+Model

RDF-76; No. of Pages 9 ARTICLE IN PRESS

HOSTE D BY Available online at [Link]

ScienceDirect
Review of Development Finance xxx (2016) xxx–xxx

Financial development and poverty reduction in developing countries: New

evidence from banks and microfinance institutions
Ficawoyi Donou-Adonsou a,∗ , Kevin Sylwester b
a Department of Economics and Finance, John Carroll University, 1 John Carroll Boulevard, University Heights, OH 44118, United States
b Southern Illinois University, Carbondale, United States

Abstract
The literature on financial development and growth has received a lot of attention over the past two decades. Unlike growth, not much of
consideration has been given to poverty reduction. Moreover, most of the past studies focus on bank and stock market development. The advent of
microfinance institutions (MFIs) lets to think about the potential role MFIs can play in a countrywide economy. In this study, we consider to what
extent banks and MFIs reduce poverty. We apply the instrumental variables approach, namely the fixed-effects two-stage least squares, to a panel
of 71 developing countries over the period 2002–2011. Using credit to GDP as the main financial development indicator, the results indicate that
banks reduce poverty when poverty is measured by the headcount ratio and poverty gap. As for the squared poverty gap, there is no significant
effect of banks. On the other hand, MFIs do not appear to have any impact on poverty regardless of the measure of poverty employed. These results
imply that while banks have some ability to reduce poverty, MFIs do not, at least at the aggregate level. Our results are robust to the use of assets
to GDP as an alternative measure of financial development.
© 2016 Africagrowth Institute. Production and hosting by Elsevier B.V. All rights reserved.

JEL classification: O11; G21

Keywords: Banks; Microfinance; Poverty; Developing countries

1. Introduction wide measures of financial development, such as money and

quasi money, market capitalization, or private credit in their
Numerous studies have considered interactions between empirical work. Such indicators fail to capture how different
financial development and economic growth, including the institutions within the financial sector influence poverty.
direction of causality between the two. Some research has also This omission could be especially important when examining
considered poverty and financial development. These studies microfinance institutions (MFIs) since they were promoted for
include Honohan (2004), Beck et al. (2007), Odhiambo (2010), the specific purpose of providing financial services to the poor,
and Jeanneney and Kpodar (2011). Studies such as Jalilian especially credit, in order to alleviate poverty. Microfinance first
and Kirkpatrick (2002) consider a “trickle-down” approach. gained prominence during the 1970s with organizations such
Building upon the framework of Beck et al. (2000) and Dollar as the Grameen Bank in Bangladesh and the work of advo-
and Kraay (2002), Jalilian and Kirkpatrick (2002) first consider cates like Mohammad Yunus. The scope of MFIs has since
how financial development affects economic growth and then grown manifold. From 2002 to 2011, the gross loan portfolio
examine to what extent growth reduces poverty. A common of MFIs in developing countries increased by more than 1700%
element of these studies is that they have considered economy and its number of active borrowers increased by 400%.1 MFIs
have been most prevalent in South Asia, especially compared to
other poor regions such as Africa and Latin America although
∗
they have grown in these regions as well. They also often serve
Corresponding author. Tel.: +1 216 397 4970; fax: +1 216 397 1728.
E-mail addresses: fdonouadonsou@[Link] (F. Donou-Adonsou),
ksylwest@[Link] (K. Sylwester).
Peer review under responsibility of Africagrowth Institute. 1 MIX Market data from [Link].

[Link]
1879-9337/© 2016 Africagrowth Institute. Production and hosting by Elsevier B.V. All rights reserved.

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
2 F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx

rural communities where banks often cannot be found. Although 2. Literature review
MFIs also make loans from their deposits as do banks, their
financing can also come from investor borrowing, from equity, Greenwood and Jovanovic (1990) develop a model where
and from grants. Since they often serve poor communities, loans financial intermediaries analyze imperfect information and
to individuals are generally much smaller than typical bank loans channel funds from savers to borrowers. Their model includes a
and often not collateralized. Therefore, given that banks and participation cost, a lump-sum fee that agents must pay to partic-
MFIs serve different clienteles, make different types of loans, ipate in the financial sector. This fee effectively keeps the poor
and are financed from different sources, they capture distinct from taking advantage of opportunities in the financial sector.
aspects of financial development and so could have different Not only would the poor not benefit, but the income distribution
impacts upon poverty alleviation. could even widen between low and high income agents. This
The objective of this paper is to compare traditional banks implication is endorsed by Stiglitz (1993) for whom financial
to microfinance institutions as to what extent each contributes market failure is the fundamental cause of poverty in develop-
to poverty reduction. In this sense, we coincide with Jeanneney ing countries. Applying this model to our case, the participation
and Kpodar (2011) in that we characterize financial development fee would likely be lower for MFIs and so they would be bet-
as access to financial services in the “banking system” com- ter able to obtain credit, invest, and escape poverty. Of course,
posed of traditional banks and MFIs. Studies such as Khandker whether it is sufficiently low so as to benefit the poor is another
(2005) and Mahjabeen (2008) have considered to what extent question.
microfinance has lowered poverty at the local level. More Such considerations have not been examined at a macro-
recent theoretical work (Ahlin and Jiang, 2008; Yusupov, 2012; economic level where, as stated, economy wide measures of
Buera et al., 2012) has suggested the potential for macroecono- financial development are used. Jalilian and Kirkpatrick (2002),
mic effects of microfinance. For example, Donou-Adonsou and Beck et al. (2008), and Jeanneney and Kpodar (2011) have
Sylwester (2015) find that microfinance loan growth increases used the trickle-down approach – an indirect effect of financial
economic growth and total factor productivity in developing development on poverty reduction through economic growth –
countries. In this paper, we take the potential for MFIs to have to investigate financial development and poverty reduction in
macroeconomic impacts seriously, especially given the rapid developing countries and find that financial development fosters
growth in MFIs, and examine to what extent this rise has been growth which then reduces poverty. For instance, Jalilian and
able to reduce poverty at a national level. Comparing this effect Kirkpatrick (2002) argue that by widening financial services
to that from traditional banks helps to place any impact in better access to the poor, their income will grow, which eventually
context as well as to see if MFIs do, indeed, play an enhanced will reduce poverty. For example, an insurance service pro-
role in reducing poverty. vided to the poor can better protect them against income shocks.
An example of such a comparison comes from Thanvi (2010) Other studies have investigated the direct relationship between
for the Cooch Behar District of West Bengal, India. Thanvi financial development and poverty reduction or the income dis-
describes a shift from banks to MFIs due to the unavailability tribution. These studies include Honohan (2004), Jalilian and
(in part or total) of loans from banks. For Thanvi, MFIs supple- Kirkpatrick (2005), Beck et al. (2007), Perez-Moreno (2011),
ment the role of banks by reaching the unreached. In this way, Jeanneney and Kpodar (2011), and Sehrawat and Giri (2015)
one might infer that MFIs should be able to reduce poverty to a although they differ both in terms of what proxies for finan-
larger extent than banks. However, Thanvi also documents that cial development they use as well as in their outcome variable
MFIs charge higher interest rates. One MFI, Bardhan, charges (headcount ratio, poverty gap, Gini coefficient, etc.).
an effective rate of 24%, twice that charged by banks. These Given the purported role of MFIs in assisting lower income
higher rates raise questions regarding how effective MFIs are at households, various studies have focused upon these institutions
reducing poverty. and examined to what extent they can help raise living standards
This study employs an instrumental variables approach to a among the poor. Several studies have found beneficial effects
panel of 71 developing countries over the period 2002–2011. upon consumption or income (Khandker, 2005; Kondo et al.,
The results indicate that banks reduce poverty when poverty 2008; Berhane, 2009; Collins et al., 2009; Imai and Azam, 2011;
is measured by the headcount ratio or the poverty gap but Berhane and Gardebroek, 2011), housing conditions (Berhane,
not when poverty is measured by the squared poverty gap. 2009; Berhane and Gardebroek, 2011), village-level wages and
On the other hand, MFIs do not appear to have any impact investment in agriculture (Kaboski and Towsend, 2012), savings
on poverty regardless of the measure employed. While the (Kondo et al., 2008; Dupas and Robinson, 2009), and health
results suggest that banks play a role in reducing poverty, and food security (Stewart et al., 2010). Other studies remain
MFIs do not appear to have done so, at least at the aggregate skeptical. For instance, Chowdhury (2009) casts doubt on the
level. effectiveness of microfinance as a poverty alleviation tool given
The paper is structured as follows: Section 2 provides a more the profit-seeking nature of financial institutions. He argues that
detailed description of the literature. Section 3 describes the data microfinance, though it provides a safety net and can help smooth
and outlines the methodology. Section 4 presents and explains consumption, needs its borrowers to have business skills and
the results. In Section 5, we provide a robustness check using marketing information for loans to expand businesses and create
an alternative measure of financial development, and Section 6 jobs. Likewise, Copestake and Williams (2011) argue that MFIs
offers concluding discussion. by themselves cannot bring sustainable growth and reduction in

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx 3

poverty as obstacles such as selection bias weaken the positive lower leverage abilities. Although we focus upon credit, we
effects of MFIs on households’ welfare. will also consider another financial measure, namely the ratio of
These empirical studies examine the effects of MFIs at the assets to GDP, as a robustness check. Assets to GDP, contrary to
local level, but to what extent might MFIs influence poverty private credit to GDP, include credit to the public sector as well.
at the national level? Even if the effects of any one MFI All two financial development variables have been extensively
are only felt locally, countrywide effects could still arise if used in the literature.2 For each of these variables, we consider
MFIs were located in many communities as in many develop- one measure for banks and another measure for MFIs. Compar-
ing countries. Another possibility for macroeconomic effects ing results between the two measures can provide better context
could occur through spillovers, especially if increasing con- for what extent MFIs influence poverty compared to traditional
sumption or investment spurs additional job creation. Buera banks. Given that banks’ data are deflated, we also deflate the
et al. (2012) build an economy-wide model of entrepreneur- microfinance data as described in Beck et al. (2013).3
ship wherein MFIs service the poor who cannot borrow in the A weakness of our approach is that loans from what one gen-
“traditional” financial sector. They then explore to what extent erally labels as microcredit – and, hence, – loans from MFIs – are
MFIs influence output, capital, total factor productivity, wages, combined with loans from traditional banks as MIX relies upon
and interest rates, finding in some cases that MFIs not only raise the characteristics of the borrower to denote a loan as “micro-
output but decrease disparities between rich and poor. Ahlin and credit” and not the lender. This shortcoming is not fatal for our
Jiang (2008) and Yusupov (2012) also find that MFIs promote purposes because we want to examine to what extent finance
development on a wider scale. directed toward the poor (regardless from what source) lowers
poverty.
3. Data and methodology As for variables capturing the extent of poverty, we consider
the poverty headcount ratio, the poverty gap, and the squared
3.1. Data poverty gap. The first two variables come from Poverty and
Equity Database published by the World Bank and the last
Banks are defined as financial institutions that accept deposits variable comes from PovcalNet, also published by the World
and make loans. In this study, we consider deposit money banks Bank. The World Development Indicators provides data for the
commonly studied in the finance-growth literature. These banks Gini coefficient and real GDP per capita ($PPP). In Table 3, we
accept demand deposits, saving deposits, and time deposits provide more details on all these variables.
(Beck et al., 2013), and are composed of commercial banks The data covers 71 developing countries from 2002 to 2011.
and other financial institutions such as thrift institutions and The initial year is chosen due to data availability. The summary
credit unions. Microfinance, on the other hand, is defined as statistics are provided in Table 1, the correlation coefficients in
institutions that primarily provide financial services to the poor. Table 2, and the countries in the sample are listed in appendix.
They may be compared to credit unions in terms of structure An important feature of these correlation coefficients is that
and activities. However, microfinance institutions differ from the poverty measures are weakly negatively correlated with the
banks in that they receive most of their funding from exter- financial development variables. It is also important to point out
nal loans, grants, or investors. They also differ from banks that the magnitudes of the correlations are higher for the meas-
in that they mostly make small loans called “microcredit” ures of bank activity. One last feature of the correlation table
to the poor. Referring to Microfinance Information Exchange is the weak negative correlation between banks and MFIs with
(MIX), a microfinance institution “can be a nonprofit organi- respect to their respective credit and assets.
zation, regulated financial institution or commercial bank that
provides microfinance products and services to low-income
clients.” Microfinance data comes from the MFI Profiles and 3.2. The empirical model and methodology
Reports from MIX. MIX recognizes six legal statuses for
microfinance institutions: banks, credits/cooperatives, non-bank We use the growth-poverty model suggested by Ravallion
financial institutions, non-governmental organizations, rural (1997) and Ravallion and Chen (1997). Adams and Page (2005)
banks, and others. Data regarding traditional banks comes from have used this model to investigate the impact of international
Beck et al. (2013). migration and remittances on poverty in developing countries.
Our main measure capturing the roles of banks and MFIs We adopt a similar approach that controls for income and its
are their respective private credit as a percentage of GDP since distribution to investigate the impact of financial development
loans represent the key financial service offered by most insti- on poverty. The empirical model is given by:
tutions, especially in developing countries. Thus, higher levels
of credit imply higher levels of financial services, and therefore log povit = αi + β1 log μit + β2 log git + β3 log xit + εit (1)
higher levels of financial intermediation. Looking at Table 1,
bank credit averages 27.34% of GDP whereas that for micro-
finance credit is 1.12%. The reason for this large difference 2 See for instance Levine et al. (2000) and Beck et al. (2013).
lies in the difference between the two institutions’ customers. 3 Beck et al. (2013) provide the following deflation method:
Banks generally fund larger enterprises with much higher lever- {(0.5)*[Ft /Pet + Ft−1 /Pet−1 ]}/[GDPt /Pat ], where F is credit or assets; Pe
age abilities, while MFIs fund smaller enterprises with much is end-of-period CPI, and Pa is average annual CPI.

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
4 F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx

Table 1
Descriptive statistics.
Variable Obs. Mean Std. dev. Min Max

Poverty headcount at $1.25 a day PPP (%) 324 14.30 18.59 0.00 87.72
Poverty gap at $1.25 a day PPP (%) 324 5.22 8.26 0.00 52.76
Squared poverty gap at $1.25 a day PPP (%) 314 3.67 6.97 0.00 52.76
Per capita GDP ($PPP-constant 2011) 700 6088.77 4794.86 492.61 22,569.81
Gini index 325 42.05 9.71 16.23 67.40
Credit bank (%GDP) 710 27.34 22.51 0.55 121.49
Credit MFI (%GDP) 710 1.12 1.87 0.00 13.70
Asset bank (%GDP) 687 35.24 25.98 0.63 131.49
Asset MFI (%GDP) 709 2.09 4.22 0.00 48.58
Rule of law 620 3.14 1.08 1.00 6.00
Ethnic tensions 620 3.72 1.24 0.00 6.00

Table 2
Correlation coefficients.
Credit bank Asset bank Credit MFI Asset MFI Headcount Pov. gap Sq. pov. gap Gini P.c. GDP Rule of Ethnic
index law tensions

Credit bank 1
Asset bank 0.93 1.00
Credit MFI −0.03 −0.12 1.00
Asset MFI −0.04 −0.11 0.97 1.00
Headcount −0.10 −0.12 −0.04 −0.08 1.00
Poverty gap −0.10 −0.12 −0.02 −0.05 0.96 1.00
Squared poverty gap −0.23 −0.24 −0.06 −0.06 0.86 0.93 1.00
Gini index 0.08 0.11 0.02 −0.05 0.39 0.43 0.19 1.00
Per capita GDP 0.11 0.19 −0.33 −0.28 −0.52 −0.44 −0.47 −0.15 1.00
Rule of law 0.05 0.02 −0.04 −0.02 −0.14 −0.14 −0.16 −0.67 0.18 1.00
Ethnic tensions 0.09 −0.02 0.00 0.01 0.26 0.23 −0.05 0.08 −0.19 −0.26 1

Table 3
Variable description and source.
Variable Description Source

Poverty headcount Poverty headcount ratio at $1.25 a day 2005 $PPP. It is the percentage of the population living on less Poverty and equity database
than $1.25 a day at 2005 international prices
Poverty gap Poverty gap at $1.25 a day 2005 $PPP. It is the mean shortfall from the poverty line expressed as a Poverty and equity database
percentage of the poverty line
Squared poverty gap Squared poverty gap at $1.25 a day 2005 $PPP, defined as a % of poverty line. It is an indicator of PovcalNet
poverty severity.
Per capita GDP PPP (constant 2011 international $) World development indicators
Gini index Measures income inequality. An index of 0 represents perfect equality, while an index of 100 implies World development indicators
perfect inequality
Credit bank Deposit money banks credit, defined as a % of GDP Beck et al. (2013)
Credit MFI Measured by the gross loan portfolio, defined as a % of GDP Market information exchange
Asset bank Deposit money banks assets, defined as a % of GDP Beck et al. (2013)
Asset MFI Defined as a % of GDP Market information exchange
Rule of law Measures law and order tradition of the country. It ranges from 0 (weak tradition) to 6 (strong International country risk guide
tradition)
Ethnic tensions Measures the degree of tension within a country attributable to racial, nationality, or language International country risk guide
divisions. It ranges from 0 (high tensions) to 6 (minimal tensions)

Note: We use the deflation method proposed by Beck et al. (2013) to deflate MFI credit and assets. The deflation formula is given by
{(0.5)*[Ft /Pet + Ft−1 /Pet−1 ]}/[GDPt /Pat ], where F is credit or assets; Pe is end-of-period CPI, and Pa is average annual CPI.

where pov is the measure of poverty in country i at time t; measured by bank development (bank credit) and microfinance
μ represents the mean per capita income measured by per development (MFI credit); αi denotes country fixed-effects;
capita GDP ($PPP); g is income inequality measured by the and ε is the error term. In Eq. (1), the coefficients (βi ) are
Gini coefficient; x denotes a financial development indicator as elasticities.

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx 5

Table 4
OLS estimates of the effects of financial development (measured by private credit/GDP) on poverty.
Poverty headcount Poverty gap Squared poverty gap

1 2 3 1 2 3 1 2 3

Per capita GDP −1.732*** −2.396*** −1.707*** −2.444*** −2.907*** −2.449*** −2.006*** −1.753** −1.992**
(−4.17) (−4.32) (−3.46) (−6.45) (−5.70) (−5.49) (−3.77) (−2.17) (−2.63)
Gini index 4.593*** 4.516*** 4.581*** 4.245*** 4.098*** 4.247*** 3.374*** 3.444*** 3.366***
(4.24) (3.88) (4.18) (4.87) (4.63) (4.84) (3.60) (3.63) (3.60)
Credit bank −0.531** – −0.529** −0.326* – −0.326* 0.151 – 0.154
(−2.61) (−2.57) (−1.67) (−1.66) (0.61) (0.64)
Credit MFI – −0.026 −0.006 – −0.009 0.001 – 0.010 −0.005
(−0.45) (−0.12) (−0.18) (0.03) (0.09) (−0.04)
Constant 1.157 5.498 0.974 6.787 10.308* 6.827 4.277 2.301 4.170
(0.19) (0.8) (0.16) (1.42) (1.75) (1.32) (0.72) (0.28) (0.56)
Within R2 0.49 0.46 0.49 0.53 0.51 0.53 0.22 0.22 0.22
# of countries 68 68 68 68 68 68 66 66 66
# of obs. 309 309 309 299 299 299 270 270 270

Note: All variables are expressed in logs. The estimation is based on the fixed-effects method for which we report the within R-squared. t-statistics in parentheses
are based on standard errors that are robust to heteroskedasticity. ***, **, and * denote significance at 1%, 5%, and 10%, respectively. The number of observations
is reduced in the table because of missing values for poverty variables (see summary statistics for details).

In Eq. (1), and consistent with Ravallion (1997), per capita racial, nationality, or language divisions. Lower ratings imply
income or economic growth is expected to lower poverty while high tensions, while higher ratings are given to countries where
income inequality is expected to have a positive effect on tensions are minimal. Higher values for the rule of law vari-
poverty. As for the financial development indicator, which is able denote a greater adherence to law and order. In addition to
the addition to this model, its relationship with poverty is not these two instruments, we also include the first and second lags
clear-cut in the literature. Nevertheless, following Jalilian and of financial development indicator as instruments. Singh et al.
Kirkpatrick (2002), Beck et al. (2008), and Jeanneney and (2011) use the first and second lags of money and quasi money
Kpodar (2011), who find that financial development fosters measured as (M2)/GDP and domestic credit/GDP, respectively,
growth which then reduces poverty, we expect financial indi- when analyzing the determinants and macroeconomic effects of
cators to reduce poverty. remittances.
To mitigate endogeneity concerns, one sometimes takes first To gauge the validity of the aforementioned instruments,
differences as in Ravallion and Chen (1997) since differences we run the Sargan-Hansen test of overidentifying restrictions.
are less persistent over time than are levels. However, we refrain Under the null hypothesis, the instruments are valid. We will also
from doing so as needing two observations for poverty or the include these six instruments in the regression model and esti-
Gini coefficient would greatly reduce the sample size. Despite mate by OLS. If the instruments are valid, then their coefficient
this concern, we first estimate equation (1) using the fixed-effects estimates should be zero.
technique (OLS), which assumes that our independent variables
are exogenous. To then address endogeneity concerns, we also 4. Results
use instruments and estimate via two-stage least squares (2SLS).
The first instrument we use is ethnic tensions.4 Beck et al. (2003) We begin with the fixed-effects results without using instru-
have used this variable and find significant negative correlation ments. The results are reported in Table 4. Column (1) reports
between ethnic fractionalization and private credit. This corre- the results with bank credit only, column (2) with MFI credit
lation could be explained by the fact that greater ethnic diversity only, and column (3) with both variables included in the same
implies the adoption of policies and institutions geared toward model. Controlling for per capita income and the Gini index
power and control and not toward creating an open and compet- that are negatively significant as expected, the results indicate
itive financial system. The second instrument we use is the rule bank credit reduces poverty as far as the poverty headcount and
of law used by Levine et al. (2000). Banks sign a lot of contracts, poverty gap are concerned. As for the squared poverty gap, we
and a country having a tradition of establishing law and order is do not see any significant effect of bank credit. MFI credit, on
likely to boost its financial sector development. Both ethnic ten- the other hand, does not appear to have significant effects on
sions and rule of law range from zero to six points according to poverty reduction regardless of what poverty measure is used.
the International Country Risk Guide database. Ethnic tensions However, given the potential for financial development to be
measure the degree of tension within a country attributable to endogenous, these OLS estimates may be biased.
Table 5 reports the fixed-effects 2SLS results when using rule
of law, ethnic tensions, and the first two lags of both bank credit
4 Ethnic tensions differs from ethnic heterogeneity in that the former takes into and MFI credit to instrument for bank credit and MFI credit.
account actual events and disturbances within the country arising from ethnic The results in Table 5 indicate that bank credit reduces poverty
heterogeneity. as measured by the headcount or the poverty gap. The use of

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
6 F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx

Table 5
2SLS estimates of the effects of financial development (measured by private credit/GDP) on poverty.
Poverty headcount Poverty gap Squared poverty gap

1 2 3 1 2 3 1 2 3

Per capita GDP −2.052*** −2.842*** −2.265*** −2.327*** −2.971*** −2.547*** −0.987 −1.410** −1.155*
(−4.13) (−5.03) (−3.94) (−5.07) (−5.68) (−4.81) (−1.60) (−2.10) (−1.68)
Gini index 3.398*** 3.809*** 3.472*** 4.112*** 4.508*** 4.193*** 3.74*** 3.966*** 3.860***
(4.04) (4.41) (4.07) (5.19) (5.56) (5.22) (4.18) (4.30) (4.16)
Credit bank(1) −0.648*** – −0.685*** −0.552*** – −0.585*** −0.256 – −0.312
(−3.30) (−3.31) (−2.88) (−2.91) (−1.08) (−1.13)
Credit MFI(2) – −0.003 0.053 – 0.007 0.054 – 0.019 0.075
(−0.05) (0.73) (0.10) (0.80) (0.14) (0.51)
Constant 8.775 12.038* 10.559* 7.014 9.360 8.806 −4.648 −2.558 −3.341
(1.51) (1.87) (1.67) (1.29) (1.56) (1.49) (−0.68) (−0.35) (−0.46)
Within R2 0.42 0.39 0.41 0.50 0.47 0.50 0.26 0.26 0.25
# of countries 58 58 58 58 58 58 57 57 57
# of obs. 214 214 214 206 206 206 185 185 185
Sargan–Hansen p-value 0.178 0.703 0.271 0.061 0.721 0.177 0.386 0.623 0.503

Note: All variables are expressed in logs. The estimation is based on the fixed-effects 2SLS method for which we report the within R-squared. t-statistics in parentheses
are based on standard errors that are robust to heteroskedasticity. ***, **, and * denote significance at 1%, 5%, and 10%, respectively. The number of observations
is reduced in the table because of missing values for poverty variables (see summary statistics for details). The sample is further reduced because ICRG does not
report data for some countries like Benin, Cambodia, Nepal, etc. (1) Credit bank is instrumented using Rule of law, Ethnic tensions, 1st and 2nd lags of credit bank,
Per capita GDP, and Gini index as instruments. (2) Credit MFI is instrumented using Rule of law, Ethnic tensions, 1st and 2nd lags of credit MFI, Per capita GDP,
and Gini index as instruments.

natural logarithms for both poverty and bank credit implies that with the squared poverty gap. In theory, this instrument should
the coefficient on bank credit can be interpreted as an elasticity. be dropped. However, we keep it in the instrument list when
For the headcount index, a 10% increase in bank credit reduces considering the squared poverty gap because the results – not
poverty by about 6.5–6.9%, whereas the same 10% increase in reported here but available upon request – do not qualitatively
bank credit reduces poverty by 5.5–5.9% as measured by the change when excluding the first lag of bank credit from the
poverty gap. For the former, this means that a 10% increase in instrument list.
bank credit reduces the fraction of the population living on less To sum up, our results indicate that bank credit reduces
than $1.25 a day by about 6.7%. For the latter, the difference in poverty when poverty is measured by the headcount ratio
income between the $1.25 per person threshold and the actual and poverty gap. These results support those from Honohan
income of the poor diminishes by roughly 5.7%. (2004), Jalilian and Kirkpatrick (2005), Beck et al. (2007),
Compared to the OLS estimates, the IV elasticities are sta- Jeanneney and Kpodar (2011), and Sehrawat and Giri (2015),
tistically stronger (1% significance versus 5% and 10% for the who also find that financial development lowers poverty. How-
OLS estimates) and larger in magnitude. Just like the OLS esti- ever, these conclusions are tempered in that we find no significant
mates, the coefficients from the IV estimation are not significant effect when using the squared poverty gap to measure poverty.
when considering the squared poverty gap although the sign MFI credit, on the other hand, does not appear to have any
goes from positive to negative. As for MFI credit, the IV results impact on poverty regardless of the measure we consider,
confirm those from Table 4 in that MFI credit does not have any suggesting that any effect upon poverty reduction is at most
significant effect on poverty reduction. Per capita income and small. We are not the first to be skeptical of the ability of
Gini coefficient have the expected signs as in Ravallion (1997) MFIs to lower poverty as Chowdhury (2009) raises similar
and Adams and Page (2005). More importantly, the magnitudes doubts.
of their elasticities are consistent with those in Ravallion (1997) From our results, one can thus wonder how banks show
and Adams and Page (2005). Only for the squared poverty gap poverty reduction effects (at least as shown by the headcount
is income per capita not statistically significant when the model and poverty gap measures) whereas MFIs do not, especially
includes bank credit only as its measure of financial develop- since MFIs focus upon helping the poor. One possible way to
ment. explain the effect of bank credit on poverty is through invest-
On the validity of the instruments, the overidentification tests ments in infrastructure. In fact, many developing countries still
generally indicate that the instruments are valid. We observe, have a lot of room for improvements in infrastructure and such
however, a rejection at the 10% significance level when regress- projects could create various spillovers. A construction com-
ing the poverty gap on the measure of bank credit by itself. As pany, for instance, may seek funding from banks to build roads.
this may cast some doubt on how valid our instruments are, we This road construction will require the hiring of many people,
take a step further and regress the three poverty measures on the including the poor, which will lead to a decrease in poverty.
proposed instruments. The results reported in Table 6 indicate There could be also an indirect effect on poverty reduction as
that only the first lag of bank credit is significantly associated highlighted by Gachassin et al. (2010) in that access to roads will

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx 7

Table 6
OLS estimates of the effects of instruments on poverty.
Poverty headcount Poverty gap Squared poverty gap

1 2 3 1 2 3 1 2 3

Per capita GDP −2.194*** −2.691*** −2.360*** −2.431*** −2.961*** −2.757*** −0.965 −1.526** −1.168
(−4.97) (−4.90) (−4.61) (−4.53) (−4.49) (−4.19) (−1.32) (−2.08) (−1.44)
Gini index 3.490*** 3.636*** 3.516*** 4.149*** 4.467*** 4.298*** 3.699*** 3.856*** 3.774***
(2.74) (2.98) (2.65) (3.31) (3.55) (3.23) (3.40) (3.44) (3.42)
Rule of law 0.323 0.289 0.328 0.264 0.187 0.253 0.429 0.442 0.402
(1.37) (1.12) (1.42) (0.82) (0.68) (0.87) (1.17) (1.23) (1.04)
Ethnic tensions 0.075 0.162 0.106 0.224 0.334 0.250 0.292 0.214 0.293
(0.38) (0.62) (0.52) (0.99) (1.16) (1.06) (0.90) (0.60) (0.90)
1st lag credit bank −0.119 – −0.139 0.059 – 0.056 −0.887* – −0.923**
(−0.31) (−0.37) (0.16) (0.15) (−1.94) (−2.09)
2nd lag credit bank −0.608 – −0.613 −0.739 – −0.791 0.758 – 0.735
(−1.21) (−1.18) (−1.35) (−1.42) (1.30) (1.27)
1st lag credit MFI – 0.035 0.054 – 0.014 0.032 – −0.035 0.019
(0.62) (1.36) (0.21) (0.71) (−0.34) (0.17)
2nd lag credit MFI – −0.059 −0.019 – −0.001 0.032 – 0.053 0.038
(−1.07) (−0.40) (−0.03) (0.74) (0.79) (0.55)
Constant 9.420 10.807 10.862 7.561 8.821 10.099 −5.859 −1.849 −4.058
(1.37) (1.74) (1.55) (0.99) (0.248) (1.23) (−0.74) (−0.22) (−0.46)
Within R2 0.45 0.39 0.45 0.53 0.48 0.53 0.28 0.27 0.29
# of countries 58 58 58 58 58 58 57 57 57
# of obs. 214 214 214 206 206 206 185 185 185

Note: All variables are expressed in logs. The estimation is based on the fixed-effects method for which we report the within R-squared. t-statistics in parentheses
are based on standard errors that are robust to heteroskedasticity. ***, **, and * denote significance at 1%, 5%, and 10%, respectively. The number of observations
is reduced in the table because of missing values (see summary statistics for details).

facilitate market access by reducing transportation costs. More- 5. Robustness check

over, roads can shorten distances to feeder roads and increase
labor productivity in agriculture as shown in Uganda by Fan In this section, we check how robust our results are by using an
and Zhang (2008). It is obvious that large companies, and to a alternative measure of financial development, namely assets as a
lesser extent medium companies, will not go to microfinance percentage of GDP. This variable is also popular among financial
institutions to seek funding given the size of MFIs’ portfolio. development indicators in the literature (see for instance, Beck
A second possible explanation as to why traditional banks et al., 2013). As mentioned earlier, the assets to GDP variable
could be more successful in eliminating poverty is that some has the advantage to include credit to the public sector, con-
banks facing competition from MFIs have also expanded lend- trary to credit to GDP, which only includes credit to the private
ing to the poor. In fact, some banks offer microcredit services, sector by banks and other financial intermediaries. For conve-
and some small scale businesses may find bank loans cheaper as nience, we skip the OLS estimates to focus on the IV estimates.
MFIs are believed to charge on average higher interest rates. Table 7 reports the fixed-effects 2SLS estimates when financial
Thanvi (2010) finds that an MFI in India charges twice the development is measured by the assets to GDP ratio.
effective rate charged by banks. Finally, from the supply side, The results are qualitatively similar to those of credit in
Holden and Prokopenko (2001) argue that the quality of the Table 5. However, the magnitudes of the elasticities are higher.
loan portfolio of MFIs is sometimes poor because of inadequate An increase in bank assets by 10% reduces the proportion of
management and deficiencies in control of their activities. As a people living on less than $1.25 a day by about 8% (7.8–8.4%).
consequence, it is difficult for MFIs to reach efficiency levels Likewise, the same increase of 10% in bank assets induces about
to cover their costs. As for the demand side, Chowdhury (2009) an 8% (7.6–8.4%) decline in the amount of income separating
points out that borrowers may be lacking business skills and those in poverty from the $1.25 threshold. As before, estimates
marketing information for loans to expand businesses and cre- are not significant when using the squared poverty gap. Turning
ate jobs. Borrowers possessing these skills might then be more to microfinance, MFI assets do not appear to have any impact
successful at obtaining financing from traditional banks. on any of the three measures of poverty. As before, income per
A final reason is the much greater size of banks. As shown capita and the Gini coefficient have the expected signs and are
in Table 1, bank credit and assets are many times larger than statistically significant. Also, it is important to highlight that
that of microfinance institutions, providing much greater poten- in all regressions, we fail to reject the null hypothesis of over-
tial for their changes in lending to affect the poor. Moreover, identifying restrictions at conventional levels of significance.
lending from MFIs is almost always “small”, implying that any All together, these results do not systematically differ from the
effects could be contained locally and so not impact poverty at credit results in the previous section, although the elasticities of
the country level. poverty with respect to bank assets appear to be a bit larger.

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
8 F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx

Table 7
2SLS estimates of the effects of financial development (measured by assets/GDP) on poverty.
Poverty headcount Poverty gap Squared poverty gap

1 2 3 1 2 3 1 2 3

Per capita GDP −2.211*** −2.699*** −2.515*** −2.437*** −2.921*** −2.826*** −1.312** −1.515* −1.653*
(−4.36) (−4.26) (−3.77) (−3.27) (−5.01) (−4.68) (−2.08) (−1.86) (−1.87)
Gini index 3.111*** 3.758*** 3.161*** 3.719*** 4.487*** 3.798*** 3.33*** 4.009*** 3.421***
(3.54) (4.33) (3.55) (4.53) (5.49) (4.56) (3.56) (4.28) (3.57)
Assets bank(1) −0.778*** – −0.839*** −0.762*** – −0.835*** −0.295 – −0.383
(−3.30) (−3.30) (−3.27) (−3.32) (−1.08) (−1.13)
Assets MFI(2) – −0.026 0.055 – −0.003 0.070 – 0.035 0.086
(−0.35) (0.71) (−0.05) (0.99) (0.27) (0.53)
Constant 11.839* 10.968 14.568** 10.290* 8.989 13.695** −0.095 −1.799 2.908
(1.95) (1.62) (2.01) (1.82) (1.43) (2.05) (−0.01) (−0.23) (0.32)
Within R2 0.42 0.39 0.41 0.51 0.47 0.50 0.26 0.25 0.25
# of countries 57 58 57 57 58 57 56 57 56
# of obs. 210 214 210 202 206 202 181 185 181
Sargan–Hansen p-value 0.478 0.438 0.651 0.296 0.712 0.315 0.147 0.691 0.221

Note: All variables are expressed in logs. The estimation is based on the fixed-effects 2SLS method for which we report the within R-squared. t-statistics in parentheses
are based on standard errors that are robust to heteroskedasticity. ***, **, and * denote significance at 1%, 5%, and 10%, respectively. The number of observations
is reduced in the table because of missing values for poverty variables (see summary statistics for details). The sample is further reduced because ICRG does not
report data for some countries like Benin, Cambodia, Nepal, etc. (1) Assets bank is instrumented using Rule of law, Ethnic tensions, 1st and 2nd lags of assets bank,
Per capita GDP, and Gini index as instruments. (2) Assets MFI is instrumented using Rule of law, Ethnic tensions, 1st and 2nd lags of assets MFI, Per capita GDP,
and Gini index as instruments.

6. Conclusion possible policy recommendation is to lower such fees in order to

allow more people to take advantage of financial opportunities
This paper examines the relationship between financial devel- that banks can provide. Such fees could be a cause of a market
opment as measured by the size of either traditional banks failure, which according to Stiglitz (1993) is the fundamental
or MFIs and poverty reduction. Specifically, we compare the cause of poverty in developing countries.
extent to which each one of these financial institutions con- The microfinance results seem to be consistent with
tributes to alleviate poverty. To that effect, we employ a poverty Chowdhury (2009), who casts doubt on the effectiveness of
model developed by Ravallion (1997) that regresses poverty on microfinance as a means to alleviate poverty. Even if theo-
income per capita and the Gini index, in which we include finan- ries (Buera et al., 2012) and evidence (Donou-Adonsou, 2014;
cial development indicators. Applying the fixed-effects 2SLS Donou-Adonsou and Sylwester, 2015) suggest that MFIs have
technique to a panel of 71 developing countries from 2002 to macroeconomic effects such as growth effect, our study clearly
2011, our results indicate that bank loans have poverty reduc- implies that the so-called trickle-down effect in the financial
tion effects when poverty is measured by poverty headcount and development literature may not work for MFIs as it does for
poverty gap. When poverty is measured by squared poverty gap, banks. We do not suggest that the international community
bank loans do not have significant impacts. MFI loans, on the should not support MFIs, but our results do suggest that even if
other hand, do not appear to significantly reduce poverty across they are beneficial, MFIs are not panaceas for reducing poverty.
all the three measures of poverty. These results are robust when We urge further work in this area to better understand find-
financial development is measured by bank assets or MFI assets ings. The data only extend back to 2002, preventing use of a
as a percentage of GDP. More importantly, banks’ impact on long time series. Perhaps our sample period is not long enough
poverty is higher. Our results suggest that banking development to adequately measure the effect of MFIs on poverty. As stated
can help combat extreme poverty, but may fail to reach the poo- previously, due to data availability our measures of the size of
rest, while microfinance development is still at infant stage and MFIs rely upon the characteristics of the borrower. Ideally, one
has certainly a long way to go to alleviate poverty. Many studies would like to examine to what extent results would change if
have reported poverty reduction effects of MFIs at the micro MFIs were defined based on the characteristics of the lender.
level, but these effects are yet to appear at the aggregate level. Finally, we consider the potential for banks and MFIs to have
A few studies have reported that financial development countrywide effects, but macro effects could arise at regional
reduces poverty (Honohan, 2004; Jalilian and Kirkpatrick, 2005; (i.e. provincial levels) even if they are not countrywide. These
Beck et al., 2007; Jeanneney and Kpodar, 2011; Sehrawat and limitations warrant continued study of this issue.
Giri, 2015), and our bank results are consistent with the liter-
ature, at least when poverty is measured by poverty headcount Appendix.
and poverty gap. In the model of Greenwood and Jovanovic
(1990), there is a fee the poor need to pay to participate in the The countries in the sample include: Albania, Armenia,
formal financial market. Failing to pay this fee excludes the poor Azerbaijan, Bangladesh, Benin, Bolivia, Brazil, Burkina Faso,
from taking advantage of opportunities in the financial sector. A Burundi, Cambodia, Cameroon, Chad, China, Colombia, Congo

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]
+Model
RDF-76; No. of Pages 9 ARTICLE IN PRESS
F. Donou-Adonsou, K. Sylwester / Review of Development Finance xxx (2016) xxx–xxx 9

(Dem.), Congo (Rep.), Costa Rica, Cote d’Ivoire, Domini- Fan, S., Zhang, X., 2008. Public expenditure, growth and poverty reduction in
can Republic, Ecuador, Egypt, El Salvador, Ethiopia, Georgia, rural Uganda. Afr. Dev. Rev. 20 (3), 466–496.
Gachassin, M., Najman, B., Raballand, G., 2010. The impact of roads on poverty
Ghana, Guatemala, Haiti, Honduras, India, Indonesia, Jor-
reduction: a case study of Cameroon. World Bank Policy Research Working
dan, Kazakhstan, Kenya, Kyrgyzstan, Macedonia, Madagascar, Paper Series., pp. WPS5209.
Malawi, Mali, Mexico, Moldova, Mongolia, Morocco, Mozam- Greenwood, J., Jovanovic, B., 1990. Financial development, growth, and the
bique, Namibia, Nepal, Nicaragua, Niger, Pakistan, Panama, distribution of income. J. Polit. Econ. 98, 1076–1107.
Peru, Philippines, Romania, Russia, Senegal, Serbia, Sierra Holden, P., Prokopenko, V., 2001. Financial development and poverty allevia-
Leone, South Africa, Sri Lanka, Swaziland, Syria, Tanzania, tion: Issues and policy implications for developing and transition countries.
IMF Working Paper WP/01/160.
Thailand, Togo, Tunisia, Turkey, Uganda, Ukraine, Venezuela, Honohan, P., 2004. Financial development, growth, and poverty: How close are
Vietnam, Yemen, Zambia. the links. World Bank Policy Research Working Paper (3203). World Bank.
Imai, K.S., Azam, M.S., 2011. Does Microfinance Reduce Poverty in
References Bangladesh? New Evidence from Household Panel Data. Discussion Paper
DP2010-24. Kobe University.
Jalilian, H., Kirkpatrick, C., 2002. Financial development and poverty reduction
Adams, R.H., Page, J., 2005. Do international migration and remittances reduce
in developing countries. Int. J. Finance Econ. 7 (2), 97–108.
poverty in developing countries? World Dev. 33 (10), 1645–1669.
Jalilian, H., Kirkpatrick, C., 2005. Does financial development contribute to
Ahlin, C., Jiang, N., 2008. Can micro-credit bring development? J. Dev. Econ.
poverty reduction? J. Dev. Stud. 41 (4), 636–656.
86, 1–21.
Beck, T., Demirguc-Kunt, A., Laeven, L., Levine, R., 2008. Finance, firm size, Jeanneney, S.G., Kpodar, K., 2011. Financial development and poverty reduc-
and growth. J. Money Credit Bank. 40 (7), 1379–1405. tion: can there be a benefit without a cost? J. Dev. Stud. 47 (1),
143–163.
Beck, T., Demirgüç-Kunt, A., Levine, R., 2003. Law, endowments, and finance.
J. Financ. Econ. 70 (2), 137–181. Kaboski, J., Towsend, R., 2012. The impact of credit on village economies. Am.
Econ. J.: Appl. Econ. 4 (2), 98–133.
Beck, T., Demirgüç-Kunt, A., Levine, R., 2007. Finance, inequality and the poor.
J. Econ. Growth 12 (1), 27–49. Khandker, S.R., 2005. Microfinance and poverty: evidence using panel data from
Beck, T., Levine, R., Loayza, N., 2000. Finance and the sources of growth. J. Bangladesh. World Bank Econ. Rev. 19 (2), 263–286.
Financ. Econ. 58, 261–300. Kondo, T., Orbeta, A., Dingcong, C., Infantado, C., 2008. Impact of microfinance
Beck, T., Demirguc-Kunt, A., Levine, R., Cihak, M., Feyen, E., 2013. Financial on rural households in the Philippines. IDS Bull. 39 (1), 51–70.
Development and Structure Dataset (updated April 2013). Policy Research Levine, R., Loayza, N., Beck, T., 2000. Financial intermediation and growth:
Working Paper 2146. World Bank, Washington, DC. causality and causes. J. Monet. Econ. 46, 31–77.
Berhane, G., Gardebroek, C., 2011. Does microfinance reduce rural poverty? Mahjabeen, R., 2008. Microfinancing in Bangladesh: impact on households,
Evidence based on household panel data from Northern Ethiopia. Am. J. consumption and welfare. J. Policy Model. 30, 1083–1092.
Agric. Econ. 93 (1), 43–55. Odhiambo, N.M., 2010. Is financial development a spur to poverty reduction?
Berhane, G., (Doctoral thesis) 2009. Econometric Analysis of Microfinance Kenya’s experience. J. Econ. Stud. 37 (3), 343–353.
Group Formation, Contractual Risks and Welfare Impacts in Northern Perez-Moreno, S., 2011. Financial development and poverty in developing
Ethiopia. Wageningen University. countries: a causal analysis. Empir. Econ. 41 (1), 57–80.
Buera, F.J., Kaboski, J.P., Shin, Y., 2012. The Macroeconomics of Microfinance. Ravallion, M., 1997. Can high-inequality developing countries escape absolute
Working Paper (17905). NBER. poverty? Econ. Lett. 56 (1), 51–57.
Chowdhury, A., 2009. Microfinance as a Poverty Reduction Tool – A Critical Ravallion, M., Chen, S., 1997. What can new survey data tell us about recent
Assessment. United Nations, Department of Economic and Social Affairs changes in distribution and poverty? World Bank Econ. Rev. 11 (2), 357–382.
(DESA) Working Paper (89). Sehrawat, M., Giri, A.K., 2015. Financial development, poverty and rural-urban
Collins, D., Morduch, J., Rutherford, S., Ruthven, O., 2009. Portfolios of the income inequality: evidence from South Asian countries. Qual. Quant., 1–14.
Poor: How the world’s poor live on $2 a day. Princeton University Press, Singh, R.J., Haacker, M., Lee, K.W., Le Goff, M., 2011. Determinants and
Princeton. macroeconomic impact of remittances in Sub-Saharan Africa. J. Afr. Econ.
Copestake, J., Williams, R., 2011. What is the Impact of Microfinance, and 20 (2), 312–340.
What Does this Imply for Microfinance Policy and for Future Impact Stud- Stewart, R., Van Rooyen, C., Dickson, K., Majoro, M., De Wet, T., 2010. What is
ies? Oxford Policy Management, Oxford, Retrieved from: [Link] the impact of microfinance on poor people? A systematic review of evidence
[Link]/sites/opml/files/impact [Link]. from Sub Saharan Africa. Technical Report, EPPI-Centre Social Science
Dollar, D., Kraay, A., 2002. Growth is good for the poor. J. Econ. Growth 7, Research Unit. University of London.
195–225. Stiglitz, J., 1993. The role of the state in financial markets. In: Proceedings of the
Donou-Adonsou, F., (Doctoral dissertation) 2014. Growth, Poverty, and Capital World Bank Annual Conference on Development Economics, pp. 19–52.
Structure Effects of Financial Development. Southern Illinois University at Thanvi, R.K., 2010. Microfinance Outreach: Comparing Banks and MFI (Band-
Carbondale. han) in Cooch Behar District. Centre for Microfinance Research. Bankers
Donou-Adonsou, F., Sylwester, K., 2015. Macroeconomic effects of microfi- Institute of Rural Development, Available from: [Link]
nance: evidence from developing countries. J. Econ. 41 (1), 21–35. doc/Comparing%20Banks%20and%[Link].
Dupas, P., Robinson, J., 2009. Savings Constraints and Microenterprise Devel- Yusupov, N., 2012. Microcredit and development in an occupational choice
model. Econ. Lett. 117, 820–823.
opment: Evidence from a Field Experiment in Kenya. NBER Working Papers
(14693). National Bureau of Economic Research.

Please cite this article in press as: Donou-Adonsou, F., Sylwester, K., Financial development and poverty reduction in developing countries:
New evidence from banks and microfinance institutions. Rev. Dev. Finance (2016), [Link]