Content uploaded by Zewdu Getachew
Author content
All content in this area was uploaded by Zewdu Getachew on Aug 09, 2023
Content may be subject to copyright.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
1
Impact of Membership to Certified Coffee Marketing Cooperatives on the Income of
Smallholder Farmers in Jimma Zone of Oromia Region, Ethiopia
Zewdu Getachew, Fekadu Beyene, Jema Haji and Tesfaye Lemma
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
100
Impact of Membership to Certified Coffee Marketing
Cooperatives on the Income of Smallholder Farmers
in Jimma Zone of Oromia Region, Ethiopia
Zewdu Getachew1*, Fekadu Beyene1, Jema Haji2, Tesfaye
Lemma1
1 Haramaya University, School of Rural Development and
Agricultural Innovation, Ethiopia
2 Haramaya University, School of Agricultural Economics and
Agri-Business, Ethiopia
* Corresponding Author’s Email: zewdug2001@gmail.com.
Article History
Received 5th July 2023
Received in Revised Form 17th July 2023
Accepted 26th July 2023
How to cite in APA format:
Getachew, Z., Beyene, F., Haji, J., & Lemma, T. (2023). Impact
of Membership to Certified Coffee Marketing Cooperatives on
the Income of Smallholder Farmers in Jimma Zone of Oromia
Region, Ethiopia. International Journal of Economics, 8(1),
100–126. https://doi.org/10.47604/ijecon.2047
Abstract
Purpose: Coffee is one of the most important agricultural
commodities with a significant contribution to the growth and
well-functioning of Ethiopia’s economy, and to the livelihoods of
millions of smallholder farmers and laborers. Despite its
importance, smallholder coffee production and marketing
performance have been unsatisfactory due to various reasons. The
introduction of voluntary coffee certification schemes such as
Fairtrade (FT) and Organic (Org) certification schemes through
cooperatives are viewed as mechanisms to overcome the
constraints smallholder farmers face in accessing high value
coffee markets and earn better income. However, the impacts of
these schemes on the livelihoods of smallholder farmers were not
analyzed yet. The main purpose of this study was to estimate the
impact of joining (FT, Org or dual FT-Org) certified coffee
cooperatives on gross annual incomes earned by member farmers.
Methodology: The study employed cross-sectional data collected
from randomly selected sample smallholder coffee farmers
through a semi-structured questionnaire. Descriptive and simple
inferential statistical tests (e.g., frequency, percentage, mean, t-
and chi2-tests), and PSM methods were employed to analyze the
data.
Findings: Results of the descriptive statistics depict that 234
(62.07%) of the total 377 samples farmers were members of
certified coffee marketing cooperatives. Among the cooperative
members, 83 (35.47%), 84 (35.90%) and 67 (28.63%) were
members of FT, Org and dual FT-Org certified coffee marketing
cooperatives, respectively. The results of the binary probit model
however show that the decisions to join certified coffee marketing
cooperatives was significantly influenced by sex, marital status,
total livestock holding size, total coffee land size (ha), log total
quantity of coffee produced (kg), credit access, and walking
distances to development agent’s office, coffee marketing center
and all-weather road in minutes, respectively. The PSM analysis
results show that membership to certified coffee marketing
cooperatives has a positive and significant impact on average
annual gross income (ETB) earned. The average gross annual
income earned by coop member farmers was ETB 14639.15,
which is by 36.51% higher than their counterpart non-coop
member farmers. The difference is statistically significant at 1%
probability level.
Unique Contribution to Theory, Practice and Policy: The
study recommended that Cooperatives should be encouraged to
establish credit and saving units in their internal structure and/or
work in collaboration with other saving and credit providing
institutions (such as Cooperative Bank of Oromia) to be able to
provide demand-driven credit services to member farmers
Keywords: Smallholder Farmers, Certified Coffee Marketing
Cooperative, Gross Annual Income, Propensity Score Matching
Method, Ethiopia
©2023 by the Authors. This Article is an open access article
distributed under the terms and conditions of the Creative
Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/)
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
101
INTRODUCTION
Coffee is one of the most important agricultural commodities with a significant contribution to
the growth and well-functioning of the economy, and the social stability of Ethiopia
(Alemseged, 2012). It also serves as the main source of income to tens of millions of small-
scale coffee farming populations, workers and traders (Alemseged, 2012; Abu and Tefera,
2013). While millions are dependent on the coffee sector (farming, picking, transportation, etc.),
coffee also is the principal export commodity of the country (Tadesse and Yalem, 2014). It for
example accounted for 47% of the total agricultural export value and 34% of total commodity
export value of the country in 2017/18 (GAIN, 2019). The crop contributes approximately 10%
of the total GDP of the country (Hirose, 2014).
The coffee sub-sector however is dominated by smallholder farmers, supplying a significant
share (more than 90%) of the country’s total coffee supply, while the remaining is contributed
by plantation (commercial) coffee farmers (Petty et al., 2004; Babur, 2009; Stellmacher and
Grote, 2011). Despite their importance, smallholder farmers face major problems in production,
supply and marketing of coffee. The domestic coffee marketing system is not fair and efficient
which has problems in product assembling, storing, handling, processing, quality inspection
and grading, and in having fair and transparent trading system (Bastin and Matteucci, 2007;
ECEA, 2008; Minten et al., 2015).
In the early 2000s, due to the interplay between increasing poverty of coffee smallholders in
major producer countries and growing demands for healthier and more socially and
environmentally-friendly produced coffee in larger consumer countries, certification of
cooperatives has gradually gained wider significance worldwide (Petit, 2007; Stellmacher and
Grote, 2011). Moreover, certification schemes are expected to significantly contribute to
production of healthy and traceable coffee to consumers and improving the livelihoods and
welfare of smallholder coffee farmers by enhancing their incomes through premium prices and
stabilizing it through minimum prices (Stellmacher et al., 2010; Stellmacher and Grote, 2011;
Ferris et al., 2014; Minten et al., 2015; Fikadu et al., 2017).
Despite the introduction and expansion of different smallholder and cooperative-based product
certification schemes in the coffee sector of Ethiopia nearly two decades ago, there are not
conclusive evidences on the impacts of cooperative-based coffee certification schemes on
member farmers in Ethiopia in particular (Stellmacher et al., 2010; Jena et al., 2012; Amsaya
et al., 2015; Amsaya, 2015; Minten et al., 2015; Fikadu et al., 2017). Specifically, there exist
little and conclusive empirical evidences on the impacts of certified (e.g., Organic, Fairtrade or
dual Organic-Fairtrade) certified coffee marketing cooperatives on member farmers. Many of
the empirical studies carried out to investigate impacts of joining certified coffee marketing
cooperatives on the livelihoods and incomes of smallholder farmers came up with mixed and
inconsistent results which differ depending on the specific contexts (Tium, 2013; Jena et al.,
2012; Jena et al., 2015; Jena et al., 2015; Amsaya, 2015; Fikadu et al., 2017).
Against this backdrop, this research paper aims to analyze factors influencing the decisions by
smallholder farmers to join (Fairtrade, Organic or dual Fairtrade-Organic) certified coffee
marketing cooperatives and the impact of membership on gross annual income earned by the
member farmers in Jimma zone of Oromia region, Ethiopia.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
102
MATERIALS AND METHODS
Description of the Study Area
Ethiopia is a federal country divided into 11 regional states and 2 city administrations. Each
region is subdivided into zones, and the zones into woredas (districts), which are roughly
equivalent to a county in the United States or UK. Woredas, in turn, are divided into peasant
associations (PAs), or kebeles, an administrative unit consisting of a number of villages
(Dercon et al., 2008).This study was conducted in Gomma and Limmu Kossa districts of Jimma
zone, Oromia National Regional State of the country (see Figure 1). The zone is located in the
southwestern part of the country about 360km away from Addis Ababa, the capital city of the
country. The zone extends between 7013’- 8o56’ North latitudes and 35049’-38038’ east
longitudes. Jimma zone is one of the major coffee producing areas with about 105,140 hectares
of land covered with coffee, which includes small-scale farmers’ holdings as well as state and
privately owned plantations (Berhanu et al., 2015; Dagne et al., 2015). About 30-45% of the
people in Jimma Zone directly or indirectly benefit from the coffee industry (Anwar, 2010).
However, coffee is widely produced in eight districts, namely, Gomma, Manna, Gera, Limmu
Kossa, Limmu Seka, Seka Chokorsa, Kersa and Dedo districts. Gomma and Limmu Kossa
districts are among top coffee producing districts in Jimma zone. The majority of smallholder
farmers in the districts are engaged in engaged in coffee production and marketing as their
main livelihood acclivity and means of cash income (BoCPA, 2018).
Figure 1: Map of the Study Area
Source: Tibebu Kasawmar, Staff of Addis Ababa University (2021)
In Jimma zone, there were 516 farmers’ associations and 288 agricultural service cooperatives
with 125468 male and 3307 female farmers, making up a total of 128775 member farmers in
2010 (BoFED, 2011). Based on the information obtained through personal contacts from
Jimma zone offices of agriculture and cooperatives’ promotion agency in 2018, there were
about 60, 472 smallholder coffee farmers in Gomma and Limmu Kossa districts in 2017/18
coffee season. Of the total 60, 472 farmers, 18251 (30.18%) were members of primary
cooperatives registered for Fairtrade, Organic or dual Fairtrade-Organic certified coffee
marketing (see Table 1).
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
103
Table 1: Selected Districts, Coffee Certification Schemes, Cooperatives, Kebeles and
Samples of the Study
District
Cert
scheme
Coop
Kebeles
with
access
to cert
coop
Population and sample sizes
by cooperative membership
status
Combined size
Members
Non-
members
Pop
Sample
Pop
Sample
Pop
Sam
ple
Gomma
Fairtrade
cert
Choche
Gudda
Bulbulo
146
5
705
49
851
39
Choche
Lemi
588
20
313
22
901
41
Ilbu
Ilbu
469
16
592
41
1061
48
Omo
Beko
Omo
Beko
878
29
281
20
1159
52
Omo
Guride
685
23
645
45
1330
60
Subtotal
3
5
2766
93
2536
177
5302
240
Limmu
Kossa
Organic
cert
Tencho
Tencho
414
14
30
8
444
20
Shogole
Gena
Denbi
257
8
155
39
412
19
Chime
Chime
291
10
39
10
330
15
Mito
Gundub
Mito
Gundub
394
13
27
7
421
19
Chefe
Ilfeta
Chefe
Ilfeta
414
14
96
24
510
23
Subtotal
5
5
1770
59
347
24
2117
96
Dual
Org-FT
cert
Babu
Babu
465
16
28
2
493
22
Debello
Debello
368
12
36
3
404
18
Harewa
Jimate
Harewa
Jimate
304
10
41
3
345
16
Kacho
Tirtira
Kacho
Tirtira
620
21
50
3
670
31
Subtotal
4
4
1757
5 9
155
11
1912
87
Total
12
14
6293
211
3038
212
9331
423
Source: Cooperative promotion agencies of Jimma zone, Gomma and Limmu Kossa districts
(2019)
Sampling Method and Data
The question of how large a sample to take arises early in the planning of any survey or
experiment. This is an important question that should not be treated lightly. Sample size
determination however can be done based on whether the samples are needed for estimating
population mean or percentage or proportion under study (Kothari, 2004; Daniel and Cross,
2013). Since the aim of this study is to estimate the impact of cooperatives on gross annual
income earned by the proportion of member farmers, the initial sample size (n0) was determined
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
104
by using the formula for estimating the population proportion or percentage as given by the
above-mentioned authors as follows:
(1)
n0 = 384
Where, n0 is the initial desired sample size; Z = 1.96 for a 95% desired confidence level, p is
the estimated population percentage or proportion (which is usually set at 0.5 for a population
proportion which is unknown a priori), q = 1-q = 1-0.50 = 0.50, and d = ±0.05 for a 95%
desired precision level, respectively.
Then, a total of 423 sample farmers including 10% more samples in compensation for possible
drop out of respondents or missing or incomplete survey questionnaires were selected from 14
kebeles covered by certified coffee marketing cooperatives for conducting the sample survey.
Finally, a multistage stage sampling technique was employed to select 211 cooperative member
and 212 non-cooperative member sample farmers based on probability proportional to the size
of the respective coop member and non-coop member farmers in the total population in the two
districts.
Methods of Data Analysis
Impact Evaluation
The outcome variables (impact indicators) are those that can express the effect of the treatment
(cooperative membership in this case). Several approaches exist on how to evaluate the impacts
of a treatment on the performance of an individual or an organization. The common practices,
among these, are either quantitatively measuring the output or directly asking the performance
levels based on different scales. Prior to taking a measurement on the variable of interest, it is
important to determine the indicator that captures the impact under investigation (Dagne et al.,
2015). Empirical studies however used agricultural yield and productivity level achieved, sales
volume and prices received, and crop-specific and gross-annual incomes earned by households
as indicators of economic impacts of cooperatives on member farmers (Zekarias and D’Haese,
2016; Fikadu et al., 2017; Fikadu et al., 2020). In this study, as such, gross annual income
earned (ETB) is used as an indicator of economic impact of of membership to certified coffee
marketing cooperatives on member farmers in 2017/18 coffee season as it is one of the most
commonly used proxy variables to analyze poverty and welfare impacts of certification
schemes (Jena et al., 2012; Tium, 2013; Fikadu et al., 2020).
Econometric Analysis of Factors Influencing the Decisions to Join Cooperatives
The decisions to join (Fairtrade, Organic or dual Fairtrade-Organic) certified coffee marketing
cooperative can be modeled using the random utility framework (Berhanu, 2012; Degnet and
Mekbib, 2013). According to this framework, the actual utility level gained from membership
to a certified coffee marketing cooperative by the member household is unknown. However,
the household chooses to be member of a cooperative if the utility gained from membership
(Uim) is larger than the utility of non-membership (Uin). The utility gain, (Uim - Uin) of
cooperative membership can then be expressed as a function of observed characteristics (Z) in
the latent variable model as follows:
iii
C
*
(2)
2
2
0d
pqz
n
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
105
Where C*i is an indicator of the latent cooperative membership and εi is the disturbance term.
In turn, the observed dependent variable, which depicts cooperative membership status (Ci),
where Ci=1 for member of a cooperative and Ci = 0 for non-member of a cooperative, is
related to C*i as follows:
(3)
The choices of the explanatory variables included in Z is guided by relevant economic theories
and previous empirical studies on factors influencing decisions to join agricultural cooperatives
by smallholder farmers in developing countries that include Ethiopia (Wollni and Zeller, 2007;
Bernard et al., 2008a; Bernard and Spielman, 2009; Francesconi and Heerink, 2011;
Fischer and Qaim, 2012; Degnet and Mekbib, 2013). The definitions and hypothesized
influences of the socio-economic, institutional and infrastructure related explanatory variables
on the decisions to join certified coffee marketing cooperatives thus are as given in Table 2.
Table 2: Econometric Model Variables Hypothesized to Influence Decisions to Join
Cooperatives
Variable name
Description
Hypothesized
influence
Dependent variable: Coop_(1= Yes; 0
otherwise)
Membership status of a household to (Fairtrade, Organic or
dual Fairtrade and Organic) certified coffee marketing
cooperative (1= Yes; 0 otherwise)
Independent variable:
Sex (Sex)
Sex of household head (1 = male, 0= female)
-
Age (Age)
Age of the household head in years.
+
Marital status (Marstat)
Marital status of the household head (1= Married, 0 = Single,
divorced or Separated).
-
Literacy level (Literacy)
Literacy level of the household head (1 Literate= reads and/or
writes, 0 Illiterate= cannot read and/or write at all)
+
Family size (FamS_ME)
Total number of family members (in adult equivalent).
+
Total livestock holding size
Total livestock holding size in tropical livestock unit (TLU).
-
Total land size under coffee production
(LANDCP)
Land size land under coffee production in ha.
+
Coffee farming experience
(CFarming_EXP)
Full years of experience in coffee farming.
+
Access to extension services
(Access_extension)
It refers to whether the farmer had access to extension services
(1= Yes, 0 otherwise).
+
Social capital (SCapital)
Membership to traditional rural organizations (e.g., idir1or ekub
(rotating saving and credit association) (1= Yes, 0 otherwise).
+
Access to credit (Access_Credit)
It refers to whether the household head had access to credit
services (1= Yes, 0 otherwise)
+
Off-farm income sources
(Off_farm_Income)
Participation in off-farm income earning activities (such as
serving as daily laborers on others farms) (1= Yes, 0 otherwise).
-
Non-farm income sources
(Nonfrm_Income)
Participation in non-farm income earning activities (such as
carpentry and non-farm labour markets) (1= Yes, 0 otherwise).
-
Distance to development agents’ offices
(DISTDAWM)
Walking distance traveled to reach to development agents’
offices in minutes.
+
Distance to coffee marketing centers
(DISTTRCMC)
Walking distance traveled to reach to coffee marketing centers
in minutes .
-
Distance to cooperative’s office
(DISTCOOPWM)
Walking distance traveled to reach to cooperative’s office in
minutes.
+
Distance to all-weather road
(DISTTRAWR)
Walking distance traveled to reach to the nearest all-weather
road in minutes.
-
1
Is a traditional association which provides insurance for members during death and other accidents (Degnet and
Mekbib, 2013).
0> if 1 otherwise 0
{i
C
i
C
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
106
Estimation of Impact of Cooperatives Using the PSM Method
To test if there was significant difference in the means of the annual gross incomes earned per
household between members and non-members of certified coffee marketing cooperatives the
propensity score matching (PSM) method was used. According to Caliendo and Kopeinig
(2008), there are steps in implementing PSM. These are estimation of the propensity scores,
choosing a matching algorithm, checking on common support condition, testing the matching
quality and sensitivity analysis. In what follows, we discuss each step one by one. The first step
in the implementation of the PSM method is to estimate the predicted probability (propensity
score) that a household would be member of a certified coffee marketing cooperative,
conditional on observed covariates of the household using binary logit model (Caliendo and
Kopeinig, 2008). The propensity score then was estimated as:
P (Zi) = Prob (Zi = 1|Xi) (5)
where P is the propensity score, Zi is the ith household, the P (Zi) the propensity score of the ith
household and Prob (Zi = 1|Xi) is the probability of the ith household to join certified coffee
marketing cooperative membership conditional on observed personal, household, farm and
location characteristics, Xi. These household level factors influencing the decision of a
household to join a certified coffee marketing cooperative were identified based on relevant
economic theories and previous empirical studies (see Table 2).
Once the propensity scores to join certified coffee marketing cooperatives are estimated, the
next step is to match the propensity scores of the treatment group with that of the non-treatment
or control group to identify households from both groups with similar propensity scores using
appropriate matching estimators. In this step thus each member household of a certified coffee
marketing cooperative was matched with that of the non-cooperative member household with
similar propensity score values, in order to estimate the average treatment effect for the treated
(ATT). Though various matching methods exist, the nearest neighbor matching (NNM),
caliper or radius matching (CM) and kernel-based matching (KBM) methods are the most
widely used matching methods (Caliendo and Kopeinig, 2008). However, the NNM and KBM
methods are the most commonly used matching methods as they enable to ensure that members
are matched with the non- members over a common region of the matching variables. Any
remaining bias in the matching estimator can thus be attributed to unobserved characteristics
(Jalan and Ravallion, 2003).
Checking overlap and finding common support region between the treatment and control
groups is the third step in PSM matching (Bryson et al., 2002). The common support region is
the area which contains the minimum and maximum propensity scores of treatment and control
group households, respectively. Comparing the incomparable must be avoided, i.e., only the
subset of the comparison group that is comparable to the treatment group should be used in the
analysis (Caliendo and Kopeinig, 2008). No matches can be made to estimate the average
treatment effects on the ATT parameter when there is no overlap between the treatment and
non-treatment groups. In this study, the KBM method was used to pair cooperative members
to similar non-members using the estimated propensity scores. The data was analyzed using
alternative matching estimators to check the robustness of the results (Degnet and Mekbib,
2013).
In the fourth step, the average treatment effect on the treated (ATT), which in our case is the
impact of membership to certified coffee marketing cooperatives on member farmers, is
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
107
estimated. Following Becker and Ichino (2002), the average treatment effect on the treated
(ATT) was estimated as follows:
ATT = E (Y1 -Y0|X, M=1) = E (Y1|X, M = 1) – E (Y0|X, M = 1) (6)
Where ATT denotes the effect of certified coffee marketing cooperative on average gross
annual income of smallholder member farmers, Y1 and Y0 denote outcomes of members and
non-members in certified coffee marketing cooperative, respectively, X is a vector of observed
characteristics of the a sample farmer that may influence his/her decision to join certified coffee
marketing cooperative and/or the expected outcome of membership or non-membership to such
a cooperative. The Xs are used as explanatory variables, and M denotes cooperative
membership decision (M = 1, if a farmer joined the cooperative or = 0, otherwise).
There is however fundamental problem when estimating ATT, given Equation (6), that it is
impossible to observe a person’s outcome for with and without the treatment at the same time.
While it is possible to observe the post-intervention outcome, E (Y0|X, M =1), however, the
counterfactual outcome of the ith household when she/he does not get the treatment is not
observable in the data. A solution to this problem is to construct the unobserved outcome which
is called the counterfactual outcome that households would have experienced, on average, had
they not joined the cooperatives (Rosenbaum and Rubin, 1983), and this is the central idea of
matching.
The counterfactual outcome, E (Y0|X, M= 0), then is constructed by replacing the unobserved
outcome value (missing value) of a cooperative member farmer, E (Y0|X, M= 1), with the
expected outcome value (observed outcome value) of the matched non-cooperative member
farmer who had similar observable characteristics with the cooperative member farmers.
Therefore, Equation (6) can be re-written as:
ATT = E (Y1 -Y0|X, M=1) = E (Y1|X, M=1) – E (Y0|X, M = 0) (7)
The conditional average effect of treatment on the treated however has a problem, if the number
of the set of conditioning variables (X’s) is high, and thus the degree of complexity for finding
identical households both from members and non-members of certified coffee marketing
cooperative becomes difficult to reduce the dimensionality problem in computing the
conditional expectation, Rosenbaum and Rubin (1983) showed that instead of matching on the
base of X’s one can equivalently match treated and control units on the basis of the “propensity
score” defined as the conditional probability of receiving the treatment given the values of X’s.
Therefore, Equation (7) was used to estimate the effect of membership to certified coffee
marketing cooperative on gross annual average income of cooperative member farmers.
To use PSM for estimating ATT, however, two important assumptions must be satisfied. The
effectiveness of matching estimators as a feasible estimator for impact evaluation however
depends on two fundamental assumptions (Rosenbaum and Rubin, 1983). These assumptions
are the conditional Independence Assumption (CIA) and the assumption of common support
condition. The CIA imposes a restriction that choosing to join a cooperative is purely random
for similar individuals. As a consequence, this assumption excludes the familiar dependence
between outcomes and membership to a cooperative that lead to a self selection problem
(Heckman et al., 1998). The conditional average effect of treatment on the treated has a
problem, if the number of the set of conditioning variables (X’s) is high, and thus the degree
of complexity for finding identical households both from members and non-members of
certified coffee marketing cooperative becomes difficult. To reduce the dimensionality
problem in computing the conditional expectation, Rosenbaum and Rubin (1983) showed that
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
108
instead of matching on the base of X’s one can equivalently match treated and control units on
the basis of the “propensity score” defined as the conditional probability of receiving the
treatment given the values of X’s.
The assumption of the common support region implies that the P(x) lies between 0 and 1, where
P(x) denotes the propensity scores of both members and non-members of certified coffee
marketing cooperatives in our case. This restriction implies that the test of the balancing
property is performed only on the observations whose propensity score belongs to the common
support region of the propensity scores of treated and control groups (Becker and Ichino, 2002).
Individuals that fall outside the common support region would be excluded in the treatment
effect estimation. This is an important condition to guarantee improving the quality of the
matching used to estimate the ATT. Moreover, implementing the common support condition
ensures that a person with the same X values (explanatory variables) has a positive probability
of being both member and non-member of a certified coffee marketing cooperative (Heckman
et al., 1999). This implies that a match may not be found for every individual sample.
Bootstrap standard errors were used to test the statistical significance of the estimated ATT in
order to account for the variation caused as a result of the matching process. Finally, the
robustness of the evaluation results was tested for their sensitivity for the hidden variables that
may affect cooperative membership decision of households.
RESULTS AND DISCUSSION
Descriptive Statistics
Table 3 presents descriptive statistics of categorical pre-treatment socio-economic
characteristics of the sample farmers. Results show that 234 (62.07%) of the total 377 samples
farmers were members of certified coffee marketing cooperatives. Amongst the cooperative
members, 83 (35.47%), 84 (35.90%) and 67 (28.63%) were members of Fairtrade, Organic and
dual Fairtrade and Organic certified coffee marketing cooperatives, respectively. Regarding
marital status, 340 (90.19 %) were married household heads. In terms of literacy level, 319
(84.61%) were literate (i.e., they can read and/or write). On the hand only 75 (19.88%) of the
total sample farmers had social capital or networks (i.e., member of idir, senbete or ekub, etc.).
With regard to access to extension services, 260 (68.96%) of the total sample farmers reported
they had access to extension services related to coffee production, farm management and post-
harvest handling practices. Development agents and cooperative officials were reported as the
main sources of the extension services. On the other hand, 329 (87.27%) of the total sample
farmers responded that they participated in various training programs related to coffee
production and marketing activities. The results on income earning sources indicate that on-
farm income was the sole source of income to all of the sample farmers. Only 232 (61.54%)
and 48 (12.73%) of the sample farmers earned incomes from off-farm and non-farm income
sources, respectively. However, the proportion of coop member farmers (15.81%) who earned
income from the non-farm income sources is significantly greater than the proportion of non-
coop member farmers (7.69%) who earned income from the non-income sources. The
difference is significant at 10% probability level.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
109
Table 3: Descriptive Statistics of Categorical Pre-Treatment Characteristics of the
Sample Farmers
Pre-treatment variable
Pooled sample
(N=377)
Cooperative membership status
Chi2
test
Non-members
(N=143)
Members
(N=234)
Freq.
%
Freq.
%
Freq.
%
Coop membership status
(Yes if a Member)
No
143
37.93
114
43.02
29
25.89
0.002**
Yes
234
62.07
151
56.98
83
74.11
Total
377
100.00
143
100.00
234
100.00
Sex (Yes if Male)
No
21
5.57
8
5.59
13
5.56
0.987
Yes
356
94.43
135
94.41
221
94.44
Total
377
100.00
143
100.00
234
100.00
Religion (Yes if Muslim)
No
63
16.71
29
20.28
34
14.53
0.147
Yes
314
83.29
114
79.72
200
85.47
Total
377
100.00
143
100.00
234
100.00
Marital status (Yes if
Married)
No
36
9.55
18
15.59
18
7.69
0.117
Yes
341
90.45
125
87.41
216
92.31
Total
377
100.00
143
100.00
234
100.00
Literacy level (Yes if
Literate)
No
58
15.39
21
14.69
37
15.81
0.769
Yes
319
84.61
122
52.31
197
84.19
Total
377
100.00
143
100.00
234
100.00
Social capital (Yes if
Member of idir or coop)
No
302
80.12
124
86.71
178
76.07
0.012**
Yes
75
19.88
19
13.19
56
23.93
Total
377
100.00
143
100.00
234
100.00
Contact with DA
(Yes if the sample had
contact)
No
34
9.02
11
7.69
23
9.83
0.482
Yes
343
90.98
132
92.31
211
90.17
Total
377
100.00
143
100.00
234
100.00
Access to extension
services (Yes = if the
sample farmer had access
to extension)
No
17
31.04
7
4.90
10
4.27
0.077*
Yes
260
68.96
136
95.10
224
95.73
Total
377
100.00
143
100.00
234
100.00
Access to training
(Yes = if the sample
farmer had access to
training)
No
48
12.73
21
14.69
27
45.09
0.374
Yes
329
87.27
122
85.31
207
54.91
Total
377
100.00
143
100.00
234
100.00
Off-farm income sources
(Yes = 1)
No
145
38.46
57
39.86
88
37.61
0.413
Yes
232
61.54
86
60.14
146
62.39
Total
377
100.00
143
100.00
234
100.00
Non-farm income
sources
(Yes = 1)
No
329
87.27
132
92.31
197
84.19
0.097*
Yes
48
12.73
11
7.69
37
15.81
Total
377
100.00
143
100.00
234
100.00
Source: Own Survey Data (2019).; Key: ***, **, And * Refer To Probability at 1%, 5% and
10% Probability Level, Respectively.
Table 4 below presents the results of descriptive statistics of the relevant continuous
socioeconomic and other proximity to offices and infrastructure related characteristics of the
sample farmers. As shown in table 4, the age of the total sample respondents ranged from 19
to 100 years with a mean age of 43.17 years. The average family size was 3.63 in adult
equivalent. On average, a sample farmer had 4.77 livestock holding size in tropical livestock
unit and 2.18 ha of agricultural land, respectively. Cereals, flowering and oil crops, coffee, khat
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
110
(Katha edulis), vegetables and sugar cane were the main crops cultivated by the sample farmers.
However, fruit crops such as orange, mango, avocado, papaya and sugar apple) were also
grown on mini plots of land. Regarding coffee production, the sample farmers on average had
1.15 ha of land under coffee production in 2017/18 G.C. coffee season, with the average 18.86
years of coffee farming experience. However, a sample farmer needed to walk on average for
29.40, 36.15 and 55.14 minutes to reach to his/her nearest coffee plots, development agent (DA)
office and nearest coffee marketing center, respectively. There is significant difference in the
means of the walking distances traveled in minutes to reach to the DA’s office between coop
member and non-coop member households at 5% probability level. The implication is that the
non-cooperative member farmers needed to walk for longer time than their counter but
cooperative member farmers needed to reach to DA’s office.
On the other hand, the sample farmers on average needed to walk for 36.32 minutes to reach
to all-weather road. However, there is significant difference in the walking time needed to reach
to all-weather road between non-coop member and coop member farmers at 5% probability
level. The implication is that coop member farmers are relatively closer to all-weather roads
than the non-coop member farmers. The results on income earned indicate (in Table 3) that
the sample farmers earned incomes from three income sources, viz., on-farm, off-farm and non-
farm income sources), respectively. They earned on average ETB 13253.20, 6784.79 and
5440.10 gross annual income from on-farm, off-farm and non-farm income earning sources,
respectively. The total average gross annual income earned by the sample farmers was ETB
41745.64. However, the gross annual incomes earned by the coop member and non-coop
member farmers were ETB 48358.36 and 30924.81, respectively. But there is no significant
difference between the average gross annual incomes earned between the two groups.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
111
Table 4: Descriptive Statistics of Continuous Pre-Treatment Characteristics of the
Sample Farmers
Pre-
treatment
variable
Combined
(N=377)
Cooperative membership status
Combined
difference in
t-test
Non-members
(N=143)
Members (N=234)
Mean
SD
Mean
SD
Mean
SD
Mean
SDa
Age in years
43.17
11.19
42.44
12.39
43.62
10.39
-1.179
1.24
0.84
Coffee
farming
experience
in years
18.86
8.615
18.62
9.84
19.00
7.79
-0.382
0.97
0.66
Family size
in adult
equivalent
3.60
1.60
3.50
1.60
3.67
1.61
-0.170
0.17
0.84
Livestock
holding size
in TLU
4.77
2.42
4.68
2.37
4.82
2.46
-0.144
0.27
0.70
Total land
size owned,
hectare
2.18
2.016
1.71
1.37
2.46
2.28
-0.759
0.19
0.19
Coffee land
size, hectare
1.15
1.24
0.93
1.145
1.286
1.274
1.149
0.13
1.00
Walking
distance to
coffee plot
in minutes
29.40
25.79
31.18
28.04
28.31
25.79
2.864
2.25
0.148
Walking
distance to
DA’s office
in minutes
36.15
26.43
39.48
28.19
34.11
25.14
5.368
2.32
0.028**
Walking
distance to
coffee
marketing
center in
minutes
55.14
38.93
55.79
38.68
54.74
39.16
1.054
3.62
0.400
Walking
distance to
all-weather
road in
minutes
36.32
35.32
41.11
35.89
33.37
34.71
7.736
3.21
0.020**
On-farm
income
earned in
ETB
13253.
20
11823.57
9435.93
6611.61
15585.97
13584.76
-6150.04
1255.91
1.000
Off-farm
income
earned in
ETB
6784.7
9
5055.365
5677.65
4226.52
7380.934
5377.46
-1703.28
797.208
0.972
Non-farm
income
earned in
ETB
5440.1
0
6261.213
2377.14
1196.04
6599.054
6993.055
-4221.91
1625.85
0.985
Total
income
earned in
ETB
41745.
64
34987.41
30924.81
25857.88
48358.36
38106.86
-
17433.55
3522.98
1.000
Source: Own Survey Data (2019).
***, ** and * Stand For Probability at the 1%, 5% and 10% Levels, Respectively.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
112
aSTD for mean difference =
Econometric Analyses
Econometric Model Results of the Determinants of the Decisions to Join Cooperatives
As shown in Table 5, the binary probit model was used for identifying factors influencing
decisions to join cooperatives. The model sufficiently fitted the data as the Wald chi-square
(LR χ2 (14) = 47.83) is significant at 1% probability level indicating that the null hypothesis of
no explanatory power of the model was strongly rejected. The pseudo-R2 is 0.108, which is
moderately low, indicating that there was no systematic difference in the distribution of
covariates between cooperative member and non-cooperative member farmers. While the
coefficients of the covariates are used to report their relationships on the decisions to join
certified coffee marketing cooperatives, their P>|z| values are used for explaining the
probability levels influences on the decisions to join certified coffee marketing cooperatives.
The marginal effects of the dummy variables after probit model estimation are used to report
their effects on the decisions to join cooperatives. However, the coefficients of the continuous
variables are used to explain their respective effects on the decisions to join the cooperatives.
The results in Table 6 show that the decisions to join (FT, Org or dual FT-Org) certified coffee
marketing cooperatives were significantly influenced by sex, marital status, total livestock
holding, total land size under coffee production (ha), log total quantity of coffee produced in
kg, access to credit, and walking distances to DA’s office and nearby coffee marketing center
and all-weather road in walking minutes, respectively. Sex of the household head had a
negative and statistically significant relationship with the decision to join cooperative at 10%
probability level. The finding of this study is consistent with our hypothesis. The finding of
this study is consistent with the findings of previous empirical evidences (e,g., Bernard et al.,
2008; Dagne et al., 2017 and Fikadu et al., 2017) in which gender of the household head
negatively and significantly influenced the decisions to join agricultural cooperatives by rural
households in Ethiopia.
Marital status of the household head (i.e., being married) had a positive and significant
relationship with the decision to join certified coffee marketing cooperative at 5% probability
level. The finding is against the hypothesis but consistent with previous empirical studies.
Dagne et al. (2015) for example found that marital status had a negative and significant
influence on the decisions to join agricultural cooperatives among rural households in Ethiopia.
The empirical evidences are plausible since married households expected to have more access
to information regarding cooperatives owing to their better social capital than the unmarried
household heads. Thus, married household heads are more likely to join cooperatives than the
unmarried household heads. The finding on social capital also depicts that farm households
having a social capital (who were members of idir or ekub) were more likely to join certified
coffee marketing cooperatives than the household heads with no such social capital.The total
livestock holding size (as measured in tropical livestock unit, TLU) negatively and statistically
significantly influenced the decision to join cooperative at 5% probability level. This finding
agrees with its hypothesized influence and empirical evidences (Fikadu et al., 2017; Manda et
al., 2020). Fikadu et al. (2017) find that livestock holding in TLU had a negative and significant
influence on the decisions to join certified coffee marketing cooperatives at 1% probability
level. Similarly, Manda et al. (2020) find that livestock ownership positively and significantly
influenced the decision to join agricultural cooperatives at 5% probability level.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
113
On the other hand, land size under coffee production (in ha) had a positive and statistically
significant relationship with the decision to join cooperatives at 1% probability level. The
finding is in agreement with the hypothesize. Previous empirical evidences also support this
finding (Bernard et al., 2008; Bernard and Spielman, 2009; Zekarias and D’Haese, 2016).
Bernard et al. (2008) and Bernard and Spielman (2009) indicate that landholding size had a
positive and significant influence on the probability of participation in agricultural cooperatives
among smallholder farmers in Ethiopia. However, they report that the squared landholding size
rather had a negative and significant influence on the decisions to join cooperatives, reflecting
what they call the middle class effect of landholding size on the decision to join cooperatives.
According to the authors, though the probability of cooperative participation increases for each
additional hectare of landholding, its marginal effect on cooperative membership decision
decreases with the amount of land after a maximum is reached.
The result on credit access indicates that credit access negatively and statistically significantly
influenced the decision to join cooperatives at 10% probability level. The finding is against the
hypothesized influence and previous empirical evidences (Manda et al., 2020). Manda et al.
(2020) shows that access to credit influenced positively and significantly the decision to join
agricultural cooperatives in Zambia at 5% probability level. Against these findings, Jena et al.
(2015) find that access to credit from cooperatives positively and significantly influenced the
decisions to join cooperatives registered for FT and Org certified coffee marketing in Jinotega,
Nicaragua. The walking distance traveled to reach to DA’s office negatively and significantly
influenced the decisions to join certified coffee marketing cooperatives at 5% probability level.
Distance to DA’s office has a direct implication on not only access but also frequency of access
to information by smallholder farmers as the DA is the main source of such information to the
farmers. Thus, farmers farther away from DAs offices are less likely to have sufficient
information about farmers’ organizations and so do they less likely join cooperatives.
On the other hand, the distance traveled to reach to the nearest coffee marketing center had a
positive and significant effect on the decision to join certified coffee marketing cooperatives at
5% probability level (see Table 5 above). This finding is in agreement with the hypothesized
influence of the variable. The finding is plausible since farmers are more like to join
cooperatives with coffee marketing or collection centers nearby their villages (ibid.). Previous
empirical studies (e.g., Jena and Grote, 2015; Zekarias and D’Haese, 2016; Musa and Hiwot,
2017) also support the finding of this study. However, walking distance to all-weather road in
minutes had a negative and statistically significantly influence on the decisions to join certified
coffee marketing cooperatives at 10% probability level. This finding is consistent with the
hypothesis. Previous empirical evidences (Degnet and Mekbib, 2013). Degnet and Mekbib
(2013) found similar findings in which road distance had a positive and significant effect on
the decision to join agricultural cooperatives by smallholder farmers in Ethiopia.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
114
Table 5: Binary Probit Model Results of Determinants of the Decisions to Join
Cooperatives
Variable
Coef.
z
P>|z|
dy/dx*
The dependent variable = Decision to join cooperative (1= Yes, 0 = No)
The independent variable:
Sex* (1= Male, 0 = Female)
-0.628
-1.69
0.091*
-0.198
Age in years
0.012
1.28
0.201
0.004
Marital status* (1= Married, 0 Otherwise)
0.703
2.05
0.041**
0.273
Family size in adult equivalent
-0.063
-1.11
0.267
-0. 023
Total livestock holding in tropical livestock unit (TLU)
-0.070
-2.01
0.044**
-0.026
Total land size under coffee production in ha
0.318
2.43
0.015**
0.117
Coffee farming experience in years
-0.018
-1.44
0.151
-0.007
Log total quantity of coffee produced in kg (log_TQCH)
0.511
1.83
0.067*
0.189
Social capital* (1= Member of idir, senbete or ekub, etc.,
0 otherwise)
0.274
1.36
0.174
0.098
Access to extension services* (1= Yes, 0 otherwise)
0.459
1.22
0.222
0.178
Access to credit* (1= Yes, 0 otherwise)
-0.353
-1.85
0.064*
-0.134
Walking distance to DA’s office in minutes
-0.007
-2.24
0.025**
-0.003
Walking distance to nearest coffee marketing center in
minutes
-0.008
2.83
0.005**
0.003
Walking distance to all-weather road in minutes
-0. 004
-1.40
0.162
-0.002
Constant
-1.612
-1.65
0.098*
Number of obs.
336
Wald chi2 (14)
47.83
Prob > chi2
0.000***
Pseudo R2
0.108
Log likelihood
-197.852
Source: Own Survey Data (2019).
*dy/dx is for discrete change of dummy variable from 0 to 1, and shows marginal effect of the
variable after probit model.
Note: ***, ** and * represent 1%, 5% and 10% probability level, respectively.
Impact of Cooperatives on Gross Annual Income Earned
After estimating the propensity scores to join (FT, Org or dual FT-Org) certified coffee
marketing cooperatives using the binary probit model, the next step is to determine the common
support region. The results in Table 6 depict that the propensity scores of both treated and
control groups of the sample households vary between 0.213 and 0.996 (with mean =0.628) for
coop member households (treatment group) and between 0.080 and 0.928 (with mean =0.540)
for non-coop member households (control group). Then, the range of the propensity scores in
the common support region for both the treatment and control groups was selected based on
the minima and maxima selection criteria. The basic criterion of this approach is to delete all
observations whose propensity score is smaller than the minimum and larger than the maximum
in the opposite group (Caliendo and Kopeinig, 2008). Thus, based on these selection criteria,
the common support region for the two groups would then lie between 0.213 and 0.928
propensity scores. In other words, the sample households in treatment and control groups with
estimated propensity scores less than 0.213 and greater than 0.928 were excluded from further
matching exercise. As a result of this restriction, of the total 336 (211 treated and 125 control)
sample farmers considered for propensity score analysis, 320 sample farmers (195 from the
treated and all 125 from the control groups) were retained and all of the remaining 16 sample
farmers from the treated group were discarded from further analysis (See figures 2 to 4).
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
115
Table 6: Distribution of the Estimated Propensity Scores to Join Cooperatives
Group
Obs.
Mean
Std. Dev.
Min
Max
Obs on
support
Coop member
HHs
211
.680
.180
.213
.996
195
Non-coop
member HHs
125
.540
.152
.080
.928
125
Total
336
.628
.183
.080
.996
320
Source: Own survey data (2019)
HHs = Households
Figure 2 portrays the distribution of the households with respect to their estimated propensity
scores. Most of the treatment households are found in the right side and partly in the middle.
On the other hand, most of control households are found in the left side of the distribution. In
general, the graph shows that there is wide area in which the propensity scores of cooperative
member households are similar with that of non-cooperative member households.
Figure 2. The Kernel Density Distribution of Propensity Scores of Coop Member and Non-
Coop Member Households
Source: Own Survey Data (2019)
Figure 3: The Propensity Scores of Coop Members in Common Support before Matching.
Source: Own Survey Data (2019)
NCMHHS = Non-Cooperative Member Households
0.5 11.5 22.5
Density
0.2 .4 .6 .8 1
psmatch2: Propensity Score
Total sample households
Coop member households
Non-coop member households
Pscore before matching
kernel = epanechnikov, bandwidth = 0.0519
Kernel density estimate
0.5 11.5 2
Density
.2 .4 .6 .8 1
psmatch2: Propensity Score
Coop members
NCMHHs in common support
Pscore of CMHHs in common support before matching
kernel = epanechnikov, bandwidth = 0.0557
Kernel density estimate
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
116
Figure 4: Propensity Scores of Non-Coop Members in Common Support before Matching
Source: Own Survey Data (2019)
CMHHs = Cooperative member households; NCMHHs = Non-cooperative member
households
Figures 3 and 4 depict the distribution of estimated propensity scores, with and without the
imposition of the common support condition, for coop member and non-coop member
households, respectively.
After estimating the propensity scores and determining the common support region, the next
step is finding an appropriate matching estimator. In this regard, alternative matching
estimators can be employed in matching the cooperative members and comparison households
in the common support region. The final choice of a matching estimator was done by taking
different criteria such as equal means test referred to as the balancing test (Dehejia and Wahba,
2002), pseudo R2 and matched sample size. Specifically, a matching estimator which balances
all explanatory variables (i.e., results in insignificant mean differences between the two groups),
a model which bears a lower pseudo R2 value and results in larger matched sample size is a
preferable matching algorithm (Alemu, 2010;Yemisrach et al., 2011).
Table 7 depicts the results of the matching qualities of the three matching methods. Thus, based
on the above mentioned performance criteria, the kernel matching method (bw =0.1) was used
for identifying the common support region for both coop member and non-coop member
households because it retained the highest number of matched sample households. Thus, the
kernel matching method as a whole is suitable for bootstrapping standard errors of the average
treatment effects (Alemu, 2010; Tihitina, 2011; Tium, 2013).
Thus, using the kernel matching method (bw =0.1), 336 sample households (211 from
cooperative member and 125 from non-cooperative member households) out of the total 377
sample households were retained for further matching exercise.
0.5 11.5 22.5
Density
0.2 .4 .6 .8 1
psmatch2: Propensity Score
CMHHs
NCMHHs in common support
Pscore of NCMHHs in common support before matching
kernel = epanechnikov, bandwidth = 0.0521
Kernel density estimate
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
117
Table 7: Performance of Matching Qualities of the Different Estimators (Algorithms)
Matching estimator
Matching performance criteria
Balancing
test*
Pseudo-
R2
Mean bias
Matched
sample size
Nearest
neighbor
matching
Neighbor (1)
10
0.054
4.34
324
Neighbor (2)
10
0.048
4.32
337
Neighbor (3)
12
0.033
4.32
337
Neighbor (4)
11
0.026
4.32
336
Neighbor (5)
12
0.031
4.32
337
Kernel
matching
Bandwidth (0.01)
11
0.026
4.32
317
Bandwidth (0.1)
11
0.026
4.32
336
Bandwidth (0.25)
11
0.026
4.32
336
Radius
caliper
matching
Caliper (0.10)
11
0.026
4.32
234
Caliper (0.25)
11
0.025
4.32
250
Caliper (0.50)
11
0.025
4.32
250
Source: Own Survey Data (2019)
*Number of explanatory variables with no statistically significant mean differences between
the matched groups of participant and non-participant households.
After choosing the best performing matching algorithm, the next job is to check the balancing
properties of the propensity scores and covariates between the treated and control groups by
using the selected matching algorithm which is kernel matching with 0.1 bandwidth. Table 8
shows the balancing tests of the propensity scores and covariates of the matching groups before
and after matching. As such, the results indicate that the propensity scores and covariates of
the matching groups after matching were insignificant making it possible to estimate the
average treatment effect of cooperative membership on the gross annual income earned by
cooperative member households.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
118
Table 8: Balancing Tests of the Propensity Scores and Covariates of the Matching
Groups
Covariate
Matching
group
Kernel-based matching (bw: 0.1)
Mean
%bias
t-test
Treated
Control
t
p>|t|
_ pscore
Unmatched
0.64584
0.60348
-3.942
1.000
Matched
0.64487
0.64521
-0.3
-0.04
0.965
Sex
Unmatched
0.9444
0.9441
0.0003
0.987
Matched
0.94811
0.9500
-0.8
0.09
0.930
Age
Unmatched
43.61966
42.44056
-0.9928
0.839
Matched
44.104
43.341
6.9
0.71
0.475
Literacy
Unmatched
0.8419
0.8531
0.0865
0.769
Matched
0.83491
0.84151
-1.9
-0.18
0.854
Family size, AE
Unmatched
3.66692
3.49762
-0.9946
0.8397
Matched
3.7456
3.8293
-5.4
-0.56
0.577
Livestock holding size,
TLU
Unmatched
4.82208
4.75279
-0.2556
0.601
Matched
4.8141
4.8195
-0.2
-0.02
0.981
Coffee farming experience,
years
Unmatched
19.00427
18.62238
-0.4172
0.662
Matched
18.769
18.976
-2.5
-0.24
0.812
Extension access
Unmatched
0.9573
0.9510
0.0796
0.778
Matched
0.95283
0.97547
-10.2
-1.25
0.211
Training access
Unmatched
0.8846
0.8531
0.7911
0.374
Matched
0.88208
0. 90755
-7.7
-0.85
0.394
Off-farm income source
Unmatched
0.3846
0.3427
0.6713
0.413
Matched
0.3868
0.3632
4.8
0.500
0.617
Non-farm income source
Unmatched
0.1624
0.0979
3.1046
0.078*
Matched
0.15566
0.12925
7.8
0.78
0.438
DISTDAWM
Unmatched
34.1111
39.4790
1.9200
0.028**
Matched
34.863
36.835
-7.2
-0.78
0.437
DISTCOOPWM
Unmatched
35.5769
39.9475
1.4862
0.069
Matched
36.34
35.114
4.2
0.47
0.635
Source: Own Survey Data (2019).
Key: ***, **, and * refer to probability at 1%, 5% and 10% probability level, respectively.
Based on the above matching tests, all the propensity scores and covariates of the matching
groups after matching were selected for estimating the average treatment effect of cooperative
membership on gross annual total income earned because their mean values were statistically
insignificantly different between the treatment and control groups of the sample farmers. Table
9 presents the results on the average treatment effect of membership to certified coffee
marketing cooperative on gross annual income earned by cooperative member farmers
computed using the kernel PSM method (bandwidth = 0.1). The results show that membership
to certified coffee marketing cooperative had a positive and significant effect on the average
annual gross income earned (ETB) by the member farmers. The average gross annual income
earned by the coop member farmers was ETB 14639.15, which is by 36.51% higher than that
of the non-cooperative member farmers. The difference is statistically significant at 1%
probability level.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
119
Table 9: Average Treatment Effect of Coop Membership on Gross Annual Income
Earned (ETB) Using Kernel Matching Method (Bw= 0.1)
Outcome variable
PSM group
Mean
Treated
Control
Difference
S.E.b
T-stat
Gross annual income earned (ETB)
48903.79
34264.64
14.639.15
3697.22
4.53***
Source: Computation Based on Own Survey Data (2019)
Key:*** indicate statistical significance at 1% probability level.
bBootstrapped standard errors (S.E.) obtained after 100 replications.
The bootstrapped standard errors of the mean of the treatment effect (i.e., cooperative
membership) on gross annual income earned in Ethiopian Birr, obtained after 100 replications,
indicate that there are significant differences in the variances of the average of gross annual
income earned between cooperative member and non-cooperative member household (see
Table 10).
Table 10: Bootstrapped Statistics of Standard Errors of the Average Treatment Effect
Variable
Reps
Observed
Bias
Std. Err.
[95% Conf. Interval]
Tot income earned (ETB)
100
6561.62
830.12
4471.3
-2310.415
15433.65
(N)
362.452
(P)
-1407.292
(BC)
Source: Computation Based on Own Survey Data (2019)
Note: Reps = Replications; N = normal’ P = percentile; BC = bias-corrected
Last but not the least, the result on the Rosenbaum bounding sensitivity analysis of the average
treatment effect on the treated (ATT) (i.e., the effect of certified coffee marketing cooperative
membership) on average gross annual income earned by the coop member households,
indicates that the observed treatment effect of certified coffee marketing cooperative
membership on the average gross annual income earned was insensitive to selection,
unobservable or hidden biases (see Table 11).
Table 11: Sensitivity Analysis of the ATT Using the Rosenbaum Bounding Method
Gamma
sig+
sig-
1
0
0
2
0
0
3
0
0
4
0
0
5
2.6e-14
0
6
3.2e-12
0
7
1.0e-10
0
8
1.3e-09
0
9
1.0e-08
0
10
5.2e-08
0
Source: Computation Based on Own Survey Data (2019).
*gamma - log odds of differential assignment due to unobserved factors, sig+ - upper bound
probability level, sig- - lower bound probability level
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
120
CONCLUSION AND RECOMMENDATIONS
The study analyzed factors influencing the decisions to join FT, Org or dual FT-Org certified
coffee marketing cooperatives and its impact on gross annual income earned by the member
farmers in two major coffee growing district of Jimma zone, southwestern Ethiopia. The binary
probit model results show that the decision to join (FT, Org or duak FT-Org) certified coffee
marketing cooperatives was negatively and significantly influenced by sex, total livestock
holding size, credit access, and walking distances to DA’s office, nearby coffee marketing
center and all-weather road in minutes, respectively. However, marital status, total coffee land
size (ha), log total quantity of coffee produced (kg) had positive and significant influences on
the decisions to join these cooperatives. Regarding the impact of cooperative membership, the
PSM results show that certified coffee marketing had positive and significant effect on average
gross annual income earned by the member farmers.
It is thus important to strengthen existing certified coffee marketing cooperatives so that
members continue to derive income benefit from the cooperatives. No members residing in
kebeles covered by such types of cooperatives should be encouraged to join such cooperatives
and derive membership benefits. Moreover, we call for promotion of the establishment
cooperatives registered for coffee certification schemes in kebeles without such cooperatives
by smallholder coffee farmers so that they earn better income.
In order to increase the likelihood of non-members joining existing certified coffee marketing
cooperatives however both the public sector, the cooperatives themselves and other
stakeholders should work hard to identify and address gender-and marital status-based factors
that influence cooperative membership decisions. Moreover, cooperatives are advised to
incorporate credit or loan schemes in order to attract non-members non-members join them.
On the other hand, concerned stakeholders (the cooperatives themselves and development
agents’ (DAs’)) should provide information regarding the relevance of cooperatives specially
for smallholder farmers relatively living far away from the cooperatives’ and development
agents’ (DAs’) offices using various communication channels.
On the other hand, cooperatives should establish harvested coffee collection centers nearby
farmers’ residences so as to attract non-members join the cooperatives. Moreover, factors
hindering decisions to join certified coffee marketing cooperatives by smallholder farmers with
relatively smaller coffee land size (ha) and lower coffee yield should be identified and
addressed in order to encourage such farmers join the cooperatives and derive membership
benefits such as better gross annual income.
Cooperatives should also be encouraged to establish credit and saving units in their internal
structure and/or work in collaboration with other saving and credit providing institutions (such
as Cooperative Bank of Oromia) to be able to provide demand-driven credit services to
member farmers (Zekarias and D’Haese, 2018; Minten et al., 2018).
Acknowledgments
The authors are grateful to the Ministry of Education for funding the research through
Haramaya University. We also owe special thanks to the officials of bureaus of agriculture and
cooperative promotion agencies of Iimma zone, and Gomma and Limmu Kossa districts of
Jimma zone for generously providing the information needed throughout the study. We are
highly indebted to the farmers, enumerators and supervisors for their wholehearted
participation in the study and provision of invaluable information in the course of data
collection. Last but not the least, we wish to thank friends, colleagues and staff of the
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
121
Department of Rural Development and Agricultural Extension at Haramaya University and
Madda Walabu University for their encouragements and invaluable supports which made this
research possible.
Funding
This research was funded by the Ministry of Education through Haramaya University, Ethiopia.
Competing interests
The authors declare no competing interests.
Citation information
Cite this article as: Impact of certified coffee marketing cooperatives on income of smallholder
farmers in Jimma Zone of Oromia region, Ethiopia, Zewdu Getachew, Fekadu Beyene, Jema
Haji & Tesfaye Lemma, Cogent Food & Agriculture (22), …
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
122
REFERENCES
Abu Tefera and Teddy Tefera, 2013. Coffee Annual Report. Ethiopia. Global Agricultural
Information Network (GAIN). USDA Foreign Agricultural Service. GAIN Report
Number: ET-1302. Date: Accessed on 22/07/2016.
Alemseged Assefa. 2012. Ethiopia’s Coffee Export Performance Report - Fiscal Year 2011-12
by Alemseged General Manager, ECEA Addis Ababa.
Alemu Tolemariam. 2010. Impact Assessment of Input and Output Market Development
Interventions by IPMS Project: The Case of Gomma Woreda. MSc. Thesis Research.
Haramaya University.
Amsaya Anteneh, R. Muradian and R. Ruben. Chapter 4: Impact of multiple certification on
smallholder coffee farmers’ livelihoods: evidence from southern Ethiopia. In ‘Ruben R.
and P. Hoebink. 2015. Editors. Coffee certification in East Africa: Impact on farms,
families and cooperatives Wageningen Academic.Publishers.
http://www.tandfonline.com/doi/pdf/10.1080/19439342.2015.1073893?needAccess=t
rue.
Bastin, A. and N. Matteucci. 2007. Financing Coffee Farmers in Jimma Zone, Ethiopia:
Challenges and Opportunities, International Conference on Rural Finance Research:
Moving Results into Policies and Practice, FAO Headquarters, Rome, Italy, 19-21
March 2007.
Becker, S. O. and A. Ichino. 2002. Estimation of average treatment effects based on propensity
scores. The Stata Journal, 2(4): 358–377.
Berhanu Tsegaye, Ali Mohammed, and Essubalew Getachew, 2015. Effect of Sun Drying
Methods and Layer Thickness on Quality of Selected Natural Arabica Coffee Varieties
at Jimma, South west Ethiopia. Discourse Journal of Agriculture and Food Sciences.
www.resjournals.org/JAFS ISSN: 2346-7002 Vol. 3(9): 135-147.
Bernard, T., Alemayehu Seyoum and Eleni Gabre-Madhin. 2008a. Impact of Cooperatives on
Smallholders’ Commercialization Behavior: Evidence from Ethiopia. Agricultural
Economics, 39 (2): 147–61.
Bernard, T. and D. Spielman. 2009. Reaching the Rural Poor through Rural Producers
Organizations? A Study of Agricultural Marketing Cooperatives in Ethiopia. Food
Policy, 34 (1): 60–69.
Bernard T., 2010. Cooperatives for Staple crop Marketing: evidence from Ethiopia.
Washington, USA. IFPRI research monograph; 164. http://www.ifpri.org/sites/
default/files/publications/rr164.pdf.
BoCPA (Bureau of Cooperative Promotion Agency) of Jimma zone. 2018. Jimma zone. Jimma.
BoFED (Bureau of Finance and Economic Development). 2019. Physical and Socio-Economic
Profile of Jimma zone. Jimma.
Bryson, A., Dorsett, R., and S. Purdon, 2002. The Use of Propensity Score Matching in the
Evaluation of Active Labor Market Policies. Working Paper 4. Department for Work
and Pensions, London.
Caliendo, M., and S. Kopeining. 2008. Some Practical Guidance for The Implementation Of
Propensity Score Matching. Journal of Economic Surveys, 32: 31–72.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
123
Dagne Mojo, C. Fischer and Terefe Degefa, 2015. Social and environmental impacts of
agricultural cooperatives: evidence from Ethiopia. International Journal of Sustainable
Development & World Ecology.
http://www.tandfonline.com/doi/pdf/10.1080/13504509.2015.1052860?needAccess=t
rue.
Dagne Mojo, C. Fischer , Terefe Degefa, 2017. The determinants and economic impacts of
membership in coffee farmer cooperatives: recent evidence from rural Ethiopia.
Journal of Rural Studies, 50: 84-94. ScienceDirect. Elsevier Ltd.
Daniel, W.W. and C. L. Cross. 2013. Biostatistics: A Foundation for Analysis in the Health
Sciences. 10th edition. John Wiley & Sons, Inc. Library of Congress Cataloging-in-
Publication Data. ISBN 978-1-118-30279-8 (cloth). Printed in the United States of
America.
Delelegne A. Tefera, J. Bijman and M. A. Slingerland. 2017. Agricultural cooperatives in
Ethiopia-Evolution, function and impact. Review Article. Journal of International
Development, 29, 431–453.
Dercon S., D. O. Gilligan, J. Hoddinott and Tassew Woldehanna. 2008. The Impact of
Agricultural Extension and Roads on Poverty and Consumption Growth in Fifteen
Ethiopian Villages. Food Consumption and Nutrition Division. IFPRI Discussion Paper
00840.
ECEA (Ethiopian Commodity Exchange Authority). 2008. Analysis of Coffee Supply,
Production, Utilization and Marketing Issues and Challenges in Ethiopia.
Understanding Commodities to be Traded at Ethiopia Commodity Exchange. Volume
I. Policy Analysis and Economic Research Team. Addis Ababa.
Ferris, S., P. Robbins, R. Best, D. Seville, A. Buxton, J. Shriver and E. Wei, 2014. Linking
Smallholder Farmers to Markets and the Implications for Extension and Advisory
Services. MEAS Discussion Paper Series on Good Practices and Best Fit Approaches
in Extension and Advisory Service Provision.
Fikadu Gutu, Wondaferahu Mulugeta and Sisay Tolla. 2020. Economic Impact of Fair-Trade
Certification on Small-Scale Coffee Producers in Ethiopia. Journal of International
Trade, Logistics and Law, Vol. 6, Num. 1, 2020, 52-60.
Fikadu Mitiku, Y. de Mey., J Nyssen and M. Maertens. 2017. Do Private Sustainability
Standards Contribute to Income Growth and Poverty Alleviation? A Comparison of
Different Coffee Certification Schemes in Ethiopia. http://www.mdpi.com/2071-
1050/9/2/246/pdf.
Fischer, E. and Qaim, M. 2011. Smallholder Farmers and Collective Action: What Determines
the Intensity of Participation? Proceedings of the German Development Economics
Conference, Berlin, No. 28.
Francesconi G. and N.Heerink. 2010, ‘Ethiopian agricultural cooperatives in an era of global
commodity exchange: does organizational form matter?’ Journal of African Economies,
20(1): 1-25.
Heckman, J.J., Ichimura, H. and Todd, P. 1998. Matching as an Econometric Evaluation
Estimator. Review of Economic Studies, 65 (2): 261-94.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
124
Heckman, J. J., LaLonde, R. and Smith, J. 1999. The Economics and econometrics of Active
Labor Market Programs. In “Ashenfelter, O. and Card, D. 1999. Editors. Handbook of
Labor Economics, (3): 1865-2097. Amsterdam: North-Holland.”
Holloway G., C. Nicholson and C. Delgado, 2000. Agroindustrialization through Institutional
Innovation: Transactions Costs, Cooperatives and Milk-Market Development in the
Ethiopian Highlands. MSSD Discussion Paper No. 35.
Jalan J. and M. Ravallion. 2003. Estimating the Benefit Incidence of an Antipoverty Program
by Propensity-Score Matching. Journal of Business & Economic Statistics, 21 (1):19-
30.
Jawtuch J. B. Oehen, and U. Nihhli. 2011. Environmental, Social, and Economic Impacts of
Sustainability Certification in the Agricultural Sector-– The Current State of Empirical
Research. Standards And Regulations. FiBL & IFOAM (2011): The World of Organic
Agriculture 2011. Bonn and Frick.
Jena, P. R., Stellmacher, T. and U. Grote. 2012. The Impact of Coffee Certification on Small-
Scale Producers’ Livelihoods: Evidence from Ethiopia. Selected Paper prepared for
presentation at the International Association of Agricultural Economists (IAAE)
Triennial Conference, Foz do Iguaçu, Brazil.
Jena, P. R., T. Stellmacher and U. Grote. 2015. Can coffee certification schemes increase
incomes of smallholder farmers? Evidence from Jinotega, Nicaragua. Environment,
Development and Sustainability, 19: 45–66. Springer. DOI: 10.1007/s10668-015-9732-
0.
Kherallah M. and J.F. Kirsten. 2002. The new institutional economics-Applications for
agricultural policy research in Developing countries. Agrekon, (41) 2.
Manda J., M. G. Khonjeb, A. D. Alenec, A. H Tufa, T. Abdoulayed, M. Mutenjee, P. Setimelae
and V. Manyong. 2020. Does cooperative membership increase and accelerate
agricultural technology adoption? Empirical evidence from Zambia. Technological
Forecasting & Social Change, 158, 120160.
Minten B., Mekdim Dereje, Ermias Engeda, and Seneshaw Tamru. 2015. Who Benefits from
the Rapidly Increasing Voluntary Sustainability Standards? Evidence from Fairtrade
and Organic Coffee in Ethiopia. Agriculture in an Interconnected World. International
Conference of Agricultual Economics. August 8-14. Milan, Italy.
https://ageconsearch.umn.edu/record/212708/files/MintenDerejeEtAl_ValueChain_IA
AE2015.pdf.
Minten B., Mekdim Dereje, Ermias Engida, Seneshaw Tamru. 2018. Tracking the Quality
Premium of Certified Coffee: Evidence from Ethiopia. World Development, 101:119-
132. https://doi.org/10.1016/j.worlddev.2017.08.010.
MoT (Ministry of Trade). 2012. Proposal for Coffee Platform in Ethiopia. Draft. United
Nations Development Programme.
Musa Hasen and Hiwot Mekonnen. 2017. The impact of agricultural cooperatives membership
on the wellbeing of smallholder farmers: empirical evidence from eastern Ethiopia.
Agricultural and Food Economics (2017) 5:6. DOI 10.1186/s40100-017-0075-z.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.8, Issue 5, No.5. pp 100 - 125, 2023
www.iprjb.org
125
Nilsson J., G. T. Svendsen and Large and G. L. Svendsen. 2012. Are Complex Agricultural
Cooperatives Losing Their Social Capital? Article in Agribusiness · March 2012. DOI:
10.1002/agr.21285.
Petit, E. 2007. Ethiopia’s Coffee Sector-A Bitter or Better Future. Journal of Agrarian Change,
(7)2: 225–263. Blackwell Publishing Ltd, Henry Bernstein and Terence J. Byres.
Petty, C., Seaman, J. and Majid, N. with F. Grootenhuis. 2004. Coffee and Household Food
Security-A Study of Coffee and Household Economy in Two Districts of Ethiopia.
Funded by DFID, Save the Children UK.
Rosenbaum, P. R. and D. B. Rubin. 1983. ‘The central role of the propensity score in
observational studies for causal effects’. Biometrica, 70 (1): 41-55.
Shrestha R. K. 2015. Building social capital within the framework of agricultural cooperatives
development in rural Nepal. A thesis submitted for the degree of Doctor of Philosophy
at the School of Agriculture and Food Sciences, the University of Queensland.
Stellmacher T., U. Grote and J. Volkmann. 2010. Protection of biodiversity through coffee
certification? The case of forest coffee in Bench Maji and Kaffa Zone, Ethiopia.
Certification for forest coffee, Ethiopia. TEEBcase available at: TEEBweb.org.
Stellmacher, T. and U. Grote, 2011. Forest Coffee Certificationin Ethiopia: Economic Boon or
Ecological Bane? ZEF Working Paper Series, ISSN 1864-6638. Department of Political
and Cultural Change. Center for Development Research, University of Bonn.
Tadesse Meskela and Yalem Teshome, 2014. From Economic Vulnerability to Sustainable
Livelihoods: The Case of the Oromia Coffee Farmers Cooperatives Union in Ethiopia.
Tihitina Abebe, 2011. Impact Assessment of Input and Output Market Development
Intervention of the IPMS Project: The Case of Meiso Woreda, Oromia National
Regional State, Ethiopia. MSc. Thesis Submitted to the Department of Agricultural
Economics, School of Graduate Studies, Haramaya University.
Tium Gebrehiwet. 2013. Impact of Fairtrade Certification on Smallholder Coffee Producers’
Income: The Case of Gimbo District, South Western Ethiopia. M.Sc. Thesis. Haramaya
University.
Wollni, M. and M. Zeller. 2007. Do farmers benefit from participating in speciality markets
and cooperatives? The case of coffee marketing in Costa Rica. Agricultural Economics
37, 243–248.
Yemisrach Getachew, Moti Jaleta and Birhanu G/medihin. 2011. Impact of Input and Output
Market Development Interventions on Input Use and Net Income of Households in
Ethiopia. Journal of Economics and Sustainable Development, (2)9. www.iiste.org.
Zekarias Shumeta and M. D’Haese. 2016. Do Coffee Cooperatives Benefit Farmers? An
Exploration of Heterogeneous Impact of Coffee Cooperative Membership in
Southwest Ethiopia. Research Article. International Food and Agribusiness
Management Review, (19) 4: 37-52.
http://www.wageningenacademic.com/doi/pdf/10.22434/IFAMR2015.0110.
06/02/2017.
Zekarias Shumeta and M. D’Haese. 2018. Do Coffee Farmers Benefit in Food Security from
Participating in Coffee Cooperatives? Evidence from Southwest Ethiopia Coffee
Cooperatives. Food and Nutrition Bulletin, 39(2): 266-280.
