Muayila Kabibu Edited by James

Acknowledgements
We would like to thank my Promoter, Professor Eric Tollens, for having spent the last couple
of years in mentoring and guiding our doctoral training program. Thanks also for the many
hours and days he spent editing this doctoral draft. Our sincere thanks go to all of the
doctoral committee members, in particular to Professor Erik Mathijs, Professor Miet
Maertens, Professor Mwabila Malela, Professor Jan Diels and Professor Marysse Stefaan, for
their comments and corrections. We would also like to thank the staff of the Division of
Agricultural and Food Economics for their assistance during our doctoral training, in
particular to our colleagues, the doctoral students and researchers: Wouter Achten, Koen
Dillen, Kidane Egziabher, Abebe Ejigu Alemu, Jorge Alberto Cusiscanqui Giles, Anneleen
Kenis, Josaphat Mugabo, Basil Mugonola, Alice Nakiyemba Were, Cristal Taboada, Mesfin
Tilahun Gilaye and Benny Vande Velde.
We would like to thank the Church Development Service (Evangelischer Entwicklungsdienst

– EED) of Germany for the financial assistance and scholarship granted for my Master’s and
Doctoral degrees. Our sincere gratitude goes to the International Office, K.U.Leuven, for
providing financial support throughout the Expat-Allowance program during the three years
of our doctoral training program in Leuven. Special gratitude goes to the authorities of the
Protestant University of Congo, in particular to the Rector Professor Daniel Ngoy Boliya, the
Academic Secretary Albert Afumba Wandja and Roger Massamba Massambe, the
Administrative Secretary Léon Nondo Maneng and the Financial Director Raymond Ngwala,
for providing financial support and materials in data collection and processing, and for their
confidence in me and encouragement.
Special appreciation and thanks go to our lovely father, Bernard Kabibu Bimvulu, and my
mother, Astride Mujinga Tshioto, for having encouraged and advised me throughout the
entire period of my study from my childhood on. In a similar way, we would like to express
our gratitude to my brothers, sisters and cousins, in particular to Dr. John Bimvulu Kabibu,
Delice Katshiki Kabibu, Rose Mabudi Kabibu, Pierre Tshioto Kabibu, Irene Mujinga Kabibu,
Esther Ngamba Munyi Kabibu, Bernadette Kayengi Kabibu, Dany Henriette Musau Kabibu,
Emmanuel Kande Mukala, Daddy Tshigoma Mpanda and Christian Nsangamina.
i
Many thanks also go to our colleagues from the Protestant University and the University of
Kinshasa for their support and encouragement, in particular to Ghislain Batakela, Nathalie
Kibonge, Jean-Pierre Nganda Afumba, Paul Honoré Kalenga, Lena Phoba, N’shue Mokime,
Frank Mulamba, Luck Bondala, Patrick Bakengela Shamba, Prosper Ciamala, Gerome
Mutombo, Dani Kadi, Doudou Ewuli, Paty Kalay, Prince Leta, Mabi Lukusa and Frederic Kalala
Tshimpaka. In the same way, our gratitude goes to our friends, Jean Jacques Mamba, Didier
Ilolo, Chantal Landu, Patricia Nzanzi, Efika Tiniya Lenoir, Efika Mukaji Muenyi Lenoir, John
Mudianyi, Karim Nondo, Gisèle Kanjinga, Hubert Bambila, Alphonse Kalala, Dieudonné
Nyembwe, Prince Tshivuadi Kabengele, Jule Muya, Donat Olela, Appolinaire Biloso, Buluya
Yousse, Honoré Kanumbedi and Gabriel Mamba Tshidinda, for their encouragement and
moral support.
- ii -
Table of Contents
Acknowledgements i
Table of Contents iii
Abbreviations v
Abstract viii
Chapter 1 1
General Introduction 1
1.1 Background information 1
1.2 Microfinance 5
1.3 Problem statement 8
1.4 Conceptual frameworks 10
1.5 Objectives and hypotheses 13
1.6 Study area 15
1.7 Data and research methodology 16
1.7.1 Data collection 16
1.7.2 Research methodology 21
1.8 Overall dissertation structure 24
Chapter 2 29
Determinants of Farm Household Credit Constraints in the Hinterland of Kinshasa 29
2.1 Introduction 29
2.2 Literature review 30
2.3 Data and methodology 34
2.4 Results and discussions 39
2.4.1 Results of household’s responses 39
2.4.2 Results of estimations 41
2.4.3 Discussion of results 44
2.5 Conclusions 49
Chapter 3 52
Credit Constraints and Efficiency of Cassava Farmers in the Hinterland of Kinshasa 52
3.1 Introduction 52
3.2 Review of literature 53
3.4.1 Efficiency scores distributions 60
3.4.2 Factors affecting cassava production efficiency differentials 61
- iii -
3.4.3 Credit constraints conditions and cassava production efficiency 63
3.4.4 Discussion of the results 71
3.5 Conclusions 76
Chapter 4 79
Impact of Credit Constraints on Farm-Household Welfare in the Hinterland of Kinshasa 79
4.1 Introduction 79
4.2 Review of literature 80
4.3 Methodology and data 82
4.3.1 Data 82
4.3.2 Research methodology 82
4.4.1 Results 86
4.5 Conclusions 91
Chapter 5 93
Performance of Microfinance Institutions in the Democratic Republic of Congo 93
Does Credit Towards Farmers Matter? 93
5.1 Introduction 93
5.3.1 Results 99
5.4 Conclusions 110
Chapter 6 115
Summary and Implications 115
6.1 Summary 115
6.2 Policy implications and recommendations for further research 120
- iv -
Abbreviations
CADIM Centre d’Appui au Développement Intégral/Mbankana

RIFIDEC Regroupement des Institutions du Système Financier Décentralisé du Congo)
DEA Data Envelopment Analysis
MFIs Microfinance Finance Institutions
PSM Propensity Score Matching
DRC Democratic Republic of Congo
ANOVA Analysis of Variance
VRC Variable Returns to Scale
CRS Constant Returns to Scale
FC Franc Congolais
GDP Gross Domestic Product
USD United States Dollars
FAO Food and Agriculture Organization
UCCEC Union de Coopératives Centrales d’Epargne et de Crédit
COOPEC Coopératives d’Epargne et de Crédit
NGOs Non-government Organizations
USAID United States Agency for International Development
GTZ Deutsche Gesellschaft für Technische Zusammenarbeit
BTC Belgian Technical Cooperation
DID Développement International de Desjardins
UNDP United Nations Development Programme
FINCA Foundation for International Community Assistance
ATT Average Treatment Effect on Treated
SE Scale Efficiency
CIA Conditional Independency Assumption
UNSD United Nations Statistics Division
ROA Returns of Asset
OSS Operational Self-Sufficient
ADA Appui au Développement Autonome
-v-
UK United Kingdom
ULB Université Libre de Bruxelles
IRES Institut de Recherches Economiques et Sociales
CEPAS Centre d’Etudes pour l’Action Sociale
FASE Faculté d’Administration des Affaires et Sciences Économiques
List of Figures
Figure 1.1 Food-production index (base period, 1999-2001) 2

Figure 1.2 Share of agriculture in the national economy 2
Figure 1.3 Credit constraints and their consequences on households: conceptual
framework 13
Figure 1.4 Location of the study area 16
Figure 1.5 Schematic presentation of research methodology 24
Figure 1.6 Schematic presentation of dissertation organization 25
Figure 2.1 Empirical approach used to identify farm household credit constraints 35
Figure 3.1 Distribution of average efficiency score 65
Figure 3.2 Distribution of CRS technical efficiency for credit: constrained versus
unconstrained 65
Figure 3.3 Distribution of VRS technical efficiency for credit: constrained versus
unconstrained 66
Figure 3.4 Distribution of scale efficiency for credit: constrained versus unconstrained 66
Figure 3.5 Distribution of propensity of being credit: constrained (histogram) 69
Figure 3.6 Distribution of propensity score of being credit: constrained (boxplot) 70
Figure 5.1 Average efficiency score 100
Figure 5.2 Efficiency in the provision of saving and credit 101
Figure 5.3 Efficiency in the production of financial performance 102
Figure 5.4 Efficiency in the production of social performance 103
Figure 5.5 Efficiency in the production of microfinance double bottom performance 104
Figure 5.6 Technical efficiency score (CRS) for MFIs with versus without agricultural credit
104
Figure 5.7 Technical efficiency score (VRS) for MFIs with versus without agricultural credit
105
- vi -
Figure 5.8 Scale efficiency for MFIs with versus without agricultural credit 105
List of Tables
Table 1.1 Distribution of sample size 17

App. Table 1.1 Statistic description of quantitative variables 25
App. Table 1.2 Statistic description of dummy variables 27
Table 2.1 Definition, measures and expected signs of covariates 39
Table 2.2 Number of loan applications by lenders 40
Table 2.3 Volume of loan demanded and supplied by lenders (in FC) 41
Table 2.4 Results of logit model of credit constraints 42
Table 2.5 Results of probit model of credit constraints 43
App. Table 2.1 Statistic description of variables used in the regression 50
App. Table 2.2 Multicollinearity diagnostic 50
Table 3.1 Distribution of technical and scale efficiency in cassava production 61
Table 3.2 Determinants of cassava production efficiency differentials 62
Table 3.3 Distribution of cassava production efficiency scores 64
Table 3.4 Balancing properties 68
Table 3.5 Impact of credit constraints on cassava production efficiency 70
Table 3.6 Simulation-based sensitivity analysis 71
App. Table 3.1 Statistic description of variables used in Chapter 3 78
Table 4.1 Impact of credit constraints on consumption per adult equivalent per
day 88
Table 4.2 Simulation-based sensitivity analysis 88
App. Table 4.1 Statistic description of variables used 92
Table 6.1 Descriptive statistics of inputs and outputs 99
Table 5.2 Results from ANOVA and Mann-Whitney tests 106
App. Table 5.1 Distribution of efficiency of MFIs 110
App. Table 5.2 Efficiency scores of MFIs with versus without agricultural credit
provision 112
- vii -
Abstract
This dissertation has two main objectives. The first is to investigate and identify some key
factors affecting farm household credit constraints in the hinterland of Kinshasa and to
assess their effects on the efficiency of cassava production and on the distribution of
household welfare. The second is to measure the efficiency of Microfinance Institutions
(MFIs) operating in the Democratic Republic of Congo (DRC). More specifically, the
dissertation examines whether credit provision to farm-households affects the efficiency
scores of MFIs. Using cross-sectional data from household surveys, Data Envelopment
Analysis (DEA) and Propensity Score Matching (PSM) are applied to measure efficiency in
cassava production and the impact of credit constraints on household welfare, respectively.
Truncated regression is applied to identify the key drivers of cassava production efficiency.
Household consumption is used as a proxy for welfare distribution. Probit and logit models
are applied to identify household socio-economic and demographic characteristics affecting
the probability of facing credit constraints. DEA is also applied to compute the efficiency of
MFIs, while a standard ANOVA test and Mann-Whitney test are applied to investigate
whether agricultural credit affects the efficiency and the performance of MFIs in the DRC.
The results from households’ responses reveal that that a lack of collateral, inappropriate
loan term conditions, the inadequacy of microcredit technology, a higher level of agricultural
risks and high interest rates, combined with low returns in farming activities, explain the
limited access of farmers to credit. The results from the estimation of logit and probit
models show that household size, association participation, household size, landholding
property and remittances are covariates associated with credit status constraints conditions.
The findings also show that credit constraints have had a significant and negative impact on
household welfare distribution and on efficiency in cassava production. Landholding
property, education, association participation and farm size are found to the key drivers of
cassava production efficiency. The findings also show that credit constraints have a
significant and negative impact on household welfare distribution and cassava production
efficiency scores.
- viii -
The findings show that some of the MFIs operating in the DRC reach a double bottom
efficiency line. This suggests that it is possible to find a good balance between social and
financial performance of MFIs. Thus, it is possible for MFIs both to promote a social mission
and achieve optimal financial sustainability. As the majority of MFIs are operating in the
zone of increasing efficiency scales and under variable returns to scale, the introduction of
management best practices and human capital empowerment issues would increase their
efficiency. Furthermore, the increase of MFI assets and operations would also improve their
efficiency. The study has found in addition that the financing of farming activities is not
significantly associated with the efficiency scores of distributions.
- ix -
Chapter 1
General Introduction
1.1 Background information
With an under-developed industrial sector, agricultural activities remain the principal

component of economic development in the DRC. Consequently, it is absurd to think of
development without adequate agricultural policies and strategies that foster the
improvement of agricultural productivity and performance. This assertion is supported and
justified by several reasons: (a) agriculture accounts for about 45% of GDP and 20% of export
earnings; (b) agricultural activities are the mainstay of livelihoods and a source of income for
over 60 % of the population; (c) the majority of the Congolese population is affected by food
insecurity; and (d) domestic industry is still at a primitive stage and needs a dynamic and
productive agriculture to provide essential raw materials.
Despite the adoption of many agricultural reforms and policies, the agricultural sector in the
DRC remains dualistic with a dominance of small traditional family farms. Two farming
systems, namely the traditional farming system and the modern farming system, coexist.
Technically, the traditional farming system, also called itinerant agriculture, is characterized
by relatively long fallow periods, depending on whether the region is under-populated or
over-populated. High yielding varieties to improve production and other inputs for land
regeneration are not used in the traditional farming system. This production system
continues to resort to shifting cultivation and depends on family labor. In the traditional
production system, women play a dominant role as the participation of men is limited to
pre-farming tasks (cutting trees, clearing land, incineration, etc.). The possibilities of access
to resources, such as loans, improved inputs and modern tools are very limited. The harvest
is stored under random environmental conditions, which lead to extensive and damage.
Before 1960, the colonial authorities tried to promote traditional agriculture by using the so-
called paysannat. The goal of paysannat was to replace the extensive farming system, a
source of land destruction, with a more intensive farming system, ensuring the sustainability
and improvement of land productivity. However, a few years after political independence,
the paysannat system, which was still at the experimental stage. The attempts made to
-1-
revitalize the traditional agricultural system did not lead to significant progress. Thus,
traditional farming remains essentially unchanged: the techniques of traditional production
remain rudimentary and many farmers continue to use poor yielding varieties. The modern
farming system, introduced during the colonization period, was composed of large farms
mainly specializing in export crops and owned by foreign companies. These companies had
access to resources and were the principal recipient of the results of agronomic research,
which was carried out according to their needs. Most of them went bankrupt after the
Zairianisation of the 1970s, although those which still survive continue to farm in the same
manner.
Providing enough food for a growing population still remains the main challenge of the
Congolese agricultural sector. The DRC is one of the Sub-Saharan African countries where
agricultural production has been trailing population growth for many years. The food
production index has been largely unstable and has shown a steady decline since 1969
(Figure 1.1). Thus, the DRC, which achieved food self-sufficiency during the first years of
political independence, has become a net food importer and depends on food aid. FAO
(2007) statistics (Figure 1.2) show that food imports represented more than 70% of total
food consumed and 70% of total imports between 1995 and 2007.
Figure 1.1 Food-production index (base period, 1999-2001)
Source FAO, 2007.
-2-
Figure 1.2 Share of agriculture in the national economy
Source : FAO, 2007.
The lack of infrastructures in rural areas, the inefficiency of agricultural research, the
underdevelopment of productive forces, the high imperfection of markets and especially the
lack of financial institutions in rural areas are the main causes of the low productivity and
underperformance of agriculture in the DRC. The improvement of performance of this sector
cannot be done without using modern equipment and inputs such as improved seeds,
fertilizers, herbicides, insecticides, etc. These new technological changes require sufficient
capital, which is unfortunately not available to smallholder farmers.
The provision of financial services to poor farmers in the DRC, as in many developing nations,
has often been perceived as difficult. The difficulties of financing agricultural activities are
the high transaction costs, the low returns from agriculture and the high covariant risk.
González-Vega (2003) reported several difficulties constraining the access to credit of farm-
households. These difficulties are related to:
(i) high transaction costs for both borrowers and lenders, which increase the cost of
the loans beyond interest rates and depend on distance (physical, cultural, and
social), (ii) information problems, which create uncertainty for the lender about the
ability and willingness to repay of the borrower and may thus result in adverse
selection and moral hazard, (iii) incentive problems, which emerge from the
conflicting interests of lenders and borrowers and, under hidden information, may
result in moral hazard, (iv) enforcement problems, which increase the costs for
lenders, and (v) covariance problems, as the activities funded may be influenced by
systemic shocks (González-Vega, 2003).
-3-
In the DRC, as in many developing nations, different governments, with the support of
donors and international organizations, have taken measures to correct the imperfections of
rural financial markets and improve farmers’ access to credit, but with little success. On the
eve of political independence, State-owned banks and rural finance programs were created
to improve farm’s access to credit and thereby improving the productivity and the efficiency
of family-based agriculture (Wampfler, 2000; Wampfler et al. 2002).
However, the credit provided by these institutions and finance programs reached the
smallholder farmers only with difficulty. However, these financial institutions and programs
went bankrupt after a few years of functioning due to poor management by the banks and
the inefficiency of the credit technologies used. Moreover, the provision of credit was
concentrated on cash crop agriculture and was less adapted to the needs of smallholder
farmers. The interest rates applied were very low and did not cover the costs and ensure
financial sustainability. Likewise, loan terms and conditions, such as collateral, were
inappropriate. The problem of management was very important; risk management tools
were non-existent and transaction costs were very high (Wampfler, 2000; Wampfler et al.
2002). Given the failure of State-owned banks and finance programs, Credit and Saving
Cooperatives were created to solve the problem. However, it quickly appeared that credit
provided by the cooperatives was mainly directed towards consumption smoothing and
small business than to agricultural production. Indeed, more than 70% of cooperative loan
portfolios were allocated to small business, while only approximately 5% were granted for
agricultural production investment (UCCEC, 1999). Since the beginning of the 1990s, the
activities of the Credit and Saving Cooperatives have been reduced drastically, leading to a
loss of 80% of customers and 70% of funds invested in deposit banks (UCCEC, 1999).
The experience of microfinance in the DRC started with the implementation of microcredit
programs by Non-Governmental Organizations (NGOs). NGOs provided microcredit to poor
populations excluded from the traditional system with reference to the welfare framework
and with the aim of improving their well-being. Currently, several NGOs are operating in the
field of microfinance. The emergence of NGOs was made possible thanks to the technical
and financial support of international organizations, such as the United States Agency for
International Development (USAID), the Deutsche Gessellschaft für Technische
-4-
Zusammenarbeit (GTZ), the Belgian Technical Cooperation (BTC), the DESJARDINS of Canada
(DID), and the United Nations Development Program (UNDP).
Recently, new actors have appeared in the microfinance sector operating with respect to the
commercial framework. Two main trends have been observed: the implementation of
specific financial institutions working under the commercial framework (FINCA, PROCREDIT);
and the development of microfinance products within commercial banks or their
participation in microfinance organizations (The International Bank of the Credit). Despite
the proliferation of MFIs and programs operating in the DRC, many farmers and poor
populations continue to face limited access to credit (Muayila, 2006). Thus, there is a need
for a better understanding of the mechanisms leading to the credit constraints experiences
by this section of the population in the context of the development of microfinance, and
their implications for households, in order to assist policy formulation.
1.2 Microfinance
Microfinance is the provision of financial services for poor people, who have socio-economic
characteristics that exclude them from traditional banks (Labie, 1999; Helms, 2006; Schmidt
and Zeitinger, 1994; Armendariz and Morduch, 2007). It is about the extension of small loans
and other financial services, such as savings accounts, remittances, insurance, etc. The
success of Asian and South American MFIs has led international development agencies to
believe that microfinance is a new paradigm capable of contributing efficiently to the
provision of credit for poor populations and thus to the reduction of poverty. Microfinance is
thought to reduce poverty by promoting self-employment and entrepreneurship and
overcoming liquidity constraints, thus allowing consumption and income smoothing
(Armendariz and Morduch, 2007).
The literature discusses various institutional forms of microfinance organizations, including

informal, semi-formal and formal finance bodies; these vary across countries depending on
the level of financial development, economic development and regulation policy. In the
informal finance sector, transactions are generally short-term and based upon personal
relationships, cultural links and geographical proximity. The lending in the informal market is
-5-
characterized by: repayment as a function of borrower cash-flow, loan maturity adapted to
the client’s requirements, small loan size and high interest rates (Seibel, 1996; DeBriey,
2003, 2005; Albert, 1997; Nzemen, 1988; Lelart, 1989). Social pressure, the subordination of
a new loan to the refunding of the former one and the cultural and geographical proximity
applied in informal microfinance markets allow for a reduction in the risks of adverse
selection and moral hazard, as well as a reduction in transaction costs (Albert, 1997;
Aryeetey and Udry, 1997; DeBriey, 2003). The adverse selection occurs in credit market
when potential borrowers who are better risks or have better products do not apply for
loans, while those with higher risks apply. The lenders lack good information about the
riskiness of the potential borrowers’ projects, thus cannot distinguish risky borrowers from
safer borrowers (Stiglitz and Weiss, 1981). Thus, lenders have to rely on proxy indicators to
screen potential borrowers. Moral hazard in lending is the risk that the borrowers will not
use the money as was intended, may take unnecessary risks or not be vigilant in reducing
risk. It refers to situations where lenders cannot observe either the effort made, the action
taken by the borrower, or the realization of project returns (Armendariz and Morduch,
2007).
The most known informal structure is the tontine. The operations of tontines are very limited
because in most cases, they depend on local resources, and they are only adapted to
homogenous environments. The advantages of tontines are: easy implementation in local
communities, and the greater possibility of combining financial provision with other
collective activities, such as working in group project for the village (Labie, 1999; DeBriey,
2003, 2005). Despite the proliferation of informal lenders in many developing nations, their
contribution to the financing of micro enterprises remains marginal due to a lack of real
financial intermediation, the difficulty of combining savings and credit, the insufficiency of
the financial resources, the applied high interest rate and the non-adaptation to the returns
of micro enterprises (Adams, 2002; Yaron, 1992).
Most known semi-formal organizations of microfinance are NGOs and cooperatives. The
cooperatives are a form of organization functioning under the principle of “one person-one
vote”. The organization of these cooperatives is based upon respect for the principles of
cooperation and on the fact that the borrowers do not have other attractive alternatives of
financing. The cooperatives collect the savings of members and/or non-members to provide
-6-
loans. The creation of cooperatives rests on the existence of financial needs which cannot be
satisfied by the market (“market failure”) or by the public authorities (“State failure”). The
funds of cooperatives come from savings, loans, public subsidies and donations (Hugon,
1996). Most cooperative loans are collateralized by savings. This mechanism of collateral
leads to the exclusion of poor people without savings. The fact that the members are at the
same time owners and clients enables a convergence of interest which is likely to reduce
adverse selection and moral hazard (DeBriey, 2003, 2005).
NGOs target the poorest of the poor and provide micro credit on the basis of social or
welfare imperative. For NGOs, micro credit is meaningful only if it is provided under the
framework of the fight against poverty and to improve the poor’s well-being (Woller et al.
1999). In other words, NGOs perceive micro credit as an effective instrument to improve the
conditions of the poor population. The technology of micro credit applied by most NGOs
consists of granting small loans to the individuals or/and to the group, typically comprised of
three to seven neighbors (Armendariz and Morduch, 2007). Group lending consists of
“arrangements by individuals without collateral who get together and form groups with the
aim of obtaining loans” (Armendariz and Morduch, 2007). This lending technology has been
pioneered by microfinance leaders like Bangladesh's Grameen Bank and Bolivia's BancoSol
(Armendariz and Morduch, 2007). The advantages of the NGOs are their high ability to
collect subsidies, encourage credit technology innovations and pursue outreach rather than
financial profitability (Labie, 1999). However, the most important disadvantages of NGOs are
their lack of good governance, the fact that they do not collect savings and the difficulty of
managing growth.
The failure of many microcredit programs designed by NGOs in terms of promoting people’s
welfare has amplified concerns about their performance and sustainability. Thus, many
authors have called for the application of a commercial imperative to ensure the provision of
financial services towards poor populations in a sustainable way (Woller et al. 1999; DeBriey,
2005). This new framework has led to the formation of formal microfinance structures also
known as regulated MFIs. The number of regulated MFIs regulated financial institutions has
increased during the last two decades, especially due to the expansion of the scope of NGOs
into regulated institutions through downscaling. Formal MFIs resort to a new approach
known as the “institutionalist approach” or the “commercial approach” (Woller et al. 1999;
-7-
DeBriey, 2005). This approach has two objectives: outreach and the perpetuation of the
financial sustainability of institutions. For this approach, MFIs must freely apply interest
rates allowing them to cover costs and optimize profit, in order to face the challenges of
growth and to provide financial services in a sustainable way (Woller et al. 1999; DeBriey,
2005). This approach is based on the assumption that what is important for micro
enterprises is access to credit rather than the cost of it (DeBriey, 2003). Following Wampfler
(2000), there is a trend and gradual consensus around three concepts: sustainability of
financial market, effective financial intermediation and viable and financially sustainable
institutions.
To sum up, the common feature of microfinance is the technology of credit based on the
concepts of social capital and geographic proximity with clients, which are necessary
requirements for building a better relationship with clients. This credit technology is likely to
allow the majority of problems, such as administrative costs, transaction costs, moral hazard
and adverse selection, to be addressed. Lending technology applied in microfinance resorts
to non-traditional approaches to collateral requirements, such as village banking groups.
This approach provides a win-win institutional setting (Armendariz and Morduch, 2007). The
collaterals required may be material or moral, but in most cases are based on social
pressure as well as the motivation of receiving a new loan. The interest rates applied in
microfinance are generally equal or higher than those applied in the traditional banks
(Labie, 1999). The basic assumption in microfinance is that the problem of information,
which results in adverse selection, moral hazard, information costs and barriers to access of
credit, is resolved because the borrowers are well known as a result of their social and
geographic proximity.
1.3 Problem statement
Despite the development of MFIs in the developing nations, the gaps between supply and
demand of agricultural credit remain large, thus many farmers from these nations continue
to face credit constraints. Credit constraints have been a subject of much concern in the
DRC as well as in many developing nations. A few years after political independence, credit
was provided under the paradigm of direct formal finance towards farmers, which was
-8-
supposed to lead to sustainable agricultural growth. Thus, specialized agricultural
development banks and rural finance programs were created and promoted. However, after
a few years of functioning, most of them failed due to problem of credit default.
Consequently, huge numbers of rural households have to face credit constraints.
To solve the problem, MFIs have emerged to meet the unsatisfied demands for rural
household loans. However, despite the increasing number of MFIs, credit constraints
continue to be the main characteristic of rural households in the DRC. Only a few MFIs are
working with smallholder farmers and many of them prefer to transact with customers
engaged in off-farm activities (Muayila, 2006). Previous studies have focused on the
difficulties of financing agriculture through the microfinance paradigm. These difficulties
include the problem of high covariate risks related to agriculture (Ellis, 1988; Wampfler,
2000; Wampfler et al. 2002), MFIs’ limited resources to fund the important, long term needs
of agriculture (Wampfler, 2000), the high transaction costs related to the financing of
agricultural activities (Gurgand et al. 1996), and the strong demand of sectors with high
returns and low risks (Lapenu, 2001).
However, in the DRC, despite the high prevalence of credit constraints towards farm-
households, this topic remains poorly documented. This study aims at contributing to the
literature on credit constraints and their effects in farm household production efficiency and
welfare, by providing empirical evidence from the hinterland of Kinshasa. This study also
tests whether the provision of credit to farmers affects the efficiency of MFIs operating in
the DRC. The dissertation will address the following questions:
1. What are the main difficulties in the provision of microcredit towards farm-households?
2. What are the key drivers affecting the probability that a farm-household is credit
constrained?
3. Do credit constraints affect the production efficiency of farm-households?
4. What are key drivers of technical efficiency of farm-households?
5. Do credit constraints influence farm-household welfare distribution?
6. Is there a good balance between financial sustainability and outreach aims of MFIs?
-9-
7. Is there a substantial difference in efficiency distribution between MFIs with and without
the provision of microcredit towards farm-households?
1.4 Conceptual frameworks
The provision of credit involves adverse selection and moral hazard risks resulting from
information problem, which does not allow lenders to better screen potential borrowers
(Stiglitz & Weiss 1981). In addition, providing credit towards farm-households is difficult
because of the high risks related to agriculture, the low returns of agricultural activities, the
more volatile cash flows of farm-households and high transaction costs (Ellis, 1988; Pischke,
1994; Gurgand et al. 1996; Klein et al. 1999; Wampfler, 2000). Furthermore, MFIs, especially
small scale and local financial institutions working in rural areas, do not often have access to
external resources from market and donors, and thus have to face liquidity constraints. A
common remedy to cope with the information problems and the risks related to financing
small business is to collateralize loans. The collateral requirements provide incentives for
borrowers to take good decisions and allow lenders to address moral hazard and adverse
selection problems. As many farm-households from developing nations are too poor and
have insufficient physical capital endowments to collateralize loans, MFIs extend collateral
requirements to non-physical capital, such as social capital. They also resort to a close
lending relationship by implementing local financial agencies. This lending technology is
made possible by their geographic and cultural proximity with the potential borrowers.
Close lending combined with group lending technology allows MFIs to better screen
potential clients and reduce transaction costs.
The conceptual framework (Figure 1.3) is constructed on the basis of a model developed by
Zeller (1995) and the household portfolio model proposed by Chen and Dunn (1996). This
conceptual framework recognizes that household characteristics providing information on
household capital endowments play an important role in determining household access to
credit (Zeller, 1995). Household capital endowments include physical capital, human capital
and social capital owned by households. Household physical capital represents physical
goods and financial capital, such assets, land, tangible assets, remittances, cash and saving
- 10 -
owned by household. Human capital represents skills, knowledge, management capacity and
the ability to take production decisions promoting wealth, and is thought to be a function of
household demographic composition, such as size, dependency ratio, education, age, gender
of household head, etc. It provides information on the capacity of household to use credit in
optimal way. Social capital, for example membership of a social organization, is an indicator
of the capacity of households to respond to socio-economic engagements. Furthermore,
microfinance resorts to group lending technology, and social capital, such as membership, is
used as a mechanism of supervision and collateral, which allows MFIs to address moral
hazard and adverse selection problems.
These three forms of capital determine the household repayment capacity and
creditworthiness as they can serve as collateral (Zeller, 1995). Therefore, households owning
less of these three capital forms are expected to have a lower risk-bearing capacity and will
then be more exposed to credit constraints. The conceptual framework also assumes that
local institutional characteristics are important in determining household credit status
because they determine the transaction costs for both lender and borrower, and thus
impact both the decision of lenders to grant credit and demand for loans by households. The
lack of infrastructure and institutions, such as roads and market, is more likely to increase
the transaction costs related to credit and prevent household from making good production
and consumption decisions.
Based on the household portfolio model, the conceptual framework shows the interaction
between household capital endowments and activities such as production investment and
consumption (Chen and Dunn, 1996). Household pool resources are used to support
investment for production and consumption activities, and the surplus resources from
household activities are returned to the resource pool (Zeller, 1995; Chen and Dunn, 1996).
For example, household human capital can allow the adoption and efficient use of more
productive technology, which in turn increases productivity and income for consumption
smoothing. Social capital, such as participation in a farmer’s organization, can result in more
efficient production processes through mutual learning and information exchange. At the
same time, physical capital can help households exploit economies of scale (Zeller, 1995).
Household resources can be directly used to support the consumption of food and non-food
items, such as clothing, health care, and education, which in turn contributes to
- 11 -
improvements in household welfare. The consumption activities of household do not
directly contribute to the accumulation of household income and assets, but they are
important in maintaining and increasing resource productivity and efficiency. Additional
resources from loans increase the resources of households and can be used to support any
household activity. Thus, credit constraints prevent opportunities and investment by limiting
the ability of households to use modern inputs (Freeman et al. 1998). This in turn negatively
affects production efficiency (Croppenstedt et al. 2003). Indeed, empirical studies have
revealed that credit constraints affect risk aversion, as well as technology choice and
adoption, since constrained households are more likely to invest in less risky and less
productive technologies rather than in more risky and more productive one (Dercon, 1996;
Guirkinger and Boucher, 2008; Eswaran and Kotwal, 1986). Furthermore, credit may be used
to finance household consumption and might affect household welfare in different ways,
such as consumption and income smoothing, risk-coping and liquidity effects (Maldonado,
2004). Thus, households with credit constraints will tend to have lower consumption and
welfare levels compared to those without credit constraints.
The performance of MFIs is evaluated with reference to the efficiency frontier as a

conceptual framework. Furthermore, the selection of variables for the measurement of the
efficiency of MFIs is based on the outreach and sustainability conceptual framework
proposed in Yaron (1992) and the financial production model (Korsah et al. 2001; Soteriou
and Zenios, 1999). Outreach refers to social performance of MFIs or their capacity to reach
the poor population, while sustainability refers to financial performance. This conceptual
framework sought to show the convergence, or the complementarily, between financial
sustainability and the social performance of MFIs. Furthermore, given the many constraints
related to the financing of agriculture, one could assume that the provision of credit towards
farm-households reduce the efficiency of MFIs. However, as the majority of MFIs in the
study are working in commercial framework and resort to close lending technology, it might
be possible to expect that the provision of credit towards farm-households has no strong or
significant effects on the efficiency of MFIs.
- 12 -
Figure 1.3 Credit access and its consequences on households: conceptual framework
Local institutionnel & Household Capital

community characteristics
- Human capital
- Physical capital
Financial Access to
- Social capital
market Credit
Macro economic Household Activities

policy & conditions Consumption Investment
Food/ Non -food

consumption
Welfare & Resource Income

health productivity & assets
1.5 Objectives and hypotheses
This study examines credit constraints and their consequences on farm-households in the
hinterland of Kinshasa. It also addresses the issue of evaluating the efficiency of MFIs
operating in the DRC with respect to the financing of agricultural activities. The specific
objectives of this study can be summarized as follows:
1. To identify the difficulties to the provision of micro credit towards farm-households;
2. To explore the key drivers affecting the probability that a farm household is credit-
constrained;
- 13 -
3. To measure production efficiency of farm-households and explore the key determinants
of technical efficiency scores differentials;
4. To evaluate the impact of credit constraints on technical efficiency of farm-households;
5. To evaluate the effect of credit constraints on farm-household welfare distribution;
6. To measure the efficiency of MFIs and investigate the trade-off between financial
sustainability and social performance;
7. To test whether the provision of agricultural credit affects the efficiency distribution of
MFIs with and without the provision of agricultural credit.
Finally, with regard to the findings, we draw some recommendations that can be useful in
the process of designing new strategies and policies to improve the access of farm-
households to micro credit, and to improve agricultural production efficiency and household
welfare.
Based on the conceptual frameworks developed in section 1.4, the hypotheses sustaining
this study can be summarized as follows:
1. The nature of agricultural activity constrains farm household access to microcredit;
2. Limited financial resources at the levels of MFIs reduce their capacity to respond to the
magnitude of demand for microcredit from farm-households;
3. Lending technology applied in the MFIs constrains farm-household participation in the

microcredit market;
4. Capital endowments reduce the probability that a farm-household is credit-constrained;
5. Lack of good local institutional environment increases the probability that a farm
household is credit-constrained;
6. Household’s capital endowments positively affect the technical efficiency of farm-

households;
7. Good local institutional environment positively affects the technical efficiency of farm-
households;
8. Credit constraints reduce the technical efficiency of farm-households;
- 14 -
9. Credit constraints decrease the scale efficiency of farm-households;
10. Credit constraints reduce the farm-household welfare level;
11. There is a good balance between social and financial efficiency of MFIs;
12. There is no substantial difference in efficiency distribution between MFIs with and
without the provision of agricultural microcredit.
1.6 Study area
The study was conducted in the hinterland of Kinshasa, which consists of all the
neighborhoods Kinshasa. This area is dominated by agriculture and its related activities. The
hinterland is a vast area of Kinshasa located in the Commune of Maluku and on the Batékés
Plateau (Figure 1.4), which has an area bordering on 7000 km². The Commune of Maluku is
limited to the north by the Congo River, Congo-Brazzaville and the territory of Kwamouth
(the Province of Bandundu); in the east by the territory of Bagata and Kenge (the Province of
Bandundu); in the south by the territories of Kasangulu, Kisantu and Kimvula (the Province of
Bas-Congo); and in the west by the Commune of N’sele (Pauwels, 1993; De Saint Moulain
and Kalombo, 2005).
The vegetation of Kinshasa consists of savanna strewn with shrubs and forest galleries.
Following urban pressure, Kinshasa is localized on hills and the Kwango Plateau (Lubini,
1988). The land of the Batéké Plateau is characterized by a tropical climate and low soil
fertility (Crabbe, 1980). The hydrographic network of Kinshasa is made up of rivers (Kalamu,
Gombe, Makelele, Funa, D’jili, N’sele, Mayi-Ndombe and Bombo-Lumene) taking their
sources from hills. These rivers run from south to north and join the Congo River at the level
of the Malebo Pool. Following the Köppen classification, the climate of the Batéké Plateau is
type AW4 (Wet Tropical Climate Soudanien) and characterized by two seasons, a dry season,
which extends from mid-May to mid-September and a wet season starting in mid-September
and being ending in mid-May (Bultot, 1950).
The attribution of the arable land in the Plateau of Batéké occurs according to both
traditional land tenure and Congolese legislation. Nsombo (2005) underlines the communal
- 15 -
nature of the land in Téké community and culture. However, population growth and the
development of the commercial economy have brought up the problem of land tenure in the
hinterland of Kinshasa. Among the population living in this area, the Téké people represent
the oldest group and form the majority. They are concentrated in the south and upstream of
the Malebo Pool (Mutamba, 1989). Three ethnic groups, the Suku, Yaka and Yansi, are
located in this zone.
Figure 1.4 Location of the study area
Source: De Saint Moulin and Kalombo (2005)
1.7 Data and research methodology
1.7.1 Data collection

Data used in this study were collected from households and MFI surveys. The household
survey was conducted in villages located in the area between Menkao, Dumi and Mbankana,
and where Téké represent the main share of the population. Three groups of villages were
selected for the survey. The first group includes Menkao, Bita and Kingankadi. The second
consists of Dumi, Mutiene and Inzolo. The last one comprises Mbankana, Quatrième
Cité/CADIM and Kinzono. To simplify the analysis, and given the small size of the sample,
these groups were named, respectively, the Menkao group, the Dumi group and the
Mbankana group. The survey took place from February to March 2008. Taking into account
- 16 -
the lack of reliable demographic data at the communal level, the first step in the survey
consisted of counting the households per village among the three groups selected for the
study: 1319 households were listed. The sample selection was made by quota with a rate of
15%. This resulted in a sample of 202 households randomly selected from each of the three
designated groups of villages. Table 1.1 shows the distribution of the sample size.
Table 1.1 Distribution of sample size
Village Number of Households Rate Households Selected

Menkao 304 15% 46
Bita 77 15% 12
Kingakata 58 15% 9
Dumi 230 15% 35
Mutiene 90 15% 14
Inzolo 110 15% 17
Mbankana 283 15% 43
4ème Cité 117 15% 18
Kinzono 50 15% 8
Total 1319 15% 202
The household data was collected from a cross-sectional survey in the three groups of
selected villages. A household was defined as a social unit sharing the same residence,
resources and income. Households consist of groups of people living in the same house and
under the responsibility of a head recognized by all members with particular characteristics
(Ntoto, 2001). A structured questionnaire was used to obtain data at household level. The
questionnaire consisted of a wide range of questions regarding the household’s
characteristics, including age, gender and education of the head of the household as well as
social capital, etc. The questionnaire also consisted of data directly related to cassava
production, such as land used, labor (family and hired), cassava planting materials and
cassava output. The questionnaire was pre-tested in order to correct mistakes eliminate
irrelevant and add relevant information. The household survey was conducted by
investigators selected on the basis of communication skills, all of whom had masters’
degrees in agricultural sciences or economics and knowledge of cultural and social traditions
of the study area. The data collection procedure was supervised and questionnaires were
- 17 -
examined to ensure complete responses. Thus, uncompleted questionnaires were detected
and omissions rectified by revisiting the respondents.
App. Tables 1.1 and 1.2 show the results of the descriptive statistics of data collected from
households. Given the low number of households selected in the same villages (see Table
1.1.), the villages were regrouped into three village areas on the basis of geographical
distance for the empirical analysis of data. There three groups were called the Mbankana
area, the Dumi area and the Menkao area. The average age of household head studied is
about 45 years old, with a range of 22 to 66 years old. There is no statistically significant
difference between the average age of household head in the three selected areas of study.
The average age of households head in Mbankana is 10.36, while that of Dumi is 9.49 and
9.59 for Menkao. The average number of years of formal education for household head is
about 9.45 (SD=3.85) with a maximum of 18 years (corresponding to university level) and a
minimum of 0 (no formal education). There is no strong and significant difference in the
education of household heads between the Mbankana area (9.65), the Dumi area (8.97) and
the Menkao area (9.65).
Turning to the gender of the household head, the results of the survey show that the
majority of households studied (91%) are led by male heads in the hinterland of Kinshasa.
About 92% of households of Mbankana are led by male heads, against 88% for Dumi and
93% for Menkao. The proportion of male heads between the three village areas is not
statistically significant. Regarding household size, the average household size is about five
people with a maximum of 11 persons and minimum of one person. There is a statistically
significant (5%) difference in means between households living in the three selected village
areas. The households of Mbankana have on average 4.57 persons, compared to Dumi with
5.38 persons and Menkao with 4.93 persons. The economic dependency rate appears higher
in the hinterland of Kinshasa (50%). There is a significant difference in means between the
dependency ratio of the three selected areas of the study. The average dependency ratio is
estimated at 0.55 for households living in the Mbankana area, which is higher than for the
Dumi area (0.44) and the Mbankana area (0.49). The mean difference is significant at 5%.
The results from the descriptive analysis also show that 42% of households have access to
remittances. There is a substantial and significant difference in access to remittances
- 18 -
between households living in Mbankana, Dumi and Menkao. Households living in Mbankana
area tend to have a high level of access to remittances. About 53% of households in
Mbankana have access to remittances, against 32% for households in Dumi and 39.7 for
those in Menkao. Furthermore, the results of the survey show that about 50% of households
from the sample have access to extension services. The proportion of households with
access to extension services in Mbankana (0.55) appears to be higher than those of
households of Dumi (0.43) and Menkao (0.52), but the difference is not statistically
significant. The results of descriptive analysis show that about 53% of households in the
sample are members of social networks during the survey. There is no difference in network
participation between households living in the areas of Mbankana (50.7%), Dumi (52.3%)
and Menkao (55.9). The estimated percentage of credit-constrained households is about
71% for the sample households, while it is about 69% for households of the Mbankana area,
80% for the Dumi area and 64.7% for the Menkao area. The prevalence of credit constraints
tend to be higher among households in the Dumi area than in Mbankana and Menkao, but
the difference is not found to be statistically significant. The results also show that about
40% of household participate in off-farming activities. The proportion of households with
off-farming activities is higher in the Dumi area than in Menkao (44%) and Mbankana (40%).
The results also show that agricultural production in the hinterland of Kinshasa is dominated
by small scale farm-households. The average farm size is about 13 hectares. The average
farm size is about 15 hectares in Dumi, 14 hectares in Menkao and 12 hectares in Mbankana;
and the difference in mean between villages is not statistically significant. About 48% of
farm-households have landholding property (52% in the Mbankana, 43% in Dumi and 48% in
Menkao). The cassava production is the most cultivated crop in the hinterland of Kinshasa.
The average cassava production is estimated at about 20074 kilograms. The range is:
minimum, 2000 kilos; maximum, 81400 kilos. The average cassava production is higher in
Mbankana (21956 kilos) than the average cassava production in Dumi (19438 kilos) and
Menkao (18773 kilos), but the difference appears not to be statistically significant. The main
factors used in cassava production are land, labor and cassava planting materials. The
average cassava planting materials used is about 761 kilos for the sample. That of Mbankana
is 782 kilos higher than 752 for Dumi and 750 for Menkao. The average land used for cassava
production is about 1.64 hectares. The range is: minimum, 0.20 hectare; maximum, 8.14
- 19 -
hectares. The average land used for cassava production is about 1.79 hectares for
Mbankana, 1.62 for Dumi and 1.60 for Menkao; and the difference in mean between villages
is not statistically significant. Most of the farmers in the study area use both family and hired
labor. The average labor used for cassava production is about 278 man-days. The average
labor in Mbankana is estimated at 304 man-days, higher than in Dumi (266) and Menkao
(264); but these differences are not statistically significant.
The descriptive statistics show that household expenditure devoted to food consumption
varies greatly among villages. On average, food consumption is estimated at about 483 FC
per person for the whole sample. Household food consumption is reflected in certain village
patterns. The highest proportion of the mean food consumption is recorded in Menkao (549
FC) compared to Mbankana (428 FC) and Dumi (470 FC). The difference in mean is
statistically significant at 5%. The average food consumption is twice as high as the average
non-food consumption expenditures (237 FC). The average food consumption for three
groups of villages selected appears to be not statistically significant. The highest proportion
of mean non-food consumption expenditure is reached in the Menkao area (243 FC)
compared to with 239 FC for Mbankana are and 228 FC for Dumi area (228 FC). On average,
household total consumption per person is estimated at about 786 FC (SD=464 FC). The
highest level of total consumption is estimated at 3729 FC while the minimum value of total
household consumption is estimated at 210 FC. No significant difference in average total
household consumption per person is observed between the Mbankana area (724 FC), the
Dumi area (768 FC) and the Menkao area (862 FC). The average welfare index-1USD is
slightly over 1, indicating the average consumption per person and per day is over the
poverty line of 1USD. However, the average index-2USD is under the poverty line of 2USD
and estimated at about 0.71. There is no statistically difference in the mean welfare index
between selected village areas.
Data from MFIs were collected from RIFIDEC (Regroupement des Institutions du Système
Financier Décentralisé du Congo) and the Mix Market. The RIFIDEC is a nonprofit
organization created in 2000 with a grant from the German Technical Cooperation (GTZ).
RIFIDEC counts more than 300 microfinance organization members, of which 44 are
approved by the Central Bank of Congo. The main aim of RIFIDEC is to strengthen the ability
- 20 -
and professionalism of its members in the provision of financial services to the population
excluded from the standard bank system. To carry out this objective, it is involved in
different activities, such as promoting the development of microfinance, ensuring the
defense of the collective interests of its members, carrying out administrative control and
the management of the members, organizing technical assistance and training courses,
defining professional deontology rules, and supporting the solidarity and cooperation
between its members.
Mix Market is a nonprofit organization based in Washington and provides information for
approximately 2000 MFIs in all regions of the developing world. The information provided
covers social and financial performance, operational, product, client and social performance
data. Due to the lack of financial data, 32 microfinance organizations, members of RIFIDEC,
were selected in our sample for the year 2008. The data were extracted from financial
statements and additional information was collected from the annual reports or provided by
the microfinance organizations themselves. In addition, the sample was supplemented by
four MFIs that are not members of RIFIDEC; data from these four organizations were drawn
from “Microfinance Exchange Information, Mix market” and cover the year 2008.
1.7.2 Research methodology
The prevalence and the existence of credit constraints at household level are measured
using the direct approach. The direct method identifies credit constraints based on the
observation of household participation in credit markets by asking direct questions to
households about their credit status. The direct approach permits identification both of the
existence of credit constraints as well as their classification. This approach is the most
pertinent and most suited to addressing the issue of the identification of farm household
credit constraints. Numerous empirical studies have resorted to such direct methods of
identification (Kochar, 1997; Mushinski, 1999; Boucher, 2002; Maldonado, 2004; Bali, 2002;
Diagne et al. 2001).
The probit and logit models are used to identify key drivers affecting the probability of being
credit-constrained. The analysis of farm household credit constraints conditions gives a
- 21 -
qualitative dependent variable, which is a dummy variable taking value 1 if the household is
credit-constrained and 0 otherwise. Henceforth, the probit and the logit models are applied
to estimate the determinants of the probability that a farm household is credit-constrained.
These models are popularly employed in explaining farm household credit constraints and
access to credit (Nuryartono et al. 2005; Boucher et al. 2005; Waqar et al. 2008; Rahji and
Adeoti, 2010; Petrick, 2003).
The DEA is applied to measure technical efficiency scores in cassava production as well as
the efficiency of MFIs. The DEA was introduced by Charnes et al. (1978) for efficiency
analysis on the basis of the frontier concept. Farrell (1957) analyzed production efficiency
into two components: technical efficiency (TE), which measures the firm’s performance in
producing maximal output with a given set of inputs, and allocative efficiency (AE), which
evaluates quantitatively a firm’s ability to choose an optimum combination of a set of inputs
at given prices (Coëlli, 1996). The popularity of DEA in financial institution analysis is due to
the fact that it does not require an a priori assumption about the functional form of the
production function.
Assessing the efficiency of a Decision Making Unit (DMU) is performed on the basis of
observed data. The DEA application avoids the possibility of misspecification of the
production technology (Bauer et al. 1998). Furthermore, compared to other performance
measurements, DEA focuses on the outliers and allows the identification of firms with best
practices. Moreover, contrary to other frontier approaches, such as the Stochastic Frontier
Approach, DEA allows the computing of optimization problems in the context of multi-inputs
and outputs. Hence, the DEA approach applies to data on various inputs and outputs, and
allows an examination of the magnitude of inefficiency (Coëlli et al. 2005). Seiford and Thrall
(1990) asserted that the kind of mathematical programming procedure used by DEA for the
estimation of efficiency scores is comparatively robust. However, the disadvantage of DEA is
its assumption on the random errors, which is assumed to do not exist. Thus, DEA assumes
that all deviations from the frontier of best practices are attributed to inefficiency (Mester,
1996; Bauer et al., 1998). Furthermore, the standard formulation of DEA creates a separate
linear program for each DMU and incident factors, such as variation of organization
structure, climate, geographical location, soil type, economic conditions, and measurement
errors, are not accounted for in the DEA approach (Kumar, 2006).
- 22 -
The PSM proposed by Rosenbaum and Rubin (1983) is applied to the assignment of credit
constraint impacts on households. Credit constraints are non-random because of lender
selection and borrower self-selection. Indeed, given the agency relationship between
borrower and lender and the asymmetry of information related to credit contract, a lender
may select borrowers upon the basis of some observable characteristics, which provide the
information on the capacity of borrower to repay credit and cope with risk. There also might
be a geographic selection of borrowers based on the transaction costs related to the
distance. Moreover, the borrower may self-select to participate in a credit program based on
their risk aversion, preferences, motivations and resources. The PSM allows the issue of
selection bias to be addressed by comparing the mean outcomes of treated observations
and control units which have the similar pretreatment characteristics. The difference in
mean outcomes of the two groups is then attributed to the treatment. Two important
assumptions are required for applying the propensity score matching: these are known as
the balancing property and conditional independence assumption (CIA). Testing for a
balancing property is important for the comparability of constrained and unconstrained
groups and to ensure that the observations within each group are in fact similar. The CIA
allows the assertion that being credit-constrained is random and uncorrelated with
household welfare indicators if a set of observable characteristics are controlled.
The truncated regression (TR) is applied to explore the key factors affecting the TE
differentials among cassava farmers. The choice of a TR model is dictated by the nature of
the TE measure, which is truncated at 1, and by the findings of academic literature (Battese
and Coëlli, 1995; Simar and Wilson, 2007). Finally, the standard ANOVA test and the Mann-
Whitney test are applied to test whether there is a substantial difference in mean efficiency
scores between MFIs with agricultural credit provision versus those without agricultural
credit provision.
- 23 -
Chapter 1
Methods Logit & Probit PSM TR DEA
Chapter 2 Chapter 4 Chapter 3 Chapter 5
Data Household survey Microfinance survey
Chapter 6
Figure 1.5 Schematic presentation of research methodology
1.8 Overall dissertation structure
Besides the general introduction (Chapter 1), and conclusions and policy implications
(Chapter 6), this dissertation is organized as follows: Chapter 2 focuses on the determinants
of credit constraints for farm-households. Chapter 3 measures technical and scale efficiency
in the cassava production and the impact of credit constraints on technical and scale
efficiency scores. It also identifies key factors associated with TE scores differentials among
farm-households. Chapter 4 discusses the effect of credit constraints on a household’s
welfare distribution. Finally, Chapter 5 evaluates the efficiency of MFIs operating in the DRC
with respect to the financing (or not) of agricultural activities. This structure is presented in
Figure 1.6.
- 24 -
Framework Chapter 1
Household level Microfinance

level
Credit Chapter 2
conditions
Impact of credit Chapter 3 Chapter 4

conditions
Performance of Chapter 5
microfinance
Implications Chapter 6
Figure 1.6 Schematic presentation of dissertation organization
App. Table 1.1 Descriptive statistic of quantitative variables
Variable N Mean SD Min Max F (P.value)

Education 202 9.43 3.85 0 18 0.683(0.506)
Mbankana area 69 9.65 3.55 0 18
Dumi area 65 8.97 4.12 0 17
Menkao area 68 9.65 3.91 0 15
Farm size 202 13.34 11.45 2.5 60 1.943(0.146)
Mbankana area 69 12.72 8.74 5 100
Dumi area 65 15.58 17.00 5 30
Menkao area 68 11.84 5.69 3 100
- 25 -
Household size 202 4.95 1.84 1 11 3.371(0.036)
Mbankana area 69 4.57 1.81 1 10
Dumi area 65 5.38 1.72 2 9
Menkao area 68 4.93 1.92 2 11
Household head age 202 45.05 9.80 22 66 0.514(0.599)
Mbankana area 69 44.14 10.36 28 66
Dumi area 65 45.85 9.49 25 65
Menkao area 68 45.21 9.59 22 65
Dependency ratio 202 0.50 0.22 0 0,9 3.650(0.027)
Mbankana area 69 0.49 0.23 0 0.9
Dumi area 65 0.44 0.22 0 0.9
Menkao area 68 0.55 0.21 0 0.9
Land used for cassava 202 1.67 1.02 0.20 8.14 0.671(0.512)
Mbankana area 69 1.79 1.02 0.60 5.20
Dumi area 65 1.62 1.16 0.45 8.14
Menkao area 68 1.60 0.86 0.20 4.20
Labor used for cassava 202 278.51 231.29 29.41 1544.11 0.650(0.523)
Mbankana area 69 304.27 269.47 29.41 1544.11
Dumi area 65 266.15 203.58 60.24 1097.82
Menkao area 68 264.19 214.83 45.91 1334.00
Cassava materials 202 761.62 722.60 74 5542 0.041(0.959)
Mbankana area 69 782.01 914.92 74 4070
Dumi area 65 752.09 665.98 88 2542
Menkao area 68 750.05 540.90 140 5542
Cassava production 202 20074.80 14964.09 2000 81400 0.860(0.424)
Mbankana area 69 21956.66 16511.11 2000 78000
Dumi area 65 19438.46 14528.36 5000 81400
Menkao area 68 18773.52 13700.55 4500 81000
Food consumption 202 483.48 287.68 100 1864 3.196(0.043)
Mbankana area 69 549.59 349.92 175 1220
Dumi area 65 470.83 218.30 109 1467
- 26 -
Menkao area 68 428.49 266.02 100 1864
Non-food consumption 202 237.19 236.55 15.23 2353.33 0.07(0.932)
Mbankana area 69 243.35 181.55 27.50 930.00
Dumi area 65 228.40 303.24 17.08 2353.33
Menkao area 68 239.34 215.54 15.23 1151.66
Total consumption 202 785.94 464.47 210.99 3729.49 1.60(0.205)
Mbankana area 69 862.87 464.26 283.97 2703.65
Dumi area 65 768.45 490.18 210.99 3729.49
Menkao area 68 724.59 434.42 253.60 2559.20
Welfare index 1 USD 202 1.40 0.82 0.37 6.65 1.60(0.205)
Mbankana area 69 1.54 0.82 0.50 4.82
Dumi area 65 1.37 0.87 0.37 6.65
Menkao area 68 1.29 0.77 0.45 4.57
Welfare index 2 USD 202 0.71 0.41 0.18 3.32 1.60(0.205)
Mbankana area 69 0.77 0.41 0.25 2.41
Dumi area 65 0.68 0.43 0.18 3.32
Menkao area 68 0.64 0.38 0.22 2.28
App. Table 1.2 Descriptive statistic of dummy variables
Variable N % Khi 2 (P.value)

Network participation 202 53.0 0.383(0.826)
Mbankana area 69 50.7
Dumi area 65 52.3
Menkao area 68 55.9
Remittance 202 42.1 6.476(0.039)
Dumi area 65 32.3
Menkao area 68 39.7
Off-farming 202 40.6 0.849(0.654)
- 27 -
Dumi area 65 44.6
Menkao area 68 36.8
Extension 202 50.5 2.172(0.338)
Dumi area 65 43.1
Menkao area 68 52.9
Credit status 202 71.3 3.950(0.139)
Dumi area 65 80.0
Menkao area 68 64.7
Gender of household head 202 91.1 1.363(0.506)
Dumi area 65 87.7
Menkao area 68 92.6
Landholding property 202 48.0 1.120(0.571)
Dumi area 65 43.1
Menkao area 68 48.5
- 28 -
Chapter 2
Determinants of Farm Household Credit Constraints in the Hinterland of Kinshasa
2.1 Introduction
Despite the proliferation of MFIs operating in the DRC and especially in Kinshasa, it is
surprising to note that credit constraints continue to be one of the main characteristics of
farm-households. To the best of our knowledge, there is a lack of academic studies from the
DRC focused on this topic. Furthermore, the literature on microfinance is hampered by the
lack of household level data to identify farm-households who are credit-constrained, while
numerous studies have focused on access to microcredit (e.g., Kodjo et al. 2003; Omonona
et al. 2008; Yehuala, 2008; Diagne, 1999; Diagne and Zeller, 2001). There is little empirical
evidence on the profile of farm-households who are more likely to experience microcredit
rationing (Maldonado, 2004; Boucher and Carter, 2002; Zeller, 1994). Furthermore, these
few studies on microcredit rationing towards farm-households have come to diverse
conclusions regarding the importance and magnitude of factors that result in the farm
household being rationed by MFIs. In addition, these previous studies have used data from
large MFIs receiving financial support from donors, while there is a lack of empirical
evidence on the drivers affecting the probability that a farm household is rationed by small
MFIs, which generally face high liquidity constraints.
The aim of this study is to identify the key drivers affecting the probability that a farm
household is rationed by small microfinance organizations. Using evidence from the
hinterland of Kinshasa, which is characterized by the exclusive existence of small scale MFIs,
we firstly, propose an empirical approach of elicitation of microcredit rationing toward farm-
households, which allows the identification of farm-households which are credit-constrained
and those which are unconstrained. This approach also helps to identify the major
constraints related to financing of agriculture by MFIs. Secondly, we use information on
institutional and household characteristics to construct a model explaining the probability of
a household being credit-constrained. This chapter is organized as follows. In the next
section 2.2, a review of the literature on credit constraints and rationing is given. The
- 29 -
literature review is restricted to the understanding of credit rationing using the contribution
of neoclassical, contractual and conventional theoretical constructions and to the specific
constraints related to the financing of agriculture. In section 2.3, the data and the research
methodology are presented. In section 2.4, the results are presented and discussed. Section
2.5 provides a conclusion.
2.2 Literature review
Neoclassical models define the market as a place of coordination of economic agents,

interacting through the system of the market price, which is determined by an internal
instance. For neoclassic economists, the market price contains all the information allowing
economic agents to make optimal and rational decisions (DeBriey, 2003; 2005). The
neoclassical economists make economic analyses under the assumption of the market
clearing equilibrium, where the quantity supplied equals the quantity demanded. They
assume that if the quantity demanded is not equal to the quantity supplied, an adjustment
will be made and the equilibrium will be automatically restored through the market price. In
terms of the credit market, the neoclassical construction underlines that interest rates lead
to equilibrium between supply and demand of loan. Moreover, as markets are assumed to
be working clearly and properly, the interest rate contains all the information. Therefore, the
adverse selection and moral hazard problems, as well as quantity credit rationing, should not
exist (Kochar, 1997; Maldonado, 2004; DeBriey, 2003; 2005). Consequently, transactional
costs related to credit are supposed to be null as the lender knows the risks related to the
potential borrowers without cost and can discriminate between them through market
interest rates (Kochar, 1997; Maldonado, 2004; DeBriey, 2003; 2005).
The neoclassical construction is often viewed as restrictive and divorced from reality,
especially for developing nations where there are significant problems related to market
information, such as high moral hazard and high adverse selection (DeBriey, 2003; 2005). In
contrast to what is assumed in the neoclassical model, empirical studies have established
the existence of credit constraints and quantity rationing among economic agents,
particularly poor farm-households and small business (Kochar, 1997; Mushinski, 1999;
Boucher and Carter, 2002; Maldonado, 2004). Credit constraints are particularly important
- 30 -
in developing nations. There are significant difficulties surrounding credit transactions, such
as incentive incompatibilities, contract enforcement problems, a lack of legal and judicial
framework, undeveloped physical and institutional infrastructures and the absence of
efficient and sustainable financial institutions (González-Vega, 2003).
An important theoretical advance in the understanding of credit constraints and rationing

mechanisms was reached by the interpretation of contractual theory and especially the
agency theory (DeBriey, 2003; 2005). Following Jensen and Meckling (1976), the agency
relation is characterized by a situation in which the principal engages the agent to execute
tasks involving the delegation of decision-making power. The contract between the agent
and principal often features an informational asymmetry creating the agent’s opportunistic
behavior to the detriment of the principal (Coase, 1937; Jensen and Meckling, 1976; Jensen,
1983; Fama and Jensen, 1983). This behavior is likely to reduce the economic efficiency in
credit market transactions, such as adverse selection and moral hazard, leading to high
transactional costs.
Regarding credit contracts, DeBriey (2003) noted that the relation between borrower and
lender can be considered as an agency relation in which the lender grants a part of his funds
to the borrower, and the borrower promises to repay upon the basis of conditions well
defined in advance. The agency problem results in the fact that the interests of lenders and
borrowers do not always conform. Thus, the presence of information asymmetry may lead
the lender to ration some potential borrowers using collateral requirements, interest rates
and other loan contract terms and conditions. Stiglitz and Weiss (1981) have argued that,
within a context of incompleteness of information, a credit market with equilibrium could be
characterized by rationing due to the existence of informational asymmetry between lenders
and borrowers. Stiglitz and Weiss (1981) defined a situation of equilibrium with rationing as
a phenomenon appearing when only some of potential borrowers applying for a loan receive
the full amount they have requested, even if they would all accept to repay at equilibrium
interest rates. They also showed that credit rationing appears when there are some
individuals who apply for a loan and are incapable of obtaining it at any level of interest, but
who would be able to obtain it in the context of increased quantity supply. For Stiglitz and
- 31 -
Weiss (1981), an increase in interest rates to adjust the supply of a loan to its demand could
be a source of adverse selection and moral hazard risks.
Another important theoretical advance in the understanding of credit rationing mechanisms

was reached through the interpretation of conventional approaches. Here, the analysis of
social and economic phenomena is carried out by applying a multidisciplinary theoretical
framework utilizing economics, philosophy, sociology, etc. The conventional approaches
recognize various mechanisms of coordination between contracted parties, resorting to so-
called “conventions”. The conventions are defined as common knowledge, regularity of
behavior as well as collective landmarks, avoiding the explanation of phenomena (Boltanski
and Thevenot, 1987; 1991; Duvernay, 1989). The conventional approaches assume the
limited rationality of the agent, depending on the available information (Isaac, 1996; Gomez,
1995). In the light of conventional approaches, Honoré (1995) explains credit rationing by
the conventions used by lenders to cope with risks in an uncertain environment. For Honoré
(1995), the impossibility of setting up applications, allowing the risk evaluation and some
subjective and objective common knowledge, could explain credit rationing.
The provision of credit towards farmers and rural households in developing nations, such as
the DRC, is particularly problematic because of high transaction costs for both borrowers
and lenders. The transaction costs represent for borrowers a set of costs related to the
access to loans and their repayment. As mentioned in Labie (1999), the transaction costs for
borrowers are related to the loan application, documentation, transportation, time spent in
monetary equivalent, etc. Theoretically, it is assumed that the more the transaction costs,
the less potential borrowers will apply for loans (Calkins et al. 1992). For lenders, transaction
costs are related to the evaluation of the quality of the borrowers, the lack of infrastructure
in rural areas and the seasonal characteristics of agricultural production (Lapenu, 2001).
Another constraint on the financing of agricultural activities is related to the difficulty of

applying market interest rates to agricultural credit programs. Indeed, the financing of
agriculture involves high transaction costs, leading lender to apply high interest rates to
cover costs to enable them to grant credit sustainably. However, the application of high
interest rates to cover high transactional costs may be very risky for lenders as it increases
- 32 -
moral hazard and adverse selection risks. Moreover, given the farmer’s objective, which is to
optimize profit or minimize the costs of production (Broussard, 1987; Fei and Ranis, 1964),
an increase in interest rates could result in a decrease in loan applications given the low
returns on agricultural activities. Likewise, the application of the low interest rates to adapt
credit to the low profitability of agriculture could lead to the decreasing of loan provisions.
The loan conditions applied in MFIs (short maturity periods, frequent repayments, collateral
required and small loan amounts) are among constraints related to the provision of micro
credit towards farmers. Indeed, given short loan terms and the frequency of repayment
applied in microfinance, farmers would be likely to abandon participation in the credit
market. In addition, lenders could be less interested in financing agricultural activities
because of the incompatibility of lending technology and lack of long term resources needed
in agriculture. Kodjo et al. (2003) noted that “the longer the maturity of the loan, the more
the farmer has a chance to obtain credit”. Concerning loan collateral, it is generally assumed
in microfinance that social capital plays the role of collateral. However, empirical studies
have demonstrated the difficulty of making this system of guarantees work efficiently.
Consequently, the absence of the physical collateral is more likely to be a determining factor
for a farmer not to apply for a loan and for lenders to discriminate against poor farmers
without physical collateral (Kodjo et al. 2003).
Agricultural activities, especially in developing nations, are characterized by high covariance

risks which constrain the provision of credit towards farmers (González-Vega, 2003) and
farmer’s loan applications. In effect, beyond moral hazard and adverse selection risks, which
are often reported in the literature to explain the difficulty of financing small business,
farming activities have to face specific and various risks. Given high risks related to
agricultural activities, honest and more risk-adverse farmers might give up on the existing
agricultural credit market. The high risks related to agricultural activities could also lead
lenders to restrict or abandon the provision of credit towards farmers, given the lack of
physical collateral and insurance at farmer’s household level.
- 33 -
2.3 Data and methodology
The data used in this chapter were collected from farm-households in the hinterland of
Kinshasa. For a detailed description of sample and data collection, refer to Chapter 1. The
description of data used in this chapter is reported in the App. Table 2.1. The existence of
credit constraints was identified using the direct approach. We directly asked questions to
households about their credit constraints conditions. We asked households if they had
applied for a loan or not. Households who did not apply for a loan because of reasons
including the high interest rates, the high transaction cost of credit, the inappropriate loan
maturity period and frequency of repayment, the lack of collateral required, the lack of
information on credit provision, the high risk related to agricultural activities, the low returns
of agricultural activities, the shame of being in debt and the belief that MFIs do not lend to
the poor are classified as credit-constrained (Fig. 2.1). Households who apply for a loan and
do not receive the full amount are classified as partially constrained. The possible reasons
for being partially credit-constrained are the lack of collateral, the lack of funds at the level
of the financial institution, the high amount of credit demanded or applied for and the
procedure not being respected by borrowers. Households who do not apply for a loan
because they do not need it, as well as those who apply and receive the full amount applied
for, qualify as credit unconstrained.
- 34 -
Figure 2.1 Empirical approach used to identify farm household credit constraints
Did you apply ?
Yes No
Obtained ? Why ?
Yes No
Lack of collateral
Full amount ? Why ?
Loan maturity
Small credit amount

ye No Lack of collateral
s
High risks
Lack of funds
High interest rates
Why ? High demand
Shame of debt
Low returns
Bad experience
Procedure
Not for farmer
Lack of information
Constrained
Low returns
Transactional costs
Unconstrained No need
To determine the key factors affecting household probability of being credit-constrained, we

applied logit and probit models. Thus, we assumed an underlying latent response variable
d * which is likely to be a linear function of a set of independent variables X . That is,
d i*  X   i (2.1)
Where a set of parameters  reflects the impact of changes in the probability of being
credit-constrained. However, d * is not observed: we only observed a dummy dependent

variable d , which takes value 1 if household is credit-constrained and 0 otherwise. So, we
observed d  1 if d i*  c and 0 otherwise. Therefore,
Pr(d i*  c)  Pr( x '    i  c)  Pr( i  c  x '  ) (2.2)
If the probability function F (.) is symmetric (as with normal and logistic distribution), then
- 35 -
Pr(d i*  c)  F ( x '  ) (2.3)
Several probability functions F (.) are employed in the literature. Those commonly used are
the normal cumulative distribution function (.) , which implies the application of the probit
model and the logistic cumulative distribution function (.) giving rise to the logit model.
Under the probit model, it is assumed that the error term follows a standard cumulative
distribution. In contrast, under the logit model, the error term follows the logistical
distribution. The density function of the probability under probit model and logit model is
given by, respectively:
t 2
1 x'
Pr(d i  1)  ( x'  ) 
2 
e 2
dt (2.4)
ex 
'
Pr(d i  1)  ( x  ) 
'
(2.5)
1 ex 
'
The coefficients obtained from the probit and logit model are difficult to interpret because
they measure the change in the unobservable d* associated with a change in one of the
explanatory variables. The alternative measure is what is called the marginal effect ( MEi ) ,
that is,
P(d i  1) F ( X )
Probit model, MEi   (2.6)
X i X i
P(d i  1) e X
Logit model, MEi   F ( X )  i   i (2.7)
X i (1  e X ) 2
Three groups of explanatory variables used as potential covariates are expected to influence
the probability that a farm household is credit-constrained. The first group refers to
household composition and demographic characteristics, such as household size (HHSIZE),
dependence ratio (DRATIO), the age of the household head (HHAG), the level of formal
education of the household head (HHEDUC), the gender of the household head (HHMALE)
and household network participation (ASSOCIATION). These variables are used as proxy for
household social and human capital. Large size households almost always have significant
credit demands which cannot be entirely funded by MFIs. Thus, one may expect household
size to increase the probability of being credit-constrained. However, as more adult
- 36 -
households would have higher income-generating capability, it is possible to expect a
positive relationship.
Households with large dependency ratio are often expected to face labor constraints and to
consume more than they produce. Thus, the dependency ratio is assumed to increase the
probability of being credit-constrained. Household heads will be more likely to have many
contacts, opportunities to obtain employment, and good production management and skills,
which, in turn, will help taking healthier production decision. Thus, they will be expected to
be less constrained. Older farmers are expected to use the credit in an optimal way. In
addition, informal lenders may grant loans to older persons to obtain power and social
prestige within village communities (Calkins et al. 1992). Furthermore, older people have
many personal connections that would serve as social collateral. Thus, one could expect the
age of household head to be inversely related to the probability of being credit-constrained.
Women lack control over economic resource and men often have more exposure to
information (Kodjo et al. 2003). Therefore, households led by females will be more exposed
to credit constraints. However, given the targeting of women by NGOs and the idea that
women repay more reliably than men, one could expect farm-households led by males to be
more exposed to credit rationing.
The second group refers to household physical and financial capital of households. This
groups includes landholding property (LANDPROPERTY), farm size (FARMSIZE) and
remittances (REMITTANCES). Households with large cultivated land will be more likely to
face labor constraints, and thus will need external funds. Lending institutions, especially
informal lenders, typically resort to the landholding property of the borrower to enforce
contracts and collateralize loan (Besley, 1995b). Thus, households with landholding property
will be less exposed to credit constraints. Households with access to remittances will have a
high capacity to cope with risks related to agricultural activities, and will adopt more
productive technologies, allowing higher income generation (Mendola, 2008; Pleitez-Chavez,
2004). This in turn, will reduce the risk of being credit-constrained.
The last and third group refers to the village of residence, which is used to capture the effect
of local institutional characteristics, such as distance to market, density of population,
- 37 -
number of civil society groups and NGOs, access to transport, extension services, etc. Three
dummy variables were constructed and used for empirical analysis. These include villages
located in the area of Mbankana (GROUPMBAKA), Dumi (GROUPDUMI) and Menkao
(GROUPMENKAO). The decision to use groups of villages instead of the village itself was due
to the low density of population in some selected villages. As 15% of households from each
village were randomly selected, it appeared that most of the villages with a low density of
population had less than 5% of selected sample households. These groups of villages are in
the same climatic zone with the same temperature and duration of dry period within the
growing season. However, there are many differences in terms of market access and the
number of operating local institutions. Menkao and villages located in the area of Menkao
are more likely better integrated to market as there is a relative big local market in the area.
Furthermore, the area is close to Kinshasa, and thus may reach the markets of Kinshasa at
minimum cost in terms of distance, time and cost of the transport. However, there are a
limited number of NGOs and civil society groups. Compared to Menkao and the villages in
this area, Dumi and the villages located in the Dumi area have small market and little bite far
from Kinshasa’s markets. To reach Kinshasa markets and institutions, they have to spend
much more time and money compared to the households living in the Menkao area. In
addition, there limited number of NGOs and other local institutions. Mbankana is about 150
kilometers from Kinshasa, thus reaching Kinshasa’s markets for households living in this area
is really costly, in terms of distance, time and transport costs. However, compared to the
Dumi and Menkao areas, there are many civil society institutions and NGOs, such CADIM and
World Vision, with extension services and development projects.
- 38 -
Table 2.1 Definition, measures and expected signs of covariates
Expecte
Variables Unity or type Explanation
d signs
HSIZE Continuous Size of household +
Head household Male = 1

HHMALE Binary +/-
Head household Female = 0
HHAG Continuous Age of household head in years -
DRATIO Continuous Dependency ratio +
Years of formal education of household

HHEDUC Continuous -
head
With landholding property (2007) = 1

LANDPROPERTY Binary -
Without landholding property(2007) = 0
With remittance = 1
REMITTANCES Binary -
Without remittance = 0
ASSOCIATION Binary Member (2007) = 1 No Member (2007) =0 -
SIZEFARM Continuous Size of farm in hectare +
GROUPMBAKA Binary Yes =1, No =0
GROUPDUMI Binary Yes =1, No =0
GROUPMENKAO Binary Yes =1, No =0
2.4 Results and discussions
2.4.1 Results of household’s responses
Table 2.2 presents the distribution of the number household loan applications with
reference to the categories of lenders. Of the 120 households who applied for a loan, 60%
made their loan application in the semi-formal sector of microfinance, including NGOs (45%),
closed credit and saving cooperatives (13%) and open credit and saving cooperatives (12%).
We observed that the semi-formal MFIs have a high level of credit rationing. The rate of
- 39 -
credit rationing is about 63% for NGOs, 56% for closed credit and saving cooperatives and
42% for open credit and saving cooperatives. It is important to note that, despite the high
level of credit rationing among those semi-formal financial institutions, they still represent
an important part of the credit market in the hinterland of Kinshasa. In light of Table 2.2, we
note that 40% of households who apply for loans made their requests in the informal
microfinance sector: 18% of loan applications were made to money lenders, and 12% to
tontines and other informal lenders. Money lenders rationed 59% of loan applications
received while tontines and others informal lenders rationed 42% of loan applications.
Table 2.2 Number of loan applications by lenders
Nature of Financial agency Demands Demands Demands

applied rationed non-rationed
Money lenders 22 13 9
NGOs 54 34 20
Closed Credit and Saving Cooperatives 16 9 7
Open Credit and Saving Cooperatives 14 6 8
Tontines and others informal lenders 14 6 8
In the light of Table 2.2, we find that credit rationing is very important both in informal and
formal sector of microfinance. As one can observe (see Table 2.3), the amounts of credit
demanded and supplied are very small. The average loan applied to NGOs is estimated at
155487 FC (about 300 USD), 92750 FC (about 200 USD) for money lenders, 100000 FC (about
200 USD) for closed credit and saving cooperatives, 55571 FC (about 100 USD) for open
credit and saving cooperatives and 105000 FC (less than 200 USD) for tontines and others
informal lenders. This also holds true for the actual loans supplied. The average loan is
estimated at 109363 FC (about 200 USD) for the NGOs, 90670 FC (less than 200 USD) for
money lenders, 87857 FC (less than 200 USD) for closed credit and saving cooperatives,
96071 FC (less that 200 USD) for tontines and other informal lenders and less than 50 USD
for open cooperatives.
- 40 -
Table 2.3 Volume of loan demanded and supplied by lenders (in FC)
Categories Demand of microcredit Access to microcredit
Means Min. Max Means Min Max.
Money lenders 92750 15000 270000 90670 11000 270000
NGOs 155487 25000 550000 109363 20000 440000
Closed cooperatives 100000 35000 250000 87857 20000 250000
Open cooperatives 55571 15000 165000 37458 10000 82500
Tontines and others 105000 20000 250000 96071 15000 250000

Source: Our survey
The results from household responses (Tables 2.2 and 2.3, and Figure 2.2) report a high level
of credit constraints among farm-households in the hinterland of Kinshasa. Indeed, 71% of
farm-households of the sample are classified as being credit-constrained. Among
constrained households, 38% faced liquidity constraints but did not apply for a loan, 11%
applied and were rejected and 22% are partially constrained. The analysis of the responses
(Figure 2.2) shows that the main reasons for not applying for a loan are related to the
conditions, terms and the lending technology applied by the existent microfinance
organizations. Households reported that they do apply for loans because of lack of collateral
required (79%), the short maturity of the loans granted, which is not appropriate to
agriculture (83%), the small size of credit provided (76%) and the high interest rates applied
by lenders (82). About 31% of sample households applied for loan and were rejected or
partially rationed. Those households reported that the main reasons were the lack of
collateral required (65%) and procedures not being respected by borrowers (31%).
2.4.2 Results of estimations
- 41 -
Tables 2.4 and 2.5 present the results of logit and probit estimations, respectively. We
assessed the significance of a set of covariates using the p-values associated with the
likelihood ratio  2 (.) statistic. P-values approximately equal to zero suggest that the all
parameters are jointly significant, and thus the null hypothesis can be rejected.
Table 2.4 Results of logit model of credit constraints
Variables Coefficients Std. Err. Z-statistic Marginal effect
Education -0.0229 0.0456 -0.50 -0.0042
Household head male 0.3327 0.5777 0.58 0.0660
Household head age -0.0047 0.0171 -0.28 -0.0008
Dependency ratio -0.2376 0.7511 -0.32 -0.0442
Household size 0.2668*** 0.1205 2.12 0.0496
Land property holding -1.2961*** 0.3882 3.34 -0.2312
Farm size 0.1506 0.1807 0.87 0.0294
Remittance -0.7611*** 0.3626 -2.10 -0.1454
Association -0.8051** 0.4402 -1.83 -0.1374
Mbankana Group 0.4031 0.3993 1.01 0.0725
Dumi Group 0.5789 0.4665 1.24 0.1020
Constant 1.1044 1.1366 0.97 -
Prob.>  2 = 0.002 Pseudo R 2 = 0.1427 Log pseudo likelihood = - 103.8329 Wald  2 = 29.94
Key: * p<1; ** p<.05; ***p<.01
- 42 -
Table 2.5 Results of probit model of credit constraints
Variables Coefficients Std. Err. Z-statistic Marginal effect
Education -0.0163 0.0272 -0.60 -0.0052
Household head male 0.1928 0.3522 0.55 0.0648
Household head age -0.0031 0.0103 -0.31 -0.0010
Dependency ratio -0.1592 0.4460 -0.36 -0.0510
Household size 0.1586*** 0.0652 2.43 0.0508
Land property holding -0.7556*** 0.2154 -3.51 -0.2325
Farm size 0.0834 0.0938 0.89 0.0267
Remittance -0.4804*** 0.2106 -2.28 -0.1568
Association -0.4657** 0.2433 -1.91 -0.1392
Group Mbankana 0.2269 0.2344 0.97 0.0708
Group Dumi 0.3214 0.2651 1.21 0.0987
Constant 0.7328 0.6671 1.10 -
Prob.>  2 = 0.001 Pseudo R 2 = 0.1435 Log pseudo likelihood = - 103.7358 Wald  2 = 33.49
Key: * p<.1; ** p<.05; ***p<.01
For both logit and probit models, the likelihood ratio statistic is estimated at about – 103,
2
and both regression models are significant with 1% significance. The Pseudo- R is estimated
at about 0.14 for both logit and probit models. The signs affected to the significant
- 43 -
parameters of explanatory variables are consistent with the expectations. We examined the
dataset to test for evidence of multicollinearity. Following Greene (1997), multicollinearity
may affect the precision of individual effects of explanatory variables. To test the evidence of
multicollinearity, the VIF test is computed and the results are reported in the App. Table 2.3.
In the light of these results, one might reject the hypothesis of multicollinearity as all VIF
values are less than the critical value of 10. To avoid possible endogeneity problems related
to covariates potentially endogenous, extension and off-farm activities are not considered
for the estimation of probit and logit models, while other such landholding and association
membership are lagged. The results of the probit and logit models (Tables 2.4 and 2.5)
indicate that four explanatory variables are significantly associated with farm household
credit constraints conditions. Landholding property, remittance and association membership
are negatively associated with being credit-constrained while the coefficient of household
size is positive and significant at 5%.
2.4.3 Discussion of results
We start by discussing the findings from household responses. The results of households’
responses (Figure 2.2) show some extent of credit constraints conditions due to the
inappropriate lending technology applied by MFIs. High interest rates, short loan maturity
periods and frequent repayments, the requirement for physical collateral and small loan size
are the key drivers of household decisions not to participate in the microfinance market. This
is in line with our hypothesis that the lending technology applied in the MFIs constrains
farm-household participation in the microcredit market. This finding calls for efforts on the
part of MFIs to adapt and diversify their lending methodology to meet the needs and the
specificities of agricultural activities. Loan conditions and terms (repayment periods and
frequency) should be flexible and adapted to the farm-household’s cash flow. Interest rates
should be adapted to the agricultural returns and non-physical collateral should be accepted.
Moreover, rationed farm-households also report higher demand for loans than the available
resources of MFIs to be among the key reasons of being credit rationed, confirming our
hypothesis that the limited financial resources at the levels of MFIs reduce their capacity to
respond to the scale of demand for microcredit among farm-households. This is not
- 44 -
surprising, as most MFIs with transactions in agricultural activities in the periphery of
Kinshasa are local initiatives and depend on local small savings collected. Most of them do
not have access to resources from financial capital or subsides from donors and
governments. Thus, there is a need for public policy to increase financial resources and the
assets of MFIs operating in rural areas. Likewise, farm-household responses tell us that the
high risks and low returns of agriculture activities are among the reasons for not applying for
a loan. This finding supports our hypothesis that the nature of agricultural activity constrains
farm-household access to microcredit. Therefore, agricultural policy aiming at securing
agricultural activities and improving their profitability will increase the participation of farm-
households in the credit markets. Not surprisingly, the responses of sample farm-households
conform to the literature on the constraints related to the financing of agriculture (Section
2.2).
The estimated results of logit and probit models indicate that household size, association
participation, landholding property and remittances are the key factors affecting the
probability of a farm-household being credit-constrained in the hinterland of Kinshasa. This
confirms the hypothesis that capital endowments reduce the probability that a farm-
household is credit-constrained. The household size is found to be positively associated with
the probability of being credit-constrained at the significance level of 1%. This means that an
increase in household size will increase the likelihood of being credit-constrained. The
positive effect of household size on the propensity of being credit-constrained was found in
Nuryartono et al. (2005) and Oyedele et al. (2009). Two possible arguments might explain
the observed positive direction of the relationship. Large households have an important
credit demand which cannot be entirely satisfied by microfinance, given the small loan sizes
granted. Large households in developing nations, as in the hinterland of Kinshasa, are
vulnerable and poor. The informal lenders and the formal MFIs working under the
commercial framework are less interested in financing the activities of the very poor
because of the high risk and the lack of collateral. In terms of policy implication, the positive
and significant effect of household size calls for family planning to be included as one of the
extension packages. A restrictive demographic policy is likely to reduce credit liquidity
constraints by reducing household size and dependency burdens.
- 45 -
The association participation is found to be statistically significant (at 5%) and negatively
associated with credit constraints conditions. This means that network participation
decreases the risk of being credit-constrained. This might be due to the fact that network
participation is viewed as an indicator of the ability to respond to socio-economic
engagements. Given the lack of physical collateral, due to the high level of poverty among
farm-households, some microfinance organizations, notably NGOs, use social capital as
collateral and as a mechanism to cope with moral hazard and adverse selection risks.
According to Zhou and Bankson (1996) and Coleman (2002), social capital permits a flood of
information between lenders and borrowers and reinforces social control. Thus, the policy of
promoting social economy by improving farmers’ participation in social networks, such as
farm cooperatives, by providing them with non-physical collateral and information, and by
reducing transactional costs for both borrower and lender, will reduce the prevalence of
credit constraints among farmers.
The landholding property is negatively and significantly associated with household credit
constraints conditions (at 1% of level of significance). This is in line with the assumption that
credit constraints tend to be the inverse function of household wealth. This finding is
consistent with Kochar (1997) and Nuryartono et al. (2005), who reported that farmland
owned is among the main key factors affecting credit constraints conditions. In the
hinterland of Kinshasa, land title is used by some lenders, especially the informal lenders, as
loan collateral. According to Besley (1995b), formal land property title would be used by
lenders as collateral and could allow a lowering of the risk of collateral loss since households
with landholdings are viewed as less risky, even when the loan is not collateralized with land
property. This finding suggests that public agricultural policy to improve farmers’ access to
land property will reduce the risk of being credit-constrained, as land title is used as
collateral.
Access to remittances is found to be a key driver of the probability of being credit-

constrained. Households with remittances tended to be less exposed to the credit
constraints. The negative and significant effect of remittances could be explained by the fact
that remittances may increase the capacity of households to cope with systemic risk, but
also act as a form of collateral (Pleitez-Chavez, 2004). Moreover, access to remittances
- 46 -
increases household income and helps to overcome liquidity constraints and reduce income
fluctuations (Lucas, 1987). Likewise, access to remittances would enable households to
smooth consumption (Taylor et al. 1997), adopt riskier but more productive technologies,
and allow the accumulation of capital ), adopt riskier but more productive technologies, and
allow the accumulation of capital (Mendola, 2008; Woodruff and Zenteno, 2007; Pleitez-
Chavez, 2004).
Households with a large dependency ratio are often expected to face labor constraints and
to consume more than they produce. Thus, their saving rate is expected to be null or
negative. Following the above argument, we expected that the dependency ratio could have
a positive impact on the probability of being credit-constrained. However, the estimated
results show that there is a negative and not statistically significant effect of dependency.
This implies that there is no strong evidence to support the effect of dependency ratio on
household credit constraints in the hinterland of Kinshasa. The result is not consistent with
Dong et al. (2010), which reported significant effects of dependency ratio on credit
constraints conditions. The possible explanation here could be the participation of children
in most economic activities.
The household head education is negative and insignificantly associated with a household’s
credit-constrained status. This direction of relationship is consistent with our expectations
that education could reduce the risk of being credit-constrained. This was supported by
several assumptions. Firstly, education could provide better production management and
skills, which are more likely to assist household to take better production decisions.
Secondly, more educated household heads might have a large number of contacts and
opportunities and will have a higher probability of obtaining employment and diversifying
income resources. Thirdly, more educated household heads have initial advantages, such as
better entrepreneurial abilities, that can be translated into income since lenders prefer
wealthier borrowers. Moreover, the insignificant education coefficient observed in this study
does not imply that the education policy is not important in the hinterland of Kinshasa.
Instead it does not significantly affect the chance of farm-households receiving the full
amount of the loan applied for. This finding is consistent with Nuryartono et al. (2005).
- 47 -
The age of household head is assumed to be a proxy for managerial capacity, experience,
responsibility, confidence and the capacity to repay. Older farmers are expected to use
credit obtained in an optimal way and, consequently, would be less exposed to the risk of
repayment default. Moreover, age is associated with a certain reputation in rural
communities, which might positively impact lender’s decision to grant a loan. According to
Calkins et al. (1992), informal lender could provide credit to an elder, or to a group of older
people, simply to obtain power and social prestige in the community. Furthermore, an older
individual has personal connections in the community that would serve as social collateral
for a loan application. Following the above arguments, we expected that household head
age would be negatively associated with the probability of being credit-constrained. The
results of our estimations report negative and no significant effect of household head age.
This implies that the variations in household head age across the sample do not have
significant impact on the probability of being credit-constrained. The insignificant impact of
age on the likelihood of being credit-constrained was found in Baydas et al. (1994) and
Zagarra et al. (2008). This finding does not conform to the positive and significant effects
found in Nuryartono et al. (2005), Omonona et al. (2008) and Dong et al. (2010).
The variable household head male is expected to be an ambiguous sign. We expect that its
effects could be positive because lending methodology often favors women, who are
targeted by NGOs; and because a belief in the microfinance world that women repay better
than men. However, financial institutions operating under a commercial framework tend to
lend to the richer segment of the poor. Thus, we expect that men would be less exposed to
credit rationing than women, who are poorer and more vulnerable. The results obtained
from the hinterland of Kinshasa show that the impact of the variable household head male
appears to be negative, but not statistically significant. The farm size is expected to be
positively associated with credit constraints conditions, given the lack of financial
development markets to respond to the magnitude of investment due to the size of farm.
Our findings report a positive effect of farm size as expected, but it is not statistically
significant. This means that an increase in household size by unit will not reduce
substantially the probability of being credit-constrained.
- 48 -
The variables related to the place of residence, such as the Mbankana group, the Dumi
group and the Menkao group, were introduced to test the village effect and control for the
effects of other characteristics related to local institutions, such as markets, schools, roads,
etc. The results indicate insignificant effects of villages on the probability of being credit
constraints. Thus, there is no strong evidence to support the hypothesis that the lack of good
local institutional environment increases the probability that a farm household is credit-
constrained.
2.5 Conclusions
This chapter explores the major constraints related to the financing of agricultural activities,
and identifies the key drivers of the probability of a farm household being credit-
constrained. The results of household responses reveal that the high risks and the low
returns of agricultural activities, the weak portfolio of microfinance operating in the area, as
well as the lending methodology (collateral, high interest rates, non-compatible loan terms
and procedures) applied in microfinance are the major constraints to financing agriculture by
MFIs. The findings imply public agricultural policy aimed at increasing profitability and
securing agricultural activities will reduce credit constraints by increasing the willingness of
MFIs to transact with farm-households. The finding calls for the adaptation of the lending
technology to the specificities of agriculture and farm household cash flows. A public policy
of increasing financial resources and assets of MFIs operating in rural areas will reduce credit
constraints among farm-households.
The results from logit and probit models show that household size, association participation,
remittance and landholding property are the keys factors affecting farm household credit
constraints conditions in the hinterland of Kinshasa. Those factors, excepting household size,
reduce the probability of being credit-constrained. The findings call for land policy reform to
improve farm household access to arable land and to secure its transactions. This will
provide lenders with formal collateral and enforcement of financial contracts, and therefore
increase lenders’ willingness to transact with farm-households. The findings imply that
family planning policy aimed at addressing the issue of demographic growth by reducing the
size of households will result in a decrease of credit constraints among farmers by reducing
- 49 -
liquidity constraints. The findings also suggest that there is a need for the promotion of
social organizations and local institutions and the improvement of the existing
infrastructure. Given the negative and significance effects of remittance, there is need to
promote household’s income resource diversification to increase the willingness of MFIs to
transact with farmers and cope with systematic risks related to the financing of agricultural
activities.
App. Table 2.1 Statistic description of variables used in the regression
Variable Mean SD Min Max
Education 9.430 3.859 0 18
Household head male 0.910 0.285 0 1
Household head age 45.049 9.808 22 66
Dependency ratio 0.506 0.240 0 0.9
Household size 4.950 1.849 1 11
Land property holding (2007) 0.554 0.498 0 1
Farm size 13.341 11.456 2.5 100
Credit constraints conditions 0.712 0.453 0 1
Remittance 0.420 0.494 0 1
Association (2007) 0.698 0.698 0 1
Mbankana group 0.342 0.475 0 1
Dumi group 0.322 0.468 0 1
Menkao village 0.336 0.473 0 1
App. Table 2.2 Multicollinearity diagnostic
Variables VIF Tolerance

Education 1.20 0.8319
Household head male 1.23 0.8121
Household head age 1.17 0.8582
- 50 -
Dependency ratio 1.10 0.9086
Household size 1.11 0.9002
Land property holding in 2007 1.05 0.9480
Farm size 1.03 0.9677
Remittance 1.09 0.9160
Association membership in 2007 1.12 0.8935
Mbankana group 1.37 0.7276
Dumi group 1.47 0.6822
Mean VIF 1.18 -
- 51 -
Chapter 3
Credit Constraints and Efficiency of Cassava Farmers in the Hinterland of Kinshasa
3.1 Introduction
Agriculture is an important component in the economy of the DRC and cassava is the most
important crop cultivated. This is so much so that, at 15 million tons, the DRC is ranked the
second highest cassava producer in Africa (FAO, 2007). Since the middle of the 1980s, annual
cassava per capita production and the cassava yield per hectare have fallen significantly
(Enete, 2003). In dominant discourse from the DRC, the lack of financial resources has been
recognized as one of the key factors explaining the agricultural production slowdown.
However, academic studies from the DRC on this topic are poor. In the meantime, the
literature on credit constraints from developing nations has been amply documented, and
has reported a negative impact of credit constraints on the performances of agriculture. The
credit constraints have been found to negatively affect agricultural output (Feder et al. 1990;
Petrick, 2003; Freeman et al. 1998), farm profits (Foltz, 2004; Blancard et al. 2006),
agricultural productivity (Guirkinger and Boucher, 2008; Feder et al. 1990) and farm
investments (Carter and Olinto, 2003; Kohansal et al. 2008), risk aversion, technology choice
and adoption (Dercon, 1996; Eswaran and Kotwal, 1986), resource allocation (Jabbar et al.
2002; Croppenstedt et al. 2003) and TE (Färe et al. 1990; Omonona et al. 2008; Komicha and
Öhlmer, 2006). However, the effects of credit constraints on farm household scale efficiency
are not well documented. Therefore, using evidence from cassava farmers of the hinterland
of Kinshasa, This study evaluates the effects of credit constraints on farm household scale
efficiency using evidence from the cassava farmers of the hinterland of Kinshasa. In addition,
this study fits into the body of the literature seeking to evaluate the impact of credit
constraints on technical efficiency. This study also controls for household and institutional
characteristics affecting differences in technical efficiency scores among cassava producers.
This chapter is organized as follows: introduction, review of literature, data and
methodology, results and discussions, and conclusions.
- 52 -
3.2 Review of literature
The relationship between financial development, economic growth and economic

development has been the subject of a large number of previous theoretical and empirical
studies (Benhabib, 2000; Levine, 1997; Romer, 1986; Shaw, 1973; McKinnon, 1973). These
studies have asserted credit to be fundamental in production and investment decisions. In
the same line, many previous empirical studies have demonstrated the link between access
to credit and poor growth, income and household welfare (Morduch, 1998; Khandker, 2005;
Pitt and Khandker, 1998; Coleman, 1999). The free cash-flow approach assumes a positive
effect of credit on production efficiency. Following this approach, farmers with credit access
have to face repayment obligations, leading to increased efforts and limited waste of
materials in the production process. This in turn affects the production efficiency positively.
The agency theory assumes that a farmer’s access to credit might have a negative impact on
production efficiency. The defenders of this approach argue that information and monitoring
costs related to credit affect the efficiency of farmers with access to credit. The credit
evaluation approach asserts a reverse causality between credit and production efficiency.
This approach is based on the idea that production efficiency positively affects the level of
credit insofar as financial institutions are expected to lend to the most efficient farmers.
Credit constraints conditions are found to affect agricultural efficiency through different
channels. Credit constraint conditions, as a result of market imperfections, prevent
opportunities and constrain production and optimal investment (Freeman et al. 1998).
Credit constraints might limit a farmer’s ability to use modern products such as fertilizers,
herbicides and pesticides, thereby affecting production efficiency (Croppenstedt et al. 2003;
Latruffe, 2004). Credit constraints might also affect farmer risk aversion, technology choice
and adoption. This because constrained farmers are more likely to invest in less risky and
less productive technology rather than in more risky and more productive technologies
(Dercon, 1996; Guirkinger and Boucher, 2008; Eswaran and Kotwal, 1986).
The imperfections of rural credit markets resulting in high prevalence of credit constraints
among farmers are considered to lead to heterogeneous resource allocation and different
outputs among farmers (Jabbar et al. 2002; Freeman et al. 1998; Croppenstedt et al. 2003).
The constrained farmers are assumed to be more likely to misallocate and under-invest their
- 53 -
resources than unconstrained farmers. In line with this, Binswanger and Deininger (1997)
argue that an unequal distribution of initial endowments in the context of financial market
imperfections may make the majority of farmers far from productive investments.
Feder et al. (1990) found a significant difference in productivity between credit-constrained

farmers and unconstrained farmers in China. Kohansal et al. (2008) reported the importance
of credit in increasing a famer’s propensity to invest in agricultural activities. Analyzing the
resource productivity, allocative efficiency and determinants of technical efficiency of rice
farmers in Nigeria, Ogundari (2008) showed that access to credit was one of the significant
determinants of technical efficiency among the farmers. Freeman et al. (1998) found that
the marginal contribution of credit to milk productivity was different among credit-
constrained and non-constrained farmers in Africa. Studying the effects of credit constraints
in Peruvian agriculture, Guirkinger and Boucher (2008) reported that the productivity of
credit-constrained households was a function of their endowments of productive assets and
loans obtained. Many other studies have reported the negative effect of credit constraints
on agricultural production efficiency (Ali and Flinn, 1989; Dorfman and Koop, 2005; Bravo-
Ureta and Pinheiro, 1993).
The data were collected from a household survey carried out in the hinterland of Kinshasa.
The survey questionnaire contains the necessary information needed to identify credit-
constrained households, the household and institutional characteristics as well as inputs and
outputs in the cassava production. The details on the data collection procedure and
sampling are presented in Chapter 1. The descriptive statistics of variables used in this
chapter are reported in App. Table 3.1.
We used three analytical approaches for the empirical analysis. Firstly, we ran technical and
scale efficiency scores of cassava farmers using DEA. Following Coëlli (1996), we estimated
the TE in cassava production by solving the following DEA linear program under a Constant
Return to Scale (CRS), that is:
- 54 -
TE j  Min CRS , CRS
j (3.1)
j
Subject to
Pj  P; CRS F j  F;   0
j
.
 
Where j th is a farmer drawn from n  farmers of the sample; P  is cassava produced (kg);
 CRS
j  is TE  j  of the farmer under CRS and   is a vector of weights nx1 and F  is a
set of input variables used for cassava production, including planting material (kg), labor
(man-days) and farm size (hectare). Modern inputs such as manures, pesticides, fertilizers
are not used in cassava production in the hinterland of Kinshasa. An extremely limited
number of cassava farmers use tractors; therefore, any farmer using a tractor was not
randomly drawn in the sample. In addition, small agricultural tools such as the hoe and the
spade are used in cassava production and other tools are depreciated in full (did not have an
accountable value), and therefore were not considered as inputs for the empirical
estimation of DEA efficiency scores.
The DEA CRS specification is based on the assumption that all DMUs are operating at the
frontier of efficiency. However, given agricultural input market imperfections, environmental
constraints and several other constraints faced by the cassava producers from the hinterland
of Kinshasa, this assumption appears, in all probability, unrealistic. Thus the Variable Returns
to Scale (VRS) assumption needs to be specified and is more appropriate and better fits the
data than the CRS hypothesis. However, we estimate and discuss the efficiency scores
obtained from both assumptions. The TE  VRS

j  
of cassava producers under VRS is run by
adding the convexity constraint  j  1 to the CRS DEA linear program (Coëlli, 1996).
The efficiency scores under VRS may be equaled or be greater than those obtained under
the CRS assumption. Moreover, the Scale Efficiency (SE) can be obtained by calculating the
ratio  CRS
j /  VRS
j . The SE can be either Increasing Returns to Scale (IRS) or Decreasing Returns
to Scale (DRS). 
In general,  VRS
j 
,  CRS
j  and (SE) range from 0 to 1. Thus,
0   j  1 and 0  SE  1 . The farmer is technically efficient if  j  1 and reaches the SE
if SE  1 .
- 55 -
The production inputs selected to estimate the efficiency score of cassava production are
land, labor and plant materials. Land is the total land area used for cassava production,
including land owned, rented and obtained through gift, and is measured in hectares (ha).
Labor is composed of the family labor force and external labor supply, measured in man-
days. Plant materials include those obtained from self-production, bought or received from
other external sources, and are measured in kilograms. Moreover, the efficiency scores are
computed under output-orientation. This choice is explained by the fact that the cassava
producers from the hinterland of Kinshasa do not have control over productive forces
because of market imperfections and the dependence of agriculture on the environment
and natural conditions. The efficiency scores are computed using the DEAP software
developed by Coëlli (1996).
Secondly, we explored key drivers of TE differences among cassava producers by estimating

a TR model. The cassava production efficiency is used as a dependent variable, while
household and institutional characteristics are used as explanatory variables. The choice of a
TR model is dictated by the nature of the TE measure, which is truncated at 1, and by the
findings of the academic literature (Battese and Coëlli, 1995; Simar and Wilson, 2007). Thus,
we computed truncated model to estimate the influence of the above covariates on the TE
score. That is,
N
Ei     i X j  i (3.2)
k 1
Where Ei is the value of the TE index obtained from DEA, X is a set of potential
determinants of technical efficiency differentials,  is a set of unknown parameters to be

estimated that captures the effect of covariates on technical efficiency, and  is an error
term that independently follows normal distribution with (0,  2 ) .
The selected covariates used for estimation are household head age, household head years
of schooling, household size, gender of household head, landholding property, farm size,
household participation in social and economic organizations, and villages located in the
area of Mbankana and villages located close to Dumi. The household head age is used as a
proxy of experience and is expected to increase production efficiency. Household head
education is measured by the number of years of schooling, and this is also expected to
- 56 -
increase TE by affecting managerial skills, which would result in efficiency variations among
farmers. Nevertheless, several empirical studies have reported mixed effects (Bravo-Ureta
and Pinheiro, 1993; Battese et al. 1996; Ali and Flinn, 1989). Family labor is an important
source of labor supply in the hinterland of Kinshasa, as in many developing nations. In the
context of labor markets imperfections and credit constraints, as with the case of study area,
farm-households with inadequate family labor will face farm labor limits. Thus, we expected
that the size of household will be positively associated with TE; and household dependency
ratio will be negatively correlated with production efficiency.
The relationship between farm size and productivity has been much debated among
researchers in developing nations. The studies carried out on this topic showed mixed
results. A significant number of studies have reported a positive relationship between farm
size and productivity (Bhandari, 2006; Deolalikar, 1981; Fan and Chan-Kang, 2005). Several
other studies found an inverse relationship, explained by the relative advantage of using
more family labor by small farms, which reduces the monitoring and supervision costs of
hired labor (Sen, 1962; Berry and Cline, 1979; Barrett, 1996). Thus, farm size is expected to
be positively or negatively associated with TE in cassava production. It is expected that
landholding property holdings will have a positive effect on TE distribution as these may be a
source of motivation for producers. It is also expected that network participation will be
positively associated with TE scores. The villages of residence are expected to be either
positively or negatively correlated with TE distributions.
Thirdly, we calculated the effect of credit constraints on cassava production efficiency by

applying a matching approach. By applying this approach, we assumed that credit
constraints could be similar to a treatment, and then causal effects may be estimated by
running the Average Treatment effect on Treated (ATT). Following Becker and Ichino (2002),
one may estimate the causal effect of credit constraints conditions on production efficiency
by resolving the following equations:
ATT  EE1i  E0i / Di  1 (3.3)
= E Ei1 / Di  1  EEi 0 / Di  1
For sample, ATT is given by:
- 57 -
1 n
ATT   ( E1i  E0i / Di  1)
n i 1
(3.4)
1 n
  ( E1i / Di  1)  ( E0i / Di  1)
n i1
Where Ei1 / Di  1 is the outcome variable (the production efficiency scores) applicable to
the constrained farm-households, and Ei 0 / Di  1 is the outcome of what would have
happened to the same farm-households if they had not been constrained. One could
observe that it is not possible to compute ATT using the equations (3.3) or (3.4) because
Ei 0 / Di  1 is not observed. To solve this problem of missing data, the matching approach
as proposed by Rosenbaum and Rubin (1983) was used: this imposes the CIA, implying that
given observable control variables, the assignment to the treatment (here credit-
constrained) is random and independent of the outcome (production efficiency). That is,
Ei 0 , Ei1   D / X i (3.5)
Where X is a set of control variables or conditional variables, and where  denotes

independence. This condition is also known as the “ignorable treatment assignment”
(Rosenbaum and Rubin, 1983) or the “selection on observables” (Heckman and Robb, 1985).
The idea underlying CIA is that if enough information on X is known, it is possible to
eliminate the correlation between Ei 0 , Ei1  and Di by conditioning on X . Thus, the ATT can
then be computed as averaging the conditional effect over the conditioning variables. That
is,
ATT  E X E( Ei1 / X i , Di  1)  E( Ei 0 / X i , Di  0/ Di  1 (3.6)
One way to compute equation (3.6) is to match constrained farm-households and

unconstrained farm-households on their observable characteristics X . However, in practice,
doing so is not possible as the number of observable characteristics is huge. Given the huge
number of household characteristics X i , ATT could be difficult to estimate. To address this
problem, and following previous studies, we applied the PSM. PMS avoids the dimensional
problem which results from the huge number of observable characteristics (Becker and
Ichino, 2002). In this study, the propensity score is defined as an index function showing the
probability of a given household to face credit constraints, given by:
- 58 -
P( X )  Pr(D  1 / X ) (3.7)
PSM resorts to the assumption that if interest variable Ei is independent of the treatment D ,
given a set of conditional variables X, it is also independent of the treatment D , given the
propensity score of treatment P(X). That is,
Ei  D / X  Ei  D / P( X ) (3.8)
By applying PSM, we match the outcomes of the treated group and the control group upon
the basis of a single index P(X) instead of all variables of a set X. Therefore, ATT may be
computed using the following equation:
ATT  EP ( X ) E ( Ei1 / P( X ), Di  1)  E ( Ei 0 / P( X ), Di  0/ Di  1(3.9)
As the propensity score is a continuous variable, it becomes hard to expect a lot of

observations within the same region. Therefore, exact matching is not possible. To avoid this
problem, Becker and Ichino (2002) proposed four matching methods, including kernel
matching, radius matching, nearest neighbors matching and stratification matching. In this
study, we apply nearest neighbors matching with replacement, that is:
1   I  
ATT 
NT
 E T
i   w E ij
C
j   T  Ei    wij E j 
T C
(3.10)
iT  jC i  N  iT iT jC i  
1 1

NT
E
iT
i
C

NT
w E
j
C
j
Where: T = vector of treated individuals; C = vector of control group; EiT = observable
outcome welfare indicator of constrained households matched group; EiC = observable
outcome TE of control group matched; N C and N T are respectively the observations in the
control group and that of the treated group and C (i)  min j Pi  Pj (3.11). For empirical
estimation, we used the pmatch2 software, developed by Leuven and Sianesi (2003) and set
up as a Stata software application, for the empirical estimation of ATT. A simulation-based
sensitivity analysis for matching estimators was done using the Stata program Sensatt
proposed by Nannicini, which allows running the sensitivity analysis of matching estimators
as proposed by Ichino et al. (2006).
- 59 -
3.4.1 Efficiency scores distributions
Table 3.1 presents the computed technical and SE scores of the cassava farmers. The results
reveal that TE scores in cassava production are very low and many farmers are working in
the inefficiency zone. Indeed, under the CRS assumption, the mean TE score of the farmers
from the sample is estimated at 0.272 (SD=0.186). Moreover, 45% of farmers have TE scores
ranging from 0.20 to 0.40 and about 36% of households have the technical efficiency score in
the interval of 0.41 and 0.60. Only 3% of households from the sample might be viewed as
technically efficient.
- 60 -
Table 3.1 Distribution of technical and scale efficiency in cassava production
Technical efficiency Technical efficiency Scale efficiency

(CRS) (VRS) ( SE)
Interval Frequency % Frequency % Frequency %
0.00-0.20 92 45.5 81 40.1 0 0
0.21-0.40 73 36.1 70 34.7 3 1.5
0.41-0.60 24 11.1 27 13.4 4 2
0.61-0.80 7 3.5 10 5 27 13.4

0.81-1.00 6 3 14 6.9 168 83.2
Average 0.272 0.318 0.902
Maximum 1 1 1
Minimum 0.03 0.032 0.254
S.D 0.185 0.231 0.13
A similar feature is observed for VRS TE scores. The results from the VRS assumption show
that the average TE score is estimated at 0.318. About 40% of the farmers from the sample
are in the zone of TE with scores ranging from 0.00 to 0.20; 37% of farmers range from 0.21
to 0.40. Only 7% of the sampled farm-households are technically efficient. The findings
report that the majority of farmers (92% under CRS and 88% under VRS) have TE scores less
than or equal to 0.60. The estimated SE shows that 83% of cassava producers are close to
the scale efficiency line. The average scale efficiency score is very high and estimated at
0.90.
3.4.2 Factors affecting cassava production efficiency differentials
Table 3.2 presents the results of TR applied to identify key factors affecting TE scores
differences among farm-households.
- 61 -
Table 3.2 Determinants of cassava production efficiency differentials
CRS model VRS model
Variables Coefficients Z-statistic Coefficients Z-statistic
Education 0.0060** 1.94 0.0094*** 2.55
Household head male -0.0348 -0.83 -0.0307 -0.62
Household head age 0.0004 0.37 0.0007 0.54
Household size -0.0096 -1.57 -0.0056 -0.78
Land property holding 0.0673*** 3.01 0.0632*** 2.40
Farm size 0.0098 1.03 0.0192* 1.71
Association 0.0698*** 2.84 0.0645*** 2.25
Mbankana group 0.0095 0.36 0.0227 0.73
Dumi group -0.0032 -0.12 -0.0104 -0.32
Constant 0.1621*** 2.19 0.1126 1.28
Prob.>  = 0.004 Prob.>  = 0.009

2 2
Log likelihood = 90.80 Log likelihood = 59.7

Wald  = 23.70 Wald  = 21.86
2 2
Key: * p<.1; ** p<.05; ***p<.01
The likelihood ratio tests indicate that all variables in both specifications models (VRS and
CRS), taken together, have a statistically and significant effect on TE scores. The results of
the VRS model reveal four covariates likely to be significantly and positively associated with
TE scores. Landholding property, the household head’s number of years of formal schooling
and association are significant at 1%, while farm size is significant at 10%. Under CRS
assumption, farm size is not statistically significant. The effects of household head age,
household head gender and villages of residence are not found to be statistically significant
- 62 -
associated with TE scores. In other words, this means that TE is distributed independently of
them.
3.4.3 Credit constraints conditions and cassava production efficiency
Results of descriptive analysis
Table 3.3 reports the distributions of technical and SE scores of both groups of farm-
households (constrained and unconstrained). For a better display of differences in efficiency
score distributions between constrained and unconstrained households, we provide figures
on the distributions of average technical and SE scores (Figure 3.1), the distribution of TE
scores under CRS (Figure 3.2), the distribution of technical efficiency scores under the VRS
assumption (Figure 3.3) and the distribution of scale efficiency scores (Figure 3.4).
The findings reveal that although TE scores are very low for both constrained and
unconstrained households, there are substantial gaps between the two groups as the scores
for unconstrained households are twice as high as constrained households under both CRS
and VRS assumptions. Under the CRS assumption, the average TE score is 0.202 for
constrained households and 0.445 for unconstrained. The gap between unconstrained and
constrained farm-households is 0.243 on average. About 95% of constrained households
have TE scores under 0.410, against 47% for unconstrained. Furthermore, among
constrained households no one is close to the frontier of the TE line, while 10% of
constrained households have TE scores close to the frontier of best practice.
Similarly, TE scores under VRS are slightly improved for both groups, but the gap between
the two groups remains substantial. The average TE score for constrained households is
estimated at 0.246 and that of unconstrained is estimated at about 0.497. The estimated
VRS TE score show that 86% of constrained households do not reach the score of 0.41,
against 44% for unconstrained. Only 3.5% of constrained households are technically efficient
against 15.5% for unconstrained households.
- 63 -
Table 3.3 Distribution of cassava production efficiency scores
Sample Constrained Unconstrained

Efficiency Scores
CRS Frequency % Frequency % Frequency %
0.00-0.20 92 45.5 87 60.4 5 8.6
0.21-0.40 73 36.1 50 34.7 23 39.7
0.41-0.60 24 11.1 4 2.8 20 34.5
0.61-0.80 7 3.5 3 2.1 4 6.9
0.81-1.00 6 3 0 0 6 10.3
Value Value Value
Average 0.272 0.202 0.445
Maximum 1 0.796 1
Minimum 0.03 0.03 0.063
S,D 0.185 0.113 0.212
VRS Frequency % Frequency % Frequency %
0.00-0.20 81 40.1 78 54.2 3 5.2
0.21-0.40 70 34.7 47 32.6 23 39.7
0.41-0.60 27 13.4 12 8.3 15 25.9
0.61-0.80 10 5 2 1.4 8 13.8
0.81-1.00 14 6.9 5 3.5 9 15.5
Value Value Value
Average 0.318 0.246 0.497
Maximum 1 1 1
Minimum 0.032 0.031 0.063
S.D 0.231 0.181 0.244
SE Frequency % Frequency % Frequency %
0.00-0.20 0 0 0 0 0 0
0.21-0.40 3 1.5 3 2.1 0 0
0.41-0.60 4 2 3 2.1 1 1.7
0.61-0.80 27 13.4 17 11.8 10 17.2
0.81-1.00 168 83.2 121 84 47 81
Value Value Value
Average 0.902 0.896 0.915
Maximum 1 1 1
Minimum 0.254 0.254 0.516
S.D 0.130 0.135 0.114
The results reveal a high level of SE scores for both groups and no substantial differences
between the two groups of households. The average SE is 0.896 for constrained households
and 0.915 for unconstrained households. About 84% of households have SE scores ranging
from 0.81 to 1.00, against 81% for constrained households. There are no unconstrained
- 64 -
households working in the region of SE scores ranging from 0.00 to 0.40, while 4% of the
constrained households are in this zone.
Figure 3.1 Distribution of average efficiency score
Figure 3.2 CRS technical efficiency: constrained versus unconstrained farm-households
- 65 -
Figure 3.3 VRS technical efficiency: constrained versus unconstrained farm-households
Figure 3.4 Distribution of scale efficiency : constrained versus unconstrained households
Results of PSM
Household credit constraints conditions are considered as a treatment variable, which is a

binary choice taking value 1 if a given household faces credit constraints and 0 otherwise. It
is used as a dependent variable to compute the propensity score using the logit model. The
conditional variables used for the estimation of the propensity score of being credit-
constrained are household size, dependency ratio, the age of household head, the
household head’s formal education, the gender of the household head, property
landholding, farm size, associations and villages of residence defined by Mbankana, Dumi
and Menkao groupings. The villages of residence are used as proxy to control the
- 66 -
institutional characteristics, such as the market, local institutions, etc. The outcomes
variables are TE scores under both CRS and VRS assumptions and SE. The results of logit
model are presented and discussed in Chapter 2. The distributions of the estimated
propensity score are reported in Figures 3.5 and 3.6.
The estimation of matching parameters requires that the balancing properties might be
satisfied. The balancing properties are tested to ensure that there are no substantial
significant differences in the observable characteristics or covariates between the matched
treated and control groups. In others words, the balancing properties assumption allows
testing whether or not treated and control households matched differ substantially in
observable characteristics. The results (Table 3.4) reveal that the propensity score is well
performed and balancing conditions are satisfied, as there is no significant difference in
means for most of covariates, indicating that the statistical differences between the mean
values of covariates are lower and not significant for the matched sample than unmatched
sample. Moreover, the distributions of propensity score (figures 3.5 and 3.6) show that there
are sufficient overlaps.
After computing the propensity score and checking if the balancing conditions are satisfied,
the average effect on treated (ATT) is calculated using nearest neighbor matching with
replacement to ensure that each treatment observation is matched to the nearest control
observation. Table 3.6 reports the results of estimated ATT or the average impact of credit
constraints on cassava production efficiency.
To address the robustness of estimated ATT, we tested the CIA. This assumption (CIA) is
intrinsically not directly testable, but Ichino et al. (2006) proposed a method to address the
issue of robustness of ATT to the possible failure of CIA. Thus, following Ichino et al. (2006),
we applied a method with confounder calibrated to mimic observable binary covariates and
the neutral confounder to test the validity of CIA. Table 3.8 reports the results of simulation-
based sensitivity analysis. The results show that under the deviation from CIA, with
characteristics behaving like neutral confounder, association, landholding property and
remittances, the ATT obtained using both CRS and VRS outcomes do not differ substantially
from the baseline ATT. This is the evidence that our matching estimators computed to
- 67 -
evaluate the causal effect of credit constraints on TE scores in the context of hinterland of
Kinshasa are robust. Thus, we accept the hypothesis that credit constraints reduce TE scores.
The baseline estimate for SE scores shows no significant ATT equal to -0.012 points. The
results of simulation-based-sensitivity analysis reveal that, with the confounder behaving like
associations, landholding property, remittances and neutral confounder, the baseline ATT
differs substantially from the simulated ATT. For example, in the simulation with the
confounder behaving like associations, the ATT is brought down to 0.004. This is evidence
that our findings are not strong to support the possible impact of credit constraints on
cassava production SE scores in the hinterland of Kinshasa.
Table 3.4 Balancing properties
Mean % reduction T-test
Variables Sample Treated Control % bias bias T p>t
Education Unmatched 9.3681 9.5862 -5.6 -0.36 0.717
Matched 9.3478 9.7313 -9.9 -75.8 -0.79 0.433
Household head male Unmatched 0.9236 0.8793 14.8 1.00 0.320
Matched 0.9130 0.9449 -10.7 28.0 -0.94 0.349
Household head age Unmatched 44.993 45.190 -1.9 -0,13 0.898
Matched 45.322 46.490 -11.5 -494.5 -0.89 0.372
Dependency ratio Unmatched 0.5039 0.5118 -3.2 -0.21 0.833
Matched 0.5069 0.5083 -0.6 82,4 -0.05 0.964
Household size Unmatched 5.2014 4.3276 46.8 3.10 0.002
Matched 4.8435 5.0545 -11.3 75,9 -0.84 0.399
Land property holding Unmatched 0.4791 0.7413 -55.5 -3.48 0.001
Matched 0.5739 0.5704 0.7 98,7 0.05 0.958
Farm size Unmatched 13.837 12.112 16.7 0.97 0.334
Matched 13.152 13.781 -6.1 63,5 -0.45 0.651
Remittance Unmatched 0.3611 0.5689 -42.3 -2.74 0.007
- 68 -
Matched 0.4173 0.5255 -22.0 48,0 -1.64 0.101
Associations Unmatched 0.6736 0.7586 -18.8 -1.9 0.053
Matched 0.7478 0.7252 5.0 73,4 0.39 0.699
Mbankana group Unmatched 0.3333 0.3620 -6.0 -0.39 0.699
Matched 0.3739 0.3924 -3.9 35,4 -0.29 0.774
Dumi group Unmatched 0.3611 0.2241 30.3 1.89 0.060
Matched 0.2956 0.2869 1.9 93,7 0.14 0.885
Before matching After matching

2
Pseudo R 2 0.143 Pseudo R 0.016
LR  2 (p-value) 34.75(0.000)***
LR  2 (p-value) 4.95(0.933)
Figure 3.5 Distribution of propensity of being credit: constrained (histogram)
Constrained Unconstrained
4
3
Density
2
1
0
0 .5 1 0 .5 1
- 69 -
Figure 3.6 Distribution of propensity score of being credit: constrained (boxplot)
Constrained Unconstrained
1
.8
.6
.4
.2
Table 3.5 Impact of credit constraints on cassava production efficiency
Outcome Group Average for Average for ATT T-test

treated controls
TE CRS Unmatched 0.202 0.445 -0.242 -10.52
Matched 0.195 0.446 -0.251 -6.71
TE VRS Unmatched 0.245 0.496 -0.251 -8.04
Matched 0.242 0.502 -0.260 -5.79
SE Unmatched 0.896 0.914 -0.018 -0.92
Matched 0.891 0.904 -0.012 -0.54
Legend: * p<.1; ** p<.05; ***p<.01
- 70 -
Table 3.6 Simulation-based sensitivity analysis
Outcome Selection ATT

effect effect
TE (CRS)
Neutral confounder 3.610 1.051 -0.208
With confounder calibrated to land property 7.525 0.328 -0.268
With confounder calibrated to association 71.832 0.756 -0.218
With confounder calibrated to remittances 8.733 0.446 -0.258
TE ( VRS)
With confounder calibrated to associations 61.769 0.669 -0.215
SE
Neutral confounder 1.458 1.108 0.004
With confounder calibrated to land property 6.612 0.308 0.014
With confounder calibrated to associations 2.255 0.744 0.004
3.4.4 Discussion of the results
The results of DEA-estimated technical and SE scores suggest some relevant and important
findings. A few farm-households have reached the frontier of best practice and therefore
can be viewed as efficient. This means that these efficient farmers apply their resources
relatively better than the other share of farmers from the sample with similar inputs. These
farmers viewed as efficient have a higher ability to use resources in producing cassava than
their counterparts, given the available technology represented by the best practice frontier.
The mean values of the TE score (VRS and CRS) in the sample are very low and the majority
- 71 -
of farmers are operating in the region of inefficiency, and thus may be viewed as not using
their resources in an optimal way compared to the efficient farmers from the sample. In
other words, the results reveal a high level of technical inefficiency among farm-households
from the sample, which needs to be reduced in order to produce at the best possible TE
level. Given the high level of inefficiency in cassava production, there are significant
possibilities for improving efficiency in the hinterland of Kinshasa. The average TE scores
(27% under CRS and 32% under VRS) indicates that farm-households from the sample would
substantially improve their productions with the same inputs. On average, TE scores may be
increased by 73% under CRS and by 68% under VRS assumptions. Although TE is a relative
concept, ours findings are consistent with empirical studies done in Nigeria, which reported
high levels of average technical efficiency scores and small variance between cassava
farmers (Edeh and Awoke, 2009; Ogboma et al. 2007; Adeyemo et al. 2010; Iheke, 2008; Onu
and Edon, 2009; Ogundari and Ojo, 2007).
By comparing technical and SE scores, it is clear that the scale efficiency scores are higher
than the TE scores under both CRS and VRS assumptions. Therefore, the inefficiency
observed from the sample is more likely to be associated with inappropriate farming
systems and lack of farming best practices rather than the farm size. The mean value of the
SE score is greater than that of the TE score, suggesting that technical inefficiency scores
make a greater impact on cassava production inefficiency rather than scale inefficiency. This
means that the inefficiency in cassava production results more from the use of inadequate
farm management practices than the farm size. This finding implies that the improvement of
farm management practices of inefficient farmers will increase the cassava production
efficiency and will allow a high level of output with the same inputs. From an agricultural
policy viewpoint, the diffusion of optimal farm management practices in cassava production
through the extension services will assist in improving the efficiency scores of farm-
households. Moreover, this finding calls for research aiming at understanding the
management practices of efficient farm-households to assist the formulation of agricultural
policy and extension packages.
Since TE is a relative concept and the wide variations in technical efficiency scores are
unconditional, many more lessons may be learned by conditioning them on a set of relevant
household and institutional characteristics. The results of TR models show that landholding
- 72 -
property, associations, formal education of household head and farm size are the keys
drivers of TE score differentials between farm-households. Thus, the hypothesis that
household capital endowments positively affect the technical efficiency of farm-households
can be confirmed.
The positive coefficient of landholding property means that households with landholding
property are more technically efficient than their counterparts without landholding
property. It should be noted that in the hinterland of Kinshasa, farmland conflicts are greater
than in the rest of the country. The attribution of land property rights involve both
traditional authorities and the State, actors who are often in permanent conflict. In many
cases, these conflicts are resolved to the detriment of small landholders. This in turn results
in an increased rural population without land, who are then obliged to grow their crops on
the same plot for consecutive larger number of seasons. Land conflicts are more likely to
limit farmers’ efforts to maximize outcome and to affect their risk aversion, leading to low
efficiency. These results are in line with previous studies in sub-Saharan African that have
suggested land tenure reform to promote agricultural intensification and productivity
growth (Feder and Nishio, 1999; Besley, 1995a; Gavian and Fafchamps, 1996; Jacoby and
Minten, 2007). The result calls for government intervention in land administration to
overcome land tenure insecurity and increase household access to arable land.
Household head education appears to be among the key drivers positively affecting TE
scores obtained both under CRS and VRS assumptions. This indicates that TE increases with
the formal schooling of the farm household head. Farm-households led by more educated
heads are more technically efficient than those led by the less educated. Several empirical
studies have come to a similar conclusion (Bravo-Ureta and Pinheiro, 1993; Ali and Flinn,
1989; Battese et al. 1996). The positive sign associated with household head education could
possibly be due to the fact that education is more likely to improve the quality of decision-
making. In addition, education is an important determinant in adopting good farming
practices. This finding implies that public policy aiming at investing in education and
information will reduce cassava production inefficiency.
Farm size tends to be significantly associated with TE of cassava farmers in the hinterland of
Kinshasa. The results suggest that there is a positive effect of farm size on TE. This means
- 73 -
that the efficiency increases with farm size and that farm size has an increasing effect on TE.
The results suggest that the public policy aiming at increasing the farm size will reduce
cassava production inefficiency. The positive sign related to farm size is consistent with other
empirical studies which have reported a positive relationship between farm size and
efficiency (Bhandari, 2006; Deolalikar, 1981; Fan and Chan-Kang, 2005).
Household participation in associations is found to be an important determinant of TE. The

results suggest that households participating in associations are more likely to be efficient
compared to their non-participating counterparts. This could be due to the fact that
associations play an important role in the adoption and the sharing of farm practices.
Participation in associations generates information on new technologies and reduces input
constraints such as labor and planting materials. Moreover, participation in farm
organizations facilitates the cooperation and the sharing of tactics and information. This
finding calls for public policy and nonprofit organizations operating in the hinterland of
Kinshasa to support farm associations and work for the increased participation of farm-
households in social organizations such as cooperatives. The variables related to village areas
are used to control for local institutional and community characteristics. The results from TR
show these variables are not statistically significant. Thus, there is no strong evidence to
support the hypothesis that good local institutional environment positively affects the
technical efficiency of farm-households.
Credit constraints conditions seem to be among the key factors affecting TE scores. The
results of simple descriptive analysis indicate a wide gap in TE scores between constrained
and unconstrained households. The unconstrained households have higher TE scores than
constrained households. Moreover, the ATT nearest neighbor estimator is negative and
significant at 1%, indicating that credit constraints have a negative and significant effect on
TE scores under both CRS and VRS assumptions. Credit constraints tend to reduce average
TE in cassava production by 0.251 under CRS and by 0.260% under VRS, revealing
considerable inefficiency due to credit constraints. This supports our hypothesis that credit
constraints reduce the TE of farm-households. However, the findings reveal that credit
constraints have negative and not statistically significant effect on the SE of farm-
households. This suggests that credit constraints are not important in explaining the
- 74 -
differences in SE distributions. Thus, the hypothesis that credit constraints reduce the scale
efficiency of farm-households can be rejected.
Given the huge proportion of credit-constrained households in the hinterland of Kinshasa

and the negative effect of credit constraints on TE, removing credit constraints will
substantially reduce inefficiency in cassava production. This finding calls for agricultural
credit policy and measures to address credit constraints problem and satisfy adequately the
magnitude of farm household credit demands. This in turn will increase production through
by increasing TE.
The negative and significant effect of credit-constrained conditions found in this study is
consistent with several empirical studies conducted in developing nations (Ali and Flinn,
1989; Guirkinger and Boucher, 2008; Feder et al. 1990; 1989; Kohansal et al. 2008; Ogundari,
2008; Dorfman and Koop, 2005; Bravo-Ureta and Pinheiro, 1993). For example, Ali and Flinn
(1989) found a positive and significant effect of access to credit on TE among basmati rice
producers in Pakistan. Guirkinger and Boucher (2008) found that unconstrained farmers had
high levels of productivity compared to credit-constrained households; they explained this
difference by the farmers’ possession of productive assets and the credit granted by
informal lenders. Iheke (2008) reported a positive and significant effect of credit on TE of
cassava farmers in South Eastern Nigeria.
As mentioned in the literature review, credit constraints negatively affect TE through risk
aversion and technology choice and adoption. This because credit-constrained farm-
households are likely to invest in less risky and less productive rather than in more risky and
more productive technology. This explanation also holds true in this analysis. Indeed, the
limited financial resources faced by constrained households in the hinterland of Kinshasa
might affect their risk behavior in adopting more productive technologies even if they are
provided by NGOs without repayment obligations. This is plausible as most of the
constrained households do not have access to the extension services provided by NGOs.
Furthermore, the negative effect of credit constraints on the cassava production efficiency in
the hinterland of Kinshasa might be explained by the fact that constrained households have
a low purchasing power to make purchases from local markets and apply the high yielding
- 75 -
varieties (HYV) of cassava planting materials, which are more resistance to Cassava Mosaic
Disease (CMD).
Another possible argument is that non-rationed farm-households might increase efforts to
attain the maximum possible output and minimize the waste of inputs in the production
process in order to face repayment obligations and social pressure from MFIs. The credit
constraints have negative and no statistically effects on farm household SE. This suggests
that credit constraints are not important in explaining the differences in SE distributions.
3.5 Conclusions
This chapter aims at evaluating the impact of credit constraint on cassava production
efficiency. The study also explores the main drivers of cassava production efficiency
differentials. DEA was computed to estimate technical and SE scores of cassava production.
A TR model was applied to determine key factors associated with cassava production
efficiency. Propensity Scores Matching was applied to assess the effect of credit constraints
on the efficiency of cassava production.
The results reveal high levels of technical inefficiency in the cassava production: the average
TE is about 0.318 under VRS and 0.272 under CRS, indicating the possibility of increasing the
current level of cassava output. The results also suggest that TE scores of sample farm-
households may be increased by about 0.73 under CRS and by about 0.68 under VRS. The
average SE (0.902) is markedly higher in comparison to TE. This means that the resource
misallocation makes a greater impact on the cassava production inefficiency than farm size.
Household participation in associations appears to be among the key factors positively

affecting cassava production efficiency as farmers participating in associations are more
technically efficient than their counterparts. Households led by a head with more years of
formal education tended to be more technically efficient than those led by less well-
educated household heads. Landholding property is found to be positively correlated with
technical efficiency scores, indicating that households with landholding property are more
efficient than their counterparts. Farm size is another covariate positively associated with
- 76 -
cassava production efficiency, indicating that farms of relatively big size are more technically
efficient than small farms.
Credit constraints conditions are found to negatively affect the technical efficiency. The
average technical efficiency under VRS is estimated at 0.246 for constrained and 0.497 for
unconstrained households. Under CRS, the average technical efficiency score is estimated at
about 0.202 for constrained and about 0.445 for unconstrained households. Moreover, the
results from the matching approach reveal that there is a negative and significant impact of
credit constraints on technical efficiency, which is estimated at about 0.251 % under VRS and
0.260 under CRS. These findings are consistent with our hypothesis that credit constraints
reduce the technical efficiency of constrained farm’s households. However, the effect of
credit constraints on scale efficiency is not statistically significant. This rejects the hypothesis
that credit constraints reduce the scale efficiency of farm-households.
The implications are that the public policy to improve education and information will
enhance resources used in agricultural production by increasing technical efficiency.
Moreover, the results call for public policy to improve farm-households access to sufficient
credit and landholding property in order to reduce the inefficiency in cassava production in
the hinterland of Kinshasa. Our findings imply that there is a policy need to improve
household participation in the social economy, such as farm organizations and cooperatives,
and to improve efficiency in cassava production.
- 77 -
App. Table 3.1 Statistic description of variables used in Chapter 3
Variable Mean SD Min. Max.
Education 9.430 3,859 0 18
Household head male 0.910 0,285 0 1
Household head age 45.049 9,808 22 66
Dependency ratio 0.506 0240 0 0,9
Household size 4.950 1,849 1 11
Land property holding (2007) 0.554 0,498 0 1
Farm size 13.341 11,456 2,5 100
Credit constraints conditions 0.712 0,453 0 1
Remittance 0.420 0494 0 1
Association (2007) 0.698 0,698 0 1
Mbankana group 0.342 0,475 0 1
Dumi group 0.322 0,468 0 1
Menkao village 0.336 0,473 0 1
Cassava production 20074.800 14964,09 2000 81400
Cassava materials 761.623 722.607 74.074 5541.667
Labor 278.516 231.291 29.411 1544.118
Land used for cassava production 1.675 1.025 0.200 8.14
- 78 -
Chapter 4
Impact of Credit Constraints on Farm-Household Welfare in the Hinterland of Kinshasa
4.1 Introduction
Despite its large and various natural resources, the DRC remains one of the poorest
countries in the world. Kalonji (2005) noted that poverty in DRC and particularly in Kinshasa
has become a mass phenomenon. Since the beginning of the 1990s, the average per capita
income has fallen under 100USD and the poverty incidence has been over 85% (Kalonji,
2005). Moreover, food insecurity has been increasingly alarming and a matter of much
concern. The FAO data reported that 17 million people were under-nourished in the DRC
(Tollens, 2004). In contrast with the 1980s, the poverty incidence has increased in Kinshasa
compared to Lubumbashi, Mbuji-mayi, Boma and Matadi (Ministère du Plan et Commerce,
2002). Using household surveys from the peripheries of Kinshasa, Nkwembe (2006) and
Muayila (2004) found that farm-households were more exposed to poverty incidence,
poverty depth and poverty severity than non-farm-households.
Credit constraints on farm-households have been presented in dominant discourses to be

among the key drivers of rural poverty in the DRC. However, to the best extent of our
knowledge, academic research and studies on the cost of credit constraints on rural poverty
using evidence from DRC remain poor. Furthermore, many empirical studies from
developing nations have focused the effect of microcredit on income, consumption
smoothing and household welfare (Khandker and Faruque, 2003; Coleman, 1999; Morduch
and Haley, 2002; Pitt and Khandker, 1998; Wright, 2000; Khandker, 2001; 2003; Arun et al.
2006; Aroca and Hewings, 2009; Negash, 2008). Only a few studies have directly evaluated
the impact of credit constraints on rural household welfare (Baiyegunhi et al. 2010; Li and
Zhu, 2010). This chapter aims at contributing to the literature on credit constraints and rural
poverty using evidence from the hinterland of Kinshasa. This study examines how far credit
constraints affect farm-household economic welfare.
The remainder of this chapter is organized as follows. Section 4.2 presents the review of
literature on the relationship between credit and household welfare. Section 4.3 describes
- 79 -
the data and research methodology. The estimated results are reported and discussed in
section 4.4. Finally, section 4.5 summarizes the main findings and draws policy implications.
4.2 Review of literature
Credit markets in most developing nations operate inefficiently and are recognized as
imposing severe financial constraints on households (Stiglitz and Weiss, 1981; Kochar, 1997;
Foltz, 2004). However, it is surprising to note that the effect of credit constraints on farm-
household economic welfare is poorly documented. Only a few empirical studies have
addressed the issue of the evaluation of the impact of credit constraints on household
welfare (Li and Zhu, 2010; Baiyegunhi et al. 2010). For example, using survey data from
Chinese rural households, Li and Zhu (2010) found that credit constraints had significant and
negative effects on the income and consumption. In addition, Baiyegunhi et al. (2010) found
that relaxing credit constraints among rural households from the Eastern Cape Rural Finance
Corporation would improve the welfare of credit-constrained households.
Literature shows different channels through which access to credit affects household
welfare distribution. Credit may be used for consumption and income smoothing as the
consumption smoothing could become less costly when households have access to credit
(Deaton, 1992; Maldonado, 2004). Credit may be used as an effective tool to manage and
cope with risk; this is known as the risk-coping effect (Maldonado, 2004). Access to credit
may increase purchasing power and lead to an increased purchase of goods (the liquidity
effect), or may impact household welfare through the income and wealth effect (González-
Vega, 2003). Better access to credit by households may affect their welfare distribution by
reducing the costs of trading and of participating in markets and could allow a more efficient
allocation of resources, thereby increasing incomes (Maldonado, 2004). Credit could also
allow households achieving more rewarding and stable consumption paths over time
(Maldonado, 2004).
During the two last decades, there has been an increased focus in the academic literature on
the impact of microfinance on poverty reduction. This increased interest on the impact of
microfinance might be viewed as the part of the paradigm by which microcredit can
- 80 -
contribute to the reduction of poverty and the improvement of the well-being of poor
households in developing nations. This paradigm is purely and simply a demonstration of a
virtual sequence: microcredit – pro-poor growth – poverty reduction, inspired by the
“financial liberalization, growth and poverty reduction paradigm”. The idea behind the
microfinance paradigm is that by providing microcredit to poor households, the credit helps
them improve their welfare, smooth income and consumption flows and improve other
benefits, such as health care and education, allowing poor households to escape poverty and
transform their lives (Khandker and Faruque, 2003; Coleman, 1999; Morduch and Haley,
2002; Pitt and Khandker, 1998).
However, empirical studies conducted on the impact of microfinance using rigorous

methodologies reveal disagreement and mixed results. The most cited case is reported in
Pitt and Khandker (1998) and Morduch (1998). Using data collected from 87 villages in 29
sub-districts and from group loan clients of three microfinance programs in Bangladesh
(Grameen Bank, BRAC, and RD-12), and applying a fixed effect research design, Pitt and
Khandker (1998) reveal that every additional “taka” lent to a woman resulted in additional
0.18 taka on annual household expenditures compared to male borrowers and an 18%
return to income from borrowing.
However, Morduch (1998) pointed out serious concerns with the statistical fixed effect
model applied in Pitt and Khandker (1998). He argued that this method of correcting for
program placement bias exacerbates the bias. Then he analyzed the same data using
difference-in-difference. The findings from Morduch analysis reported no significant
consumption differentials between participants and non-participants. They reveal that the
only significant impact of microfinance was associated with vulnerability reduction rather
than poverty reduction. Using household data from the northeast of Thailand, Coleman
(1999) also found that access to microcredit had no significant effect on improving
household welfare, using data from data from 10 villages served by BRAC for participants
and eligible non-participants, and by applying the Heckman procedure to correct the
selection bias. Zaman (1999) found that microfinance significantly reduces poverty and
vulnerability. He also reported that the impact of microfinance on poverty was large when a
loan reached a certain cumulative level, such as 200USD.
- 81 -
Applying the PSM as proposed in Rosenbaum and Rubin (1983) to correct the selection bias,
Arun et al. (2006), Aroca and Hewings (2009) and Negash (2008) reported positive effects of
microfinance on household consumption. Many other empirical studies showed positive
effects of microfinance on household economic welfare (Wright, 2000; Khandker, 2001;
2003; Khandker and Faruque, 2003). In contrast to the empirical studies that have reported
positive and significant effects of microfinance on household welfare and poverty reduction,
some empirical studies have revealed that microcredit is not intended for the poorest and
most vulnerable (Hulme and Mosley, 1997; Evans et al. 1999, Navajas et al. 2000). Goetz and
Sen (1996) found that microfinance inflicted extreme pressure, especially on women, who
usually experience difficulties repaying loans (Goetz and Sen, 1996). To sum up, the mixed
results from the impact of microfinance on poverty reduction indicate that microfinance
paradigm is far from achieving unanimity on the relation between microcredit and poverty
reduction. The mixed results stress the complexity and the difficulty to correct the selection
bias and data limitations affecting the outcome of research. Despite these mixed results,
most empirical studies support the positive effects of microfinance on poverty reduction.
4.3 Methodology and data
4.3.1 Data
The dataset used in this chapter was collected in the hinterland of Kinshasa and consisted of
202 households. For a detailed description of the study areas and data collection, see
Chapter 1. The App. Table 4.1 presents the descriptive statistic of variables used in this
chapter.
4.3.2 Research methodology
To estimate the causal effect of credit constraints on farm-household welfare, we applied a

matching approach. Thus, if we denote the welfare level (here consumption) of a
constrained household by W1 , and outcome for unconstrained by W0 , and the credit
- 82 -
constraints conditions or treatment variable by Di , which is a dummy variable taking value 1
if the household is constrained and 0 otherwise, the ATT may be computed as :

ATT  E (W1  W0 / D  1) (4.1)
= E W1 / D1  1  E W0 / D1  1
For sample, ATT is given by:

1 n
ATT   (W1i  W0i / Di  1) (4.2)
n i 1
1 n
  (W1i / Di  1)  (W0i / Di  1)
n i 1
One could observe that it is impossible to calculate ATT using the above equations as
(W0i / Di  1) is not observed. Hence, the challenge is to estimate this counterfactual. If credit
constraints conditions were random, (W0i / Di  0) could be used as proxy, but this is only
possible in case of experimental data and absence of selection bias.
To avoid this problem of missing data, we apply the PSM proposed by Rosenbaum and Rubin
(1983). We assume that being credit-constrained could be similar to a “treatment”.
Therefore, the average treatment effect (ATT) may be estimated as a proxy of causal
inference of credit constraints on household welfare distributions. By applying PSM, we
compare the distributions of household welfare indicators for constrained households versus
that for unconstrained households with a similar propensity of being credit-constrained or
treated. Then the difference between matched treated and control units is attributed to the
cost of credit constraints.
Given that it is very hard to expect many observations within the same region, the
computational of the exact matching estimator is not possible. To avoid this problem and
following Dehejia and Wahba (2002), we apply the nearest neighbors matching with
replacement so to assure that each matched treatment unit is compared to the matched
nearest control unit, allowing for a bias reduction. Following Becker and Ichino (2002), the
ATT under the nearest neighbor matching is obtained by calculating the following algorithm:
- 83 -
1  
ATT 
NT
 W i
T
  w W ij j
C

iT  jC i  (4.3)
I  
T 
 Wi T    wijW jC 
N  iT iT jC i  
1 1

NT
W
iT
i
C

NT
w W j j
C
Where: T = vector of treated individuals; C = vector of control group; Wi T = observable
outcome welfare indicator of constrained households matched group; Wi C = observable
outcome welfare indicator of control group matched; C i vector of households in the control
group matched to treated group with Pi ; N C and N T are respectively the observations in the
control group and that of the treated group and C (i)  min j Pi  Pj (4.8)
The first step in the implementation of the PSM procedure is the estimation of the
propensity score of being credit-constrained using the logit or probit models. The dependent
variable is household credit constraints conditions, which is a binary choice taking value 1 if
household faces credit constraints and 0 otherwise. The conditional variables are those
related to household and institutional characteristics, including the household size, the
dependency ratio, the remittances, the age of the household head, the gender of household
head, the landholding property, the farm size, associations, and the villages of residence.
The outcome variables are household welfare indicators, measured in terms of

consumption. The use of consumption to measure welfare rather than income is explained
by theoretical and empirical reasons. Firstly, household consumption level is more stable
over time than income because of small transitory components of consumption. Secondly,
transitory income fluctuations affect the precision of the cross-sectional data on welfare.
Thirdly, household income is more difficult to measure than household consumption or
expenditure. Fourthly, estimated income for a period less than one year, especially for
agricultural households, may be affected by seasonal variation. Fifthly, seasonal patterns in
consumption are less marked because of consumption smoothing. Sixthly, household
income is difficult to construct for households with informal activities. Finally, questions on
income may be viewed as more sensitive than those on consumption (UNSD, 2005).
- 84 -
Moreover, daily household consumption per adult equivalent is calculated to allow the
comparability of welfare of households with different size and composition. The household
consumption per day and per adult equivalent is obtained by dividing total consumption of
household per day by the number of adult equivalents in the household. Literature provides
a wide range of equivalence scales (see Atkinson et al. (1995). However, the most commonly
used in empirical studies are the OECD equivalence scale, also called the Oxford scale, the
OECD modified equivalent scale and the FAO equivalence scale.
The OECD equivalence scale assigns a value of 1 to first household adult member, 0.7 to
each household additional person and 0.5 to the children of less than 15 years old. The OECD
modified equivalence scale was adopted in the late 1990s. It assigns a value of 1 to
household head, 0.5 to each additional adult person and 0.3 to child of less than 15 years
old. FAO constructs the adult equivalence scale by assigning a value of 1 to the household
head, 0.8 to each adult household member, and 0.5 to each child less than 15 years old.
These approaches address the issue of equating the needs of children with those of adults by
converting them into the adult equivalence, as there might be lower child needs (food and
non-food) consumption. The main problem of these approaches is that in many developing
nations, as in the area of this study, food consumption and expenditure represent a large
part of the household budget, which leads to lower economies of scale. Furthermore,
because of the importance of food consumption in household total expenditures, it does not
appears realistic to believe that additional adult in household will represent a part of the
needs of the first adult needs, particularly food consumption.
Therefore, the equivalence scale applied in this study consists of assigning a value of 0.5 to
children of less than 15 years old, as in the Oxford scale equivalence. As mentioned in Sylla
(2002), we assign the value of 1 to adults members of household. Household consumption is
classified in food consumption and non-food consumption. Non-food consumption is
obtained by adding durable and non-durable goods. Only the amount of durable goods
consumed during the period of survey is accounted for the calculation of total non-food
consumption. Food consumption is obtained by adding the value of the food obtained from
the farm and food bought in the market. Following Johnson (2004), food from farm
- 85 -
production and other self-service production are evaluated by asking how much households
would have to pay if they were bought in the local markets. To provide a better
understanding of how households fit the minimum standard of living, the household total
consumption per adult equivalent and per day was divided by the poverty line (1 USD and 2
USD) to obtain welfare index 1USD and welfare index 2USD.
4.4.1 Results
This section reports the results of the propensity score matching used to estimate the
impact of credit constraints on farm-household welfare. The results of the logit estimation of
the propensity score are reported and discussed in Chapter 2. The distribution of the
propensity score is reported in Chapter 3. The computation of ATT requires balancing in the
covariates chosen to estimate the propensity score. We address balancing properties by
testing for equality of means between all treated and control units matched for all selected
covariates used. The results are reported in Chapter 3 and show that after the correction of
bias, the treated and untreated groups matched do not differ substantially in observables
covariates. In most cases, the differences in covariates for unmatched samples (treated and
control group) are significant, while the difference between matched samples are very low
2
and are not significant. Moreover pseudo R obtained after matching is not significant
indicating that the covariates are well performed and then ATT may be computed. The
distributions of propensity score (Figures 3.5 and 3.6) show sufficient overlaps.
As mentioned in the research methodology design, ATT is run using nearest neighbor
matching with replacement, which assures that each matched treated unit is compared to
the matched nearest control unit. We only matched observations in the common support
region, which is an area where the propensity scores of the treated observations is not
higher than the maximum or less than the minimum propensity score of the control (Becker
and Ichino, 2002). Table 4.1 shows the estimated ATT in the common support region. The
results reveal negative and significant (at 1% of significance) for all ATTs, except that of non-
food consumption.
- 86 -
To check the robustness of the baseline ATT to the possible failure of the CIA, which is an
important requirement for matching estimators, we carry out simulation-based sensitivity
analysis as proposed in Ichino et al. (2006). Thus, we apply a method with a neutral
confounder and with a confounder calibrated to mimic observable and significant binary
covariates, such as land holding property, remittances and association participation. The
results of simulation-based sensitivity analysis are reported in Table (4.2).
Table 4.1 Impact of credit constraints on consumption per adult equivalent per day (FC) as
indicator for household welfare
Outcome Group Average for Average for ATT T-test

treated controls
Food consumption Unmatched 416.660 649.367 -232.706*** -5.58
Matched 427.817 686.960 -259.143*** -3.80
Non-food consumption Unmatched 202.122 324.279 -122.156*** -3.41
Matched 208.288 249.960 -41.671 -0.65
Total consumption Unmatched 680.688 1047.260 -366.572*** -5.42
Matched 698.376 1006.648 -308.272*** -2.72
Welfare index 1 USD Unmatched 1.215 1.870 -0.654*** -5.42
Matched 1.247 1.797 -0.550*** -2.72
Welfare index 2USD Unmatched 0.607 0.935 -0.327*** -5.42
Matched 0.623 0.898 -0.275*** -2.72
- 87 -
Table 4.2 Simulation-based sensitivity analysis
Outcome Outcome Selection ATT

effect effect
Food consumption
Non-food consumption
Total consumption
Welfare index 1USD
Welfare index 2USD
With confounder like: 1.522 1.022 -0.204
Legend: ***p<.01
- 88 -
The estimated results for the unmatched sample reveal substantial differences in welfare
distributions between constrained and unconstrained households. The unconstrained
households have high levels of consumption compared to constrained households. The
average difference is about 232 FC (0.41 USD) for food consumption, 122 FC (0.22 USD) for
non-food consumption, 366 FC (0.65 USD) for total consumption, 0.65 for welfare index
1USD and 0.33 FC welfare index 2USD. The difference between unconstrained and
constrained households is significant at 1% for all selected indicators of welfare. Moreover,
food consumption represents an important part of household consumption. This
corroborates previous studies on Kinshasa and its hinterland (Nkwembe, 2006; Muayila,
2004). These unmatched and naïve estimations should not be viewed as efficient indicators
of causal effect of credit constraints conditions because of possible selection bias, due to the
fact that credit constraints conditions are not random. The bias results from the fact that
there might be lender selection and borrower self-selection (Toshio, 2007). The lenders
could grant more privileges to the richest of the poor, who are perceived as less risky than
the poorest of the poor (Hulme and Mosley, 1996; Evans et al. 1999; Navajas et al. 2000).
Thus, some observable socio-economic and demographic characteristics of households, as
well as the transaction cost related to the distance, could be applied by lenders to select
potential borrowers. The self-selection operated by the borrower could be bound by a set of
decisional factors, such as education, risk aversion, motivation, incentives, resource and
confidence that might be associated with participation in the credit market and welfare
level.
As shown in Table 4.1, the results of matched samples also reveal the substantial difference
between matched treated and control units in welfare distribution. ATT for all outcomes
variables are significant at 1%, except for non-food consumption. The ATT estimated for food
consumption shows that credit constraints reduce significantly per capita food consumption
per day by 259FC (0.46 USD) on average. When total per capita consumption is used as
outcome variable, credit constraints have on average a reducing effect of 308 FC (0.55 USD).
Similarly, when welfare index 1USD and welfare index 2USD are considered as outcome
variables, the average impact of credit constraints is estimated respectively at about -0.550
- 89 -
and -0.274. However, when non-food consumption is applied as outcome variable, ATT is
non-significant and estimated at about -41FC (-0.074 USD).
The findings of simulation-based sensitivity analysis show that when food consumption is
used as outcome variable and confounders behaving like association, remittances and
neutral confounder are simulated, baseline ATT (-259 FC ( -0.46 USD)) is brought down,
respectively to -208FC ( -0.37 USD), 221 FC (0.39 USD) and -199FC (-0.35 USD), but remains
significant. These ATT changes are due to the strong outcome effect, which is 32 for
association participation, 8 for remittances and 1 for neutral confounder. The same feature
is observed for total consumption, welfare index 1 USD and welfare index 2USD. To sum up,
the results of the sensitivity analysis report some changes in ATTs when binary confounders
behave like association, remittances and landholding property and neutral confounder are
simulated. These changes are not economically significant and do not bring ATT down to
non-significant values. Most of these changes might be due to the high outcome effects.
Thus, the findings of sensitivity analysis might be viewed as evidence of the robustness of
estimated matching estimators to the possible failure of CIA.
The negative and significant effect of credit constraints on food consumption, total
consumption, welfare index 1USD and welfare index 2USD confirm our hypothesis that the
credit constraints negatively affect farm-household welfare. Similar results were found in Li
and Zhu (2010), who reported a negative and significant impact of credit constraints on
consumption smoothing using survey data from Chinese rural households. The negative and
significant effect of credit constraints on household welfare implies that policy to remove
credit constraints among farm-households will assist in increasing food consumption or
expenditure, total consumption as well as farm-household welfare. This is in line with
Baiyegunhi et al. (2010), who also found that relaxing credit constraints among rural
households from the Eastern Cape Rural Finance Corporation will improve the welfare of
credit-constrained households. These findings echo numerous empirical studies revealing
the positive effect of household access to credit on food consumption or expenditure
(Negash, 2008; Arun et al. 2006; Aroca and Hewings, 2009; Wright, 2000; Khandker, 2001;
Khandker and Faruque, 2003; Pitt and Khandker, 1998; Khandker, 2003).
- 90 -
However, the effect of credit constraints on non-food consumption is negative but non-
statistically significant. This suggests that credit constraints conditions reduce household
non-food consumption, but the effect is not important in magnitude. This finding is in line
with Negash (2008), who reported a non-significant effect of microcredit on non-food
consumption, but a significant effect on total consumption in Ethiopia. He found that the
effect of microcredit on total consumption was positive and significant, while non-food
consumption effect was found to be non-statistically significant. He then explained this
finding by the higher proportion of food consumption in the composition of household total
consumption.
This explanation holds true in the case of the hinterland of Kinshasa too, as the same feature
is reported in Table 4.1 and in Nkwembe (2006) and Muayila (2004). Another plausible
explanation of the non-significant impact of credit constraints on non-food consumption
could be related to the loan size. Loans provided by MFIs in the hinterland of Kinshasa are
too small and do not allow household funding non-food expenditures such as education,
clothing, clearing and personal care etc. Moreover, given the high level of poverty and food
insecurity in the hinterland of Kinshasa, the additional resources obtained from credit or
other external sources are more likely to be used in priority to food consumption rather that
to non-food consumption. This finding calls for the implementation of well-designed and
sufficient financial provision towards farm-households, responding to the production needs
as well as non-food consumption.
4.5 Conclusions
This chapter aims at estimating the effect of credit constraints on farm-household welfare in
the hinterland of Kinshasa. The propensity score matching is applied to compute the impact.
The outcomes show that on average credit constraints significantly reduce household food
consumption by 259 FC, total consumption by 308 FC, welfare index 1USD by 0.550 and
welfare index 2USD by 0.275. The impact of credit constraints on non-food consumption is
negative (-41 FC) but insignificant. These results confirm our hypothesis that credit
constraints conditions negatively affect household economic welfare. Thus, findings suggest
that provision of well-designed credit towards farm-households will contribute substantially
- 91 -
to the improvement of household consumption and therefore would enhance a farm-
household’s economic welfare.
App. Table 4.1 Statistic description of variables used
Variable Mean SD Min Max
Education 9.430 3.859 0 18
Household head male 0.910 0.285 0 1
Household head age 45.049 9.808 22 66
Dependency ratio 0.506 0.240 0 0,9
Household size 4.950 1.849 1 11
Land property holding (2007) 0.554 0.498 0 1
Farm size 13.341 11.456 2.5 100
Credit constraints conditions 0.712 0.453 0 1
Remittances 0.420 0.494 0 1
Association (2007) 0.698 0.698 0 1
Mbankana group 0.342 0.475 0 1
Dumi group 0.322 0.468 0 1
Menkao group 0.336 0.473 0 1
Per capita food consumption per day 483.477 287.687 100 1864.286
Per capita non-food consumption per day 237.197 236.554 15.238 2353.333
Per capita total consumption 785.941 464.477 210.994 3729.498
Welfare index 1USD 1.403 0.829 0.376 6.659
Welfare index 2USD 0.701 0.414 0.188 3.329
- 92 -
Chapter 5
Performance of Microfinance Institutions in the Democratic Republic of Congo
Does Credit Towards Farmers Matter?
5.1 Introduction
Microfinance is a provision of financial services to the population excluded from the

traditional bank system because of their socio-economic characteristics. It has been praised
as a new paradigm for overcoming poverty in developing countries and is built on the
premise that it would create an inclusive financial sector for poor populations, and therefore
would help to develop their assets. Considering the difficulty for traditional banks to finance
agricultural activities, due to the high transaction costs, high risks and low profitability of this
sector (Wampfler, 2000; Lapenu, 2001; Lesaffre and Pesche, 2002), as well as the failure of
agricultural banks and rural finance programs in several developing countries, MFIs appear
to be important actors able to provide sustainable financial services for poor people, and
certainly farmers and rural households (Kodjo et al. 2003; Tollens, 2004). The lending
technology applied in microfinance allows the majority of problems related to financing
small businesses, particularly farming activities, to be addressed (Kodjo et al. 2003; Tollens,
2004; Narajan and Pritchett, 1999). However, the failure of many microcredit programs of
NGOs with reference to the welfare of people has amplified concerns about their
performance and sustainability. The debate has taken place between institutionalists and
welfarists (Conning, 1999; Woller et al. 1999; DeBriey, 2003; 2005).
The institutionalists, also known as defenders of the commercial approach, are largely
concerned with financial self-sufficiency and sustainability of MFIs rather than their social
responsibility (DeBriey, 2003; 2005; Conning, 1999; Woller et al. 1999). For them, the
introduction of a commercial framework in MFIs would lead to their integration into the
private capital market, consolidate governance and consequently enhance financial
performance (Cornée, 2006). They assert that a sustainable MFI has the capacity to survive
without external subsidies from donors and governments (Cornée, 2006; DeBriey, 2003;
2005). On the other hand, the proponents of the welfare approach, especially NGOs, argue
that the raison d’être of microfinance is the provision of financial services to the poor.
- 93 -
Furthermore, they argue that commercialization of microfinance could increase the
exclusion of the poorest of the poor from the provision of financial services (DeBriey, 2003;
2005; Conning, 1999; Woller et al. 1999). Thus, the adherents of the welfare approach assert
that sustainability should be a secondary issue.
In spite of this debate, the provision of sustainable financial services to the poor populations
has been recognized as a big challenge for MFIs. Achieving this requires a better trade-off
between social and financial performance. In other words, if MFIs would like to continue
providing financial services to the poor in a sustainable way, they must be themselves viable
and sustainable. The latter assumption is based on the conceptual framework developed by
Yaron (1994): “outreach” and “sustainability”. Outreach measures social performance, while
sustainability is described by the financial performance. In recent years, studies on the
performance of microfinance were done with reference to this conceptual framework
(Cornée, 2006; Gutiérrez-Nieto et al. 2009). In the DRC, we note a lack of rigorous analysis
and serious empirical studies on the performance of microfinance. To the best of our
knowledge, the analysis of efficiency in the microfinance industry in the DRC remains poor. A
few existing empirical studies were conducted through the analysis of financial ratios or one-
dimensional econometric analyses (Bambila, 2009; Aliango, 2004; BCC, 2001; Bondobondo,
2002; Kalala, 1988; Kikassa, 1981). This study evaluates the performance of microfinance
organizations, using evidence from the DRC. This study tests whether there is a good balance
between financial performance and social responsibility of MFIs.
The determinants of the efficiency of MFIs have been a subject of many empirical studies
(Martinez-Gonzalez, 2008; Paxton, 2007; Cull et al. 2007; Hudon and Traca 2011; Gutiérrez-
Nieto et al. 2009). However, to the best of our knowledge, the impact of the provision of
agricultural microcredit on the efficiency of MFIs is not well documented. There is a lack of
evidence on whether it is possible for MFIs to provide financial services to farm-households
and be at the same time sustainable. Therefore, this study focuses on the relation between
agricultural credit and efficiency of MFIs. It investigates whether it is possible for MFIs to
finance farm-households and achieve a good performance at the same time.
- 94 -
Financial institutions performance analysis has been traditionally measured through financial
ratios (Tucker and Miles 2004; Koveos & Randhawa, 2004). The financial ratios analysis is
certainly fertile for some questions, but appears inappropriate since one seeks an aggregate
indicator of the efficiency, which captures long term performance (Sherman and Gold, 1985;
Soulama, 2008; Cornée, 2006). The activities and decisions of financial institutions have a
multidimensional aspect that could not be examined and evaluated through financial
analysis of ratios (Athanassopoulos and Ballantine, 1995; Gregoriou et al. 2005). Moreover,
the analysis of performance through only financial ratios can be arbitrary (Yeh, 1996).
During recent years, there has been a trend towards measuring the efficiency of financial
institutions, certainly of microfinance organizations, using the frontier models. The most
used was a nonparametric method, known as the DEA. Emouznejad and Podinovski (2004)
have identified more than 4000 studies in journals and book chapters; for example, Favero
and Papi (1995), Taylor et al. (1997), Rangan et al. (1988), Sherman and Gold (1985) and
Berger and Humphrey (1997) have applied DEA to the measurement of bank efficiency and
performance. There has frequently also been a trend towards measuring microfinance
performance by applying DEA analysis (Soulama, 2008; Cornée, 2006; Gutiérrez-Nieto et al.
2009).
This study applies DEA to compute technical and scale efficiency of MFIs operating in the
DRC. Following Coëlli (1996), we assume that there are k inputs I  and m outputs R 
for n microfinance organizations where u is the output weight and v is the input
weight. The DEA efficiency score for each MFI under CRS was obtained by maximizing the
following dual linear program:
Min  ,  (5.1)
Subject to  R0    j R j  0
I 0    j I j  0
j
- 95 -
0
Where:  is a scalar denoting the efficiency score and  is a Nx1 vector of constraints.
The efficiency scores are assumed to range from 0 to 1. The Unit Decision Making (UDM) is
efficient if it achieves the efficiency scores equal to 1. However, CRS model is only
appropriate if all firms are operating at their optimal scale (Coëlli, 1996). This assumption is
unrealistic because of input constraints and imperfect competition (Coëlli, 1996). The
previous empirical studies showed the fact that institutions face non-constant returns to
scale in most cases (McAllister and MacManus, 1993; Wheelock and Wilson, 1999).
To deal with this criticism, the VRS assumption is introduced. VRS does not assume that all
DMUs are operating at optimal scale. The introduction of VRC is done by incorporating a
convexity constraint to equation (5.1), so that N1'   1 . This additional constraint results in a
convex hull that envelops data points more compactly, resulting in efficiency scores greater
or equal to those from CRS model. By calculating a CRS DEA along with the VRS DEA,
technical efficiency scores for each method can be compared, and the ratio illustrates the
SE (Coëlli, 1996).
The specification of DEA for efficiency score estimation requires some decisions to be taken
on inputs and outputs, the orientation approaches and a hypothesis on SE. The definition of
inputs and outputs in the financial sector has been a subject of controversy in the literature.
Two models have emerged: the intermediation model and the production model (Korsah et
al. 2001). The first considers financial institutions as an intermediary between savers and
borrowers. In this approach, deposits are used as inputs and loans as output (Korsah et al.
2001). However, the second considers financial institutions as the producers of deposits and
loans using available inputs. The production approach uses deposits and loans as outputs
(Korsah et al. 2001; Soteriou and Zenios, 1999; Gutiérrez-Nieto et al. 2009). The problem
here is to establish whether the deposit should be regarded as an output, as suggested in
the production model, or an input, as suggested in the intermediation model.
In this study, the production model and outreach-performance framework (Yaron, 1992) are
used for the selection of the inputs and outputs of the MFIs. To provide these financial
services in a sustainable way, MFIs must be autonomous, sustainable and financially viable
- 96 -
(Soulama, 2008). The provision of these services requires the use of physical, financial and
human resources. The MFIs are different from traditional banks in that they have a social
and financial mission, known in the literature on microfinance as “double bottom
performance”.
There is no consensus in the literature on the measurement of the social performance of

microfinance. However, many previous empirical studies have applied several indicators as a
proxy for the depth and breadth of microfinance outreach (Zeller et al. 2002; Lapenu et al.
2004). The most useful proxies were the percentage of active women borrowers, the
average loan balance per borrower and the average deposit balance per saver (Lapenu et al.
2004). Regarding the measurement of financial performance, empirical studies have
traditionally applied financial performance indicators, such as the Returns on Asset (ROA),
the Returns on Equity (ROE) and the Operational Self Sufficiency (OSS), as outputs for the
estimation of DEA financial efficiency scores (Soulama, 2008; Cornée, 2006; Gutiérrez-Nieto
et al. 2009). ROA, ROE and OSS are used in the analysis of microfinance performance to
measure the institutional ability to acquire capital at a reasonable cost and to invest these
funds in a profitable way.
With reference to the outreach and sustainability conceptual framework suggested by Yaron
(1994), and data constraints, the percentage of Active Women Borrowers (W), the number
of total Active Borrowers (nL), the number of total Savers (nS), the ROA and OSS, total Loan
Portfolio (L), and total Deposits (S) are selected as outputs; while the total Assets (A) and the
total number of Personnel (P) are selected as inputs for the estimation of DEA efficiency
scores.
The percentage of active women with access to micro credit is used as social output in the
DEA estimation of social efficiency scores by Cornée (2006) and Gutiérrez-Nieto et al.,
(2009). It appears also as one of the indicators of microfinance outreach in the Mixmarket
dataset. The total number of active borrowers and total number of active savers are applied
in several empirical studies and the Mixmarket dataset as proxies of microfinance outreach.
The Return on Assets and Operational Self-Sufficiency are traditionally used as outputs in
many studies on banks and MFI performance (Soulama, 2008; Cornée, 2006; Gutiérrez-Nieto
- 97 -
et al. 2009). The total number of Personnel is applied in DEA estimation of banks
performance by several authors (Dekker and Post, 2001; Cook et al. 2000; Sherman and
Gold, 1985) and microfinance performance (Cornée, 2006; Gutiérrez-Nieto et al. 2009). Total
asset is used as input in the estimation of efficiency score of MFIs in Cornée (2006), Soulama
(2008) and Gutiérrez-Nieto et al. (2009).
Four empirical DEA models are specified to compute the efficiency scores and the results are
then compared. The first DEA specification (LS-AP) measures the efficiency in the provision
of micro saving and microcredit. The second DEA model specification (ROA-OSS-AP)
associates two overall financial indicators (ROA and OSS) with (P) and (A) for the estimation
of overall financial efficiency scores. The third specification (W-nL-nS-A-P) links (W), (nL) and
(nS) to (A) and (P) to measure social efficiency scores. The last and fourth DEA specification
(W-ROA-OSS-AP) joins (ROA), (OSS) and (W) to (A) and (P) for the measurement of
microfinance double bottom efficiency scores. This specification measures the ability of MFIs
to reach the poorest of the poor and being at the same time sustainable. Following Berg et
al. (1993), microfinance efficiency scores are computed under CRS and VRS assumptions to
allow for the comparison of efficiency scores obtained from both assumptions.
- 98 -
Table 6.1 Descriptive statistics of inputs and outputs
Without agricultural
Total Sample With agricultural credit credit
Variables Average DS Average DS Average DS
Savings (million USD) 2.872013 11.36378 3.552335 12.60312 2.386069 10.68667
Credit (million USD) 0.7908861 2.547126 1.048816 2.511908 0.6066502 2.617524
ROA 5.63597 8,397314 2.83117 2.705053 7.61582 10.406987
OSS 153.5472 77.90932 126.921 19.48778 172.3421 97.23231
% women borrowers 44.44598 29.04801 46.34187 26.66122 43.02406 31.32421
Number of savers 2966.4 6476.566 4521.867 9414.118 1799.8 2532.85
Number of borrowers 1665.486 6360.077 3476.333 9543.563 307.35 862.2856
Assets (million USD) 3.160864 10.99603 1.647104 3.929048 4.242121 14.06573
Total personnel 23.13889 57.94981 33.86667 76.65215 15.47619 40.19903
5.3.1 Results
Figure 5.1 presents the average distribution of TE score (CRS and VRS) and SE score for all
DEA models specified. The results from all DEA models specified, except that of the social
efficiency model, show that the TE score under VRS is higher than CRS and SE.
- 99 -
Figure 5.1 Average efficiency score
Efficiency in the provision of saving and credit
The TE (CRS) in the provision of micro saving and microcredit ranged from 0.02 to 1.00 with
an average of 0.43(SD=0.29). The majority of MFIs have a low TE score (CRS) and hence are
inefficient. Indeed, only 11 percent of MFIs from the sample are at the frontier of TE (CRS),
whereas more than 60% of MFIs have a technical efficiency score (CRS) ranging from 0.02 to
0.50. The distribution of the TE score (VRS) in the provision of micro-saving and microcredit
is greater than that obtained from the CRS assumption. The average TE (VRS) was estimated
at 0.65(SD=0.27) and only 25 percent of MFIs are located on the frontier of best practice,
and thus are technically efficient. More than 60 percent have a TE score (VRS) higher than
0.50.
The average SE is estimated at 0.62, and only 11 percent of MFIs are scale efficient.
However, 39 percent of MFIs have a SE score higher than 0.75. These results suggest that
inefficiency in the provision of the micro-saving and microcredit is technical inefficiency as
well as scale inefficiency. Therefore, with respect to the existing production technology, TE
(CRS) and technical efficiency (VRS) could be increased by 0.57 and 0.35 of score on average,
respectively. The increase of TE could be made possible by improving management skills
- 100 -
(human resources management, procedure, asset management, etc,) as well as by increasing
inputs (assets and number of personnel). However, by increasing the level of inputs and
management, the SE will increase by 0.38 on average.
Figure 5.2 Efficiency in the provision of saving and credit
Efficiency in the production of financial performance
The distribution of TE scores in the production of overall financial performance (ROA and
OSS) shows that the average technical efficiency scores is estimated at 0.68 (SD=0.26) and
0.89 (0.11), respectively, under CRS and VRS assumptions. About 28% of MFIs (under CRS)
and 38% (under VRS) are on the frontier of best practices. Thus, these institutions would be
regarded as financially sustainable and self-sufficient. The average SE (0.76(SD=0.25)) is
greater than the average TE (CRS) and lower than the average technical efficiency (VRS).
Moreover, only 31% of the MFIs are scale efficient; more than 57% reach SE score higher
than 0.76, while more than 20 percent of MFIs have a scale efficiency score ranging from
0.18 to 0.50.
These results suggest that the average TE (CRS) and technical efficiency (VRS) related to the
maximization of overall financial performance would be improved by 0,11(CRS) and
0,32(VRS) if resources were allocated optimally. The fact that SE scores remain lower than TE
- 101 -
(VRS) indicates that the inefficiency is more scale inefficiency than technical inefficiency
(VRS). The comparison between SE and TE (CRS) shows that the inefficiency is more technical
inefficiency than scale inefficiency. However, this assumption is less plausible because many
MFIs are operating under VRS rather than CRS assumptions.
Figure 5.3 Efficiency in the production of financial performance
Efficiency in the production of social performance
The average TE score (CRS) with regard to outreach production, and particularly to the
access of the poorest of the poor to financial services, is estimated at 0.68 (SD=0.31),
whereas technical efficiency (VRS) accounted for about 0.75 (SD=0.29). Furthermore, 31
percent and 42.9 percent of MFIs are on the frontier of the social efficiency line. These MFIs
are regarded as the most efficient in the production of outreach indicators, given the
existing technology and resources used. In addition, more than 50 percent of the MFIs from
the sample are operating in the region of a TE (CRS) lower than 0.50, as against 26 percent
for TE (VRS).
The average SE (0.91 (SD=0.16)) under the social performance model is greater than TE for
both VRS and CRS assumptions. Furthermore, none of the MFIs in the sample have a SE
score lower than 0.25. About 31% of MFIs are operating on SE, and more than 50 percent of
MFIs had a scale efficiency score ranging between 0.76 and 0.99. In the light of the above
- 102 -
results, one can note that the inefficiency in the production of social performance of MFIs
from the sample is technical inefficiency and resulted from the misallocation of resources
due to the lack of management skills rather than the size of operation. Therefore, by
improving the management of resources or using inputs in an optimal way, TE will be
increased on average by 0.11 under VRS and 0.32 under CRS.
Figure 5.4 Efficiency in the production of social performance
Efficiency in the production of microfinance double bottom performance
Taking into account both the missions of microfinance, outreach and financial sustainability,
measured respectively by the access of the poorest to financial services provision (W) and
the indicators of overall financial performance, such as ROA and OSS, the results obtained
from DEA show that MFIs do better in double bottom performance, rather than in the
previous DEA models. Indeed, about 50 percent (VRS) and 39 percent (CRS) of MFIs are
technically efficient in the production of the so-called microfinance double bottom.
Technically, these MFIs make a better trade-off between social responsibility and financial
sustainability, given the available production technology. The average TE score (CRS) is
estimated at 0.79(SD=0.22) and technical efficiency score (VRS) is 0.91 (SD=0.11). There are
no MFIs with less than 0.50 of TE score (VRS), whereas they are 10% under CRS. The average
SE is estimated at 0.86 (SD=0.19). In addition, 32 percent, 21 percent and 7 percent of MFIs
- 103 -
are respectively in the intervals of (0.76-0.99), (0.51-0.75) and (0.26-0.50) SE scores. About
40 percent of the MFIs achieve the optimal SE. The inefficiency appears to be more technical
inefficiency under CRS and more scale inefficiency when the VRS assumption is assumed.
Figure 5.5 Efficiency in the production of microfinance double bottom performance
Agricultural credit and efficiency of microfinance
Table 3 shows the statistical description of technical and SE scores with reference to
agricultural credit provision. One can observe small differences in the distribution of scores
between MFIs with agricultural credit versus MFIs without agricultural credit. To know
whether these differences are statistically significant, the analysis of the variance (ANOVA)
and the nonparametric (Mann-Whitney) tests are carried out. The results of these tests are
described in Table 5.2. With the reading of these results, one notes that for all DEA models
specified, financing agriculture is not associated with technical and SE score distributions.
Thus, microcredit provision towards agricultural producers does not have significant effect
on the efficiency and performance of MFIs.
- 104 -
Figure 5.6 Technical efficiency score (CRS) for MFIs with versus without agricultural credit
Figure 5.7 Technical efficiency score (VRS) for MFIs with versus without agricultural credit
Figure 5.8 Scale efficiency for MFIs with versus without agricultural credit
- 105 -
Table 5.2 Results from ANOVA and Mann-Whitney tests
MODELS ANOVA MANN-WHITNEY

F (dl) Probability Z Probability
Efficiency in the provision of saving and credit ( SL-AP)
CRS 0.186 (1-34) 0.669 -0.128 0.8978
VRS 0.040 (1-34) 0.842 -0.340 0.7342
SE 0.103 (1-34) 0.751 -0.321 0.7481
Efficiency in the production of financial performance (ROA-OSS-AP)
CRS 0.016 (1-27) 0.899 0.022 0.9822
VRS 0.101 (1-27) 0.753 0.638 0.5238
SE 0.049 (1-27) 0.827 0.090 0.9284
Efficiency in the social performance (W-nL-nS-AP)
CRS 0.039 (1-33) 0.845 0.203 0.8390
VRS 0.119 (1-33) 0.732 -0.417 0.6769
SE 0.327 (1-33) 0.571 0.643 0.5200
Efficiency in the production of double bottom performance (W-ROA-OSS-AP)
CRS 0.066 (1-26) 0.799 0.335 0.7374
VRS 0.311 (1-26) 0.582 -0.074 0.9407
SE 0.662 (1-26) 0.423 0.527 0.5984
- 106 -
The TE score (VRS) is greater than the technical efficiency score (CRS) in all DEA models
specified, except in the SL-AP model. Thus, VRS assumption leads to an increase in the level
of TE. This confirms what has been reported in the existing literature (Coëlli, 1996). In
addition, VRS is more plausible than CRS. Indeed, about 75% in the case of the provision of
saving and credit model, 62% in the overall financial performance model, 53% in the social
performance model and 50% in double bottom performance model are operating under
non-constant returns to scale.
The analysis of TE scores obtained from four DEA models specified shows the superiority of
double bottom efficiency model. Indeed, the average TE (CRS) is estimated at 0.79 in the
double bottom model, against 0.68, 0.68 and 0.43 respectively in social performance model,
overall financial performance model and saving and credit provision model. Furthermore,
the average TE (VRS) was 0.91 in the double bottom model, 0.75 in the overall financial
performance model, 0.89 in the social performance model and 0.65 in the saving and credit
provision model. The same trend is observed for SE. This suggests that, contrary to what has
been often affirmed in conferences and in the existing literature, which present the social
objectives and financial sustainability as conflicting, the findings from this study show that
there is a good balance between both objectives. Indeed, many MFIs reach the maximum of
the poor population (women in this case) and achieve at the same time financial
sustainability, measured in this study by ROA and OSS. In other words, the optimization of
overall financial performance resulting from introduction of the commercial framework into
MFIs does not always compromise their social responsibility. This finding is in line with our
hypothesis that there is good balance between outreach and financial performance, and
Christen et al. (1995), Otero and Rhyne (1994) and Meyer (2002), who reported that
outreach and financial sustainability are complementary since an increase in the number of
clients creates economies of scale and therefore reduce costs and help MFIs to reach
financial sustainability. However, this is not in line with Hulme and Mosley (1996), who
reported an inverse relationship between outreach and financial sustainability of MFIs.
- 107 -
By comparing the estimated TE scores to SE scores, it is observed that for all DEA models
specified, except the social performance model, the average technical efficiency score (VRS)
is higher than the scale efficiency score. Indeed, 0.65 of the TE score is higher than 0.62 of
the SE score under the saving and credit production model; 0.89 of the technical efficiency
score (VRS) is higher than 0.76 of the scale efficiency score under financial performance
model; and 0.91 of the technical efficiency (VRS) is higher than 0.86 of the scale efficiency
under double bottom production model. These results suggest that the inefficiency observed
is more scale inefficiency (resulting from the size and level of operation) than technical
inefficiency (VRS).
One possible explanation for this could be the fact that local MFIs, which constitute the
majority of microfinance organizations operating in the DRC, rely exclusively on micro-saving
and do not have access to external financial resources. The lack of appropriate and long term
resources, as well as the low level of working capital, does not allow them to capture the
scale economy. Yet, the highest level of SE, compared to TE (VRS) in the social production
model, suggests that the inefficiency is more technical inefficiency (VRS) resulting from
misallocation of resources rather than scale inefficiency due to the size of institution.
However, regarding SE and TE (CRS) in all models specified, the results show that the
inefficiency is more technical inefficiency than scale inefficiency. The above results have very
strong policy implications for the improvement of microfinance performance operations in
the DRC. The policy to enhance the performance of MFIs requires the introduction of best
practices, best methods of management, the development of new lending technologies and
the improvement of human resource quality, as well as their integration in the financial
system and private capital.
The results from Table 4 suggest that the efficiency scores of MFIs operating in the
agricultural sector are not significantly different from those without the provision of
agricultural credit. The standard ANOVA and the Mann-Whitney statistical tests are both
non-significant. Agricultural credit is not associated with the performance and efficiency of
MFIs operating in the DRC. Thus, MFIs could finance farming activity without negatively
affect their sustainability. This conforms to the hypothesis that there is no substantial
difference in efficiency distribution between MFIs with or without the provision of
agricultural microcredit. This also confirms what has been reported in Tollens (2004) and
- 108 -
Kodjo et al. (2003). The finding is consistent with the result found in Pakistan by Tanvir and
Zulfiquar (2007), who reported a non-significant decline in the efficiency of commercial
banks after the start of intensive agricultural lending.
The observed non-significance in mean efficiency score between MFIs with agricultural
lending and without agricultural lending could be explained by many factors, including the
methodology and technological innovations used in the microfinance. Indeed, the provision
of microfinance services to rural and agricultural households, mainly credit, is based on the
personal relationships that MFIs maintain with their customers. These relationships
consolidated and encouraged by geographical and cultural proximity. Furthermore, the
incentives used in microfinance, such as social capital (social pressure) and the subordination
of new credit line, as well as the geographical proximity, allows MFIs to face the risk of anti-
selection and moral hazard, and therefore to reduce transaction costs associated with the
operation of loans. They also allow MFIs to lower the screening and enforcement costs.
Many MFIs operating in the DRC, especially those which have financial operations in rural
areas have been created under the cooperative and mutual forms. Following Ledgerwood
(1999), in this kind of local financial organization, the members are at once owners and
customers. The double qualification of members generates the convergence of interest and
therefore reduces the agency problem.
In contrast to the traditional financial institutions (banks), MFIs, and particularly those which
have financial operations in rural areas, realize an economy of costs due to their lower levels
of fixed and administrative costs. Moreover, the credit policy, especially the orientation and
diversification of the loan portfolio allows MFIs to face covariant risks and low profitability in
agricultural activities. Indeed, the loan portfolios of most rural MFIs are more diversified and
directed at the small non-farm economy, with less risk and relatively high profitability
compared to agricultural activities. Furthermore, the choice of the borrowers is made in
many cases on the basis of human capital, physical and financial endowment, which reduces
the credit risk.
- 109 -
5.4 Conclusions
This study aims to measure the technical and scale efficiency of MFIs operating in the DRC,
subject to the provision or no provision of microcredit towards farm-households. The sample
is extracted from RIFIDEC and Mixmarket dataset and consists of 36 microfinance
organizations with available data. The majority of these organizations are members of
RIFIDEC. Using production orientation, the DEA is used to calculate efficiency scores under
both constant returns to scale and VRS assumptions. Four DEA empirical models are
specified (social performance model, overall financial performance model, provision of micro
saving and microcredit model and double bottom model). The mean efficiency scores of
MFIs working and not working with farmers are compared using standard ANOVA and Mann-
Whitney statistics to test whether the financing of agricultural activities results in negative
effects on microfinance performance and efficiency.
The analysis shows that many of the MFIs operating in the DRC reach double bottom
technical efficiency line. This suggests that a good balance between social and financial
performance of microfinance. As the majority of MFIs are operating in the zone of increasing
efficiency scale and under variable returns to scale assumption, the introduction of best
practices of management and improvement of human capital will increase their efficiency.
Furthermore, an increase in MFIs assets and operations would result in the improvement of
their efficiency. The study has also found that financing of farming activities is not
significantly associated with the social and financial efficiency of MFIs.
App. Table 5.1 Distribution of efficiency of MFIs
Niveau TE (CRS) TE (VRS) (SE)

d’efficacité
Frequency Percentage Frequency Percentage Frequency Percentage
MODEL SL-AP
0.00-0.25 11 30.6 2 5.6 6 16.7
0.26-0.50 13 36.1 11 30.6 7 19.4
0.51-0.75 6 16.7 10 27.8 9 25.0
0.76-0.99 2 5.6 4 11.1 10 27.8
- 110 -
1.00 4 11.1 9 25.0 4 11.1
Total 36 100 36 100 36 100
Mean 0.43 0.65 0.62
SD 0.29 0.27 0.30
Maximum 1.00 1.00 1.00
Minimum 0.02 0.11 0.10
MODEL RO-AP
0.00-0.25 1 3.4 - - 1 3.4
0.26-0.50 7 24.1 - - 5 17.2
0.51-0.75 9 31.0 4 13.8 6 20.7
0.76-0.99 4 13.8 14 48.3 8 27.6
1.00 8 27.6 11 37.9 9 31.0
Total 29 100 29 100 29 100
Mean 0.68 0.89 0.76
SD 0.26 0.12 0.25
Maximum 1.00 1.00 1.00
Minimum 0.18 0.63 0.24
MODEL WnLnS-AP
0.00-0.25 5 14.4 3 8.6 -
0.26-0.50 7 20.0 6 17.1 3 8.6
0.51-0.75 4 11.4 6 17.1 1 2.9
0.76-0.99 8 22.9 5 14.3 20 57.1
1.00- 11 31.4 15 42.9 11 31.4
Total 35 100 35 100 35 100
Mean 0.68 0.75 0.91
SD 0.31 0.29 0.16
Maximum 1.00 1.00 1.00
Minimum 0.11 0.13 0.16
MODEL WRO-AP
- 111 -
0.00-0.25 - - - - - -
0.26-0.50 3 10.7 - - 2 7.1
0.51-0.75 10 35.7 3 10.7 6 21.4
0.76-0.99 4 14.3 11 39.3 9 32.1
1.00- 11 39.3 14 50.0 11 39.3
Total 28 100 35 100 28 100
Mean 0.79 0.91 0.86
SD 0.22 0.11 0.19
Maximum 1.00 1.00 1.00
Minimum 0.27 0.64 0.36
App. Table 5.2 Efficiency scores of MFIs with versus without agricultural credit provision
Mean SD Min Max,
Model SL-AP
TE (CRS)
With agricultural credit provision (15) 0.45 0.33 0.04 1.00
Without agricultural credit provision (21) 0.40 0.26 0.02 1.00
Total (36) 0.42 0.29 0.02 1.00
TE ( VRS)
Total (36) 0.65 0.27 0.11 1.00
SE
Total (36) 0.62 0.29 0.10 1.00
Model RO-AP
TE (CRS)
- 112 -
Total (29) 0.68 0.25 0.17 1.00
TE (VRS)
Total (29) 0.89 0.11 0.63 1.00
SE
Total (29) 0.76 0.25 0.24 1.00
Model WnLnS-AP
TE ( CRS)
Total (35) 0.68 0.30 0.10 1.00
TE ( VRS)
Total (35) 0.74 0.29 0.01 1.00
SE
Total 0.90 0.16 0.39 1.00
Model WRO-AP
TE (CRS)
- 113 -
Total (28) 0.79 0.21 0.26 1.00
TE ( VRS)
Total (28) 0.91 0.11 0.64 1.00
SE
Total (28) 0.86 0.19 0.36 1.00
- 114 -
Chapter 6
Summary and Implications
6.1 Summary
This dissertation explores the major constraints on the provision of microcredit towards
farm-households and the keys drivers affecting the probability that a farm-household is
credit-constrained. In addition, this study measures the impact of credit constraints on farm-
household production efficiency and welfare. Furthermore, this study addresses the issue of
measuring the efficiency of MFIs, and tests whether there is a substantial difference in
efficiency score distribution between MFIs with and without the provision of agricultural
credit. The data are cross-sectional and were collected from farm-households in the
hinterland of Kinshasa and MFIs surveys. The data from households are collected in villages
located in the area between Menkao, Dumi and Mbankana, while data from MFIs are
collected from RIFIDEC and Mixmarket. The results of analysis are reported in four chapters
with specific objectives and analysis. Chapters 2, 3 and 4 focus on the first general objective
of this study, which consists of understanding credit constraints and their consequences on
farm-households of the hinterland of Kinshasa. While Chapter 5 treats the second general
objective, which is related to the measurement of performance of MFIs with and without the
provision of agricultural microcredit.
Chapter 2 proposes an empirical approach to the measurement of credit based on the direct
approach, and explores the major difficulties to financing smallholder farm-households by
the strategy of MFIs. Using probit and logit models, this study identifies household and
institutional characteristics affecting the probability of being credit-constrained. The findings
from the responses of households reveal that low returns and high systematic risks related
to the agricultural activities are among factors constraining farm-household’s participation in
microcredit programs. This confirms hypothesis that the nature of agricultural production
constrains farm-household participation in credit market. This conforms to some previous
studies (Wampfler, 2000; Lesaffre and Pesche, 2002).
- 115 -
The results from the response of households also reveal that households are partially or
totally rationed because of insufficient resource of MFIs. This may be due to the fact that
MFIs working in the hinterland of Kinshasa are not subsided or linked to financial markets.
Thus, the hypothesis that limited financial resources of MFIs reduce their capacity to
respond to the magnitude of farm-household demand for credit is accepted. Furthermore,
households from sample point out short loan terms, frequency of repayment non-adapted
to farm-household’s cash flows, lack of collateral requirements, high interest rates and small
loan size provided by MFIs as among the key factors explaining a household’s decision not
apply for a loan. This confirms the hypothesis that lending technology applied in the MFIs
constrains farm-households from joining microcredit programs. Similar conclusion is
reported in Atieno (2001), who reveals that lending terms and conditions significantly
determine the access to microcredit. This is also in line with some previous studies, which
argue that the difficulty to grant credit towards farm-households is due to lending
technology and policy, particularly the high interest rates (Yehuala, 2008; Lepanu, 2001;
Evans et al. 1999). However, these findings challenge the conventional wisdom that what is
matter for poor households is the access to credit rather than the related costs
(Ledgerwood, 1999; Woller et al. 1999).
The results from logit and probit models show that household network participation
significantly decreases the probability of being credit-constrained. This might be explained
by the fact that social networks facilitate the flow of information, which allows a reduction
of transaction costs for both borrower and lender (Okten and Osili, 2004). In addition, given
the lack of physical collaterals and group lending technology applied by some MFIs, social
network participation has to play an important role in collateralizing loans. The results also
reveal that landholding significantly reduces the probability that a farm-household is credit-
constrained. A similar conclusion is found in Rahji and Adeoti (2010) and Kochar (1997). The
results of analysis also show that remittances have a negative and significant effect on the
propensity of being credit-constrained. This might be explained by the fact that resources
from remittances decrease household liquidity constraints, helping it to cope with income
shocks and increasing its willingness to agree credit contract terms. The findings also reveal
that household size increases the probability that a farm-household is credit-constrained.
- 116 -
This finding is in line with Nuryartono et al. (2005), Oyedele et al. (2009) and Rahji and
Adeoti (2010).
Looking at the covariates significantly affecting the probability of being credit-constrained, it

is clear that farm-household capital endowments (physical and non-physical) are important
in explaining the probability of being credit-constrained. These capital endowments provide
information on risk-bearing and repayment capacity of households, and play an important
role in determining household credit constraints status. Thus, the hypothesis that household
capital endowments affect the probability that a farm-household is credit-constrained is
accepted. Given that many households from the study area are very poor and lack these
capital endowments to secure loans, this finding might be convincing evidence that MFIs
from hinterland of Kinshasa are not clearly pro-poor. This finding challenges the assumption
shared in the dominant literature that MFIs help to correct the imperfections of credit
markets by increasing the access to credit of poor households, particularly farm-households
(Kodjo et al. 2003; Navajas et al. 2000). Furthermore, this conclusion contrasts with
outcomes of large microfinance programs developed by NGOs under the welfarist
framework, such as those implemented in Bangladesh, which effectively target the poorest
of the poor and have a higher rates of outreach (Khandker, 2005; Pitt and Khandker, 1998).
This may be unique in the case of DRC and possibly in other developing nations, where the
microfinance sector is dominated by local and small size MFIs working under commercial
logical framework and facing greater liquidity constraints.
The results of logit and probit models show no strong effect of economic dependency ratio
on the probability that a farm-household is credit-constrained, while Dong et al. (2010)
reported a significant effect. As in Nuryartono et al. (2005), this study shows that the level of
education of household head is negatively, but non-statistically and significantly associated
with the probability of being credit-constrained. The effect of age of household head is
negative, but is not statistically significant. This is in line with Baydas et al. (1994), Zagarra et
al. (2008) and Rahji and Adeoti (2010), but does not conform with Nuryartono et al. (2005),
Omonona et al. (2008) and Dong et al. (2010), who report positive and significant effects.
The Household head male has a positive effect, but non-statistically significant, implying that
gender of household does not account for targeting borrowers. The villages of residence
used to capture the effect of local institutional characteristics, such as market, extension
- 117 -
service, road, are not important in explaining the probability that a farm-household is credit-
constrained. Therefore, one can reject the hypothesis that local institutional characteristics
affect the probability that a farm-household is credit-constrained.
Chapter 3 evaluates the impact of credit constraints on the technical and the SE of farm-
households of the hinterland of Kinshasa. It also identifies the key drivers explaining
differences in TE score distribution. The DEA is applied to compute technical and SE scores.
Then, the PSM is applied to compute the impact of credit constraints on farm-household
production efficiency. In addition, the TR model is computed to explore the keys factors
responsible for the variations in TE scores among cassava farm-households. The findings
from the DEA reveal high levels of cassava production inefficiency among farm-households
in the sample. The average SE score is three times as big as the TE score under both CRS and
VRS assumptions. The results from DEA also show a substantial difference in the TE of
cassava producing households. These findings are not in line with previous empirical studies
from Nigeria, which report high levels of TE scores and small variances among farm-
households producing cassava (Edeh and Awoke, 2009; Ogboma et al. 2007; Adeyemo et al.
2010; Iheke, 2008; Onu and Edon, 2009; Ogundari and Ojo, 2007).
The results from TR model show that landholding property is an important factor increasing
TE score of farm-households. This supports previous studies from Sub-Saharan African that
have called for land tenure reform to improve agricultural productivity (Feder and Nishio,
1999; Besley, 1995a; Gavian and Fafchamps, 1996; Jacoby and Minten, 2007). The results
from TR also show that the level of formal education of the household head increases the TE
of farm-households. Similar findings are reported in some previous empirical studies, such in
Bravo-Ureta and Pinheiro (1993), Ali and Flinn (1989), Battese et al. (1996). Farm size tends
to be positively and significantly associated with TE score distribution. This is consistent with
empirical studies, which have reported a positive relationship between farm size and
agricultural efficiency (Bhandari, 2006; Deolalikar, 1981; Fan and Chan-Kang, 2005).
Household network participation is also found to be an important factor increasing TE
scores. A similar conclusion is given in the Idiong (2007), who reported positive effect of
farm-household participation in association on their TE scores. The technology adoption is
one of the mechanisms through which social networks effect production decision. Thus, the
results from the TR model confirm the hypothesis that household capital endowments affect
- 118 -
TE scores of farm-households. However, the results of TR model show that the village of
residence, used to capture the effect of local institutions characteristic, is non-statistically
significant. This rejects the hypothesis that local institutional characteristics affect TE of
farm-households.
The results from PSM show that credit constraints have a negative effect on TE. This
confirms hypothesis that credit constraints reduce the TE of farm-households; and suggests
that the removing of credit constraints among farm-households will substantially increase TE
score of farm-households. This is consistent with previous empirical studies that have
reported a reducing effect of credit constraints on the TE of farmers (0monona et al. 2008;
Komicha and Öhlmer, 2006; Färe et al. 1990). However, the results PSM also show that
credit constraints do not statistically and significantly affect the SE of farm-households in the
hinterland of Kinshasa. The effect of credit constraints on SE is found to be negative as
expected, but the magnitude of causal inference remains non-statistically significant. Thus,
the hypothesis that credit constraints reduce the SE is rejected. This suggests that that
improving farm-household access to microcredit will not improve substantially the SE of
farm-households.
Chapter 4 evaluates the effects of credit constraints on farm-household welfare. The PSM is
applied for impact assignment. The findings reveal that, on average, credit constraints
significantly reduce household food consumption, total expenditures and welfare index
(1USD and 2USD). This confirms the hypothesis that credit constraints decrease household
economic welfare. Similar results were found by Li and Zhu (2010) and Baiyegunhi et al.
(2010), who report a negative effect of credit constraints on household welfare. These
findings are also in line with the body of previous empirical studies that have demonstrated
a positive effect of microfinance on household consumption and income (Negash, 2008;
Arun et al. 2006; Aroca and Hewings, 2009; Wright, 2000; Khandker, 2001; Khandker and
Faruque, 2003; Pitt and Khandker, 1998; Khandker, 2003). The results suggest that removing
credit constraints among farm-household, for example through the strategy of microfinance,
will substantially improve the economic welfare of households, and thus will have a reducing
effect on rural poverty, by improving food consumption. However, the effect of credit
constraints on non-food consumption is found to be negative, but not statistically significant.
Similar results were found in (Negash, 2008), who report insignificant effects of access to
- 119 -
microfinance on non-food consumption. This insignificant effect of credit constraints on non-
food consumption may possibly be due to small loan size and lack of good rural markets and
infrastructure.
Chapter 5 estimates the efficiency of microfinance organizations operating in the DRC, and
tests whether there is a substantial difference in efficiency distribution between MFIs with
and without the provision of agricultural credit. The DEA is used to compute technical and SE
sores of MFIs from the sample. The findings from DEA reveal that the TE scores are higher
than SE scores (VRC) for all DEA models specified, except social performance DEA model.
This suggests that the inefficiency in MFIs from the DRC is more scale inefficiency than
technical inefficiency. However, by comparing the SE score and TE (CRS), it appears that the
inefficiency is more technical inefficiency than scale inefficiency. Furthermore, the results
from all DEA models specified show that many MFIs from the sample are working on the
frontier of double bottom efficiency line. They also show a higher level of average TE in
double bottom efficiency DEA model. This finding confirms hypothesis that there is a good
balance between outreach and financial sustainability of microfinance, and challenges what
is often found stated in the mainstream literature: that social performance and financial
sustainability are conflicting objectives (DeBriey, 2003; 2005; Ledgerwood, 1999; Woller et
al. 1999). However, this finding is not in line with Tariq and Mohd (2010), who show that
there is no trade-off between efficiency and outreach in case of sample of MFIs from India.
In addition, the results reveal that TE score distributions do not substantially differ between
MFIs with and without the provision of agricultural credit. This supports the hypothesis that
there is non-substantial difference in efficiency distribution between MFIs with and without
the provision of agricultural microcredit.
6.2 Policy implications and recommendations for further research
The findings of this dissertation yield some policy implications and recommendations for
further research. The results establish the existence of a high prevalence of credit
constraints among farm-households. They attest that constrained households are less
efficient and have low levels of consumption compared to unconstrained households. This
- 120 -
calls for interventions to remove credit constraints and to improve production efficiency as
well as household economic welfare. However, the effects of credit constraints on the non-
food consumption and SE of farm-households are not found to be statistically significant.
Thus, further research based on large sample size and panel data may be needed to test the
cause inference between credit constraints, SE and non-food consumption.
Household responses to credit market participation point out the limited financial resources
at the level of MFIs among the main constraints on the financing of farm-households.
Furthermore, the analysis of the performance of MFIs in DRC reveals that most of them are
operating under the scale inefficient zone. These findings call for the strengthening of
financial capacity of MFIs and assets. They imply, among other things, new ways to think
about the role of the public sector, donors and traditional financial institutions in financing
the small scale and local MFIs. There is also need to think how small scale MFIs could be
integrated into the financial private capital.
Household’s responses show that lending technologies (short loan maturity and repayment
frequency, high interest rates, lack of collateral, small size of loan) applied by MFIs are not
adapted to the needs and specificities of agricultural activities as well as the economic
profile of farm-households. Therefore, there is a need to apply the appropriate terms and
conditions to loan contracts. The provision of credit towards farm-households needs to be
done with respect to their cash flows. The repayment frequency and loan maturity should be
adapted to farm-household cash flows. The collateral should be diversified and extended to
non-physical assets, such as land property and social pressure. There is need for MFIs to set
up lending technology and innovations to provide adequate medium and long term credit
adapted to the specific needs of agriculture. There also is need of further research,
particularly on financial product, on terms and conditions of loan contract, which correspond
to the needs of farm-households.
The constrained households report that the profitability of agricultural activities and
associated high risk are among the key factors of a household’s decision to not apply for a
loan. This calls for public policy and investment to increase farmer’s access to new
technology and innovations, such as HYV, which will improve agricultural profitability. MFIs,
especially NGOs with micro credit program, have to play an important role by providing high
- 121 -
yielding input credit and various mechanisms, such as insurance, to cope with risks.
Furthermore, public investment with support of donors might create funds of risk-guarantee
to cope with systemic risk. This could be a strong signal in the direction of MFIs, and could
change their attitude vis-à-vis agricultural activities.
The household participation in social and economic organizations is found to be an

important determinant of household credit constraints conditions. It is more likely to reduce
significantly the probability of a household being credit-constrained, and is an important
factor increasing TE of farm-households. Thus, interventions to promote social networks will
reduce credit constraints and enhance the production efficiency of farm-households. The
findings reveal that landholding property is negatively and significantly associated with being
credit-constrained. Furthermore, landholding property is found to be correlated with TE.
Thus, agricultural public policy to improve farm-household access to landholding property
will reduce credit constraints among farmers, by lowering risk collateral, and will enhance
the production efficiency of farm-households.
Public investment in human capital is likely to improve agricultural efficiency since the
education level of the household head is found to be positively correlated with TE of farm-
households. Given that the education of household is considered as a proxy for
management, information and good decision-making, the public investment in the extension
services will contribute to the improvement of production efficiency. The findings reveal that
household size is strongly and positively associated with credit conditions. Therefore, public
family planning policy to limit household size is needed, by including in the extension
programs, a package of actions to inform farm-households about the importance of small
household size and by promoting the use of contraceptives. These measures will, among
other things, reduce liquidity constraints and therefore credit constraints.
The research findings of this study reveal that MFIs working in the hinterland of Kinshasa are
not pro-poor. Household capital endowments play an important in determining the
probability that a farm-household is credit-constrained. It would be important to go deeper,
by testing how the profile and performance of MFIs affect the probability that a farm-
household is rationed. Thus, further research may be needed to examine how institutional
- 122 -
design and the logical framework affect the outreach of MFIs and the poor’s access to
microcredit.
The research findings confirm that there is a good balance between financial sustainability
and outreach of microfinance. This finding might be viewed as an empirical reference for
further scientific research on the trade-off between social responsibility and the
sustainability of microfinance. However, this study is not able to explain why some MFIs are
efficient and other not: further research may be needed to explore the keys drivers affecting
the efficiency of microfinance.
The findings of this research study show that the provision of microfinance towards farm-
households does not have a statistically significant effect on the efficiency of MFIs. This can
be used as a point of reference for further investigations on the impact of agricultural credit
on the financial sustainability of MFIs. Since the sample size of MFIs used in this study is
small, and data are cross-sectional, there is need for a better understanding of this causal
relationship in the dynamic and multivariate perspective, using for example the panel data
from large geographical areas.
- 123 -
References
Adams, W., 2002. Microfinance et financement de l’agriculture: mariage difficile. Dialogue

29: 5–11.
Adeyemo, R., Oke, J.T.O. and Akinola, A.A., 2010. Economic efficiency of small scale farmers
in Ogun State, Nigeria. Tropicultura 28(2): 84–88.
Albert, L., 1997. Les mécanismes collectifs de crédit: une nouvelle situation miracle? Cahier
de la Faculté de Sciences Economiques de Namur 195. Namur: FNDP.
Ali, M. and Flinn, J.C., 1989. Profit efficiency among Basmati rice producers in Pakistan
Punjab. American Journal of Agricultural Economics 71: 303–310.
Aliango, M., 2004. Le taux de pénétration des Institutions de Microfinance congolaises.
Mémoire de licence, Département d’Administration des Affaires, Université
Protestante au Congo, R.D. Congo.
Armendariz, B. and Morduch, J., 2007. The Economics of Microfinance. MIT Press,
Cambridge.
Aroca, P. and Hewings, G.J.D., 2009. Microcredit impact assessment: The Brazilian and
Chilean cases. Panoroma Socioeconόmico 27(39): 100–112.
Arun, T., Imai, K. and Sinha, F., 2006. Does the microfinance reduce poverty in India?
Propensity score matching based on national-level household data.
Development Economics and Public Policy, Paper Series 17. University of
Manchester.
Aryeetey, E. and Udry, C., 1997. The characteristics of informal financial markets in Sub-
Saharan Africa. Journal of African Economies 6(1): 161–203.
Athanassopoulos, A.D. and Ballantine, J.A., 1995. Ratio and frontier analysis for assessing
corporate performance: evidence from the grocery industry in the UK. Journal
of the Operational Research Society 46(4): 427–440.
Atieno, R., 2001. Formal and informal institutions’ lending policies and access to credit by
small-scale enterprises in Kenya: An empirical assessment. African Economic
Research Consortium, Research Paper 111. University of Nairobi.
Atkinson, A.B., Rainwater, L. and Smeeding, T.M., 1995. Income distribution in OECD
countries: Evidence from the Luxembourg income study. Social Policy Studies
18. Paris: OECD.
- 124 -
Baiyegunhi, L.J.S., Fraser, G.C.G. and Darroch, M.A.G., 2010. Credit constraints and
household welfare in the Eastern Cape Province, South Africa. African Journal
of Agricultural Research 5(16): 2245–2252.
Bali Swain, R., 2002. Credit rationing in rural India. Journal of Economic Development 27 (2):
1–21.
Bambila, M.H., 2009. Microfinance et réduction de la pauvreté en R.D.C. Quel rôle pour la
coopération au développement? Mémoire de D.E.S, Département des Sciences
Sociales, Université Libre de Bruxelles, Belgique.
Banque Centrale du Congo, 2001. Rapport annuel, Kinshasa, R.D. Congo.
Barrett, C.B., 1996. On price risk and the inverse farm size productivity relationship. Journal
of Development Economics 51(2): 193–216.
Battese, G.E. and Coëlli, T., 1995. A model for technical inefficiency effects in a stochastic
frontier production function for panel data. Empirical Economics 20: 325–332.
Battese, G.E., Malik, S.J. and Gill, M.A., 1996. An investigation of technical inefficiencies of
production of wheat farmers in four districts of Pakistan. Journal of Agricultural
Economics 47: 37–49.
Bauer, P.W., Berger, A.N., Ferrier, G.D. and Humphrey, D.B., 1998. Consistency conditions for
regulatory analysis of financial institutions: A comparison of frontier efficiency
methods. Journal of Economics and Business 50 (2): 85–114.
Baydas, M.M., Richard, M.L. and Aguilera-Alfred, N., 1994. Credit rationing in small-scale
enterprises: Special microenterprise programs in Ecuador. Journal of
Development Studies 31(2): 279–209.
Becker, S.O. and Ichino, A., 2002. Estimation of average treatment effects based on
propensity scores. Stata Journal 2 (4): 358–377.
Benhabib, J., 2000. The role of financial development in growth and investment. Journal of
Economic Growth 5(2): 341–360.
Berg, A., Forsund, A., Hjalmarsson, F. R. L. and Suominen, M., 1993. Banking Efficiency in
the Nordic Countries, Journal of Banking and Finance17 (23): 371–
Berger, A.N. and Humphrey, D.B., 1997. Efficiency of financial institutions: International
survey and directions for future research. European Journal of Operational
Research 98: 175–212.
- 125 -
Berry, R.A. and Cline, W., 1979. Agrarian Structure and Productivity in Developing Countries.
Johns Hopkins University Press, Baltimore.
Besley, T., 1995a. Property rights and investment incentives: Theory and evidence from
Ghana. Journal of Political Economy 103(5): 903–937.
Besley, T., 1995b. Non-market institutions for credit and risk sharing in low-income
countries. Journal of Economic Perspectives 9(3): 115–127.
Bhandari, R., 2006. Searching for a weapon of mass production in Nepal: Can market-
assisted land reforms live up to their promise? Journal of Developing Societies
22: 111–143.
Binswanger, H.P. and Deininger, K., 1997. Explaining agricultural and agrarian policies in
developing countries. Journal of Economic Literature 35(4): 1958–2005.
Blancard, S., Boussemart, J.P., Briec, W. and Kerstens, K., 2006. Short-and long-run credit
constraints in French agriculture: A directional distance function framework
using expenditure-constrained profit function. American Journal of Agricultural
Economics 88(2): 351–364.
Boltanski, L. and Thevenot, L., 1987. Les économies de la grandeur. Cahiers du centre
d’études de l’emploi 31. Paris: Presse Universitaire de France.
Boltanski, L. and Thevenot, L., 1991. De la Justification: Les économies de la grandeur.
Gallimard, Paris.
Bondobondo, G., 2002. Etudes des lieux du secteur de la microfinance en R.D. Congo.
Kinshasa: RIFIDC.
Boucher, S. and Carter, M.R., 2002. Risk rationing and activity choice: Moral hazard
constrained credit markets. Agricultural and Applied Economics, Staff Paper
445. University of Wisconsin-Madison.
Boucher, S., 2002. Endowments and credit market performance: An econometric exploration
of non-price rationing mechanisms in rural credit markets in Peru. Ph.D. Thesis,
Department of Agricultural and Applied Economics, University of Wisconsin,
USA.
Boucher, S.C., Guirkinger, C. and Trivelli, C., 2005. Direct elicitation of credit constraints:
Conceptual and practical issues with an empirical application to Peruvian
agriculture. American Agricultural Economics Association Annual Meeting,
Providence-Rhode Island, 24 – 27 July.
- 126 -
Bravo-Ureta, B.E. and Pinheiro, A.E., 1993. Efficiency analysis of developing country
agriculture: A review of the frontier function literature. Agricultural Resource
Economics Review 28: 88–101.
Broussard, J.M., 1987. Economie de l’Agriculture. Economica, Paris.
Bultot, F., 1950. Cartes des régions climatiques du Congo belge établies d'après les critères
de KOPPEN. Communication du bureau climatologique 2. Bruxelles: INEAC.
Calkins, P., Lariviere, E.S., Martin, F. and Nteziyaremye, A., 1992. Facteurs de succès et
d’échec du crédit agricole dans les pays économiquement moins développés.
Développement International Desjardins, Quebec.
Carter, M. and Olinto, P., 2003. Getting institutions right for whom? Credit constraints and
the impact of property rights on the quantity and composition of investment.
American Journal of Agricultural Economics 85(1): 173–186.
Charnes Α., Cooper, W. and Rhodes, E., 1978. Measuring the efficiency of decision making
units. European Journal of Operational Research 2(6): 429–444.
Chen, M. A. and Dunn, E., 1996. Household Economic Portfolios. AIMS Project Report,
USAID/G/EG/MD. Washington, D.C.: Management Systems International.
Christen P., Rhyne E., Vogel R. C. and McKean C. 1995. Maximizing the outreach of micro
enterprise finance: An analysis of successful micro finance programs. Program
and Operations Assessment Report. USAID, Washington DC.
Coase, R.H., 1937. The nature of the firm. Economica 4(16): 386–405.
Coëlli, T.J., 1996. A guide to DEAP version 2.1: A Data Envelopment Analysis (Computer
program). Working Papers 8/96. Brisbane: CEPA.
Coëlli, T.J., Rao, D.S.P., O’Donnel, C.J. and Battese, G.E., 2005. Introduction to Efficiency and
Productivity. Springer, New York.
Coleman, B.E., 1999. The impact of group lending in Northeast Thailand. Journal of
Development Economics 60(1): 105–141.
Coleman, J.S., 2002. Social capital and poverty: A microeconomic perspective. In Thierry Van
Bastelaer (ed.), The Role of Social Capital in Development. Cambridge
University Press, Cambridge: 19–41.
Conning, J., 1999. Outreach, sustainability and leverage in monitored and peer-monitored
lending. Journal of Development Economics 60: 51–77.
- 127 -
Cook, W.D., Hababou, M. and Tuenter, H., 2000. Multicomponent efficiency measurement
and shared inputs in Data Envelopment Analysis: an application to sales and
service performance in bank branches. Journal of Productivity Analysis 14 (3):
209–224.
Cornée, S., 2006. Analyse de la convergence entre performances financières et
performances sociales: Application de la méthode Data Envelopment Analysis
sur 18 Institutions de Microfinance Péruviennes. Mémoire de Master,
Département des Sciences de gestion, Université de Rennes 1, France.
Crabbe, M., 1980. Le Climat de Kinshasa. AGCD, Bruxelles.
Croppenstedt, A., Demeke, M. and Meschi, M.M., 2003. Technology adoption in the
presence of constraints: the Case of fertilizer demand in Ethiopia. Review of
Development Economics 7(1): 58–70.
Cull R., Demirguc-Kunt, A. and Morduch, J., 2007. Financial performance and outreach: A
global analysis of lending microbanks. Economic Journal 117: 107–133.
De Saint Moulin, L. and Kalombo, T., 2005. Atlas de l’Organisation Administrative de la RDC.
CEPAS, Kinshasa.
Deaton, A., 1992. Household saving in LDCs: Credit markets, insurance and welfare.
Scandinavian Journal of Economics 94(2): 253–273.
DeBriey, V., 2003. Elaboration d’un cadre d’évaluation de performance des institutions de
microfinance. Thèse de doctorat, Institut d’Administration et de Gestion,
Université Catholique de Louvain, Belgique.
DeBriey, V., 2005. Plein feu sur la microfinance en 2005. Regards économiques 28. Louvain-
la-Neuve: IRES.
Dehejia, R.H. and Wahba, S., 2002. Propensity score matching methods for non-
experimental causal studies. The Review of Economics and Statistics 84 (1):
151–161.
Dekker, D. and Post, T., 2001. A quasi-concave DEA model with an application for bank
branch performance evaluation. European Journal of Operational Research
13(2): 296–311.
Deolalikar, A.B., 1981. The inverse relationship between productivity and farm size: a test
using regional data from India. American Journal of Agricultural Economics 63:
275–279.
- 128 -
Dercon, S., 1996. Risk, crop choice and savings: evidence from Tanzania. Economic
Development and Cultural Change 44(3): 485–513.
Diagne, A. and Zeller, M., 2001. Access to credit and its impact on welfare in Malawi.
Research Report 116. Washington, DC: IFPRI.
Diagne, A., 1999. Determinants of household access to and participation in formal and
informal credit markets in Malawi. Discussion Paper 67. Washington DC: IFPRI.
Diagne, A., Zeller, M. and Sharma, M., 2001. Empirical measurement of households’ access
to credit and credit constraints in developing countries: Methodological issues
and evidence. Discussion Paper 90. Washington DC: IFPRI.
Dong, F., Lu, J. and Allen M.F., 2010. Effects of credit constraints on productivity and rural
household income in China. Working Paper 10-WP 516. Center for Agricultural
and Rural Development, Iowa State University.
Dorfman, J.H. and Koop, G., 2005. Current developments in productivity and efficiency
measurement. Journal of Econometrics 126: 233–240.
Dunford, C., 2000. The holy grail of microfinance: Helping the poor and sustainable? Small
Enterprise Development 1: 40–44.
Duvernay, E.F., 1989. Conventions de qualité et formes de coordination. Revue économique
40(2): 329–359.
Edeh, H.O. and Awoke, M.U., 2009. Technical efficiency analysis of improved cassava farmers
in Abakaliki local government area of Ebonyi State: A stochastic frontier
approach. Agricultural Journal 4(4): 171–174.
Ellis F., 1988. Peasant Economics: Farm Households and Agrarian Development. Cambridge
University Press, Cambridge.
Emrouznejad A. and Podinovski V., 2004. Data envelopment analysis and performance
management. International Symposium of DEA, Aston Business School, Aston
University UK, 5 – 6 September.
Enete, A., 2003. Resource use, marketing and diversification decisions in cassava producing
households of Sub-Saharan Africa. Ph.D. Thesis, Division of Agricultural and
Food Economics, Katholieke Universiteit Leuven, Belgium.
Eswaran, M. and Kotwal, A., 1986. Access to capital and agrarian production organization.
Economic Journal 96: 482–498.
- 129 -
Evans, T.G., Adams, A.M., Rafi, M. and Norris, A.H., 1999. Demystifying nonparticipation in
microcredit: A population-based analysis. World Development 27(2): 419–430.
Fama, E. and Jensen, M., 1983. Agency problems and residual claims. Journal of Laws and
Fan, S. and Chan-Kang, C., 2005. Is small beautiful? Farm size, productivity, and poverty in
Asian agriculture. Agricultural Economics 32(1): 135–146.
Färe, R., Grosskopf, S. and Lee, S., 1990. A nonparametric approach to expenditure
constrained profit maximization. American Journal of Agricultural Economics
72(2): 574–81.
Farrell, M.J., 1957. The measurement of productive efficiency. Journal of the Royal Statistical
Society 120(3): 253–290.
Favero, C.A. and Papi, L., 1995. Technical efficiency and scale efficiency in the Italian banking
sector: a non-parametric approach. Applied Economics 27(4): 385–395.
Feder G., Lau, L.J., Lin, J.Y. and Luo, X., 1990. The relationship between credit and
productivity in Chinese agriculture: A microeconomic model of disequilibrium.
American Journal of Agricultural Economics 72(5): 1151–1157.
Feder, G. and Nishio, A., 1999. The benefits of land registration and titling: Economic and
social perspectives. Land Use Policy 15(1): 143–69.
Feder, G., Lau, L.J., Lin, J.Y. and Luo, X., 1989. Agricultural credit and farm performance in
China. Journal of Comparative Economics 13: 508–525.
Fei J.C.H. and Ranis G., 1964. Development of Labor Surplus Economy: theory and policy.
Homewood, Richard D. Irwin, Illinois.
Foltz, J.D., 2004. Credit market access and profitability in Tunisian agriculture. Agricultural
Economics 30(3): 229–240.
Food and Agriculture Organization of the United Nations, FAO (2007). Online statistical
database, Rome. www.fao.org.
Freeman, A.H., Ehui, S. and Jabbar, M., 1998. Credit constraints and smallholder dairy
production in the east African highlands: Application of switching regression
model. Agricultural Economics 19: 33–44.
Gavian, S. and Fafchamps, M., 1996. Land tenure and allocative efficiency in Niger. American
Journal of Agricultural Economics 78(2): 460–71.
- 130 -
Georgescu-Roegen, N., 1960. Economic theory and agrarian economics. Oxford Economic
Papers 28: 1–40.
Goetz, S.M., and Sen, G.R., 1996. Who takes the credit? Gender, power, and control over
loan use in rural credit programs in Bangladesh. World Development 24(1): 45–
63
Gomez, P.Y., 1995. Théorie des conventions: une formalisation statistique. Cahiers de
Recherche 9502. Lyon: ESC.
González-Vega C., 2003. Deepening rural financial markets: Macroeconomic, policy and
political dimensions. Paving the way forward for rural finance. International
Conference on Best Practices, Washington DC, 2 – 4 June.
Greene, W., 1997. Econometric Analysis. 3rd Edition, Prentice Hall, Englewood Cliffs, New
Jersey.
Gregoriou, G., Messier, J. and Sedzro, K., 2005. On assessment of the relative efficiency of
Credit Union branches via Data Envelopment Analysis. Information Systems and
Operational Research 42(4): 281–297.
Guirkinger, C. and Boucher, S.R., 2008. Credit constraints and productivity in Peruvian
agriculture. Agricultural Economics 39(3): 295–308.
Gurgand, M., Peterson, G. and Yaron, J. (1996). Rural Finance in Sub-Saharan Africa, Savings
and Development 2: 135-169
Gutiérrez-Nieto, B., Serrano-Cinca, C. and Mar Molinero, C., 2009. Social efficiency in
microfinance. Journal of Operational Research Society 60: 104–119,
Heckman, J. and Robb, R., 1985. Alternative methods for evaluating the impact of
interventions. In Heckman, J. and Singer, B., (eds.), Longitudinal Analysis of
Labor Market Data. Cambridge University Press, New York: 156–245.
Helms B., 2006. Access for all: Building inclusive financial systems. The World Bank,
Washington DC.
Honoré L., 1995. L’octroi de crédit à l’entreprise: les modalités de la rationalisation de la
décision. Cahiers de recherche du Groupe ESC-Lyon, Lyon. Cahiers de recherche
Hudon, M. and Traca, D., 2011. On the efficiency effects of subsidies in microfinance: An
empirical inquiry. World Development 39(6): 966–973.
Hugon, P., 1996. Les innovations financières informelles et semi-formelles en Afrique
subsaharienne. Mondes en Développement 24(94): 17–28.
- 131 -
Hulme, D. and Mosley P., 1996. Finance Against Poverty. Volume 2, Routledge, London.
Hulme, D. and Mosley, P., 1997. Finance for the poor or poorest? Financial innovation,
poverty and vulnerability. In Wood, G.D. and Sharif, I. (eds.), Who needs credit?
Poverty and finance in Bangladesh. Zed Books, London and New York: 97–129.
Ichino, A., Mealli, F. and Nannicini, T., 2006. From temporary help jobs to permanent
employment: What can we learn from matching estimators and their
sensitivity? Discussion Paper 5736, London: CEPR.
Idiong, I., 2007. Estimation of farm level technical efficiency in small scale swamp rice
production in Cross River State of Nigeria: A stochastic frontier approach, World
Journal of Agricultural Sciences 3(5): 653–658.
Iheke, O.R., 2008. Technical efficiency of cassava farmers in South Eastern Nigeria: A
stochastic frontier approach. Agricultural Journal 3(2): 152–156.
Isaac, H., 1996. L’apport des théories des conventions à l’analyse de la gestion de la qualité
dans les services. Cahier de recherche 35. Paris: CREPA.
Jabbar, M.A., Ehui, S.K. and Von Kaufmann, R., 2002. Supply and demand for livestock credit
in Sub-Saharan Africa: Lessons for designing new credit schemes. World
Development 30(6): 1029–1042.
Jacoby, H.G. and Minten, B., 2007. Is land titling in Sub-Saharan Africa cost-effective?
Evidence from Madagascar. World Bank Economic Review 21(3): 461–485.
Jensen, M.C. and Meckling, W.H., 1976. Theory of the firm, managerial behavior, agency
costs and ownership structure. Journal of Business 3: 305–360.
Jensen, M.C., 1983. Organization theory and methodology. The Accounting Review 58 (2):
319–339.
Johnson, D., 2004. Measuring consumption and consumption poverty: Possibilities and
issues. Reconsidering the federal poverty measure. American Enterprise
Institute, Washington DC, 18 November.
Kalala, K., 1988. La contribution des coopératives d’épargne et de crédit au développement
de l’agriculture paysanne: Le cas du Zaïre. Thèse de doctorat, Département
d’Economie Agricole, Université de Kyoto, Japon.
Kalonji, N., 2005. Le DSRP et la problématique de la lutte contre la pauvreté en République
Démocratique du Congo. Conférence-Débat tenue à l’Université de Kinshasa,
Kinshasa, 18 Juin.
- 132 -
Khandker, S. and Faruque, R.R., 2003. The impact of farm credit in Pakistan. Agricultural
Khandker, S., 2001. Does microfinance really benefit the poor? Evidence from Bangladesh.
Asia and Pacific Forum on Poverty: Reforming Policies and Institutions for
Poverty Reduction. Manila: Asia Development Bank, 5–9 February.
Khandker, S., 2003. Micro-finance and poverty: Evidence using panel data from Bangladesh.
Working Paper 2945, Washington DC: World Bank Policy Research.
Khandker, S., 2005. Microfinance and poverty: Evidences using panel data from Bangladesh.
Word Bank Economics Revue 19(2): 263–286.
Kikassa, M., 1981. Les dix ans du Mouvement Coopératif d’Épargne et de Crédit au Zaïre.
Zaïre-Afrique 159: 543–565.
Klein, B., Meyer, R. L., Hannig A., Burnett J. and Fiebig, M., 1999. Better practices in
agricultural lending. Food and Agricultural Organization and Deutsche
Gesellschaft für Technische Zusammenarbeit, Rome.
Kochar, A., 1997. An empirical investigation of credit constraints in rural credit markets in
India. Journal of Development Economics 53: 339–371.
Kodjo, M.K., Abiassi, E.H. and Allagbe, M.C., 2003. Financement de l’agriculture béninoise
dans un contexte de libéralisation: contribution de la microfinance. Cahier de
recherche 03–4. Cotonou: ELIFID.
Kohansal, M.R., Ghorbani, M. and Mansoori, H., 2008. Effect of credit accessibility of farmers
on agricultural investment and investigation of policy options in Khorasant-
Razavi Province. Journal of Applied Sciences 8(23): 4455–4459.
Komicha, H.H. and Öhlmer, B., 2006. Effect of credit constraints on production efficiency of
farm households in Southeastern Ethiopia. Ethiopian Journal of Economics
15(1): 1–34.
Korsah, K.B., Nyarko, E.K. and Tagoe, N.A., 2001. Impact of financial sector liberalization on
competition and efficiency in the Ghanaian banking industry. Research Paper
01(2). Rome: International Labour Organization.
Koveos, P. and Randhawa, D., 2004. Financial services for the poor: assessing microfinance
institutions, Managerial Finance. Emerald Group Publishing Limited 30 (9): 70–
95.
- 133 -
Kumar, S., 2006. A decomposition of total factor productivity growth: A regional analysis of
Indian industrial manufacturing growth. International Journal of Productivity
and Performance Management 55(3–4), 311–331.
Labie, M., 1999. La Micro-Finance en Questions. Limites et choix organisationnels, Luc Pire,
Bruxelles.
Lapenu, C., 2001. Adéquation entre l’offre des MFI et les besoins de l’agriculture familiale. Le
financement de l’agriculture familiale dans le contexte de la libéralisation.
Quelle contribution de la microfinance? Séminaire Internationale sur le
financement de l’agriculture familiale, Dakar, 21–24 Janvier.
Lapenu, C., Zeller, M., Greeley, M., Chao-Béroff, R. and Verhagen, K., 2004. Performances
sociales: une raison d’être des institutions de microfinance et pourtant encore
peu mesurées, Quelques pistes. Mondes en Développement 32 (126): 57–74.
Latruffe, L. ,2004. Farm Technical Efficiency, Credit and Investment in a Transition Country: A
Microeconomic Analysis in Poland. PhD thesis, University of Paris 10, France.
Ledgerwood, J., 1999. Microfinance handbook: An institutional perspective. The World Bank,
Washington DC.
Lelart, M., 1989. L'épargne informelle en Afrique. Les tontines béninoises. Revue Tiers
Monde 30(118): 271–298.
Lesaffre, L. and Pesche, D., 2002. Microfinance et sécurisation du crédit aux exploitations
agricoles. Séminaire Internationale sur le Financement de l’Agriculture
Familiale, Dakar, 21–24 Janvier.
Leuven, E. and Sianesi, B., 2003. PSMATCH2: Stata module to perform full Mahalanobis and
propensity score matching, common support graphing, and covariate
imbalance testing. Department of Economics, Boston College.
Levine, R., 1997. Financial development and economic growth: view and agenda. Journal of
Economic Literature 35(2): 688–726.
Li, R. and Zhu, X., 2010. Econometric analysis of credit constraints on Chinese rural
households and welfare loss. Applied Economics 42(13): 1615–1625.
Lubini, A., 1988. Les Sols de Jachères et les Pâturages du Centre Agricole de Mbankana.
Fondation Hanns-seidel, Kinshasa.
Lucas, R.E.B., 1987. Emigration to South African mines. American Economic Review 77(3):
313–330.
- 134 -
Maldonado, J.H., 2004. Relationships among poverty, financial services, human capital, Risk
coping and natural resources: Evidences from El Salvador and Bolivia. Ph.D.
Thesis, Department of Agricultural, Environment and Development Economics,
Ohio State University, USA.
Martínez-González, A., 2008. Technical efficiency of microfinance: Evidence from Mexico.
Master’s thesis, Department of Agricultural, Environment and Development
Economics, Ohio State University, USA.
McAllister, P.H. and McManus, D., 1993. Resolving the scale efficiency puzzle in banking.
Journal of Banking and Finance 17(2–3): 389–406.
McKinnon, R.I., 1973. Money and Capital in Economic Development. Brookings Institution,
Washington DC.
Mendola, M., 2008. Migration and technological change in rural households: Complements
or substitutes. Journal of Development Economics 85(1–2): 150–175.
Mester, L., 1996. A study of bank efficiency taking into account risk-preferences. Journal of
Banking and Finance 20(6): 1025–1045.
Meyer, R.L., 2002. Track record of financial institutions in assisting the poor in Asia. ADB
Institute Research Paper 49. Manila: Asian Development Bank
Ministère du plan et commerce, 2002. Rapport annuel 2002. Kinshasa , République
Démocratique du Congo.
Morduch, J. and Haley, B., 2002. Analysis of the effects of microfinance on poverty
reduction. Wagner Working Paper 1014. New York: New York University.
Morduch, J., 1998. Does microfinance really help the poor? New evidence from flagship
programs in Bangladesh. New York University, New York.
Muayila, K., 2004. Evaluation de la pauvreté dans les milieux périurbains de Kinshasa. Revue
de la Faculté d’Administration des Affaires et Sciences Économiques numéro et
page ?
Muayila, K., 2006. Financement de l’agriculture paysanne par les organisations de
microfinance. Déterminants de l’offre et la demande de micro crédit par les
ménages agricoles dans la périphérie de Kinshasa. Mémoire de DEA, Institut
d’Etudes du Développement, Université de Liège, Belgique.
Mushinski, D., 1999. An analysis of offer function of Banks and credit Unions in Guatemala.
Journal of Development Studies 36(2): 87–111.
- 135 -
Mutamba, M., 1989. Makoko, roi des Bateke. CRP, Kinshasa.
Narajan, D. and Pritchett, L., 1999. Cents and sociability: Household income and social capital
in rural Tanzania. Economic Development and Cultural Change. 47(4): 871–897.
Navajas, S., Schreiner, M., Meyer, R.L., González-Vega, C. and Rodriguez-Meza, J., 2000.
Microcredit and the poorest of the poor: Theory and evidence from Bolivia.
World Development 28: 333–346.
Negash, Z., 2008. Microfinance Loan Delivery, Utilization and Impact: with specific reference
to Tigray, North Ethiopia. Ph.D. Thesis, Division of Agricultural and Food
Economics, Katholieke Universiteit Leuven, Belgium.
Nkwembe, U., 2006. Problématique de la pauvreté des ménages agricoles ruraux et urbains
dans la périphérie de la ville de Kinshasa, Essai d’analyse du phénomène et de
ses implications sur la sécurité alimentaire. Thèse de doctorat, Département
d’Agronomie, Université Catholique de Louvain, Belgique.
Nsombo, M.B., 2005. Utilisation actuelle des terres dans le rayon du centre d'appui au
développement intégré de Mbankana au plateau des Batéké. Mémoire de
DESS, Département d’Aménagement et Gestion Intégrés des Forêts et
Territoires Tropicaux, Université de Kinshasa, R.D. Congo.
Ntoto, M., 2001. Budget de consommation des ménages: structure et déterminants, cas de
quelques quartiers pauvres de la ville de Kinshasa, Kisenso, Kindele et Makala.
In Tollens, E. and Kankonde, M. (eds.), Sécurité alimentaire au Congo-Kinshasa,
analyse, production et consommation. Harmattan, Katholieke Universiteit
Leuven, Paris: 367– 404.
Nuryartono, N., Zeller, M. and Schwarze, S., 2005. Credit rationing of farm households and
agricultural production: Empirical evidence in the rural areas of Central
Sulawesi, Indonesia. Tropentag, Stuttgart-Hohenheim, 11–13 October.
Nzemen, M., 1988. Théorie de la Pratique des Tontines au Cameroun. L'Harmattan, Paris.
Ogboma, M.C., Asumugha, G.N., Anyaegbunam, H.N., Okoye, B.C., Onwumere, J. and
Akinpelu, O.A., 2007. Resource use efficiency in cassava processing in two
selected areas of Abia State, Nigeria. Pakistan Journal of social sciences 4 (6):
778–781.
- 136 -
Ogundari, K. and Ojo, S., 2007. An Examination of technical, economic and allocative
efficiency of small farmers: the case study of cassava farmers in Osun State of
Nigeria. Bulgarian Journal of Agricultural Sciences 13: 185–195.
Ogundari, K., 2008. Resource productivity, allocative efficiency and determinants of technical
efficiency of rainfed rice farmers: a guide for food security policy in Nigeria.
Agricultural Economic-Czech 54(5): 224–233.
Okten, C. and Osili, U.O., 2004. Social networks and credit access in Indonesia. World
Development 32(7): 1225–1246.
Omonona, B.T., Akinterinwa, A.T. and Awoyinka, Y.A., 2008. Credit constraint and output
supply of cowan farmers in Oyo State, Nigeria. European Journal of Social
Sciences 6(3): 382–390.
Onu, J.I. and Edon, A., 2009. Comparative economic analysis of improved and local cassava
varieties in selected local government areas of Tabara State, Nigeria. Journal of
social sciences 19(3): 213–217.
Otero, M. and Rhyne, E., 1994. The New World of Micro-enterprise Finance: Building Healthy
Financial Institutions for the Poor. Kumarian Press, Connecticut.
Oyedele, G.A., Akintola, J.O. and Rahji, B.T., 2009. Credit constrained condition of farm
households and profitability of agricultural production of Nigeria agriculture.
Agricultural Journal 4(4): 192–201.
Pauwels, L., 1993. Nzayilu N’ti, guide des arbres et arbustes de la région de Kinshasa-
Brazzaville. Jardin Botanique National de Belgique, Meise.
Paxton, J., 2007. Technical efficiency in a semi-formal financial sector: The case of Mexico.
Oxford Bulletin of Economics and Statistics 69: 57–74.
Petrick, M., 2003. Empirical measurement of credit rationing in agriculture: a methodological
survey. IAMO Discussion Paper 45. Halle (Saale): IAMO.
Pitt, M.M. and Khandker, S., 1998. The impact of group-based credit programs on poor
households in Bangladesh: Does the gender of participants matter? Journal of
Political Economy 106(2): 958–996.
Pleitez-Chávez, R., 2004. Remittances as a strategy to cope with systemic risk: Panel results
from rural households in El Salvador. Ph.D. Thesis, Department of Agricultural,
Environment and Development Economics, Ohio State University, USA.
- 137 -
Rahji, M.A.Y. and Adeoti, A.I., 2010. Determinants of agricultural credit rationing by
commercial banks in South-Western Nigeria. International Research Journal of
Finance and Economics, Issue 37.
Rangan, N., Grabowski, R., Aly, H.Y. and Pasurka, C., 1988. The technical efficiency of US
banks. Economics Letters 28(2): 169–175.
Romer, P.M., 1986. Increasing returns and long-run growth. Journal of Political Economy
94(5): 1002–1037.
Rosenbaum, P.R. and Rubin, D.B., 1983. The central role of the propensity score in
observational studies for causal effects. Biometrika 70: 41–55.
Schmidt, R.H. and Zeitinger C.P., 1994. Critical issues in small and microbusiness finance.
Working Paper 1. Frankfurt: Internationale Projekt Consult (IPC) GmbH.
Seibel, H.D., 1996. L’évolution de la Microfinance. Défis Sud, Bruxelles.
Seiford, L.M. and Thrall, R.M., 1990. Recent developments in DEA: the mathematical
approach to frontier analysis. Journal of Econometrics 46(12): 7–38.
Sen, A.A., 1962. An aspect of Indian agriculture. Economic Weekly 14: 243–246.
Shaw, E.S., 1973. Financial Deepening in Economic Development. Oxford University Press,
New York.
Sherman, H.D. and Gold, F., 1985. Bank branch operating efficiency: Evaluation with data
envelopment analysis. Journal of Banking and Finance 9(2): 297–315.
Simar, L. and Wilson, P.W., 2007. Estimation and inference in two-stage, semi-parametric
models of production processes. Journal of Econometrics 136: 31–64.
Soteriou, A.C. and Zenios, S.A., 1999. Using data envelopment analysis for costing bank
products. European Journal of Operational Research 114: 234–24.
Soulama, S., 2008. Efficacité et inefficience à l’échelle des institutions de microfinance au
Burkina-Faso. Journées Internationales de Micro-intermédiation, Laboratoire
d’Orléans, Orléans, 13–14 Mars.
Stiglitz, J.E. and Weiss, A., 1981. Credit rationing in markets with imperfect information.
American Economic Review 65: 283–300.
Sylla, M.B., 2002. Evaluation de la pauvreté en Guinée-Bissau (2001–2002). Bissau: Institut
National de statistique et Banque mondiale.
- 138 -
Tanvir, A. and Zulfiquar, A.G., 2007. Role of agricultural credits and efficiency of commercial
banks in Pakistan. International Journal of Agriculture and Biology 9(6): 921–
924.
Tariq, M. and Mohd, A.I., 2010. Technical efficiency of microfinance institutions in India: A
stochastic frontier approach. MPRA Paper 25454, University Library of Munich.
Taylor, W.M., Thompson, R.G., Thrall, R.M. and Dharmapala P.S., 1997. DEA/AR efficiency
and profitability of Mexican banks: A total income model. European Journal of
Operational Research 98: 175–212.
Tollens, E., 2004. Les défis: Sécurité alimentaire et cultures de rente pour l’exportation.
Principales orientations et avantages comparatifs de l’agriculture en RD Congo.
Table Ronde sur l’Agriculture en RDC, Kinshasa, 19–20 Mars.
Toshio, K., 2007. Impact of microfinance on rural households in the Philippines. Asian
Development Bank, Mandaluyong City.
Tucker, M. and Miles, G., 2004. Financial performance of microfinance institutions: a
comparison to performance of regional commercial banks by geographic
region. Journal of Microfinance 6(1): 41–54.
Union Centrale de Coopératives d’Epargne et Credit (UCCEC), 1999. Rapport annuel,
Kinshasa. UCCEC, Kinshasa.
United Nations Statistics Division (UNSD), 2005. Handbook on poverty statistics: Concepts,
methods and policy use. Special project on poverty statistics. UNSD, New York.
Wampfler, B., 2000. Le financement de l’agriculture dans un contexte de libération: quelle
contribution de la microfinance ? Techniques Financière et Développement 59:
61–70.
Wampfler, B., Doligez, F. and Poursat, C., 2002. Quel mode intervention pour améliorer la
contribution de la microfinance au financement de l’agriculture. Séminaire
Internationale sur le financement de l’agriculture familiale, Dakar, 21–24
JanvierJanuary.
Waqar, A., Zakir, H., Sial, M.H. and Ijaz H., 2008. Agricultural credit constraints and
borrowing behavior of farmers in rural Punjab. European Journal of Scientific
Research 23(2): 294–304.
Wheelock, D.C. and Wilson, P.W., 1999. Technical progress, inefficiency, and productivity
change in U.S banking. Journal of Money, Credit and Banking 31(2): 21–234.
- 139 -
Woller, G., Dunford, C. and Woodworth, W., 1999. Where to microfinance? International
Journal of Economic Development 1: 29–64.
Woodruff, C. and Zenteno, R., 2007. Migration networks and microenterprises in Mexico.
Journal of Development Economics 82(2): 509–528.
Wright, G., 2000. Microfinance Systems: Designing Quality Financial Services for the poor.
University Press Ltd., Dhaka and Zed Books, London and New York.
Yang, D. and Choi, H., 2007. Are remittances insurance? Evidence from rainfall shocks in the
Philippines. The World Bank Economic Review 21(2): 219–248.
Yaron, J., 1992. Rural finance in developing countries. The World Bank, Washington DC.
Yaron, J., 1994. Successful rural financial institutions. The World Bank Research Observer
9(1): 49–70.
Yeh, Q., 1996. The application of Data Envelopment Analysis in conjunction with financial
ratios for bank performance evaluation. The Journal of the Operational
Research Society 47(8): 980–988.
Yehuala S., 2008. Determinants of smallholder farmers access to formal credit: the case of
Metema Woreda, North Gondar, Ethiopia. Master’s thesis, Department of Rural
Development and Agricultural Extension, Haramaya University, Ethiopia.
Zaman, H., 1999. Assessing the Poverty and Vulnerability Impact of Microcredit in
Bangladesh: A case Study of BRAC. The World Bank, Washington DC.
Zegarra, E., Escobal, J. and Aldana, U., 2008. Titling, credit constraints, and rental markets in
rural Peru: Exploring channels and conditioned impacts. Inter-American
Development Bank, Washington DC.
Zeller, M. and Meyer, R.L., 2002. The Triangle of Microfinance, Financial Sustainability,
Outreach and Impact. Johns Hopkins University Press, Baltimore.
Zeller, M., 1994. Determinants of credit rationing: A study of informal lenders and formal
credit groups in Madagascar. World Development 22(12): 1895–1907.
Zeller, M., 1995. The demand for financial services by rural households: Conceptual
framework and empirical findings. Quarterly Journal of International
Agriculture 34: 149–170.
Zeller, M., Lapenu, C. and Greeley, M., 2003. Measuring social performance of microfinance
institutions: A proposal. Social Performance Indicators Initiative (SPI), Final
report. Argidius Foundation and CGAP, Washington DC.
- 140 -
Zeller, M., Sharma, M., Henry, C. and Lapenu, C., 2002. An operational tool for evaluating
poverty outreach of development policies and projects. In Zeller, M. and
Meyer, R.L. (eds.), The Triangle of Microfinance. Johns Hopkins University Press,
Baltimore and London: 172–195.
Zhou, M. and Bankston, C.L., 1996. Social capital and the adaptation of the second
generation: The case of Vietnamese youth in New Orleans. In Alejandro, P.
(ed.), The New Second Generation. Russell Sage Foundation, New York: 197–
232.
- 141 -

Muayila Kabibu Edited by James

Uploaded by

Copyright:

Available Formats

Muayila Kabibu Edited by James

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Muayila Kabibu Edited by James

Uploaded by

Copyright:

Available Formats

Acknowledgements

We would like to thank the Church Development Service (Evangelischer Entwicklungsdienst

CADIM Centre d’Appui au Développement Intégral/Mbankana

Figure 1.1 Food-production index (base period, 1999-2001) 2

Table 1.1 Distribution of sample size 17

1.1 Background information

With an under-developed industrial sector, agricultural activities remain the principal

Figure 1.1 Food-production index (base period, 1999-2001)

Source FAO, 2007.

Source : FAO, 2007.

The literature discusses various institutional forms of microfinance organizations, including

1.3 Problem statement

3. Do credit constraints affect the production efficiency of farm-households?

4. What are key drivers of technical efficiency of farm-households?

5. Do credit constraints influence farm-household welfare distribution?

1.4 Conceptual frameworks

The performance of MFIs is evaluated with reference to the efficiency frontier as a

Local institutionnel & Household Capital

Macro economic Household Activities

Food/ Non -food

Welfare & Resource Income

1.5 Objectives and hypotheses

1. To identify the difficulties to the provision of micro credit towards farm-households;

4. To evaluate the impact of credit constraints on technical efficiency of farm-households;

5. To evaluate the effect of credit constraints on farm-household welfare distribution;

1. The nature of agricultural activity constrains farm household access to microcredit;

3. Lending technology applied in the MFIs constrains farm-household participation in the

4. Capital endowments reduce the probability that a farm-household is credit-constrained;

6. Household’s capital endowments positively affect the technical efficiency of farm-

8. Credit constraints reduce the technical efficiency of farm-households;

10. Credit constraints reduce the farm-household welfare level;

1.6 Study area

Figure 1.4 Location of the study area

Source: De Saint Moulin and Kalombo (2005)

1.7 Data and research methodology

1.7.1 Data collection

Table 1.1 Distribution of sample size

Village Number of Households Rate Households Selected

1.7.2 Research methodology

Methods Logit & Probit PSM TR DEA

Chapter 2 Chapter 4 Chapter 3 Chapter 5

Data Household survey Microfinance survey

Figure 1.5 Schematic presentation of research methodology

1.8 Overall dissertation structure

Household level Microfinance

Impact of credit Chapter 3 Chapter 4

Figure 1.6 Schematic presentation of dissertation organization

App. Table 1.1 Descriptive statistic of quantitative variables

Variable N Mean SD Min Max F (P.value)

App. Table 1.2 Descriptive statistic of dummy variables

Variable N % Khi 2 (P.value)

Determinants of Farm Household Credit Constraints in the Hinterland of Kinshasa

2.2 Literature review

Neoclassical models define the market as a place of coordination of economic agents,

An important theoretical advance in the understanding of credit constraints and rationing

Another important theoretical advance in the understanding of credit rationing mechanisms

Another constraint on the financing of agricultural activities is related to the difficulty of

Agricultural activities, especially in developing nations, are characterized by high covariance

Did you apply ?

Key: * p<1; p<.05; *p<.01

Key: * p<.1; p<.05; *p<.01