Determinants of Sustainability and Outreach of Ethiopian Microfinance Institutions

Determinants of Sustainability and Outreach of Ethiopian
Microfinance Institutions: The Case of Operational

Characteristics of Microfinance Institutions
Second Draft
A.K.Vashisht**
Karamjeet Singh++
Letenah Ejigu Wale$$
May, 2010
**
Professor of Finance, University Business School, Panjab University, Chandigarh, India.
E-mail:[email protected]
++
Reader, Accounting and Finance, University Business School, Panjab University, Chandigarh, India.
$$
PhD Research Scholar, University Business School, Panjab University, Chandigarh, India
1
Abstract
Microfinance has spread around the world. Yet microfinance practitioners estimate that 500
million poor people worldwide demand financial services, while MFIs reach only 15 to 70
million of them. Besides, lack of greater outreach, MFIs is plagued by poor financial
sustainability. The majority of MFIs could not survive without subsidies. Hence the two goals
of microfinance i.e. reaching to the poor and being financially self reliant are great
challenges to the industry as a whole. In this connection, this paper tries to uncover the
impact of operational characteristics of MFIs on their sustainability and outreach
performance. Operational characteristics include variables such as capital structure
decisions, asset allocation; interest rate charged, cost efficiency, productivity and portfolio
quality measures. A host of other control variables are also included.
Data from 15 MFIs for the year 2003-2007 is taken from the Association of Ethiopian
Microfinance Institutions (AEMFI) annual reports. Linear panel data models are used for the
analysis.
The result of the study reveals that there is a clear tradeoff in serving the poor clients and
being financially self-sufficient. But the MFIs are actually following an optimization strategy
in achieving the sustainability and outreach objectives.
In this connection, we have found that MFIs that have a poverty focus (MFIs with small
average loan size) charge high interest rate, have high labor cost to asset ratio and are small
in size. But the seemingly high interest rate of these MFIs arise when we compare it with that
of the commercially oriented MFIs interest rate, which might create some tension in the local
microfinance market with increasing competition in the future. When we compare with the
global level, the interest rate of these MFIs is even low and hence creates no tension.
The commercially oriented MFIs (large average loan size MFIs) charge lower interest rate
and are large in size.
Interest rate and saving mobilization are found to have a positive impact on sustainability.
Capital cost to asset affect the sustainability of subsidy free institutions (based on FSS
results) whereas firm size matters for the sustainability of subsidy dependent institutions
(based on OSS results). For breadth of outreach regression, we have not found much of the
proposed variables to be significant. Hence we propose other researcher to explore this issue
further.
Key terms: Microfinance, Sustainability, Outreach, Operational Characteristics
JEL Code: G21
2
I. Introduction
Microfinance generally defined as financial services such as savings, credit, insurance and
payments products to low income clients including the self employed has a long history
(Ledgerwood, 1999; Rhyne and Otero, 2006). However formal microfinance can be traced
back to the prior works of Grameen Bank in Bangladesh and ACCION International in Latin
America in the late 1970s (Ledgerwood, 1999; Christen, 1997).
Over the years, microfinance has not only acquired an additional dimension as a tool for
financial system development, it has also recorded impressive growth (Ledgerwood, 1999;
Woller and Schreiner, 2006). The range of products currently provided by the microfinance
industry has widened, the repayment rates have been maintained at close to 100 percent, the
number of loans per borrower has increased significantly and several microfinance
institutions (MFIs) are reportedly financially sustainable and profitable (Cull et al, 2007;
Rhyne and Otero, 2006).
In terms of scale of outreach, the number of savers and borrowers and the value of loan
portfolio have increased exponentially. Citing a publication by Consultative Group to Assist
the Poorest (CGAP), Ledgerwood and White (2006) reports that the current combined loan
portfolio of MFIs worldwide is approximately US$ 15 billion and they claim that
microfinance is believed to be growing annually at between 15 and 30 percent.
Ethiopian microfinance has made remarkable progress over the past decade, reaching almost
two million clients in a country of 77 million people. Nevertheless, financial services for the
low-income population, poor farmers and MSMEs are still characterized by limited outreach,
high transaction costs for clients, a generally weak institutional base, weak governance and a
nominal ownership structure as well as dependence on government and mother NGOs.
(Pfister et al, 2008).
As the microfinance industry has evolved and rapidly expanded both globally, questions
regarding sustainability and outreach have come to the fore. On the tradeoff between
sustainability and outreach front, for example, Morduch (1999) and Cull et al (2007) ask
whether microfinance can meet the full promise of reducing poverty without ongoing
subsidies. They also observe that high repayment rates recorded by MFIs cannot be
translated easily into profitability. Buckley (1997) questions whether MFIs are any different
from past smallholder rural and cooperative finance of the 1960s and 1970s, suggesting that
they may not be sustainable without either substantial donor subsides or a shift towards less
poor clients.
On breadth of outreach, Ledgerwood and White (2006) observes that the microfinance
industry has seen impressive growth for longer than a decade, yet still reaches only a small
percentage of its potential market worldwide.
3
On depth of outreach, despite that literature suggest that microfinance contribute towards
alleviating poverty, there is a wide literature in the area which argues that microfinance does
not reach the poorest of the poor (Woller, 2002; Hashemi and Rosenberg, 2006). Hashemi
and Rosenberg (2006) opine that microfinance excludes the poor.
So the pertinent research question to ask is: “What steps can we take to make microfinance
available to more poor people and doing so on lasting basis?
Empirical research on sustainability and outreach using econometric analysis has rarely been
undertaken in recent times. Studies conducted in different parts of the world are mostly done
in the 1990s and the literature seems saturated. Some of the prominent research of this time
includes Christen et al, (1995), Gurgand et al, (1994), Chaves and Gonzalez-Vega (1996),
Paxton and Fruman (1998), Conning (1999), Sharma and Zeller (1999). One of the most
recent and comprehensive study is that done by Cull et al (2007).
In Ethiopia most studies on MFIs are on impact evaluations and few focus on sustainability
and outreach. The studies by Amha (2007), Kidnae (2007), Kereta (2007) to name just a few,
focuses on sustainability and outreach. But these studies have serious limitations to
generalize about the sustainability and outreach of Ethiopian MFIs. The first problem is they
cover small number of MFIs. Secondly, they use descriptive analysis without statistical test
of significance. They also didn’t attempt to reveal the determinants of sustainability and
outreach. The current study is an improvement over those studies made in Ethiopia by using
15 MFIs and a panel data framework. Hence, the objective of this study is to reveal the
impact of operational characteristics of MFIs (discussed in the literature review section) on
their sustainability and outreach. Allied with this, we will also test the relationship between
sustainability and outreach, which is a highly debated topic in the microfinance literature.
The rest of the paper is organized as follows. Section two discusses the relevant literature;
section three addresses the data and methodology. Results of the study are discussed in
section four. Section five discusses some robustness tests to change in the measure of
variables and the last section six concludes.
II. Review of Related Literature
The literature review discussed below includes the definition and measures of sustainability
and outreach, the summary findings on the impact of some operational and control variables
on the sustainability and outreach and the empirical findings on the relationship between
sustainability and outreach.
4
2.1. Definition & measures of sustainability and outreach
Sustainability
Various authors (Conning, 1999; Chaves and Gonzalez Vega, 1996; Humle and Mosley,
1996) have defined sustainability, all having a common sprit on their definition. To mention,
Conning (1999:52) argues that “in most discussions, sustainability is taken to mean full cost
recovery or profit making and is associated with the aim of building microfinance institutions
that can last into the future without continued reliance on government subsidies or donor
funds.” For the purpose of this research we operationally define sustainability as the ability of
MFIs to operate continually without relying on subsides.
There are four measures of sustainability. These include the Subsidy Dependence Index
(SDI), self sufficiency measures, adjusted profitability ratios and the arrears rate. Efficiency
and productivity measures are also measures of sustainability. Based on the extant literature
(Cull et al, 2007; Okumu, 2007; Mersland and Storm, 2007) we used self sufficiency
measures and some form of adjusted profitability ratio (ROA) as measures of sustainability.
Self-sufficiency measures are generated by dividing the total of all considered incomes by the
total of all considered expenses (SEEP Network and Calmeadow, 1995). There is apparent
consensus in the literature that Operational Self sufficiency (OSS) and Financial Self
Sufficiency (FSS) are the preferred measures of self sufficiency. (Barres, 2006). The financial
self-sufficiency ratio indicates the institution’s ability to operate without ongoing subsidy,
including soft loans and grants. Unlike the financial self-sufficiency ratio, OSS is not
adjusted for subsidy.
Outreach
Conning (1999: 52) defined outreach as the term “typically used to refer to the efforts by
microfinance organizations to extend loans and financial services to an ever wider audience
(breadth of outreach) and especially toward the poorest of the poor (depth of outreach).
Schreiner (1999:2) refers to outreach as proxies for the benefit of microfinance in terms of
the number of clients or the average loan amount. Specifically he classified outreach in to six
dimensions: worth to clients, cost to clients, depth, breadth, length and scope.
For the purpose of this research we narrowly define outreach in terms of breadth and depth
and use the definition given by Conning (1999). The narrow definition is adopted because
based on Schreiner (1999:2) framework measuring worth and cost of outreach is difficult.
Length of outreach is not a new term and it is easily captured by the age of the institution
which we will consider later. Finally, there is no adequate literature on scope of outreach and
hence this variable is not used.
5
Christen (1997) uses average loan size divided by GDP per capita income as a measure of
depth of outreach but notes that, although it is widely used, it has not been systematically
tested. This can be misleading because of the heterogeneity of loan products in terms of
maturity periods and purposes, and therefore may not reflect the target market and the level
of the poverty of the clients served.
Like other authors (Cull et al, 2007; Okumu, 2007; Mersland and Storm, 2007) we used the
number of borrowers served by the MFIs as a measure of breadth of outreach. Depth of
outreach is measured by the average loan size and the percentage of women borrowers served
by MFIs.
2.2. Determinants of sustainability and outreach
Based on the extant literature, the determinants of sustainability and outreach of MFIs can be
categorized in to two: operational characteristics (a term that we coined) of the MFIs and
corporate governance characteristics. For limiting the scope of the current study we will focus
on operational characteristics of MFIs as determinants of sustainability and outreach.
Corporate governance effects will be considered in the future.
1. Operational characteristics
The operational characteristics discussed below include capital structure variables, asset
allocation; interest rate charged, cost efficiency, productivity and portfolio quality measures.
Capital structure variables

A significant amount of literature on microfinance has placed much emphasis on the sources
of funds (capital structure) as a major determinant of sustainability and outreach (Rhyne and
Otero, 1992; Christen, 1997; Ledgerwood and White, 2006).
The source of capital for the MFIs can be deposits mobilized from voluntary and compulsory
savers, commercial debt, share capital, donor subsidies and retained profits. Due to lack of
data we will not cover the effect of subsidy on the sustainability and outreach of MFIs. Hence
the literature in this issue is ignored. Besides we will not cover the effect of retained earnings
on sustainability and outreach. It is assumed that MFIs don’t distribute their profit in the form
of dividend as they have poverty alleviation motive. Hence earnings retained in one period
are expected to have a growth impetus on next period sustainability and outreach. This can be
studied using dynamic models and is beyond the scope of the present paper. Hence we will
only consider the effect of savings mobilized technically called deposit, commercial debt and
share capital on MFIs sustainability and outreach.
Rhyne and Otero (1992) argues that extensive outreach by the MFIs can be achieved and
sustained through saving mobilization and access to commercial loans. However a recent
study by Bogan (2008) reveals that there is no significant relationship between deposits to
6
asset ratio and debt to assets with either sustainability or outreach. Contrary to this he found
that share capital to asset is negatively related to OSS.
Asset allocation
The use of funds by MFIs for their major activity i.e. lending to poor borrowers is presumed
to have a positive effect on breadth of outreach as more borrowers are served. The usual
proxy for use of funds is Gross Loan Portfolio (GLP) to assets ratio. The effect of use of
funds on sustainability can be mixed as follows: an increase in GLP to assets results in higher
operating revenue and improves sustainability. However, an increase in GLP could also leads
to a decrease in operating revenue because as more loans are disbursed and left uncollected,
less revenue is generated.
In a study by Okumu, (2007) GLP to assets is negatively related with OSS but positively
related to outreach. The negative relationship between GLP to assets and OSS means that
MFIs must recover the disbursed loans as the greater the loan portfolio or high non-
performing loan portfolio, the lower the sustainability. Still we can question on whether
greater GLP to assets leads to lower repayment and whether lower repayment in fact reduce
sustainability. The positive relationship with outreach means that the more money is with the
clients, the more people are reached and hence the greater the outreach. Hence Okumu (2007)
finding call for some sort of for achieving the sustainability and outreach objective.
Lending interest rate

Lending interest rate affects sustainability and outreach through three broad channels. Firstly,
an increase in interest rate affect sustainability directly and thereby breath of outreach
assuming the profit resulting from an increase in sustainability is invested in expanding
breadth of outreach. Evidence from Asia and Latin America has shown that financial
institutions that charge commercial rates have attained sustainability and reached millions of
low income clients. Those that charge subsidized lending rates can’t achieve wider outreach
(Chaves and Gonzalez-Vega (1996)
The second channel through which interest rate affect sustainability and outreach is through
the repayment rate. In the first case, it is implicitly assumed that the MFI can increase the
lending rate progressively without any problem. This is often not true. As Stiglitz & Weiss
(1981) argue in imperfect financial market, at higher interest rate borrowers with good
projects are unlikely to borrow and mainly bad borrowers may be attracted. This can
exacerbate the default rate and cause the flow of revenue to the institution to reduce.
Therefore, although an empirical question, in the long term higher lending interest rates by
reducing the repayment rates could lead to lower sustainability and breadth of outreach. In
this case, lending rates are negatively related to sustainability and breadth of outreach.
The third channel through which interest rates affect sustainability and outreach is through
the demand for loan. In the first channel it is argued that interest rate is positively related to
sustainability and outreach. This argument assumes that the demand for credit is given or
highly inelastic and there is no credit rationing arising from information asymmetry in credit
markets. (Stigiltz &Weiss, 1981). Following classical arguments, lending rates are negatively
7
related to the demand for loans. Lower demand for credit leads to a reduction in outreach and
revenue generated by the institutions. The extent to which a higher lending interest rate
discourage borrowing and leads to lower level of sustainability and outreach, however
depends on the elasticity of demand for credit and the availability of alternative source of
external financing (Morduch, 1999)
Summarizing the above discussion in the first case, lending rate positively affects
sustainability and outreach while in the latter two cases it negatively affects them.
The study by Okumu (2007), Crombrugghe et al (2008) and Cull et al (2007) shows that real
effective lending rate positively affects sustainability. Okumu (2007) also found that real
effective lending rate is not a significant determinant of breadth of outreach partly because
the demand for credit in Uganda with a large number of the population served by the informal
sector is inelastic or highly inelastic. In addition, a number of authors have argued that access
to finance is not often constrained by interest rates but by other factors such as availability of
financial services (Robinson, 2001)
Cost
Cost includes financial costs, operating costs and loan losses. Cost affects operational self
sufficiency both directly and indirectly. For instance, an increase in costs leads to a decrease
in operational self sufficiency and by extension it leads to a decrease in breadth of outreach
and vice versa. This is a direct effect of cost on operational self sufficiency and indirect effect
on outreach. Costs can also affect sustainability and outreach through their effects on the
demand for loans. As cost increase, interest rate increase and hence demand for loan
decrease.
A study by Okumu (2007) reveals that unit cost of loan disbursed is negatively related with
sustainability and breadth of outreach. In a similar study by Cull et al (2007) found that
capital costs to assets negatively affects profitability, whereas labor cost to asset is
insignificant.
Makame (2008) shows that a higher cost per borrower, an efficiency measure in the
microfinance literature, result in larger loan size.
Borrowers per field officer

Borrower per field officer is a productivity measure in the microfinance literature. The most
important role of a low ratio of borrowers per field officer should be the possibility for field
officers to effectively educate, select and monitor borrowers thereby enhancing the
repayment rate and sustainability. Contrary to this connotation, Crombrugghe et al, (2008)
and Mersland and Storm (2007) found that borrower per field officer is positively and
significantly related to financial self sufficiency. This indicates that increasing the number of
borrowers for a given amount of field officers would contribute to profit.
Repayment rate
8
Empirical research has underlined the contribution of high repayment performance in the
success or failure of MFIs. Chaves and Gonzalez-Vega (1992) attribute the success of a
significant number of MFIs in Indonesia to high repayment or low arrears rates. Yaron
(1992), Gurand et al (1994), Hulme and Mosley (1996) have also underscored the importance
of high repayment or low default rates in institutional performance.
However as the experience of Grameen Bank shows high repayment are only a necessary but
not sufficient condition for sustainability. Portfolio at Risk (PAR), an inverse measure of the
repayment rate, was used by Bogan (2008) study. But he didn’t find any significant impact on
2. Control Variables
Many control variables that are presumed to have an effect on the sustainability and outreach
of MFIs are discussed in the literature. This includes lending method, age and size of the
MFIs, ownership type and other. But in this research, we will address only age and size of the
MFIs as we don’t have data on other control variables.
Age
Age of MFIs is one of the control variables that affect performance. The age of the
organization also affects sustainability and outreach through accumulated experience from
learning by doing, the development of operating systems, experience and training of staff and
the level of scale attained (SEEP Network and Calmeadow, 1995; Morduch, 1999; Okumu,
2007).
On the effect of age of MFIs on outreach, Mersland and Storm (2007) found positive effect of
age on breadth of outreach and loan size. However, increase in average loan with age is not a
sign of mission drift as Christen (2001) suggests. The age of the MFI-Client relationships
may imply that future loans following successful repayment increase compared to past and
present loans. Mission drift instead is a shift in the composition of new clients or a
reorientation from poorer to wealthier clients among existing clients. (Cull et al, 2007)
Cull et al (2007) also found positive relation between financial performance and age of MFI.
However, Mersland and Storm (2007) found negative relationship between age of MFI and
ROA. Hence the findings are mixed.
Firm size
The study by Bogan (2008), Mersland and Storm (2007), and Cull et al (2008) shows that log
of assets (a firm size measure) is positively and significantly related to financial performance.
This indicates that the larger institutions increased self sufficiency is likely associated with
the delivery of services to a larger group of clients or with extending credit in the form of
larger loans to clients although the channel through which this effect to be materialized is not
clear. Allied with this Bogan (2008) found that log of assets is positively related to number of
borrowers.
9
On the relationship between loan size and size of the MFIs, Hudon and Traca (xxx) have
found a negative relationship implying that larger MFIs are pro-poor whereas Cull et al
(2007) found positive relationship implying the reverse.
2.3 Relationship between sustainability and outreach
As noted by Woller and Schreiner (2006) the often expressed fear in the debate of the
relationship between sustainability and outreach is that a focus on financial self sufficiency
will divert MFIs attention and resources away from their core objective of poverty alleviation
and core poor market (“mission drift”). This fear is supported by Mersland and Storm,
(2007), Okumu, (2007) and is attributed to several reasons:
The first is cost related. It is argued that the poor tends to concentrate in rural areas, which is
isolated with very poor physical infrastructure and financial market. (Schreiner,1999).
The second argument is related to risk. Apart from engaging in economic activities perceived
to be very risky i.e. agriculture, low income earners tend to lack the physical collateral
required by the traditional formal financial institutions (Zeller, 1994).
Related to the issue of risk and cost is the information asymmetries prevalent in poor
communities. Absence of credit information means chances of moral hazard and adverse
selection are very high (Navajas et al, 2003).
Finally, there is the question of loan size. Hulme and Mosley (1996) find that there is a
negative relationship between loan size and cost of administration. The smaller the loan size,
the higher the cost of administering the loan. As a result, high costs mean low revenue unless
there are substantial economies of scale.
Despite the above arguments, which suggest existence of a negative relationship between
expending access to financial services for low-income earners and financial sustainability,
Some authors believe that such an inherently dichotomous relationship does not exist (Woller
and Schreiner, 2006; Schreiner, 1999, Christen, 1997). Crombrugghe et al (2008) also
showed that there is no significant relationship between sustainability and depth of outreach
as measured by the percentage of women borrowers.
The study by Makame(2008) in East African MFIs shows that the breadth of outreach
appears to be negatively related with ROA, which implies that an expanded operation to
encompass wider client base might necessarily compromise the ROA, this again may be a
result of taking abroad more clients at the expense of observing their creditworthiness.
Furthermore, the breadth of outreach is observed to have a negative relationship with average
loan, implying that the higher the average loan the lower the breadth of outreach and vice
versa.
III. Data and Methodology
Data
10
There are 30 MFIs in Ethiopia as of 2009. But not much time series data is available for
many of the MFIs. So we are forced to use only 15 MFIs which have five years data as a
sample for the study. Although the sample can’t be representative of the whole MFIs owing
to its non-randomness, the 15 MFIs, with a five year data from 2003-2007, are presumed to
give a good picture of the 12 year old Ethiopian MF industry.
All the data are quantitative and taken from the annual reports of the Association of Ethiopian
Microfinance Institution (AEMFI). Some benchmark figures are taken from the Microfinance
Information eXchange (MIX) data base. For conformity with our study time period of 2003-
2007 we have taken benchmark figures for these periods.
The Econometric model

The model used to reveal the impact of operational characteristics of MFIs on their
sustainability and outreach is static panel data regression and the choice of the fixed effect
versus random effect is selected by the Hausman test.
With respect to functional form selection, the literature doesn’t recommend a certain
specification. Some authors used the log models (Okumu, 2007) while other used level form
(Cull et al, 2007). We have checked the appropriateness of both models using residual
normality tests and found that the log model seems to fit to the data. Hence we used all
variables in log forms.
The basic two-way panel data model looks like as follows:
lnSit = α + βilnOCit + βjlnCit + µi + γ t + εit --------------------------(1)

lnBreadthit = α + βilnOCit + βjlnCit + µi + γ t + εit ------------------(2)
lnDepthit = α + βilnOCit + βjlnCit + µi + γ t + εit --------------------(3)
Where
lnSit = Sustainability indicators such as Financial Self Sufficiency(FSS), Operational
Self Sufficiency (OSS) and Adjusted ROA for MFI i in period t.
lnBreadthit = Breadth of outreach indicator used is the number of borrowers for MFI i
in period t
lnDepthit = Depth of outreach indicators used are average loan size (AvLnSz) and
percent of women borrowers served by MFI i in period t
lnOCit = vector for operational characteristics of MFI i in period t. This includes
Capital to assets, Deposit to assets, Debt to assets, GLP to assets, interest
rates, capital costs to assets, labor costs to assets, borrowers per field officer,
and Portfolio at Risk (PAR). We have also included some of the dependent
variables in other regression to check the relationship between
lnCit = vector for control variables for MFI i in period t. This includes MFIs age and
size.
µi = time-invariant unobserved heterogeneity for MFI i
11
γ t = time dummies to check the dynamics in sustainability and outreach
εit = time varying error term for MFI i in period t
IV. Results
1. Summary Statistics and Correlations
The mean, standard deviation and one sample t tests with African and worldwide MIX
benchmark results are described in Table 1.
[Insert Table 1]
For most variables we can use the MIX benchmark data and compare it with AEMFI data as
the variables are measured in percentages. However, this comparison can’t be done for assets
and average loan size variables. These variables are measured in local currency (Birr terms)
in AEMFI data whereas the MIX figures are in dollar terms. Our effort to covert the Birr
figures in to dollar can’t succeed because of difficulty in finding suitable exchange rate data
for the period 2003-2007. Besides this, there is no suitable MIX benchmark data available for
debt to assets, capital costs to assets and labor cost to assets.
However, one feature of MIX benchmark data has to be seriously thought of. The benchmark
figure in the different reports for the same year is inconsistent. For example in the 2003-2005
benchmark report and 2004-2006 benchmark report, we get data about 2004 for specific
variables such as the number of borrowers, the percentage of women clients etc for the
African MFIs and the global average. But the problem is the two series of reports have
different figures for 2004 for a single variable say number of borrowers. This created a lot of
frustration on us. We have tried to inquire the source of these variations from the MIX people
themselves, but we didn’t get any response for our queries. We used the figures for earlier
series of report for each year to add consistency. Hence, the use of the benchmark figures is
only to provide some insight about the performance of Ethiopian MFIs as compared to
African and global benchmarks, not as a serious analysis based on factual data.
When we come to the other variables we got the following results:
The MFIs have an average age of 7 years which is significantly lower than that of African
and worldwide benchmarks. This tells us Ethiopian MFIs are still young.
With respect to capitalization, Ethiopian MFIs are significantly financed by shareholder funds
in a significant way as compared to many African and worldwide MFIs. The average capital
to assets ratio is 45% which is by far greater than the 31.52 % of Africa and 30.48% for
worldwide MFIs. This capital to asset ratio contains donations and this might partly also
indicate the heavily subsidized nature of Ethiopian MFIs. This shows that debt as a source of
finance is not used by many of the MFIs. This is clearly because of the underdevelopment of
the commercial debt market in the country.
12
Deposits to assets ratio stood at 10% and this figure is at par with the African MFIs. The
worldwide figure of zero can’t be genuine one and hence ignored.
GLP to assets ratio is about 70% and this is at par with African MFIs but fell short of the
worldwide benchmarks.
The average number of borrowers served by the MFIs in the period is 82,796 which is by far
larger than both the African and worldwide benchmarks. This is very encouraging
achievement in raw terms. But as compared to the poor population in each of the benchmarks
it may be low as Ethiopia is one of the least developed country in the world having many
poor people.
With respect to targeting women borrowers, Ethiopian MFIs fell short of both African and
worldwide standards. They stood at 46% whereas the figure for Africa and the world is
61.68% and 64.4% respectively. Hence we can say depth of outreach is small.
Ethiopian MFIs have a significantly lower profitability rate as measured by ROA as

compared to the benchmarks. The figure stood at -3% whereas ROA for Africa is -1.18% and
that of the world standard is 1.48%.
With respect to operational self sufficiency, Ethiopian MFIs are better than African MFIs but
at par with the world average. The reverse holds true for FSS. They are at par with African
MFIs but fall short of the worldwide standards. So a problem of sustainability especially in its
true form, FSS, is a clear problem for Ethiopian MFIs. The overall picture is that Ethiopian
MFIs are lagging in terms of sustainability as compared to the benchmarks.
Somewhat surprisingly, the interest rate charged by Ethiopian MFIs is lower than that of the
benchmark. This may be one cause of poor financial performance. It is 20%. Whereas the
average rate for Africa and worldwide standards stood at 29.22% and 32.4% respectively. We
can roughly infer Ethiopian MFIs are pro-poor as compared to the benchmark because of
their low interest rate. But this conclusion can be simply wrong because rate of return for
projects and cost structures differ across countries. In Ethiopia rate of return for projects are
low because of the underdevelopment of the economy.
In terms of productivity the MFIs are better than the benchmarks. The loan officers serve
more borrowers. But this still may be a bad performance as quality of outreach decrease when
more borrowers are served by one loan officer.
Portfolio at risk is at par with the African standard but fell short of the world. It is 5% while
the world wide figure is 2.42%
The correlations among variables shows that significant high correlations exist between total
assets and labor cost to assets, number of borrowers and assets, number of borrowers and
labor cost to assets, FSS and ROA, OSS and FSS, average loan size and GLP to assets. This
inter-correlations form the basis of modeling discussed below to address for multicollneaity
issues.
13
[Insert Table 2]
However, as can be seen in table 2 the correlation between average loan size and the
percentages of women borrowers although in the expected direction significant is so small in
value (-0.134). This may tell us the two measures can’t be used to measure the same
theoretical construct i.e. the poverty profile of clients. Hence care has to be taken in this
regard.
2. Panel Data Results
Procedural briefs
Maximum efforts are made to follow pre and post estimation diagnostic tests. These includes
Hausman test for the selection of the fixed vs. random effect model, tests of residual
normality and outliers detection to check for model fit, and use of robust regressions to
address for hetrodecasacity. Different models are also estimated for each dependent variable
to address for multicollearity problem. We have also conducted Ramsey RESET tests for the
functional form of conditional mean.
First a hausman specification test is conducted to select from fixed and random effects
models. In the test all explanatory variables including time dummies are included. The result
of the test confirms the following. For Ln FSS, Ln OSS, and Ln Women regressions, the
random effect model seems appropriate. But tests show that the random effect model has no
significant improvement over the Pooled OLS results. So we report Pooled OLS results. For
Ln Borrower, Ln AvLnSz and Ln ROA regressions the fixed effect model is found to be
appropriate and it has clear improvement over the Pooled OLS method.
The reason the fixed effect model shows improvement over Pooled OLS in Ln Borrower, Ln
AvLnSz and Ln ROA can be attributed to omitted variables bias. The appropriateness of the
fixed effect also tells us that the omitted variables are correlated with those found in the
model. One potential omitted variable that is expected to be correlated with explanatory
variables is the amount of subsidy received by the MFIs. A priori we feel that subsidy is
correlated with interest rates, sustainability, breadth and depth of outreach, cost efficiency
and PAR. Subsidy might help MFIs to increase their breadth of outreach and to reach poorest
borrowers. However, subsidy may not have an impact on sustainability of MFIs. If subsidy is
time varying under this circumstance, it will create endogenity due to omitted variable bias
and hence instrumental variable regression should be used rather than simple panel data
models. Since we can’t have data on subsidy we can’t verify this proposition and hence a
further study shall try to uncover this issue. For Ln FSS, Ln OSS and Ln Women the model
we proposed (Pooled OLS) seems good that is why there is no unobserved heterogeneity in
the MFIs.
Our model seems appropriate as compared to instrumental variable models as many of the
omitted control variables like lending method and ownership type are time invariant and thus
removed through the within effect transformations in the fixed effect model. Had the random
effect was appropriate, their effects could have been estimated. With respect to subsidy we
feared that it is time varying and possibly create endogenity problem as it is correlated with
14
many regressors. But our measure of capital to asset, we feel, implicitly has the subsidy
component. But it should be known that we can’t disentangle the shareholders funds effect
from the donations effect.
For all regressions robust standard errors are computed to solve the problem of hetrodsecacity
and in some regression abnormality of the residuals.
To address for multicollinearity, we have used the Variance Inflation Factor criterion and the
correlations matrix. From this analysis it is found that high inter correlation and VIF values
( > = 10) exist between borrowers, assets and labor cost to assets on the one hand and
between GLP to assets and AvLnSz on the other. This forces us to adopt a three specification
approach to each of the regressions by dropping collinear variables at each stage. So in the
first case we drop both assets and borrowers and add labor cost to assets. Also we drop GLP
to assets and add AvLnSz. In the second case, we drop both borrowers and labor cost to
assets and add Assets. Also we both drop AvLnSz and add GLP to assets. In the third case,
we drop assets and labor cost and adds borrowers. Also we drop both GLP to assets and
AvLnSz.
FSS regression
As indicated previously the random effect model seems appropriate for the Ln FSS
regression. But the tests of the unobserved heterogeneity confirmed that they are not
statistically different from zero. So the pooled OLS results are used. Based on these results
the following variables are found to be significant determinant of MFIs FSS across
specifications. These include deposits to assets, average loan size, interest rate, capital cost to
assets and GLP to assets. Other variables are found to be insignificant. Some surprising
insignificant variables are borrowers, PAR, age and size of the MFIs. Specially the
insignificance of borrowers tells us that MFIs enhanced sustainability comes from extending
large loan size not from extensive breadth of outreach. Besides time dummies like age are not
significant which tells us that the MFIs don’t improve their sustainability overtime. This may
be due to the brief nature of the period considered or because of inability of the MFIs to reap
learning curve economies.
[Insert Table 3]
A brief explanation for the significant variables follows:
Deposits to assets positively impact financial self sufficiency. This is in contrast to Bogan
(2008) who doesn’t found any significant relationship between deposits to assets and
sustainability. The channel through which saving mobilization affect sustainability can be
two: one saving is a cheaper source of finance for the MFIs as compared to commercial
source of credit. Second, irrespective of the cost issue, funds available through saving will
help to expand breadth of outreach and there by sustainability. The exact channel saving
affects sustainability in this research is through the cost channel not through the borrower’s
channel. As we can see below cost matter for sustainability but borrowers doesn’t matter for
sustainability. Besides deposits to assets doesn’t affect borrowers regression i.e. even if
15
money is mobilized through savings it doesn’t help MFIs to increase their borrower’s
number. Correlation shows that deposits to assets and cost components are significantly
negatively related. Our finding helps to explain the positive role of saving mobilizations for
sustainability as deposits are lower cost source of finance for the MFIs.
A higher average loan size is associated with better financial self sustainability. Hence this
points to the tradeoff that exists between serving the poor and attaining financial self
sufficiency. This is consistent with the views of Okumu, (2007) and Mersland and Storm
(2007). This calls for a clear market segmentation in the MFIs industry as the two are
competing objectives. This creates two segments of MFIs one that is subsidized and caters for
the ultra poor and the other working on market based principles and targets the upper
segments of the poor.
As expected interest rate impact the financial self sufficiency of MFIs positively. This result
is in conformity with the findings of Okumu (2007), Crombrugghe et al (2008) and Cull et al
(2007). Hence the prediction that interest rate affect sustainability directly and positively is
confirmed. The literature discussed claim that a higher interest rate could negatively affect
sustainability through a reduced repayment rate or low demand for loans. The correlation
between interest rate and repayment is low and insignificant (-0.075) indicating that a higher
interest rate don’t deter borrower’s to pay their loans. This can be attributed to several
reasons: one borrower’s projects may be so profitable as compared to the interest rate they
pay. But this seems implausible. The other reason is the poor fear to default and they pay
what they earned (or even by borrowing from friends and moneylenders) and at the end of the
day they don’t come of the vicious circle of poverty. This seems plausible.
The correlation between interest rate and demand for loans proxied by the number of
borrowers is significant but it is modest, not high. (-0.474). This still corroborates our
primary intuition that a rise in interest rate doesn’t reduce the demand for loans. But it should
be seen that interest rate impact heavily demand for loan rather than repayment rate i.e. a
higher interest rate deter new borrowers and also force existing borrowers to stop taking
further credit. In many developing countries credit demand is inelastic. What matters is
credits availability, not its price. So the theory that interest rate can affect sustainability
negatively through the repayment rate and demand for loans channels is not confirmed by our
data. The reader can see also both the repayment rate and the borrower’s numbers are not a
significant determinant of financial self sufficiency.
A higher capital cost to assets reduces financial self sufficiency. This calls for cost
management on the part of the MFIs. This finding is consistent with the findings with
Okumu, (2007) and Cull et al (2007). Consistent with Cull et al (2007) we also found no
significant impact of labor cost to assets on MFIs sustainability. This tells us labor costs are
either small due to low salary paid to staff which is a typical reality in Ethiopian salary scales
or small number of staff hired by many MFIs or a combination of factors. The effect of
capital costs on sustainability is the direct impact discussed in the literature. Indirect impacts
of increase in cost on sustainability through a higher interest rate and a reduced demand for
loans is not confirmed. Actually one can see that the correlation between capital cost to assets
16
and interest rate is reasonably high and significant (0.545) i.e. MFIs uses cost-plus pricing.
But this link breaks because a higher interest rate doesn’t reduce demand for loans as most
clients have inelastic loan demand.
A higher GLP to assets improve financial self sufficiency. So more assets allocated to
productive purpose is beneficial to the improvement of MFIs FSS. This finding is in contrast
with the result of Okumu (2007) who finds a negative relation between GLP to assets ratio
and sustainability. Actually Okumu used OSS rather than FSS. But our results for OSS are
the same as FSS as discussed on the robustness tests section. Our results mean more loans
disbursed results in increased sustainability through increased revenue. The fear expressed
that more loans disbursed create huge repayment problems is not supported. The correlation
between GLP to assets and repayment rate is small and but significant (-0.353) and
repayment rate is not a significant determinant of sustainability. But when we disaggregate
GLP to assets itself we will get two components: average loans size and the number of
borrowers. As seen above average loan size is positively related to FSS whereas borrowers
are not. This tells us that the positive impact of GLP to assets on FSS is because of large
average loans size, not because of a large number of borrowers.
Borrower regression
For the Ln Borrower regression the fixed effect model is appropriate and the model has clear
improvement over the Pooled OLS. Based on this result, only the age of the MFI is found to
be significant determinant of MFIs borrowers’ size across specifications i.e. older MFIs serve
more borrowers. However, the R2 of this model is low. This calls for the search for many
other explanatory variables not included in this model. Some surprising insignificant
variables are deposits to assets, debt to assets, GLP to assets, interest rates, costs, average
loans size, and time dummies.
[Insert Table 4]
A discussion on the insignificant variables follows:
Rhyne and Otero (1992) attribute the extensive outreach of MFIs was achieved through
saving mobilizations and access to commercial loans. However, in our regression both
deposits to assets and debt to assets are insignificant determinants of the number of
borrowers. This shows that Ethiopian MFIs saving mobilization is so weak to impact breadth
of outreach. At the same time commercial sources of credit are very meager in Ethiopian
financial system approach and hence this source of capital can’t impact breadth of outreach.
Because source of finance includes share capital, subsidy and retained profits we can’t
attribute the increased breadth of outreach of Ethiopian MFIs to any of the sources of finance.
This reaffirms the irrelevance of capital structure to the growth of firms’ literature.
Okumu (2007) presumed that a higher GLP to assets should be associated with greater
breadth of outreach as more borrowers are served. But this proposition still can’t be satisfied
in our regression. This may be because the average loan given to borrowers may be large and
hence inflating the GLP to assets figure. If this is the case obviously more borrowers can’t be
17
reached even if GLP to assets is increasing. From this we can infer that Ethiopian MFIs are
not pro poor because they extend large average loan size, a conclusion which is also verified
in the summary statistics section.
As discussed in the Ln FSS regression although interest rates and demand for loans are
significantly correlated, the figure is modest. We infer that demand for loan is inelastic. So,
interest rates don’t curb the growth in size of borrowers as most Ethiopians are credit
constrained.
Cost also doesn’t matter for expanding the number of borrowers. This may be due to the low
cost of running a micro bank using group lending.
The most paradoxical finding is the relationship between average loan size and the number of
borrower. A priori, given a limited amount of funds, a higher average loan size will deter
MFIs to reach more borrowers. But this proposition is not confirmed by the study. Hence this
finding can’t be believed on.
Lastly, time dummies are not significant either. This may be because of the brief nature of the
period considered.
Average loan size regression
For average loan size regression the fixed effect model is appropriate and the model has clear
improvement over the Pooled OLS. Based on these results the following variables are found
to be significant determinant of MFIs AvLnSz across specifications. These include FSS,
interest rate, labor cost to assets and assets. Other variables are found to be insignificant.
Some surprising insignificant variables includes MFIs age and time dummies where
Mersland and Storm, (2007) and Cull et al, (2007) found that overtime MFIs give large loan
size. This proposition is not supported because the time period is brief or because the MFIs
stick to their original mission.
[Insert Table 5]
A brief explanation for the significant variables follows:
As explained in the Ln FSS regression result, there is a positive relation between financial
self sustainability and average loan size. This result is also confirmed here.
Interest rate is negatively related to average loan size. This should not be interpreted in a
casual sense rather it means those MFIs that target the very poor (small average loan size)
charge the highest interest rate. This is partly because they incur high cost to reach the poor
and have small borrower size. So they have to recover this large cost by charging higher
interest rate. But the question is: is the interest rate charged by this pro-poor small MFIs
extremely exploitative for the poor to remain poor? The answer is not as such. If we see some
statistics, on a worldwide level the median interest rate charged by MFIs is 26% (Rosenberg
et al, 2009) where as only three pro-poor MFIs have an average interest rate that charge
18
around 28% almost at par with the global average. Hence this should not be seen as a serious
problem.
Labor cost to assets is negatively related to average loan size. Those MFIs that target the poor
have higher labor cost to assets. This also seems paradoxical because Makame (2008) found
that those MFIs that have larger cost will have larger loan size. So Makame’s arguments are
straightforward and right. Because larger cost will erode the sustainability of MFIs, the MFIs
increase their average loan size to get profit. On the other hand, Ethiopian MFIs follow a
different strategy. To recover their high cost, they charge relatively high interest rate but want
to show that they are pro-poor by extending small loan size.
So the question is which one is a good strategy to follow to contain the increased cost if the
goal is poverty alleviation? Is it increasing average loan size or increasing interest rate? Or
decreasing both while being subsidized? Increasing average loan size is totally a mission drift
if it is the right measure for poverty profile of clients. So we can’t in the first place consider
as a MFI having a poverty alleviation goal. Increasing interest rate in the face of a poverty
alleviation goal (small loan size) is within the realm of the objective function as long as it is
not exploitative to the poor. Even if the two strategies are evil by themselves, we think
increasing interest rate is better than total mis-targeting. However, the ideal solution is to
decrease both and this can happen only if the MFI is subsidized and this was what we were
preaching in earlier sections that the microfinance industry has to be divided in to two market
segments.
Assets (a measure of the MFI size) are positively related to average loan size. This is
consistent with Cull et al (2007) finding. This means large sized MFIs have low poverty
target.
The overall picture here is those MFIs that have a poverty target charge high interest rate,
have high labor cost to assets and are small in size. For helping the poor to come of poverty
they either have to reduce their interest rate or manage their labor cost.
V. Robustness Analysis
The robustness of the findings is checked by changing the way dependent variables are
measured. It is known that sustainability is measured using FSS, but alternative measures like
OSS and Adjusted ROA are used in the literature. Hence these measures are employed to see
the consistency of the results. Alternatively depth of outreach is measured using the
percentage of women borrowers served by the MFIs.
OSS regression
The random effect model seems appropriate for the Ln OSS regression. But the tests of the
unobserved heterogeneity confirmed that they are not statistically different from zero. So the
pooled OLS results are used. Based on these results the following variables are found to be
significant determinant of MFIs OSS across specifications. These include deposits to assets,
average loan size, interest rate, assets, borrowers, and GLP to assets. Other variables are
found to be insignificant.
19
[Insert Table 6]
Deposits to assets, average loan size, interest rate and GLP to assets effect on OSS are the
same as that of FSS. So with respect to this variables our result is are robust to the way the
self sufficiency of the MFIs is defined.
While capital cost to assets affect FSS it doesn’t affect OSS. On the contrary Assets and
Borrowers affect OSS but not FSS. This is interpreted as follows. A higher capital cost to
assets affects the self sufficiency of a subsidy free institution. However, if the MFIs is
subsidy dependent capital cost to asset doesn’t have an effect on its self sufficiency. This is
because the subsidy received by the institution will cover this cost. But once subsidies are
lifted cost becomes an important determinant of sustainability. For subsidized MFIs, typically
small ones, size matters most to be sustainable rather than cost management.
ROA regression
The fixed effect model is appropriate for the Ln ROA regression and the model has clear
improvement over the Pooled OLS. Based on these results AvLnSz and GLP to assets are
found to be significant determinant of MFIs ROA across specifications. Other variables are
found to be insignificant.
[Insert Table 7]
The R2 of this model is low. This calls for the search for many other explanatory variables not
included in this model. Although correlations between ROA and FSS are high (may be
spurious), this is also a verification that the concepts of profitability as measured by ROA in
many mainstream finance based research are different from sustainability measures such as
FSS and OSS which is used in the microfinance area.
The impact of average loan size and GLP to assets on ROA is the same as those interpreted
on FSS.
Women regression
The random effect model seems appropriate for the Ln Women regression. But the tests of
the unobserved heterogeneity confirmed that they are not statistically different from zero. So
the pooled OLS results are used. We don’t found any consistent result for the women
regression across specifications.
[Insert Table 8]
The R2 of this model is low. This calls for in search for many other explanatory variables not
included in this model. This also tells us that the two concepts that are used to measure the
poverty profile of client’s i.e. average loans size and the percentage of women borrowers
served might be different and may not catch poverty profiles of clients.
VI. Concluding Remarks
20
The main objective of this study is to reveal the impact of operational characteristics of the
MFIs on their sustainability and outreach. Such operational characteristics considered were
capital structure decisions, asset allocation; interest rate charged, cost efficiency, productivity
and portfolio quality measures. To reveal these relationships the MFIs age and size are used
as control variables. The main results of the study are summarized as follows:
Sustainability regression
Summary findings reveal that sustainability of the MFIs are affected by deposits to assets,
average loan size, interest rate, capital cost to assets and GLP to assets for FSS specification
and deposits to assets, average loan size, interest rate, assets, and GLP to assets for OSS
specification. The ROA specification doesn’t yield any discernable result suggesting that it is
a different profitability measure as opposed to sustainability measures.
We found that saving mobilization, as cheaper source finance (not as a means to increase the
number of borrowers), help to increase sustainability. Hence, commercial oriented MFIs
should push for more saving mobilization efforts by studying the determinants of saving
mobilizations viewed from both the clients and the MFIs side. Trade off in serving the poor
and being financially self sufficient is also found which calls for market segmentation. An
increase in interest rate also increases sustainability directly and not indirectly through the
repayment rate and the demand for loans channels discussed in the literature. Capital costs to
assets matter for FSS but not for OSS. This tells us that cost management is a crucial task for
subsidy free institution as opposed to subsidized ones. For subsidized ones, rather than cost
management, it is economies of scale (size) that matters for sustainability. We also found a
higher GLP to assets increase sustainability but this is because of large average loans size and
not because of a large borrower’s base reaffirming the tradeoff in serving the poor and being
financially self sufficient.
Some insignificant variables in sustainability regression are also worth mentioning. This
includes PAR, age and size of the MFIs and time dummies. Repayment rate doesn’t matter
for sustainability. Because of the brief time period of the study or due to in capability of the
MFIs to reap the benefits of experience curve economies both age and time dummies
becomes insignificant and lastly size of the MFIs also doesn’t matter for sustainability of
subsidy free institutions.
Breadth of outreach regressions
When we come to breath of outreach, almost all variables expect MFIs age is insignificant.
The significance of age tells us overtime MFIs learn to expand the borrower’s size but as we
see above they don’t learn to increase sustainability. This shows they were following a
growth strategy whereas they don’t look at what they were financially doing. This is common
in a young industry. But they also should give due attention to what they actually has done in
the past (i.e. their financial performance). The insignificance of many variables tells us there
are other drivers for breadth of outreach and researchers should look at this issue in the
future. But from the variables we include in our analysis we found some surprising
21
insignificant variables such as deposits to assets, debt to assets, GLP to assets, interest rates,
costs, average loans size, and time dummies.
Rhyne and Otero (1994) proposition that saving mobilization and access to commercial
source of finance enhances breadth of outreach can’t be supported. This may be because
saving mobilization efforts are so low in Ethiopian MFIs and access to commercial credits is
limited. The insignificance of GLP to assets is the same as that of average loan size which is
very difficult to understand. Given limited funds, an increase in average loan size should have
decreased the number of borrowers. But this doesn’t happen. This has to be explored in the
future. The insignificance of interest rate tells us that demand for credit in Ethiopia is
inelastic. Cost also doesn’t matter for expanding breadth of outreach as group lending is
innovative way of cutting transaction costs.
Depth of outreach regression
In depth of outreach measures we used average loan size and the percentage of women
borrowers served. But we didn’t get any significant result in the percentage of women
borrowers’ regression and hence we left this part. When we come to the average loan size
regression, we found the following variables significant: FSS, interest rate, labor cost to
assets and assets.
Those MFIs that have poverty focus (small average loan size) charge relatively high interest
rate as compared to commercially focused MFIs (MFIs who have large average loan size).
But when we compare the average interest rate of the pro-poor MFIs with the worldwide
average figures they are low and hence we can’t treat the pro-poor MFIs as exploitative. This
however is not to mean the small MFIs should leave the issue of interest rate unattended.
With increasing competition expected in the local microfinance market in the year to come,
customers will start comparing the interest rate of each MFI while availing services.
Hence either the whole Ethiopian MF industry has to go to a higher interest rate regime to
compare with worldwide figures and thereby increase the industry’s sustainability or else the
pro-poor MFIs has to scheme out a strategy for being sustainable without the interest rate
premium that they have as compared with the commercially oriented MFIs. Otherwise it will
create frustration with the pro-poor MFIs managers with increased competition in the local
MFI market.
We have seen that MFIs can’t be pro-poor and at the same time commercial oriented because
of the tradeoff results discussed earlier. This calls for clear market segmentation strategy to
be followed along the line of tradeoff. This result can arise because of three scenarios:
 The MFIs are pro-poor but not financially sustainable.
 They are not pro-poor ( i.e..commercially oriented) but financially sustainable
 They are not pro-poor and at the same time not financially sustainable
22
When we categorize Ethiopian MFIs in these three scenarios, we can classify them into
scenario one and two with a bit mixed picture.
 The pro-poor MFIs fall in scenario one. They are pro-poor in strict sense of the term.
They have small average loan size, but they are not financially sustainable as
compared to the commercially oriented one. They charge relatively high interest rate
to be financially sustainable although it doesn’t help them much. Rather this high
interest rate overshadows their pro-poor focus. The better option to be sustainable
should have been size expansion for this group of MFIs although without client mis-
targeting ( i.e. a focus on large loan size).
 The commercial MFIs fall in scenario two. They are not pro-poor, an indication of
either unscientific client targeting strategy which marginalize the poor or a deliberate
targeting strategy exclusively made to target the upper segment of the poor
population. They have large average loan size, but are financially more sustainable as
compared to the pro-poor MFIs. Contrary to their commercial focus they charge low
interest rate to be pro-poor, although high interest rate can help them to pursue their
commercial goals perfectively. They follow the strategy to increase their
sustainability through economies of scale (rather than high interest rate) which can a
more justifiable reason in the microfinance sector. But these MFIs can even increase
interest rate due to the inelastic nature of loan demand in this country.
From the insignificant variables age is one of them. So, overtime there is no mission drift in
MFIs as they are not extending larger loan size.
Limitations of the study
Serious care should be taken in generalizing the findings of this study. This is because several
constraints affect the researcher to conduct a flow blown study. The major constraint is
resource limitation; the other being time limitation. This paper is prepared as part of PhD
study and its sole purpose is to serve the graduation requirements of the PhD. It can’t be
viewed as a consulting project which sees every dimensions of the sustainability and outreach
of MFIs and the factors that affect them. Some of the limitations include:
 We haven’t collected primary data from clients or the MFIs themselves which might
help either in corroborating or refuting the findings of this study which is based on
total secondary sources of data. Specific questions that need answers by collecting
client level primary data could have been the perception of clients on whether the
interest rates charged to them is high or low, the share of interest rate to the total cost
structure of the clients and the likes. From MFIs we could have obtained a primary
data which shows whether they have a clear client targeting policy based on poverty
studies or some versions of such a policy. But as said above, primary data is not
collected because of resource and time constraints and the purpose of the research is
to achieve a limited academic goal rather than consulting type goals.
23
 The study only looked at two dimensions of outreaches i.e. breadth and depth of
outreach. It excluded worth to clients, cost to clients, length and scope of outreach.
The worth to clients and cost to clients are excluded because there are no still
operational measures available to measure them reliably and the data to measure
them is difficult to get. Length of outreach is catched by the MFI age which is
included in the modeling process. Scope of outreach is relatively less difficult to
measure but ignored because it is not much discussed in the literature, its
determinants are not fully known and to limit the scope of the study to breadth and
depth of outreach which are frequently a research agenda in the MF literature.
 The data used cover the period 2003-2007. This might make the results of the study
not consistent with the current state of the MFIs sustainability and outreach
performance, being a bit old data. But this happens because data are not available
after 2007 in a published way from AEMFI at the time of doing the project. Although
2008 data is available now, some of the preliminary statistics shows nothing different
from those of the previous years. Hence, although a bit old data might be a cause for
concern for some MFIs, seen at the aggregate Ethiopian MFI level to which the
research is mainly focusing, this problem has little practical relevance as nothing has
changed significantly within two years (2008 and 2009).
References
Amha W. (2007) Managing growth of microfinance institutions: Balancing Sustainability and

reaching large number of clients in Ethiopia, AEMFI Occasional paper No 18, Addis
Ababa, Ethiopia
Barres I. (2006) Financial Self Sufficiency (FSS), the Micro banking Bulletin, 2006, 13: 21-
22, Microfinance Information Exchange Market Inc: Washington DC
Bogan V. (2008) Microfinance Institutions: Does Capital Structure Matter, Cornell
University
Buckley G. (1997) Microfinance in Africa: Is it either the problem or the solution, World
Development, 25 (7), 1081-1093
Chaves R.A. and Gonzalez-Vega C. (1996) Design of Successful rural financial
intermediation: Evidence from Indonesia, World Development, 24 (1), 65-92
Christen R. (2001) Commercialization and Mission Drift: The transformation of microfinance
in Latin America, Occasional Paper No 5, CGAP, Washington DC
Christen R., Rhyne E. and Vogel R. (1995) Maximizing the outreach of microenterprise
finance: Analysis of successful microfinance program, Program and operation
assessment Report 10, USAID, Washington DC
Christen R.P (1997) Banking Services for the poor: Managing for financial success, ACCION
International, Washington DC
Conning J. (1999) Outreach, Sustainability and Leverage in monitored and peer-monitored
lending, Journal of Development Economics, 60, 51-77
Crombrugghe A. D, Tenikue M and Sureda J (2008) Performance Analysis of a sample of
Microfinance Institutions in India, Annals of Public and Cooperative Economics,
79(2), 269-299
24
Cull, R., A. DemigÄuc-Kunt, and J. Morduch (2007). Financial Sustainability and outreach
and outreach: A global analysis of leading microbanks. Economic Journal, 117 (517),
107-133.
Gurgand M. Perderson G. and Yaron J. (1994) Outreach and Sustainability of six rural
financial institutions in Sub-Saharan Africa, World Bank Discussion Paper No 248,
Washington DC
Hashemi S. and Rosenberg R. (2006) Graduating the poorest in to microfinance: Linking
safety nets and financial services, CGAP Focus Note 34
Hudon M. and Traca D.(xxx) Subsidies and Sustainability in Microfinance, Solvay Business
School Working Paper, WP-CEB 06-020
Hulme D. and Mosley P. (1996) Finance against poverty, Volume 1, Routledge. London
Kereta B. (2007), Outreach and Financial Sustanability and outreach analysis of
Microfinance Institutions in Ethiopia, African Economic Conference, Addis Ababa
Kindane A. (2007) Outreach and Sustainability of the Amhara Credit and Saving Institutions
(ACSI) Ethiopia, MA thesis, Norwegian University of Life Sciences
Ledgerwood J. and White V. (2006) Transforming Microfinance institutions: Providing full
financial service to the poor, The World Bank, Washington DC
Ledgerwood, J. (1999), Sustainable Banking with the Poor Microfinance Handbook: An
Institutional and Financial Perspective, the World Bank, Washington, DC.
Makame A. H. (2008) Empirical findings on Cognitive Dissonance Around Microfinance
Outreach and Sustainability; Birmingham Business School, Birmingham University
Mersland R. and R.O. Storm (2007) Sustanability and outreach and Corporate governance in
Microfinance Institutions, Agder University, Norway
Morduch J. (1999) The microfinance promise, Journal of Development Literature, 37(4),
1569-1614
Navajas S. J, Conning J. and Claudio Gonzalez Vega R.L. (2003) Lending technologies,
competition and consolidation in the market for microfinance in Bolivia, Journal of
International Development, 15, 747-770
Okumu L.J (2007) The Microfinance industry in Uganda: Sustainability, Outreach and
Regulation, Un Published PhD Dissertation, University of Stellenbosch, South Africa.
Paxton J and Fruman C (1998) Saving-first vs. Credit First Microfinance institutions in eight
African Countries, Microfinance in Africa S.A. Breth ed. Sasakawa, Africa
Association, Mexico City
Pfister M. W., Gesesse D., Amha W., Mommartz R., Duflos W., Steel E. (2008) Access to
finance in Ethiopia: Sector Assessment study volume 2, GTZ
Rhyne E. and Otero M. (1992) Financial service for microenterprise principles and
Institutions, World Development, 20(11), 1561-1571
Rhyne E. and Otero M. (2006) The changing face of microfinance: closing the quality gap,
the Micro banking Bulletin Issue No 13:1-7 Microfinance Information Exchange Inc.
Washington DC
Robinson, M. (2001). The Microfinance Revolution: Sustainable Banking for the Poor,
Washington, DC: The World Bank.
Rosenberg R., Gonzalez A., Narain S. (2009) Are Microcredit Interest rate Excessive?
CGAP Brief Note, Washington D.C.
25
Schreiner, M. (1999) Aspects of outreach: a framework for the discussion of the social
benefit of Microfinance, Microfinance Risk Management, St. Louis
SEEP Network and Calmeadow (1995) Financial ratio Analysis of Microfinance Institution,
the SEEP Network Pact Publication, New York, January
Sharma M. and Zeller (1999) Outreach of group based Credit based credit organizations: The
case of ASA, BRAC and Proshika in Bangladesh, World Development, 27 (12), 2123-
2136
Stiglitz J. and A. Weiss (1981) Credit Rationing in markets with imperfect information,
American Economic Review, 71, 393-419
Woller G. (2002) The promise and perils of microfinance Commercialization, Small
Enterprise Development, 13 (4), 12-21
Woller G.M and M. Schreiner (2006) Poverty lending, financial self sufficiency and the six
aspects of outreach, SEEP Network Publication, New York
Yaron J (1992) Successful rural financial institutions, Discussion paper No 150, The World
Bank, Washington DC
Zeller M. (1994) Determinants of Credit rationing: A study of informal and formal credit
groups in Madagascar, World Development, 22 (12), 1895-1907
Zeller M. and Meyer R.(2002) The triangle of microfinance: Financial sustainability,
outreach and impact, Johns Hopkins University Press, Baltimore
Table 1: Summary Statistics
Mean Std. Dev. MIX 2003-2007 Africa One World One

Average sample t test sample t test
Variables (p values) (p values)
Africa World
Age 7 2 8 9.6 0.000 0.000
Assets 147958279 300550624
CapitaltoAssets 45 19 31.52 30.48 0.000 0.000
DeposittoAssets 10 8 11 0 0.2858 0.000
DebttoAssets 0 1
GLPtoAssets 70 12 67.32 78.72 0.0945 0.000
Borrowers 82796 138423 17029.4 12135.2 0.0001 0.000
Women 46 19 61.68 64.4 0.000 0.000
AvLnSz 996 433
ROA -3 5 -1.18 1.48 0.0031 0.000
OSS 125 46 108.8 117.2 0.0023 0.1237
FSS 88 31 97 108.8 0.0169 0.000
InterestRate 20 8 29.22 32.4 0.000 0.000
CapitalCosttoAssets 13 4
LaborCosttoAssets 5 2
BorrowerperLoanOfficer 446 300 271.4 224.2 0.000 0.000
PAR 5 7 4.38 2.42 0.2535 0.0003
26
Table 2: Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 1.000
2 .619* 1.000
*
3 -.409 -.367* 1.000
4 .080 .438* -.506* 1.000
5 .179 .090 -.316* .328* 1.000
6 .351* .273* -.319* .129 .075 1.000
7 .601* .983* -.395* .469* .118 .259* 1.000
8 -.354* -.423* .249* -.226 -.183 -.059 -.441* 1.000
9 .469* .589* -.201 .136 -.023 .625* .454* -.134 1.000
10 .149 .445* -.376* .454* -.060 .485* .452* -.098 .390* 1.000
* * * * *
11 .223 .534 -.210 .416 -.057 .468 .522 -.119 .479 .724* 1.000
12 .176 .456* -.419* .519* .031 .456* .471* -.095 .354* .950* .758* 1.000
13 -.285* -.524* .108 -.053 -.100 -.278* -.474* .297* -.501* .026 -.081 .098 1.000
14 -.288* -.698* .244* -.325* -.010 -.172 -.686* .262* -.387* -.579* -.422* -.508* .543* 1.000
15 -.429* -.846* .324* -.391* -.025 -.215 -.786* .415* -.646* -.403* -.473* -.370* .650* .717* 1.000
16 .301* .360* -.204 .124 -.071 .334* .360* -.130 .292* .302* .312* .252* -.349* -.451* -.502*
17 -.149 -.336* .335* -.324* .064 -.353* -.316* -.082 -.379* -.629* -.470* -.616* -.075 .378* .277*
*values significant at 5%
Key: 1 age 2 assets 3capital to assets 4 deposits to assets 5 debt to assets 6 GLP 7borrower 8
women 9avlnsz 10ROA 11OSS 12FSS 13interest rate 14 capital cost to assets 15labor cost to
assets 16 borrower per loan officer 17 PAR
27
Table 3: Ln FSS: Pooled OLS model with Robust Standard Errors
Model 1 Model 2 Model 3
Variables Coefficients P values Coefficients P values Coefficients P values
Ln Capital to Assets -0.02 0.507 0.004 0.895 0.02 0.619
Ln Deposits to Assets 0.12 0.009* 0.08 0.031* 0.11 0.008*
Ln Debt to Assets 0.03 0.698 0.01 0.890 0.09 0.371
Ln GLP to Assets Dropped Dropped 0.79 0.000* Dropped Dropped
Ln Borrower Dropped Dropped Dropped Dropped 0.05 0.137
Ln Average Loan Size 0.36 0.000* Dropped Dropped Dropped Dropped
Ln Interest rate 0.67 0.001* 0.66 0.000* 0.53 0.000*
Ln Capital costs to Assets -0.58 0.000* -0.48 0.009* -0.46 0.013*
Ln Labor costs to Assets 0.06 0.603 Dropped Dropped Dropped Dropped
Ln Borrowers per Loan Officer 0.06 0.222 -0.0008 0.984 0.05 0.339
Ln PAR -0.04 0.180 -0.04 0.237 -0.08 0.037*
Ln Age -0.24 0.128 -0.36 0.002* -0.42 0.005*
Ln Assets Dropped Dropped 0.06 0.138 Dropped Dropped
Y2004 0.001 0.982 0.05 0.302 0.09 0.195
Y2005 0.03 0.727 0.05 0.276 0.22 0.012*
Y2006 0.15 0.204 0.14 0.036* 0.36 0.002*
Y2007 0.06 0.693 0.09 0.371 0.31 0.019*
Constant 1.14 0.324 -0.43 0.761 3.3 0.003*
R2 0.78 0.000* 0.84 0.000* 0.73 0.000*
*Shows values sig at 5%
Table 4: Ln Borrowers Model: Fixed Effect Model with Robust Standard Errors
Model 1 Model 2
Variables Coefficients P values Coefficients P values
Ln Capital to Assets -0.05 0.789 -0.05 0.784
Ln Deposits to Assets 0.34 0.68 0.03 0.647
Ln Debt to Assets -0.11 0.293 -0.16 0.141
Ln GLP to Assets Dropped Dropped 0.35 0.213
Ln FSS 0.07 0.701 -0.2 0.316
Ln Average Loan Size -0.26 0.095 Dropped Dropped
Ln Interest rate -0.03 0.868 0.3 0.066
Ln Capital costs to Assets -0.05 0.743 -0.15 0.402
Ln Borrowers per Loan Officer 0.13 0.141 0.13 0.135
Ln PAR -0.06 0.188 -0.05 0.201
Ln Age 2.39 0.000* 2.15 0.001*
Y2004 -0.209 0.195 -0.2 0.186
Y2005 -0.27 0.316 -0.3 0.247
Y2006 -0.28 0.450 -0.2 0.415
Y2007 -0.32 0.473 -0.3 0.401
Constant 7.2 0.000* 4.9 0.001*
R2 0.45 0.000* 0.4 0.000*
28
Table 5: Ln Av Ln Sz : Fixed Effect Model with Robust Standard Errors

Ln Capital to Assets 0.004 0.968 0.05 0.657 0.03 0.746
Ln Deposits to Assets -0.03 0.449 -0.012 0.760 0.02 0.674
Ln Debt to Assets 0.15 0.344 0.11 0.312 -0.03 0.766
Ln Borrowers Dropped Dropped Dropped Dropped -0.25 0.094
Ln FSS 0.43 0.020* 0.5 0.006* 0.47 0.031*
Ln Interest rate -0.52 0.004* -0.64 0.000* -0.61 0.006*
Ln Capital costs to Assets 0.1 0.554 0.2 0.208 0.05 0.796
Ln Labor costs to Assets -0.4 0.002* Dropped Dropped Dropped Dropped
Ln Borrowers per Loan Officer -0.15 0.045* -0.12 0.097 -0.13 0.133
Ln PAR -0.02 0.478 -0.002 0.941 -0.06 0.162
Ln Age 0.37 0.377 -0.53 0.300 1.1 0.016*
Ln Assets Dropped Dropped 0.45 0.000* Dropped Dropped
Y2004 -0.002 0.980 0.089 0.409 -0.02 0.824
Y2005 0.05 0.793 0.183 0.318 0.01 0.936
Y2006 0.016 0.949 0.12 0.592 -0.04 0.869
Y2007 0.158 0.618 0.216 0.448 0.07 0.832
Constant 7.13 0.000* -0.5 0.742 7.5 0.000*
R2 0.59 0.000* 0.43 0.000* 0.33 0.000*
Table 6: Ln OSS: Fixed Effect Model with Robust Standard Errors

Ln Capital to Assets 0.17 0.177 0.17 0.14 0.20 0.148
Ln Deposits to Assets 0.13 0.010* 0.09 0.022* 0.12 0.008*
Ln Debt to Assets -0.04 0.739 -0.06 0.641 0.00 0.993
Ln Borrower Dropped Dropped Dropped Dropped 0.11 0.004*
Ln Interest rate 0.38 0.009* 0.33 0.000* 0.2 0.017*
Ln Capital costs to Assets -0.31 0.032* -0.19 0.259 -0.18 0.284
Ln Labor costs to Assets -0.14 0.201 Dropped Dropped Dropped Dropped
Ln Borrowers per Loan Officer 0.05 0.380 0.04 0.349 0.08 0.150
Ln PAR -0.02 0.619 -0.004 0.907 -0.04 0.242
Ln Age -0.13 0.313 -0.37 0.012* -0.42 0.006*
Ln Assets Dropped Dropped 0.12 0.004* Dropped Dropped
Y2004 0.09 0.242 0.15 0.025* 0.204 0.010*
Y2005 -0.06 0.534 -0.008 0.904 0.14 0.091
Y2006 0.19 0.094 0.22 0.022* 0.42 0.001*
Y2007 0.17 0.246 0.204 0.099 0.42 0.005*
Constant 1.94 0.227 -1.08 0.566 2.7 0.028*
R2 0.58 0.000* 0.66 0.000 0.56 0.000*
29
Table 7: Ln ROA model: Fixed Effect Model with Robust Standard Errors

Ln Capital to Assets 0.34 0.157 0.39 0.089 0.48 0.018*
Ln Deposits to Assets -0.02 0.742 -0.01 0.906 -0.017 0.847
Ln Debt to Assets 0.17 0.617 -0.07 0.758 0.26 0.317
Ln Borrower Dropped Dropped Dropped Dropped -0.37 0.270
Ln Interest rate 0.56 0.068 0.62 0.031* 0.15 0.635
Ln Capital costs to Assets -0.49 0.150 -0.65 0.028* -0.69 0.031*
Ln Labor costs to Assets 0.14 0.660 Dropped Dropped Dropped Dropped
Ln Borrowers per Loan Officer 0.21 0.197 -0.19 0.106 0.12 0.442
Ln PAR 0.06 0.356 0.005 0.937 -0.003 0.967
Ln Age -1.4 0.129 -1.04 0.256 -0.4 0.552
Ln Assets Dropped Dropped -0.015 0.956 Dropped Dropped
Y2004 0.44 0.137 0.48 0.058 0.49 0.113
Y2005 0.61 0.155 0.59 0.090 0.78 0.087
Y2006 0.85 0.112 0.77 0.079 1.07 0.048*
Y2007 0.85 0.178 0.95 0.075 1.2 0.040*
Constant -5.7 0.229 -3.05 0.549 4.8 0.035*
2
R 0.17 0.000* 0.26 0.000* 0.0057 0.000*
Table 8: Ln Women Regression: Pooled OLS Model with Robust Standard Errors

Ln Capital to Assets -0.27 0.340 -0.23 0.420 -0.22 0.402
Ln Deposits to Assets -0.17 0.483 -0.25 0.416 -0.25 0.417
Ln Debt to Assets -0.06 0.846 0.1 0.817 0.1 0.813
Ln Borrower Dropped Dropped Dropped Dropped 0.04 0.809
Ln FSS 0.03 0.920 -0.076 0.792 -0.07 0.787
Ln Interest rate -0.47 0.296 -0.06 0.834 -0.07 0.802
Ln Capital costs to Assets 0.05 0.866 0.25 0.518 0.25 0.529
Ln Labor costs to Assets 0.48 0.020* Dropped Dropped Dropped Dropped
Ln Borrowers per Loan Officer 0.06 0.611 0.034 0.815 0.03 0.821
Ln PAR -0.18 0.044 -0.174 0.038* -0.17 0.047*
Ln Age -0.79 0.079 -1.01 0.205 -1.03 0.220
Ln Assets Dropped Dropped 0.03 0.820 Dropped Dropped
Y2004 0.08 0.697 0.07 0.712 0.08 0.684
Y2005 0.1 0.695 0.119 0.627 0.13 0.595
Y2006 0.29 0.335 0.315 0.268 0.33 0.262
Y2007 -0.04 0.926 -0.05 0.904 -0.03 0.935
Constant 6.6 0.007* 6.05 0.038* 6.3 0.006*
2
R 0.25 0.03* 0.22 0.06* 0.22 0.04*
30
31

Determinants of Sustainability and Outreach of Ethiopian Microfinance Institutions

Uploaded by

Copyright:

Available Formats

Determinants of Sustainability and Outreach of Ethiopian Microfinance Institutions

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determinants of Sustainability and Outreach of Ethiopian Microfinance Institutions

Uploaded by

Copyright:

Available Formats

Determinants of Sustainability and Outreach of Ethiopian

Microfinance Institutions: The Case of Operational

Letenah Ejigu Wale$$

Key terms: Microfinance, Sustainability, Outreach, Operational Characteristics

JEL Code: G21

II. Review of Related Literature

2.2. Determinants of sustainability and outreach

Capital structure variables

Lending interest rate

Borrowers per field officer

2.3 Relationship between sustainability and outreach

III. Data and Methodology

The Econometric model

The basic two-way panel data model looks like as follows:

lnSit = α + βilnOCit + βjlnCit + µi + γ t + εit --------------------------(1)

1. Summary Statistics and Correlations

When we come to the other variables we got the following results:

Ethiopian MFIs have a significantly lower profitability rate as measured by ROA as

2. Panel Data Results

A brief explanation for the significant variables follows:

A discussion on the insignificant variables follows:

Average loan size regression

A brief explanation for the significant variables follows:

VI. Concluding Remarks

Breadth of outreach regressions

Depth of outreach regression

 The MFIs are pro-poor but not financially sustainable.

 They are not pro-poor ( i.e..commercially oriented) but financially sustainable

Limitations of the study

Amha W. (2007) Managing growth of microfinance institutions: Balancing Sustainability and

Table 1: Summary Statistics

Mean Std. Dev. MIX 2003-2007 Africa One World One

Model 1 Model 2 Model 3

Model 1 Model 2 Model 3

Model 1 Model 2 Model 3

Model 1 Model 2 Model 3

You might also like