 52(9): 308-313.
Bozoglu, M. and V. Ceyhan. 2007. Measuring technical efficiency and exploring the
inefficiency determinants of vegetable farms in Samsun Province, Turkey. Journal of
Agricultural Systems, 94(3): 649-556.
Bravo-Ureta, B. L. and L. Rieger. 1991. Dairy farm efficiency measurement using
stochastic frontiers and neoclassic duality. American Journal of Agricultural
Economics, 73(2): 421-428.
Bravo-Ureta, B. E. and E. R. Evenson, 1994. Efficiency in agricultural production:
the case of peasant farmers in Eastern Paraguay. Agricultural Economics, 10(1): 27-
37.
Breffle, W. and E. R. Morey. 2000. Investigating preference heterogeneity in a
repeated discrete choice recreation demand model of Atlantic salmon fishing. Marine
Resource Economics, 15(1): 1-20.
Brock ,W. A. and A. Xepapadeas. 2003. Valuing biodiversity from an economic
perspective: a unified economic, ecological and genetic approach. American
Economic Review, 93(5): 597-614.
Brookfield, H. 2001. Exploring Agro diversity. New York: Columbia University
Press.
262
Brookfield, H., Padoch, C., H. Parsons and M. Stocking. 2002. Cultivating
Biodiversity: Understanding, Analysing and Using Agricultural Diversity. The
United Nations University, ITDG Publishing.
Brookfield, H. and M. Stocking. 1999. Agrodiversity: definition, description and
design. Global Environmental Change, 9(2): 77-80.
Brown, G. M. 1990. Valuation of Genetic Resources. In: The Preservation and
Valuation of Biological Resources, Orians, G.H., Brown, G.M., W. E. Kunin and J.
E. Swierzbinski. (Eds.), University of Washington Press, pp: 203-228.
Brush, S. B. 1995. In situconservation of landraces in centers of crop diversity. Crop
Science, 35(2): 346-354.
Brush, S. B., J. E. Taylor and M. Bellon. 1992. Technology adoption and biological
diversity in Andean potato agriculture. Journal of Development Economics, 39(2):
365-387.
Bunning, S. and C. Hill. 1996. Farmers’ rights in the conservation and use of plant
genetic resources: a gender perspective. Rome, FAO.
Burki, A. A and H. N. Shah. 1998. Stochastic frontier and technical efficiency of
farms in irrigated areas of Pakistan's Punjab. Pakistan Development Review, 37(3):
275-291.
Cameron, A. C. and P. K. Trivedi. 1998. Regression Analysis of Count Data,
Econometric Society Monograph No.30: Cambridge University Press.
263
Carlsson, F., P. Frykblom and C. Liljenstolpe. 2003. Valuing wetland attributes: An
application of choice experiments. Ecological Economics, 47(1): 95-103.
Conservation on Biological Diversity. 2002. Secretariat of the convention on
biological diversity. United Nations Environmental Program.
http://www.biodiv.org/programmes/areas/agro
Central Bank of Sri Lanka. 2009. Annual Report. Colombo: Central Bank of Sri
Lanka.
Ceroni, M., S. Liu and R. Costanza. 2005. The Ecological and Economic Roles of
Biodiversity in Agroecosystems. In: Managing Biodiversity in Agroecosystems
Jarvis, D. I., C. Padoch and D. Cooper. (Eds.), Columbia University Press, NY USA.
Chavas, J. P. and M. T. Holt. 1996. Economic behavior under uncertainty: A joint
analysis of risk preferences and technology. Review of Economics and Statistics,
78(2): 329-335.
Chambers, E. A. and D. R. Cox. 1967. Discrimination between alternative binary
response models. Biometrika, 54(3-4): 573-578.
Champ, P. A., T. C. Brown and K. J. Boyle. 2004. A Primer on Nonmarket
Valuation. Dordrecht: Kluwer Academic Publishers.
Christensen, L. R., D.W. Jorgenson and L. J. Lau. 1973. Transcendental logarithmic
production frontiers. The Review of Economics and Statistics, 55(1): 28-45.
264
Ciriacy-Wantrup, S. 1947. Capital returns from soil conservation practices. Journal
of Farm Economics, 29:1181-1196.
Coelli, T. J. 1995. Recent developments in frontier estimation and efficiency
measurements. Australia Journal of Agricultural Economics, 39(3): 219-245.
Coelli, T. J. and G. E. Battese. 1996. Identification of factors which influence the
technical inefficiency of Indian farmers. Australian Journal of Agricultural
Economics, 40(2): 103-128.
Coelli, T. J. 1996. A guide to FRONTIER version 4.1: A Computer Program for
Stochastic Frontier Production and Cost Function Estimation, CEPA Working Papers
No. 7/96, Department of Econometrics University of New England: Armidale.
Coelli, T. J., Rao, D. S. P., C. J. O'Donnell and G. E. Battese. 2005. An Introduction
to Efficiency and Productivity Analysis. New York: Springer.
Czech, B. 2003. Technological progress and biodiversity conservation: a dollar
spent, a dollar burned. Conservation Biology, 17(5):1455-1457.
Dattalo, P. 2008. Determining Sample Size: Balancing Power, Precision and
Practicality. New York: Oxford University Press.
Davis, R. 1963. The Value of Outdoor Recreation: An Economic Study of the
Marine Woods. Ph.D. dissertation, Department Economics, Harvard University.
265
Department of Census and Statistics. 2002 .Census of Agriculture, Small Holding
Sector,Agriculture and Environmental Statistics Division Department of Census and
Statistics, Colombo, Sri Lanka.
Department of Census and Statistics in Sri Lanka. 2010. Statistical Abstract.
Agricultural and Environmental Statistics Division. Department of Census and
Statistics: Colombo.
De Janvry, A., M. Fafchamps and E. Sadoulet. 1991. Peasant household behaviour
with missing markets: some paradoxes explained. The Economic Journal, 101(409):
1400-1417.
Di Falco, S. and C. Perrings. 2003. Crop genetic diversity, productivity and stability
of agroecosystems. A theoretical and empirical investigation. Scottish Journal of
Political Economy, 50(2): 207-216.
Di Falco, S. and J. P. Chavas. 2009. On crop biodiversity, risk exposure and food
security in the highlands of Ethiopia. American Journal of Agricultural Economics,
91(3): 599-611.
Diwakar, P. A. and F. H. Johnsen. 2009. Valuation of crop genetic resources in
Kaski, Nepal: farmers' willingness to pay for rice landraces conservation. Journal of
Environmental Management, 90(1): 483-491.
266
Drucker, A. G., V. Gomez and S. Anderson. 2001. The economic valuation of farm
animal genetic resources: a survey of available methods. Ecological Economics,
36(1): 1-18.
Drucker, A. G., M. Smale and P. Zambrano. 2005. Valuation and sustainable
management of crop and livestock biodiversity: a review of applied economics
literature. CGIAR System-wide Genetic Resources Programme, International Food
Policy Research Institute. Kenya: Nairobi.
Evenson, R. E., D. Gollin and V. Santaniello. 1998. Agricultural Values of Plant
Genetic Resources. CAB International, Wallingford.
Fafchamps, M. 1992. Cash crop production, food price volatility and rural market
integration in the third world. American Journal of Agricultural Economics, 74(1):
90-99.
FAO. 2007. The State of Food and Agriculture 2007. Part I: Paying farmers for
environmental services. Rome.
FAO. 1999. Multifunctional character of agriculture and land. Conference
Background Paper No. 1. FAO, Rome.
FAO. 1999a. The Strategic Framework for FAO 2000-2015. Food and Agriculture
Organisation of the United Nations. Rome.
267
FAO. 1999b. Agricultural biodiversity, multifunctional character of agriculture and
land conference, Background Paper No. 1. Maastricht, Netherlands.
Feder, G. and D. Umali. 1993. The adoption of agricultural innovations: a review.
Technological Forecasting and Social Change, 43(3-4): 215-239.
Farrell, M. J. 1957. The measurement of productive efficiency. Journal of the Royal
Statistical Society, General, 120 (3): 253-290.
Franks, J. R. 1999. In situ conservation of plant genetic resources for food and
agriculture: a UK perspective. Land Use Policy, 16(2): 81-91.
Freeman III, A. M. 2003. The Measurement of Environmental and Resource Values:
Theory and Methods, (2nd
Eds.), Resources for the Future, Washington, D.C.
Ganesh, R. J. and S. Bauer. 2006. Determinants of rice variety diversity on
household farms in the Terai region of Nepal. Paper presented at the International
Association of Agricultural Economists Conference, Gold Coast, Australia.
Garber-Yonts, B. 2001. A Choice Experiment Analysis of Public Preferences for
Conservation of Biological Diversity in the Oregon Coast Range, Unpublished
Doctoral Dissertation, Oregon State University, USA.
Gauchan, D. and M. Smale. 2003. Choosing the “Right Tools” to assess the
economic costs and benefits of growing landraces: an illustrative example from Bara
district, Central Terai, Nepal. Plant Genetic Resources Newsletter,134:41-44.
268
Gauchan, D. 2004. Conserving Crop Genetic Resources On-Farm: The Case of Rice
in Nepal, Unpublished Doctoral Dissertation, University of Birmingham, UK.
Gauchan, D., Smale, M., Maxted, N., Cole, M., Sthapit, B. R., D. Jarvis and M. P.
Upadhyay. 2005. Socioeconomic and agroecological determinants of conserving
diversity on-farm: the case of rice genetic resources in Nepal. Nepal Agricultural
Resource Journal, 6: 89-98.
Ghaouti, L., Z. Vogt-Kaute and W. Link. 2008. Development of locally-adapted
fiber bean cultivars for organic conditions in Germany through a participatory
breeding approach. Euphytica, 162(2):257-268.
Greene, W. H. 1997. Econometric Analysis.(3rd
Eds.), New Jersey: Prentice Hall.
Greene, W. H. 2000. Econometric Analysis.(4th
Eds.), New Jersey: Prentice Hall.
Greene, P. E. and V. Srinivasan. 1990. Conjoint analysis in marketing: new
developments with implications for research and practice. The Journal of Marketing,
54(4): 3-19.
Grootendorst, P. V. 1995. A comparison of alternative models of prescription drug
utilisation. Health Economics, 4(3): 183-198.
Gujarati, D. 2003. Basic Econometrics. (4th
Eds.), New York: Graw-Hill Higher
Education.
269
Hadgu, K. M., Rossing, W.A.H., L. Kooistra and A. H. C. van Bruggen. 2009.
Spatial variation in biodiversity, soil degradation and productivity in agricultural
landscapes in the highlands of Tigray, northern Ethiopia. Food Security, 1:83-97.
Hadley, D. 2006. Patterns in technical efficiency and technical change at the farm-
level in England and Wales, 1982-2002. Journal of Agricultural Economics, 57(1):
81-100.
Hadri, K. and J. Whittaker. 1999. Efficiency, environmental contaminants and farm
size: testing for links using stochastic production frontiers. Journal of Applied
Economics, 2: 337-356.
Haji, J. 2006. Production efficiency of smallholders’ vegetable-dominated mixed
farming system in eastern Ethiopia: a non-parametric approach. Journal of African
Economics, 16(1): 1-27.
Hanemann, M. 1984. Welfare evaluations in contingent valuation experiments with
discrete responses. American Journal of Agricultural Economics, 66(3): 332-341.
Hanemann, M. and B. Kanninen.1999. The statistical analysis of discrete-response
CV data. In: Valuing Environmental Preferences, Bateman, I. and K. Willies (Eds.),
Oxford University Press.
Hanemann, W. M., J. Loomis and B. Kanninen. 1991. Statistical efficiency of double
bounded dichotomous choice contingent valuation. American Journal of Agricultural
Economics, 73: 1255-1263.
270
Hanley, N., MacMillan, D., Wright, R. E., Bullock, C., Simpson, I., D. Parsisson and
B. Crabtree. 1998. Contingent valuation versus choice experiments: estimating the
benefits of environmentally sensitive areas in Scotland. Journal of Agricultural
Economics, 49(1): 1-15.
Hausman, J. and D. McFadden. 1984. Specification tests for the multinomial logit
Model. Econometrica, 52(5): 1219-1240.
Heisey, P. W., Smale M., D. Byerlee and E. Souza. 1997. Wheat rusts and the costs
of genetic diversity in the Punjab of Pakistan. American Journal of Agricultural
Economics, 79(3): 726-737.
Hensher, D. A., J. M. Rose and W. H. Greene. 2005. Applied Choice Analysis: A
Primer. Cambridge University Press, Cambridge, UK.
Hengsdijk, H., Guanghuo, W., Van den Berg, M. M., Jiangdi, W., Wolf, J., Changhe,
L., R. P. Roetter and H. V. Keulen. 2007. Poverty and biodiversity trade-offs in rural
development: a case study for Pujiang county, China. Agricultural Systems,
94(3):851-886.
Hensher, D. A., J. M. Rose and W. H. Greene. 2005. Applied Choice Analysis: A
Primer. Cambridge University Press: Cambridge.
Hilbe, J. M. 2005. Negative Binomial Regression. New York: Cambridge University
Press.
271
Hilbe, J. M. 2011. Negative Binomial Regression.(2nd
Eds.), New York: Cambridge
University Press.
Hoang, V. and T. Coelli. 2009. Measurement of agricultural total factor productivity
growth in corporating environmental factors: a nutrients balance approach. Working
Paper, Centre for Efficiency and Productivity Analysis, The University of
Queensland, Australia.
Hole, D. G., Perkins, A. J., Wilson, J. D., Alexander, I. H., P.V. Grice and A. D.
Evans. 2005. Does organic farming benefit biodiversity? Biological Conservation,
122(1): 113-130.
Howard-Borjas, P. 1999. Some implications of gender relations for plant genetic
resources management. Biotechnology and Development Monitor, 37(3): 2-5.
Idiong, I. C. 2007. Estimation of farm level technical efficiency in small-scale
swamp rice production in cross river state of Nigeria: a stochastic frontier approach.
World Journal of Agricultural Sciences, 3(5): 653-658.
Isakson, S. R. 2007. Uprooting diversity? Peasant farmers’ market engagements and
the on-farm conservation of crop genetic resources in the Guatemalan highlands.
Working Paper 122, Political Economy Research Institute, University of
Massachusetts Amherst.
IUCN. 2007. The 2007 red list of threatened fauna and flora of Sri Lanka, IUCN,
Colombo, Sri Lanka.
272
Jeremy Carew-Reid, 2002. Biodiversity Planning in Asia. IUCN, Gland, Switzerland
and Cambridge, UK.
Jondrow, J., Lovell, C. A. K., I. S. Materov and P. Schmidt. 1982. On the estimation
of technical inefficiency in the stochastic frontier production function model. Journal
of Econometrics, 19(2-3): 233-243.
Johnson, F. R., K. E. Mathews and M. F. Bingham. 2000. Evaluating welfare-
theoretic consistency in multiple-response, stated-preference surveys. Triangle
Economic Research Technical Working Paper, No T-0003.Triangle Economic
Research, Durham.
Johnson, K. H., Vogt, K. A., Clark, H, J., O. J. Schmitz and D. J. Vogt. 1996.
Biodiversity and the productivity and stability of ecosystems. Trends in Ecology and
Evolution, 11(9): 372-377.
Karagiannis, G., P. Midmore and V. Tzouvelekas. 2002. Separating technical change
from time varying technical inefficiency in the absence of distributional assumptions.
Journal of Productivity Analysis, 18(1): 23-38.
Kassie, G. T., A. Abdulai and C. Wollny. 2009. Valuing traits indigenous cows in
central Ethiopia. Journal of Agricultural Economics, 60(2): 386-401.
Keller, G. B., H. Mndiga and B. L. Maass. 2006. Diversity and genetic erosion of
traditional vegetables in Tanzania from the farmer’s point of view. Plant Genet
Resource, 3(3): 400-413.
273
Kessels, R., P. Goos and M.Vandebroek. 2006. A comparison of criteria to design
efficient choice experiments. Journal of Marketing Research, 43(3): 409-419.
Kikulwe, E. M., Birol, E., J. Wesseler and J. Falck-Zepeda. 2011. A latent class
approach to investigating demand for genetically modified banana in Uganda.
Agricultural Economics, 42 (95): 547-560.
Kodde, D. A. and A. C. Palm.1986. Wald criteria for jointly testing equality and
inequality restrictions. Econometrica, 54(5): 1243-1248.
Kontoleon, A. 2003.Essays on Non-Market Valuation of Environmental Resources:
Policy and Technical Explorations. Unpublished Doctoral Dissertation, University of
London.
Kontoleon, A. and M. Yabe. 2003. Assessing the impacts of alternative ‘opt-out’
formats in choice experiment studies: consumer preferences for genetically modified
content and production information in food. Journal of Agricultural Policy and
Resources, 5: 1-43.
Kotagama, H. B. 2002. Financing environmental conservation: a case study in Sri
Lanka. Sri Lanka Journal of Agricultural Economics, 2(3): 12-18.
Krutilla, J. V. 1967. Conservation reconsidered. American Economic Review, 57(3):
777-786.
Kuhfeld W. 2005. Marketing research methods in SAS. Available from:
http://support.sas.com/techsup/technote/ts722.pdf.
274
Kumbhakar, S. C. and C. A. K. Lovell. 2000. Stochastic Frontier Analysis.
Cambridge: Cambridge University Press.
Kurosaki, T. and M. Fafchamps. 2002. Insurance market efficiency and crop choices
in Pakistan. Journal of Development Economics,67(2): 419-453.
Lancaster, K. 1966. A new approach to consumer theory. Journal of Political
Economy, 74(2): 132-157.
Latruffe, L., Balcombe, K., S. Davidova and K. Zawalinska. 2004. Determinants of
technical efficiency of crop and livestock farms in Poland. Applied Economics,
36(12): 1255-1263.
Latruffe, L., Balcombe, K., S. Davidova and K. Zawalinska. 2005. Technical and
scale efficiency of crop and livestock farms in Poland: does specialization matter?
Agricultural Economics, 32(3): 281-296.
Layton, D. F. and G. Brown. 2000. Hetergeneous preferences regarding global
climate change. The Review of Economics and Statistics, 82(4): 616-624.
Layton, D. and G. Brown. 1998. Application of stated preference methods to a public
good: Issues for discussion. Paper presented at the NOAA Workshop on the
Application of Stated Preference Methods to Resource Compensation, Washington,
DC.
Li-zhi, G. 2003.The conservation of Chinese rice biodiversity: genetic erosion,
ethnobotany and prospects. Genetic Resources and Crop Evolution,50(1): 17-32.
275
Louviere, J., M. Fox and W. Moore. 1993. Cross-task validity comparisons of stated
preference choice models. Marketing Letters, 4: 205-213.
Louviere, J. J., Hensher, D. A., J. D. Swait and W. Adamowicz. 2000. Stated Choice
Methods: Analysis and Applications. Cambridge University Press, Cambridge.
Luce, D. 1959. Individual Choice Behavior. New York: John Wiley.
Lukas, B. A. 2007. Marketing Research. New South Wales: McGraw-Hill Higher
Education.
Lusk, J. L., J. Roosen and J. A. Fox. 2003. Demand for beef from cattle administered
growth hormones or fed genetically modified corn: a comparison of consumers in
France, Germany, the United Kingdom and the United States. American Journal of
Agricultural Economic, 85(1): 16-29.
Lynn, S. J., J. W. Rhue and J. P. Green. 1988. Multiple personality and fantasy
proneness: is there an association or dissociation? British Journal of Experimental
and Clinical Hypnosis, 5: 138-142.
Maddala, G. S. 1999. Limited Dependent and Qualitative Variables in Econometrics.
Cambridge University Press, Cambridge.
Maddala, G. S. 2000. Introduction to Econometrics.(3rd
Eds.), Upper Saddle River,
NJ, USA: Prentice-Hall, Inc.
276
Maikhuri, R. K., K. S. Rao and R. L. Senwal. 2001. Changing scenario of Himalayan
agroecosystems: loss of agrobiodiversity, an indicator of environmental change in
Central Himalaya, India. The Environmentalist, 21(1): 23-39.
Manski, C. 1977. The structure of random utility models. Theory and Decision, 8(3):
229-254.
Mattison, E. H. A. and K. Norris. 2005. Bridging the gaps between agricultural
policy, land-use and biodiversity. Trends in Ecology and Evolution, 20(11): 610-616.
Matson, P. A., Parton, W. J., A. G. Power and M. J. Swift. 1997. Agricultural
intensification and ecosystem properties. Science, 277(5325): 504-509.
McFadden, D. 1974. Conditional Logit Analysis of Qualitative Choice Behaviour.
In: Frontiers in Econometrics, Zarembka, P. (Eds.), New York: Academic Press.
McFadden, D. and K. Train. 2000. Mixed MNL models of discrete response. Journal
of Applied Economics, 15(5): 447-470.
Meeusen, W. and J. van den Broeck. 1977. Efficiency estimation from Cobb-
Douglas production functions with composed error. International Economic Review,
18(2): 435-444.
Meng, E. C. H. 1997. Land Allocation Decisions and In Situ Conservation of Crop
Genetic Resources: The Case of Wheat Landraces in Turkey. Unpublished Doctoral
Dissertation, University of California at Davis, California, USA.
277
Meng, E. C. H., Smale, M., M. Bellon and D. Grimanelli. 1998. Definition and
measurement of crop diversity for economic analysis. In: farmers, gene banks and
crop breeding: economic analyses of diversity in wheat, maize and rice, M. Smale
(Eds.), Dordrecht and Mexico, D.F.: Kluwer and International Maize and Wheat
Improvement Centre.
Meng, E. C. H., Smale, M., Rozelle, S. D., R. Hu and J. Huang. 2003. Wheat Genetic
Diversity in China: Measurement and Cost. In: Agricultural Trade and Policy in
China: Issues, Analysis and Implications, Rozelle S. D. and D. A. Sumner. (Eds.),
Ashgate, Burlington, Vermont, USA, pp. 251-267.
Minot, N., B. Baulch and M. Epprecht. 2006. Poverty and inequality in Vietnam-
spatial patterns and geographic determinants. IFPRI Research Report No.148.
International Food Policy Research Institute: Washington, DC.
Ministry of Environment and Natural Resources in Sri Lanka.2007. National Focal
Point. Ministry of Environment and Natural Resources. Colombo.
Ministry of Forestry and Environment in Sri Lanka. 1999. Forest Sector
Statistics.Ministry of Environment and Natural Resources. Colombo.
Minot, N., B. Baulch and M. Epprecht. 2006. Poverty and inequality in Vietnam-
spatial patterns and geographic determinants. IFPRI Research Report No148.
International Food Policy Research Institute, Washington, DC
Mogas, J., P. Riera and J. Bennett. 2002. A comparison of contingent valuation and
choice modeling: estimating the environmental values of Catalonian forests.
278
Occasional paper no. 1, National Centre for Development Studies, Australian
National University.
Morey, E. and K. Rossmann. 2003. Using stated-preference questions to investigate
variations in willingness to pay for preserving marble monuments: classic
heterogeneity, random parameters, and mixture models. Journal of Cultural
Economics, 27(3-4): 215-229.
Mozumder, P. and R. P. Berrens. 2007. Inorganic fertilizer use and biodiversity risk:
an empirical investigation. Ecological Economics, 62(3-4): 538-543.
Mulatu, E. and K. Belete. 2001. Participatory varietal selection in lowland sorghum
in eastern Ethiopia: impact on adoption and genetic diversity. Experimental
Agriculture, 37(2): 211-229.
Nagarajan, L., M. Smale and P. Glewwe. 2007. Determinants of millet diversity at
the household-farm and village-community levels in the dry lands of India: the role
of local seed systems. Agricultural Economics, 36(2): 157-167.
Ndjeunga, J. and C. H. Nelson. 2005.Toward understanding household preference for
consumption characteristics of millet varieties: a case study from western Niger.
Agricultural Economics, 32(2): 151-165.
Negri, V. 2003. Landraces in central Italy: where and why they are conserved and
perspectives for their on-farm conservation. Genetic Resources and Crop Evolution,
50(8): 871-885.
279
Ogunyýnka, E. O. and I. A. Ajibefun. 2004. Determinants of technical inefficiency in
farm production: Tobit analysis approach to the NDE Farmers in Ondo State,
Nigeria. International Journal of Agriculture and Biology, Pakistan, 6(2): 355-558.
Ouma, E., A. Abdulai and A. Drucker. 2007. Measuring heterogeneous preferences
for cattle traits among cattle-keeping households in East Africa. American Journal of
Agricultural Economics, 89(4): 1005-1019.
Parikh, A. and K. Shah. 1995. Measurement of technical efficiency in the north-west
frontier province of Pakistan. Journal of Agricultural Economics, 45 (1): 133-137.
Pascual, U. and C. Perrings. 2007. Developing incentives and economic mechanisms
for in situ biodiversity conservation in agricultural landscapes. Agriculture,
Ecosystems and Environment, 121(3): 256-268.
Perrings, C. 2001. The Economics of Biodiversity Loss and Agricultural
Development in Low Income Countries. Environment Department, University of
York, Heslington.
Pimentel, D., Houser, J., Preiss, E., White, O., Fang, H., Mesnick, L., Barsky, T.,
Tariche, S., J. Schreck and S. Alpert. 1997. Water resources: agriculture, the
environment, and society. BioScience, 47 (2): 97-106.
Portney, P. R. 1994. The contingent valuation debate: why economists should care.
Journal of Economic Perspectives, 8(4): 3-17.
280
Poudel, D. and F. H. Johnsen. 2009. Valuation of crop genetic resources in Kaski,
Nepal: farmers’ willingness to pay for rice landraces conservation. Journal of
Environmental Management, 90(1): 483-491.
Primack, R. 1993. Essential of conservation biology, Sinauer Sinauer Associates,
Inc. Publishers: Sunderland, MA.
Rana, R. B., Garforth, C., Sthapit, B. R., A. Subedi and D. I. Jarvis. 2005. Influence
of socioeconomic and cultural factors on management of rice varietal diversity in
Nepal. In: On Farm Conservation of Agricultural Biodiversity in Nepal, Sthapit, B.
R., Upadhyay, M.P., Shrestha, P.K. and D. I. Jarvis (Eds.), International Plant
Genetic Resources Institute, Rome, Italy.
Revelt, D. and K. Train. 1998. Mixed logit with repeated choices: households'
choices of appliance efficiency level. The Review of Economics and Statistics, 80(4):
647-657.
Roesslera, R., Drucker, A. G., Scarpad, R., Markemanna, A., Lemkea, U., L.T.
Thuye and A. V. Záratea. 2008. Using choice experiments to assess smallholder
farmers' preferences for pig breeding traits in different production systems in North-
West Vietnam. Ecological Economics, 66(1): 184-192.
Rolfe, J., J. J. Bennett and J. Louviere. 2000. Choice modelling and its potential
application to tropical rainforest preservation. Ecological Economics,35(2): 289-302.
281
Romstad, E., Vatn, A., P. K. Rorstad and V. Soyland. 2000. Multifunctional
Agriculture: Implications for Policy Design. Department of Economics and Social
Sciences Report No. 21. Oslo: Agricultural University of Norway.
Ruto, E., G. Garrod and R. Scarpa. 2008. Valuing animal genetic resources: a choice
modeling application to indigenous cattle in Kenya. Agricultural Economics, 38(1):
89-98.
Salvatore, D., M.Bezabih and Y. Mahmud. 2010. Seeds for livelihood: crop
biodiversity and food production in Ethiopia. Ecological economics, 69 (8): 1695-
1702.
Sanjeeva, S. 2003. Gap Analysis of Biodiversity Conservation in Sri Lanka: A
Framework for Action. Report Prepared for the Ministry of Environment and Natural
Resources, Colombo.
Scarpa, R., Drucker, A. G., Anderson, S., Ferraes-Ehuan, N., Gomez, V., C. R.
Risopatron and O. Rubio-Leonel. 2003. Valuing genetic resources in peasant
economies: the case of ‘hairless’ Creole pigs in Yucatan. Ecological Economics,
45(3): 427-443.
Scarpa, R. and J. M. Rose. 2008. Designs efficiency for nonmarket valuation with
choice modelling: how to measure it, what to report and why. The Australian Journal
of Agricultural and Resource Economics, 52(3): 253-282.
Scarpa R., Ruto, E. S. K., Kristjanson, P., Radeny, M., A. G. Drucker and J. E. O.
Rege. 2003.Valuing indigenous cattle breeds in Kenya: an empirical comparison of
282
stated and revealed preference value estimates. Ecological Economics, 45(3):409-
426.
Seeth, H. T., Chachnov, S., A. Surinov and J. Von Braun. 1998. Russian poverty:
muddling through economic transition with garden plots. World Development,26(9):
1611-1623.
Seyoum, E.T., G. E. Battese and E. M. Fleming. 1998. Technical efficiency and
productivity of maize producers in eastern Ethiopia: a study of farmers within and
outside the Sasakawa- Global 2000 Project. Agricultural Economics, 19(3): 341-348.
Shapiro, K. H. 1983. Efficiency differentials in peasant agriculture and their
implications for development policies. Journal of Development Studies, 19(2): 77-
190.
Sharma, R. C., Chaudhary, N. K., Ojha, B., Yadav, L., M. P. Pandey and S. M.
Shrestha. 2007. Variation in rice landraces adapted to the low lands and hills in
Nepal. Plant Genetic Resources: Characterization and Utilization, 5(3): 120-127.
Shehu, H. E., J. D. Kwari and P. M. Bzugu. 2007. An exploratory survey of soybean
production as influenced by soil nutrient status in Northeastern Nigeria. Journal of
Agronomy, 6(4): 576-580.
Sherlund, S. M., C. B. Barrett and A. A. Adesina. 2002. Smallholder technical
efficiency controlling for environmental production conditions. Journal of
Development Economics, 69(1): 85-101.
283
Singh, K., Ram, P., V. Singh and S. K. Kothari. 1986. Effect of dates of planting and
nipping on herb and oil yield of Mentha arvensis. Indian Journal of Agronomy, 31:
128-130.
Slingenberg, A., Braat, L., van der Windt, H., L. Eichler and K. Turner. 2009. Study
on understanding the causes of biodiversity loss and the policy assessment
framework. European Commission Directorate-General for Environment, Rotterdam:
The Netherlands.
Smale, M., M. Bellon and A. Aguirre. 2001. Maize diversity, variety attributes, and
farmers’ choices in Southeastern Guanajuato, Mexico. Economic Development and
Cultural Change,50(1): 201-225.
Smale, M.,Hartell, J., P. W.Heisey and B. Senauer. 1998. The contribution of genetic
resources and diversity to wheat production in the Punjab of Pakistan. American
Journal of Agricultural Economics, 80(3): 482-493.
Smale, M., Meng, E., J. P. Brennan and R. Hu. 2002. Determinants of spatial
diversity in modern wheat: Examples from Australia and China. Agricultural
Economics, 28(1):13-26.
Swanson, T. 1996. Global values of biological diversity: the public interest in the
conservation of plant genetic resources for agriculture. Plant Genetic Resources
Newsletter, 105: 1-7.
Swanson, T. 1997. Global action for biodiversity. Earthscan, London.
284
Swanson, T. 1999. The Underlying Causes of Biodiversity Decline: An Economic
Analysis, IUCN, Gland, Switzerland.
Stevenson, R. E. 1980. Likelihood functions for generalized stochastic frontier
estimation. Journal of Econometrics, 13(1): 57-66.
Taylor, E. and I. Adelman. 2003. Agricultural household models: genesis, evolution
and extensions. Review of the Economics of the Household, 1(1): 33-58.
Thrupp, L.A. 1998. The central role of agricultural biodiversity: trends and
challenges. In Conservation and sustainable use of agricultural biodiversity. Manila,
CIP-UPWARD in partnership with GTZ, IDRC, IPGRI and SEARICE.
Thurstone, L. L. 1927. A law of comparative judgment. Psychological Review, 34
(4): 278-286.
Tilman, D., Matson, G.K., P.A. Naylor and S. Polasky. 2002. Agricultural
sustainability and intensive production practices. Nature, 418:671-677.
Tsegaye, B. 1997. The significance of biodiversity for sustaining agricultural
production and role of women in the traditional sector: the Ethiopian experience.
Agriculture, Ecosystems and Environment, 62 (2-3): 215-227.
UNEP, 1995. Global Biodiversity Assessment. Cambridge, Cambridge University
Press
285
Van Dusen, E. 2000. In Situ Conservation of Crop Genetic Resources in Mexican
Milpa Systems.Unpublished Doctoral Dissertation.University of California, Davis,
USA.
Van Dusen, E. 2005. Understanding the Factors Driving on Farm Crop Genetic
Diversity: Empirical Evidence from Mexico. In: Agricultural Biodiversity and
Biotechnology in Economic Development, Cooper, J., L. M. Lipper and D.
Zilberman (Eds.), Springer: US, pp. 127-145.
Van Dusen, E. and J. E. Taylor. 2005. Missing markets and crop diversity: evidence
from Mexico. Environment and Development Economics, 10(4): 513-531.
Van Dusen, E., D. Gauchan and M. Smale. 2005. On farm conservation of rice
biodiversity in Nepal. Presented at the American Agricultural Economics
Association Annual Meeting, Providence, Rhode Island.
Vangelis, T., C. J. Pantzios and F. Christos. 2001. Technical efficiency of alternative
farming systems: the case of Greek organic and conventional olive-growing farms.
Journal of Food Policy, 26(6): 549-569.
Villano, R. and E. Fleming. 2006. Technical inefficiency and production risk in rice
farming: evidence from central Luzon Philippines. Asian Economic Journal, 20(1):
29-46.
Widawsky, D. and S. Rozelle. 1998. Varietal diversity and yield variability in
Chinese rice production. In: Farmers, Gene Banks, and Crop Breeding: Economic
286
Analyses of Diversity in Wheat, Maize, and Rice, Smale, M. (Eds.), Kluwer,
Dordrecht and CIMMYT, Mexico, pp. 159-172.
Wijesinghe, L., Gunatilleke, I., Jayawardana, S. D. G., S. W. Kotagama and C.V.S
Gunatilleke. 1993. Biological Conservation in Sri Lanka: A National Status
Report.IUCN Sri Lanka Country Office. Colombo.
Wilson, P., D. Hadley and C. Asby. 2001. The influence of management
characteristics on the technical efficiency of wheat farmers in eastern England.
Agricultural Economics, 24(3): 329-338.
Wilson, C. and C. Tisdell. 2006. Globalization, concentration of genetic material and
their implication for sustainable development. In: Leading Economic and Managerial
Issues Involving Globalisation, Aurifeille, J., S. Svizzero and C. Tisdell. (Eds.),
Nova Science Publishers, Inc., United States of America, New York, pp. 251-262.
Winkelmann, R. 2008. Econometric Analysis of Count Data. (5th
Eds.): Springer.
Winters, P., L. H. Hintze and O. Ortiz. 2005. Rural development and the diversity of
potatoes on farms in Cajamarca, Peru. In: Valuing Crop Biodiversity: On-Farm
Genetic Resources and Economic Change. Smale, M. (Eds.), CABI Publishing,
Wallingford, UK.
Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA, USA: Massachusetts Institute of Technology.
287
World Conservation Monitoring Centre. 1992. Global Biodiversity: Status of the
Earth's Living Resources. London: Chapman and Hall.
Wunsch, D. 1986. Survey research: determining sample size and representative
response. Business Education Forum, 40(5): 31-34.
Xu, Y. 1997. Contextual tonal variations in Mandarin. Journal of Phonetics, 25(1):
61-83.
Xu, X. and S. R. Jeffrey. 1998. Efficiency and technical progress in traditional and
modern agriculture: evidence from rice production in China. Agricultural Economics,
18(2): 157-165.
Zander, K. K. and A. G. Drucker. 2008. Conserving what's important: using choice
model scenarios to value local cattle breeds in East Africa. Ecological Economics,
68(1-2): 34-45.
288
Appendix A (1): Defining agricultural biodiversity
Source: FAO, 1999a
Biodiversity
Agricultural
biodiversity
Mixed agro-ecosystems
Crop species/varieties
Livestock and fish species
Plant/animal germplasm
Soil organisms in cultivated areas
Biocontrol agents for crop/livestock pests
Wild species as landraces or with breading
Cultural and local knowledge of diversity
289
Appendix A (2): TEV of agricultural biodiversity on small-scale farm
290
Appendix A (3): Defining TEV of agricultural biodiversity on small-scale farms
Biodiversity
components
Use values Non-use values
Direct use
values
Indirect use values Option value Bequest value Altruistic
value
Existence value Cultural value
Crop
diversity
Output,
quality and
quantity of
food, cash
income,
productivity
gains
Improvement
of function
such as eco-
system
productivity,
soil and
water cycle
quality, habitat
protection
Its potential to provide
economic benefits to
human society in the
future, such as being
inputs to improvement
of many varieties and
breeds
Some individuals
may value the fact
that the future
generations will
have the
opportunity to
enjoy an
environmental
asset, such a
picturesque
landscape
Others may
be concerned
that the good
is available
for others in
this
generation,
whether or
not they use it
themselves
Individuals may
value the simple
fact that an
environmental
asset exists,
whether or not it
is used by these
individuals
The traditional or
indigenous
knowledge
associated with
certain crop
varieties, seed or
breed management
or farming
techniques
Cultural values
embedded in
traditional
varieties, i.e.
landraces, with
which traditional
Sri Lankan dishes
are cooked
Agro-
diversity
Landrace
cultivation
Developing
resistance,
Improvement of
function such as
eco- system
Productivity, soil and
water cycle quality,
habitat
protection
Option values of
exploration and
insurance value,
Livestock
diversity
Output,
quality and
quantity of food,
cash income
Increase soil quality Its potential to provide
economic benefits to
human society in the
future, such as being
inputs to improvement
of many varieties and
breeds Organic
production
Productivity
gains
Increase soil and
water quality
291
Appendix B: Number of described species in the World
Group Number of described species
Bacteria and blue-green algae 4,760
Fungi 46,983
Algae 70,900
Bryophytes (mosses and liverworts) 17,000
Gymnosperms (conifers) 750
Angiosperms (flowering plants) 250,000
Protozoans 30,800
Sponges 5,000
Corals and Jellyfish 9,000
Roundworms and earthworms 24,000
Crustaceans 38,000
Insects 751,000
Other Arthropods and minor invertebrates 132,461
Mollusks 50,000
Starfish 6,100
Fishes (teleosts) 19,056
Amphibians 4,184
Reptiles 6,300
Birds 9,198
Mammals 4,170
Total 1,435,662
Source: World Conservation Monitoring Centre (1992)
292
Appendix C: Biodiversity wilderness areas in the world
California Floristic Province
Madrean Pine-Oak Woodlands
Mediterranean
Basin
Caucasus
Irano-
Mountains of Central Asia
Mountains of Southwest China
Japan
Mesoamerica
Caribbean Islands
Anatolian Himalaya
Western
Ghats
IIInnndddooo--- Burma
Philippines
Polynesia-
Polynesia -
Micronesia
Tumbes- Chocó- Magdalena
Cerrado
Guinean Forests
of West Africa
Horn of Africa
and Sri Lanka
Sundaland
Wallacea
Micronesia East Melanesian Islands
New Zealand
Tropical Andes
Chilean
Winter Rainfall- Valdivian Forests
Atlantic Forest
Succulent
Karoo
Cape Floristic Region
Madagascar and the Indian Ocean Islands
Coastal Forests of Eastern Africa
Maputaland- Pondoland-Albany Wilderness areas
Southwest Australia New Zealand
Source: World Conservation Monitoring Centre (1992)
293
Appendix D (1): Topography in Sri Lanka
Source: Adopted as Dela (2007)
294
Appendix D (2): Major climatic zones in Sri Lanka
Source: Ministry of Environment and Natural Resources in Sri Lanka (2007)
295
Appendix E: Protected areas under department of wildlife in Sri Lanka
Source: Ministry of Environment and Natural Resources in Sri Lanka (2007)
296
Appendix F: List of protected areas of Sri Lanka
Sanctuaries Area
(ha)
Date of
declaration Protected area Area
(ha)
Date of
declaration
Chundikulam 11,149.10 25/02/1938 Parapuduwa Nuns' Island 189.6 17/08/1988
Wilpattu North 632 25/02/1938 Kahalla-Pallekele 21,690 1/07/1989
Telwatta 1,424.50 25/02/1938 Sigiriya 5,099 26/01/1990
Weerawila-Tissa 4,164.20 27/05/1938 Bellanwila-Attidiya 372 25/07/1990
Katagamuwa 1,003.60 27/05/1938 Bar Reef 30,669.9
0 3/04/1992
Polonnaruwa 1,521.60 27/05/1938 Nimalawa 1,065.80 18/02/1993
Tangamale 131.5 27/05/1938 Madunagala 995.2 30/06/1993
Mihintale 999.6 27/05/1938 Muthurajawela block I 1,028.60 31/10/1996
Kataragama 837.7 27/05/1938 Muthurajawela block II 256.8 31/10/1996
Anuradhapura 3,500.50 27/05/1938 Anawilundawa 1,397 11/06/1997
Udawatta Kele
Sanctuary
104 29/07/1938 Elahera-Girithale 14,035.2
0 13/01/2000
Rocky Islets 1.2 25/10/1940 Dahaiyagala 2,685.10 7/06/2002
Peak Wilderness
Sanctuary
22,379.10 25/10/1940 Tabbowa 2,193.30 19/07/2002
Kurulu Kele (Kegalle) 113.3 14/03/1941 Rumassala 170.7 3/01/2003
Pallemalala 13.7 23/10/1942 Kiralakele 310 8/09/2003
Welhilla Kategilla 134.3 18/02/1949 Eluwiliyaya 186 11/09/2009
Kokkilai 1,995 18/05/1951 Kaudulla-Minneriya 8,777.30 1/06/2004
Senanayake Samudra 9,324 12/02/1954 Kirama 45.7 6/10/2004
Gal Oya North-East 12,432 12/02/1954 Kudumbigala 6,533.90 20/02/2006
Gal Oya South-East 15,281 12/02/1954 Rekawa - 25/05/2006
Giant's Tank 4,330.10 24/09/1954 Godawaya - 25/05/2006
Vavunikulam 4,856.20 21/06/1963 Bundala - Wilmanna 3,339.40 30/06/2006
Sakamam 616.4 21/06/1963 Maduganga 2,300 17/07/2006
Padawiya Tank 6,475 21/06/1963 Nature reserves
Naval Headworks
Sanctuary 18,130 21/06/1963 Triconamadu 25,019.3
0 24/10/1986
Great Sober Island 64.7 21/06/1963 Riverine 824.1 31/07/1991
Little Sober Island 6.5 21/06/1963 Minneriya-Girithale
Kimbulwana Oya 492.1 21/06/1963 II block 1,923.60 25/06/1993
Mahakanadarawa Wewa 519.3 9/12/1966 III block 4,745.30 7/07/1995
Madhu Road 26,677 28/06/1968 IV block 8,335.50 1/09/1997
Seruwila-Allai 15,540 9/10/1970 Wetahirakanda 3,229 7/06/2002
Paretitivu Island 97.1 18/05/1973 Strict nature reserves
Honduwa Island 8.5 19/11/1973 Hakgala 1,141.60 5/02/1938
Buddhangala 1,841.30 1/11/1974 Yala 28,904.7
0 10/03/1939
Ravana Falls 1,932 18/05/1979 Ritigala 1,528.10 7/11/1941
Medinduwa 0.8 6/06/1980 Kalametiya lagoon 2,525.20 1/11/1984 Sri Jayawardenapura
birds sanctuary
449.2 9/01/1985 Victoria-Randenigala-
Rantambe 42,087.30 30/01/1987
Maimbulkanda -
Nittambuwa 25.1 8/06/1988
Source: Department of Survey in Sri Lanka (2007).
297
Appendix G: Map showing survey areas in Sri Lanka
298
Appendix H: Questionnaire used in the survey
Note: This questionnaire was translated into Sinhalese for the final survey
Agricultural biodiversity, Poverty and
farm level efficiency: Survey
A Survey by K.M.R. Karunarathna PhD candidate
Queensland University of Technology Australia
We greatly appreciate your participation in this survey
Good morning/ afternoon/ evening. My name is …………………., I am conducting this survey
on behalf of Ms Muditha Karunarathna who is a PhD Student at the Queensland
University of Technology, Australia. Her thesis is on agricultural biodiversity, poverty and
farm level efficiency in Sri Lanka. We have selected a sample of farmers to represent your
area and your farm has been chosen as part of the sample. I am visiting you today for this
survey.
By participating in this survey you will be assisting us to better understand and identify
the value of agricultural biodiversity in farms in Sri Lanka. Please be assured that this is
purely a research project and we do not represent any business or product or a
government institution. No government action will be involved as a result of your
participation in this study. We assure you that all the information that you provide us will
remain confidential. Please feel free to give any answer that you think is correct or
appropriate. We would appreciate it very much if you could spend some time with us and
answer some questions to the best of your ability. The survey should not last longer than
one and half hours. Would you be willing to take part in this survey? Yes ….1 No……...2 Note: If No, the enumerator will leave the farm Questionnaire No:……. Village: …………………...................... District: …………….............................. Date of Interview:…………............ Enumerator Name:………................ Time Started: ………….................... Time Finished:………......................... Muditha Karunarathna can be contacted in the next few months at the following address: Department of Economics and Statistics University of Peradeniya Sri Lanka TP: 81-239 2622 (office)/071-806 1246 (mobile) Email: [email protected], [email protected]
299
Part A: General Information on Farm Characteristics
Interviewer: The following questions relate to the general information about your
farm. Firstly, we would like to find out about your farm, and the methods you use to
cultivate them. Please concentrate only on the last cultivation season.
1. What is the size of your farm? Please state in acres ....................
2. For how long have you cultivated your land? Number of years: ………………
3. How far is your house from your farm? Number of kilometres:...........................
4. Could you please tell us the number of separate plots that you have used for the
following?
Crops (No.):................................
Livestock1(No.):.........................
Poultry2(No.)...............................
Mix-both crops and livestock and/or poultry (No.)............
5. What is the most important factor that you would consider when making
investment decisions on your farm?
1. Revenue 4. Water availability
2. Market prices 5. Household consumption
3. Capital availability 6. Other (specify)...................
6. How would you rank your farm with respect to its soil fertility?
1. Excellent 2. Good 3. Average 4. Poor
7. How would you rank your farm with respect to its land shape?
1. Very steep 2. Average 3. Flat
8. What is the extent of the irrigated area of your farm. ................% (state as a
percentage)?
1 Livestock refers to one or more domesticated animals raised in an agricultural setting to produce commodities such as food,
fibre and labour (e.g. cattle, cow, pig , goat... etc.). The term "livestock" does not include poultry or farmed fish. 2 Poultry is the category of domesticated birds that people keep for the purpose of collecting their eggs, or killing for their meat
and feathers (chickens, ducks ...etc.)
300
9. Have you received adequate water during the last season from the irrigation canal?
1. Yes, all the farm needs have been met
2. Yes, part of the farm needs have been met
3. No, did not receive any irrigated water
4. My farm does not rely on irrigation
10. Could you please tell us the total land area (acres) that you have used for
agricultural purposes during the last season under the following headings?
11. Do you think that the age of your farm has an influence on the productivity of
your farm?
1. Yes 2. No 3. Don’t know
12. Do you use the farm to do the following: (Please tick relevant box/boxes)
Types of farm Starting year
1 Grow crops only
2 Livestock and poultry only
3 Mix (both crops and livestock and/or poultry)
If you tick number 1 please go to section B and answer all questions except 4-5
If you tick number 2 please go to section B and answer all questions except 1,2 and 3
If you tick number 3 please go to section B and answer all questions
Owned Rented out Rented in Other
301
Part B : Information on Agricultural Biodiversity and Farm Level Efficiency
Interviewer: In this section, we are interested in getting some information about the
different components of agricultural biodiversity and the level of efficiency on your
farm.
Note: The enumerator will first give a broad introduction on diverse farming
systems, practiced in different areas in Sri Lanka and will then narrow down his
attention to the farming system in the survey areas.
1. Could you please provide us the following information with regards to the crops
you have cultivated, input you have used and the market prices that you have
received on your farm during the last season?
Crop Area
(m2)
Traditional
variety or
not (Yes/No)
Fertilizer
(Code)
Pesticides
(Yes/No)
Production
(kg.)
Market
price
(Rs)
Market
value
(Rs)
HH
consumption
(%)
1..........
2..........
3..........
4..........
5..........
6..........
Total
Fertilizer code: 0- no fertilizer, 1- chemical , 2- organic
302
2. Could you please let me know the amount of labour days used to cultivate the
above mentioned crops under following categories?
Items Hired labour (days) Family labour (days)
Preparing the Land
Cultivating the crops
Applying pesticides and fertilizers
Harvesting
Others…… (specify)
3. Please provide me details of your expenditure on the following items used to
cultivate the above mentioned crops:
Items Quantity Rs.
Tractor
Seeds
Pesticides
Fertilizer
Others
4. Could you please provide us the following information about livestock and poultry
production on your farm during the last season?
Livestock
and poultry
No. of
head
Area
(m2)
Traditional
breed or not
(Yes/No)
Production
(kg/litter/no)
Market
price
(Rs.)
Market
value
(Rs.)
HH
consumption
1........................
2........................
Total
303
5. Please provide me your expenditure on livestock and poultry under the following:
Items Quantity Rs.
Livestock
Labour
Feed
Veterinary
Other
Poultry
Labour
Feed
Veterinary
Other
6. What is the most common way of marketing your agricultural products?
Co-op Village trader/shop Village pola Town
Crops
Livestock
Poultry
What is the distance to the nearest town? ………………...........(in km)
What is the distance to the second nearest town? ………………(in km)
7. Have you been satisfied with the prices that you have received during the last
season?
Satisfied Not satisfied Don’t know
304
8. Could you please let us know what prices you were expecting and what prices you
obtained for the three most important crops and livestock/poultry sold in the market
during the last season?
9. Could you please provide us with the maximum and minimum prices you have
received for the three main crops and for livestock/poultry products you have sold in
the last two seasons?
10. What is the distance to the nearest market (km)? ...............................
11. Do you have any facilities to access alternative markets? Yes/No
If Yes, what is the distance to the alternative market(km)?.........................
12. Do you directly sell your farm product in the market? Yes/ No
If No, how do you sell them?........................................................................
Crops Expected
price (Rs.)
Actual
price (Rs.)
Livestock/
poultry
Expected
price (Rs.)
Actual price
(Rs.)
1............... 1..................
2............... 2..................
3............... 3...................
Crops Maximum
price (Rs.)
Minimum
price (Rs.)
Livestock/
poultry
Maximum
price (Rs.)
Minimum
price(Rs.)
1.......... 1............
2.......... 2............
3.......... 3............
305
13. Do you participate in agricultural extension services ?Yes/No
If Yes, how many times did you participate in the last season? ..............
14. Have you received any subsidies for agricultural production? Yes/No
If Yes, what is the approximate amount(Rs.)?.............................................
15. How do you finance your farm cultivation?
1. Savings
2. Money borrowed from private individuals
3. Money borrowed from traders
4. Money borrowed from the financial institutions
5. Other……………...(please specify)
16. What is the amount of family expenditure covered from farm production (where
Applicable)? 1. Crops …... (%) 2. Livestock....... (%) 3. Poultry........
(%)
17. How much money will you be investing on your farm next season?
Rs…………..(approximate amount)
18. Assume that your profits will increase by any of the
percentages shown below. Taking this into consideration by how much will you
increase your farm investment?
Profits % 0 50 100 More
Investment %
306
Water sources and use on the farm
19. Where is your farm located within the field canal? (please tick the appropriate
box)
20. What is your main source of water on your farm used for cultivation? (tick the
appropriate box)
If you tick number 1 please go to question 21.1
If you tick number 2 please go to questions 21.2
If you tick number 3 please answer all questions (21.1, 21.2 and 21.3)
21.1. Agrowell
Please provide me details about pumping water from the agrowell to your farms
a. How do you pump water from the agrowell?
1. Using my own pump 2. Hired pump
b. For how many hours is the pump used per day?
Number of hours (approximately):………………
c. For how many days per month is the pump used?
Number of days per season (approximately):…………
d. How long is the cultivation season during the Yala/Maha season?
Number of months per season (approximately):……………….…..……
e. What is the size of the water pump (h/power):……………………….…
Note: Enumerator will check pump size by examining the pump
Head Middle Tail
1.Agrowell 2. Field canal 3.Both
307
21.2. Field canal
Please give me details about using water from the field channel to your farm
a. How do you obtain water to your farm from the field canal?
Yala Maha
Water flows continuously throughout the season
A rotational system (water access is restricted)
Any other system (specify)………..
b. For how many hours is water taken per day?
Number of hours (approximately):…………
c. For how many days per month is water taken?
Number of days (approximately):…………
d. How long is the cultivation season during the Yala/Maha season?
Number of months:…………
21.3. What proportion of your total water requirements do you obtain from different
sources?
Please state the approximate percentage: Agrowell ………...
Field canal ..…....…
Rain water……….
308
Part C: Evaluating Poverty, Income and Expenditure
Interviewer: Now we are going to ask you about your income and expenditure. The
main purpose of obtaining this information is to evaluate the relationship between
your farming system and your farm income. In addition to that, we are interested in
seeing whether there is a way to improve your farm income by changing existing
agricultural practices.
1. How do you rank the availability of food in your household in a typical year?
1. We have enough food for consumption
2. Very rarely we have insufficient food
3. Very often we are running short of food
2. In your view do you think that your household is healthy?
1. Yes 2. No
If Yes, what is the reason?……………………………………………….…..
If Not, what is the reason?……………………………………………….…..
3. Socio-economic status/income level of the household. (Note: This assessed by
observation of the enumerator. The enumerator will take photographs that define
socio-income status)
1. Luxury/ Upper middle class
2. Ordinary
3. Small house/Cottage
309
4. We now ask questions related to facilities available in your house. Could you
please let us know whether you have the following facilities in your house?
Facilities Yes No
1 Telephone
2 Electricity
3 Pipe water
4 Vehicle road to the house
5 Water sealed toilet
6 Attached bathroom
5. In this question we are asking about the capital assets that you own. Could you
please provide us all the capital assets with their purchased values
Assets Quantity Approximate Value (Rs) Year of purchase
Tractors
Threshing machine
Water pump
Vehicles
Motorcycle
Other(Specify)…
6. Did you purchase any land or/and houses over the last 5 years
1. Yes 2. No If Yes value:..........
7. Monthly Income (family): From farm:
1. Crop (Rs.)............ 2. Livestock (Rs.)................ 3. Poultry (Rs.).......
Other sources (Rs.): ......………………...
310
8. Household expenditure last month
Items Rs.
1 Household living
2 Education
3 Health
4 Other
9. Does any member of the household receive a pension or direct welfare payment?
1. Yes 2. No
If Yes, please indicate the number of person(s) and the nature of such
contribution
No of Persons Amount (Rs)
Pension
Samurdi
Other
10. Could you please provide us the details of your total debt up to last month
Amount(Rs)
1 Debt owing to the informal sector
2 Debt owing to the formal sector
11. How would you classify the economic status of your household relative to others
in this village? (put the appropriate number):...........
a. Much better than most people (rich)
b. Better than most people (relatively well off)
311
c. About average
d. Below average
e. Much worse than average (very poor)
f. Don’t know / Not sure
12. Which one of the statements below is true for your household? Please choose
only one.
a. We can hardly make ends meet.
b. We can only afford the necessities
c. We do not have any financial problems, however we do not live in luxury
d. We have enough money to live a comfortable life
e. We live a comfortable life, sometimes we can afford luxury goods
f. We live in luxury
Part D: Conservation of Agricultural Biodiversity: Choice Experiment Survey
Enumerator: This part of the questionnaire is about farmers’ preference on
agricultural biodiversity on farms. We are interested in how you, as a farmer as well
as a consumer, perceive agricultural biodiversity and its different characteristics.
In this part we would like to find out the important of different components of
agricultural biodiversity and your own farm preferences using different attributes
level. Therefore, with the help of several farm producers and agricultural scientists
we have identified five components of agricultural biodiversity and generated several
312
(imaginary) farm profiles using differing levels of these characteristics. Farm
characteristics and their levels include:
1. Crop species diversity. This is measured by the total number of crop varieties that
are grown on a small-scale farm setting. For example, a farm with tomatoes, beans
and carrots has in total three different crops. We will present you with four levels of
crop diversity which involve 3, 7, 10, and 15 different varieties.
2. Mix crop and Livestock diversity: This is designed to indicate whether you prefer
an integrated crop and livestock/ poultry production system over a system that is
specialised in crops or livestock/poultry.
3.Organic production. This indicates whether or not a small-scale farm employs
organic methods of production. For example, when a farmer sells small-scale farm
crops that are produced entirely by employing organic methods, these products are
certified as organic. Consider your imaginary farm. Decide whether or not you prefer
a farm in which you produce crops with entirely organic methods.
4. Landrace cultivation. This shows whether or not you prefer to have a farm in
which a landrace is grown as opposed to none. A landrace is defined as a crop variety
that was grown by farmers, such as you or your ancestors, before the agricultural
modernisation programs commenced during the 1970s.
5. Economic importance of small-scale farms. This is defined as the expected
proportion (in percentage terms) of annual household food expenditure reduction
313
through food production in the small-scale farm. It indicates the importance of the
contribution of the small-scale farm production to your household budget. The
percentages that will be presented to you are 5%, 10% and 15%.
6. Estimated cost in terms of additional labour days. This is defined as a percentage
of additional labour requirements under different policy options. It indicates the
additional costs that you have to bear when you are accepting a new policy. The
percentages that will be presented to you are 10%, 20% and 30%.
The first four attributes reflect the various attributes of agricultural biodiversity
found in the farms in Sri Lanka. The sixth factor represents benefits that farmers can
receive in terms of net revenue changes under different policy options. The last
factor is the monetary attribute in terms of additional labour costs that farmers have
to use under different policy options.
We have placed the generated hypothetical farms in pairs on a series of cards, and we
would like you to indicate out of the pair, which type of farm you prefer in each card.
Now, please imagine you will cultivate a hypothetical farm. The following questions
will each present you with two different farms: farm (A) and farm (B), in each case
the farm is equal to half an acre. Could you please compare each farm in the
following cards I will be presenting to you and tell me which one you prefer in each
case?
314
Question 1
Assuming that the following farms were the ONLY choices you have, which one
would you prefer to cultivate?
Farm Characteristics Farm
(A)
Farm
(B)
Neither
Small-
scale farm
(A) nor
Small-
scale farm
(B):
Total number of crop varieties grown on a farm 10 10
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No No
Decrease in food expenditure (in percentage) 15% 10%
Estimated cost in terms of additional labour requirement
( in percentage)
20% 10%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
Question 2
Assuming that the following small-scale farms were the ONLY choices you have,
which one would you prefer to cultivate?
Farm Characteristics Farm
(A)
Farm
(B)
Neither
Small-scale
farm (A)
nor Small-
scale farm
(B):
Total number of crop varieties grown on a farm 10 5
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No Yes
Decrease in food expenditure (in percentage) 15% 10%
Estimated cost in terms of additional labour requirement
( in percentage)
10% 30%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
315
Question 3
Assuming that the following small-scale farms were the ONLY choices you have,
which one would you prefer to cultivate?
Farm Characteristics Farm
(A)
Farm
(B)
Neither
Small-
scale farm
(A) nor
Small-
scale farm
(B):
Total number of crop varieties grown on a farm 10 10
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No No
Decrease in food expenditure (in percentage) 5% 10%
Estimated cost in terms of additional labour requirement
( in percentage)
10% 30%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
Question 4
Assuming that the following small-scale farms were the ONLY choices you have,
which one would you prefer to cultivate?
Farm Characteristics Farm
(A)
Farm
(B)
Neither
Small-
scale farm
(A) nor
Small-
scale farm
(B):
Total number of crop varieties grown on a farm 10 5
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No Yes
Decrease in food expenditure (in percentage) 20% 30%
Estimated cost in terms of additional labour requirement
( in percentage)
30% 10%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
316
Question 5
Assuming that the following small-scale farms were the ONLY choices you have,
which one would you prefer to cultivate?
Farm Characteristics Farm (A) Farm
(B)
Neither
Small-
scale
farm (A)
nor
Small-
scale
farm (B):
Total number of crop varieties grown on a farm 10 10
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No No
Decrease in food expenditure (in percentage) 5% 10%
Estimated cost in terms of additional labour requirement
( in percentage)
10% 20%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
Question 6
Assuming that the following small-scale farms were the ONLY choices you have,
which one would you prefer to cultivate?
Farm Characteristics Farm
(A)
Farm
(B)
Neither
Small-
scale farm
(A) nor
Small-
scale farm
(B):
Total number of crop varieties grown on a farm 10 5
Crops are combined with livestock/poultry production Yes No
Farm crops are produced entirely using organic methods Yes Yes
Farm has a landrace cultivation No Yes
Decrease in food expenditure (in percentage) 30% 15%
Estimated cost in terms of additional labour requirement
( in percentage)
5% 10%
I prefer to cultivate Farm (A)…...................
Farm (B)…....................
Neither Farm .......….… (please pick one option)
317
7. When answering Questions 1 to 6, which of the five implications were mostly
important to you and which were the least important? Please rank the five
implications by placing the numbers 1 to 6 in the following boxes: (1-most
important; 5-least important)
Total number of crop varieties grown on the farm
Number of animal breeds on the farm
Farm production is combined with livestock/ poultry production
Farm crops produced entirely using organic methods
Farm has at least one landrace
8. If you always chose neither option, which of the following statements most closely
described you reason for doing so?
I oppose to increase agricultural biodiversity on the farm
I don’t want to change the existing system
I believe that change will increase the risk of farm production
I didn’t know which option was best so I stuck with the current
situation
Other reasons (specify) ..............................................................
................................................................
................................................................
318
9. Thinking about Questions 1 to 6, and the information about the agricultural
biodiversity on farms presented earlier, please indicate how strongly you agree or
disagree with each of statement a) to g) below. For each statement, please circle the
number that represents your view:
AS AG NA DI DS
1. I understood the information in the questionnaire
2. I needed more information than was provided
3. The information was biased in favour of the scheme
4. The information was biased in opposition to the scheme
5. I found questions 36 to 41 confusing
6. I did not read the enclosed pamphlet in detail
7. I found questions 1-6 in part D meaningful
Strongly Agree (AS), Agree (AG), Neither agree or disagree (NA),
Disagree (DI) and Strongly Disagree (DS)
Part E: Farmers attitudes towards different components of agricultural
biodiversity
Enumerator: Now we are going to ask about your attitudes towards conservation of
agricultural biodiversity. All of the following statements relate to the agricultural
biodiversity on your farm. Please indicate the extent of which you agree or disagree
with each of statement.
1. The number of crop varieties makes the view of the landscape more beautiful
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
319
2. Traditional varieties represent our cultural heritage
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
3. Organic farming methods are better for the environment than conventional
methods
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
4. Organic food is better for me than commercial agriculture (e.g. farming using
chemical inputs) because it does not contain any chemical residues
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
5. Environmentally friendly farming practices reflects principles and values that are
important to me
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
6. Environmentally friendly farming practices help improve consumers’ perceptions
of farmers
4 3 2 1 0
Fully agree Agree Normal Disagree Strongly disagree
7. The number of crops varieties in the farm increases the crop variety diversity
4 3 2 1 0
Strongly agree Agree Normal Disagree Strongly disagree
8. Organic fertilizer increases the soil quality and productivity of the farm
4 3 2 1 0
Strongly agree Agree Normal Disagree Strongly disagree
320
9. Chemical fertilizer increases productivity but decreases soil quality
4 3 2 1 0
Strongly agree Agree Normal Disagree Strongly disagree
10. What is your general attitude towards agricultural biodiversity?
4 3 2 1 0
Very positive Positive Normal Negative Strongly negative
11. What is your attitude towards increase agricultural biodiversity?
4 3 2 1 0
Very positive Positive Normal Negative Strongly negative
Part E: General Information of Households
In this section, we seek general information about you and your household.
1. Respondent’s main occupation: 1. Farming 2. Other
2. Age of respondent: …………………
3. Gender of respondent:…………………
4. Education: Years of schooling:……… Any other education:…………………
5. Number of family members in the household:..............................
6. Number of children in the family (< 15 years):.............................
7. Number of regular income recipients in the family (public or private sector
employed). Rs ..........................................................
8. For how long have you worked on your farm? ...........................(number of years)
If yes which year?............................................
9. Do you have a business vehicle? Yes/No
If Yes what is it............................................
10. For how long are you engaged in agricultural activities?
Number of years: ………..
321
11. Could you please state the approximate percentage of your household income
spent on food consumption?..............................%
12. Are you a member of this farm organisation? Yes/ No
13. Do you think that you can easily access borrowed credit for agriculture?
Yes / No
14. For how long have you been in this village? (No of years)..................
This is the end of the interview. Thank you very much for your participation. Do
you have any general comments about this study or anything to say?
Comments:......................................................................................................................
.............................................................................................................................
Enumerators Name: ..........................……………..
Signature: ................................................................
322
Appendix I(1): A sample choice set is given to the respondent
BLOCK 01: Question 01
I prefer to cultivate Farm (A)…................... Farm (B)….................... Neither Farm .......….… (please pick one option)
Farm Characteristics
Farm (A)
Farm (B)
Neither Small-scale farm (A) nor Small-scale farm (B):
Total number of crop varieties grown on a farm
10 7
Crops are combined with livestock/poultry production
Yes No
Farm crops produced entirely using organic methods
Yes Yes
Farm has a landrace cultivation
No No
Expenditure reduction (%)
15% 10%
Estimated cost in terms of additional labour requirement ( %)
20% 10%
323
Appendix I (2): Description of the 36 choice sets of the choice experiment
Farm A
Block Q Crop Mix Org land Exp% Lcost%
1 1 3 No Yes Yes 5 30
1 2 7 No No No 15 20
1 3 15 Yes Yes No 15 10
1 4 10 Yes No Yes 10 20
1 5 10 No No Yes 5 30
1 6 10 Yes Yes No 5 20
Farm B
Crop Mix Org land Exp% Lcost%
15 Yes No No 5 30
7 Yes No Yes 10 30
3 No No No 10 10
15 No Yes Yes 10 20
10 Yes No Yes 5 20
15 No No Yes 10 30
2 1 3 Yes No Yes 15 30
2 2 10 Yes No No 10 20
2 3 15 No No Yes 10 30
2 4 7 No No No 15 20
2 5 15 Yes No Yes 10 10
2 6 7 Yes Yes Yes 5 10
3 1 3 Yes Yes Yes 10 20
3 2 15 No Yes No 15 30
3 3 10 No No Yes 15 20
3 4 7 No Yes Yes 10 30
3 5 3 No No No 10 10
3 6 3 Yes Yes Yes 15 10
4 1 15 Yes Yes No 10 10
4 2 3 Yes Yes No 10 30
4 3 7 No No No 5 10
4 4 15 No Yes Yes 10 20
4 5 7 Yes No Yes 5 20
4 6 10 Yes No Yes 5 20
5 1 15 No Yes Yes 5 10
5 2 7 Yes Yes No 5 20
5 3 3 Yes Yes No 15 30
5 4 3 No No No 10 10
5 5 10 No Yes Yes 15 10
5 6 15 Yes No No 5 30
6 1 3 No No No 5 10
6 2 10 Yes No Yes 15 10
6 3 10 No Yes No 5 30
6 4 15 Yes No No 15 30
6 5 7 Yes No Yes 10 30
6 6 7 No Yes Yes 15 20
3 Yes Yes No 10 30
7 No No No 15 20
15 No Yes No 10 30
3 Yes Yes Yes 10 20
7 No No No 5 10
15 Yes Yes No 10 10
3 No No No 5 10
3 No No No 10 10
10 Yes No Yes 15 10
10 No Yes Yes 10 15
3 Yes Yes No 15 30
7 Yes Yes No 5 20
15 No Yes Yes 5 10
7 No Yes Yes 10 30
3 Yes Yes Yes 15 10
7 Yes Yes Yes 5 10
15 Yes No Yes 10 10
3 No Yes Yes 5 30
3 Yes No Yes 15 30
10 Yes No No 10 20
15 Yes Yes No 15 10
7 No No No 15 20
7 No Yes Yes 15 20
7 Yes No Yes 5 20
15 Yes No No 15 30
10 No Yes No 15 10
10 Yes Yes No 5 20
10 No Yes No 10 20
10 No No Yes 5 30
10 No No Yes 15 20
324
Appendix J: Descriptive statistics of the sample respondents
Table J (1): Descriptive statistics (Ampara district)
Variables Average Maximum Minimum SD
Value (Rs.) 22,844 59,800 1,500 15,382
Acres 0.91 2.50 0.01 0.89
Labour days 31.15 142 2 26.33
Other expenditure (Rs.) 6,008 40,000 124.5 7,134
Capital(Rs.) 2,814 12,000 250 1,989
Age 41.71 64 20 11.51
Education 8.00 16 0 3.21
HH Size 3.73 6 2 1.34
No. Plots 1.76 6 1 1.18
Agri. Extention 0.54 1 0 0.50
Credit 0.46 1 0 0.50
MFO 0.65 1 0 0.48
Land ownership 0.86 1 0 0.35
Crop diversity 3.08 6 0 1.75
Animal diversity 1.01 3 0 1.25
Landrace cultivation 0.33 1 0 0.42
Organic 0.36 1 0 0.37
Type of farm 0.25 1 0 0.43
Note: More details about the descriptive statistics are discussed under each
chapter
325
Table J (2): Descriptive statistics (Anuradhapura district)
Variables Average Maximum Minimum SD
Value (Rs.) 24,109 56,563 3,200 13,525
Acres 0.91 2.50 0.01 0.79
Labour days 29.89 73 3 15.94
Other expenditure (Rs.) 5,836 15,000 250 3,654
Capital (Rs.) 5,129 20,035 150 3,935
Age 40.34 62 17 11.80
Education 9.00 16 0 3.98
HH Size 3.75 6 2 1.28
No. Plots 1.95 5 1 1.04
Agri. Extension 0.36 1 0 0.48
Credit 0.51 1 0 0.50
MFO 0.55 1 0 0.50
Land ownership 0.63 1 0 0.48
Crop diversity 3.14 5 0 1.30
Animal diversity 1.58 5 0 1.99
Landrace cultivation 0.37 1 0 0.38
Organic 0.36 1 0 0.37
Type of farm 0.12 1 0 0.35
Note: more details about the descriptive statistics are discussed under each
chapter
326
Table J (3): Descriptive statistics (Kurunegala district)
Variables Average Maximum Minimum SD
Value (Rs.) 29,074 59,625 5,300 13,254
Acres 1.51 2.50 0.15 0.67
Labour days 28.51 76 2 17.98
Other expenditure (Rs.) 7,360 26,000 230 5,099
Capital (Rs.) 3,401 9,800 197 2,155
Age 41.22 61 16 12.29
Education 8.00 16 0 4.15
HH Size 3.61 7 1 1.28
No. Plots 2.55 5 1 1.35
Agri. Extention 0.53 1 0 0.50
Credit 0.56 1 0 0.50
MFO 0.59 1 0 0.49
Land ownership 0.73 1 0 0.45
Crop diversity 2.92 6 0 1.80
Animal diversity 0.77 4 0 1.21
Landrace cultivation 0.34 1 0 0.35
Organic 0.36 1 0 0.40
Type of farm 0.23 1 0 0.44
Note: more details about the descriptive statistics are discussed under each
chapter
327
Appendix K: Zero inflated Poisson / negative binomial regression model
The Poisson regression model is the most basic model that explicitly takes into
account the nonnegative integer-valued aspect of the dependent count variable.
However, it has been criticized for its restrictive property that the conditional
variance equals the conditional mean. Real-life data are often characterized by
overdispersion. The negative binomial regression model is a generalization of the
Poisson regression model that allows for overdispersion by introducing an
unobserved heterogeneity term for each observation. However, in real-life data
frequently display overdispersion and excess zeros (Lambert 1992; Greene 1994).
Zero-inflated count models provide a way of modeling the excess zeros in addition to
allowing for overdispersion. In this context, there are two models namely zero
inflated Poisson regression and zero inflated negative binomial regression model.
The Poission and negative binomial probability functions, and their respective log-
likelihood functions, need to be amended to exclude zeros and at the same time
provide for all probabilities in the distribution to sum to 1. This can be done using the
following ways. Poisson regression model with a log link function:
!y
e)y(P
i
y
i
i
ii
with )exp( ii x where xi is a covariate vector and β is a vector of unknown
coefficients to be estimated. When there is a population heterogeneity or
overdispersion, a gamma mixture of Poisson variables is often assumed. This will
lead to the negative binomial modes as given in the following equation.
328
iy
iii
i
y
yuxyf
1
11
1
1
)/1()1(
)/1(),;(
/1
where α is an ancillary parameter indicating the degree of overdispersion. The model
converges to a Poisson model if α is close to 0. To account for an extra amount of
zeros, the zero-inflated Poisson (ZIP) model assumes that
)exp()1( iii for yi = 0
P (yi/xi) =
i
y
iii
y
i
)exp()1(
for yi = 1
where is the probability of being an extra zero. Thus the subjects with y = 0 is
recognized as consisting of two groups, one not subject to the Poisson process and
the other belonging to a Poisson distribution with mean but taking on the value of
zero. The zero-inflated negative binomial regression model can be constructed
similarly. However, zero-inflated negative binomial regression model can be Logit or
Probit models. The log-likelihood functions of the ZINB logit and ZINB probit
models are given below. In this case β1 signifies the binary component linear
predictor while β signifies the count component.
The log-likelihood functions of the ZINB logit can be given as follows:
/1
111 )'exp(1
1
)'exp(1
1
)'exp(1
1ln:)0(
iii
n
i xxxyif
11ln)1(ln
1ln
)'exp(1
1ln:)0(
11
ii
i
n
i
yyx
yif
329
)'exp(1
11ln
)'exp(1
1ln
i
i
i xy
x
The log-likelihood functions of the ZINB logit can be given as follows:
/1
11
1 )'exp(1
1))'(1()'ln(:)0(
i
ii
n
i xxxyif
lyyxyif iii
n
i
11ln)1(ln
1ln))'(1ln(:)0( 1
1
)'exp(1
11ln
)'exp(1
1
i
i
i xy
xn
where exp(xβ1) is the µ from the binary process, and exp(xβ) is the same with respect
to the count process. Φ represents the normal or Gaussian cumulative distribution
function. Inflation refers to the binary process. The binary process typically has
different predictors than in the count process. The important point is for the
statistician to use the model to determine which variables in the data have direct
bearing on zero counts.
Although the ZIP distribution has received considerable attention in the literature, it
remains rather inflexible, in the sense that the nonzero counts are assumed to follow
a zero-truncated Poisson distribution. In practice, count data are often over-dispersed
so that alternative distributions such as the zero-inflated negative binomial (ZINB)
may be more appropriate than the ZIP. Furthermore, it has been established that the
ZIP parameter estimates can be severely biased if the nonzero counts are
overdispersed in relation to the Poisson distribution. This is especially the case for
correlated count data, where the observations are either clustered or represent
repeated outcomes from individual subjects.
330
APPENDIX L: Maximum likelihood estimates (MLE) of parameters and point
estimates of the technical efficiency of each producer
The truncated normal distribution is first introduced by Stevenson and the
generalised version of half-normal model by allowing U to follow a truncated normal
distribution. The log-likelihood function of this model can be found in Kumbhakar
and Lovell (2000). One of the main advantages of using this type of distribution is
that it allows for a wider range of distributional shapes. However, this sort of
flexibility comes at the cost of computational complexity as there are more
parameters to estimates (Coelli et al. 2005). It contains an additional parameter μ to
be estimated (its mode) and provides a somewhat more flexible representation of the
pattern of efficiency in the data. It has the following distributional assumptions.
i. iV ~ iid ),0( 2vN
ii. iU ~ iid ),( 2uN
iii. Ui and Vi are distributed independently of each other, and of the regressors.
The truncated normal density function for Ui is given by:
,2
exp/2
1)(
2
2
uuu
u
UUf
0iU (L.1)
where Φ(.) is the standard normal cumulative distribution and μ is the mode of the
normal distribution, which is truncated below at zero. Thus fu(U) is the density
function of a normally distributed random variable with mean μ truncated below at
zero. The density function of the random variable Vi is given by:
331
,2
exp2
1)(
2
2
vv
v
VVf
,V (L.2)
It is clear that the truncated normal distribution is a two-parameter distribution
depending on placement and spread parameters μ and u . Given V and U are
assumed to be distributed independently, their joint distribution (product of their
density functions) is given as:
,
22exp
/2
1),(
2
2
2
2
vuuvu
uv
VUVUf
0iU and ,iV (L.3)
Substituting the composite error term UV into above equation, the joint
distribution of ε and U can be expressed as follows:
,22
exp/2
1),(
2
2
2
2
vuuvu
UUUg
(L.4)
The marginal density function of ε is given by the integration of Equation L.4:
dUUgf
0
,
22
2
2/122 2exp
/2
1)(
vuuuv
Uf
where vu /
1
1)(
u
f
(L.5)
where f is asymmetrically distributed, with mean E(ε) and variance V(ε):
332
2
2
1exp
22)()(
u
uUEE
and 222
221
2)( vuV
where 1/
u . Truncated normal distribution has three parameters, a
placement parameter μ and two spread parameters u and v . Using this
information we can express the density function of Yi as:
22
2
2*
*2/122 2
),(exp
///2
1)(
vu
iii
iuiuv
iy
ZXfY
ZYf
(L.6)
where by 2222* /),( uviiuivi XfYZ and 22222* / uvuv
Let us define: 222uv and
22 u . Note that γ є(0,1); if 0 then
either 02 u or 2v which results if the symmetric disturbance term Vi
dominates the truncated efficiency component Ui which in turn indicates that the
idiosyncratic error component dominates the inefficiency effects. In that situation
OLS estimation techniques are more appropriate than stochastic frontier analysis. if
1 then either 2u or 02 v which results if the variation in the
inefficiency component explains the entire variation in εi and that indicates that
stochastic production frontier is the appropriate procedure.
Given the above reparameterizations as well as the cross sectional data for a sample
of N producers, the log likelihood function can be expressed as follows:
),,,/ln(ln 2 YL
333
ii
N
i
N
iii
XYXYNYL
'(ln))'(
2
1ln2ln
2
1)/(
11
2
2
2
2/12 )1(ln
N
(L.7)
where iii XY ' with β and Xi being [1×k] vectors. The log likelihood function
can be maximised with respect to the parameters to obtain maximum likelihood
estimates of the parameters. Accordingly the partial derivatives of the above equation
with respect to the parameters in the reparameterised set Ω are:
i
N
iiii
N
i
XXXYL
(.)
(.))'(
1ln
2
2
112
iii
N
i
XXY
(.)
(.))'(
1
2
22
1
(L.8)
(.)
(.))1()'(
(.)
(.)1)'(
1
2
1ln
1
1
2/12
2
2
1
2
1222
NXYXYN
Lii
N
i
ii
N
i
(.)
(.))1()'(
(.)
(.)1ln
1
13
2/12
22
2
1
NXY
Lii
N
i
(L.10)
and
(.)
(.))1(
(.)
(.)1)'(
1ln
1
12/12
2
2
11
NXY
L N
iii
N
i
(L.11)
where (.)1 and (.)1 are the standard normal density and distribution functions
respectively, evaluated at ])1)(/([ 2/12 and (.)2 and (.)2 are the standard
normal density and distribution functions evaluated at ]))'(/([ 1 ii XY .
These first order derivatives are used to derive the MLE estimates of β, σ2, λ and μ.
(L.9)
334
In order to predict technical efficiency, we clearly need to have some information
about the Ui. The conditional distribution )/( Uf is given by:
)(
),()/(
f
UfUf
21
2
11 2
)(exp
)]/(1[2
1)/(
UUf (L.12)
)/( Uf is distributed as ),( 21iN
, where 222 /)( viui and
22221 /)( vu . Thus either the mean or the mode of )/( Uf can be used to
estimate the technical efficiency of each producer:
)]/(1[
)/()/(
1
1
11
i
iiiiUE and )/( iiUM
The point estimates of the technical efficiency of each producer can be obtained by
substituting either )/( iiUE or )/( iiUM into following Equation L.13:
iii UETE /)exp(
2
11
11
2
1exp
)]/(1[
)/(1
i
i
iiTE (L.13)
A complete detail of derivatives for all conditions can be obtained from Kumbhakar
and Lovell (2000) and Coelli et al. (2005).
θi if θi ≥0,
0 otherwise
335
Appendix M: Derivatives of elasticities using translog production function
2
7
2
6
2
543210 )()()( LnKLnLBLnLALnRLnKLnLBLnLALnY
LnKLnLBLnRLnLALnKLnLALnLBLnLALnR ****)( 1211109
2
8
LnRLnKLnRLnLB ** 1413 (M.1)
The output elasticity of capital and labor is determined by taking the partial
derivative of the production function (M.1) with respect to each of the inputs. The
output elasticities are given by: The elasticity can be derived as follows:
LnRLnKLnLBLnLALnLA
LnYLA 1110951 2
(M.2)
LnRLnKLnLALnLBLnLB
LnYLB 1312962 2
(M.3)
LnRLnLBLnLALnKLnRLnK
LnYK 141210743 2
(M.4)
LnKLnLBLnLALnRLnR
LnYR 14131184 2
(M.5)
Marginal productivity in each factor can be estimated as follows:
LA
Y
LA
Y
LnLA
LnYp LALA
LB
Y
LB
Y
LnLB
LnYp LBLB
K
Y
K
Y
LnK
LnYp KK
R
Y
R
Y
LnR
LnYp RR
336
Appendix N(1): List of crops varieties on small-scale farms
List of crop varieties
No. Sinhala name English name Traditional
varieties
Modern
varieties
1 Sahal Rice* 2 Artapale ala Potato* 3 Batala Sweet potatao* 4 Loku lunu Big-Onion* - 5 Rathu lunu Red-Onion* - 6 Miris Chilli* - 7 Sudu lunu Garlic* - 8 Gotukola Gotukola* 9 Innala Innala* 10 Talanabatu Talanabatu -
11 Thibbatu Thibbatu -
12 Pathola Snake gourd* - 13 Beetroot Beetroot* - 14 Mea Long beans* 15 Kola boonchi Green beans* - 16 Butter boonchi Butter beans* -
17 Karavila Bitter gourd* 18 Pipinna Cucumber* 19 Batu Brinjal/Eggplants* - 20 Gova Cabbage* - 21 Malu miris Capsicum* - 22 Kekiri Kekiri* - 23 Alukesel Ash plantains* 24 Alupuhul Ash pumpkin* 25 Kola elavalu Leafy vegetables 26 Rajala Rajala* -
27 Dambala Wing bean* 28 Knolkhol Knol-khol* - 29 Takkali Tomato* - 30 Vetakolu Luffa* - 31 Rabu Raddish* -
32 Bandakka Okra* - 33 Wattakka Pumpkin* 34 Kiriala Kiriala* 35 Kohila ala Kohila yams* 36 Mannokka Manioc* 37 Alakola Dioscorea* -
38 Kankun Kankun* 39 Alupuhul Ash pumpkin*
337
No. Sinhala name English name Traditional
varites
Modern
varieties
40 Murunga Drumstick 41 Mal gova Cauli flower* - 42 Rumpa Rumpa* -
43 Tora kola Tora kola -
44 Mugunuvenne Mugunuvenne* -
45 Nivithi Nivithi* -
46 Sarana Sarana -
47 Gova kola Gova kola* - 48 Kaha Turmeric* -
49 Enguru Ginger* 50 Meneri Meneri* 51 Kollu Horse grain* 52 Kurakkan Finger millet* 53 Tala Gingelly* 54 Undu Black gram* 55 Mun ata Green gram/Mungbean* 56 Parippu Dhal* 57 Kadala Chickpea* 58 Cowpea Cowpea*
59 Eringu Maize* 60 Bulath Betel* 61 Sunflower Sunflower* 62 Idal iringu Sorghum* 63 Rata kaju Ground nuts* 64 Soya boonchi Soybean* - 65 Kadala parippu Pigeonpea
Note: We only included seasonal crops in this analysis. This implies that any variety that takes more
than 6 months to harvest is excluded from the survey
* cash crops
Appendix N(2): List of livestock breeds on small-scale farms
No. Sinhala name English name Traditional
varieties
Modern
varieties
1 Elaharak Neat cattle 2 Meharak Buffalo 3 Eluva Goat 4 Ura Pig 5 Kukula Poultry
Source: Household survey data (Sep-Oct 2010), Sri Lanka