African Development Review, Vol. 27, No.

2, 2015, 117–129

Microfinance Institutions in Ethiopia, Kenya and Uganda: Loan

Outreach to the Poor and the Quest for Financial Viability

Gashaw Tsegaye Ayele

Abstract: Microfinance institutions’ (MFIs’) loan service outreach to the poor (depth) and the ensuing institutional viability
concern is an unsettled issue in the literature. Can MFIs increase the depth of their outreach whilst achieving financial viability
(viability)? Answering this question is exceedingly relevant to countries that opt for right policies towards financial inclusion. In
their microfinance operations, Kenya and Uganda ranked first and second in Africa; fifth and eighth in the world, respectively;
and Ethiopia is an emerging MFI destination. Yet, the loan outreach in these countries falls short of the uncontested huge
demand. The study introduces an approach that disintegrates the overall effect of depth on viability into direct and indirect
effects. Hausman-Taylor and Generalized Structural Equation Models are employed on unbalanced panel dataset of 31 MFIs
(2003–12) drawn from the three countries. The result implied a direct-positive effect and an indirect-negative effect running
from depth to viability. Under contained operational-expenses-per-loan-portfolio, depth could be pro-viability. Debt-to-Equity-
Ratio relate inversely with viability whereas ‘Real-Yield’ relates directly. The paper concludes that support to MFIs should be
aligned to ensure efficiency through reduced operational costs and thereby complementary depth–viability nexus can prevail.

1. Introduction
Conventional banks in sub-Saharan Africa (SSA) that own 53 percent and 60 percent shares of the total loan portfolio and the
total deposits, respectively, serve only a quarter of the total borrowers. The remaining three-quarters rely on loans from non-bank
financial intermediaries, non-governmental organizations, credit unions/financial cooperatives and other sources (CGAP-MIX,
2011). These have called governments and donors to support Microfinance Institutions (MFIs) to promote financial inclusion. A
Conference of African Ministers of Economy and Finance (CAMEF) recommended a minimum set of policies meant to advance
MFI services amid African Union member countries: ‘fostering an enabling policy and regulatory environment for microfinance
that balances increased access for poor people, financial stability, and consumer protection’ (African Press Organization, 2009 p.
1). In 2012, the global funds committed for financial inclusion reached US$29 billion and SSA is amongst the major recipients of
the funds (CGAP, 2014).
MFIs in Kenya and Uganda received huge donor support since the early days and they are considered to be MFI success stories
of Africa. For instance, from its establishment in the 1980s until 2000, the Kenya Rural Enterprise Programme (K-REP) alone
received US$23 million; from 1995 to 2001, the Kenya Women Finance Trust (KWFT) received about US$6 million (Otto et al.,
2002). Uganda received an estimated US$40 billion between 1998 and 2003. Donor aid has principally been to ensure financial
sustainability through capacity building, technical support and for on-lending (MFTransparency, 2011). In the Economist
Intelligence Unit’s (EIU, 2013) ranking of 55 top microfinance destination countries, Kenya leads the African continent and is
ranked fifth in the world. Uganda ranked second in Africa and eighth in the world. Kenya and Uganda are also known for mobile
money (see Table 1). M-Pesa and M-Shwari mobile-money service for millions of Kenyans is an inspiration to the rest of the
world. In Ethiopia, mobile banking is a phenomenon after the introduction of M-birr in 2013 (Dominque, 2013). Ethiopia is an
emerging microfinance destination with growing outreach in recent years.
Nevertheless, the industries are still less competitive, less efficient, and with limited outreach. For instance, the total asset held
in Kenya’s MFI industry grew from US$1.71 billion (twice Ethiopia’s) in 2009 to US$2.59 billion (4 times Ethiopia’s) in 2011,

*
Research Fellow at the Horn Economic and Social Policy Institute, The Horn Economic and Social Policy Institute/HESPI, Addis Ababa, Ethiopia;
Mobile: (þ251) 918-78-7482; Tel: (þ251) 115-15 32 62/65; e-mail: [email protected] or [email protected]; website: http://www.hespi.org/
© 2015 The Authors. African Development Review © 2015 African Development Bank. Published by Blackwell Publishing Ltd,
9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. 117
118 G. T. Ayele

Table 1: Financial indicators for Kenya, Uganda, SSA and World in 2011
Kenya Uganda SSA World

Loan in the past year (% age 15þ) 67 52 46 33

Loan in the past year, female (% age 15þ) 66 52 45 31
Loan in the past year, rural (% age 15þ) 67 53 46 34
Mobile phone used to receive money (% age 15þ) 66 25 14 3
Mobile phone used to receive money, rural (% age 15þ) 64 24 12 3
Mobile phone used to send money (% age 15þ) 60 20 11 2
Mobile phone used to send money, rural (% age 15þ) 57 19 9 2
Source: Author’s summary from the World Bank’s Global Findex database, 2012.

yet 80 percent of the total asset belonged to the Equity Bank. If we exclude commercial banks, the asset growth drops
significantly. The size of the MFI sector, in Kenya, without banks is one-fifth its size with banks (AMFI and MFR, 2013). This
dominance could be because banks and MFIs are not working collaboratively as evidenced by a study for West Africa (Sodokin
and Donou-Adonsou, 2010).
Low competition is also evident in the MFI sector of Ethiopia. In 2011/12, four MFIs (Amhara, Dedebit, Oromia and Omo
Credit and Saving Institutions) accounted for 75 percent of the capital, 88 percent of the loan outstanding, and 83 percent of the
assets held in the sector. The NBE 2010/11 annual report indicated: of the total MFIs, 50 percent are in the capital; of the total
credit disbursed through MFIs, Addis Ababa (with 5 percent population share) accounted for 40 percent, Tigray 20 percent (with
5 population share), Amhara 16 percent (with 20 percent population share), and Oromia 12 percent (with 25 percent population
share) (NBE, 2011). In Uganda, in 2011, Centenary Bank with over 50 percent of the industry’s gross loan portfolio served only
24 percent of MFI loan clients; whereas PRIDE Uganda and BRAC Uganda with 8 percent and 5 percent share of the gross loan
portfolio served 24 and 15 percent of the loan clients, respectively. These all support the claim that competition in the MFI
industries of Ethiopia, Kenya and Uganda are low.
Scholars have been split into two camps (the self-sustainability camp and the poverty lending camp) on their view of the effect
of donor funds on MFIs institutional financial sustainability and of the effect on outreach to the economically active poor
(henceforth ‘the poor’ as opposed to the extreme poor). Those in the poverty lending camp favor micro-lending to the poor at a
lower cost (lower interest rates) through donor and government subsidies meant to combat poverty (Millson, 2013; Cull et al.,
2007; Schreiner, 2002; Conning, 1999; Morduch, 1999a, 1999b). Morduch (1999b) argued that successful MFIs, including the
pioneering Grameen Bank, received massive donor support. Between 1985 and 1996, the effective subsidy Grameen Bank
received was US$175 million and for the same period scale of outreach grew by 12 fold. Thus, he advised not to undermine the
benefit of continued subsidization under robust cost-benefit analysis. Donor institutional support to cover the start-up costs will
eventually ignite innovations that lead to efficiency through reduced per unit cost of service delivery and increased revenue
generation capabilities (Schreiner, 2002). In contrast, the self-sustainability camp share a firm-stand against direct financing of
loan portfolios claiming that unless MFIs are sustainable through full-cost recovery, the global microfinance demand will remain
unmet, that is donors and government subsidies will be short-lived. They, however, subscribe to donors’ and governments’
institutional support to MFIs in areas of technical assistance, equity funding and so on (Rhyne, 1998; Robinson, 2001).
The empirical evidence is also mixed. Cull et al.’s (2007) cross-sectional study on 124 MFIs in 49 countries concluded that
profitability-outreach trade-off or mutuality is a matter of who the MFIs are serving. There is a possibility of profitability while
serving the poor but trade-off is the case when services are extended to the poorest clients. However, their disaggregated analysis
by lending methodologies supported outreach-viability trade-off under individual lending methodology. Wagenaar (2012) using
a 15-year panel data of 1,558 MFIs, concluded mission drift on MFIs transformed from non-profit to profit institutions.
Wagenaar claims that as MFIs transform into a for-profit institution, their average loan size increases (lending to richer clients)
and the proportion of female borrowers goes down. Kipesha and Zhang (2013) studied profitability, financial viability and
outreach nexus in East Africa (47 MFIs and 4 panel periods). The welfarist’s approach specification revealed trade-off between
profitability and outreach and no trade-off between viability and depth of outreach. The specification based on institutionalists’
approach supported no trade-off between depth and viability. Quayes’ (2012) investigation of 702 MFIs selected from 83
countries revealed complementarities between depth of outreach and financial viability. Woller and Schreiner’s (2002) study
pooling 13 village banks (including FINCA of Uganda) for over a three-year period found a statistically significant and positive
© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 119

relationship between financial self-sufficiency and depth of outreach. A study by Befekadu (2007) concluded mutualism
between outreach and viability in Ethiopia. Although the transmission mechanism from outreach indicators to financial viability
was unclear, a random effect panel estimation of 14 MFIs (2002–10) by Bayeh (2012) found a positive relationship that runs
from breadth of outreach and depth of outreach to financial viability. In sum, prior studies are different in the institutions
considered, methodologies applied, data representations and countries covered.
Can MFIs be able to improve depth of loan-outreach whilst achieving operational viability? The study analyses depth of loan
outreach and financial viability nexus; and introduces a disintegrated approach to see the direct and indirect roots through which
outreach and viability are linked. Hausman-Taylor (H-T) and Generalized Structural Equation Models are employed on
unbalanced panel dataset of 31 MFIs (2003–12) drawn from the three countries. The paper concludes that support to MFIs should
be aligned to ensure efficiency through reduced operational costs to enable complementary outreach–viability nexus. Expanded
mobile banking can be one of the ways to get there.
The paper is structured as follows: Section 2 contains background on the MFI industries. Section 3 discusses the econometric
approach while Section 4 presents the results; and Section 5 concludes the study.

2. Background of the MFI Industries and Descriptive Analysis

2.1 Ethiopia
MFIs are a 1990s phenomenon in Ethiopia. The Microfinance proclamation in 1996 marked the start of deposit taking MFIs in
Ethiopia. The sector has progressed from humanitarian orientation to combining outreach and viability missions. The
government’s hand in the industry is huge, ranging from extending institutional and portfolio supports to claiming ownership in
MFIs. An assessment study by Wiedmaier et al. (2008) on access to finance in Ethiopia indicated that government led MFIs have
managed to achieve high service outreach. This is not because the markets compel them into lending to the poor but is because of
the purposive government supported lending to the poor. Without such support, the outreach could have been lower. In 2012,
ACSI (766,386 loan clients), OCSSCO (515,890), and DECSI (380,356) led the industry in their loan outreach. Large MFIs in
Ethiopia are characterized by huge affiliation to the government. Some of these MFIs are the Amhara Credit and Saving
Institution (ACSI), Dedebit Credit and Savings Institution (DECSI), Oromia Credit and Savings Share Company (OCSSCO),
Addis Savings and Credit Institution (ADSCI) and Omo Microfinance Institution Share Company (OMO). In 2008, the
government share in ADSCI and OMO reached 97 percent and 80 percent, respectively (Wiedmaier et al., 2008). The
inexpensive funding and staff salaries can partly explain the low interest rates on loans existent in most of the MFIs in Ethiopia.
According to the National Bank of Ethiopia (NBE), the number of MFIs in the country has reached 33 at the turn of 2012. For the
same year, registered MFIs’ total capital and total asset were Birr 3.8 billion (US$190 million) and Birr 13.3 billion (US$665
million), respectively (see Table 2 for more on Ethiopia’s MFI industry financial structure and performances). ACSI’s outstanding
loan balance is 71 times higher than Aggara’s but its outstanding loan balance per borrower is 37 percent below Aggara’s.

Table 2: Registered microfinance institutions aggregate performance in Ethiopia

Growth
MFIs’ aggregate 2008/09 2009/10 2010/11 2011/12 2009/10 2010/11 2011/12

Capital (US$ m) 86.870 118.761 147.299 187.774 28 37 24

Saving deposits (US$ m) 104.937 132.948 188.954 272.530 44 27 42
Loan outstanding (US$ m) 246.807 291.225 349.599 464.482 33 18 20
Total assets (US$ m) 331.032 397.910 507.819 665.410 31 20 28
Loan outstanding to asset ratio (%) 75 73 69 70
Loan outstanding to capital ratio (%) 284 245 237 247
Capital asset ratio (%) 26 30 29 28
Saving liability to asset ratio (%) 32 33 37 41
Source: National Bank of Ethiopia (NBE) report 2011 and 2012.

© 2015 The Authors. African Development Review © 2015 African Development Bank
120 G. T. Ayele

2.2 Kenya
In the 1980s, K-REP was a pioneer that provided credit and technical assistance to other NGOs. In the 1990s the financial
sustainability of K-REP was under pressure, combining the provision of financial services (loans and deposits) with non-
financial services (training and technical assistance). And it targeted people in business and reduced the NGOs it supported from
twelve to four and later the KWFT was included as the fifth MFI supported by K-REP. Donors changed their strategy and started
to focus on funding non-financial services of micro-finance agencies, particularly on financial and institutional innovations (Otto
et al., 2002). Deposit-taking MFIs appeared in 2009 when two of the pioneering MFIs – Faulu Kenya and Kenya Women
Finance Trust – transformed to deposit-takers. Transformation of microcredit programs into a bank serving only low-income
clients is an old story that happened back in 1999 when K-Rep became the first commercial bank in Kenya to serve only low
income clients, and the first NGO in Africa to transform into a regulated financial institution (Frankfurt School of Finance and
Management, 2012; Central Bank of Kenya, 2013a).
The microfinance average outstanding loan size in Kenya is high. The second row entry in Table 3 for the year 2011 shows that
for a currency unit per capita income earned, there is a loan outstanding of 1.93 currency units, that is, a borrower can have a loan
size nearly twice her/his share from the total GNI. Excluding microfinance services provided by banks leads the whole MFI
sector average outstanding loans to fall from 193 percent to 54 percent in 2011. Such fall shows banks’ involvement in lending to
richer clients.
In 2011, the microfinance sector altogether got 58.9 percent of their total assets funded from deposits, and the figure takes a
different picture when banks are excluded: the dominant fund source becomes borrowed money (54.2 percent) followed by
compulsory deposits (22.5 percent) and voluntary saving (6.32 percent). The debt-to-equity-ratio is near 5 percent, which
indicates low equity leverage in the sector.
Operational Self-Sufficiency (OSS) excluding banks was 110 percent, 105 percent, and 105 percent for 2009, 2010 and 2011,
respectively. Comparing these figures to the total sector (133 percent in 2009 and 150 percent in 2011), shows non-bank
microfinance institutions are performing less than banks. Yet, they have, in aggregate, managed to be operationally self-
sufficient. However, the sector still relies on donations and 73.3 percent of donations are raised from international partners while
only 26.7 percent from local entities and bodies. In terms of external funding, the sector reports that 59 percent of its facilities are
domestically raised while the remaining 41 percent is raised on international capital markets (AMFI and MFR, 2013). Among
MFIs in Kenya that self-reported data to the MIX database (the leading microfinance data source globally), only 20 percent had
average loan size per borrower over US$2,000. In 2012, 90 percent of them served less than 20,000 loan clients; the only
exception is Equity Bank that reported 704,249 borrowers.
Although there is a large number of MFIs in Kenya, the number of borrowers served by each is significantly lower than in
Ethiopia. The dominant MFIs in terms of loan size and client outreach in Ethiopia have a maximum of 337 average outstanding
loans per borrower. In Kenya, the lead MFI lender, Equity Bank, serves only 704,247 clients with ten times ACSI’s outstanding
loan size (the lead MFI in Ethiopian). The comparison falls to five times, when adjusted for per capita GDP.

Table 3: Key outreach and financial indicators of Kenya’s MFI industry (FY 2009 to 2011)
Whole MFI sector Excluding banks
Indicators 2009 2010 2011 2009 2010 2011

Average disbursed loan size 1,405 1,242 1,649

Average outstanding loan per borrower as 181 157 193 54
percent of per capita GDP.
Operational self-sufficiency 133 147 150 110 105 105
Portfolio yield 23.8 24.7 34.9 34.2
Debt-to-equity ratio 4.1 4.3 6.3 5.4
Operating expenses ratio 16.7 15.6 26.7 26.7
Number of active borrowers 1,395,890 1,433,897 1,475,664
Source: Author’s construction from 2012 Annual Report on Microfinance Sector (AMFI and MFR, 2013, pp. 7-9).

© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 121

2.3 Uganda
In the 1980s, a number of NGOs developed microfinance departments and a few specialized MFIs started operations offering
microcredit collateralized by compulsory savings. In the 1990s, these organizations grew to counteract the gap left by the closure
of several large banks. The late 1990s and early 2000s witnessed a plethora of fully sponsored training and technical assistance
for MFIs, and the adoption of sustainability and profitability alongside the original social mission of the MFIs. The microfinance
providers currently regulated by the Bank of Uganda include: two commercial banks (Centenary Bank and Equity Bank); two
credit institutions (Opportunity Uganda and Post Bank Uganda); and three micro deposit taking institutions (Finance Trust,
FINCA, and Pride Microfinance).
In Uganda, SACCOs are dominant in their number and distribution. The government has a package of incentives for new
SACCOs targeting at least one SACCO for a sub-city. Government extends start-up grants, provides interest-free loans, and
subsidizes interest rates. The government support extends to providing rent-free offices and covering staff salaries for the first two
years of operation (Linthorst, 2013). The interest rate in Uganda remains so high primarily because of high operational costs. Flat
rates as opposed to declining balance are dominant. MFIs’ lending is concentrated on business or commercial loans. There was a
variety of products, including group loans (19 MFIs), school fee loans (31 MFIs), bicycle loans (13 MFIs), solar power loans (12
MFIs) and emergency loans (13 MFIs) amongst others. Mobile banking in Uganda has not only extended convenience for MFIs and
their clients but also is contributing to reduced transaction costs and improved rural market penetration (MFTransparency, 2011).

2.4 Recent Developments

Recently, the interest rates charged by MFIs and over-indebtedness of clients have garnered the academic and policy debates.
Microcredit had been laudable in recovering loans. Today, there is a general fall in loan repayment rates. This is partly attributed to
the over-indebtedness of clients that led to further impoverishment of borrowers. Reduced consumption levels, late fees, asset
seizures, loss of creditworthiness are the manifestations (Jessica and Richard, 2011). Further concerns have emerged with regard to
high interest rates. In some countries, client over-indebtedness is attributed in part to market saturation, with a narrow range of credit
products and competition among MFIs pushing lenders to make increasingly risky loans and pursue harsh collection practices.
Over-indebtedness can increase financial and social vulnerability as borrowers take new loans to repay old ones or resort to extreme
measures to make their payments, including reducing their consumption of food and selling productive assets (Ledgerwood, 2013).
A survey of 350 leading microfinance institutions which charged between 20 and 40 percent real interest rate Cull et al. (2008)
cited in Morduch (2008). Another survey by the Center for the Study of Financial Innovation (appeared in its Microfinance
Banana Skins 2014 report), emphasized over-indebtedness of microfinance clients to be the largest risk. The report further noted
that over-indebtedness is a symptom for surplus lending capacity, a lack of professionalism within MFIs, and an emphasis on
growth and profit at the expense of prudence. It is also linked to the credit risk emanating from poor client evaluation, arrears and
recovery practices and risk management systems. However, the report identified unhealthy competition in the MFI sector that led
to growth in lending and entry of new MFIs with little or no social mission. Of the seven regions covered by the survey all but two
placed over-indebtedness as the number one challenge. The exceptions were Africa, where it came third after credit risk and
governance (Center for the Study of Financial Innovation, 2014).
Uganda is infamous for high operational expenses per loan portfolio (see Figure 1). Ethiopia managed to keep it low. The
Ethiopian case could be for government stake taking in the industry to the extent of providing free expert and personnel support
and the low wage rates (wage rates in Ethiopian are one of the lowest in SSA).
The real yield in Ethiopia and Kenya is significantly below the global average of 20 percent while Uganda touched over 40
percent (with a 4 percent loan loss rate and over 5 percent write-off ratio) (see Figure 2). The loan loss rate and the write-off ratios
in Ethiopia and Kenya are below the global average. The portfolio at risk (90 days) is close to 20 percent for Uganda while
Ethiopia’s remains close to zero. Besides, the average loan size in Uganda is high, these all support that over-indebtedness in
Uganda is apparent compared to the other two nations. The cost per loan is extremely low in Ethiopia.

3. Econometric Approach
Theoretically, depth-of-outreach and financial viability are jointly determined by two opposing effects: (1) increased transaction
costs emanating from deeper outreach (administration of small loans leads to escalating transaction costs) which compromises
© 2015 The Authors. African Development Review © 2015 African Development Bank
122 G. T. Ayele

Figure 1: Operating-expense-per-loan portfolio comparison

Africa East Asia and the Pacific Ethiopia

Kenya Latin America and The Caribbean

Operating expense / loan portfolio (Weighted Average)

Uganda
.8

.6

.4

.2

0
2000 2005 2010 2015
Fiscal Year

Source: Author’s construction, data from MIX market database, 2013

financial viability; (2) increased operational revenue from increased loan portfolio (by allowing more loan clients) contributes
positively to viability. The overall effect depends on the relative importance of the two opposing effects. The researcher finds no
prior study that attempted to make causal inference (mediation) running from depth of outreach to financial viability.
In this piece of work, both the overall effect and the mediation effects are targeted. First the financial viability indicator is
regressed on depth of outreach indicator for its direct effect. Then, in the second step, an assumed path from depth of outreach to
operational cost and finally to financial viability has been quantified.
OSS is chosen as an indicator of financial viability. Average loan size per loan portfolio and proportion of women clients are
used as proxies of outreach depth. A model employing panel data is constructed and a Hausman and Taylor instrumental variable
estimation technique is used to capture the direct effect; and Generalized Structural Equation Model (GSEM) is employed as
well to quantify the mediation (path). The Breusch–Pagan LM procedure was conducted to test the poolability and the test result
did not support pooled regression and hence a panel model has been constructed.

3.1 Definition of Key Variables

Depth of Outreach
Depth of outreach is one of the indicators of outreach. It measures the extent to which MFI services reached low income people.
Direct measurement of depth of outreach is difficult and hence indirect proxies are often used. MFIs’ financial performance can
give a clue to their social performance. Poorer clients are less able to absorb larger loans (Woller, 2006). Either size of average
outstanding loan per borrower per Gross National Income (GNI) (used for instance by Wagenaar, 2012; Quayes, 2012;
Schriener, 2002; Cull et al., 2007) is common depth of outreach proxy. Lower values imply deeper outreach as poor clients take
small-sized loans. Proportion of women from the total borrowers is also another proxy for depth of outreach used.

Financial Viability
Financial viability refers to the ability of a MFI to cover its costs with earned revenue. To be financially viable, an MFI cannot rely on
donor funding to subsidize its operations. To determine financial viability, self-sufficiency indicators are calculated. (Ledgerwood, 1999,
pp. 216–17)
© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 123

Figure 2: Selected financial indicators

Real yield on gross portfolio Loan loss rate Write-off ratio

Ethiopia Kenya Uganda

2012
Uganda
2012

Ethiopia Kenya Uganda

2011 2012
2011
2011
2012 2012
Kenya

2012
2011 2011
2011
2012
Ethiopia

2012 2012
2011 2011
2011

0 20 40 60 -2 0 2 4 0 5 10

Cost per loan Portfolio at risk (90 days) Average outstanding loan
US Dollar balance (%GNI Per capita)
Uganda

2012
Uganda

2011 2012

Uganda
2012
2011
2012 2011
Kenya

2012
Ethiopia Kenya

2012

Ethiopia Kenya
2011
2011
2011
Ethiopia

2012
2012
2012
2011
2011
2011
0 100 200
0 10 20 0 100 200

Global average Maximum MFIs in MIX

market database
Real yield on gross portfolio 19.72 85%
Loan loss rate 3.5 5%
Write-off ratio 4.4 5.85
Cost per loan 241 –
Portfolio at risk (90 days) 17 58
Average outstanding loan (% GNI) 88 537

Transition to viability is from operationally unviable (unable to cover operational costs from operational revenues) to
operationally viable (able to cover operational costs from operational revenues) to financially viable (able to cover operational
costs without subsidy).
Theoretically, depth-of-outreach and financial viability are jointly determined by two opposing forces: (1) increased
transaction costs emanating from deeper outreach (administration of small loans leads to escalating transaction costs) which
compromises financial viability; (2) increased operational revenue from increased loan portfolio (by allowing more loan clients)
contributes positively to viability. The overall effect depends on the relative importance of the two opposing forces. This piece of
work aims to investigate the direct, indirect and overall effects of depth on viability. First, the financial viability indicator is
regressed on depth of outreach indicator, along with control variables, for its direct effect. Then, in the second step, an assumed
path from depth of outreach to operational cost and finally to financial viability has been quantified. However, a path is assumed
and the result is interpretable under the assumed path.
OSS is chosen as an indicator of financial viability. Average loan size per borrower per GNI and proportion of women clients
are used as proxies of outreach depth. A Hausman and Taylor instrumental variable estimation technique is used to capture the
direct effect; and a Generalized Structural Equation Model is run for the indirect effect of depth of outreach on OSS.
© 2015 The Authors. African Development Review © 2015 African Development Bank
124 G. T. Ayele

3.2 Data Source

The Microfinance Information Exchange (MIX)1 market is a source of data for the current research. Among the scholars that
used the MIX market database are Quayes (2012), Kipesha and Zhang (2013), Millson (2013), Cull et al. (2007), Kumar (2011),
and many others. Data in the MIX market is self-reported and it is organized in such a way as to classify data quality in different
levels—diamond levels. MIX uses the ‘diamonds’ system to indicate an MFI’s level of transparency. A higher diamond rank
means a more transparent MFI and more reliable data. MFIs in the sample are those with periodic reports available in the MIX
database and with three or more diamonds. It is worth acknowledging the possibility of attrition bias and loss of randomness in a
situation when the MFIs are reporting to the MIX.

Model Specification for the Direct Effect

Several panel estimation techniques are compared: Fixed Effects (FE), Random Effects (RE), Arellano-Bond dynamic panel
model (A-B), the Hausman and Taylor instrumental variable model (H-T). The FE method could avoid the potential problem of
correlation between the explanatory variables and the unobserved individual heterogeneity. However, the FE transformation
does not give estimates for the vector of time-invariant coefficients and it loses efficiency as it overlooks variations across
individuals (Wooldridge, 2002; Baltagi, 2005; Greene, 2003). Random effects transformation can remedy the shortcomings of
the FE transformation, on condition that all the explanatory variables are uncorrelated with both the idiosyncratic and
unobserved individual heterogeneity components of the composite error terms. The random effects model fails to qualify the
Hausman test.
The possibility remaining are accepting the FE estimation despite the loss of estimates for the time invariant variables and
tolerating the efficiency loss; or fixing the endogeneity problem of the RE estimation through revisiting the model to control for
omitted explanatory variables (exogenizing); or simply use instrumental variable estimation techniques. Exhausting omitted
variables was impossible, as some of the variables are unobservable. Another approach is to use the traditional instrumental
variable model (finding exogenous variables out of the model that are uncorrelated with the composite error component but
correlated with the endogenous variables) but finding appropriate instruments is difficult. Therefore, the Arellano-Bond (A-B)
and Hausman-Taylor (H-T) instrumental variable estimation procedures that use instruments within the model are compared
(Cameron and Trivedi, 2010). However, application of the A-B technique led to huge loss of observations. Finally, the H-T
estimation technique is found to be appropriate.
The H-T method is based on the random-effects transformation that leads to the following model:

       
Log OS S it ¼ a þ X 1 it d1 þ X 2 it d2 þ Z 1 it d1 þ Z 1 i g 1 þ Z 2 i g 2 þ n i þ m it ::: The H-T model

where:

X 1 it ¼ X 1it  ^u i X 1i


a and b are constant terms of the models

Table 4: Endogenous and exogenous variables included in the H-T estimation

Endogenous time Exogenous time Endogenous time
variants invariants invariants

X 2 it Z 1 it Z 2 it
Exogenous time variants X 1 it

– Operating expenses per loan portfolio – Height of outreach – Country dummies – Regulatory status of MFIs
– Real yield
– MFI experience (Age)

© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 125

The formula for ^u i is given in xthtaylor

n i ¼ ni ð1  ^u i Þ 6¼ 0


  
n i is correlated with Z 2 i and X 2 it
X_ 2it is instrument for  X 2it ; X_ 2it ¼ X 2it  X 2 i
X 1i instrument for Z 2 i


X_ 1it is instrument for X 1 it




Z 1 i as instrument for Z 1 i
X 1 is used as instrument twice: as X_ 1it and as X 1i : By using, the average of X 1i in forming instruments data from other periods
are used as instruments.
The H-T method assumes the unobserved time-invariant individual specific effect to be correlated with some of the left-hand
side variables but not with all. H-T estimation requires certain variables to be uncorrelated with individual specific unobserved
errors. Initially, such exogenous time invariant and exogenous time variants have been selected. The assumed time variant
exogenous variables play two roles: (1) the deviation from their own mean is used to estimate their own coefficients; (2) their
average over time serves to estimate the time invariant endogenous variables. The over identification restriction test has dictated
identification of endogenous variables, that is, the explanatory variables were included in the endogenous variable list initially
with rule of thumb and later a step-by-step inclusion of variables were followed observing improvements in the Sargan–Hansen
test of overidentifying restrictions. Consistency of the H-T estimators requires all regressors to be uncorrelated with the
idiosyncratic errors and subsets of the regressors to be uncorrelated with the fixed effects (Hausman and Taylor, 1981).

Model Specification for the Indirect Effect (Path or Mediation)

GSEM is used to estimate how lending to the poor reflects itself onto increased operational expenses per loan portfolio (OL)
(intermediate variable) and thereby affects financial viability. With GSEM, OL has been predicted using its determinants: debt-
to-equity-ratio; age (experience) of the MFI; height-of-outreach, outreach-to-women; country, regulatory status of the MFI, staff
salary per GNI, number of loans per loan officer, and number of loans per staff. The estimated value of OL from the GSEM is to
be multiplied by the coefficient estimate of OL in the H-T model to find the indirect effect of depth on viability. Maximum
likelihood method is used to estimate this model:
Log (OSS) ¼ f(Log (OL), control variables) … Structural equation
Log (OL) ¼ f (Log (HO), control variables)… Reduced form
where OSS, OL, and HO are Operational Self-Sufficiency, Operational expense per Loan portfolio, and Height of Outreach,
respectively.

4. Results and Discussion

Depth of outreach is measured by the average outstanding loan per client per GNI. The exact inverse of it is ‘height of outreach’
(HO). The coefficient estimate for the Log (HO) of –0.145 implies that the direct effect of a percent increase in the height of
outreach on OSS is a fall of it by 0.145 percent. However, delving deeper into the issue, the study found that depth or height of
outreach has an indirect channel to affect OSS via ‘operating cost per loan portfolio (OL)’. The indirect effect is computed by
multiplying the coefficients Log (HO) in column (2b) by the coefficient of Log (OL) in column (1) of Table 5, that is, –0.794) * –
0.523 ¼ 0.42. Therefore, the indirect elasticity of OSS for a percent increase in height of outreach is 0.42. The overall effect is the
sum of the direct and indirect effects, which is 0.42–0.145¼ 0.275. Thus, for a percentage increase in height of outreach or
decrease in depth of outreach (lending to the opulent), OSS improves by 0.42 percent via its indirect channel. The overall effect
favors the trade-off view under cost inefficiency.
Compared to banks, MFIs are less efficient and they often charge higher interest rates on loans and have a shorter repayment
schedule. Truly wealthier clients rarely become clients of MFIs, instead clients who request a larger loan could rather be those
who plan to default (high credit risks). The MFIs are less able to compete with the big banks in providing large size loans to truly
© 2015 The Authors. African Development Review © 2015 African Development Bank
126 G. T. Ayele

Table 5: H-T, GSEM, fixed effects, and random effects robust estimators (standard errors in parentheses)
Direct effect Indirect effect Fixed effect Random effect
using H-T (1) using GSEM (2) robust (3) robust (4)
(2a) (2b)
Explanatory variables Log (OSS) Log (OSS) Log (OL) Log (OSS) Log (OSS)
Log (OL) –0.523*** –0.540*** –0.498*** –0.536***
(0.0519) (0.0372) (0.0975) (0.0548)
Log (DE) –0.0639*** –0.0492** –0.000568 –0.0617* –0.0712***
(0.0211) (0.0210) (0.0256) (0.0359) (0.0266)
Real yield 0.00468*** 0.00566*** 0.00406*** 0.00618***
(0.00132) (0.00153) (0.00123) (0.00134)
New MFI 0.0289 –0.100 –0.0262 0.0142 –0.00445
(0.0649) (0.0646) (0.0811) (0.104) (0.0802)
Young MFI –0.0619 –0.0768** 0.124*** –0.0617 –0.106***
(0.0380) (0.0361) (0.0454) (0.0436) (0.0357)
Log (HO) –0.145*** –0.00193 0.794*** –0.136** –0.0554*
(0.0441) (0.0296) (0.0488) (0.0534) (0.0317)
Log (OW) 0.00357 0.0177 0.0681 0.0137 0.0185
(0.0480) (0.0386) (0.0478) (0.0426) (0.0345)
Kenya 0.300** 0.156*** 0.312*** 0.157**
(0.153) (0.0535) (0.0622) (0.0662)
Uganda 0.401*** 0.239*** 0.408*** 0.255***
(0.146) (0.0817) (0.0758) (0.0874)
Regulated MFIs 0.342* 0.0943** 0.0808 0.0813
(0.184) (0.0410) (0.0493) (0.0571)
Log (salary GNI per capita) –0.794***
(0.0543)
Log (loans per loan officer) 0.0634
(0.0482)
Log (loans per staff) –0.972***
(0.0733)
Inflation 0.00207
(0.00201)
par90 0.00161
(0.00712)
par30 0.00455
(0.00503)
Constant 6.784*** 6.362*** 8.393*** 7.112*** 6.695***
(0.362) (0.237) (0.396) (0.614) (0.288)
Observations 150 115 115 150 150
Number of MFIs 31 31 31
R-squared 0.482

wealthier and creditworthy clients. The direct positive relationship between depth and viability lies more on the less fierce
competition that a MFI faces when providing small-sized loans. This finding is consistent with Quayes’ (2012) cross-sectional
global study and Bayeh’s (2012) panel study for Ethiopia and it is against the findings of Abate et al.’s (2013) panel study for
Ethiopia and Okumu’s (2007) panel study for Uganda.
The logarithm of outreach to women in the H-T model is statistically insignificant which means the effect of being a female
client on financial viability is insignificant. The log (OL), with an estimated coefficient of –0.523 implies a fall of OSS by 0.523
percent for a percent increase in OL. The transmission mechanisms are twofold. First, an increase in operating cost per loan
portfolio obviously means growth of operational expense components in the OSS financial formula. Secondly, it could be that the
increased operational expense has also led to increased operational revenue. Untimely, what matters is the net effect. The
coefficient estimate for the logarithm of debt-to-equity-ratio (–0.064) is significant at 1 percent. The debt-to-equity-ratio does not
© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 127

Key independent variables

Log (OSS) Logarithm of Operational Self-Sufficiency
Log (OL) Logarithm of operating expense per loan portfolio after loan portfolio is converted to percent
Log (DE) Logarithm of debt to equity (the ratio is converted to percentage to be able to calculate logarithms)
Age 1–4 years of operation¼ New; 5–8 years ¼ young; over 8 years ¼ mature
Log (HO) Logarithm of Height of outreach (which is inverse of depth of outreach)
Log (OW) Logarithm of outreach to Women(logarithm of percent women from the total loan clients)
Kenya, Uganda MFIs in Kenya and Uganda compared to the reference category, Ethiopia (Dummy)
Real yield Price adjusted yield (proxy for interest rate)

necessarily indicate the ability of MFIs to leverage their capital through commercial loans and/or deposits. The MFIs are getting
subsidized loans from governments or donors. This is likely to lead to less responsible lending and inefficient financial
management and thereby cause an inverse relationship between debt-to-equity-ratio and OSS. The questions should be how to
make the debt sustainable and at the same time use it to generate revenue above its costs.
‘Real yield’ (a proxy for interest rate), with an estimated coefficient of 0.0047, is significant at 1 percent. A percentage point
increase in real yield leads to a 0.47 percent improvement in OSS (0.0047*100, as it is a log-level model for real yield). Real yield
can have two transmission channels to affect OSS. On one hand, real yield could negatively affect the client’s borrowing decision;
its significance again depends on the client’s loan demand elasticity. On the other, an increase in real yield means higher income per
unit of loan outstanding. The overall effect depends on the extent real yield affects the volume of outstanding loan and the income
from a unit of outstanding loan. In the current case, an increase in real yield has led to improved OSS, that is, the fall in loan portfolio
as a result of increased cost to borrowers is over compensated by the higher earnings per unit of loan portfolio.
The H-T result also indicated MFIs based in Uganda and Kenya are more likely to be financially viable than that of Ethiopia.
The coefficient estimate for Uganda (0.4 significant at 1 percent) and for Kenya (0.3 significant at 5 percent) are transformed into
marginal effects by a conversion formula ((eb – 1)*100, where b is the coefficient estimates). Thus, MFIs in Uganda and Kenya
are 49 and 35 percent more likely to be operationally self-sufficient than those in Ethiopia, respectively. Lastly, the GSEM
suggested a negative direct impact of average staff salary on operating expenses per loan portfolio. However, the transmission
from increased staff salary to staff productivity to operating expense to OSS requires further inquiry.

5. Concluding Remarks
The study has investigated the depth-of-loan-outreach and financial-viability nexus, quantified an assumed path running from
depth to viability and assessed existing challenges of the microfinance industries of Ethiopia, Kenya and Uganda. The study
introduces an approach that disintegrated the overall effect of depth on viability into direct and indirect effects. Hausman–Taylor
and Generalized Structural Equation Models are employed on an unbalanced panel dataset of 31 MFIs (2003–12) drawn from the
three countries.
The result implied a direct positive effect and an indirect negative effect running from depth to viability. The indirect effect
outgrows the direct and hence the overall effect is against lending to the poor. However, under contained operational-expenses-
per-loan-portfolio, depth could be kept pro-viability. Mutual depth-viability nexus is a possibility when operating cost
differentials across different loan sizes can be kept low. This signals the need for rethinking MFIs’ lending methodologies to
reduce operational expenses.
Under careful design, mobile banking could help MFIs reduce operation costs. The growing mobile subscription rate
corroborates the argument. It is not promising for MFIs in the region to establish their own mobile-banking services as it requires
huge financial, material and time investments. If several smaller MFIs pool their resources for initial investment, their collective
scale could justify the investment, if not they need to continue partnering with the mobile-banking service providers (as is the
case now). MFIs can lower costs through automatic text messages that notify customers about upcoming payments or loan
disbursements or warn of late payment notices to dismantle the depth of outreach and viability trade-off.
The debt-to-equity-ratio relates inversely with viability whereas ‘real-yield’ relates directly. The paper concludes that support
to MFIs should be aligned to ensure efficiency through reduced operational costs and thereby a complementary depth–viability
nexus can prevail. Government and donor support should target empowering MFIs to restructure their operating cost per a unit of
loan outstanding.
© 2015 The Authors. African Development Review © 2015 African Development Bank
128 G. T. Ayele

References
Abate, G., C. Borzaga and K. Getnet (2013), ‘Financial Sustainability and Outreach of Microfinance Institutions in Ethiopia:
Does Organizational Form Matter?’, European Research Institute on Cooperative and Social Enterprises (Euricse) Working
Paper, No. 56.
African Press Organization (2009), ‘Micro Finance Development, Establishment of African Investment Bank and Alternative
Funding of the African Union’, Extraordinary Conference of African Ministers of Economy and Finance (CAMEF), 14–17
December, African Union Commission, Addis Ababa.
Armendariz, B. and A. Szafarz (2011), ‘On Mission Drift in Microfinance Institutions’, Munich Personal RePEc Archive
(MPRA) Paper, No. 31041.
AMFI and MFR (Association of Microfinance Institutions and MicroFinanza Rating) (2013), ‘Annual Report on Microfinance
Sector in Kenya, 2012’, First Edition, available at: http://www.microfinanzarating.com/images/stories/2012_Report_on_the_-
Kenyan_Microfinance_Sector_final_27-02-2013.pdf (accessed 15 May 2014).
Bayeh, K. (2012), ‘Financial Sustainability of Microfinance Institutions in Ethiopia’, European Journal of Business and
Management, Vol. 4, No. 15, pp. 1–10.
Befekadu, K. B. (2007), ‘Outreach and Financial Performance: Analysis of Microfinance Institutions in Ethiopia’, African
Economic Conference, ‘Africa’s Development Challenges and Opportunities in the Global Arena', 15–17 November, United
Nations Conference Center, Addis Ababa.
Baltagi, B. (2005), Econometric Analysis of Panel Data, 3rd edn, John Wiley & Sons, New York.
Cameron, A. and Trivedi, K. (2010), Microeconometrics Using Stata, rev. edn, Stata Press, College Station, TX.
Central Bank of Kenya (2013a) ‘Bank Supervision Annual Report 2013’.
Central Bank of Kenya (2013b) ‘FinAccess National Survey 2013: Profiling Developments in Financial Access and Usage in
Kenya’.
Centre for the Study of Financial Innovation (2014), ‘Microfinance Banana Skins 2014, Facing Reality: Survey of Microfinance
Risk’, available at: http://www.citigroup.com/citi/inclusivefinance/data/2014_microfinance_banana_skins.pdf
CGAP-MIX (Consultative Group to Assist the Poor-Microfinance Information Exchange) (2011), ‘MIX Microfinance World:
Sub-Saharan Africa Microfinance Analysis and Benchmarking Report, 2010’, available at: http://www.themix.org/sites/default/
files/096_MIX_Africa%20Report_05-05-2011.pdf (accessed 20 May 2014).
CGAP and Its Partners (2014), ‘A Report on Greenfield MFIs in Sub-Saharan Africa: A Business Model for Advancing Access
to Finance’, No. 8.
Conning, J. (1999), ‘Outreach, Sustainability and Leverage in Monitored and Peer-Monitored Lending’,Journal of Development
Economics, Vol. 60, pp. 51–77.
Cull, R., A. Demirg€ucS -Kunt and J. Morduch ( 2007), ‘Financial Performance and Outreach: A Global Analysis of Leading
Microbanks’, The Economic Journal, Vol. 117, pp. 107–33.
Dominque, M. (2013), ‘Mobile Money Comes to Ethiopia’, African Business Magazine, 9 January, available at: http://
africanbusinessmagazine.com/sector-reports/technology/mobile-money-comes-to-ethiopia/
Economist Intelligence Unit (2013), ‘Global Microscope on the Microfinance Business Environment’, available at: http://www.
citigroup.com/citi/citizen/community/data/EIU_Microfinance_2013_Proof_08.pdf (accessed 14 June 2014).
Esperance, J., P. Ana and A. Mohamed (2003), ‘Corporate Debt Policy of Small Firms: An Empirical (Re) Examination’,
Journal of Small Business and Enterprise Development, Vol. 1, pp. 62–80.
Frankfurt School of Finance and Management (2012), ‘Transforming Microfinance in Kenya the Experience of Faulu Kenya and
Kenya Women Finance Trust’, Bankakademie/HfB.
Greene, W. (2003) Econometric Analysis, 5th edn, Prentice Hall, New York.

© 2015 The Authors. African Development Review © 2015 African Development Bank
Microfinance Institutions in Ethiopia, Kenya and Uganda 129

Hausman, J. and W. Taylor (1981), ‘Panel Data and Unbservable Individual Effects’, Econometrica, Vol. 49, No. 6, pp. 1377–
98.
Kipesha, E. and X. Zhang (2013), ‘Sustainability, Profitability and Outreach Tradeoffs: Evidences from Microfinance
Institutions in East Africa’, European Journal of Business and Management, Vol. 5, No. 8.
Ledgerwood, J., J. Earne and C. Nelson (eds.) (2013), The New Microfinance Handbook: A Financial Market System
Perspective, World Bank, Washington, DC.
MFTransparency (2011), ‘Promoting Transparent Pricing in the Microfinance Industry of Uganda: Country Survey’, available
at: http://www.mftransparency.org/wp-content/uploads/2012/05/MFT-RPT-107-EN-Country-Survey-Uganda.pdf (accessed
13 April 2014).
Millson, H. (2013), ‘The Trade-Off between Sustainability and Outreach: The Experience of Commercial Microfinance
Institutions’.
Morduch, J. (1999a), ‘The Microfinance Promise’, Journal of Economic Literature, Vol. XXXVII, pp. 1569–614.
Morduch, J. (1999b), ‘The Role of Subsidies in Microfinance: Evidence from the Grameen Bank’, Journal of Development
Economics, Vol. 60, pp. 229–48.
Morduch, J. (2008), ‘Can the Poor Afford Microcredit?’, Financial Access Initiative, Framing Note No. 4.
National Bank of Ethiopia (2011), ‘Annual Report 2010/11’.
National Bank of Ethiopia (2012), ‘Annual Report 2011/2012’.
National Bank of Ethiopia (2013), ‘Annual Report 2012/2013’.
Okumu, L. (2007), ‘The Microfinance Industry in Uganda : Sustainability, Outreach and Regulation’, unpublished PhD thesis,
University of Stellenbosch.
Otto, H., M. Muli and O. Milcah (2002), ‘An Evaluation of Micro-Finance Programmes’, Evaluation Report, available at: http://
www.gdrc.org/icm/country/Kenya-finalreport.pdf (accessed 23 December 2013).
Quayes, S. (2012), ‘Depth of Outreach and Financial Sustainability of Microfinance Institutions’, Applied Economics, Vol. 44,
pp. 3421–33.
Rhyne, E. (1998), ‘The Yin and Yang of Microfinance: Reaching the Poor and Sustainability’, available at: http://www.
jointokyo.org/mfdl/readings/Rhyne-yingyang.pdf (accessed 3 April 2014).
Robinson, M. (2001), ‘The Microfinance Revolution: Sustainable Finance for the Poor’, The World Bank, Washington, DC.
Schreiner, M. (2002), ‘Aspects of Outreach: A Framework for the Discussion of the Social Benefits of Microfinance’, Journal of
International Development, Vol. 14, No. 5, pp. 591–603.
Sodokin, K., and C. Donou-Adonsou (2010), ‘Banks, Microfinance Institutions and Economic Growth in the WestAfrican
Economic and Monetary Union’, African Development Review, Vol. 22, No. 4, pp. 495–510.
Wagenaar, K. (2012), ‘Institutional Transformation and Mission Drift in Microfinance’, Centre of Development Studies,
University of Cambridge.
Wiedmaier-Pfister, M., D. Gesesse, W. Amha, R. Mommartz and E. Steel (2008), ‘Access to Finance in Ethiopia Sector
Assessment Study, 2008’, Sector Assessment Study, Vol. 2, available at: http://www.microfinancegateway.org/sites/default/
files/mfg-en-case-study-access-to-finance-in-ethiopia-sector-assessment-study-volume-2-jul-2008.pdf (accessed 31 December
2013).
Woller, G. and M. Schreiner (2002), Poverty Lending, Financial Self-sufficiency and the Six Aspects of Outreach, Poverty
Lending Working Group of the Small Enterprise and Education Promotion Network.
Wooldridge, M. (2002), Econometric Analysis of Cross Section and Panel Data, MIT Press, Cambridge, MA.
World Bank (2012), Global Findex Database, 2012, World Bank, Washington DC.

© 2015 The Authors. African Development Review © 2015 African Development Bank