efficiency of microfinance institutions: a data envelopment analysis
TRANSCRIPT
Asia-Pacific Finan Markets (2010) 17:63–97DOI 10.1007/s10690-009-9103-7
Efficiency of Microfinance Institutions: A DataEnvelopment Analysis
Mamiza Haq · Michael Skully · Shams Pathan
Published online: 17 October 2009© Springer Science+Business Media, LLC. 2009
Abstract This study examines the cost efficiency of 39 microfinance institutionsacross Africa, Asia and the Latin America using non-parametric data envelopmentanalysis. Our findings show non-governmental microfinance institutions particularly;under production approach, are the most efficient and this result is consistent withtheir fulfillment of dual objectives: alleviating poverty and simultaneously achiev-ing financial sustainability. However, bank-microfinance institutions also outperformin the measure of efficiency under intermediation approach. This result reflects thatbanks are the financial intermediaries and have access to local capital market. It maybe possible that bank-microfinance institutions may outperform the non-governmentalmicrofinance institutions in the long run.
M. Haq (B)University of Queensland Business School, The University of Queensland, St. Lucia,QLD 4072, Australiae-mail: [email protected]
M. HaqRMIT University, Victoria, Australiae-mail: [email protected]
M. SkullyDepartment of Accounting and Finance, Monash University, PO Box 197, Caulfield East,Victoria 3145, Australiae-mail: [email protected]
S. PathanFaculty of Business, Technologyand Sustainable Development, Bond University, Robina, Gold Coast,QLD 4229, Australiae-mail: [email protected]
123
64 M. Haq et al.
Keywords Africa · Asia · Data envelopment analysis · Efficiency · LatinAmerica · Micro finance institutions
1 Introduction
Microfinance institutions (MFIs) provide a range of financial services to poor house-holds.1 Their worldwide growth in numbers has had a positive impact by providingthe poor with loans, savings products, fund transfers and insurance facilities. This hashelped create an encouraging socio-economic environment for many of these devel-oping countries households.
The nature of these institutions is quite different from traditional financial institu-tions (such as commercial banks). MFIs are significantly smaller in size, limit theirservices towards poor households and often provide small collateral-free group loans.Most MFIs depend on donor funds and are not-for-profit oriented organizations thatshare a common bond among the members. They also differ in their two main oper-ational objectives. First, as mentioned they act as financial intermediaries to poorhouseholds. This is known as the ‘institutionist paradigm’ (Woller et al. 1999; Murdoch2000) which affirms that MFIs should generate enough revenue to meet their operatingand financing costs. Second, they have a social goal. This can be defined as the ‘welf-arists’ paradigm’ which includes a focus on poverty alleviation and depth of outreachalong with achieving financial sustainability.2 An efficient MFI management shouldpromote these two objectives (Brau and Woller 2004).
The formal MFI institutions (bank MFIs, non bank financial institution MFIsand cooperative MFIs) are subject to prudential regulation and their activities li-censed; mainly delivering credit facilities to their members. Some of these also mobi-lize savings from non-members. In contrast, semiformal MFI institutions, typicallynon-government organization MFIs (NGO-MFIs), are usually unregulated but regis-tered under some society legislation. Table 1 shows the range of products and fundingsources each type entail. Finally, the informal MFI institutions include money lend-ers, shop keepers and pawn brokers. Unfortunately, their small size and often lack oflicensing make them difficult to identify and so they are excluded from our study. Thequestion, though, among the remaining four MFI types is whether one category mayprove more efficient than the others.
MFIs with the largest asset size are found in Asia. Asia also has the most efficientMFIs due to large population densities and lower wages (Microbanking Bulletin 2004).Other factors such as strong outreach and preservation of low operating expenses havealso helped Asian MFIs to be efficient. However, South Asian MFIs are relatively moreefficient than their counterparts in East Asian MFIs.3 This differences in efficiency
1 Poor households are defined as under the international poverty line of 1.00US$ per day. Maxwell (1999)reports the poor households generally fall under the category of income/consumption poverty, social exclu-sion, lack of capability and functioning, vulnerability, livelihood unsustainability and relative deprivation.2 Welfarists’ paradigm believes MFIs can achieve sustainability without being self-sufficient. They suggestdonor fund is the equity and donors are the social investors.3 South Asian MFIs incur on an average 25 dollars per borrower while Philippines, Vietnam, Indonesiaand Cambodia incur double the cost.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 65
Table 1 Features of micro finance institutions across Africa, Asia and the Latin America
Microfinanceinstitutions
Type, statusand reg.
Productsoffered
Percentage ofMF
Main fundingsource
Ruhunadevelopmentbank/SouthAsia/SriLanka
Formal/bank/reg. L, VS, L, LE,FT, T&C
81–90% Shareholdercapital,savings,loans andgrants
NirdhanUtthan bankLtd./SouthAsia/Nepal
Formal/bank/reg. L, VS, I 91–100% Grants, loans,savings,shareholdercapital
First Microfi-nance BankLtd./SouthAsia/Pakistan
Formal/bank/reg. L, VS, I, FT 91–100% Shareholdercapital,grants,savings
K-REP/Africa/Kenya
Formal/bank/reg. L, VS, I, FT 91–100% Savings loans,share holdercapital
Bank Solidario/Latin America/Ecuador
Formal/bank/reg. L, VS, FT 91–100% Loans, savingsand shareholdercapital
Centenary Rural DevelopmentbankLtd/(CERUDEB)/Africa/Uganda
Formal/bank/reg. L, VS, FT 91–100% Savings
Amhara Credit and SavingsInstitutions(ACSI)/Africa/Eithiopia
Formal/bank/reg. L, VS, FT 91–100% Grants andloans
Association ofCambodianLocal EconomicDevelopmentAgencies(ACLEDA)/EastAsia/Cambodia
Formal/bank/reg. L, VS, FT,T&C
91–100% Savings andshareholdercapital
Banco Sol/LatinAmerica/Bolivia
Formal/bank/reg. L, VS, FT,T&C
91–100% Loans, savingsand shareholdercapital
ChhimekBikas BankLtd(CBB)/SouthAsia/Nepal
Formal/bank/reg. L, VS, I, FT 91–100% Loans, savingsand shareholdercapital
Bank RakyatIndonesia(BRI)/EastAsia/Indonesia
Formal/bank L, VS, FT,T&C
100% Savings
GrameenBank/SouthAsia/Bangladesh
Formal/bank/reg. L, VS, I, 91–100% Grants, loans,savings andshareholdercapital
123
66 M. Haq et al.
Table 1 Continued
Microfinanceinstitutions
Type, statusand reg.
Productsoffered
Percentage ofMF
Main fundingsource
Dedebit Credit and Savingsinstitutions(DECSI)/Africa/Eithiopia
Formal/bank/reg. L, VS, I, FT 91–100% Savings
Equity Bank(EBS)/Africa/Kenya
Formal/NBFI/reg. L, VS, FT,T&C
91–100% Loans, savingsandshareholdercapital
Angkor Mikro-heranhvathoKampuchea(AMK)/EastAsia/Cambodia
Formal/NBFI/reg. L, VS 91–100% Shareholdercapital
Caja MunicipalAhorro yCredito –Arequipa(CMAC)/LatinAmerica/Peru
Formal/NBFI/reg. L,VS 91–100% Savings
Commercial MicrofinanceLimited (CMF)/Africa/Uganda
Formal/NBFI/reg. L, VS, FT 91–100% Savings
Caja Nor/LatinAmerica/Peru
Formal/NBFI/reg. L, VS, I, FT 91–100% Loans,savings,shareholdercapital
Sarvodaya Economic EnterpriseDevelopment Services(SEEDS)/South Asia/SriLanka
Formal/NBFI/reg. L, VS, I, 81–90% Grants,savings andloans
ServiciosFinancierosComunitarios(FINCOMUN)/LatinAmerica/Mexico
Formal/NBFI/reg. L, VS, 90% Savings andshareholdercapital
Africa Village Financial Services(AVFS)/Africa/Ethiopia
Formal/NBFI/reg. L, VS, 91–100% Grants,savings,loans
VYCCUSavings andCreditCooperativeSocietyLtd/SouthAsia/Nepal
Formal/coop/reg. L, VS, I, T&C 71–80% Shareholdercapital,savings,loans
Cooperative deAhorro y CreditoJardinAzuayo(COAC)/LatinAmerica/Ecuador
Semiformal/coop/unreg
L, VS,FT
91–100% Loans,savings,shareholdercapital
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 67
Table 1 Continued
Microfinanceinstitutions
Type, statusand reg.
Productsoffered
Percentage ofMF
Main fundingsource
Credit Mutuel du Senegal(CMS)/Africa/Senegal
Formal/coop/reg. L, VS 90% Shareholder capital,savings, loans, grants
AssociationCooperativade Ahorro yCreditoVincentina deR.L. (ACCOVI)/Latin America/El Salvador
Formal/coop/reg. L, VS, I 91–100% Loans, savings
Alliance de Credit et d’ Epargnepour la Production(ACEP)/Africa/Senegal
Formal/coop/reg. L, VS, FT 91–100% Loans
KapalongCooperative(KC)/EastAsia/Philippines
Semiformal/coop/unreg.
L, VS 91–100% Savings
NeighborhoodSocietyService Centre(NSSC)/SouthAsia/Nepal
Formal/NGO/reg. L, VS, I 81–90% Loans andsavings
Capital AidFund forEmploymentof the Poor(CEP)/EastAsia/Vietnam
Semiformal/NGO/unreg.
L, VS 91–100% Grants, loans,shareholdercapital andsavings
TYMFUND/EastAsia/Vietnam
Semiformal/NGO/unreg.
L, VS, I, T&C 91–100% Grants, loansand savings
Bangladesh Rural AdvancementCommittee (BRAC)/SouthAsia/Bangladesh
Formal/NGO/reg.
L, VS, T&C 90% Savings,grants loans
Association for SocialAdvancement (ASA)/SouthAsia/Bangladesh
Semiformal/NGO/unreg.
L, VS, I 91–100% Loans andsavings
Cebu Micro enterpriseDevelopmentFoundationInc(CMEDFI)/EastAsia/Philippines
Semiformal/NGO/unreg.
L, VS, I 91–100% Grants, loans,savings,shareholdercapital
All CeylonCommunityDevelopmentCouncil(ACCDC)/SouthAsia/Sri Lanka
Semiformal/NGO/unreg.
L, I, T & C 71–80% Shareholdercapital
123
68 M. Haq et al.
Table 1 Continued
Microfinanceinstitutions
Type, statusand reg.
Productsoffered
Percentage ofMF
Main fundingsource
Women DevelopmentFederation Hambantota(WDFH)/South Asia/SriLanka
Semiformal/NGO/unreg.
L, VS, I, T &C
91–100% Savings, grants, loans,shareholder-capital
Kashf Foundation/SouthAsia/Pakistan
Formal/NGO/reg.
L, VS, I, T &C
91–100% Grants, loans
Women and Associationsfor Gain both Economicand Social(WAGES)/Africa/Togo
Semiformal/NGO/unreg.
L, VS, T & C 91–100% Grants, loans,savings
Women DevelopmentFederation Wilgamuwa(Wilgamuwa)/SouthAsia/Sri Lanka
Semiformal/NGO/unreg.
L, VS 81–90% Grants,savings ,shareholdercapital
Credit du Sahel(CDS)/Africa/Cameroon
Formal/other/reg.
L, VS, LE, I,FT and T &C
71–80% Savings(80%),grants,loans,shareholdercapital
This table presents the different characteristics of the MFIs. We choose those banks that provide more than50% of microfinance services to the poor households. We first describe the features of bank-MFIs followed byNBFI-MFIs, co-op-MFIs and NGO-MFIs. We have also divided these microfinance institutions into formaland semiformal institutions. Reg. represents regulated organisation while unreg. represents unregulatedorganization. The formal institutions are licensed and regulated under the prudential banking regulation andsupervision while the semiformal institutions are unregulated but registered under some society legislation.L loans, VS voluntary savings, LE leasing, I insurance, FT fund transfer, T&C training and consulting, MFmicrofinance. Source www.mixmarket.org/en/demand (accessed in Jan 2006)
may be the result of various lending methodology applied by the Asian MFIs. ManyIndian MFIs, for example, reduce their staffing costs by lending to self-help groupsrather than to the individual borrower. Our findings show that bank-MFIs are the mostefficient under intermediation approach while NGO-MFIs are the most efficient underproduction approach.
Our study chose to apply the DEA model for several reasons. First, the DEA modelis able to incorporate multiple inputs and outputs easily. Thus, DEA is particularlywell-suited for efficiency analysis of MFIs as it considers multiple inputs and producesmultiple outputs such as alleviating poverty and achieving sustainability. Second aparametric functional form does not have to be specified for the production function.Third, DEA does not require any price information for dual cost function as is requiredfor parametric approaches.4 Fourth, DEA has the potential to provide information tothe supervisors in improving the productive efficiency of the organization. Finally,DEA presents a generalization approach because other assumptions than constantreturn to scale can be accommodated within a convex piecewise linear best practice
4 Unfortunately due to lack of price information we do not apply the parametric approaches in this study.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 69
frontier. DEA has traditionally been used for the study of non-profit organization (suchas hospital) efficiency and bank efficiency (Sherman and Gold 1985).
The rest of the paper is structured as follows. The next section covers a brief litera-ture on efficiency measurement of MFIs. Section 3 discusses the methodology used toanalyze MFI efficiency. Section 4 presents the results. Section 5 provides a summaryand finally draws the conclusion on the MFI efficiency across the regions.
2 Literature Review
There has been little research conducted on the efficiency of microfinance institu-tion. However, several studies examine various measures of MFI efficiency acrossthe regions. Study by Farrington (2000) identifies a number of accounting variablesto reflect the efficiency of the microfinance institutions. These accounting variablesare administrative expense ratio, number of loans per loan officer and loan officers tototal staff, portfolio size, loan size, lending methodology, source of funds and salarystructure as the efficiency drivers and hence as the measurements for MFI efficiency.Lafourcade et al. (2005) use cost per borrower and cost per saver as measure of effi-ciency. They found African MFIs incur highest costs per borrower but have the lowestcosts per saver. They also mention that regulated MFIs maintain higher efficiencythrough low costs per borrower and per saver. In contrast, African cooperative-MFIsare the least efficient with the highest cost per borrower. Nevertheless, cooperative-MFIs have the lowest cost per saver but unregulated MFIs have the highest. Noneof these two studies use any parametric or non-parametric approach to evaluate theefficiency of MFIs.5
Until recently, the issue of efficiency has been less commonly examined with para-metric models. Guitierrez-Nieto et al. (2006) applied data envelopment analysis (DEA)to measure the efficiency of 30 Latin American MFIs and then explored the multivar-iate analysis of the DEA results. They developed 21 specifications using two inputsand three outputs. They identified an NGO (W-Popayan), and a non-bank financialinstitution (Findesa) as the most efficient among the group of 18 MFIs. The otherkey MFI paper is that of Hassan and Tufte (2001) using a parametric approach (sto-chastic frontier analysis or SFA) found that Grameen Bank’s branches staffed by thefemale employees operated more efficiently than their counterparts staffed by themale employees. Further, Desrochers and Lamberte (2003) have also used paramet-ric approaches to study the efficiency of cooperative rural banks in the Philippines.They found that cooperative rural bank with good governance were more efficientthan their counters laced by bad governance. Leon (2001) reported that productivityof resources, governance, and business environment were the contributing factors forthe cost-efficiency of the Peruvian municipal banks.
We add to the literature by applying DEA on MFIs where profit maximization maynot be the vital interest to policymakers and regulators. Theoretically, productivity canbe measured in terms of borrowers per staff member and savers per staff member. Thus,
5 Parametric approach includes stochastic frontier approach, flexible profit function etc and non-parametricapproach includes data envelopment analysis.
123
70 M. Haq et al.
high level of MFIs efficiency here may be a result of maintaining high productivityper employee level (Microbanking Bulletin 2005). Baumann (2005) has also foundthe relation between MFI efficiency and productivity. Lafourcade et al. 2005 showthat African MFI-staffs are highly productive as the borrowers and savers numbersper staff member are highest. This high productivity may reflect their extensive use ofgroup lending as a means of achieving economies of scale.
The Grameen Bank in Bangladesh provides credit to group-borrowers and suchlending mechanism may help to increase MFI staff productivity. It has been claimed(Microbanking Bulletin 2004) that the Latin American MFIs were relatively moreefficient than all regions except Asia. This is true in case of the Latin American coop-eratives while their NGOs were the least efficient (for example, COAC, CMAC andWWB) which reflects economies of scale as increased loan size reduces their opera-tional expenses.
We follow the work of McCarty and Yaisawarng (1993) and apply a two stageanalysis where the MFIs have controllable and uncontrollable inputs. To our knowl-edge this is first such study that applies this two stage method on MFIs. This is thefirst study to measure the efficiency of MFIs using DEA on an international basisand so is important for several reasons. Firstly, individual MFI performance is mea-sured internationally by MFI type. This may assist aid agencies or governments thatfund MFIs (typically at no or low cost) to identify the most efficient in utilizing theirfunding. This is important as some claim that donor dependency breeds inefficiency(Farrington 2000). Secondly, as successful informal MFIs typically grow first intoNGO-MFIs and cooperative MFIs and then later into bank-MFIs and NBFI-MFIs,any differences between the relative efficiency of each MFI type will be identified.To the extent that these outcomes may reflect different regulatory standards, this mayprove of interest to policymakers.
3 Methodology
DEA is a piece-wise linear combination that connects the best practice observationsand forms a convex production possibility set. It was developed by Charnes et al.(1978) and applied to non-profit organizations where the objective of profit maximi-zation and cost minimization may not be considered as the vital factor. DEA also hasthe advantage of working with a small sample size and that does not require priceinformation. As Cooper et al. (2000) suggest, DEA generates cooperation betweenthe policymakers and practitioners in the choice of inputs and outputs and helps detectthe source and amount of inefficiency in each input and output.
3.1 The Model
In this section, we present the two models for measuring the cost efficiency of mi-crofinance institutions. The models differ in their treatment of uncontrollable inputs.Model 1 incorporates two-stage analysis. In the first stage it uses controllable inputsand computes the efficiency scores for the microfinance institutions. In the secondstage of analysis it unravels the effect of uncontrollable inputs from the technical effi-
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 71
ciency. Finally, Model 2 incorporates both uncontrollable and controllable variable asinputs in estimation of DEA efficiency scores.
The data envelopment method follows either a production approach or an interme-diation approach. The DEA method is based on the common assumption of constantreturn to scale6 (CRS) and the alternative assumption of variable return to scale (VRS).CRS provides the same efficiency results in both cases (Avkiran 1999; Coelli 1996).Although VRS model considers inputs increase, it does not expect increase in outputs.We apply both input and output oriented model using the production and intermedi-ation approach. So our analysis is on the basis of constant return to scale measure,Charnes et al. (1978) provided the following formula to measure efficiency.
Max
s∑
r=1
ur yr0
subject to :s∑
r=1
ur yr j −m∑
i=1
vi xi j ≤ 0
m∑
i=1
vi xio = 1; −ur ≤ −ε
−vi ≤ −ε(1)
where, yr j and xi j are positive known outputs and inputs of the jth decision makingunit (DMU) and ur and vi are the variable weights to be determined by solving theabove equation problem.
∑mi=1 vi xio = 1 ensures that it is possible to move from ratio
form to linear programming form and vice versa. ur , vi ≥ ε > 0,∀r, i, are from thenon-Archimedean conditions. This, as given in Charnes and Cooper (1962) insures:
h∗0 =
s∑
r=1
u∗r yr0 (2)
The Eq. (1) interprets as the MFIs objective is to maximize the output subject to unitinput and with the condition that virtual output cannot go beyond virtual input for anyDMU.
On the other hand, the following model provided by Banker et al. (1984) dealt withthe variable return to scale version of DEA.
maxs∑
r=1
ur yro − uo
subjects∑
r=1
ur yr j −m∑
i=1
vi xi j − uo ≤ 0
6 Constant return to scale compares each unit against all other units whereas the variable return to scalecompares each unit against other units of similar size (Avkiran 1999).
123
72 M. Haq et al.
m∑
i=1
vi xio = 1; −ur ≤ −ε
−vi ≤ −ε(3)
u∗ indicates the return to scale possibilities. An u∗ < 0 implies increasing return toscale and u∗ > 0 implies a decreasing return to scale.
Model 1 and Model 2 takes into consideration of the above methodology. However,the second stage analysis in Model 1, we use the DEA measure of efficiency scoresas a dependent variable, and regress on variables which are beyond the control of themicrofinance institutions. Since the efficiency scores are truncated from below at 1 weapply the Tobit model, which produces unbiased and consistent parameter estimationscompared to OLS regression method (McCarty and Yaisawarng 1993 and Dusanskyand Wilson 1995). Our Tobit model is specified as below:
T Ek = rkβ + µk if T E∗k > 1
= 1 if T E∗k ≤ 1 (4)
where, T Ek = technical efficiency of DMU k obtained from the first stage of Model1, T E∗
k = true but unobservable efficiency score for DMU k, rk = [1 zk] is an(1 ∗ (L + 1)) vector of uncontrollable factors plus one, β is an ((L + 1) ∗ 1)) vectorof parameters. µk is a random term, normally distributed. This represents the “puretechnical efficiency” after effects of uncontrollable factors have been unraveled.
Following McCarty and Yaisawarng (1993) we compute the DMU’s pure technical
efficiency∧µ, defined as below:
∧µ = T Ek − rk
∧β (5)
∧µ is not bounded from below at one, it can take a negative, a positive or a zero value. If∧µ is zero then DMU performs equally well compared to the average DMU with sameset of uncontrollable variables.
Our DEA score ranges between 0 and 1. Lower value of efficiency scores presents
poor performance of the DMU k. This implies lower the value of∧µ the better the
performance of DMU k.In Model 2 we ask the research question; given the controllable and uncontrollable
factors which is the most efficient DMU (microfinance institution). Hence Model 2including the uncontrollable factors is specified as follows
maxs∑
r=1
ur yro − uo
subjects∑
r=1
ur yr j −m∑
i=1
vi xi j − uo ≤ 0
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 73
m∑
i=1
vi xio = 1;−ur ≤ −ε
−vi ≤ −ε
−zi ≤ −ε
(6)
3.2 Sample
Our sample is extracted from the MIX Market database. This database providesinformation on 629 MFIs world-wide with 189 from Africa, 141 from South Asia,144 from Latin America and 66 from East Asia and the Pacific. However, our studyconsiders only those MFIs which provide not only credit facilities but also mobilizesavings from members. Second, these MFIs must also provide more than 50% of theirfinancial services (mainly loans and deposit services) to poor households. Consider-ing these factors, we have studies 39 MFIs which fitted with our criteria. This sampleconsists of 39 MFIs including 13 bank MFIs, 8 NBFI-MFIs, 6 cooperatives/creditunions MFIs and 11 NGO-MFIs, and 1 other non classified MFI from Africa, Asiaand Latin America. Finally, we choose 2004 as we find sufficient data for this yearonly to apply the data envelopment analysis. Table 1 contains information about theseMFI
It is evident from Table 1 that bank-MFIs constitutes dominant segment of the totalsample size) followed by unregulated semiformal institutions (NGO-MFIs). Micro-finance comprises from 70 to 100% of their business. Many also mobilize a consid-erable savings from both members and general public,7 for example, Bank RakyatIndonesia in Indonesia, CERUDEB and K-REP in Africa, CMAC in Latin Amer-ica and BRAC and WDFH in South Asia all show savings is their largest fundingsource.
3.3 Inputs and Outputs Specifications
The inputs and outputs variables used in our study and their detailed definitions areprovided in Table 2. Both the production and intermediation approach are used tomeasure the MFI efficiency. The production approach considers the MFI as a factoryproducing services to the poor households. So its outputs8 use the number of accountsrather than any dollar amount.
Several efficiency studies of commercial banks and bank branches, such as Shermanand Gold (1985), Ferrier and Lovell (1990), Berg et al. (1991), have used differentproduction outputs keeping the input (such as labor, fixed assets, capital) remainingthe same. Sherman and Gold (1985), for example, used the number of transactionas the output while Ferrier and Lovell (1990) used the total number of accounts andaccount size as the output.
7 These savings are not considered as a condition to receive current or future loans (MIX market database).8 Studies usually consider the number of transaction or number of branches as output. Our study is based onthe data availability in the Microfinance Information Exchange (MIX) database so we divided the numberof clients into savers per personnel and borrowers per personnel.
123
74 M. Haq et al.
Table 2 Selected inputs and outputs used under production approach and intermediation approach
Approaches Specification Definition
Production approach Inputsa. Labora Number of personnel/staffs which
is defined as the number ofindividuals actively employed bythe MFI. This includes contractemployees or advisors whodedicate the majority of time tothe MFI even if they are not onthe MFI’s roster of employees
b. Cost per borrower Operating expense/average numberof borrowers
c. Cost per saver Operating expense/average numberof savers
Outputs
a. Number of borrowers per staffmember
Total number of activeborrowers/total number ofpersonnel
b. Number of savers per staff member Total number of active savers/totalnumber of personnel
Intermediation approach Inputs
a. Total number of staffs/personnelaNumber of personnel/staffs which
is defined as the number ofindividuals actively employed bythe MFI. This includes contractemployees or advisors whodedicate the majority of time tothe MFI even if they are not onthe MFI’s roster of employees
b. Operating/administrativeexpenses
Administrative expenses excludinginterest expense
Outputs
a. Gross loan portfolio All outstanding principal for alloutstanding client loans includingcurrent, delinquent andrestructure loans but not loansthat have been written off. Itexcludes interest receivable andemployee loans
b. Total savings Total amount of voluntary savings
This table presents the choice of inputs and outputs used to measure the efficiency of different types ofMFIs. The inputs and outputs and their definitions are compiled from the MIXMarket (www.mixmarket.org,(accessed in Jan 2006)) database. The database provides information for individual microfinance institutionsin Africa, Asia and the Latin Americaa We use labor or total number of staffs as an input under both production approach and intermediationapproach. Hence the definition remains the same
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 75
Berg et al. (1991) measured activity in total loan and savings balances along withaverage size and number of accounts. Oral and Yolalan (1990) used five inputs (inphysical term) and four outputs to measure the time expended on various activities.Using a slightly different approach, Laeven (1999) and Pastor et al. (2002) introducedinterest expense and non-interest expense as the inputs in examining efficiency of 10European banks.
Our study uses the production approach with three (3) inputs: labor, cost per bor-rower and cost per saver; the later two inputs are the most important ingredients ofthe operating expense. The two outputs are savers per staff member and borrowers perstaff member. These provide productivity measurement of the MFIs which is indicatedby the number of borrowers per staff and savers per staff.
We consider labor as an input as it reflects time, effort and skills of the workforce. We use cost per borrower and cost per saver as inputs this reflects the oper-ating expenses. Thus, the MFIs would like to reduce cost and protect the borrow-ers-savers and maintain confidence, continue to develop, operate and serve a nichemarket that depends on their services. The overall productivity of the MFI’s staffin terms of managing clients, including borrowers, savers, voluntary savers andother clients. The outputs are the savers per staff member and borrowers per staffmember.
The alternative is the intermediation approach which considers MFIs as financialinstitutions providing financial services to the poor households. It considers the trans-formation of deposits and other loanable funds into credit. Berger and Humphrey(1992) argue that this approach fails to consider the cost of financial transactionsand savings deposits. Elyasiani and Mehdian (1990), Casu and Molynuex (2003) andIsik and Hassan (2003) used deposits as an input which is later on converted intocredit.
Since the MFIs provide voluntary savings facilities to their clients, we consid-ered their deposits as an output. Such consideration is supported by the studies doneby Aly et al. (1990), Berger and Humphrey (1992), Berg et al. (1993), Fried et al.(1993, 1999) and Hassan and Tufte (2001), Guitierrez-Nieto et al. (2006). We con-sidered gross loan portfolio and total savings as two outputs under the intermediationapproach. At the same time, we considered number of personnel and operating ex-penses as important inputs. We consider gross loan portfolio and savings as the outputbecause MFIs will be considered efficient if it has the capability to provide theseservices to it clients.
As our study involves a cross-country analysis, interest expenses and interest incomeare avoided due to the existence of complexities associated with inflation and interestrates in a number of countries.
4 Results
The efficiency of the MFIs is reported on a global scale under both the intermedia-tion and production approaches and the efficiency levels are compared within eachregion.
123
76 M. Haq et al.
4.1 Summary Statistics of Inputs and Outputs
The descriptive statistics for the input and output are provided region-wise in Table 3.We find that average gross loan portfolio and average savings are relatively higherwith the MFIs in the East Asian than their counterparts in the African MFIs.
The East Asian MFIs mobilized savings of some US$ 3.36 billion and operatingexpenses of US$ 0.24 billion. Their relative lowest operating expenses compared toother regions have helped them to rank second for savings mobilization. Further, therewere greater dispersion in total loan portfolio, total savings and operating expenses inEast Asia compared to the other regions.
Table 3 reports the highest average borrowers per staff member and savers per staffmember in Africa against the lowest average borrowers per staff member and sav-ers per staff member in Latin America and East Asia respectively. We find relativelyhigher borrowers per staff member and savers per staff member ratios in Africa andSouth Asia respectively. However, they are relatively lower in South Asia.
The two other variables are the cost per saver and cost per borrower. The maxi-mum cost per borrower and cost per saver are observed in Latin America and EastAsia respectively. For the sake of brevity, we do not report the amount of inputs andoutputs of individual MFIs but we do find Bank Rakyat Indonesia (BRI) mobilizesthe maximum savings and has the largest gross loan portfolio with highest operatingexpense. These demonstrate that East Asia dominates the efficiency frontier.
4.2 Efficiency Estimates of MFIs Using Model 1
The results on the efficiency estimate from running the linear programming model isobtained by considering the intermediation and production approach under the con-stant return to scale (CRS) and variable return to scale (VRS) measure. The CRS isappropriate when an MFI operates at an optimal scale and as such, this measure isnot appropriate to those MFIs which operate at sub-optimal level. The VRS measureis therefore applied to capture sub-optimal level of operations. The results based onestimates of Model 1, under both the intermediation and the production approachesare discussed in Sects. 4.2.1 and 4.2.2.
4.2.1 Intermediation Approach
Efficient MFIs are those which are able to expand their poverty outreach (output ori-ented) and able to achieve operational sustainability (input oriented). Here we discussthe result from the first stage of analysis. Table 4 shows that an African bank-MFI,DECSI, and Latin American, cooperative-MFI, COAC are the most technical efficientMFIs.9
9 Technical efficiency measures the ability of the firm to achieve maximum output for a given set of input.Scale efficiency can be referred to pure technical efficiency. This implies if CRS and VRS are conductedon same data and no difference is found in technical efficiency scores we can define this as pure technicalefficiency (Coelli 1996).
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 77
Table 3 Descriptive statistics of MFI inputs and outputs by region
Mean Maximum Minimum Standarddeviation
Africa
Intermediation approach (output)
Gross loan portfolio(in 000’ US$) 25362.29 46362.21 385.60 18348.32
Total savings(in 000’ US$) 21426.65 77740.99 31.56 27850.68
Intermediation approach (input)
Total operating expenses(in 000’ US$) 3867.56 15487.76 55.83 4617.44
Number of personnela 438 1670 39 498
Production approach (output)
Borrowers per staff members 198 405 71 112
Savers per staff members 266 779 16 259
Production approach (input)
Cost per borrower(in US$) 99 318 4.5 99
Cost per saver(in US$) 154 1175 7 342
East Asia
Intermediation approach (output)
Gross loan portfolio(in 000’ US$) 301092.31 2030015.05 322.03 762754.02
Total savings(in 000’ US$) 484328.64 3356136.42 1.08 1266401.42
Intermediation approach (input)
Total operating expenses(in 000’ US$) 35998.43 238521.06 103.37 89404.37
Number of personnela 4350 27903 34 10413
Production approach (output)
Borrowers per staff members 163 345 58 95
Savers per staff members 205 850 2 307
Production approach (input)
Cost per borrower(in US$) 41 106 10.3 35
Cost per saver(in US$) 888 5250 10 1929
South Asia
Intermediation approach (output)
Gross loan portfolio(in 000’ US$) 60284.79 337700.86 20.88 112252.38
Total savings(in 000’ US$) 28448.59 327944.01 3.10 86744.14
Intermediation approach (input)
Total operating expenses(in 000’ US$) 20799.71 170872.71 3.55 45480.27
Number of personnela 3190 18898 4 6141
Production approach (output)
Borrowers per staff members 156 307 32 90
Savers per staff members 238 1317 1 319
Production approach (input)
Cost per borrower(in US$) 28 208 6 53
Cost per saver(in US$) 102 1147 3 302
123
78 M. Haq et al.
Table 3 Continued
Mean Maximum Minimum Standarddeviation
Latin America
Intermediation approach (output)
Gross loan portfolio(in 000’ US$) 69396.03 177074.83 15102.97 64394.45
Total savings(in 000’ US$) 61346.82 180761.91 942.71 64979.61
Intermediation approach (input)
Total operating expenses(in 000’ US$) 9325.84 28790.85 929.56 9718.67Number of personnela 356 835 70 280
Production approach (output)
Borrowers per staff members 141 254 42 76
Savers per staff members 208 519 87 150
Production approach (input)
Cost per borrower(in US$) 226 384 60 111
Cost per saver(in US$) 185 423 30 138
This table presents the descriptive statistics of the inputs and outputs under intermediation approach andproduction approacha We use number of personnel/total number of staffs/labor as an input under both production approach andintermediation approach
However, in stage 1 we also conduct the variable return to scale measure. Our findingshow Bank Solidario, BRI and Grameen Bank are pure technical efficiency, under bothinput and output oriented estimations. The results are reported in the last four columnsof Table 4. Further, in the last four columns we also observe the NBFI-MFIs are lesssuccessful with the exception of CMAC from Latin America showing pure technicalefficiency. However, with the cooperatives/credit union-MFIs, COAC, an unregulatedMFI in Latin America, operates on the efficient frontier across all measures. Amongthe NGO-MFIs, two (2) in South Asia: (ASA in Bangladesh and Wilgamuwa in SriLanka) show pure technical efficiency while the remaining seven (7) NGO-MFIs areinefficient.
Table 4 shows that none of the efficient MFIs has any significant difference ininput-oriented and output-oriented DEA under the VRS measure. However, only theefficiency score of the inefficient MFIs differ between the two methods (Coelli 1996).
Furthermore, Table 5 shows that the overall mean efficiency score is the highest forthe cooperative-MFIs which is followed by bank-MFIs and NGO-MFIs.
This finding is further supported by Fig. 1, where we plot the average efficiencyscores to compare the level of efficiency for different types of MFIs. Figure A showsthe average efficiency scores among the MFIs under the intermediation approach. Thecooperative-MFIs lead in both CRS measure and input-oriented VRS measure andmarginally fall below the banks in output oriented VRS measure. However, the banksshow the overall efficiency score of 1 followed by the cooperatives. The overall leastefficiency score lies with the NGO-MFIs followed by bank-MFIs and cooperative-MFIs. Similarly, the bank-MFIs (input oriented DEA) and the NGO-MFIs (outputoriented DEA) have the maximum dispersion but the cooperative MFIs have the min-imum dispersion in respect of pure technical efficiency scores.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 79
Tabl
e4
Effi
cien
cysu
mm
ary
ofin
term
edia
tion
appr
oach
Mod
el1
Mic
rofin
ance
Inst
itutio
ns
Stag
e1
Stag
e2
Stag
e1
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
yIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Ban
ks-M
FIs
Ruh
unaD
evel
opm
ent
.215
28−0
.095
23.2
17.9
92.2
50.8
60
Nir
dhan
Utth
an.0
4137
−0.2
7535
.055
.753
.047
.873
Firs
tMic
rofin
ance
.177
31−0
.202
32.1
79.9
90.2
05.8
62
K-R
EP
.281
19−0
.050
18.3
24.8
68.5
25.5
36
Ban
kSo
lidar
io.6
604
0.15
56
1.00
0.6
601.
000
.660
CE
RU
DE
B.3
1818
0.01
415
.465
.683
.522
.609
AC
SI.5
518
0.23
64
.552
1.00
0.6
87.8
02
AC
LE
DA
.201
29−0
.130
27.3
54.5
67.4
92.4
08
Ban
coSo
l.5
259
0.02
014
.842
.624
.879
.598
CB
B.2
4525
−0.0
7121
.277
.885
.253
.971
Ban
kR
akya
t.5
696
0.12
67
1.00
0.5
691.
000
.569
Gra
mee
nB
ank
.473
100.
125
81.
000
.473
1.00
0.4
73
DE
CSI
1.00
0n.
a.n.
a.n.
a.1.
000
1.00
01.
000
1.00
0
NB
FI-
MF
Is
EB
S.3
3317
0.00
216
.533
.625
.594
.560
AM
K.0
8835
−0.2
4334
.098
.903
.089
.995
CM
AC
.855
10.
326
31.
000
.855
1 .00
0.8
55
CM
F.1
0934
−0.1
9531
.138
.787
.136
.799
123
80 M. Haq et al.
Tabl
e4
cont
inue
d
Mic
rofin
ance
Inst
itutio
ns
Stag
e1
Stag
e2
Stag
e1
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
yIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Caj
aN
or.5
617
0.03
210
.561
1.00
0.6
2.9
06
SEE
DS
.39
120.
080
9.3
91.9
97.4
43.8
80
FIN
CO
MU
N.1
2833
−0.4
2637
.132
.966
.196
.651
AV
FS.1
9130
−0.1
2524
.254
.751
.202
.946
Coo
pera
tive
-MF
Is
VY
CC
U.7
942
0.47
91
.943
.842
.933
.851
CO
AC
1.00
0n.
a.n.
a.n.
a.1.
000
1.00
01.
000
1.00
0
CM
S.5
965
0.19
45
.76
.784
.805
.740
AC
CO
VI
.430
11−0
.047
17.4
35.9
90.5
97.7
21
AC
EP
.750
30.
348
2.7
81.9
61.8
06.9
31
KC
.372
15−0
.128
26.4
00.9
30.3
82.9
74
NG
O-M
FIs
NSS
C.2
4922
−0.0
6719
.299
.834
.262
.952
CE
P.2
7920
−0.0
7222
.283
.985
.279
.999
TY
MFU
ND
.376
130.
025
12.3
90.9
64.3
81.9
87
BR
AC
.223
27−0
.125
25.6
07.3
68.6
71.3
33
ASA
.376
130.
028
111.
000
.376
1.00
0.3
76
CM
ED
FI.0
7836
−0.4
2236
.109
.709
.080
.972
AC
CD
C.3
3416
0.02
413
.426
.783
.367
.909
WD
FH.2
426
−0.0
7020
.26
.922
.244
.983
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 81
Tabl
e4
cont
inue
d
Mic
rofin
ance
Inst
itutio
ns
Stag
e1
Stag
e2
Stag
e1
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
yIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Kas
hf.2
4922
−0.1
3027
.253
.988
.254
.982
WA
GE
S.2
5821
−0.1
4429
.266
.971
.259
.996
Wilg
amuw
a.1
6332
−0.1
4730
1.00
0.1
631.
000
.163
CD
S.2
4922
−0.2
1533
.287
.866
.259
.959
Ove
rall
aver
age
0.38
3−0
.031
.510
.805
.531
.786
Min
imum
0.04
1−.
426
0.05
50.
163
0.04
70.
163
Max
imum
1.4
791
11
1
Thi
sta
ble
repo
rts
the
resu
ltsof
the
DE
Aef
ficie
ncy
scor
esun
der
inte
rmed
iatio
nap
proa
ch.
We
draw
aco
mpa
riso
nof
the
effic
ienc
ysc
ores
amon
gsa
mpl
eba
nk-M
FIs,
NB
FI-M
FIs,
Coo
pera
tive-
MFI
san
dN
GO
-MFI
sac
ross
Afr
ica,
Asi
aan
dL
atin
Am
eric
a
123
82 M. Haq et al.
Table 5 Descriptive statistics of efficiency scores (intermediation approach)
Banks-MFIs NBFI-MFIs Cooperative-MFIs NGO-MFIs
Mean
Input/output oriented (TE) .4043 .332 .538 .240
Input oriented (PTE) .559 .388 .587 .490
Output oriented (PTE) .605 .41 .605 .484
Overall mean score .523 .377 .577 .405
Minimum
Input/output oriented (TE) .041 .088 .249 .078
Input oriented (PTE) .055 .098 .283 .019
Output oriented (PTE) .047 .089 .262 .08
Overall minimum score .041 .088 .249 .019
Maximum
Input/output oriented (TE) 1 .855 1 .376
Input oriented (PTE) 1 1 1 1
Output oriented (PTE) 1 1 1 1
Overall maximum score 1 1 1 1
Standard deviation
Input/output oriented (TE) .258 .267 .261 .093
Input oriented (PTE) .360 .306 .282 .347
Output oriented (PTE) .348 .314 .289 .360
This table presents the descriptive statistics of the efficiency scores of different types of MFIs across thesample regions. The mean, minimum, maximum and standard deviation are reported for input and outputoriented technical efficiency at constant return to scale (CRS) and for input and output oriented pure technicalefficiency under variable return to scale (VRS)PTE Pure technical efficiency, TE Technical efficiency
We now turn our discussion to the second stage where we incorporate a socialeconomic factor: percentage of rural population. Hence, MFI identified as inefficientin the above discussion may not infact reflect that they are unable to reach the poorhousehold and achieve sustainability. Rather they may face a difficult challenge whichmay be beyond the MFIs control. For example, it may be difficult for the MFIs to reachall the rural population simply because there may be a group of household which mayreflect moral hazard problem. Moral hazard problem/adverse selection arises as theMFIs are unable to identify the group of borrowers’ actual intention of repaying theloan. This uncontrollable factor can lead to waste and mismanagement of resourcesfor a MFI.
Table 4 draws the difference in the rank orderings of the efficiency scores betweenfirst stage and second stage. We find that CMAC drops to third ranking from first in thesecond stage of analysis and VYCCU improves from a second place to the first place.Hence, we may have considered CMAC to be on the best practice frontier but as soonas we control for the rural population percentage, we observe CMAC’s achievementlevels are not sufficiently higher than expected given its inputs and uncontrollablefactor. K-REP, provides an example of better MFI despite being an inefficient MFI,the drop in inefficiency is relatively lower compared to the other inefficient MFIs.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 83
Comparison of Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
NGOsCooperativesNBFIsBanks
microfinance institutions
aver
age
effi
cien
cy s
core
TEInput oriented PTEOutput oriented PTE
Comparison of efficiency scores
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
aver
age
effi
cien
cy s
core
s TEPTE(input)PTE(output)
NGOsCooperativesNBFIsBanksmicrofinance institutions
A
B
Fig. 1 a Comparison of efficiency scores (intermediation approach). b Comparison of efficiency scores(production approach)
4.2.2 Production Approach
The efficiency summary of production approach in Table 6 is quite different from theintermediation approach. We first discuss stage 1 (column 1) and then discuss the puretechnical efficiency scores that we estimate under stage 1 (last four columns). Whileboth technical (column 1) and pure technical efficiencies (last four columns) are foundwith the three (3) MFIs (Ruhuna, DECSI and Wilgamuwa), only pure technical effi-ciency is found with the two (2) MFIs (ASA and COAC). A number of MFIs appearsto be efficient under production approach and these are; for instance, CEP in Vietnam,ACCDC and WDFH in Sri Lanka, CMF and WAGES in Africa.
The pure technical efficient frontier is dominated by South Asian NGO-MFIs. Largebank-MFIs like BRI, Banco Solidario and Grameen Bank, which were efficient underintermediation approach, are now inefficient under production approach. Yet, bank-MFIs such as Ruhuna, DECSI, and NGO-MFIs such as CEP and Wilgamuwa are allefficient under production approach. Under the input oriented VRS measure, FINCO-
123
84 M. Haq et al.
Tabl
e6
Effi
cien
cysu
mm
ary
ofpr
oduc
tion
appr
oach
usin
gM
odel
1
MFI
sSt
age
1St
age
2St
age
1O
utpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Inpu
t/out
put
orie
nted
CR
Ste
chni
cale
ffici
ency
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
put
orie
nted
VR
Spu
rete
chni
cal
effic
ienc
y
Ban
k-M
FIs
Ruh
una
Dev
elop
.1.
000
n.a.
n.a.
n.a.
1.00
01.
000
1.00
01.
000
Nir
dhan
Utth
an.
.338
14−0
.115
23.4
81.7
03.4
83.7
Firs
tMic
rofin
ance
.020
34−0
.252
31.0
41.4
72.1
90.1
03
K-R
EP
.102
26−0
.291
33.1
07.9
54.4
30.2
38
Ban
kSo
lidar
io.0
3331
−0.1
1421
.033
.993
.402
.082
CE
RU
DE
B.1
1325
−0.3
9536
.126
.901
.411
.276
AC
SI.4
516
−0.0
0210
.833
3.5
41.5
41.8
34
AC
LE
DA
.019
35−0
.374
35.0
63.3
05.1
43.1
35
Ban
coSo
l.0
2532
−0.1
2225
.032
.761
.299
.082
CB
B.4
1510
−0.0
3814
.478
.867
.76
.546
Ban
kR
akya
t.1
5223
−0.0
3815
.167
.913
.645
.236
Gra
mee
nB
ank
.438
80.
095
8.5
56.7
89.7
86.5
58
DE
CSI
1.00
0n.
a.n.
a.n.
a.1.
000
1.00
01.
000
1.00
0
NB
FI-
MF
Is
EB
S.2
4419
−0.1
4926
.314
.777
.591
.413
AM
K.3
8912
−0.0
0412
.533
.730
.733
.531
CM
AC
.066
27−0
.069
18.1
14.5
82.5
51.1
21
CM
F.3
3415
−0.1
7428
1.00
0.3
341.
000
.334
Caj
aN
or.0
2433
−0.1
1119
.032
.748
.309
.078
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 85
Tabl
e6
cont
inue
d
MFI
sSt
age
1St
age
2St
age
1O
utpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Inpu
t/out
put
orie
nted
CR
Ste
chni
cale
ffici
ency
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
put
orie
nted
VR
Spu
rete
chni
cal
effic
ienc
y
SEE
DS
.120
24−0
.358
34.3
13.3
82.2
31.5
18
FIN
CO
MU
N.0
1036
−0.1
1422
.020
.510
.200
.052
AV
FS.3
9111
−0.0
6217
.497
.785
.588
.665
Coo
pera
tive
MF
Is
VY
CC
U.2
6717
−0.1
8629
.412
.650
.567
.471
CO
AC
.256
180.
109
71.
000
.256
1.00
0.2
56
CM
S.1
8522
−0.0
5116
.356
.520
.505
.366
AC
CO
VI
.042
30−0
.122
24.0
56.7
54.2
74.1
53
AC
EP
.046
29−0
.190
30.0
46.9
85.3
78.1
21
KC
.474
50.
325
4.8
04.5
89.9
3.5
1
NG
O-M
FIs
NSS
C.4
487
−0.0
0513
.535
.837
.675
.663
CE
P.7
452
0.41
12
1.00
0.7
451.
000
.745
TY
MFU
ND
.578
40.
244
5.6
80.8
49.7
12.8
12
BR
AC
.340
13−0
.003
11.6
25.5
44.5
42.6
27
ASA
.787
10.
444
11.
000
.787
1.00
0.7
87
CM
ED
FI.1
9721
0.04
89
.241
.816
.504
.39
AC
CD
C.3
1116
−0.1
6727
1.00
0.3
111.
000
.311
WD
FH.2
0420
−0.2
7432
1.00
0.2
041.
000
.204
123
86 M. Haq et al.
Tabl
e6
cont
inue
d
MFI
sSt
age
1St
age
2St
age
1O
utpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Inpu
t/out
put
orie
nted
CR
Ste
chni
cale
ffici
ency
Ran
kR
esid
ual
∧ µR
ank
Inpu
tori
ente
dV
RS
pure
tech
nica
lef
ficie
ncy
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
put
orie
nted
VR
Spu
rete
chni
cal
effic
ienc
y
Kas
hf.4
339
0.16
16
.474
.914
.664
.652
WA
GE
S.6
443
0.40
83
1.00
0.6
441.
000
.644
Wilg
amuw
a1.
000
n.a.
n.a.
n.a.
1.00
01.
000
1.00
01.
000
CD
S.0
6128
−0.1
1120
.065
.940
.438
.140
Ove
rall
aver
age
0.33
−0.0
550.
490.
700.
630.
45
Min
imum
0.01
−0.3
950.
020.
200.
140.
05
Max
imum
10.
444
11
11
Thi
sta
ble
repo
rts
the
resu
ltsof
the
DE
Aef
ficie
ncy
scor
esun
derp
rodu
ctio
nap
proa
ch.W
edr
awa
com
pari
son
ofth
eef
ficie
ncy
scor
esam
ong
sam
ple
bank
-MFI
s,N
BFI
-MFI
s,C
oope
rativ
e-M
FIs
and
NG
O-M
FIs
acro
ssA
fric
a,A
sia
and
Lat
inA
mer
ica
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 87
Table 7 Descriptive statistics of efficiency scores (production approach)
Bank-MFIs NBFI-MFIs Cooperative-MFIs NGO-MFIs
Mean
Input/output oriented (TE) .316 .197 .338 .490
Input oriented (PTE) .378 .353 .543 .793
Output oriented (PTE) .545 .525 .671 .839
Overall mean efficiency .413 .358 .517 .707
Minimum
Input/output oriented (TE) .019 .01 .042 .197
Input oriented (PTE) .032 .02 .046 .241
Output oriented (PTE) .143 .20 .274 .504
Overall minimum score .019 .01 .042 .197
Maximum
Input/output oriented (TE) 1 .391 .745 1
Input oriented (PTE) 1 1 1 1
Output oriented (PTE) 1 1 1 1
Overall maximum score 1 1 1 1
Standard deviation
Input/output oriented (TE) .346 .162 .240 .292
Input oriented (PTE) .373 .326 .361 .304
Output oriented (PTE) .279 .272 .266 .227
This table presents the descriptive statistics of the efficiency scores of different types of MFIs across thesample regions under production approach. The mean, minimum, maximum and standard deviation arereported for both input and output oriented technical efficiency at constant return to scale (CRS) and forinput and output oriented pure technical efficiency under variable return to scale (VRS)PTE Pure technical efficiency, TE Technical efficiency.
MUN is the least efficient. In order to be efficient, these MFIs should reduce theirinputs by 98% as done by DECSI and Wilgamuwa.
Similarly, the output-oriented measure shows that ACLEDA in Cambodia is theleast efficient MFI. As shown on Table 7, the NGO-MFIs have the highest overallmean efficiency score followed by the cooperative-MFIs. The bank-MFIs, however,are better than the NBFI-MFIs.
Among the bank-MFIs and NGO-MFIs, there is at least one efficient MFI underboth CRS and VRS measures. There is highest dispersion in the bank-MFIs’ efficiencyscore. Figure B shows that NGO-MFIs are the most productive under all measuresfollowed by the cooperative-MFIs. The least efficient are the non-bank MFIs.
From Table 6 we can see that there are not many differences in ranking amongthe MFIs between first stage and second stage analysis. However, we find the lowestdrop among the inefficient MFIs is for ACSI, however this leads to a drop of rankingfrom 6th to 10th position. We also find Grameen bank, ACLEDA, AMK, CEP, ASAand Wages are non-movers and hence remain at the same position for both technicalefficiency and pure technical efficiency or “residual”. We can also observe that SEEDS
123
88 M. Haq et al.
show a 41% drop in their position while FINCOMUN has climbed to position 22 frombottom 36.
4.3 Efficiency Estimates of MFIs Using Model 2
Tables 8 and 9 present efficiency scores and their rank ordering from Model 2, in whichboth controllable and uncontrollable inputs are incorporated. Our findings show thatthe magnitudes of the efficiency scores are higher in Model 2 compared to Model 1.10
Our uncontrollable variable is the percentage of rural population to total population.This may also represent the urbanization rate for each region.
Table 8 presents the result of the efficiency scores under intermediation approach.The mean score of technical efficiency under constant return to scale is approximately50%, however since we consider that MFIs do not operate in optimal level so we alsoreport the variable return to scale results for the pure technical efficiency and scaleefficiency under both output and input oriented and the mean score ranges between0.65 and 0.86.
The rank orderings are quite similar to those based on residual values∧µ in Model 1.
The Pearson correlation coefficient is 84%, indicates that the two rank ordering arepositively correlated at 1% significance level. Comparing individual rankings betweenmodel 1 and model 2 we find remarkable difference which is the change in rankingfor ASA, BRI, Grameen Bank, and CMAC. These are now ranked as 1 or the mostefficient. However, the peer summary reflects that these DMUs are efficient by defaultwhich means that each DMU uses a unique combination of inputs and outputs suchthat it is compared only to itself when the efficiency score is calculated. Thus wecannot consider them to be the role model for the inefficient DMUs.
Table 9 presents the result of the efficiency scores under production approach. Themean score of technical efficiency under constant return to scale is approximately 64%,however the mean efficiency score for pure technical efficiency and scale efficiencyunder variable return to scale, both output and input oriented; ranges between 0.73and 0.88.
The rank orderings are quite similar to those based on residual values∧µ in Model 1.
The Pearson correlation coefficient is 76%, indicates that the two rank ordering arepositively correlated at 1% significance level. Comparing individual rankings betweenModel 1 and Model 2, there appears to be some differences. Bank Rakyat Indonesia,CMAC, CMF and COAC have improved in their efficiency rankings. They are now themost efficient with a score of 1. However, based on the peer analysis we find that theseMFIs cannot be identified as the role model for the inefficient MFIs as they utilize aunique combination of input and output such that it is only compared to itself.
10 Model 1 where we do not incorporate the uncontrollable variable as an input.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 89
Tabl
e8
Effi
cien
cysu
mm
ary
ofin
term
edia
tion
appr
oach
Mod
el2
MFI
sIn
put/o
utpu
tor
ient
edC
RS
tech
nica
leffi
cien
cy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
put
orie
nted
VR
Ssc
ale
effic
ienc
y
Ban
ks-M
FIs
Ruh
unaD
evel
opm
ent
0.23
529
0.35
30.
664
0.25
0.93
7
Nir
dhan
Utth
an0.
043
370.
374
0.11
60.
047
0.92
3
Firs
tMic
rofin
ance
0.17
732
0.52
50.
337
0.20
50.
862
K-R
EP
0.38
918
0.42
40.
916
0.52
50.
741
Ban
kSo
lidar
io1
11
11
1
CE
RU
DE
B0.
446
150.
465
0.95
90.
522
0.85
5
AC
SI0.
658
110.
659
0.99
90.
687
0.95
8
AC
LE
DA
0.44
316
0.44
70.
991
0.49
20.
9
Ban
coSo
l0.
862
60.
868
0.99
40.
879
0.98
1
CB
B0.
245
270.
616
0.39
80.
253
0.97
1
Ban
kR
akya
t1
11
11
1
Gra
mee
nB
ank
11
11
11
DE
CSI
1n.
a.1
11
1
NB
FI-
MF
Is
EB
S0.
505
140.
533
0.94
70.
594
0.85
AM
K0.
088
350.
446
0.19
80.
089
0.99
5
CM
AC
11
11
11
CM
F0.
109
340.
397
0.27
40.
136
0.79
9
Caj
aN
or0.
645
121
0.64
51
0.64
5
SEE
DS
0.39
170.
460.
849
0.44
30.
88
123
90 M. Haq et al.
Tabl
e8
cont
inue
d
MFI
sIn
put/o
utpu
tor
ient
edC
RS
tech
nica
leffi
cien
cy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
put
orie
nted
VR
Ssc
ale
effic
ienc
y
FIN
CO
MU
N0.
2130
10.
211
0.21
AV
FS0.
191
310.
781
0.24
40.
210.
907
Coo
pera
tive
-MF
Is
VY
CC
U0.
794
71
0.79
40.
998
0.79
5
CO
AC
1n.
a.1
11
1
CM
S0.
703
90.
760.
926
0.80
50.
873
AC
CO
VI
0.55
713
0.79
50.
70.
597
0.93
3
AC
EP
0.75
80.
781
0.96
10.
806
0.93
1
KC
0.37
220
10.
372
10.
372
NG
O-M
FIs
NSS
C0.
249
240.
776
0.32
10.
270.
925
CE
P0.
279
220.
479
0.58
30.
279
0.99
9
TY
MFU
ND
0.37
619
0.63
70.
590.
405
0.92
8
BR
AC
0.67
100.
70.
958
0.67
10.
999
ASA
11
11
11
CM
ED
FI0.
078
360.
998
0.07
80.
160.
485
AC
CD
C0.
334
210.
880.
379
0.36
70.
909
WD
FH0.
2428
0.55
40.
432
0.24
40.
983
Kas
hf0.
249
240.
533
0.46
80.
254
0.98
2
WA
GE
S0.
258
230.
596
0.43
30.
259
0.99
5
Wilg
amuw
a0.
163
331
0.16
31
0.16
3
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 91
Tabl
e8
cont
inue
d
MFI
sIn
put/o
utpu
tor
ient
edC
RS
tech
nica
leffi
cien
cy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
put
orie
nted
VR
Ssc
ale
effic
ienc
y
CD
S0.
249
240.
778
0.32
0.25
90.
959
Ove
rall
aver
age
.50
.58
.86
.73
.65
Min
imum
.043
.353
.078
.047
.163
Max
imum
11
11
1
Thi
sta
ble
repo
rts
the
resu
ltsof
the
DE
Aef
ficie
ncy
scor
esun
der
inte
rmed
iatio
nap
proa
ch.
We
draw
aco
mpa
riso
nof
the
effic
ienc
ysc
ores
amon
gsa
mpl
eba
nk-M
FIs,
NB
FI-M
FIs,
Coo
pera
tive-
MFI
san
dN
GO
-MFI
sac
ross
Afr
ica,
Asi
aan
dL
atin
Am
eric
a
123
92 M. Haq et al.
Tabl
e9
Effi
cien
cysu
mm
ary
ofpr
oduc
tion
appr
oach
usin
gM
odel
2
MFI
sIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
y
Ban
ks-M
FIs
Ruh
una
Dev
elop
.1
n.a.
1a1
1a1
Nir
dhan
Utth
an.
0.47
426
0.82
40.
575
0.48
30.
981
Firs
tMic
rofin
ance
0.13
435
0.51
70.
259
0.19
90.
673
K-R
EP
0.36
529
0.47
10.
775
0.45
60.
801
Ban
kSo
lidar
io0.
641
170.
828
0.77
40.
642
0.99
8
CE
RU
DE
B0.
428
0.46
70.
858
0.41
10.
975
AC
SI0.
507
240.
925
0.54
90.
541
0.93
8
AC
LE
DA
0.10
836
0.43
60.
248
0.14
90.
724
Ban
coSo
l0.
476
250.
826
0.57
60.
476
0.99
8
CB
B0.
7610
0.81
40.
934
0.76
1
Ban
kR
akya
t1
11a
11a
1
Gra
mee
nB
ank
0.82
88
0.96
20.
861
0.84
10.
985
DE
CSI
1n.
a.1a
11a
1
NB
FI-
MF
Is
EB
S0.
643
160.
677
0.95
0.64
40.
999
AM
K0.
656
150.
732
0.89
70.
733
0.89
6
CM
AC
11
1a1
1a1
CM
F1
11a
11a
1
Caj
aN
or0.
535
221a
0.53
51a
0.53
5
SEE
DS
0.21
533
0.68
10.
316
0.23
10.
932
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 93
Tabl
e9
cont
inue
d
MFI
sIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
y
FIN
CO
MU
N0.
423
271a
0.42
31a
0.42
3
AV
FS0.
556
210.
864
0.64
30.
588
0.94
5
Coo
pera
tive
-MF
Is
VY
CC
U0.
562
200.
861
0.65
20.
570.
985
CO
AC
11
1a1
1a1
CM
S0.
639
180.
675
0.94
70.
642
0.99
5
AC
CO
VI
0.31
931
0.81
20.
393
0.32
10.
993
AC
EP
0.36
330
0.57
20.
635
0.47
40.
767
KC
11
1a1
1a1
NG
O-M
FIs
NSS
C0.
675
140.
880.
767
0.67
50.
999
CE
P1
11a
11a
1
TY
MFU
ND
0.75
111
1a0.
751
1a0.
751
BR
AC
0.57
319
0.96
0.59
70.
586
0.97
8
ASA
0.78
79
1a0.
787
1a0.
787
CM
ED
FI0.
731
120.
999
0.73
20.
921
0.79
4
AC
CD
C0.
311
321a
0.31
11a
0.31
1
WD
FH0.
204
341a
0.20
41a
0.20
4
Kas
hf0.
7213
0.92
30.
780.
742
0.97
WA
GE
S1
11a
11a
1
Wilg
amuw
a1
n.a.
1a1
1a1
123
94 M. Haq et al.
Tabl
e9
cont
inue
d
MFI
sIn
put/o
utpu
tori
ente
dC
RS
tech
nica
lef
ficie
ncy
Ran
kIn
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Inpu
tori
ente
dV
RS
scal
eef
ficie
ncy
Out
puto
rien
ted
VR
Spu
rete
chni
cal
effic
ienc
y
Out
puto
rien
ted
VR
Ssc
ale
effic
ienc
y
CD
S0.
5123
0.77
90.
655
0.52
20.
978
Ove
rall
aver
age
.638
.859
.728
.734
.880
Min
imum
.108
0.43
60.
204
.149
.204
Max
imum
11
11
11
Thi
sta
ble
repo
rts
the
resu
ltsof
the
DE
Aef
ficie
ncy
scor
esun
derp
rodu
ctio
nap
proa
ch.W
edr
awa
com
pari
son
ofth
eef
ficie
ncy
scor
esam
ong
sam
ple
bank
-MFI
s,N
BFI
-MFI
s,C
oope
rativ
e-M
FIs
and
NG
O-M
FIs
acro
ssA
fric
a,A
sia
and
Lat
inA
mer
ica
aIn
dica
tes
the
MFI
isef
ficie
ntby
defa
ultt
hati
sit
has
aun
ique
com
bina
tion
ofin
puta
ndou
tput
sre
lativ
eto
othe
rsin
the
sam
ple,
thus
isco
mpa
red
only
toits
elf
inm
easu
ring
the
degr
eeof
prod
uctiv
eef
ficie
ncy
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 95
5 Conclusion
This study investigated the cost efficiency of MFIs (bank-MFIs, NBFI-MFIs, coop-erative-MFIs and NGO-MFIs) in Africa, Asia, and the Latin America using the dataenvelopment analysis (DEA). The MFIs were compared using the intermediation andproduction approaches to identify which MFI type is the most efficient in minimizingcosts and providing financial services to poor households.
Our findings show that under the intermediation approach four (4) out of thirteen(13) bank-MFIs are both input and output oriented, pure technical efficient and scaleefficient. Under production approach, six (6) out of twelve (12) NGO- MFIs are foundto be the most efficient. We find more MFIs show VRS pure technical efficiencythan either CRS technical efficiency or VRS pure technical efficiency under boththe intermediation and production approach. Five (5) out of twelve (12) NGO-MFIsare pure technical efficient under production approach while three (3) out of thirteen(13) bank-MFIs are pure technical efficient under intermediation approach. Only one(1) bank-MFI and one (1) cooperative-MFI under intermediation approach and two(2) bank-MFIs and one (1) NGO-MFI under production approach are CRS technicalefficient and VRS pure technical efficient.
The results discussed above may suggest that high level of cost efficiency may havedecreased due to the amount of non-performing loans specifically for bank-MFIs underthe intermediation approach. In other words, cost efficient managers are better manag-ing their loan customers and properly monitoring MFIs’ operating costs. Furthermore,the levels of efficiency have much more to do with efficient utilization of resourcesrather than scale of production.
In conclusion, we can suggest NGO-MFIs may be promoted in developing regionsas these MFIs are found to be the most efficient under production approach. This resultis not surprising given the NGO-MFIs’ dual objectives of alleviating poverty throughincreased out reach and simultaneously achieving financial sustainability. Over theyears NGO-MFIs have learnt to develop staff productivity, to increase branching anddistribution system, to build outstanding portfolio quality and to extend relationshipbanking culture to the poor. As these institutions are mostly either unregulated or lessregulated than other MFIs, policymakers should approach further NGO-MFI regula-tion with care so that this efficiency is not hampered. Nevertheless, some bank-MFIsare quite efficient in providing microfinance particularly DECSI in Africa. As morebank-MFIs are established they may have the competitive advantage as financial inter-mediaries in areas like access to local capital as well as the global financial markets.So while bank-MFIs are the most efficient type of MFI at present under intermedi-ation approach, NGO-MFIs may eventually perform better as an intermediary in thelong–run if proper regulation and supervision are in place.
Acknowledgment We thank an anonymous referee for helpful and substantial comments. We also thankJiro Akahori (Editor). The usual caveats apply.
References
Aly, H. Y., Grabowski, R., Pasurka, C., & Rangan, N. (1990). Technical, scale and allocative efficienciesin U.S. banking: an empirical investigation. Review of Economics and Statistics, 72(2), 211–218.
123
96 M. Haq et al.
Avkiran, N. K. (1999). Technical evidence on efficiency gains: The role of mergers and the benefits tothe public. Journal of Banking and Finance, 23, 991–1013.
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scaleinefficiencies in data envelopment analysis. Management Science, 30(9), 1078–1092.
Baumann, T. (2005). Pro poor microcredit in South Africa: Cost efficiency and productivity of SouthAfrican pro-poor microfinance institutions. Journal of Microfinance, 7(1), 95–118.
Berg, S. A., Førsund, F. R., & Jansen, E. S. (1991). Technical efficiency of Norwegian banks: The non-parametric approach to efficiency measurement. The Journal of Productivity Analysis, 2, 127–142.
Berg, S. A., Førsund, F. R., Hjalmarsson, L., & Souminen, M. (1993). Banking efficiency in the Nordiccountries. Journal of Banking and Finance, 17, 371–388.
Berger, A., & Humphrey, D. B. (1992). Mega mergers in banking and the use of cost efficiency as anantitrust defense. Antitrust Bulletin, 37, 541–600.
Brau, J. C., & Woller, G. M. (2004). Microfinance: A comprehensive review of the existing litera-ture. Journal of Entrepreneurial Finance and Business Ventures, 9, 1–26.
Casu, B., & Molynuex, P. A. (2003). Comparative study of efficiency in European banking. AppliedEconomics, 35, 1865–1876.
Charnes, A., & Cooper, W. W. (1962). Programming with linear fractionals. Naval Research Quar-terly, 9, 181–186.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision makingunits. European Journal of Operational Research, 2(6), 429–444.
Coelli, T. (1996). A guide to DEAP version 2.1: a data envelopment analysis (computer) program.Centre for efficiency and productivity analysis (CEPA). www.une.edu.au/econometrics/cepa.html.(accessed in Jan 2006).
Cooper, W. W., Seifordc, L. M., & Tone, K. (2000). Data envelopment analysis a comprehensivetext with models, applications, references and DEA-solver software. Boston: Kluwer AcademicPublishers.
Desrochers, M., & Lamberte, M. (2003). Efficiency and expense preference behavior in Philippines’cooperative rural banks. Centre interuniversitairesur les risque, les politiques economiques etl’emploi (CIRPÉE.) Cahier de recherche/Working paper 03-21.
Dusansky, R., & Wilson, P. (1995). On the relative efficiency of alternative modes of producing apublic sector output: The case of the developmentally disabled. European Journal of OperationalResearch, 80, 608–618.
Elyasiani, E., & Mehdian, S. M. (1990). A nonparametric approach to measurement of efficiency andtechnological change: The case of large U.S. commercial banks. Journal of Financial ServicesResearch, 4, 157–168.
Farrington, T. (2000). Efficiency in microfinance institutes. Microbanking Bulletin, 20–23.Ferrier, G. D., & Lovell, C. A. K. (1990). Measuring cost efficiency in banking: Econometric and linear
programming evidence. Journal of Econometrics, 46, 229–245.Fried, H. O., Lovell, C. A. K., & Eeckaut, P. V. (1993). Evaluating the performance of U.S. credit
unions. Journal of Banking and Finance, 17, 251–265.Fried, H. O., Lovell, C. A. K., & Yaisawarng, S. (1999). The impact of mergers on credit union service
provision. Journal of Banking and Finance, 23, 367–386.Guitierrez-Nieto, B., Serrano-Cinca, C., & Molinero, C. M. (2006). Microfinance institutions and
efficiency. OMEGA, International Journal of Management Science, 35(2), 131–142.Hassan, M. K., & Tufte, D. R. (2001). The X-efficiency of a group based lending institution: The case
of Grameen Bank. World Development, 29, 1071–1082.Isik, I., & Hassan, M. K. (2003). Financial deregulation and total factor productivity change: An
empirical study of Turkish commercial banks. Journal of Banking and Finance, 27, 1455–1485.Laeven, L. (1999). Risk and efficiency in East Asian banks. Policy Research. Working Paper 2255.
The World Bank. Financial Sector Strategy and Policy Department.Lafourcade, A., Isern, J., Mwangi, P., & Brown, M. (2005). Overview of the outreach and financial
performance of microfinance institutions on Africa. www.mixmarket.org. (accessed in Jun. 2006).Leon J. V. (2001). Decentralized efficient organizations of microfinance: The case of the Peruvian
municipal banks. Working paper series. Wittenberg University, Ohio.McCarty T. A., & Yaisawarng S. (1993). Technical efficiency in New Jersey school districts. In H.
O. Fried, C. A. K. Lovell, & S. S. Schmidt (Eds.), The measurement of productive efficiencytechniques and applications (pp. 272–287). USA: Oxford University Press.
123
Efficiency of Microfinance Institutions: A Data Envelopment Analysis 97
Maxwell, S. (1999). The meaning and measurement of poverty. Overseas Development Institute PovertyBriefing, 3, 1–6.
Microbanking Bulletin. (2005). Performance and Transparency. A survey of microfinance in South Asia.The MIX market, www.mixmarket.org (accessed in Jun 2006).
Microbanking Bulletin. (2004). Benchmarking Latin American microfinance. The MIX Market, www.mixmarket.org. (accessed in Jan 2006).
Murdoch, J. (2000). The microfinance schism. World Development, 28, 617–629.Oral, M., & Yolalan, R. (1990). An empirical study on measuring operating efficiency and profitability
of bank branches. European Journal of Operational Research, 46, 282–294.Pastor, J. T., Lozano, A., & Pastor, J. M. (2002). An efficiency comparison of European banking systems
operating under different environmental conditions. Journal of Productivity Analysis, 18, 59–78.Sherman, H. D., & Gold, F. (1985). Bank branch operating efficiency. Evaluation with data envelopment
analysis. Journal of Banking and Finance, 9, 297–315.Woller, G., Dunford, C., & Woodworth, W. (1999). Where to microfinance. International Journal of
Economic Development, 1, 29–64.
123