apce12035

W3G-apce

February 21, 2014

5:22

APCE

apce12035 Dispatch: February 21, 2014 CE: N / A

Journal MSP No.
1

No. of pages: 30

PE: Wendy
pp. 1–30

Annals of Public and Cooperative Economics 85:2 2014

2
3
4
5
6
7
8

IMPACT OF AGRICULTURAL COOPERATIVES ON
SMALLHOLDERS’ TECHNICAL EFFICIENCY: EMPIRICAL
EVIDENCE FROM ETHIOPIA

9
10
11

by
Gashaw Tadesse ABATE∗

12
13
14

University of Trento, Graduate School of Social Sciences and European Research
Institute on Cooperatives and Social Enterprises (EURICSE), Italy

15
16
17

Gian Nicola FRANCESCONI

18
19
20

International Food Policy Research Institute (IFPRI) and International Center for
Tropical Agriculture (CIAT), Dakar, Senegal

21
22

and

23
24
25

Kindie GETNET
International Water Management Institute (IWMI), Addis Ababa, Ethiopia

26
27
28
29
30
31

ABSTRACT: Using household survey data from Ethiopia, this paper evaluates
the impact of agricultural cooperatives on smallholders’ technical efficiency. We used
propensity score matching to compare the average difference in technical efficiency
between cooperative member farmers and similar independent farmers. The results
show that agricultural cooperatives are effective in providing support services that
significantly contribute to members’ technical efficiency. These results are found to be
insensitive to hidden bias and consistent with the idea that agricultural cooperatives
enhance members’ efficiency by easing access to productive inputs and facilitating extension linkages. According to the findings, increased participation in agricultural
cooperatives should further enhance efficiency gains among smallholder farmers.

32
33
34
35
36
37
38
39
40
41
42

Keywords: Agricultural cooperatives, smallholder farmers, technical efficiency, Ethiopia.
JEL classification: Q12, Q13, Q16

43
44
45
46
47
48
49
50
51

∗

We thank Shahidur Rashid for his suggestions and providing us access to the data and
Nigussie Tefera, Mark Beittel and Illana Bodini for their helpful comments. We are also grateful
to an anonymous reviewer for comments and suggestions received during the review process of
this paper. E-mail: gashawtadesse.abate@unitn.it

© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC. Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford
OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

apce12035

1

2

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

¨ die technische Effizienz von
Bedeutung von Agrargenossenschaften fur
¨
Kleinbauern: Empirischer Befund aus Athiopien
von Gashaw Tadesse Abate, Gian Nicola Francesconi und Kindie Getnet
¨
Unter Verwendung von Haushaltserhebungsdaten aus Athiopien
wird in diesem Beitrag die
Wirkung von Agrargenossenschaften auf die technische Effizienz von Kleinbauern evaluiert. Wir
wenden das Propensity-Score-Matching-Verfahren an, um die durchschnittliche Differenz der technischen Effizienz zwischen Bauern, die Mitglied in einer Genossenschaft sind, und unabhangigen
¨
Kleinbauern zu vergleichen. Die Ergebnisse zeigen, dass Agrargenossenschaften effektiv sind
bezuglich
¨
der Bereitstellung von Unterstutzungsdiensten,
¨
die signifikant zur technischen Effizienz
der Mitglieder beitragen. Es lasst
¨
sich feststellen, dass diese Ergebnisse resistent gegenuber
¨
versteckten Verzerrungen sind und konform gehen mit der Vorstellung, dass Agrargenossenschaften
die Effizienz der Mitglieder steigern, indem sie den Zugang zu produktiven Inputs erleichtern und
der Geschaftserweiterung
¨
dienliche Verbindungen erm¨oglichen. Diesen Ergebnissen zufolge sollte
eine starkere
¨
Beteiligung an Agrargenossenschaften die Effizienzgewinne von Kleinbauern weiter
steigern.

´
Impacto de las cooperativas agr´ıcolas sobre la eficacia tecnica
de las
˜
´
pequenas
explotaciones: analisis
emp´ırico en Etiop´ıa
A partir de los datos obtenidos de encuestas a los hogares en Etiop´ıa, este art´ıculo evalua
´ el impacto
de las cooperativas agr´ıcolas sobre la eficacia t´ecnica de las pequenas
˜
explotaciones. Los autores
utilizan el m´etodo denominado propensity score matching para comparar la diferencia media de
eficacia t´ecnica entre los miembros de las cooperativas agr´ıcolas y los agricultores independientes.
Los resultados muestran que las cooperativas agr´ıcolas son eficaces para proveer servicios de apoyo
que contribuyen significativamente a la eficacia t´ecnica de sus miembros. Estos resultados son insensibles a los sesgos existentes y confirman la idea de que las cooperativas agr´ıcolas incrementan
la eficacia de sus miembros facilitandoles
´
el acceso a los inputs productivos y a contactos comerciales. Los resultados ponen de manifiesto que intensificar la participaci´on en las cooperativas
agr´ıcolas acrecientan las ganancias de eficacia en los pequenos
˜ agricultores.

´
Impact des cooperatives
agricoles sur l’efficacite´ technique des petits
exploitants: Analyse empirique en Ethiopie
A partir de donn´ees d’enquˆetes de m´enages en Ethiopie, cet article e´ value l’impact des coop´eratives
agricoles sur l’efficacit´e technique des petits exploitants. Les auteurs utilisent la m´ethod dite propensity score matching pour comparer la diff´erence moyenne d’efficacit´e technique entre membres
de coop´eratives agricoles et fermiers ind´ependants. Les r´esultats montrent que les coop´eratives
agricoles sont efficaces pour fournir des services de support qui contribuent significativement a`
l’efficacit´e technique de leurs membres. Ces r´esultats sont insensibles a` des biais cach´es et confirment l’id´ee que les coop´eratives agricoles augmentent l’efficacit´e des membres en facilitant l’acc`es a`
des inputs productifs et les liaisons a` distance. Les r´esultats indiquent qu’intensifier la participation
dans les coop´eratives agricoles accroitrait les gains d’efficacit´e des petits agriculteurs.

49
50
51
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

3

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

1

Introduction

Enhancing productivity and commercialization among smallholder farmers is widely
perceived as a key strategy for rural development, poverty reduction, and food security
in Sub-Saharan Africa (World Bank 2008). For productivity gains to be achieved, smallholder farmers need to have better access to technology and improve their technical
efficiency. It is important for smallholders to have easy access to extension services in
order to optimize on-farm technical efficiency and productivity, given the limited resources available. While the private sector is gradually emerging as a contender, the
public sector remains the major provider of extension services in most of these countries
(Venkatesan and Kampen 1998). A third option for providing services to smallholder
farmers is agricultural cooperatives, which serve the dual purpose of aggregating smallholder farmers and linking them to input and output markets (Coulter et al. 1999, Davis
2008).
Given that agricultural systems in Sub-Saharan Africa are typically fragmented
into a myriad of small or micro farms over vast and remote rural areas, the role of agricultural cooperatives has become increasingly important (Wanyama et al. 2009). Despite
the turbulent history sometimes associated with post-independence and highly centralized governance regimes, agricultural cooperatives are nowadays omnipresent throughout the sub-continent. In recent days considerable public development programs or
private initiatives are channelled through cooperatives in order to overcome prohibitive
transaction and coordination costs (Pingali et al. 2005). However, it is still empirically
unclear and highly contested whether these collective organizations can deliver and
live up to their promises. Given the prominence of agricultural cooperatives, this is an
important policy question for many African countries.
Since the downfall of the Derg regime in 1991, agricultural cooperatives in Ethiopia
have become an integral part of the national strategy for agricultural transformation
(Ministry of Finance and Economic Development 2006). With varying degrees of success, agricultural cooperatives are longstanding and widespread throughout the country (Bernard et al. 2008, Bernard and Spielman 2009, Francesconi and Heerink 2010,
Francesconi and Ruben 2007, Getnet and Tsegaye 2012, Tigist 2008). The recently established Agricultural Transformation Agency (ATA) has also strongly asserted agricultural cooperatives as preferential institutions for moving smallholders out of subsistence
agriculture and linking them to emerging input and output markets. In conjunction with
promotional activities by the National Cooperative Agency, this effort has resulted in
considerable growth both in number of agricultural cooperatives and the services they
provide to their members. In June 2012, the majority of both the 40,000 primary cooperatives and the 200 cooperative unions in the country were agricultural cooperatives
engaged in input and output marketing.
By 2005, agricultural cooperatives had commercialized more than 10 per cent of the
marketable surplus in Ethiopia (Bernard et al. 2008). In recent years they are the major
suppliers of improved seeds and chemical fertilizer for all farm households (Ministry of
Agriculture and Rural Development 2010: unpublished). While their role in agricultural
inputs adoption for productivity growth is widely recognized (Abebaw and Haile 2013,
Spielman et al. 2011), the impact of technical efficiency gains among their members
remain unproven. Whether cooperative members are technically more efficient than
non-members is an open question. Agricultural cooperatives, as producer organizations,
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

Q1

apce12035

1

4

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

are mandated to supply inputs together with providing embedded support services and
for facilitating farmer linkage with extension service providers; hence, members are
expected to be technically more efficient.
This paper aims to answer this question by comparing cooperative members and
similar independent farmers within the same kebeles1 (in order to reduce potential
differences in technology and agro-ecology in which this procedure tempers possible
diffusion effects). This approach, which compares members and non-members within
the same kebeles in which the agricultural cooperatives operate, enables us to precisely
capture the efficiency gains from membership, since members receive benefits from
dividends, information, and extension services that are embedded in new technologies
and have prior access to inputs, which are directly linked with technical efficiency gains.
We used the Stochastic Production Frontier (SPF) function model to measure the
technical efficiency of sampled farm households, as it is effective in estimating the efficiency score of households that account for factors beyond the control of each individual
producer (Coelli et al. 2005, Kumbhakar and Lovell 2000). After estimating the technical
efficiency score, we applied Propensity Score Matching (PSM) techniques to estimate the
impact of membership in agricultural cooperatives on technical efficiency, drawing on
the approaches of Bernard et al. (2008), Francesconi and Heerink (2010) and Godtland
et al. (2004). Rosenbaum bounds sensitivity analysis is conducted to understand the
sensitivity of the results obtained from the matching estimates to possible unobservable
covariates. Moreover, we checked the robustness of the results following alternative estimation strategy that aimed at accounting potential bias that might arise in estimating
technical efficiency scores.
Our results consistently show a positive and statistically significant impact of
membership in agricultural cooperatives on technical efficiency at the farm level. On average, we found about a 5 per cent difference in technical efficiency between cooperative
members and non-members. The results suggest that member households are in a better
position to obtain maximum possible outputs from a given set of inputs. The results are
insensitive for a hidden bias that would double the odds of participation in cooperatives
and they are consistent with the idea that agricultural cooperatives enhance members’ efficiency by providing easy access to inputs, information, and embedded support
services.
The rest of paper is organized as follows: section 2 highlights the history and recent
development of agricultural cooperatives in Ethiopia. Section 3 presents the data source
and descriptive statistics of the variables used in the analysis. Section 4 presents the
research methodology, including discussion of the empirical strategy, estimation procedure of the propensity scores and estimation of household technical efficiency scores.
Section 5 reports the results and section 6 concludes by discussing the main findings.
2

Agricultural cooperatives in Ethiopia

Historically, agricultural cooperatives have played an important role all over the
world in providing market access, credit and information to producers. In particular,
agricultural cooperatives in the USA and Western Europe have played an important
1

Kebele is the smallest rural administrative unit in Ethiopia.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

5

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

economic role in providing competitive returns for independent farmers (Chaddad et al.
2005). Agricultural cooperatives in those countries were established as service providers
and were primarily aimed at countervailing the market power of producers’ trading
partners, preservation of market options and reduction of risk through pooling. They
have also been accorded with a range of public policy supports that has encouraged their
market coordination role in agri-business (Staatz 1987, 1989).
In Ethiopia, however, the tradition of agricultural cooperatives was completely
different from the western type of agricultural cooperatives from the initial days of
establishment to the socialist regime. During the imperial regime (1960s-1974), a period
during which cooperatives were started, agricultural cooperatives were setup in the form
of cooperative production or agricultural collectives to jointly produce commercial and
industrial crops (i.e., coffee, tea and spices). They were not in a position to operate
efficiently due to unenforceability of efforts, inequitable incentives, higher agency costs,
and slow and centralized decision-making, which are inherent problems of collective
production2 (Deininger 1995).
During the socialist regime (1974–1990) as well agricultural cooperatives continued to be extended arms of the state and were used primarily as instruments of the
government in order to control the agricultural sector and prevent the rise of capitalistic
forms of organization (Rahmato 1990). There were two types of agricultural cooperatives
during this period: production cooperatives engaged in collective production and service
cooperatives handling modern inputs, credit, milling services, selling of consumer goods,
and purchasing of farmers produce. Production cooperatives were expected to operate
over 50 per cent of the nation’s cultivable land in the same fashion of joint production and
were believed to be more cost-effective (Rahmato 1994). However, ill-conceived policies
coupled with shirking by coerced farmers resulted in lower output and underutilization of scale and deployed labours by cooperatives as compared to individual farmers.
Besides, forced formation and routine intervention from the state agents are critical
factors, which contributed to the poor record of agricultural cooperatives during this
regime (Rahmato 1993).
Subsequently, when the new mixed economic system was introduced in 1991 farmers were given the choice to work on commonly or individually owned land; the past
negative experience led most of the farmers to reallocate common lands to individual
holdings, which eventually led to the collapse of most production cooperatives (Abegaz
1994). During the transition period, despite the efforts made to create an enabling environment for agricultural cooperatives through the issuing of new regulations,3 most of
them continued to be burgled by individuals and others downsized due to competition
from the private traders following trade liberalization (Kodama 2007, Rahmato 1994).
In general, prior to 1990 agricultural cooperatives in Ethiopia were ‘pseudo’ cooperatives
both in their undertakings and membership.
During the late 1990s, the government of Ethiopia revived its interest in cooperatives and they become part and parcel of the country’s agriculture and rural development
2

See Deininger (1995) for complete historical accounts on the inefficiencies of cooperative
production systems as compared to agricultural cooperatives providing services (marketing, credit
and information) to independent farmers in Cuba, Vietnam, Nicaragua, Peru and Ethiopia in
terms of utilization of economies of scale, innovation, equity and provision of public goods.
3
Agricultural Cooperative Societies Proclamation No. 185/1994.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

6

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

strategy (Getnet and Tsegaye 2012, Ministry of Finance and Economic Development
2006). In particular, the government strongly promoted agricultural cooperatives to encourage smallholders’ participation in the market (Bernard et al. 2008). As proclaimed
in the new legal framework, this new wave of cooperative organizations was thought to
be different from previous cooperative movements. Although externally induced formation is still prevalent,4 in relative terms the new policy allows cooperatives to be diverse
and independent participants in the free market economy.
As part of the government support for cooperative promotion, cooperative governance was also reinforced through the establishment of the Federal Cooperative Commission in 2002, a public body to promote cooperatives at the national level (Bernard
et al. 2010, Francesconi and Heerink 2010, Kodama 2007). The commission was established with a plan of providing cooperative services to two-thirds of the rural populations
and to increase the share of agricultural cooperatives in input and output marketing
through the establishment of at least one primary cooperative in each kebeles. While
there is evidence that suggests a consequent growth in the cooperative movement in
Ethiopia, its coverage remains 35 per cent of kebeles, and only 17 per cent of the households living in those kebeles are members (Bernard et al. 2008).
With regards to performance, the impacts of agricultural cooperatives are less
studied. There have been only a few attempts made to understand their commercialization role in collecting and selling members’ produces and the results are mixed.
Francesconi and Heerink (2010) found a higher commercialization rate for the farmers that belong to agricultural marketing cooperatives, which suggest the importance
of organizational form in cooperative inquiries. Bernard et al. (2008) conversely found
a similar commercialization rate for the farmers that belong to cooperatives (i.e., cooperative members tend to sell an equivalent proportion of their output to market as
compared to non-members), notwithstanding the higher price obtained by the cooperatives for members per unit of output. Their role in providing a better price through
stabilizing and correcting local market in favour of the producer is also corroborated by
Tigist (2008).
Other recent studies on impact of agricultural cooperatives by Abebaw and Haile
(2013) and Getnet and Tsegaye (2012) respectively indicated better adoption of agricultural inputs and livelihood improvement among users of cooperatives as compared to
non-users. What is scarce in the literature is the impact of agricultural cooperatives on
productivity and technical efficiency of members, despite the fact that they are mainly
used as a preferential channel to access agricultural inputs (i.e., fertilizer and improved
seeds) and services (i.e., financial, training and extension). In the technical efficiency
literature there are empirical works that suggest the positive role of membership in producer organizations or cooperatives in reducing inefficiency (Binam et al. 2005, Chirwa
2003, Idiong 2007, Jaime and Salazar 2011). However, those results are merely based
on the analysis of inefficiency models without accounting for original differences among
farm households and in countries other than Ethiopia. In an effort to address this
gap, this paper made an attempt to go one step further and compare the difference in

4

In Ethiopia member initiated cooperatives account only for the 26 per cent of the total.
The remaining 74 per cent of the cooperatives are externally initiated, mostly by government and
donor agencies (Bernard et al. 2008).
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

7

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

technical efficiency between members and non-members that are similar in their observable covariates or pre-membership characteristics in the context of rural Ethiopia.

3

Data and descriptive analysis

The key variables used in this study include household characteristics; inputs
used for production; production value and village level characteristics (such as population density and availability of farmer training centres). The data used are from the
‘Ethiopia Agricultural Marketing Household Survey’, jointly carried out by the Ethiopian
Development Research Institute (EDRI), Ethiopian Institute of Agricultural Research
(EIAR) and International Food Policy Research Institute (IFPRI) between June and August 2008. This survey provided data on all the variables of interest except village level
variables, which were then obtained separately from the Central Statistical Authority
(CSA).
The ‘Ethiopia Agricultural Marketing Household Survey’ is focused on smallholders’ production and marketing patterns and covers the four most populated regions of
Ethiopia (Amhara, Oromia, SNNP5 and Tigray). The sampling procedure employed was
a three-stage stratified random sampling.6 The original sample includes 1,707 households randomly drawn from 73 Peasant Associations (PAs). From the original sample
we dropped households with missing observation on variables of interest.7 The resulting sample used in this study includes 1,638 farm households, from which we drew a
sub-sample (i.e., member and non-member farm households within cooperative kebeles)
mainly used to address our research question.
Table 1 presents a summary of demographic and geographic characteristics of
sample households used in the analysis. From the total sample households considered,
34 per cent are members of agricultural cooperatives (i.e., treatment group) and the
remaining (66 per cent) is found to be independent farm households (i.e., comparison
group). Farm households belonging to agricultural cooperatives are relatively more literate, older, more likely to have a male head and have higher household size both in
numbers and adult equivalents. In addition, members are also more likely to own radios,
televisions and mobile phones, as compared to the non-members.
As expected, members are using more productive inputs (i.e., fertilizer and improved seeds). This can be explained by ease of access, as agricultural cooperatives are
the major last-mile distributors of fertilizers and seeds, and also by the fact that members need to compensate for relatively lower fertile land. Although not reported in the
table to conserve space, the data indicates a mean difference within non-member farm
5
6

Southern Nations, Nationalities, and Peoples Regional State.
In the first stage, the Woreda’s from each region were selected randomly from a list arranged
by degree of commercialization as measured by the Woreda-level quantity of cereals marketed
(i.e., the major focus of the survey). This ensured that that Woreda’s were uniformly distributed
across the range of level of marketed cereal outputs. In the second stage, farmers’ or peasants’
associations (FAs or PAs) were randomly selected from each Woreda. For the third stage of
selection, households were randomly selected from the list provided by the PA office.
7
For example, we dropped households that report production volume without amount of seed
used or land cultivated.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

21
22

23
24
25

26
27
28

29
30
31

32
33
34

35
36
37

38
39

43
44
45

49
50
51
6.50(2.04)
1.04(0.20)
45.76 (12.28)
0.45(0.49)
6.37(2.81)
2.75(1.04)
0.55(0.49)
0.60(0.49)
0.01(0.13)
3423.4(3149.9)
96.39(136.32)
7.46(23.86)
1.37(0.94)
5.43(1.77)
1.71(1.11)
3.34(3.33)

40
41
42

Household size
Sex of HH head
Age of HH head
HH head education level
Number of plots
Number of crops
Off-farm income
Radio and/or TV ownership
Phone ownership
Value of crop produced
Fertilizer used by HHs
Improved seed used by HHs
Cultivated land size
Labor (adult equivalent)
Oxen owned by HHs
TLUa (excluding ox)

46
47
48
6.18 (2.66)
1.10 (0.30)
44.09(13.35)
0.25(0.43)
5.14(2.72)
2.34(1.04)
0.61(0.48)
0.39(0.48)
0.006(0.08)
2266.4(2437.8)
22.41(49.61)
1.70(7.53)
1.14(0.90)
5.08(2.20)
1.19(1.07)
3.22(5.29)

Min/Max
1/26
1/2
15/89
0/1
1/21
1/7
0/1
0/1
0/1
38/19380
0/650
0/100
0.15/6.75
0.74/22.12
0/8
0/69.2

Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

1/14
1/2
20/86
0/1
1/22
1/6
0/1
0/1
0/1
133/22750
0/900
0/300
0.08/7.06
1/11.69
0/8
0/31.04

Mean (Std. Dev.)

6.29(2.47)
1.08(0.27)
44.67(12.99)
0.32(0.46)
5.56(2.81)
2.48(1.06)
0.59(0.49)
0.46(0.49)
0.01(0.10)
26665.5(2758.8)
47.88(96.13)
3.68(15.51)
1.22(0.92)
5.20(2.07)
1.37(1.11)
3.26(4.71)

Mean (Std. Dev.)

1/26
1/2
15/89
0/1
1/22
1/7
0/1
0/1
0/1
37.5/22750
0/900
0/300
0.015/7.06
0.74/22.12
0/8
0/69.2

Min/Max

8

Livestock Unit.

18
19
20

February 21, 2014

a Tropical

15
16
17

Min/Max

12
13
14

Mean (Std. Dev.)

3

Indicators

9
10
11

Pooled Sample (N = 1638)

6
7
8

Non-members (n = 1074)

4
5

Members (n = 564)

1

W3G-apce

Table 1 – Demographic characteristics of sample households

apce12035
5:22

2
GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

9

2
3

Table 2 – Geographic characteristics of sample households

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

Members (n = 564)
Indicators
Distance to whether road
Distance to nearest market
Distance to Woreda capital
Population density
Access to irrigation
Soil quality
Fertile
Mediuma
Teuf
Farmer training center

Non-members (n = 1074)

Pooled Sample (N = 1638)

Mean (Std. Dev.) Min/Max Mean (Std. Dev.) Min/Max Mean (Std. Dev.) Min/Max
55.10(73.98)
67.21(69.5)
141.60(111.86)
183.2(114.6)
0.10(0.30)

0/810
5/1080
1/810
27/652
0/1

76.63(89.57)
75.63 (72.71)
154.74(111.48)
187.4(144.4)
0.09(0.28)

0/720
5/1080
2/810
27/652
0/1

69.22(85.12)
72.73(71.71)
150.22(11.75)
185.9(134.8)
0.09(0.29)

0/810
5/080
1/810
27/652
0/1

0.19(0.39)
0.65(0.47)
0.14(0.35)
0.09(0.29)

0/1
0/1
0/1
0/1

0.34(0.47)
0.49(0.50)
0.15(0.36)
0.12(0.33)

0/1
0/1
0/1
0/1

0.29(0.45)
0.55(0.49)
0.15(0.35)
0.11(0.32)

0/1
0/1
0/1
0/1

a Medium

signifies that the land owned by the household in question is a combination of both fertile and infertile
soil qualities.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

households in input use by locations. Non-member farm households residing in cooperatives’ kebeles use a higher amount of fertilizer and improved seeds as compared to
non-members living in a kebele without agricultural cooperatives. This suggests the potential presence of a spill-over effect in input use and the presence of similar technology
among members and non-members to study efficiency gains in kebeles with agricultural
cooperatives.
As shown in Table 2, farm households that belong to agricultural cooperatives are
those located at comparatively accessible locations (closer to the nearest local markets,
closer to the nearest whether roads and Woreda amenities). This can also suggest that
most of the agricultural cooperatives in Ethiopia are found in locations that are relatively
accessible. In terms of other village level characteristics, on average, members and nonmembers are located in Peasant Associations (PAs) with similar population density and
have comparable access to irrigation and Farmer Training Centres (FTC).

4

Analytical approach

43
44
45

This paper aims at measuring the average impact of membership in agricultural
cooperatives on farm households’ technical efficiency. In other words, we estimate the
Average Treatment Effect on the Treated (ATT),8 where the treatment is membership
in agricultural cooperatives and the treated are member farmers. In such types of casual inference, the estimation of treatment effects in the absence of information on the
counter-factual poses an important empirical problem. In impact evaluation literature
this is known as the problem of filling in missing data on the counter-factual (Becker

46
47
48

8

40
41
42

49
50
51

See Becker and Ichino (2002), Dehejia and Wahba (2002), Heckman et al. (1997), Rosenbaum
and Rubin (1983), Smith and Todd (2005), and Todd (2006) for detailed methodological discussion
on estimation of Average Treatment Effect on the Treated through matching procedures. We
didn’t include equations of ATT to conserve space.

© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

10

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

and Ichino 2002, Dehejia and Wahba 2002, Heckman et al. 1997, Rosenbaum and Rubin
1985). The challenge is to find a suitable comparison group with similar covariates and
whose outcomes provide a comparable estimate of outcomes in the absence of treatment.
The empirical approach in this study is twined to reduce three potential sources of
biases in the selection of a comparison group of non-member or non-cooperative farmers.
These potential biases are common in evaluations aimed at measuring ex-post impact
of projects that involve some degree of self-selection among participants. A point in case
is given by this study, which aims to evaluate the impact of membership in agricultural
cooperatives, given that participation is voluntary and based on the intrinsic preferences,
ability and motivation of the farmers, as well as considering that no baseline (i.e., exante) observations are available to assess the performance of member-farmers before
they joined a cooperative.
The first potential source of bias is given by ‘selection on observables’, which may
arise due to sampling bias, meaning that the selection of cooperative location was notrandom but determined by spatial fixed effects (i.e., village level characteristics) and
farm households characteristics. To control for selection bias associated with the fact
that participation in cooperatives was not random, we draw from similar approaches by
Bernard et al. (2008), Francesconi and Heerink (2010) and Godtland et al. (2004), and
apply Propensity Score Matching (PSM) techniques to account for differences in observed
covariates between members and non-members. Using PSM has a great importance
in providing unbiased estimate through controlling for observable confounding factors
and in reducing the dimensionality9 of the matching problem (Becker and Ichino 2002,
Rosenbaum and Rubin 1983).
With regards to placement bias, however, we argue that Ethiopia’s past and current governance of cooperative organizations minimizes the importance of farmers’ free
will and locations resource endowments, since every kebele is expected to have at least
one cooperative and participation in cooperatives means access to publicly subsidized
inputs. Hence, in most cases the establishment of agricultural cooperatives is driven
by neither location nor farm household characteristics residing in that location, but by
centrally planned governance strategies. Further supporting our argument, Bernard
et al. (2008) assume, as we do, that cooperatives are externally formed in its PSM
analysis, and found that government and development agencies initiate 74 per cent of
cooperatives in Ethiopia. Thus, in Ethiopia cooperative placement based on kebele and/or
households’ characteristics is rather negligible.
The second source of bias in selecting a comparison group is spill-over effects.
In the presence of externalities, comparing users of cooperatives with non-users in the
same kebele can increase the possibility of having spill-over effects that underestimate
the cooperative impact. On the other hand, considering a comparison group from kebele
without cooperatives can increase differences at the kebele level (i.e., difference in agroecological conditions, infrastructure and institutions) by increasing the likelihood of
selection bias. In our empirical analysis we tried to take care of both concerns. We
first consider a sample that includes members and non-members from the ‘kebeles with
9

Propensity score methods solve the dimensionality or separateness problem through creating a single composite score from all observed covariates X, which will be used for matching
(Becker and Ichino 2002, Rosenbaum and Rubin 1983, Steiner and Cook 2012).
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

11

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

cooperatives’ and then we use the whole sample to match cooperative members with
non-members from ‘kebeles without cooperatives’ as well.
The third source of bias is ‘selection on unobservable’, which arises due to differences between members and non-members in the distribution of their unobserved
characteristics (e.g., in their ability, desire, risk preference, aspiration etc.). Given the
data available we cannot control for selection on unobservable referring to farmers’ preferences, motivation or ability. Controlling for such biases requires a suitable instrument
that explains the probability of participation in agricultural cooperatives but does not
explain their outcome. In this case, however, since we employ matching and compared
members and non-members whose propensity scores are sufficiently close or have the
same distribution, we can assume that the distribution of unobservable characteristics
is the same or at least not so different for both groups independent of membership to
induce a bias (see Becker and Ichino 2002, for a discussion). Rosenbaum bounds sensitivity analysis is used to test the sensitivity of the results to possible hidden biases due
to unobservable household characteristics when this assumption is relaxed. Furthermore, the robustness of the results is checked using alternative estimation strategy that
accounts for similar potential bias that might arise in technology selection. In this strategy the technical efficiency scores are estimated after obtaining a comparable treatment
and control groups.

4.1

Estimation of the propensity score (P-score) and matching

As indicated in the previous section we deployed propensity scoring to match members of agricultural cooperatives with similar independent farm households. Hence, we
first estimated the conditional probability of becoming a member in agricultural cooperatives (i.e., propensity score) given observed household characteristics using a flexible
Probit model, where membership status in cooperatives is the dependent variable and
covariates and their quadratic terms are introduced as independent variables.10
Although the probability of participation needs to be estimated only for households living in a kebele with cooperatives for better identification of the variables that
determine participation, we also estimated the likelihood of participation for the whole
sample to understand the existence of sufficient overlap of the covariates. At large, the
coefficients and statistical significance of the covariates are similar, except for livestock
ownership, telephone ownership and households that produce barley. We mainly used
the propensity scores based on the reduced sample to estimate the average treatment
effect on the treated for two reasons. One, the opportunity to participate exists in the
restricted sample; and two, the restricted sample is the primary focus of the analysis as
it better controls local level differences that can potentially bias the impact, tempering
possible spill-over effects that are found to be negligible.
The results from the Probit estimation are summarized in Table 3. From the
results we understand that the propensity to become a member of agricultural cooperatives is high for households with large family size, experience in farming, number of farm
plots, mobile ownership, wealth (i.e., number of ox and land), and crop types produced
10

Quadratic terms are introduced in order to account for possible non-linear relationships and
to maximize the predicting power of the model (see Godtland et al. 2004, for detailed discussion).
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

12

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3

Table 3 – Determinates of participation in agricultural cooperatives

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

Members and non-members from
cooperatives’ Kebeles (reduced
sample)
Indicators
Household size
Household size2
Gender of household head
Age of household head
Household head age2
Household head literacy
Distance to the nearest road
Distance to the nearest local market
Distance to Woreda capital
Number of farm plots
Number of crops
Household access to irrigation
Household receives off-farm income
Household owns telephone
Number of ox owned
Number of ox owned2
Livestock owned other than ox (TLU)
Hectare of land held
Hectare of land held2
Household produces Teff
Household produces wheat
Household produces sorghum
Household produces barley
Household produces maize
Household produces finger melt
Constant
Number of observations
Pseudo R2
Sensitivity (in%)
Specificity (in%)
Total correctly classified (in%)

Members and non-members from
Kebeles with and without
cooperatives (whole sample)

Coefficient

(Std. Err)

Coefficient

(Std. Err)

0.201
−0.013
−0.182
0.034
−0.001
0.408
−0.001
0.001
−0.001
0.027
−0.165
−0.060
−0.157
0.987
0.259
0.033
−0.008
0.127
−0.004
0.381
0.572
−0.177
0.170
0.155
0.643
−2.369

(0.067)∗∗∗

0.206
−0.014
−0.161
0.040
−0.001
0.404
−0.001
0.001
−0.001
0.038
−0.197
−0.085
−0.139
0.521
0.252
−0.029
−0.017
0.162
−0.006
0.444
0.662
−0.180
0.240
0.137
0.762
−2.665

(0.064)∗∗∗
(0.004)∗∗∗
(0.151)
(0.018)∗∗
(0.000)∗∗
(0.077)∗∗∗
(0.000)∗∗∗
(0.000)
(0.000)
(0.016)∗∗∗
(0.105)∗
(0.123)
(0.073)∗∗
(0.342)
(0.071)∗∗∗
(0.015)∗
(0.010)∗
(0.040)∗∗∗
(0.002)∗∗∗
(0.131)∗∗∗
(0.136)∗∗∗
(0.141)
(0.131)∗
(0.135)
(0.145)∗∗∗
(0.477)∗∗∗

(0.004)∗∗∗

(0.153)
(0.019)∗
(0.000)∗
(0.078)∗∗∗
(0.000)∗∗∗
(0.000)
(0.000)
(0.016)∗
(0.109)
(0.126)
(0.075)∗∗
(0.441)∗∗
(0.073)∗∗∗
(0.015)∗∗
(0.011)
(0.041)∗∗∗
(0.002)∗∗
(0.136)∗∗∗
(0.140)∗∗∗
(0.147)
(0.135)
(0.138)
(0.149)∗∗∗
(0.488)∗∗∗
1455
0.1464
50.00
83.73
70.65

1638
0.1861
48.58
87.52
74.11

Note: g∗∗∗ Significant at 1% level, ∗∗ significant at 5% level and ∗ significant at 10% level.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

38
39
40
41
42
43
44
45
46
47
48
49
50
51

by household (i.e., teff, wheat and finger-melt). However, after certain threshold wealth,
household size and age adversely affect probability of participation. On the other hand,
farm households that have off-farm incomes, live closer to roads, and grow diverse crops
are less likely to participate in cooperatives.
The results are more or less consistent with what has been found by Bernared
et al., (2008) as predictors of participation in cooperatives. They suggest that poorer
households without any resources (i.e., land, labour, oxen etc.) and households producing
different crops than the common cereals marketed through agricultural cooperatives are
less likely to become members. They also show that wealthy households with sufficient
experience in farming and excess owned labour will not tend to be involved in collective
action, which is consistent with theoretical predications.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

13

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

0

.2

.4

Propensity Score

Non-members: Off support
Members

.6

.8

1

Non-members: On support

Figure 1 – Distributions of the propensity scores for members (treated group) and
non-members (comparison group).
The reported density distribution is for the reduced sample that includes only members and
non-members in a kebele with agricultural cooperatives
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household
Survey, 2008.

The density distribution of propensity scores for members and non-members are
presented in Figure 1. In order to improve the robustness of the estimate, the matches
are restricted to members and non-members who have a common support11 in the distribution of the propensity score. As it can be seen in the figure, the distributions appear
with sufficient common support region that allows for matching. Besides, the difference
between members and non-members in their propensity score distribution validates
the use of matching techniques to ensure comparability. From several matching techniques applicable in impact evaluation, we use two extensively applied methods (i.e.,
non-parametric kernel based matching and five nearest neighbours matching).
The non-parametric kernel regression method is used to allow matching of members with the whole sample of non-members, since the technique uses the whole sample
of the comparison with common support to construct a weighted average match for each
treated (Heckman et al. 1997, 1998). That is, the entire sample of non-members in the
comparison group is used to construct a weighted average match to each member in
the treatment group. On the other hand, the five nearest neighbours matching is used
to match each member with the mean of the five non-members who have the closest
propensity score. The imperative of nearest neighbours matching is that it compares
non-members with scores that are closer to the scores of the members.
What is more, the validity of the matching procedure relies on the extent to which
these techniques sample or construct a comparison group that resembles the treatment
group. Besides, the balancing test within blocks that are satisfied in our estimation of
the propensity score in case of both samples (see propensity score blocks in Table A1
11

Common support refers to the values of the propensity scores where both treatment (i.e.,
members) and comparison groups (i.e., non-members) are found. 8 to 13 observations that are
off-support are dropped (Tables A3 and A4).
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

14

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

and Table A2), we undertake a ‘balancing test’ that compares a simple mean (i.e., mean
equality test) of household characteristics within the treatment group to the corresponding comparison groups created by the matching techniques before and after matching
as a complement.
As reported in Table 4, the unmatched sample fails to satisfy the balancing property. Although the groups are found to be comparable in terms of access to irrigation, age
of household head and distance to market and district administration, it shows a systematic difference between members and non-members in the majority of their observed
characteristics before matching. The balancing test results after matching that compares cooperative members to the sub-set of comparison non-members selected through
five nearest neighbours matching and kernel-based matching shows no systematic or
statistical difference in observed characteristics between the two groups. Hence, the
results suggest that our comparison is valid from statistical point of view.

4.2

Measuring technical efficiency

The technical efficiency measure is intended to capture whether agricultural cooperatives enable their members in getting better access to productive inputs and services
including training on better farming practices that enhance their productive efficiency.
The stochastic frontier production model12 is used to estimate the technical efficiency
of sample households. It measures the ability of households to obtain maximum possible outputs from a given set of inputs (Coelli et al. 2005, Farrell 1957, Kumbhakar
and Lovell 2000). Such a measure is of great importance in estimating the household
efficiency score by accounting for factors beyond the control of each producer. Besides, it
helps to understand the factors that determine technical inefficiency of farm households,
since some of the factors can be influenced by policies.
Following this approach we first detected the presence of inefficiency in the
production for sample households. Estimating the stochastic production frontier and
conducting a likelihood-ratio test assuming the null hypothesis of no technical inefficiency on input-output data carried out the test. The result shows that the inefficiency
12

Unlike the deterministic approach, it is a model that incorporates household-specific random shocks that represents statistical noises due to factors beyond the control of households,
measurement errors and omission of relevant variables (Coelli et al. 2005, Kumbhakar and Lovell
2000). In other words, in stochastic production frontier the error term is composed of the symmetric error component and the technical inefficiency component that measures shortfall of output
from its maximum frontier or possible output. Hence, in this approach technical efficiency is measured as the ratio of observed output to maximum attainable output in a context characterized by
household specific random shocks (i.e., exp{V j }):
TE j =

Yj
f (X j , β).exp{V j }

Where, refers to the technical efficiency of the jth producer, Yj is the observed output, indicates
the deterministic part that is common to all producers or households, exp{V j }is a producers specific part that captures the effect of random noises or shocks on each producer. See Aigner et al.
(1977), Coelli et al. (2005), Jondrow et al. (1982), Kumbhakar and Lovell (2000), and Meeusen and
Ven den Broeck (1977) for detailed methodological discussions.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

32
33
34

35
36
37

38
39

40
41
42

43
44
45

46
47
48

49
50
51
44.80
72.11
68.26
148.58
5.38
2.42
0.08
0.61
0.002
1.22
2.80
1.51

45.76
55.10
67.21
141.6
6.37
2.75
0.10
0.55

© 2014 The Authors

Annals of Public and Cooperative Economics © 2014 CIRIEC

0.019
1.71
3.34
2.06
0.000

0.000
0.008

0.001

0.014

0.000
0.783
0.249
0.000
0.000
0.367

0.169

0.000

0.000
0.000

2.05

1.70
3.32

0.014

0.55

55.20
67.16
142.2
6.35
2.74
0.10

45.81

0.45

6.50
1.04

2.06

1.68
3.14

0.011

0.57

57.51
71.43
140.75
6.33
2.76
0.08

44.95

0.48

6.45
1.04

Non-members

0.929

0.711
0.972

0.672

0.492

0.590
0.357
0.828
0.942
0.764
0.328

0.239

0.320

0.676
0.834

Diff: P-value

2.05

1.70
3.32

0.014

0.55

55.20
67.16
142.2
6.35
2.74
0.10

45.81

0.45

6.50
1.04

Members

1.30

1.68
3.43

.017

0.56

57.7
69.95
143.56
6.21
2.72
0.09

45.41

0.46

6.46
1.03

Non-members

0.869

0.693
0.631

0.627

0.744

0.562
0.523
0.837
0.432
0.741
0.481

0.585

0.768

0.775
0.799

Diff: P-value

Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

February 21, 2014

a The reported balancing test is for the reduced sample (i.e., sample 1) that includes only members and non-members in a kebele with agricultural cooperatives.
We did similar tests for the full sample and the balancing properties are satisfied.
b Livestock owned (TLU) refers to livestock other than ox owned by the household.

0.25

29
30
31

0.45

26
27
28

Members

15
16
17

Diff: P-value

4
5

Non-members

3

6.03
1.10

23
24
25

W3G-apce

6.50
1.04

21
22

Household size
Gender of HH head (1
= Male, 2 = Female)
Household head
literacy (1 = Yes, 0 =
No)
Age of household head
Distance (minutes)
To the nearest road
To the nearest market
To Woreda capital
Number of plots held
No. of crops planted
Access to irrigation (1
= Yes, 0 = No)
Off-farm income (1 =
Yes, 0 = No)
Own telephone (1 =
Yes, 0 = No)
Number of ox owned
Livestock owned
(TLU)b
Size of farm land (ha)

18
19
20

Members

12
13
14
Kernel-based matching

9
10
11

Five nearest neighbours matching

6
7
8

Unmatched samples

1

Table 4 – Balancing test of matched samplea

apce12035
5:22

2
IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY
15

apce12035

1

16

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

component of the error term is significantly different from zero, which indicates the
presence of a statistically significant inefficiency component (i.e., Ho : Sigma_u = 0 is
rejected). The lambda (λ) value is also greater than one, indicating the significance of
inefficiency. Moreover, the value of gamma indicates that there is a 70 per cent variation
in output due to technical inefficiency. In other words, the technical inefficiency component is likely to have an important effect in explaining output among farm households
in the sample.
Once we detected the presence of technical inefficiency, we estimate a one-stage
simultaneous maximum likelihood estimate for the parameters of the Cobb-Douglas13
stochastic frontier production function to predict households’ technical efficiency scores
and to understand determinants of inefficiency. As expected, all conventional inputs
(land, labour, fertilizer, seed and number of oxen owned) are found to be significant
determinates of household production (Table 5). In particular, landholding size and
number of oxen owned are found to be the major input variables that affect output
considerably. Overall, the return to scale shows that farmers in our sample are operating
under increasing return to scale, suggesting that size may matter in the efficiency of
smallholder farmers. This result is expected in smallholder farms context and consistent
with prior studies in Ethiopia by Asefa (2012) and Haji and Andersson (2008), among
others.
The inefficiency model suggests that inefficiency of farm households is significantly linked with number of plots, diversification of crops, gender of household
head and membership in agricultural cooperatives.14 Overall, the above results are
in line with the findings of Alemu et al. (2009), Idiong (2007), and Jaime and Salazar
(2011) and comparable to the results obtained from the alternative strategy that estimate the technical efficiency scores using matched group of member and non-member
farmers.
With regard to membership in agricultural cooperatives, the result indicates that
membership reduces technical inefficiency by about 5 per cent (Table 5). Concurrently,
from the descriptive statistics we understood that the mean technical efficiency of members is significantly higher than that of non-members (i.e., 71 and 62 per cent, respectively) and the majority of the members are above the mean efficiency (i.e., 65 per cent)
of the pooled sample (Figure. 2). Besides, as is clear from Figure 2, the density of nonmembers is above that of the members on the distribution below the mean efficiency of
the whole sample. However, we cannot draw any conclusion at this stage as this difference can be partially or totally due to original differences among households. Thus,
we use matching that computes the average difference in technical efficiency scores between members and non-members in the common support region using the techniques
described above.

43
44
45
46
47
48
49
50
51

13

Cobb–Douglas stochastic frontiers are found to be adequate representations of our data as
compared to the specifications of the translog stochastic frontiers.
14 The coefficient of membership in agricultural cooperatives obtained from the inefficiency
model is comparable to the average impacts of cooperative membership on technical efficiency
resulted from matching estimators.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

17

2
3

Table 5 – Maximum Likelihood (ML) estimates of the parameters for Stochastic
Production Frontier (SPF) function and technical inefficiency determinants

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

Dependent variable: production value in Birr (logged)

Production function
ln (Land size held by household (ha))
ln (Seed used (Kg))
ln (Fertilizer used (Kg))
ln (Labor (hired in number of days))
ln (Number of oxen owned)
Constant
Return to scale (sum of elasticises)
Technical inefficiency component
Household size
Gender of household head
Age of household head
Household head read and write
Distance to local market
Number of plots held
Number of crops planted
Household access to irrigation
Household receives off-farm income
Membership in cooperatives
Household access to institutional credit
Constant
Diagnostic statistics
Sigma_v
Lambda
Gamma (γ = λ2 /(1+ λ2 )
Number of observation
Wald chi2 (5)
Prob > chi2
Log likelihood function
Likelihood-ratio test of Sigma_u = 0: chibar2(01)

Coefficient

(Std. Err.)

1.174
0.071
0.036
0.051
0.472
6.327
1.804

(0.063)∗∗∗
(0.017)∗∗∗
(0.009)∗∗∗
(0.014)∗∗∗
(0.042)∗∗∗
(0.101)∗∗∗

0.023
0.726
−0.004
−0.231
0.001
0.106
−0.620
−2.800
0.152
−0.512
0.053
−0.567

(0.026)
(0.204)∗∗∗
(0.004)
(0.148)
(0.001)∗
(0.028)∗∗∗
(0.135)∗∗∗
(1.219)∗∗
(0.141)
(0.176)∗∗∗
(0.162)
(0.439)

0.600
1.556
0.707
1638
1567.38
0.0000
−1871.810
24.80

(0.032)∗∗∗
(0.091)∗∗∗

Note: ∗∗∗ Significant at 1% level, ∗∗ significant at 5% level and ∗ significant at 10% level.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

5

Results and discussion

5.1

Average impact of agricultural cooperatives on technical efficiency

As described in the above sections, the average impact of cooperative membership
on the technical efficiency of small farmers is analysed using the reduced sample (i.e.,
sub-sample 1) that includes members and non-members from kebeles with agricultural
cooperatives and the whole sample that aimed at accounting for possible spill-over effects
(i.e., sample 2). The resulting non-parametric estimate of the Average Treatment Effect
on the Treated (ATT), average impact of membership in agricultural cooperatives on
the technical efficiency of smallholder farmers, based on the Propensity Score Matching
(PSM) methods, is reported in Table 6.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

18

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

Figure 2 – Frequency distribution of technical efficiency scores by cooperative
membership.
The reported frequency distribution is for the reduced sample (i.e., sample 1) that includes only
members and non-members in a kebele with agricultural cooperatives.
Note: TE refers to Technical Efficiency score of households.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household
Survey, 2008.

The paper mainly used the analysis based on the reduced sample as it accounts
for differences in technology and agro-ecology that can affect efficiency estimation. On
the other hand, the impact estimate based on the whole sample aimed at examining the
extent of spill-over effects. As is clear from Table 6, the diffusion effect is found to be
negligible. Meaning, the impact estimate based on the whole sample is lower15 than the
impact estimate based on the reduced sample where the possibility of diffusion effects
exists.
Consistent with the results from the descriptive statistics and the inefficiency
model of the stochastic frontier function, we found that, on average, farmers belonging
to agricultural cooperatives are more efficient than independent farmers. The results
suggest that member households are in a better position to obtain maximum possible
outputs from a given set of inputs used, by about 5 percentage points, in line with the expectation that agricultural cooperatives likely make productive technologies accessible
and provide embedded support services (i.e., training, information and extension linkages). The impact estimates are robust across different estimation methods and samples
considered. We further checked the robustness of the estimates for a specific region (i.e.,
Amhara Region), where the size of the sample allows for using matching techniques. The
results are comparable to the results from the reduced and the whole sample (i.e., about
a 5.5 per cent and 4.5 percentage points difference for kernel based and five neighbours
matching, respectively).

15

Lower average impact from the whole sample that include non-cooperative kebeles can also
indicate the presence of technology difference between cooperative and non-cooperative kebeles,
strengthening our decision to focus on cooperative kebeles in order to reduce potential differences
in technology, as it should be accounted to compare differences in technical efficiency due to
cooperative membership.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

19

2
3

Table 6 – Effect of cooperative membership on technical efficiency of smallholders

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

Kernel-based matching
ATT

Std. Err.
(0.008)∗∗∗

Five nearest neighbors matching
ATT

Reduced sample: (%
5.64
5.70
Difference in TE)
Whole sample: (% Difference
5.42
(0.009)∗∗∗
4.55
in TE)
Check for robustness: observations limited to Amhara region only
Reduced sample
4.82
(0.012)∗∗∗
4.11
Whole sample
5.30
(0.010)∗∗∗
4.02

Std. Err.

Number of Obs.

(0.010)∗∗∗

1455

(0.010)∗∗∗

1638

(0.011)∗∗∗
(0.012)∗∗∗

385
431

Note: Reduced sample includes members and non-members only from kebeles with agricultural cooperatives;
Whole sample includes the whole sample (i.e., members and non-members from kebeles with and without agricultural cooperatives). TE refers to households’ Technical Efficiency score. Bootstrap with 100 replications is used
to estimate the standard errors.
∗∗∗ Significant at 1% level.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

Nonetheless, the above results rely heavily on the assumption of unconfoundeness
or conditional independence16 (i.e., once the factors affecting participation are taken into
account, the condition of randomization restored) and are not robust against ‘hidden
bias’. If there are unobserved variables which affect participation in cooperatives and
technical efficiency simultaneously, unobserved heterogeneity affecting the robustness
of the estimates might arise (Becker and Caliendo 2007, Keele 2010, Rosenbaum 2002,
Rosenbaum and Rubin 1983).
We assess the presence of this problem using Rosenbaum bounds sensitivity analysis when the key assumption is relaxed by a quantifiable increase in uncertainty. As
reported in Table 7, the results are found to be insensitive to a bias that would double
the odds of participation (self-selection) in agricultural cooperatives but sensitive to bias
that would triple the odds. The magnitude of hidden bias, which would make our finding
of a positive and significant effect of membership in agricultural cooperatives on technical efficiency questionable or spurious, should be higher than = 2.5 and = 2.6 for
the reduced sub-sample and whole sub-sample, respectively. Hence, we deduce that the
strength of the hidden bias should be sufficiently high to undermine our conclusion of
positive and significant impact of membership in agricultural cooperatives on technical
efficiency based on the matching analysis.

5.2

Robustness check

43
44
45

Besides the Rosenbaum bounds sensitivity analysis for hidden bias presented in
Table 7, we check the robustness of the results following alternative estimation strategy

46
47
48

16

49
50
51

Unconfoundedness in our case means that participation in agricultural cooperatives does
not depend on households’ technical efficiency, after controlling for the variations in technical efficiency induced by differences in observable covariates. It is a strong assumption that implies that
participation is based on observable characteristics and that variables simultaneously influencing
participation and technical efficiency are observable.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

20

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

Table 7 – Rosenbaum Bounds sensitivity analysis for hidden bias
Critical value of hidden bias ( )
1
1.10
1.20
1.30
1.40
1.50
1.60
1.70
1.80
1.90
2
2.10
2.20
2.30
2.40
2.50
2.60
2.70
2.80
2.90
3

TE (Sample 1) Sig+ (max)

TE (Sample 2) Sig+ (max)

<0.0000001
<0.0000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
0.000011
0.000085
0.000489
0.002134
0.007333
0.020519
0.048091
0.09674
0.170595
0.268689
0.384324
0.506814
0.624664

<0.0000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
<0.000001
0.000012
0.000084
0.000443
0.001824
0.006039
0.016554
0.038524
0.077759
0.1387
0.222264
0.32474
0.43839

Note: Reduced sample includes members and non-members only from kebeles with agricultural cooperatives;
Whole sample includes the whole sample (i.e., members and non-members from kebeles with and without agricultural cooperatives). TE refers to households’ technical efficiency score.
The sensitivity analysis is for one-sided significance levels. measures the degree of departure from random
assignment of treatment or a study free of bias (i.e., = 1).
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

used by Mayen et al. (2010) and Crespo-Cebada et al. (2013) to address the same problem
of correcting potential selection bias in measuring technical efficiency difference between
two groups using PSM. In this approach the stochastic frontier model is estimated on
sub-samples of cooperative non-members and members that are obtained from PSM. The
strategy is aimed at addressing potential bias that may arise in estimating technical
efficiency scores using unmatched samples, as the technology use can be affected by the
same selection bias like that of membership in cooperatives.
Thus, before estimating the technical efficiency scores, we constructed statistically
comparable non-members using PSM. Single-nearest-neighbour matching technique is
used to pair each cooperative member with a non-member that has the closest propensity
score.17 Figure 3 shows the distribution of the propensity score for sub-sample members
and non-members obtained from the matching. As expected, the propensity score distribution of the PSM sub-sample of non-members closely resembles that of members
in terms of their propensity to membership, compared to the distribution in Figure 1.
Furthermore, as it is a matched sub-sample, there are no farm households that are
off-support in either of the groups (Figure 3).
17

Similar probability model and specification presented in section 4.1 and Table 3 is used to
estimate the propensity scores.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

21

2
3

Table 8 – Means and standard deviations of technical efficiency: PSM sub-sample

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

Members

Reduced sample
Whole sample

Non-members

Mean

Std. Err.

Mean

Std. Err.

Difference in Means

68.37
67.17

0.58
0.60

61.08
62.03

0.74
0.73

7.29∗∗
5.13∗∗

Note: Reduced sample includes members and non-members only from kebeles with agricultural cooperatives;
Whole sample includes the whole sample (i.e., members and non-members from kebeles with and without agricultural cooperatives).
∗∗ Significant at 1% level.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

Next we estimated the technical efficiency scores of the farm households using
stochastic frontier model on the two different sub-samples obtained from PSM (i.e.,
PSM sub-sample that include members and non-members in cooperative kebeles and
PSM sub-sample that also include non-members in non-cooperative kebeles). The results
from the stochastic frontier analysis are presented in Table 8.18 For the whole sample
we found the technical efficiency of cooperative members to be 67.17, which is 5.13 percentage points higher than for non-members. When we account for potential technology
differences across locations by restricting the sample to farm households only living in
cooperative kebeles, we found that cooperative members 7.29 per cent more efficient
compared to non-members. Overall, the 5 to 7 percentage points efficiency gap found
from alternative estimation strategy is comparable with the results obtained from ATT
reported in Table 6.
In all, although the magnitude or economic significance is not as high as expected,
the results obtained from the two alternative estimation strategies suggested that participation in agricultural cooperatives resulted in technical efficiency gains among smallholder farmers. We consider that this efficiency difference can be due to greater benefit
of agricultural cooperatives in farm technology/inputs adoption by lowering costs and
improving members’ access to productive inputs and services (Abebaw and Haile 2013,
Getnet and Tsegaye 2012). As presented in Table A3, we also found considerable impact
of cooperatives membership in use of farm inputs (i.e., fertilizer and improved seeds).
Moreover, benefits of cooperatives in linking smallholders to extension services can be
also the sources of this efficiency gaps between members and non-members, as recent
study by Rodrigo (2012) found a positive effect of agricultural cooperatives in increasing farmers involvement in agricultural extension programs in Ethiopia that results in
productivity growth among members.
5.3

Impact heterogeneity

The above results obtained from the alternative estimation strategies assume
a homogenous treatment effects among cooperative member households. However,
18

As indicated in section 4.2 the coefficients of the production parameters, inefficiency correlates and diagnostic statistics obtained from the SPF estimation using the matched sample
are more or less similar to the one resulted from the estimation based on the whole unmatched
sample.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

22

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3

Reduced sample

4
5
6
7
8
9
10
11
12
13
14
15
16
17

0

.2

.4

Propensity Score

.6

Non-member: On support

18
19
20

.8

1

Member: On support

Whole sample

21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

0

.2

.4

Propensity Score

Non-member: On support

.6

.8

1

Member: On support

Figure 3 – Distributions of the propensity scores for members and non-members: PSM
sub-sample.
Note: Reduced sample includes members and non-members only from kebeles with
agricultural cooperatives; Whole sample includes members and non-members from kebeles
with and without agricultural cooperatives.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household
Survey, 2008.

treatment impacts can vary within cooperative members, as households are distinct
in their socio-economic realities. In order to understand potential impact heterogeneity
within members, we graph the distribution of cooperatives’ impact on members level
of technical efficiency using the results obtained from Kernel matching estimates (i.e.,
the difference between actual observed technical efficiency and corresponding matched
values obtained from the estimation of ATT).
While the impacts are normally distributed, we observe some variations of
membership impact on technical efficiency within members across the two samples
(Figure A1). For large proportion of members, involvement in cooperatives results in
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

23

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

about 5–15 per cent efficiency gains as compared to non-members. For the remaining
few member households we notice both efficiency gains and losses ranging from 20–40
per cent as compared to their counterparts. We further regress technical efficiency gains
due to membership in cooperatives obtained from Kernel matching estimates by household characteristics, with the purpose of understanding the determinates or correlates
of observed impact variations within members.
The results from the regression suggests that the impact of membership in cooperatives on technical efficiency significantly increases with cultivated land size, application
of improved seeds and access to irrigation and farmer training centre and decreases with
distance to market, off-farm income and sex of household head (Table A4). It implies that
technical efficiency gains from cooperative membership is better responsive for member
households with large and irrigated land holding and resides in villages with farmer
training centres. The lower impact of cooperatives membership for members away from
local market on the other hand can be due to higher costs of accessing the services
provided by the cooperatives, as most of the cooperatives in Ethiopia are located closer
to nearest markets (Bernard et al 2013). Conversely the results indicate that household head literacy, access to media, as measured by radio ownership and application of
fertilizer does not explain variations in efficiency gains within members.

6

Conclusions

Over the past decade and a half, agricultural cooperatives in Ethiopia have
strongly promoted as instrument to transform subsistence agriculture by preserving
market options and increasing farmers’ income, as they are believed to be efficient
in internalizing transaction costs, reducing the variability of farmers’ income through
risk pooling and countervailing opportunistic behaviours (Hogeland 2006, Staatz 1987).
Though many variations in the agricultural cooperatives model can be distinguished,
typical agricultural cooperatives in Ethiopia combine both agricultural supply and marketing activities. Currently, agricultural cooperatives market more than 10 per cent of
farmers’ produce and supply farm inputs for all farm households irrespective of membership. Although their share in input and output marketing shows how vibrant the
cooperatives are in supporting agricultural transformation, empirical studies on their
efficiency and productivity impacts are very limited.
Using household data drawn from the Ethiopia Agricultural Marketing Household
Survey in 2008, this paper aims to understand the impact of membership in agricultural
cooperatives on technical efficiency in a context where membership incentives can result
in efficiency gains. We assume that the establishment of cooperatives in Ethiopia has
been independent of community and household level characteristics due to negative experiences in the past and current policies on cooperative formation (i.e., one cooperative
for each kebele). Moreover, we assume that difference in technology between members
and non-members is insignificant, as agricultural cooperatives in Ethiopia are required
to supply basic farm inputs for all farm households. In addition, the role of spill-over
effects cannot be underestimated. With these assumptions, we used Propensity Score
Matching techniques to compare the average technical efficiency difference between cooperative member households and independent farm households living within the same
kebele in which agricultural cooperatives operate.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

24

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Our results consistently indicate a positive and significant impact of agricultural
cooperatives on members’ levels of technical efficiency. On average members are better
situated to get maximum possible output from a given set of inputs used, by at least 5
per cent. These results are in line with the predicted role of agricultural cooperatives
in improving efficiency by providing easy access to productive inputs and embedded
support services such as training, information, and extension on input application. The
robustness of the findings is demonstrated by similar results obtained from different
approaches and techniques. However, as compared to the results of the descriptive
statistics, the impact based on the average treatment effect is lower, which indicates the
existence of variation or heterogeneity across households within members.
In general, the efficiency gains from membership in agricultural cooperatives
emerged from the analysis has important policy implications. It suggests that besides their progressive role in input and output marketing, agricultural cooperatives
in Ethiopia are effective in providing embedded supportive services, significantly contributing to members’ technical efficiency. Therefore, promoting agricultural cooperatives as complementary institutions to public extensions services should further enhance
smallholders’ technical efficiency.

21
22
23
24
25
26
27
28
29
30
31

REFERENCES

ABEBAW D. and HAILE M.G., 2013, ‘The impact of cooperatives on agricultural technology adoption: Empirical evidence from Ethiopia’, Food Policy, 38, 82–91.
ABEGAZ G., 1994, ‘Rural land use issues and policy: overview’, in D. RAHMATO (ed.)
Land Tenure and Land Policy in Ethiopia After the Derg, Proceedings of the Second Workshop of the Land Tenure Project WP-8, the Centre for Environment and
Development, University of Trondheim, Norway, pp. 21–34.

32
33
34

AIGNER D.J., LOVELL C.A.K. and SCHMIDT P., 1977, ‘Formulation and estimation
of stochastic frontier production function models’, Journal of Econometrics, 6, 21–37.

35
36
37

ALEMU B.A., NUPPENAU E.A. and BOLLAND H., 2009, ‘Technical efficiency across
agro-ecological zones in Ethiopia: The impact of poverty and asset endowments’, Agricultural Journal, 4, 202–207.

38
39

ASEFA, S., 2012. Analysis of technical efficiency of crop producing smallholder farmers
in Tigray, Ethiopia, MPRA Paper No. 40461.

40
41
42

BECKER S. and CALIENDO M., 2007, ‘Sensitivity analysis for average treatment effects’, The Stata Journal, 7, 71–83.

43
44
45
46
47
48
49
50
51

BECKER S. and ICHINO A., 2002, ‘Estimation of average treatment effects based on
propensity scores’, The Stata Journal, 2, 358–377.
BERNARD, T., GASHAW, T.A. and SOLOMON, L., 2013, ‘Agricultural cooperatives
in Ethiopia: Results from the 2012 ATA baseline survey’, International Food Policy
Research Institute, Washington DC.
BERNARD, T. and SPIELMAN D., 2009, Reaching the rural poor through rural producer
organizations: A study of agricultural marketing cooperatives in Ethiopia, Food Policy,
34, 60–69.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

25

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

BERNARD T., SPIELMAN D., TAFFESSE A.S., and GABRE-MADHIN E., 2010, Cooperatives for Staple Crop Marketing: Evidence from Ethiopia, Washington, DC: International Food Policy Research Institute (IFPRI) Research Report 166.
BERNARD T., TAFFESSE A.S. and GABRE-MADHIN E., 2008, ‘Impact of cooperatives
on smallholders’ commercialization behaviour: evidence from Ethiopia’, Journal of
Agricultural Economics, 39, 147–161.
BINAM J.N., TONYE J. and WANDJI N., 2005, ‘Source of technical efficiency among
smallholder maize and peanut farmers in the Slash and Burn agricultural zone of
Cameroun’, Journal of Economic Cooperation, 26, 193–210.
CHADDAD F., COOK M.L. and HECKELEI T., 2005, ‘Testing for the presence of financial constraints in US agricultural cooperatives: An investment behaviour approach’,
Journal of Agricultural Economics, 56, 385–397.
CHIRWA E.W., 2003, Sources of technical efficiency among smallholder maize farmers
in Southern Malawi, Working Paper 01, Wadonda Consultancy.
COELLI T., RAO D.S.P., O’DONELL C.J. and BATTESE G.E., 2005, An Introduction to
Efficiency and Productivity Analysis, New York, NY, Springer.
COULTER J., GOODLAND A., TALLONTIRE A. and STRINGFELLOW R., 1999, Marrying farmer cooperation and contract farming for service provision in a liberalizing
Sub-Saharan Africa, ODI: Natural Resources Perspectives 48.
CRESPO-CEBADA, E., PEDRAJA-CHAPARRO, F. and SANTIN, D., 2013, ‘Does school
ownership matter? An unbiased efficiency comparison for region of Spain’, Journal of
Productivity Analysis (forthcoming-published online on March, 2013).
DAVIS K., 2008, ‘Extension in Sub-Saharan Africa: Overview and assessment of past
and current models and future prospects’, Journal of International Agricultural and
Extension Education, 15 (3), 15–28.
DEHEJIA R. and WAHBA S., 2002, ‘Propensity score-matching methods for nonexperimental causal studies’, Review of Economics and Statistics, 84, 151–161.
DEININGER K., 1995, Technical Change, Human Capital, and Spillovers in United
States Agriculture 1949–1985, New York, NY, Garland Publishing.
FARRELL M.J., 1957, ‘The measurement of productive efficiency’, Journal of the Royal
Statistical Society, 120, 253–282.
FRANCESCONI G.N. and HEERINK N., 2010, ‘Ethiopian agricultural cooperatives in
an era of global commodity exchange: Does organizational form matter?’ Journal of
African Economies, 20, 1–25.
FRANCESCONI G.N. and RUBEN R., 2007, Impacts of collective action on smallholders’
commercialization: Evidence from dairy in Ethiopia’, paper presented at the I Mediterranean conference of agro-food social scientists. 103rd EAAE Seminar ‘Adding Value
to the Agro-Food Supply Chain in the Future Euromediterranean Space’, Barcelona,
Spain.
GETNET K. and TSEGAYE A., 2012, ‘Agricultural cooperatives and rural livelihoods:
Evidence from Ethiopia’, Annals of Public and Cooperative Economics, 83, 181–198.
GODTLAND E.M., SADOULET E., DE JANVRY A., MURGAI R. and ORTIZ O., 2004,
‘The impact of farmer field-schools on knowledge and productivity: a study of potato
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

Q2

apce12035

1

26

W3G-apce

February 21, 2014

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

farmers in the Peruvian Andes’, Economic Development and Cultural Change, 53,
63–92.
HAJI, J. and ANDERSSON, H., 2008, Determinants of efficiency of vegetable production in smallholder farms: the case of Ethiopia, Food Economics-Acta Agriculturale
Scandinavica C (3–4), 125–137.
HECKMAN J., ICHIMURA H., SMITH J. and TODD P., 1998, ‘Characterizing selection
bias using experimental data’, Econometrica, 66, 1017–1098.
HECKMAN J., ICHIMURA H., SMITH J. and TODD P., 1997, ‘Matching as econometric
evaluation estimator: evidence from evaluating a job-training program’, Review of
Economic Studies, 64, 605–654.
HOGELAND J.A., 2006, ‘The economic culture of U.S. agricultural cooperatives’, Culture
and Agriculture, 28, 67–79.
IDIONG I.C., 2007, ‘Estimation of farm level technical efficiency in small-scale swamp
rice production in cross-river state of Nigeria: A stochastic frontier approach’, World
Journal of Agricultural Sciences, 3, 653–658.
JAIME M.M. and SALAZAR C.A., 2011, ‘Participation in organization, technical efficiency and territorial differences: A study of small wheat farmers in Chile’, Chilean
Journal of Agricultural Research, 71, 104–113.
JONDROW J., LOVELL C.A.K., MATEROV I. and SCHMIDT P., 1982, ‘On the estimation of technical inefficiency in the stochastic frontier production model’, Journal of
Econometrics, 19, 233–238.
KEELE L., 2010, An overview of rbounds: an R package for Rosenbaum bounds
sensitivity analysis with matched data. Accessed June 2012, available at
http://www.personal.psu.edu/ljk20/rbounds%20vignette.pdf.
KODAMA Y., 2007, ‘New role of cooperatives in Ethiopia: The case of Ethiopian coffee
farmers’ cooperatives’, African Study Monographs, 35, 87–108.

32
33
34

KUMBHAKAR S.C. and LOVELL C.A.K., 2000, Stochastic Frontier Analysis, New York,
NY: Cambridge University Press.

35
36
37

MAYEN, C.D., BALAGATS, J.V. and ALEXANDER, C.E., 2010, ‘Technology adoption
and technical efficiency: organic and conventional dairy farms in the United Sate’,
American Journal of Agricultural Economics, 92 (1), 181–195.

38
39

MEEUSEN W. and VAN DEN BROECK J., 1977, ‘Efficiency estimation from CobbDouglas production functions with composed error’, International Economic Review,
18, 435–444.

40
41
42
43
44
45
46
47
48
49
50
51

MoFED (Ministry of Finance and Economic Development), Ethiopia (2006) Ethiopia:
building on progress – A Plan for Accelerated and Sustained Development to End
Poverty (PASDEP) (2005/06–2009/10), Volume I (Main Text), Addis Ababa, Ethiopia.
PINGALI P., KHWAJA Y. and MEIJER M., 2005, Commercializing small farms: reducing transaction costs, ESA Working Paper 05–08: The Food and Agriculture Organization (FAO).
RAHMATO D., 1990, Cooperatives, state farms and smallholder production, in S. Pausewang, Fantu Cheru, S. Brune, and Eshetu Chole (eds) Ethiopia: Rural Development
Options, pp. 100–110, Zed Books, London.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

27

2
3
4
5
6
7
8

RAHMATO D., 1993, ‘Agrarian change and agrarian crisis: state and peasantry in postrevolution Ethiopia’, Journal of International African Institute, 63, 36–55.
RAHMATO D., 1994, Land, peasants and the drive for collectivization in Ethiopia, in
T.J. Bassett and D.E. Crummey (eds) Land in African Agrarian System, Madison,
University of Wisconsin Press, pp. 274–297.

9
10
11

RODRIGO M.F., 2012, Do cooperatives help the poor? Evidence from Ethiopia’, Paper
presented at the Agricultural and Applied Economics Association’s (AAEA) annual
meeting, Seattle, Washington DC.

12
13
14

ROSENBAUM P.R., 2002, Observational Studies (2nd ed.), New York, NY: Springer.

15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

ROSENBAUM P.R. and RUBIN D.B., 1983, ‘The central role of the propensity score in
observational studies for casual effects’, Biometrika, 70 (1), 41–55.
ROSENBAUM P.R. and RUBIN D.B., 1985, ‘Constructing a control group using multivariate matched sampling methods that incorporate the propensity score’, American
Statistician, 39, 33–38.
SMITH J. and TODD P., 2005, ‘Does matching overcome Lalonde’s critique of nonexperimental estimators?’ Journal of Econometrics, 125, 305–353.
SPIELMAN D.J., KELEMWORK D. and ALEMU D., 2011, Seed, fertilizer, and agricultural extension in Ethiopia, Working Paper 020, International Food Policy Research
Institute: Ethiopia Strategic Support Program (IFPRI-ESSP).
STAATZ J.M., 1989, Farmer cooperative theory: recent developments, Research Report
84, U.S. Department of Agriculture, Agricultural Cooperative Service (ACS).
STAATZ J. M., 1987, Farmers’ incentives to take collective action via cooperatives:
transaction costs approach, in J. Royer (ed.) Cooperative Theory: New Approaches,
USDA-ACS Service Report 18, pp. 87–107.
STEINER P.M. and COOK D.L., 2012, Matching and propensity scores, chapter forthcoming, in: Little, T. D., ed., The Oxford Handbook of Quantitative Methods 1. Accessed June 2012, available at http://itp.wceruw.org/
Spring%2011%20seminar/Steiner%201.pdf.
TIEGIST L., 2008, Growth without structures: the cooperative movement in Ethiopia,
in P. Develtere, I. Pollet and F. Wanyama (eds) Cooperating out of Poverty: the Renaissance of the African Cooperative Movement, International Labor Office and World
Bank Institute pp. 128–152.
TODD P.E., 2006, Matching estimators’, Accessed June 2012, available at
http://athena.sas.upenn.edu/˜petra/papers/mpalgrave2.pdf.
VENKATESAN V. and KAMPEN J., 1998, Evolution of agricultural services in SubSaharan Africa: trends and prospects, discussion paper 390, Washington DC: The
World Bank.
WANYAMA F.O., DEVELTERE P. and POLLET I., 2009, ‘Reinventing the wheel?
African cooperatives in a liberalized economic environment’, Annals of Public and
Cooperative Economics, 80 (3), 361–392.
WORLD BANK, 2008, World Development Report: Agriculture for development, World
Bank, Washington, DC: World Bank.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

28

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3

Appendix

4
5

03

15
16
17

02

18
19
20

32
33
34
35
36
37
38
39
40
41
42
43
44
45

0

20

40

04

(b) Whole sample: members and non-members from Kebeles with and without
agricultural cooperatives

03

29
30
31

-20

02

26
27
28

-40

Percentage difference in technical efficiency between observed and estimated

01

23
24
25

0

21
22

0

12
13
14

01

9
10
11

(a) Reduced sample: members and non-members only from Kebeles with agricultural
cooperatives

04

6
7
8

-40

-20

0

20

40

Percentage difference in technical efficiency between observed and estimated

Figure A1 – Distribution of cooperative membership impacts based on the results from
Kernel matching estimates.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household
Survey, 2008.

46
47
48
49
50
51
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

5:22

IMPACT OF AGRICULTURAL COOPERATIVES ON SMALLHOLDERS’ TECHNICAL EFFICIENCY

29

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

Table A1 – Propensity scores blocks for members and non-members in Kebeles with
agricultural cooperatives (only observations within common support) –reduced sample
Block of Pscore
0.026
0.2
0.3
0.4
0.45
0.5
0.6
0.7
0.8
Total

Members

Non-members

Total

43
60
96
37
46
92
82
67
41
564

248
196
174
73
47
76
46
19
4
883

291
256
270
110
93
168
128
86
45
1447

Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

18
19
20

Table A2 – Propensity scores blocks for members and non-members in Kebeles with and
without agricultural cooperatives (only observations within common support) –whole
sample

21
22

Block of Pscore

23
24
25
26
27
28
29
30
31

0.015
0.2
0.3
0.4
0.5
0.6
0.8
Total

Members

Non-members

Total

54
65
97
76
76
149
47
564

448
206
153
120
68
58
8
1061

502
271
250
196
144
207
55
1625

Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

1

W3G-apce

February 21, 2014

30

5:22

GASHAW TADESSE ABATE, GIAN NICOLA FRANCESCONI AND KINDIE GETNET

2
3

Table A3 – Average impact of cooperative membership on agricultural input adoptions

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

Kernel-based
matching
Indicator

Reduced
sample
Fertilizer (total
amount in kg)
Fertilizer (kg/ha)
Improved seed
(total amount
in kg)
Whole sample
Fertilizer (total
amount in kg)
Fertilizer (kg/ha)
Improved seed
(total amount
in kg)

Five nearest neighbors
matching

ATT

Std.
Err.

ATT

Std. Err.

Number of Obs.

48.66

(6.74)∗∗∗

49.55

(7.73)∗∗∗

1455

31.32
4.45

(4.88)∗∗∗
(1.22)∗∗∗

32.78
4.40

(5.49)∗∗∗
(1.39)∗∗∗

1455
1455

46.13

(6.81)∗∗∗

44.06

(7.46)∗∗∗

1638

30.42
4.52

(4.66)∗∗∗
(1.18)∗∗∗

29.67
4.48

(6.26)∗∗∗
(1.29)∗∗∗

1638
1638

Note: Reduced sample includes members and non-members only from Kebeles with agricultural cooperatives;
Whole sample includes the whole sample (i.e., members and non-members from Kebeles with and without agricultural cooperatives). Bootstrap with 100 replications is used to estimate the standard errors.
∗∗∗ Significant at 1% level.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.

Table A4 – Correlates of variations in impact of cooperative membership on technical
efficiency within members
Dependent variable: Technical efficiency gain from
membership
Indicator

Reduced sample

Whole sample

HH head age
HH head gender
HH head literacy (1 = Yes, 0 = No)
Distance to market (Minutes)
Access to irrigation (1 = Yes, 0 = No)
Receives off-farm income(1 = Yes, 0 = No)
Radio ownership
Land cultivated (ha)
Number of plots
Number of Oxen
Reside in village with FTC(1 = Yes, 0 = No)
Improved seed(Amount used in Kg)
Fertilizer (Amount used in Kg)
Constant
Number of Obs.
R-Squared

0.000 (0.76)
−0.047 (2.19)∗∗
−0.002 (0.27)
−0.000 (1.68)∗
0.231 (25.18)∗∗∗
−0.033 (4.01)∗∗∗
0.012 (1.26)
0.015 (2.86)∗∗∗
−0.003 (1.56)
−0.006 (1.24)
0.037 (2.66)∗∗∗
0.000 (1.95)∗
−0.000 (0.14)
0.095 (2.77)∗∗∗
559
0.37

0.000 (0.46)
−0.055 (2.58)∗∗
0.004 (0.42)
−0.000 (1.51)
0.238 (27.47)∗∗∗
−0.035 (4.21)∗∗∗
0.012 (1.25)
0.015 (2.92)∗∗∗
−0.003 (1.42)
−0.004 (0.90)
0.042 (2.86)∗∗∗
0.000 (1.88)∗
−0.000 (0.26)
0.099 (2.86)∗∗∗
549
0.39

Note: ∗∗∗ Significant at 1% level, ∗∗ significant at 5% level and ∗ significant at 10% level.
t-statistics in parenthesis.
Source: Authors’ calculations, based on data from Ethiopia Agricultural Marketing Household Survey, 2008.
© 2014 The Authors
Annals of Public and Cooperative Economics © 2014 CIRIEC

apce12035

W3G-apce

February 21, 2014

5:22

Journal: APCE
Paper: apce12035
Dear Author
During the copy-editing of your paper, the following queries arose. Please respond to
these by marking up your proofs with the necessary changes/additions. Please write
your answers on the query sheet if there is insufficient space on the page proofs. Please
write clearly and follow the conventions shown on the corrections sheet. If returning the
proof by fax do not write too close to the paper’s edge. Please remember that illegible
mark-ups may delay publication.

Query No.

Description

Q1

{?Tiegist in refs}

Q2

{?further details}

Remarks

apce12035

W3G-apce

February 21, 2014

5:22

1

USING e-ANNOTATION TOOLS FOR ELECTRONIC PROOF CORRECTION
Required software to e-Annotate PDFs: Adobe Acrobat Professional or Adobe Reader (version 7.0 or
above). (Note that this document uses screenshots from Adobe Reader X)
The latest version of Acrobat Reader can be downloaded for free at: http://get.adobe.com/uk/reader/
Once you have Acrobat Reader open on your computer, click on the Comment tab at the right of the toolbar:

This will open up a panel down the right side of the document. The majority of
tools you will use for annotating your proof will be in the Annotations section,
pictured opposite. We’ve picked out some of these tools below:

1. Replace (Ins) Tool – for replacing text.

2. Strikethrough (Del) Tool – for deleting text.

Strikes a line through text and opens up a text
box where replacement text can be entered.
How to use it

Strikes a red line through text that is to be
deleted.
How to use it



Highlight a word or sentence.



Highlight a word or sentence.



Click on the Replace (Ins) icon in the Annotations
section.



Click on the Strikethrough (Del) icon in the
Annotations section.



Type the replacement text into the blue box that
appears.

3. Add note to text Tool – for highlighting a section
to be changed to bold or italic.

4. Add sticky note Tool – for making notes at
specific points in the text.

Highlights text in yellow and opens up a text
box where comments can be entered.
How to use it

Marks a point in the proof where a comment
needs to be highlighted.
How to use it



Highlight the relevant section of text.





Click on the Add note to text icon in the
Annotations section.

Click on the Add sticky note icon in the
Annotations section.



Click at the point in the proof where the comment
should be inserted.



Type the comment into the yellow box that
appears.



Type instruction on what should be changed
regarding the text into the yellow box that
appears.

USING e-ANNOTATION TOOLS FOR ELECTRONIC PROOF CORRECTION

5. Attach File Tool – for inserting large amounts of
text or replacement figures.

6. Add stamp Tool – for approving a proof if no
corrections are required.

Inserts an icon linking to the attached file in the
appropriate pace in the text.
How to use it

Inserts a selected stamp onto an appropriate
place in the proof.
How to use it



Click on the Attach File icon in the Annotations
section.



Click on the Add stamp icon in the Annotations
section.



Click on the proof to where you’d like the attached
file to be linked.





Select the file to be attached from your computer
or network.

Select the stamp you want to use. (The Approved
stamp is usually available directly in the menu that
appears).



Click on the proof where you’d like the stamp to
appear. (Where a proof is to be approved as it is,
this would normally be on the first page).



Select the colour and type of icon that will appear
in the proof. Click OK.

7. Drawing Markups Tools – for drawing shapes, lines and freeform
annotations on proofs and commenting on these marks.
Allows shapes, lines and freeform annotations to be drawn on proofs and for
comment to be made on these marks..

How to use it


Click on one of the shapes in the Drawing
Markups section.



Click on the proof at the relevant point and
draw the selected shape with the cursor.



To add a comment to the drawn shape,
move the cursor over the shape until an
arrowhead appears.



Double click on the shape and type any
text in the red box that appears.

For further information on how to annotate proofs, click on the Help menu to reveal a list of further options: