impact of coffee certification on small holder coffee ... · to be so, by means, most commonly, of...
TRANSCRIPT
IMPACT OF COFFEE CERTIFICATION ON SMALL
HOLDER COFFEE FARMING IN EMBU COUNTY, KENYA
LUCY W. MURIITHI
A103/200072/2010
A Thesis submitted in Partial fulfillment of the requirements for
the Degree of Masters of Science in Agribusiness Management
and Trade in the School of Agriculture of
Kenyatta University
June 2016
ii
DECLARATION
I Lucy Wanjiku Muriithi declare that this thesis is my original work and has not
been presented for a degree in any other University.
Signature …………………………………… Date ………………………………
Lucy Wanjiku Muriithi (A103/20072/2010)
Department of Agribusiness Management and Trade
SUPERVISORS
We confirm that the work reported in this thesis was carried out by the candidate
under our supervision.
Signature ………………………………… Date…………………………………
Dr. Ibrahim Macharia
Department of Agribusiness Management and Trade
Kenyatta University
Signature…………………………………. Date…………………………………
Dr. Elijah Gichuru
Coffee Research Institute
Kenya Agricultural and Livestock Research Organization
iii
DEDICATION
I dedicate this work to my loving husband Mr. Newton Mwaniki, my children
Jermaine Kiama and Abigail Nkatha and to my loving parents.
iv
ACKNOWLEDGMENT
I express my sincere gratitude to Coffee Research Institute through the Coffee
Leaf Rust Project (CFC/ICO/40) for financing my studies, sincere gratitude go to
my supervisors Dr.Ibrahim Macharia and Dr. Elijah Gichuru for their guidance
and support during the course of the study. I am also grateful to my colleagues
and Coffee Research Institute staff for their support. Special thanks to Mr.
Kennedy Gitonga and the staff of Economics and Research Liaison Departments
for their dedication in data collection and analysis.
I thank my husband Mr. Newton Mwaniki for his support and inspiration during
this study, special thanks to my children Jermaine Kiama and Abigail Nkantha for
their patience and love.
Ultimately, special thanks to my parents, my brothers and sisters for the moral
support. I thank the Almighty God for the gift of life, sound mind and also for
granting me the opportunity.
v
TABLE OF CONTENTS
TABLE OF CONTENTS
TITLE PAGE…………………………………………………………………….i
DECLARATION................................................................................................... ii
DEDICATION...................................................................................................... iii
ACKNOWLEDGMENT ..................................................................................... iv
TABLE OF CONTENTS ..................................................................................... v
LIST OF TABLES ............................................................................................... ix
LIST OF FIGURES .............................................................................................. x
OPERATIONAL DEFINITIONS OF KEY CONCEPTS AND TERMS ...... xi
ABBREVIATIONS AND ACRONYMS ........................................................... xii
ABSTRACT ........................................................................................................ xiv
CHAPTER ONE: INTRODUCTION ................................................................. 1
1.0 BACKGROUND INFORMATION .................................................................. 1
1.1. CERTIFICATION............................................................................................ 3
1.2 STATEMENT OF THE PROBLEM ................................................................. 4
1.3 RESEARCH OBJECTIVES ............................................................................. 6
1.3.1 MAIN OBJECTIVE .................................................................................... 6
1.3.2 SPECIFIC OBJECTIVES ............................................................................ 6
1.4 RESEARCH HYPOTHESES ........................................................................... 6
1.5 SIGNIFICANCE OF THE STUDY .................................................................. 7
1.6 LIMITATIONS AND ASSUMPTIONS OF THE STUDY .............................. 7
1.7 CONCEPTUAL FRAMEWORK ..................................................................... 8
CHAPTER TWO: LITERATURE REVIEW .................................................. 10
vi
2.0 INTRODUCTION........................................................................................... 10
2.1 APPROACHES TO IMPACT MEASUREMENT ......................................... 10
2.1.1 EXPERIMENTAL DESIGNS ..................................................................... 11
2.1.2 QUASI-EXPERIMENTAL DESIGNS ....................................................... 11
2.1.2.1 BEFORE AND AFTER APPRAISAL ..................................................... 12
2.1.2.2 WITH AND WITHOUT APPRAISAL .................................................... 12
2.1.2.3 DIFFERENCE IN DIFFERENCE ............................................................ 12
2.1.3 ESTIMATING PROPENSITY SCORE USING BINARY RESPONSE
LOGIT MODEL ................................................................................................... 14
2.2 CERTIFICATION STANDARDS .................................................................. 15
CHAPTER THREE: RESEARCH METHODOLOGY ................................. 31
3.0 INTRODUCTION........................................................................................... 31
3.1 STUDY AREA ................................................................................................ 31
3.2 STUDY DESIGN ............................................................................................ 33
3.3 SAMPLE AND SAMPLING DESIGN ........................................................... 33
3.4 RESEARCH INSTRUMENTS ....................................................................... 34
3.5 DATA ANALYSIS ......................................................................................... 35
3.5.1 FACTORS THAT INFLUENCE SMALL HOLDER COFFEE FARMER‟S
DECISION TO PARTICIPATE IN CERTIFICATION ...................................... 35
3.5.2 PROPENSITY SCORE MATCHING METHOD ....................................... 38
3.5.3 IMPACT OF CERTIFICATION ON COFFEE PRODUCTIVITY ............ 40
3.5.4 IMPACT OF CERTIFICATION ON COFFEE PRICE .............................. 41
CHAPTER FOUR: RESEARCH FINDINGS AND DISCUSSIONS ............ 42
4.1 INTRODUCTION........................................................................................... 42
4.2 DESCRIPTIVE STATISTICS ........................................................................ 42
4.3 FACTORS THAT INFLUENCE FARMERS‟ DECISION TO PARTICIPATE
IN COFFEE CERTIFICATION ............................................................................ 46
4.3.1 LOGISTIC REGRESSION .......................................................................... 46
vii
4.3.1.1. REGRESSION DIAGNOSTICS ............................................................. 46
4.3.1.1.1. NORMALITY....................................................................................... 46
4.3.1.1.2. MULTICOLLINEARITY..................................................................... 47
4.3.1.1.3. HETEROSCEDASTICITY .................................................................. 48
4.3.2 ODDS RATIO RESULTS ........................................................................... 50
4.4 PROPENSITY SCORES AND COVARIATES ............................................. 52
4.5 ASSESSING THE IMPACT OF CERTIFICATION ON FARM LEVEL
COFFEE PRODUCTIVITY ................................................................................. 55
4.5.1 QUANTITY OF COFFEE PRODUCED .................................................... 55
4.5.2 DISTANCE, FARM SIZE, ACQUISITION AND NUMBER OF COFFEE
TREES .................................................................................................................. 56
4.6 THE IMPACT OF CERTIFICATION ON COFFEE PRICES........................ 57
4.6.1 IMPACT PROPENSITY ESTIMATE ON COFFEE PRICES ................... 57
4.6.2 IMPACT PROPENSITY ESTIMATE ON INCOME FROM COFFEE AND
OTHER CROPS.................................................................................................... 58
4.6.3 IMPACT PROPENSITY SCORE ON INCOME FROM OTHER CROPS 58
4.6.4 IMPACT PROPENSITY ESTIMATE ON INCOME FROM COFFEE ..... 59
4.7 SENSITIVITY ANALYSIS ............................................................................ 60
CHAPTER FIVE: DISCUSSION ...................................................................... 62
5.0 INTRODUCTION........................................................................................... 62
5.1 FACTORS THAT INFLUENCE SMALL HOLDER COFFEE FARMER‟S
DECISION TO PARTICIPATE IN CERTIFICATION ....................................... 62
5.2 IMPACT OF CERTIFICATION ON COFFEE PRODUCTIVITY ................ 63
5.3 IMPACT OF CERTIFICATION ON COFFEE PRICES ................................ 64
CHAPTER SIX: SUMMARY, CONCLUSION AND
RECOMMENDATIONS .................................................................................... 66
6.0 INTRODUCTION........................................................................................... 66
6.1 CONCLUSION ........................................................................................... 66
viii
6.2 RECOMMENDATIONS ............................................................................ 68
REFERENCES .................................................................................................... 69
APPENDICES ..................................................................................................... 75
APPENDIX 1: HISTOGRAM OF PROPENSITY SCORES ........................... 75
APPENDIX 2: PERFORMANCE OF DIFFERENT MATCHING
ALGORITHMS (LIKELIHOOD RATIO TEST) ............................................. 76
APPENDIX 3: VARIABLES ............................................................................ 78
APPENDIX 4: EMBU COUNTY MAP ............................................................ 81
APPENDIX 5: SMALL SCALE COFFEE FARMERS QUESTIONNAIRE ... 82
APPENDIX 6: COOPERATIVE SOCIETY QUESTIONNAIRE .................... 86
ix
LIST OF TABLES
Table 1.1: Coffee (cherry) production and prices before and after certification .....4
Table 2.1: Literature review……………………………………………………...20
Table 4.1: Descriptive statistics of coffee farming households in Embu North
district Sub County ............................................................................................... 43
Table 4.2: Test for multicollinearity ..................................................................... 47
Table 4.3: Logistic regression results for coffee farmers in Embu County .......... 50
Table 4.4: Odds ratios results ................................................................................ 52
Table 4.5: Performance of different matching estimators .................................... 53
Table 4.6: Chi-square test for significance ........................................................... 54
Table 4.7: ATT for coffee production in kilograms for farmers in Embu County
between 2006 and 2007 ........................................................................................ 55
Table 4.8: Distance, farm size, acquisition and number of coffee trees for coffee
farmers in Embu County ....................................................................................... 56
Table 4.9: ATT for the prices of coffee in Kenya shillings for farmers in Embu
County between 2006 and 2007 ............................................................................ 57
Table 4.10: ATT for combined income from coffee and other crops for farmers in
Embu County between 2006 and 2007 ................................................................. 58
Table 4.11: ATT for income from other crops in Kenya shillings for coffee
farmers in Embu County between 2006 and 2007 ................................................ 59
Table 4.12: ATT for income from coffee in Kenya shillings for farmers in Embu
County between 2006 and 2007 ............................................................................ 60
Table 4.13: Sensitivity analysis using the Rosenbaum bounding approach to check
for selection bias ................................................................................................... 61
x
LIST OF FIGURES
Figure 1.1: Conceptual framework in the study…………………………………..9
Figure 3.1: Study area ........................................................................................... 32
Figure 4.1: Histogram of propensity scores .......................................................... 75
xi
OPERATIONAL DEFINITIONS OF KEY CONCEPTS AND TERMS
Certification - Certification is the process through which an organization grants
recognition to an individual, firm, process, service, or product that meets certain
established criteria. Certification also means that a state of affairs has been stated
to be so, by means, most commonly, of a document self-described as a certificate.
Certification and verification standards- Are used as a mean of communicating
information about the quality, traceability, social, environmental and financial
conditions surrounding the production of goods or provision of services.
Certified farmers- These are farmers who have undertaken certification
standards through their cooperatives societies.
Farm household- Defined as a social entity that collectively makes productive
and consumptive decisions and often eats from the same granary.
Small scale coffee farmers/small holder farmers- They are defined as farmers
with small land unit under coffee between 0-5 acres and they deliver coffee as
individuals but they pulp, mill and market coffee collectively in a cooperative.
Wet Mill/ Coffee Factory-This is a unit where primary processing of coffee is
undertaken.
xii
ABBREVIATIONS AND ACRONYMS
ATT Average treatment effect on certified group
CAFÉ Coffee and Farmers equity
CBK Coffee Board of Kenya currently Coffee Directorate
CIDIN Centre for International Development Issues Nijmegen
COSA Committee on Sustainability Assessment
CRI Coffee Research Institute
FAO Food and Agriculture Organization of the United Nations
FCS Farmers‟ Cooperative Society
FLO Fair Labeling Organization
FME Free Market Environmentalism
FT Fair Trade
GAP Good Agricultural Practices
GDP Gross Domestic Product
GPP Good Processing Practices
HH House Hold
ICA International Coffee Agreement
ICO International Coffee Organization
IFOAM International Federation of Organic Agriculture Movements
IISD International Institute for Sustainable Development
ISEAL International Social & Environmental Accreditation and
Labeling Alliance
ITC International Trade Center
xiii
Kgs Kilograms
Ksh Kenya shillings
MoA Ministry of Agriculture, Livestock and Fisheries
NGOs Non-Governmental organizations
ORCA Organic and Resource Conserving Agriculture
Sq. Km Square Kilometers
USD United State Dollars
WTO World Trade Organization
VIF Variance Inflation Factor
xiv
ABSTRACT
The coffee sector plays a significant role to Kenya‟s economy. The sector
contributes to foreign exchange earnings, household incomes, employment and
food security. The industry supports about 700,000 households comprising of 535
coffee cooperative societies for the development and marketing of coffee. A new
concept in marketing all over the world has been the adoption of certification
standards which are becoming increasingly popular in Kenya; currently there are
about four certification standards. Whereas adhering to them is beneficial only
few coffee societies have been certified. Thus the key objective of this study was
to assess the impact of coffee certification on small holder coffee farming in
Embu County in Kenya. The specific objectives were to analyze factors that
influence small holder coffee farmers‟ decision to participate in certification in
Embu County, to assess the impact of certification on farm level productivity and
to evaluate the impact of certification on coffee prices. To achieve these
objectives, data was collected from 238 certified coffee farms (Households) and
242 non certified coffee farms (Households) using a questionnaire. Logit model
was used to establish the factors that influenced farmers to participate in
certification. Propensity score matching method was used to solve the selection
bias. Results showed that factors such as price and income from coffee, gender of
the household head and farmers‟ perception positively influenced participation in
certification. On the impact of certification on coffee productivity the study
showed that the certified group produced more coffee in some years studied than
non-certified group. However the study did not find any impact of the
certification program on coffee prices. Certified farmers received higher coffee
prices in 2007/2008, but gain disappeared until the year 2010/2011 where the
gains were significantly high. The study concluded that the results were consistent
with the hypothesized relationship that there were factors that affected farmers‟
decision to participate in certification and there were mixed results in hypothesis
two and three. The study recommended that farmers need to be empowered with
information on certification, future studies need to consider impact of certification
on other farm enterprises in addition to coffee and further research on social,
environmental and socio-economic impact assessment needs to be done using
emerging business evaluation models and social return to investment.
1
CHAPTER ONE: INTRODUCTION
1.0 Background Information
Coffee is one of the most important commodities in the world today. It is the
second most traded commodity after petroleum and a vital source of export
earnings for any of the developing countries that grow it (Rice & Jennifer, 1999).
Coffee remains a major cash crop and top foreign exchange earner for the Kenyan
economy and is ranked 5th
contributor to GDP after horticulture, tourism, tea, and
diaspora remittance. The industry contributes about 1% national GDP, about 8%
of the total agricultural export earnings and up to 25% of the total labor force
employed in agriculture (Affa, 2013).
Kenya mainly produces highly valued Arabica coffee based on varieties (SL28,
SL 34, K7, Ruiru 11 and Batian). It is estimated that area under coffee is
approximately 113,500 ha out of which 80% is occupied by small holder
(cooperatives) and the industry supports about 800,000 households comprising of
535 Coffee Cooperative Societies and 4,000 Coffee Estates. Due to its effective
forward and backward linkages the industry supports about 5 million people
subsequently contributing to food security, employment and generally household
welfare within the coffee growing areas thus providing a boost to the Kenya‟s
overall economy through multiple effects (Affa, 2013).
2
Kenya‟s coffee cooperative system was formed after the end of World War II and
is regulated by the government under the Cooperatives Act. This act requires
small holders with less than five acres of coffee to come together and form coffee
cooperative societies where they pulp mill and sell their coffee collectively. The
societies vary greatly in size, where merging and splitting are common. Some
cooperatives have only one wet mill whilst others have more. Factories typically
provide services to 300 to 800 members of a society (CIDIN, 2014).
Coffee consumers on the other hand are mostly found in the developed
economies, with the Nordic countries showing some of the highest per capita
consumption. Sweden, for example, is among the world‟s top coffee consuming
nations, with an annual per capita consumption of around 9kg of coffee beans
equivalent, or about 3.4 cups of coffee per person per day (ICO, 2015).
Coffee markets link producers and consumers in developed and developing
countries and is an important vehicle for sustainable development. Recognizing
this, consumers, NGOs, the private sector and donor agencies have taken a
growing interest in promoting sustainable production and trading practices along
coffee supply chains. These have taken the form of sustainability labels (such as
Fair trade certification), business-to-business standards and other certification
initiatives, and Voluntary Sustainability Initiatives that provide assurances to
consumers and other supply-chain stakeholders that production is in accordance
with sustainable development objectives (Potts & Sanctuary,2010).
3
1.1. Certification
Increased awareness among coffee consumers of the impact of their consumption
habits on the people and environment in coffee producing countries has resulted
in implementation of certification programs in the coffee sector as an assurance of
good practices in production and marketing of coffee (Mercy et al., 2010).
Sustainable certification initiative creates incentives for farms and firms to
improve their environmental and socio-economic performance (Giovannucci &
Ponte, 2005). Certification enables the consumer to differentiate among goods and
services based on their environmental and social attributes. This improved
information facilitates price premiums for certified products, and these premiums,
in turn, create financial incentives for farms and firms to meet certification
standards.
An example of societies that have undergone certification is Ndumberi
cooperative society and the performance in terms of production and prices before
and after certification is shown in Table 1.1
4
Table 1.1: Coffee (cherry) production and prices before and after
Certification
Before certification After certification
Year Production/
Kg
Price/
Kg
Year Production/
Kg
Price/ Kg
2003/04 746,888 18.20 2006/07 939,947 31.55
2004/05 405,825 21.50 2007/08 1,060,410 33.15
2005/06 489,846 26.10 2008/09 1,268,358 37.40
2009/10 1,357,392 60.15
2010/11 1,476,851 102.45
Source: (Ndumberi FCS, 2011)
However, the complexity of certification mechanisms, their reliance on
cooperative formation to make them economically viable, and the tangible
livelihood benefits (and disadvantages) remain poorly understood by the small
scale coffee farmers in Kenya (Mercy et al., 2010).
1.2 Statement of the problem
Certification of coffee in Kenya has been suggested as an important tool that
every farmer needs to embrace because of its potential benefits, such as increased
market access, increased production efficiency and sustainable production (CBK,
2010).
As these certification initiatives penetrate mainstream markets, their economic
effects are significant. Yet the question persists: Do these initiatives improve
5
livelihoods, trade, or the environment? To date, the nature and distribution of
these impacts remain largely unknown. Data on the impacts of different initiatives
exists but it has been often piecemeal or anecdotal, leaving the major questions of
overall sustainability and global effects unanswered. The absence of a more
expansive and rigorous information base leaves policy makers, consumers, supply
chain decision-makers and, worst of all, producers, increasingly challenged as
they attempt to determine when and where investment in such initiatives is
warranted (Potts & Sanctuary, 2010).
Although a fast-growing academic literature examines sustainable certification,
little is known on whether it actually affects farms‟ and firms‟ environmental and
socioeconomic performance. Relatively few studies specifically aim to evaluate
the impacts of certification, and many of those that do rely on crude methods that
do not correct for selection effects or are likely to bias results for other reasons
(Blackman & Rivera, 2010).
In Kenya the findings from (CIDIN, 2014) on impact of certification show
conflicting results on the benefits of certification. While findings from (Mercy et
al., 2010) show that certification increased coffee prices, incomes and
productivity therefore it is not clear whether certification is beneficial or not and
assessing the impact of coffee certification on small holder farming in Embu
County would be important as this will add to the knowledge base on
performance of certification programs.
6
1.3 Research objectives
1.3.1 Main Objective
The purpose of the study is to assess the impact of coffee certification on small
holder coffee farming in Embu County.
1.3.2 Specific Objectives
The specific objectives of the study were to:-
(i) To identify and analyze factors that influence small holder coffee farmers
decision to participate in certification in Embu County.
(ii) Assess the impact of certification on coffee productivity in Embu County.
(iii) Determine the impact of certification on coffee prices in Embu County.
1.4 Research Hypotheses
(i) There are factors that influence small holder coffee farmers‟ decision to
participate in certification.
(ii) Certified farmers attain increased coffee productivity than the non-
certified farmers.
(iii) Certified farmers receive higher coffee prices than the non-certified
farmers.
7
1.5 Significance of the study
Coffee certification has been a relatively new approach focusing both small
holder coffee farmers as well as large and medium estates in Kenya and the
practices are not yet fully understood by all the coffee stakeholders. It is therefore
expected that the findings of this study will provide basic information to
institutions interested in promotion of certification standards and identify areas of
coffee certification that need further research so that advantages and
disadvantages of certification are well understood and to make appropriate
recommendations to the stakeholders.
The results will also provide useful insights to coffee certification bodies and
coffee farmers in Kenya.
1.6 Limitations and assumptions of the study
This study investigated the impacts of coffee certification on small holder coffee
producers against non-certified ones in Embu County by comparing certified
cooperatives and non-certified cooperatives. The study covered all certification
standards in general, but did not analyze the individual certification standards.
The key assumptions made were that certified small holder farmers had
accumulated auditable records and would give full disclosure and coffee sector
and national politics were favorable.
8
1.7 Conceptual framework
The study is conceptualized on randomization theory where all cases are chanced
over a finite universe of possibilities (Oscar, 1955). It involves a comparison of
certified (treatment) and non-certified farmers (intervention groups), which are
alike in all important aspects except for certification, and this eliminates selection
bias, balances the group with respect to many known and unknown confounding
variables (Dehejia & Wahba, 2008).
Fig 1.2 shows the relationship between the independent and dependent variables
in the study. The independent variables were age of the coffee farmers, gender of
the coffee farmers, education level of the coffee farmers, income from coffee,
distance to the factory, awareness of the farmers and perception of the farmers
while the dependent variable was coffee certification.
9
Figure 1.2: Conceptual framework modified from Tina et al., 2009
Education level
Awareness of
farmers
Age of the farmers
Gender of coffee
farmers
Distance to the
factory
Income from coffee
Coffee productivity
Price of coffee
Farmers‟ perception
Coffee
certification
Improved income
Improved coffee
prices
Better farming
knowledge
Improved coffee
productivity in
coffee and other
farm enterprises
Independent variable
Dependent variable
Outcomes
10
CHAPTER TWO: LITERATURE REVIEW
2.0 Introduction
This chapter reviews literature on impact assessment studies on coffee
certification. A crucial review on the impact assessment methods used and
findings from these studies and identifying the research gap.
2.1 Approaches to impact measurement
If one could observe the same individual at the same point in time, with and
without the intervention, this would effectively account for any observed or
unobserved intervening factors and the problem of endogeneity do not arise
(Ravallion, 2005). Since this is not happening in practice, something similar is
done by identifying non-participating groups identical in every way to the group
that receives the intervention, except that the non-participating groups do not
receive the intervention.
To know the effect of a project on a participating individual, we must compare the
observed outcome with the outcome that would have resulted had that individual
not participated in the project. However, as stated earlier two outcomes cannot be
observed for the same individual. In other words, only the factual outcome can be
observed. Thus, the fundamental problem in any social project evaluation is the
missing data problem (Ravallion, 2005).
11
The hypothetical question in this impact evaluation exercise is “what would have
happened to a household if the household would not have participated in any
certification program?” In literature, there are several methods available to
estimate the impacts or effects of interventions or development programs.
2.1.1 Experimental Designs
Experimental designs are generally considered the most robust evaluation
methodologies. By randomly allocating the intervention among eligible
beneficiaries, the assignment process itself creates comparable treatment and
control groups that are statistically equivalent to one another, given appropriate
sample sizes (Baker, 2000).
2.1.2 Quasi-Experimental Designs
Quasi-experimental (non-random) methods can be used to carry out an evaluation
when it is not possible to construct treatment and comparison groups through
experimental design. These techniques generate comparison groups that resemble
the treatment group, at least in observed characteristics, through econometric
methodologies, which include matching methods, double difference methods,
instrumental variables methods and reflexive comparison. When these techniques
are used, the treatment and comparison groups are usually selected after the
intervention by using non-random methods (Baker, 2000).
The methodological challenge in non-experimental evaluation method, is
examining outcome response of an intervention. It involves filtering the effect of
12
intervention only from that of the factors that affect individuals (Foster, 2003).
There are different econometric approaches that have been used to avoid or
reduce this problem.
2.1.2.1 Before and after appraisal
It addresses changes in outcomes over a specified time period. An example is
where a baseline is compared with an ex-post survey.
2.1.2.2 With and without appraisal
This is where differences are estimated between the treatment and a control
group. In this approach, the situation amongst the control group is the
counterfactual to the situation attained in the target or treatment group.
2.1.2.3 Difference in difference
A combination of the “before and after” with the “with and without” approaches
gives a difference in difference estimator. It compares the change in outcome in
the treatment group before and after the intervention to the change in the
outcomes in the control group. The change in the control group is an estimate of
the true counterfactual i.e. what would have happened to the intervention group if
the intervention had not been implemented. The “difference in difference”
estimator requires data panel which is often unavailable particularly from rural
households in Africa.
13
The absence of historical data encourages studies on impact assessment that use
cross sectional data to estimate the difference or observed changes between the
treatment and control group.
2.1.2.4 Propensity score matching
Propensity score matching technique is increasingly used to deal with the problem
of unobserved differences in an evaluation. The approach solves the “selection”
problem (Heckman et al., 1998; Rosenbaum and Rubin, 2002). Matching involves
pairing treatment and comparison units that are similar in terms of their
observable characteristics. When the relevant differences between any two units
are captured in the observable (pre-treatment) covariates, which occurs when
outcomes are independent of assignment to treatment conditional on pre-treatment
covariates, matching methods can yield an unbiased estimate of the treatment
impact (Dehejia & Wahba, 2008).
The idea behind matching is to select a group of non-beneficiaries in order to
make them resemble the beneficiaries in everything, but the fact of receiving the
intervention (certification). If such resemblance is satisfactory, the outcome
observed for the matched group approximates the counterfactual, and the effect of
the intervention is estimated as the difference between the average outcomes of
the two groups (Caliendo & Kopeinig, 2008).
The method of matching has an intuitive appeal because by constructing a control
group and using difference in means, it mimics random assignment. The crucial
14
difference with respect to an experiment is that in the latter, the similarity between
the two groups covers all characteristics, both observable and unobservable, while
even the most sophisticated matching technique must rely on observable
characteristics only (Rosenbaum & Rubin, 2002). The fundamental assumption
for the validity of matching is that, when observable characteristics are balanced
between the two groups, the two groups are balanced with respect to all the
characteristics relevant for the outcome. This study utilized the propensity score
matching method.
2.1.3 Estimating propensity score using binary response Logit model
The propensity score was obtained using the Logit model to predict the
probability of participation of household (Gujarati, 2004). Logit model was also
used to estimate propensity scores using households pre-intervention
characteristics (Rosenbaum & Robin, 2002).Matching was then performed using
propensity scores of each observable characteristics, which must be unaffected by
certification. These characteristics included covariates variables that influenced
the participation decisions and the outcome of interest. The coefficients were used
to calculate a propensity score, and participants matched with non-participants
based on having similar propensity-scores.
15
2.2 Certification Standards
This section reviews available literature on certification which includes the type
of certification standards available in Kenya, their benefits and their
commonalities.
Certification has been defined as a way of communicating information about the
quality, traceability, social, environmental and financial conditions surrounding
the production of goods and services (Fair Trade International, 2009).
Increased awareness among coffee consumers of the impact of their consumption
habits on the people and environment of coffee producing countries has resulted
to development of initiatives in the coffee sector which seeks to assure consumers
of good practices in production. Such initiatives in Kenya were first introduced in
the floriculture and horticultural industries and more recently in the tea and coffee
industries. Certification standards advocate for good practices in an endeavor to
protect the consumer, the environment as well as the producer. In Kenya, there are
six main coffee certification and verification standards as set by the International
Federation of Organic Agriculture Movements (IFOAM). These standards are
based on living ecological systems and cycles such as health of soil, plant, animal,
human and planet fairness both in terms of the environment and life opportunities;
and protecting the wealth and well-being of current and future generations (Fair
Trade, 2010).
16
The UTZ certificate was the first to be introduced in the Kenyan coffee industry
and currently there are five other certification standards that are being
implemented namely: Fair Trade, Common Code for Coffee Community (4Cs),
Rain Forest Alliance, Nespresso AAA and Cafe Practices. It is expected that more
certification standards will be introduced in Kenya as competition intensifies for
the high quality coffees produced in the region (Giovannucci & Ponte, 2005).
Good inside coffee certification (UTZ) is a worldwide certification program that
sets the standard for responsible coffee production and sourcing. UTZ which
means “good” inside in Maya language gives an assurance of the social and
environmental quality in coffee production (Burns & Blowfield, 2000).
Fair Trade certification empowers small-scale farmers organized in cooperatives
to invest in their farms and communities, protect the environment, and develop
the business skills necessary to compete in the global marketplace. Fair-trade
certification aims to improve the livelihoods and well-being of small producers by
improving their market access, strengthening their organizations, paying them a
fair price, and providing continuity in trading relationships (Fair Trade, 2009).
Common Code for Coffee Community (4C Association) is an initiative from the
coffee sector for the coffee sector and addresses all that makes their living from
coffee. Members work jointly towards improving economic, social and
environmental conditions through more sustainable and transparent practices for
all who make a living in the coffee sector (Beckman, 1998).
17
Rainforest Alliance works with farmers to improve their livelihood, health and
well-being of their communities. The standard is built on the three criteria of
sustainability i.e. environmental protection, social equity and economic viability.
These criteria are designed to protect biodiversity, deliver financial benefits to
farmers, and foster a culture of respect for workers and local communities
(Giovannucci & Pierrot, 2010).
The Rainforest Alliance works directly with farmers to teach and encourage good
land-use practices. Specifically, the Alliance requires the maintenance or
restoration of a certain percentage of natural forest cover, and no impact on
natural bodies and flows of water. Certain destructive activities are prohibited.
Rainforest Alliance also promotes training, safe working conditions, sanitation
and health for farm workers (Giovannuccci, 2006).
Nespresso AAA has been working to protect coffee ecosystems by promoting
sustainable agricultural best practices in ecosystem conservation, wildlife
protection and water conservation. The Nespresso AAA Sustainable Quality
Coffee Program sets out to ensure the cultivation of highest quality coffee in ways
that are environmentally sustainable and beneficial to farming communities (Melo
& Wolf, 2005).
Coffee and Farmers Equity Practices is a voluntary supply chain program that
provides purchasing preference to coffee suppliers who supply coffee beans that
18
are grown, processed, and traded in an economically, socially, and
environmentally responsible manner (Jaffee, 2007).
Benefits of certification
The expected benefits from such certification are: strengthening of farmer
organizations in terms of good governance and increased efficiency in provision
of technical as well as commercial services; greater accessibility of farmers to
technical services, farm inputs, credit and hence higher productivity, higher
producer prices and higher enterprise and farm incomes, higher disposable
incomes, and consequently greater investments on-farm and in other
areas/activities that improve the welfare of household members (Mercy et al.,
2010).
Key commonalities in certification standards
All certification standards share common principles relating to traceability, social,
environmental and economic aspects. Traceability tracks coffee from tree to cup
and the flow of payments back to producer. It tracks source of coffee, production
conditions, trade up to consumption. Coffee should be traceable from the field
through processing and finally to the market. This requires, exhaustive record
keeping detailing all production, processing and marketing activities.
Environmental principle:- Coffee production systems should not impact
negatively on the environment. Environment includes: flora, fauna, wildlife, soil,
19
air, and water. All standards advocate need for pollution control (Giovannucci &
Ponte, 2005).
Social principle:- Workers welfare including wages, working hours, living
conditions, basic education etc. Child labor, safety at work, discrimination, gender
equality, sexual harassment and worker‟s rights. Living conditions especially the
housing, provision of clean portable water and sanitary facilities.
Quality of coffee:- Certification standards demand certain quality levels for
coffee. Quality is determined through samples and assumption that adhering to the
standards will result in an improved coffee quality (Giovannucci & Ponte, 2005).
The gaps on available literature on certification standards were reviewed in Table
2.1 below. Certification standards available in Embu County are the UTZ and
Fair-trade certification (CBK, 2010). Societies in Embu County have explored fair
trade and UTZ certification programs.
20
Authors
and Year
Study title Objectives Analytical Method Key findings Gaps
Giovannuci
& Potts
(2008).
The COSAP
project: A
multi-
criteria cost
benefit
analysis of
sustainable
practices in
coffee
To assess both
direct and
indirect costs
and benefits of
sustainability
standards in all
the 3 areas
required (Econ-
Environ-Social)
Application of analysis of
variance(ANOVA) to assess statistical
relevance
Multi-criteria analysis to provide basic
outcomes along core sustainability
criteria
Certified farms observed to
be better off than their
counter parts but the gap is
narrow, more than 60% of
all certified farms visited
perceived their
participation in a
sustainability initiative as
having a positive economic
impact on their farms
Large sample size
are needed to
extract more
statistically
significant results
Ruerd &
Guillermo
(2008).
How
standards
compete:
Comparativ
e impact of
Coffee
Certification
in Northern
Nicaragua
To assess the
comparative
performance of
voluntary and
private
standards for
the welfare of
individual small
holder families
Propensity score match and difference
analysis with nearest neighbor and
kernel techniques to identify unbiased
impact effects 315 farmers in
Nicaragua that produce coffee under
Fair trade, Rainforest and Cafe
practices or deliver for independent
traders, the effects were compared on
income, production and investment.
Fair trade provides better
prices compared to
independent producers but
private labels outcompete
fair trade in terms of yields
and quality performance
While fair-trade
can be helpful to
support initial
market
incorporation
private labels offer
more suitable
incentives for
quality upgrading
Table 2.1: Literature review analysis
21
Authors and
Year
Study title Objectives Analytical Method Key findings Gaps
Bolwig
et al., (2009).
The
economics of
small holder
organic
contract
farming in
Tropical
Africa
To examine the
revenue effects
of certified
organic
contract
farming for
small holders
and of adoption
of organic
agriculture
farming
methods in a
tropical African
context
A standard OLS
regression.
full information
maximum likelihood
estimate of Heckman
selection method
There are positive
revenue effects both
from participation in
the scheme and more
modesty from applying
organic farming
techniques. Organic
certification can boost
net coffee revenue by
75% on average.
The usefulness of further research
on the economics of organic
farming techniques in tropical
Africa. Which techniques are
most readily adopted and why/
which generate the highest returns
and why?
The other concern is a
comparison of the design features
of the plethora of new types of
small holder contract farming
schemes that are emerging in
tropical Africa in response to
increased market differentiation
in developed countries, in terms
of their incentive effects for small
holder
22
Authors and
Year
Study title Objectives Analytical Method Key findings Gaps
Blackman &
Naranjo
(2010).
Does Eco-
certification
Have
Environmental
Benefit
Evaluation of
the
environmental
impacts of
organic coffee
certification in
Central
Certification
Propensity Sore
Matching
Organic certification
improves coffee grower‟s
environmental performance,
significantly reduces
chemical input use and
increases adoption of some
environmentally friendly
management practices
Certification standards are
likely to entail significant costs
for producers. Absent high price
premiums or other benefits
from certification and these
costs will discourage
certification and this is reflected
in the small number of certified
organic producers in the sample
Blackman &
Rivera
(2010).
The evidence
Base for
Environmental
and social
Economic
Impacts of
Sustainable
Certification
Assess the
evidence base
on the
environmental
and social
economic
impacts of
sustainable
certification of
agricultural
commodities,
tourism
operations, fish
and forest
products
Identify studies of
sustainable
certification,
searched digital
database, citations in
relevant studies and
library catalogues
Only six studies attempt to
construct a credible
counterfactual, two of these
studies find out that
certification has significant
socio-economics benefits and
one study has a significant
environmental impact. Three
studies find out that
certification has minimal
socio economic benefits or
generally a net cost
Although a considerable
literature examines the link
between coffee certification and
the Socio-economic and
environmental characteristics of
farm households, only six
studies attempt to construct a
credible counterfactual and
therefore can be considered
tests of certification‟s causal
impact. Most farm-level coffee
studies simply compare average
characteristics of a sample of
certified and non-certified
farmers.
sustainable certification.
23
Authors and
Year
Study title Objectives Analytical Method Key findings Gaps
Potts &
Sanctuary
(2010).
“Sustainable
Markets are
growing is
Sustainability
keeping
Pace?”
A Perspective
on sustainable
coffee markets
To document
the current state
of a markets for
sustainable
coffee, globally
and within
Sweden, as
well as the
drivers behind
such markets,
highlights key
opportunities
for further
improving the
sustainability of
global market
Survey and analysis of
existing literature
The first study found out
that Fair trade certification
is positively correlated with
coffee volume sold and
price obtained. But less
consistently correlated with
indicators of educational
and health status. The
second study that Fair trade
farmers have lower incomes
and productivity than
convectional farmers.
There is lack of coherence
and coverage of the different
assessment processes applied
across the studies.
The few existing studies on
the impacts of sustainability
initiatives do not provide
representative coverage and
reach uncertain conclusions.
More representative
counterfactually –based
research is needed, in order
to ensure maximum
usefulness.
24
Authors and
Year
Study title Objectives Analytical Method Key findings Gaps
Mercy et al.,
(2010).
The impact
of
certification
on small
holder
coffee
farmers in
Kenya: The
case of UTZ
Certification
program
Estimate the impact
of certification on
income, wealth
expenditure of farm
households
Assess the
economic situation,
willingness to
invest, risk attitude
and loyalty to their
cooperatives
Propensity score
marching approach
The results showed that there were
some differences between the treatment
and control groups that were important
indications of the impact of the UTZ
certification program.
Success that cut across all Household
involved in the two cooperatives that
had been UTZ certified was a higher
price for coffee.
Households that are UTZ certified sold
more coffee than non-certified counter
parts had higher house hold savings and
made more land investment, in Kiambu
cooperatives received more credit, had
more off farm income and more capital
related investments.
Input costs in coffee in the treatment
group were higher than control for
Households in Nyeri and vice versa in
Kiambu Districts.
The certification
program has been in
existence for short
period
25
Authors and Year Study title Objectives Analytical
Method
Key findings Gaps
Stellmacher &
Grote (2011).
Forest coffee
Certification in
Ethiopia
Are there
differences between
the forest
production and
forest management
in certified and
non-certified
cooperatives?
To what extents do
the forest coffee
producers receive
net benefits from
certification?
To what extent are
the forest coffee
producers aware of
and involved in
certification?
Descriptive
analysis
Empirical data shows that farmers
undertake considerable interventions in
the forest ecosystem in order to
increase their coffee yields e.g. by
cutting trees which promotes
degradation of the forest ecosystem and
biodiversity and occurs irrespective of
certification
Empirical data also illustrate practical
difficulties of certification for the
season for some cooperatives did not
pay significantly higher producer
prices than non- certified groups
Certification is not actively promoted
nor understood by those who are
certified. None of the interviewed
members could answer to the question
what certification actually mean.
More in depth
research is
needed on the
underlying
economic and
institutional
incentives of
certification,
with regard to
the sustainable
use of
conservation of
coffee forest
ecosystems and
biodiversity.
26
Authors and
Year
Study title Objectives Analytical Method Key findings Gaps
International
Trade
Centre (ITC)
(2011).
Impact of
private
standards
on
producers
in
developing
countries
A systematic
literature
review to
assess the
resources
tackling the
socioeconomic
and
environmental
impacts of
private
standards at
the producer
level in
developing
countries
Descriptive analysis of
the research including the
type and timing of
publications, the topics
and geographies covered,
methodologies applied
and
Analyzing the literature
using a systematic review
approach
Direct impact of participating in
private standards in terms of price
and profits received by producers
tended to be positive though it was
not a uniform conclusion some
studies found a negative impact on
net income for producers while the
increased earnings did not
compensate for the additional costs
and increased labour involved in
complying with standards requisites
Knowledge base that
exists today in
certification is very
thin, sparse and fragile
in terms of scope
method and depth of
coverage.
Many of the studies in
the field lack a
convincing and
consistent
methodology.
27
Authors
and Year
Study title Objectives Analytical Method Key findings Gaps
Cohn &
O‟Rourke
(2011).
Agricultural
Certification
as a
Conservation
Tool in Latin
America
Impact of
certification of small
holder coffee farmers
in Western Elsavador
A case study using
Semi-structured
interviews were
conducted with
coffee extension
agents, development
project coordinators,
representatives from
cooperative societies
in Elsavador
Certification poses a
greater challenge than
other voluntary schemes
however because it is
often marketed on the
basis of the very
conservation goals it
proposes to achieve.
Consumers and other
supply chain actors care
very little about
conservation outcomes.
Thus eco-certification
schemes are unlikely to do
much but stamp a green
seal of approval on
business as usual
Agricultural certification to
achieve conservation goals is
difficult to design, implement
and evaluate
28
Authors
and Year
Study title Objectives Analytical Method Key findings Gaps
Bennett &
Franzel
(2013).
Can Organic
and
Resource-
conserving
Agriculture
improve
Livelihoods?
Assess the
capacity of
organic and
resource-
conserving
agriculture
to improve
the
livelihoods
of poor
small
holders in
Africa
Reviews of studies done on ORCA.
The methodology for this report was to
identify using internet search engines
including web of science, Google and
goggle scholar, following leads from
other sources studies about the
livelihood effects of ORCA systems on
small-holders in developing countries,
analysis is done to understand the
factors contributing to the likelihood
that small holder farmers adopting
ORCA systems could sustainably
improve their livelihoods.
Results show that ORCA
often outperformed
conventional agriculture
with respect to yield, net
income and food security.
Yield improved upon
conversion to ORCA in 16
out the 25 cases that
reported on it and net
income improved in 19 out
of 23 such cases
Little is known about
the performance of
ORCA initiatives and
the determinants of
their success or
failure, particularly in
developing countries,
research is needed in
the following areas.
Assessing costs,
benefits and impacts
on livelihoods.
Research on social
capital asset, human
and acknowledge
asset, natural capital,
agronomic assets and
contract farming
29
Authors
and Year
Study title Objectives Analytical Method Key findings Gaps
Centre for
Internation
al
Developme
nt Issues
Nijmegen,
(2014).
The impact
of coffee
certification
on small
holder
farmers in
Kenya,
Uganda and
Ethiopia
Several surveys were
done in 2009 and 2013;
case studies were also
done through focus group
discussions. The study
combined with and
without assessment of
certification by
comparing Fair Trade,
Utz and non-certified
cooperatives, and before
and after analysis of
certification by
comparing baseline with
ex-post survey
The results found in the quantitative data are
ambiguous. Involvement in Fair trade
certification does not influence production
volumes in one case (Kiambaa versus Mecari)
and in the other case (Rugi versus Kiama)
negatively influences coffee production
volumes, compared to non-certification.
Utz certified farmers showed higher
production at baseline (2009) compared to NC
farmers, but at end line (2013) these effects
disappear
In prices Fair Trade farmers(Kiambaa)received
higher prices compared to Non Certified
farmers in both years and the difference
between the two grew significantly over the
years
Other Fair-trade certified farmers (Rugi)
received lower prices over time.
The findings
from the study
are quite broad
and more
research is
needed to closely
examine the
effects of
certification.
30
Having reviewed the available literature on coffee certification, there is very little
research verifying improved sustainable development outcomes at the farmer or
field levels. Existing impact research on certification initiatives in the coffee
sector is scarce and incomplete. The few existing counterfactual studies on the
impacts of such initiatives do not provide representative coverage, and reach
uncertain conclusions (Potts & Sanctuary, 2010). In their survey of existing
literature only six studies were found to have applied counterfactual analysis
capable of reporting on impacts associated with participation in certification
initiatives Therefore, this study was undertaken to enhance the knowledge base on
impact of certification among small holder coffee farmers of Embu County.
31
CHAPTER THREE: RESEARCH METHODOLOGY
3.0 Introduction
This chapter outlines the theoretical background of the analytical procedures used.
The selection of the study area, the sampling design, the research instruments as
well as the tools used in the research and analysis are presented.
3.1 Study area
The study was carried out in Manyatta and Runyenjes sub counties of Embu
County. Each of the Sub counties has 12 coffee societies. Embu County is located
at the foothill of Mount Kenya, and covers an area of 2,818 square Kilometers.
The county receives substantial rainfall with average annual precipitation of
1206mm. The wettest season is experienced between March and July while the
hottest comes between January and mid-March. Temperatures are estimated at an
average of between 9°C - 28°C.Much of the land is largely arable and is well
watered by a number of rivers and streams. Agriculture is the main driver of the
economy in this county with over 70% of the residents being small scale farmers.
Tea, coffee and cotton have been the main cash crops (MoA, 2011).
32
Coffee growing by counties
Figure 3.1: Study area
33
3.2 Study design
This study adopted the descriptive research design to yield both qualitative and
quantitative data in order to interpret effects of coffee certification on household
characteristics. Descriptive surveys was used when collecting information about
households‟ attitude, opinions, habits or any of the variety of social and education
factors (Kombo & Tromp,2009).
This study was structured to provide results that objectively demonstrate
certification effects on coffee productivity and prices.
3.3 Sample and sampling design
The study comprised of small holder coffee farmers of Embu County. Multi stage
sampling procedure was used. Firstly, Embu County was purposively sampled out
of the 47 counties since it had the highest number of certified societies. Six co-
operative societies were purposively sampled three that were certified and three
that were not certified to act as control but with similar characteristics such as
membership and the number of wet mills. The six selected cooperatives included
Muramuki, Rianjagi and Kamurai which were certified while Ivinge,
Kithungururu and Kirindiri were the control (non-certified) group. Further simple
random sampling was applied in each of the six cooperatives to select a total of
480 households and this was determined using the sampling methodology of
(Mugenda & Mugenda, 2003).
34
n=
Where
n= required sample size
z = table value from the normal table
p = probability of success.
q = (1-p) probability of failure.
e = desired precision 5% (standard value of 0.05).
n=
= 384
Because of estimating the propensity score matching the sample was increased to
480farmers; Muramuki (82), Rianjagi (71), Kamurai (85), Ivinge (80),
Kithungururu (80) and Kirindiri (82).
Data was collected through single farm visit interviews using structured
questionnaires administered to respondents by enumerators in November 2012.
Respondents were identified with the assistance of the cooperative society‟s staff.
3.4 Research instruments
Both primary and secondary data were used in this study. Primary data was
collected through a structured questionnaire with open and closed ended questions
were used. The secondary data was obtained from the existing literature, which
included; published reports, journals, magazines and books, society‟s sales report,
35
certification licenses and audit reports. Quantities of cherry deliveries by the
members were obtained from monthly and annual entries of the societies and
clean coffee records were obtained from the society and the coffee millers. The
researcher personally visited the selected societies, factories and members and
administered the questionnaire to obtain primary data.
3.5 Data analysis
The data collected was examined, coded and categorized using Stata software
version 12. Descriptive statistics was used to explore the underlying features in
the data on coffee certification and its impact on small scale farmers of coffee.
Descriptive statistics was used to assess the households‟ characteristics in order to
determine the general performance between the non- certified and the certified
group. Further, the t test was used to determine if the differences between the two
groups were statistically significant at the 1% - 10% significance levels.
3.5.1 Factors that influence small holder coffee farmer’s decision to
participate in certification
The Logit model was used to estimate the propensity scores using household
characteristics. These characteristics included covariate variables that seemed to
influence the participation decisions and the outcome of interest. The coefficients
were used to calculate propensity scores, and certified small holder coffee farmers
matched with non-certified small holder coffee farmers based on having similar
36
propensity scores. In estimating the Logit model (Gujarati, 2004) the dependent
variable was certification, which took the value of one if a household was
certified and zero otherwise. The mathematical formulation of Logit model was as
follows:
(1)
Where, pi was the probability of participation for the ith
household and it ranges
from 0-1
Z is a function of N-explanatory variables which is also expressed as:
+ ∑ + (2)
Where,
į= 1, 2, 3… n
= intercept
= regression coefficients to be estimated or Logit parameter
= a disturbance term, and
= certification characteristics which were specified as:-
x1 – Gender of household head
x2– Age of household head
x3 – Education level of household head
x4– Distance from the coffee factory (km)
x5– Farmers perception
x6– Price of coffee (Kshs/kg)
37
x7 – Income from coffee (Kshs)
x8 – Awareness level
Hypothesis one which predicts there are factors that influence small holder coffee
farmer‟s decision to participate in certification, would be rejected in the event that
the coefficients above did not significantly influence the decision of the farmer to
participate in certification.
The probability that a household is none certified is
1 - pi
(3)
The odd ratio can be written as:
- =
= (4)
- is the odds ratio in favor of farmers participating in the certification
program. This is defined as the ratio of the probability that a household/family
will participate in certification to the probability that a household will not
participate in certification.
Lastly, by taking the natural log of equation above the log of odds ratio can be
written as:
= ln(
-
) =
+ ∑ + (5)
38
The results were presented as descriptive statistics of dependent and independent
variables as used in the study as well as empirical results of logistics regression
analysis.
3.5.2 Propensity score matching method
One of the critical problems in non-experimental methods is the presence of
selection bias which could arise mainly from the non-random selection of
participant households that make evaluation problematic (Heckman et al., 1998).
An important problem of causal inference is how to estimate treatment effects in
observational studies, situations (like an experiment) in which a group of units is
exposed to a well-defined treatment, but (unlike an experiment) no systematic
methods of experimental design are used to maintain a control group. It is well
recognized that the estimate of a causal effect obtained by comparing a treatment
group with a non-experimental comparison group could be biased because of
problems such as self-selection or some systematic judgment by the researcher in
selecting units to be assigned to the treatment (Dehejia & Wahba, 2002).
Propensity scores are an increasingly common tool for estimating the effects of
interventions in non-experimental settings. The approach solves the “selection”
problem (Rosenbaum & Rubin, 1983) by identifying from among the non-target
group, households with similar pre-treatment characteristics X as those of the
target group. Any differences in outcomes in the target and control groups are
assigned to the intervention.
39
In order to achieve objectives 2 and 3, propensity score matching was employed
to know the impact of certification on different outcome variables. It is chosen
among other non-experimental methods because the treatment assignment is not
random and considered as second-best alternative to experimental design in
minimizing selection biases (Baker, 2000).
Propensity score matching entails forming matched sets of treated (certified) and
untreated (non-certified) subjects who share a similar value of the propensity
score (Rosenbaum & Rubin, 1985). The most common implementation of
propensity score matching is one-to-one or pair matching, in which pairs of
treated and untreated subjects are formed, such that matched subjects have similar
values of the propensity score. The matching of households in treatment and
control groups is based on a balancing score b(x) which is a function of the
covariates X. The balancing score used is based on the likelihood of participation
in a development program given the observed characteristics X (Mercy et al.,
2010).
Propensity scores are estimated from the initial conditions using a logit model
specified as follows:
Log (w =1x) =q (b +dX +e) c
40
Where:
wc = Dichotomous variable taking a value of one if household belongs to a
treatment group and zero otherwise
q = Represents a normal cumulative distribution function
b &d = Parameters to be estimated
X = Household characteristics that are hypothesized to influence households
belonging to a treatment group e is an error term
3.5.3 Impact of certification on coffee productivity
In order to assess the impact of certification on coffee productivity, the propensity
score matching methodology was used to adjust for the selection bias and estimate
the counterfactual effects (Rosenbaum et al., 1983) by identifying households that
have similar pre-treatments as the target group. PSM was used to extract
comparable pair of treatment-comparison households in a non-random program
setup. The hypothesis that certified farmers did not have increased coffee
productivity than the non-certified farmers would be rejected if there was no
significant difference on productivity of coffee between certified farmers and the
non-certified farmers.
41
3.5.4 Impact of certification on coffee price
The impact of coffee certification on coffee prices was also analyzed using the
propensity score matching technique. To achieve this, the average treatment effect
on certification group was calculated. The calculated values were then tested
using the t statistic at the 10, 5 and 1% levels of significance to determine if the
difference between the prices of the certified (certification) and the non-certified
(control groups) were statistically significantly different. The hypothesis that
certified farmers did not receive higher coffee prices than the non-certified
farmers would be rejected if there was no significant difference between the
prices of coffee received by certified farmers and the non-certified farmers.
42
CHAPTER FOUR: RESEARCH FINDINGS AND DISCUSSIONS
4.1 Introduction
This section presents the statistical analysis of the study results. It consists of four
subsections. The first subsection presents the results of descriptive statistics of the
different variables considered under the study. The second subsection gives the
propensity score matching and the effect of treatment on coffee production and
prices. The third subsection is the logistic regression results that utilize the Logit
model to assess the factors that significantly influence participation in coffee
certification. In the fourth subsection, the results of sensitivity analysis are
presented.
4.2 Descriptive Statistics
The results of this analysis are presented in Table 4.1. Household characteristics
such as marital status, age of the household head and distance to the factory
between the certified and non-certified farmers were found to be significantly
different. Characteristics such gender of the household head and education level
though exhibited differences in means; were not statistically significantly
different.
43
Table 4.1: Descriptive statistics of coffee farming households in Embu North district Sub County
Total Certified Non-certified
Variable Mean n Mean n Mean n t-Stat
Household characteristics
Gender of respondent (1 = Male, 0 = Female)
0.61
480
0.59
238
0.63
242
0.64a
Head of household (1 = Husband, 0 = Wife) 0.79 478 0.80 237 0.79 241 0.53a
Marital status of farmer (1 = Married, 0 = Single) 0.94 479 0.97 237 0.91 242 -1.38a *
Age of household head (1 = 18-30 years, 2=31-40 years,
3 = 42-50 years, 4 = 51-60 years, 5 = Over 60 years) 3.59 477 3.32 235 3.85 242 -4.7
a ***
Education of household head (1 = None, 2 = Primary,
3 = Secondary, 5 = Tertiary) 2.31 479 2.35 237 2.28 242 1.01
a
Distance to the factory (km) 2.01 478 1.88 236 2.15 242 -1.26 *
Land and Acreage
Size of coffee farm (acres) 1.99 480 1.95 238 2.03 242 -0.80
Acquired land (acres) 1.12 480 1.10 238 1.14 242 -1.29*
Number of mature trees planted in 2007 312.31 480 280.42 238 343.68 242 -3.7**
44
Total Certified Non-certified
Variable Mean n Mean n Mean n t-Stat
Number of mature trees planted in 2012 279.80 480 276.80 238 282.74 242 1.96
Coffee area in acres in 2007 0.63 313 0.59 71 0.64 242 -2.25**
Coffee area in acres in 2012 0.54 313 0.58 71 0.52 242 -1.36*
Planted coffee in last 5 years 1.65 480 1.64 238 1.65 242 -0.15
Number of trees planted 35.55 480 44.63 238 26.84 242 2.39*
Coffee production (cherry) in Kilograms
Cherry produced in 2006/2007 469.41 480 398.11 238 539.53 242 -2.54**
Cherry produced in 2007/2008 344.59 480 401.32 238 288.80 242 2.06**
Cherry produced in 2008/2009 486.59 480 482.32 238 490.79 242 -1.13
Cherry produced in 2009/2010 534.44 480 574.97 238 494.59 242 1.20
Cherry produced in 2010/2011 288.68 480 348.60 238 229.75 242 2.51**
Coffee prices (cherry) in Kenya shillings
Price per kg in 2006/2007 25.46 480 20.97 238 19.85 242 6.3***
Price per kg in 2007/2008 29.14 480 30.86 238 27.46 242 7.1***
Price per kg in 2008/2009 31.78 480 31.42 238 32.14 242 -2.2**
Price per kg in 2009/2010 51.52 480 51.52 238 50.58 242 2.21**
45
Total Certified Non-certified
Variable Mean n Mean n Mean n t-Stat
Price per kg in 2010/2011 72.28 480 72.28 238 64.19 242 9.95***
Coffee income in 2006/2007 14905 480 18807 238 11067 242 -1.40
Coffee income in 2007/2008 10342 480 12256 238 8459 242 2.63*
Coffee income in 2008/2009 15603 480 15177 238 15022 242 0.34
Coffee income in 2009/2010 28358 480 30919 238 25840 242 -1.84*
Coffee income in 2010/2011 21685 480 28613 238 14871 242 3.86**
Other crops 2006/2007 26093 313 41846 71 21471 242 2.90**
Other crops 2007/2008 32632 480 43313 238 22127 242 1.83*
Other crops 2008/2009 34531 480 47112 238 22158 242 2.53**
Other crops 2009/2010 37721 480 52656 238 23034 242 2.34**
Other crops 2010/2011 34586 480 45759 238 23898 242 3.00**
Certification
Heard of coffee certification (1 = Yes, 0 = No) 0.61 480 0.99 238 0.33 242 -17a ***
Where heard of certification (1 = Management,
2 = Neighbour, 3 = Media, 4 = Government officer 1.22 286 1.07 214 1.67 72 -6.7
a ***
* Significant at 10% ** Significant at 5% *** Significant at 1%
a test for continuous variables t-stat, dummy variables chi-square
46
4.3 Factors that influence farmers’ decision to participate in coffee
certification
The research design sought to establish the factors that influenced farmers‟
participation in coffee certification in Embu County.
4.3.1 Logistic Regression
A multiple logistic regression was performed to estimate the relationship between
participation in coffee certification and the independent variables under study
namely; household head, age of household head, education level, distance from
the factory, farmers‟ perception, coffee income and awareness level.
4.3.1.1. Regression diagnostics
Before running the regression model, the following regression diagnostics were
carried out in order to ensure that the requirements for regression analysis were
met. This included testing for normality, multi-collinearity and Hetero-
scedasticity as discussed below.
4.3.1.1.1. Normality
Logistic regression requires that the assumption of normality be met. The Shapiro
Wilk statistics was used to determine if the residuals response variable and the
explanatory variables followed the normal distribution. The results of the Shapiro
Wilk statistics confirmed that the residuals were normally distributed (W =
0.9997, p = 1.000).
47
4.3.1.1.2. Multicollinearity
Multicollinearity means that some of the explanatory variables are not
independent but are correlated. When multicollinearity is present it becomes
difficult to assign the change in the dependent variable precisely to one or the
other of the explanatory variables (Gujarati, 2004).
Tests to determine if the data met the assumption of collinearity were carried out
using the Variance Inflation Factor (VIF) and these indicated that
multicollinearity was not a concern (1.02 ≤ VIF ≤ 1.3). There were therefore no
two variables that were highly correlated and the logistic regression analysis was
carried out without further analysis on individual variables (Belsley et al., 1980).
Table4.2: Test for multicollinearity
Coffee certification (0 = non certified, 1 = Certified) VIF 1/VIF
Average rate paid per kg of coffee 1.36 0.74
Age of the household head (1 = Husband, 0 = Wife) 1.23 0.81
Distance to the coffee factory (km) 1.21 0.82
Education (1 = None, 2 = Primary, 3 = Secondary, 4 = Tertiary) 1.20 0.83
Have you ever heard of coffee certification? (1 = Yes, 0 = No) 1.17 0.86
Gender of household head (1 = Male, 0 = Female) 1.12 0.80
Average income coffee in (Kshs) 1.04 0.96
Consider other crops profitable than coffee? (1 = Yes, 0 = No) 1.02 0.98
Mean VIF 1.17
48
4.3.1.1.3. Heteroscedasticity
Heteroscedasticity means that the variance of the residuals is non-constant. A test
of homoscedasticity of error terms was carried out to determine whether the
logistic regression model's ability to predict the response variable (certification)
was consistent across all values of the explanatory variables. This was performed
using the Breusch-Pagan/Cook - Weisberg results which failed to support the
presence of heteroscedasticity ( = 35.42, p = 0.0000).
The logistic results presented in Table 4.3 below indicate that the estimate logistic
regression model is good fit for the propensity matching score (Pseudo R2 = 0.50).
From these results, there was statistically significant evidence that participation in
the coffee certification program is positively influenced by the following
explanatory variables: gender of household head, distance to the factory (both
significant at 10% level of significance), price of coffee and farmers‟ awareness
level (both significant at 1% level of significance). The study thus finds that
households headed by wives, furthest from the factory and who were aware of the
existence of coffee certification had a higher likelihood of participating in the
certification program. Further, the results indicate that certified farmers received
higher prices for each kilogram of cherry delivered to the factory.
On the other hand, the age of household head was found to negatively affect the
participation of farmers in the certification program. Younger farmers were
49
therefore more likely to participate in certification than their older counterparts.
This result was statistically significant at 5% level of significance.
The regression parameters as presented in Table 4.3 when inserted into the
regression model leads to the predicted logistic regression equation below.
ln (
) = -3.85 + 0.72x1 – 0.31x2 – 0.23x3 + 0.22x4 + 0.10x5 + 0.22x6 + 0.10x7–
4.23x8
Where ln(
) = is the log (OR) in favor of participation in coffee certification
where p is the probability of a farmer participating in the coffee certification.
50
Table 4.3: Logistic regression results for coffee farmers in Embu County
Covariates Coefficients Std. Error Z P>|z|
Gender of household head 0.72 0.39 1.83 0.07*
Age of the household head -0.31 0.12 -2.51 0.01**
Education level of household
head -0.23 0.20 -1.13 0.26
Distance to the coffee factory
(km) 0.22 0.11 1.88 0.06*
Farmers‟ perception 0.10 0.29 0.34 0.74
Price of coffee (Kshs/kg) 0.22 0.04 5.82 0.00***
Income from coffee (Kshs) 0.01 0.01 1.33 0.18
Farmers‟ awareness level 4.23 0.42 -10.05 0.00***
Cons -3.85 1.91 -2.02 0.04
N 472.00
LR (8) 323.96
Prob > 0.00 ***
Pseudo R2 0.50
Log Likelihood -165.15
* Significant at 10% ** Significant at 5% *** Significant at 1%
4.3.2 Odds ratio results
Table 4.4 gives the calculated odds ratios of the above regression parameters. The
log odds of households headed by men participating in coffee production was
found to be 2.06 (p = 0.07) when all other variables are held constant. This
51
implied that the families headed by wives had higher odds of participating in the
coffee certification.
The odds ratio associated with a one-year increase in the age of a household head
was 0.73 (p = 0.01), younger farmers had higher odds of participating in the
coffee certification as compared to their older counterparts.
Distance to the factory had an odds ratio of 1.24 (p = 0.06). Farmers who are
close to the factory were found to have higher odds of participating in the
certification program. Similarly, farmers who had the perception that, other crops
were more profitable than coffee had lower odds of participating in the
certification. The odds of coffee price and income from coffee was found to be
1.25 (p = 0.00) and giving a clear indication that certified farmers received higher
prices for coffee than the non-certified farmers. The odds of farmers‟ awareness
level were 0.01 which implies that farmers who were informed on the existence of
the coffee certification had higher odds of participating in certification.
52
Table 4.4: Odds ratios results
Covariates Odds Ratio Std. Error Z P>|z|
Gender of household head 2.06 0.81 1.83 0.07*
Age of household head 0.73 0.09 -2.51 0.01**
Education level of household head 0.80 0.16 -1.13 0.23
Distance to the coffee factory 1.24 0.14 1.88 0.06*
Farmers perception 1.10 0.32 0.34 0.74
Coffee price 1.25 0.05 5.82 0.00***
Income from coffee 1.00 0.00 1.33 0.18
Awareness level 0.01 0.01 -10.05 0.00***
Cons 0.02 0.04 -2.02 0.04**
N 472.00
LR (8) 323.96
Prob > 0.00 ***
Pseudo R2 0.50
Log Likelihood -165.15
* Significant at 10% ** Significant at 5% *** Significant at 1%
4.4 Propensity Scores and Covariates
The propensity score matching methodology was used to adjust for the selection
bias and estimate the counterfactual effects (Rosenbaum et al., 1983) by
identifying households that have similar pre-treatments as that of the target group.
53
Different algorithms were employed in matching the certification and control
groups; from which the final matching procedure was selected using the equal
mean tests criterion (Dehajia & Wahba, 2002) and the pseudo-R2.
Table 4.5: Performance of different matching estimators
Performance criteria
Sample size Balancing test Pseudo-R2
Nearest Neighbour
NN(1) 13 0.15 334
NN(2) 18 0.08 343
NN(3) 13 0.40 320
Kernel matching
Band width 0.01 08 0.13 244
Band width 0.25 14 0.24 289
Band width 0.50 18 0.09 256
Radius caliper
Radius 0.01 12 0.21 278
Radius 0.25 10 0.25 242
Radius 0.50 14 0.07 263
In these criteria, it is suggested that a matching estimator be used if it results in
insignificant mean differences between the non-certified and certified, gives a
large sample size and a lower pseudo-R2after performing the matching. The
results in Table 4.6 show the results of the different matching algorithms, nearest
54
neighbour, Kernel and Radius caliper matching. Following the above criteria, the
nearest neighbour algorithm was found to be the best estimator and thus was used
to check the balance of propensity scores and covariates, the t-test was used to test
the equality of means while the chi-square test in Table 4.6 of the same analysis
was used to test the joint significance of the variables that were used in the
analysis.
Table 4.6: Chi-square test for significance
Sample Pseudo R2 LR P > Mean Bias Med Bias
Raw 0.499 326.80 0.000 47.2 17.3
Matched 0.217 1270000.00 0.000 39.4 11.8
The propensity scores were estimated on certified and non-certified farmers and
the Pseudo R2 before and after matching compared using the method of Sianesi,
(2010). The presented values of Pseudo R2 for the unmatched (raw) and matched
results in Table 4.7 indicate that both the certified farmers and the non-certified
farmers have the same distribution after the matching procedure. Thus the
matching algorithm balanced the characteristics in both the comparison groups.
The results presented in the next sections, estimation of average treatment effect
on certified group for the households‟ certification program are based on the
above algorithm.
55
4.5 Assessing the impact of certification on farm level coffee productivity
The second objective was to assess the impact of certification on coffee
productivity per household. This was done by computing the ATT for coffee
production.
4.5.1 Quantity of coffee produced
Table 4.7 presents the results of ATT for coffee production from which it was
found that at the 10% level of significance, the certified farmers produced
significantly more coffee than the non-certified farmers in the year 2008/2009.
However, in the year 2010/2011, non-certified farmers produced significantly
more coffee than the certified farmers.
Table 4.7: ATT for coffee production in kilograms for farmers in Embu
County between 2006 and 2007
Year Quantity of coffee cherry produced in kilograms
Certified Non-certified Difference SE t-stat
2006/2007 419.53 352.86 66.67 86.57 0.77
2007/2008 363.27 264.71 98.55 161.75 0.61
2008/2009 534.96 417.98 116.98 114.99 1.02*
2009/2010 439.69 463.27 -23.57 114.19 -0.21
2010/2011 195.16 263.51 -68.35 72.44 -0.94*
* Significant at 10%
56
4.5.2 Distance, farm size, acquisition and number of coffee trees
The study considered distance, farm size, acquisition and number of mature
coffee trees owned by farmers as factors that contribute to the quantity of coffee
produced. Table 4.8 presents the results of each of these estimates. It was found
that at the 10% level of significance, the variables size of farm, coffee area in
2012 and number of coffee trees were statistically significantly different between
the certified and non-certified farmers. In particular, certified farmers had large
farms, larger acreage under coffee and more coffee trees than the non-certified
farmers.
Table 4.8: Distance, farm size, acquisition and number of coffee trees for
coffee farmers in Embu County
Variable Certified Non-certified Difference SE t-stat
Distance to factory 1.80 1.77 0.03 0.22 0.12
Size of farm 2.02 1.80 0.23 0.18 1.27*
Land acquisition 1.14 1.14 0.00 0.08 0.00
Coffee area in 2007
(acres)
0.584 0.495 0.089 0.106 0.85
Coffee area in 2012
(acres)
0.563 0.404 0.160 0.091 1.77*
Number of coffee
trees
55.87 33.65 22.21 14.66 1.52*
* Significant at 10%
57
4.6 The impact of certification on coffee prices
The third objective was to evaluate the effect of certification on coffee prices.
This was done by computing the ATT for average income from coffee and other
crops per household.
4.6.1 Impact propensity estimate on coffee prices
Table4.9: ATT for the prices of coffee in Kenya shillings for farmers in
Embu County between 2006 and 2007
Year Price per kg (Kenya shillings)
Certified Non-certified Difference SE t-stat
2006/2007 18.00 19.47 -1.47 0.25 -5.94**
2007/2008 33.00 29.87 3.14 0.95 3.30**
2008/2009 33.00 33.34 -0.34 0.49 -0.70
2009/2010 50.00 54.56 -4.56 1.52 -3.00**
2010/2011 90.55 62.94 27.61 1.86 14.85**
** Significant at 5%
There was statistically significant evidence that certified farmers received a higher
price of coffee in the years 2007/2008, and 2010/2011 but in the years 2006/2007
and 2009/10 the non-certified farmers received higher coffee prices than the
certified farmers. However, in year 2008/2009 the price between the certified and
the non-certified farmers was not statistically significant.
58
4.6.2 Impact propensity estimate on income from coffee and other crops
Table 4.10 shows that the certified farmers earned 68.82% more income from
other crops than the non-certified farmers. This result was found to be statistically
significant. Certified farmers also earned 5.69% more income from coffee than
the non-certified farmers. This result is however not statistically significant.
Table 4.10: ATT for combined income from coffee and other crops for
farmers in Embu County between 2006 and 2007
Year Average income (Kenya shillings)
Certified Non-certified Difference SE t-stat
2006/2011 50,228.57 15,663.10 34,565.47 11,633.13 2.97**
2007/2011 15,369.94 14,495.20 874.74 3,580.06 0.24
** Significant at 5%
4.6.3 Impact propensity score on income from other crops
Table 4.11 shows the empirical results of the analysis of the ATT for the income
from other crops planted by farmers besides coffee. The analysis for the income
from other crops between the years 2006 and 2011 indicated that the certified
households earned 68.34%, 69.86%, 69.05%, 68.24% and 67.71% more income
from other crops than the non-certified households in the years 2006/2007,
2007/2008, 2008/2009, 2009/2010 and 2010/2011 respectively. From the t-stat
column, it can be concluded that all the results were statistically different at the
5% level.
59
Table 4.11: ATT for income from other crops in Kenya shillings for coffee
farmers in Embu County between 2006 and 2007
Year Income (Kenya shillings)
Certified Non-certified Difference SE t-stat
2006/2007 47,413.88 14,527.76 32,886.12 11,528.50 2.85**
2007/2008 49,403.67 14,891.84 34,511.84 11,708.67 2.95**
2008/2009 49,958.78 15,464.27 34,494.49 11,667.82 2.96**
2009/2010 50,875.10 16,159.18 34,715.92 11,716.12 2.96**
2010/2011 53,491.43 17,272.45 36,218.98 11,741.13 3.08**
** Significant at 5%
4.6.4 Impact propensity estimate on income from coffee
Table 4.12 shows that certified farmers generally earned more income from coffee
than the non-certified farmers in the years 2006/2007, 2007/2008, 2008/2009 and
2010/2011.However, the non-certified farmers earned more income from coffee
in year 2009/2010 than the certified farmers. At 5% level of significance none of
the results were found to be statistically significant therefore not different.
60
Table 4.12: ATT for income from coffee in Kenya shillings for farmers in
Embu County between 2006 and 2007
Year Income from coffee (Kenya shillings)
Certified Non-certified Difference SE t-stat
2006/2007 7,551.55 7,207.340 344.16 1709.06 0.20
2007/2008 11,987.76 8,353.96 3,633.80 5347.04 0.68
2008/2009 17,653.65 14,108.17 3,545.49 3864.85 0.92
2009/2010 21,984.69 26,794.62 -4809.93 6684.59 -0.72
2010/2011 17,672.03 16,011.86 1,660.17 4810.94 0.35
4.7 Sensitivity analysis
Sensitivity analysis was performed using the Rosenbaum bounding approach
(Rosenbaum & Rubin, 2002) as a validity check for unobserved selection bias. All
the variables that influence participation in coffee certification and the outcome
variables were observed simultaneously.
The results of the sensitivity analysis are presented in Table 4.13 above. In the
calculation of the results, the certified and non-certified farmers were allowed to
differ in their odds of being certified up to 100%. All the p-values at different
levels of ey (log odds of differential due to unobserved factors) were found to be
statistically significant implying that significant covariates influencing both
certified and non-certified farmers and the outcome variables were considered.
61
Therefore, it can be concluded that the average treatment on the certification
(ATTs) were not affected by unobserved selection bias.
Table4.13: Sensitivity analysis using the Rosenbaum bounding approach to
check for selection bias
Variable ey = 1.00 e
y = 1.25 e
y = 1.50 e
y = 1.75 e
y = 2.00
Household head 3.80E-05 5.90E-10 3.30E-03 7.50E-03 3.40E-07
Marital status 8.00E-04 6.80E-04 6.70E-02 3.20E-07 8.70E-04
Age of household head 1.20E-07 1.80E-05 2.60E-08 8.00E-08 3.00E-04
Education level 8.30E-10 3.20E-06 7.70E-05 4.90E-10 3.90E-11
Years of coffee farming 3.20E-03 1.80E-05 9.00E-04 4.80E-05 3.10E-12
Distance to the factory 6.80E-04 2.40E-08 3.60E-11 5.00E-07 1.20E-05
Farm size 3.40E-07 2.60E-07 5.20E-06 8.90E-07 6.20E-03
Mature coffee trees in 2007 4.60E-09 4.70E-05 6.40E-12 7.00E-12 2.40E-08
Mature coffee trees in 2012 3.70E-11 4.60E-04 7.70E-05 2.10E-06 1.30E-07
Coffee area in 2007 7.70E-04 3.60E-08 5.80E-08 2.10E-04 5.30E-12
Coffee area in 2012 6.00E-02 3.40E-05 4.20E-05 6.00E-08 4.00E-08
Average income from coffee 1.00E-08 5.00E-04 3.60E-07 1.70E-08 7.60E-05
Quantity of coffee produced 1.70E-04 4.50E-10 8.00E-05 4.90E-11 6.10E-03
Rate paid per kg of coffee 2.80E-03 7.10E-07 3.00E-06 7.90E-07 5.50E-07
62
CHAPTER FIVE: DISCUSSION
5.0 Introduction
5.1 Factors that influence small holder coffee farmer’s decision to participate
in certification
Logistic regression analysis was done to estimate the relationship between
participation in certification program and the independent variables under study
namely: gender of the household head, age of the household head, education of
the household head, distance from the coffee factory, farmer‟s perception, price of
coffee, income from coffee and awareness level.
Gender of the household head and distance from the factory (Table 4.3), were
found to positively influence decision to participate in certification. Household
that were headed by women were more likely to participate in certification. While
women do most of the work in the coffee farms, face unequal treatment in
leadership and discrimination, yet men receive all the coffee payments.
Certification programs such as Fair Trade strive to help women realize their full
potential through empowerment trainings which promotes female cooperative
membership (Fair trade, 2010).
Age and education of household head had a negative impact in the participation of
certification program. These findings were consistent with Mercy et al., (2010)
that although not significant, education level and age of household head had an
inverse relationship to participation in the certification programs.
63
Awareness of certification programs had a positive impact on participating in the
coffee certification program which means farmers who were aware of the
program were more likely to participate than farmers who had no information.
This is because though through certification rigorous capacity building and
trainings are done on good agricultural practices, certification programs tend to
select potential partners in areas where farmers' cooperatives are effective. Where
cooperatives are not effective many farmers are scarcely informed about
certification and its different aspects (CIDIN, 2014).
5.2 Impact of certification on coffee productivity
The results from the study showed that the certified farmers produced more coffee
in early years compared to their non-certified counterparts; these effects then
disappeared as the non-certified farmers started learning from the certified
farmers and their coffee production increased even though they were not certified.
The findings from CIDIN (2014) are inconsistent. Involvement in certification
does not influence production volumes in one case (Kiambaa FCS versus Mecari
FCS) and in the other case (Rugi FCS versus Kiama FCS) certification negatively
influences coffee production volumes, compared to non-certified farmers.
Certified farmers showed higher production at baseline (2009) compared to non-
certified farmers, but at end line (2013) these effects disappear.
The findings were not consistent with Mercy et al., (2010) that participation in the
certification programs increased prices (and incomes) and productivity,
64
households belonging to the certified cooperative sold/ produced more coffee than
their non-certified counterparts (Mercy et al., 2010).
The findings were also not consistent with Fort & Ruben (2008) that yield levels
of Fair trade certified farmers are slightly higher than their non-certified
counterparts.
5.3 Impact of certification on coffee prices
The results of this study found out that certified farmers received higher coffee
prices in 2007/2008, but gains disappeared until the end line year 2010/2011
where they were significantly high.
The findings agreed with CIDIN (2014) where initial gains from certification are
usually high, but these tend to disappear once other non-certified farmers catch up
in the process. Most initial gains from trade, therefore, gradually disappear due to
spatial externalities. It points to important certification effects in the beginning of
the coffee life cycle that tends to even out over time.
Farmers tend to be better off financially when participating in certification
standards. The direct impact of participating in certification standards in terms of
price and incomes received by producers tended to be positive. However, this is
not a uniform conclusion. Jaffee (2008) found mixed evidence on the net income
for producers, where the increased earnings did not compensate for the additional
costs and increased labour involved in complying with standards requisites.
65
The overall net impact of certification, however, may or may not be completely
visible for the producer when exporters, donors or NGOs temporarily cover
certification costs, as markets mature, there is a risk that increased supply of
certified products may create increased competition to find buyers, certifications
become „commoditized‟ and premiums diminished or eliminated (Nebel et al.,
2005).
The findings did not agree with Arnould et al., (2009) that Fair trade farmers
obtain higher prices than non-Fair trade farmers. Certified farmers garner an
increased share of coffee prices relative to non-certified farmers.
The findings coincided with Fort & Ruben (2008) that there was also lack of a
real price difference between fair-trade and non-fair-trade producers, but were not
consistent with Bacon (2005) that Fair Trade certified farmers received higher
prices than the non-certified farmers.
66
CHAPTER SIX: SUMMARY, CONCLUSION ANDRECOMMENDATIONS
6.0 Introduction
The chapter presents the conclusion and recommendations for the study the
conclusions are based on the findings for each objective.
6.1 Conclusion
This study focused on the impact of certification among small holder coffee
farmers in Embu County and the following are the conclusions that were made
from this study:
1. The findings showed that there were factors that influenced farmers‟ decision
to participate in certification programs. In particular three explanatory
variables were found to significantly influence participation in the program.
These variables were gender of household head, price of coffee and awareness
level. Households that were headed by females were more likely to participate
in the program than those headed by males. Younger farmers were also more
likely to participate in certification as compared to their older counterparts.
The prices of coffee and households‟ awareness level both had a high
statistically significant impact on the participation in the program. In the
latter, high coffee prices were found to be associated with an increase in the
participation of households in the program, while lower coffee prices were
associated with lower participation in the program. On the other hand,
67
households that had more knowledge of certification programs were more
likely to participate in certification programs. The results were consistent with
the hypothesized relationship.
2. There were inconsistencies in the results of impact of certification on coffee
productivity. Whereas results from the study showed that the certified farmers
produced more coffee in some years, other years the non-certified farmers
produced more coffee than the certified farmers.
3. Certified farmers received higher coffee prices in 2007/2008, but gains
disappeared until the end line year 2010/2011 where the gains were
significantly high. There are remarkable prices received by certified farmers
in some years compared to the non-certified farmers, and there are also
remarkable prices received by the non-certified farmers compared to certified
farmers in some years.
68
6.2 Recommendations
1. On awareness level, policy makers need to focus on training and
empowering coffee farmers and coffee stakeholders on the different
certification programs available in Kenya, so that farmers they are aware
and can make decisions on whether to be certified or otherwise and also
on the type of certification to implement.
2. The study focused on impact of certification on coffee farming, future
studies need to consider impact of certification on other farm enterprises
in addition to coffee.
3. This study focused on socio-economic impact of participation in
certification standards. Further research on social, environmental and
socio-economic impact assessment needs to be done using emerging
business evaluation models and social return to investment.
69
REFERENCES
Agricultural Fisheries and Food Authority (AFFA). Found at
http/www.agricultureauthority.go.ke. (Accessed on 10th
January, 2015).
Arnould, E., A. Plastina, & D. Ball. (2009). Does Fair Trade delivers on its core
value proposition? Effects on income, educational attainment, and health in three
countries. Journal of Public Policy and Marketing 28(2): 186–201.
Bacon, C. (2005). Confronting the coffee crisis: Can Fair Trade, organic, and
specialty coffees reduce small-scale farmer vulnerability in northern Nicaragua?
World Development 33(3): 497–511.
Baker, J.L. (2000). Evaluating the impact of development projects on poverty. A
handbook for Practioners. Washington D.C. World Bank.
Beckman, B. (1998). Fair trade and trade development, Leisa Magazine,
London.
Bell, J. (1993). Doing your research project: A guide for first time researchers in
education and social science, Open University Press, Buckingham, England.
Belsley, D.W., Kuh, E. & Welsch, R.E. (1980), Regression Diagnostics,
Statistica Neerlandica 34, pp. 161-164.
Bennett, M. & Frazel, S. (2013). Can organic and resource-conserving
agriculture improve livelihoods? International Journal of Agricultural
Sustainability, 11:3, pp.193-215 Geneva.
Blackman, A. & Rivera, J. (2010). The evidence base for environmental and
socio economic mpacts of ‘susta nabl ‘c rt f cat on. Washington, D.C.
Resources for the Future.
70
Blackman, A. & Naranjo, M.A. (2010). Does eco-certification has
environmental benefits? Organic coffee in Costa Rica. Working paper.
Washington, DC: Resources for the Future.
Bolwig, S. P. Gibson, & Jones, S. (2009). The economics of small holder
organic contract farming in tropical Africa. World Development 37(6), pp. 1094–
1104.
Bredahl, J., Helles, f., Nebel, G. & Quevedo, L. (2005). Development and
economic significance of for st c rt f cat on: th cas of FSC n Bol v a’, Forest
Policy and Economics, vol. 7, no. 2, pp. 175-186.
Burns, M. & Blowfield M. (2000). Approaches to ethical trade: Impact and
lessons learned. www.nri.org/NRET/burns_final.pdf. (Accessed on 22nd
December, 2011).
Caliendo, M. & Kopeinig, S. (2008). Some Practical Guidance for the
Implementation of Propensity Score Matching, Journal of Economic Surveys, 22
(1), pp. 31-72.
CIDIN (Centre for International Development Issues Nijmegen) (2014). The
Impact of Coffee Certification on Small holder Farmers in Kenya, Uganda and
Ethiopia. Radboud University Nijmegen, Netherlands.
Coffee Board of Kenya (2011). Annual reports, Government printers, Nairobi.
Coffee Board of Kenya (2013). Found at http//www.coffeeboardkenya.co.ke
(Accessed on 10th
June, 2013).
Cohn, S. and O’Rourke, D. (2011). Agricultural certification as a conservation
tool in Latin America, Journal of Sustainable Forestry, 30: pp. 158-156.
71
Dehejia, R. H. & Wahba, S. (2002). Propensity score matching methods for non-
experimental causal studies. The Review of Economics and Statistics, 84
(1):151-161.
Fair Trade International (2009). Fair trade-Labeling Organizations
International Report Kaiser-Friedrich, Bonn, Germany.
Fair Trade International (2010). Fair-trade Labeling Organizations
International. Found at http//www.fairtrde.net. (Accessed on 10th
January, 2012).
Fort, R., & R. Ruben. (2008). The impact of fair trade on banana producers in
Northern Peru. In R. Ruben (ed.), The impact of Fair Trade. Netherlands:
Wageningen Academic Publishers. Chapter 2.
Foster, E. (2003). Propensity score matching. An illustrative analysis of dose
response. Journal of Medical Care, Lippincott Williams & Wilkins, Inc.
41(10):1183-1192.
Giovannucci, D. (2006). Salient trends in organic standards: The opportunities
and challenges for developing countries. In Standards and trade: Challenges and
opportunities for developing country agro-food trade course. Washington, DC.
Giovannucci, D. & J. Pierrot (2010). Is Coffee the Most Popular Organic Crop?
Global Survey on Organic Agriculture. Bonn, German.
Giovannucci, D. & Potts, J. (2008). The COSA Project: A multi-criteria cost-
benefit analysis of sustainable practices in coffee. International Institute for
Sustainable Development (IISD), Winnipeg, Manitoba.
Giovannucci, D. & Ponte, S. (2005). Standards as a new form of social
contract? Sustainability initiatives in the coffee industry. Food Policy30: 284–
301.
72
Gravetter, F. J., & Forzano, L. B. (2006). Research methods for the behavioral
sciences (2nd edition). Australia: Thomson Wadsworth.
Gujarati, D.N (2004). Essentials of Econometrics.4th
edition. New Delhi
McGraw-Hill Companies.
Heckman, J., H. Ichimura, J. Smith, & Todd, P., (1998). Characterizing
selection bias using experimental data. Econometrica,66: 1017–1098.
International Coffee Organization (2013). Found at http//www.ICO.org
(Accessed on 16th
December, 2015).
International Trade Center (ITC) (2011). The impacts of private standards on
producers in Developing Countries: Technical Paper Doc. No. Mar-11-2-1.E,
Geneva
Jaffee, D. (2007). Brewing Justice: Fair trade coffee, sustainability and survival.
Berkeley. University of California Press,
Kenya Bureau of Standards (KEBS) (2013). Code of Practice: Coffee KS
2366:2013
Kempthorne, O. (1955). The Randomization Theory of Experimental Inference.
Journal of the American Statistical Association, 50(271), 946–967.
http://doi.org/10.2307/2281178 (Assessed on 5thApril, 2016).
Kilian, B., Jones, C., Pratt, L. & Villalobos, A. (2006). Is sustainable
agriculture a viable Strategy to improve farm income in Central America. A case
study on coffee, Vol.59 No3, pp. 322–330.
Kombo K.D. &Tromp A.L.D. (2006). Proposal and Thesis Writing: An
Introduction, Nairobi, Pauline‟s Publications.
73
Kothari, C.R. (2004). Research Methodology: Methods and Techniques: New
Delhi New Age International Publishers.
Melo, C. J. & Wolf, S.A. (2005). Empirical assessment of eco-certification.
Organization and Environment, 18(3), 287–317.
Mercy, W. K. L., Mose, F. Ricardo & Ruben, R. (2010). The Impact of
Certification on small holder Coffee Farmers in Kenya: Th cas of UTZ’
certification program Presentation prepared for the Agricultural Economists
Association of South Africa Conference, Cape Town, South Africa pp. 1-14.
Milford, A. (2004). Coffee, Cooperative and Competition: The Impact of Fair
Trade. Bergen: Michelsen Institute.
Ministry of Agriculture publications (2011). Nairobi, Government Printers.
Mugenda, A.G. &Mugenda, O.M. (2003). Research Methods, Quantitative and
Qualitative Approaches. Nairobi, Acts Press.
Ndumberi coffee society Coffee sales reports (2011). Ministry of Cooperative
Development and Marketing, Nairobi, Government Printers.
Nebel, G., Quevedo, L.,Bredahl Jacobsen J., &Helles F. (2005). „Development
and economic significance of forest certification: the case of FSC in Bolivia‟,
Forest Policy and Economics, vol. 7, no. 2, pp. 175-186.
Potts, J., & Sanctuary M (2010). Sustainable markets are growing –is
sustainability keeping pace. A perspective on sustainable coffee markets, Issue
Brief 03, Stockholm.
Ravallion, M. (2005). Evaluating anti-poverty programs: Policy research
working paper 3625, World Bank, Washington D.C.
Rice, P.D. & Jennifer, M. (1999). Sustainable Coffee at the Crossroads.
Consumer‟s Choice Council: Washington D.C.
74
Rosenbaum, P.R. & Rubin, D.B. (2002). The central role of the propensity score
in observational studies for causal effects. Biometrika, 70(1), pp. 41-55.
Ruerd, R. & Guillermo, Z. (2008). How standards compete: Comparative
impact of coffee certification in Northern Nicaragua. Centre for International
Development Issues, Amsterdam, Radboud University.
Sianesi, B. (2010). "An introduction to matching methods for causal inference
and their implementation in Stata,” Available http:// ideas.respec.org/ (Assessed
on 6th January, 2014).
Stellmacher, T. & Grote, U. (2011). Forest coffee certification in Ethiopia:
Economic Boon or Ecological Bane? Working paper series 76. ISSN 1864-6638,
Bonn Germany
Tina, B., Manfred Z. & Thomas, O. (2009). Justified hopes or utopian
thinking? The suitability of coffee certification schemes as a business model for
small scale producers. International Association of Agricultural Economists
Conference, Beijing, China.
75
APPENDICES
Appendix1: Histogram of Propensity Scores
Figure 4.1: Histogram of propensity scores
0 .2 .4 .6 .8Propensity Score
Untreated Treated: On support
Treated: Off support
76
Appendix 2: Performance of Different Matching Algorithms (Likelihood
Ratio Test)
Matching algorithm Sample Pseudo R2 LRchi2 P > chi2
Nearest Neighbour
NN(1) Matched 0.123 22.28 0.0000
Unmatched 0.194 35.21 0.0000
NN(2) Matched 0.068 12.36 0.0021
Unmatched 0.194 35.21 0.0000
NN(3) Matched 0.087 15.86 0.0004
Unmatched 0.194 35.21 0.0000
NN(4) Matched 0.057 10.38 0.0056
Unmatched 0.194 35.21 0.0000
NN(5) Matched 0.093 16.92 0.0002
Unmatched 0.194 35.21 0.0000
Caliper matching
0.0000
Caliper(0.01) Matched 0.130 23.56 0.0000
Unmatched 0.194 35.21 0.0000
Caliper(0.05) Matched 0.098 17.71 0.0001
Unmatched 0.194 35.21 0.0000
Caliper(0.10) Matched 0.103 18.65 0.0001
Unmatched 0.194 35.21 0.0000
Caliper(0.50) Matched 0.132 24.01 0.0000
Unmatched 0.194 35.21 0.0000
77
Kernel matching
Kernel(0.10) Matched 0.194 12.41 0.0020
Unmatched 0.035 35.21 0.0000
Kernel(0.25) Matched 0.194 6.34 0.0420
Unmatched 0.046 35.21 0.0000
Kernel(0.50) Matched 0.194 8.31 0.0157
Unmatched 0.025 35.21 0.0000
Kernel(0.75) Matched 0.194 4.52 0.1044
Unmatched 0.123 35.21 0.0000
78
Appendix 3: Variables
Variable Description Measurement Levels
q3_socty Society of the interviewee Nominal
1 = Ivinge
2 = Kamurai
3 = Kirindiri
4 = Kithungururu
5 = Muramuki
6 = Rianjagi
q4_fctry Factory of the interviewee String
q5_sex Gender of the respondent Nominal 1 = Male
2 = Female
q6_age Age of the respondent Ordinal
1 = 18-30 years
2 = 31-40 years
3 = 41-50 years
4 = 4 51-60 years
5 = > 60 years
q7_head Gender of household head Nominal 1 = Husband
2 = Wife
q8_mrge Marital status of the head
of the household Nominal
1 = Married
2 = Single
q9_age Age of the household head Ordinal
1 = 18-30 years
2 = 31-40 years
3 = 41-50 years
4 = 4 51-60 years
5 = > 60 years
q10_educ Education level of
household head Ordinal
1 = None
2 = Primary
3 = Secondary
4 = Tertiary
q11_yrs Number of years in coffee
farming Ordinal
1 = < 10 years
2 = 10-20 years
3 = > 20 years
q12_dfct Distance to the factory Continuous
79
q13_size Farm size (acres) Ordinal 1 = <1 Acre
2 = 1-2 Acres
3 = 3-5 Acres
4 = > 5 Acres
q14_acqr Land acquisition Nominal 1 = Inherited
2 = Bought
3 = Leased
q15_tr07 Number of mature coffee
trees 2007
Discrete
q16_tr12 Number of mature coffee
trees 2012
Discrete
Change Change in number of
coffee trees from 2007 to
2012
Discrete
carea_07 Land under coffee in 2007
(acres)
Continuous
carea_12 Land under coffee in 2012
(acres)
Continuous
q17_plcf Planted coffee in the last 5
years Discrete
1 = Yes
2 = No
q18_tres Number of trees planted Discrete
q19_crps Other crops planted Nominal String
q20_prft Other crops are more
profitable than coffee Nominal
1 = Yes
2 = No
q20a_rsn Reasons String
q25_cert Heard of coffee
certification Nominal
1 = Yes
2 = No
q25a_cert Where heard of
certification Nominal
1 = Management of the
coffee society
2 = Neighbor
3 = Media
4 = Government
extension officer
80
q26_cert Farm certified Nominal 1 = Yes
2 = No
q27_std Certification standard Nominal
1 = UTZ certified
2 = Fair-trade
3 = Rainforest
4 = 4c
q28_std Heard of certification String
q29a_cer Reasons for not certifying String
q31_serv Other services Nominal
1 = input
(fertilizer/agrochemicals)
2 = credit
3 = school fees advances
4 = hospital bills
q32_dur Payment duration String
81
Appendix 4: Embu County Map
82
Appendix 5: Small Scale Coffee Farmers Questionnaire
1. Name of the respondent………………………………………
2. Date of Interview………………………………………………
3. Society of the interviewee………………………………………
4. Factory of the interviewee………………………………………
Household characteristic
5. Gender of the respondent
Man=1 Woman =0
6. Age of the respondent
18-30=1 31-40=2 41-50=3 51-60=4 61 and above=5
7. Who is the head of the household?
1=husband 0=wife
8. Marital status of the head of the household?
1=married 0= single
9. Age of the household head
18-30=1 31-40=2 41-50=3 51-60=4 61 and above=5
10. What is the education level of the household head?
No education =1 primary education=2 secondary education =3 tertiary
education =4other (specify) ……………………………………………
11. How many years have you been farming coffee?
1=< 10 years 2=10-20 years 3=20 years and above
83
Accessibility
12. What is the distance to the factory where coffee is delivered?
..............................
13. What is the size of your farm?
< 1 acre =1, 1-2 acres =2 2-5 acres =3 above 5 acres=5
14. How did you acquire the farm?
Inherited =1 Bought =2 leased =3
15. How many matures trees did you have in 2007? ............................
16. How many coffee mature trees do you have now?
…………………………..
17. Have you planted coffee in the last five years?
Yes =1 No=0
18. No of trees planted………………………………………
19. What other five important crops do you grow in your farm?
20. Do you consider the above crops/ livestock more profitable than coffee?
Yes=1 No=0
If no give reasons……………………………..
21. What income did you get from these crops in the last five years?
Crop Size of land in acres
84
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
22. How much coffee (cherry) did you produce in the last 5 years?
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
23. How much were you paid for a kg of Cherry Delivered to the factory in
the last five years?
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
24. What is the income you got from coffee for the last five years?
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
Certification
25. Have you ever heard of coffee certification?
Yes =1 No =0
(a) If yes where did you first hear about certification?
1=management of the coffee society 2=neighbor 3=media
4=government extension officer
26. Has your farm been certified?
Yes =1 No= 0
27. If your farm is certified what certification standard?
UTZ certified =1
85
Fair-trade = 2
Rainforest =3
4c (common code for coffee community) =4
Other specify………………..
28. If no give reasons for not pursuing certification
1=never heard of certification standards
2=certification is expensive
Others reasons specify………………………………………………….
29. Are there any challenges you have experienced from certification?
30. Has certification improved coffee prices?
31. Apart from processing your coffee which other services do you get from
your society?
1= input (fertilizer/agrochemicals)
2=credit
3= school fees advances
4= hospital bills
Other
specify………………………………………………………………………
32. How long from the time you deliver your cherry does it take to be paid?
………………………………………………………………….months/days.
THANK YOU FOR YOUR TIME
86
Appendix 6: Cooperative Society Questionnaire
1. Name of the respondent……………………………………………………………
2. Name of the coffee society………………………………………………………
3. Date of Interview…………………………………………………………………
4. How many registered members do you have.............................................................
Male ……….. Female ……………… Total …………………..
5. How has been coffee production in (Kgs of cherry) for the last 5 years?
6. What was the highest price you received for your coffee in the last 5 years? US $/
50 Kg/ Ksh /Kg
7. What was the lowest price you received for your coffee in the last 5 years? US $/
50 Kg/ Ksh/Kg
8. What were your payouts to farmers in the last 5 years
9. Has your society undergone certification?
Yes= 1 No=0
10. If No to question 10 give reasons
1=Never heard of certification
2= certification is expensive
Other reasons specify……………………………………….
11. If yes which certification standard?
UTZ certified =1
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011
87
Fair-trade = 2
Rainforest =3
Common code for coffee community (4c)=4
Other specify………………..
12. If yes when was your society certified?
> 5 years =1 5 years =2 < 5years =3
13. If yes how much ( Kg of clean coffee) was sold as certified coffee in the last 5
years
14. If certified what is the cost of certification?
15. Has certification improved coffee prices?
16. Has certification improved coffee market?
17. Has certification increased membership?
18. What are other benefits that your cooperative society gained from certification?
19. Are there any challenges associated with certification?
20. What kind of services does the cooperative society provide to its members?
Training on Good Agricultural Practices=1
Input (fertilizer/agrochemicals) =2
Credit =3
School fees advances=4
Hospital bills=5
Other specify……………………………………………………………………….
21. Do you monitor implementation of good Agricultural practices amongst your
members?
Yes =1 No =0
THANK YOU FOR YOUR TIME
2006/2007 2007/2008 2008/2009 2009/2010 2010/2011